Kevin Zhou All Handouts

knzhou.github.
io
Advice For Introductory Physics

In this file I answer some frequently asked questions about learning physics and entering physics
competitions. For general logistical questions, see the USAPhO FAQ on the official AAPT website.
For advice for how to continue after finishing introductory physics, see this file.
What should I know before I start learning physics?

In the American system, people typically learn physics in two stages. First, they take a year-
long algebra-based introductory course, which covers all subjects (mechanics, electromagnetism,
thermodynamics, a hint of modern physics), typically given in 10th or 11th grade, and corresponding
to AP Physics 1 and 2. Those interested in learning more typically take a second, calculus-based
introductory course, covering mechanics and electromagnetism, corresponding to AP Physics C.
To succeed in an algebra-based physics course, you should have a good grasp of algebra and
trigonometry, have good “number sense”, and know how to read graphs. (In terms of formal courses,
you should be taking Algebra II or higher at the same time.) If you don’t have this stuff down cold
(e.g. if you take more than one second to recall the value of sin 30◦ ), then everything will be much
harder, because a two-step problem will feel like it’s twenty steps, as you scramble to remember
math you’ve half-forgotten. It’s like trying to learn the guitar while hopping on one leg.
What should I know before I start entering physics competitions?

For people coming from a math background, the most important thing to remember is that physics
competitions aren’t like math competitions. The reason is that the typical American 10th grader
has taken ten years of math in school and zero years of physics. If you’re a bright student that likes
math, math competitions are a fun way of extending the knowledge you’ve spend a decade building
– you already have the foundations set.
If you’ve done well on math competitions, it’s tempting to jump directly into physics competitions
with the same attitude. After all, physics is just made of equations, which are math, right? If
you haven’t taken a solid year-long introductory physics course already, this attitude will make
you crash and burn. It typically results in people memorizing big lists of equations, without being
able to answer the most basic conceptual questions, and making ridiculous mistakes like confusing
tension T for time T because they’re the same letter. Without introductory physics under your belt,
you’re in the same position as a 1st grader is in math, trying to do a math competition without
even knowing how to add.
Another important difference is the role of more advanced classes. Richard Rusczyk famously
wrote in The Calculus Trap about how the standard math curriculum (calculus, multivariable
calculus, linear algebra) often just teaches a few calculational skills, without emphasizing the
problem solving skills needed in math competitions.
This is true, but physics is different. Math competitions focus on topics like Euclidean geometry
which rarely come up in higher mathematics, but can be scaled up to arbitrary difficulty; thus,
advanced classes don’t usually help. By contrast, physics competitions were invented to spark
interest in higher physics. Climbing from the F = ma to the IPhO will take you on a tour through
some of the greatest ideas in physics, from the problems that Newton solved to recent Nobel prizes.
A decent theoretical physics graduate student would know how to solve IPhO problems, and that’s
a good thing – it means you are learning important things about reality by doing them.
1
knzhou.github.io
So if you’ve learned advanced topics like relativity and quantum mechanics on your own, don’t
hesitate to jump into competitions; you’ll be rewarded for your deeper knowledge. And if you find
these subjects interesting and are debating whether they would be worth doing, just jump in! It’s
all good stuff, because it’s physics, and physics is fun.
How do I start learning physics?

The most common way to start learning physics is from your high school physics teacher!
What if I don’t have a physics teacher yet, and want to start by myself?
You’re in luck, because there are better resources for learning physics independently now than ever
before! I’ll list a few at the end of this answer. However, I want to start with some warnings. These
days, it’s easy to find good resources, but it’s even easier to find bad resources, which always vastly
outnumber the good, and you can end up wasting vast amounts of time.
First, if you’re just starting out, I strongly advise against using any resource that isn’t designed
as a cohesive whole. For example, the popular websites Brilliant and Expii have lots of neat
problems. But at this point, their physics curricula aren’t developed in a complete and logical
manner. The problems have wildly different notation, conventions, and difficulty, and units tend
not to be self-contained, often requiring knowledge from later units.
This especially applies to learning from Wikipedia. It has a lot of useful information, but if
you ever get confused reading it, e.g. if two definitions don’t seem to be compatible, or if a step
in a derivation doesn’t seem right, you should never, ever try to resolve it by opening up twenty
Wikipedia tabs. The answer is simply not going to be there, and you’ll just magnify your confusion.
YouTube videos have related problems. You can search for any topic in physics and find hundreds
of videos where a guy records himself explaining it off the top of his head. The problem is that
most of these people have only learned the basics the previous day, often by skimming Wikipedia.
Because they’re just talking off the top of their heads, their videos tend to be vague, inaccurate,
stuffed with filler, and way too long – YouTube pays them by the minute. Sometimes students
instinctively try to fix this by cranking the video speed up to 3x, which I think is almost always
a mistake. If you ever get the urge to do this, it probably means the video carries too little new
information to be worth watching, either because it’s too basic or just bad.
When I was a kid, I followed the usual procedure for information gathering taught to me in
public school: Google the term, open the top ten links, and then open all the links in those pages.
The typical result was that I’d get lost all day on an issue that should have taken five minutes,
wondering in despair why somebody didn’t just organize the material consistently. Only later did I
realize that this is literally what books and courses are! In fact, in the cases where YouTube and
Wikipedia explanations are complete and reliable, they’ve usually been copied line by line from a
book. If the book is a source of illumination, these secondary resources are just the shadows it casts
in various directions. There’s no need to stay in the shadows when you can go right to the source.
Of course, books and courses also vary widely in quality, and it’s important to avoid getting
stuck on a poor one. To understand why, you have to consider how good textbooks are created
in the first place. Usually, a teacher will start a course using an existing textbook. If they care
enough, they’ll consider a wide variety of approaches, then gradually synthesize a new one for their
lectures, based on their preferences; perhaps it will be more modern, more mathematically rigorous,
or more intuitive than the others. Then they’ll start typing up lecture notes, and once those get
refined enough, they can drop the textbook and have the students read the notes directly. Over
2
knzhou.github.io
many years, students will find errors and confusing spots in the notes, which the teacher fixes up,
while accruing a large bank of classroom-tested, interesting questions from the annual problem sets
and exams. Finally, the teacher staples all the materials together, and a new textbook is born. All
of the books and courses I recommend in this document were made this way.
There are two active ingredients in the process. The first is the students, who act as dedicated
test-readers, pushing the teacher to improve their materials year after year. Books that aren’t student
tested tend to be plagued with issues, such as constant typos, trivial or nonsensical problems, huge
jumps in difficulty, and crucial omissions. The second is the teacher’s deep expertise. To write a
good book, the teacher must know far more than what is actually contained in the book. This lets
them identify the big picture, understand problem solving strategies, create new problems, and see
the limitations of the usual formulas. Without this kind of expertise, books can still be clear, but
they’ll be missing something. They’ll tend to have lots of unoriginal plug-and-chug problems, rigid
advice that only works on such problems (“never use rotating frames”, “always begin by writing
F = ma for every particle”), and generalizations that don’t actually hold in the real world.
Therefore, a reliable way to find good books and courses is to look for those that have been
refined over a long period of time, by one or two professors, teaching a dedicated course at a good
university. So now we can finally get to some course recommendations!
• If you would like to get started with algebra-based physics, a good first goal is to pass AP
Physics 1/2. (Don’t worry about F = ma preparation until later.) Two good starting points
are the videos by Flipping Physics and Khan Academy, which have been thoroughly tested
and refined. If you’d like more structure, find an AP Physics course either online or in person
nearby. If you’re confident enough to study on your own, see the books recommended below.
• To learn calculus, you can get started with MIT OCW’s 18.01 course. You can also go through
any one of the nearly identical standard calculus books on the market, such as Stewart’s, which
all cover everything you need and more.1
• Once you know basic calculus, such as derivatives and single integrals, you’re ready to start
calculus-based physics. My top recommendation is Yale’s Fundamentals of Physics courses.
MIT OCW also has introductory physics courses, titled 8.01 and 8.02, but they have some
drawbacks. Walter Lewin’s old lectures are full of cool demonstrations, but they’re short on theory;
they would work better as a supplement if you’re interested. Meanwhile, the current 8.01 course is
broken up into 5 minute tidbits, which frankly makes it feel like a high school course to me, and
the 8.02 course materials are incomplete. EdX used to have a lot of great free options, but they’re
mostly shut down now, as its new owners try to figure out how to make money from them, and I
don’t think the new ones are nearly as good.
The main reason it’s so hard to find good video lectures for introductory physics is that in
the past decade, most top universities switched to teaching these courses with active learning,
1
Mathematicians often complain that these books aren’t rigorous enough, and prefer books like Spivak’s. But
these books are meant to train mathematicians, not physicists. Spivak is great for that purpose, but it only has
a single chapter on actually performing specific integrals, and it starts with proving basic propositions like 0 < 1
and 1 + 1 6= 0. If you’re interested in that kind of thing, you can start reading Spivak without any prior calculus
background. (Another good starting point is the Art of Problem Solving calculus book, which has a good balance of
proof sketches and concrete problems.) However, you won’t need any experience with rigorous proofs to get started
in physics. After all, Newton didn’t care about rigor. When the mathematical foundations of calculus were invented,
physicists had already been using it to solve problems for generations.
3
knzhou.github.io
where lecture is replaced with group problem solving, and students do background reading at home.
Education research has shown that this works better for the average student, who would otherwise
zone out during traditional lectures. If you’re motivated enough to be self-studying, that probably
doesn’t apply to you, so you shouldn’t feel bad about using lectures. But it does show that lectures
aren’t necessary, and you can do everything by just following good books and thinking hard. Book
recommendations for introductory physics are listed in a separate section below.
How can I tell if I understand algebra-based introductory physics?

I’ll let you in on a secret: there are standard benchmark exams used in physics education research
which have been designed over years to measure exactly this, for the purpose of evaluating new
teaching methods. Examples include the Force Concept Inventory and the Conceptual Survey of
Electricity and Magnetism. (Of course, they only work for research if people haven’t seen them
beforehand, but I think there are few enough people reading this that it won’t matter.)
Find these exams online. If you understand basic mechanics and electromagnetism, you should
be able to get above 90% on the FCI within 30 minutes, and above 80% on the CSEM within 45
minutes. If you can’t do this, you likely have misconceptions that you should resolve before doing
anything else! The newly redesigned AP Physics 1 and 2 exams are also a good benchmark; these
cover mechanics and everything else, respectively. If you can’t comfortably score a 5 on AP Physics
1, your mechanics isn’t in good shape.
What are some good introductory books at each level?

There’s a robust ecosystem of physics textbooks, with many good options.
• For algebra-based physics, commonly used books are listed in AIP’s survey of physics teachers.
Some examples of decent books, in very roughly increasing order of difficulty, include:
– Hewitt, Conceptual Physics.

– Serway and Faughn, Holt Physics.
– Serway and Vuille, College Physics.2
– Cutnell, Johnson, Young, and Stadler, Physics.
– Knight, Jones, and Field, College Physics: A Strategic Approach.
– Giancoli, Physics: Principles with Applications.
Judging from reviews and survey data, Hewitt is a good option for a typical high school course,
while Giancoli is good for an honors high school course, such as for AP Physics 1 and 2. However,
these books are all pretty similar, so you shouldn’t worry if you happen to have a different one.
None of these books are enough for physics competitions, but they’ll set a good foundation. To
start at this level, you should at least be simultaneously enrolled in an Algebra II math course.
If you’re comfortable with calculus, you could also just skip directly to calculus-based physics.
2
Note that Serway, Giancoli, and Knight also have other textbooks meant for calculus-based introductory physics.
You might be confused why the books with the word “college” in the title are actually the less advanced, algebra-based
ones. The reason is that college introductory physics has been gradually watered down for generations. In the 60s,
Halliday, Resnick, and Krane was used for standard courses in average universities. Now, there are college physics
courses at or below the level of high school algebra-based physics! These courses use “College Physics” books to say,
“look, I’m totally not a high school course”. Similarly, “College Math” books are around the level of Algebra II.
4
knzhou.github.io
• For basic calculus-based physics, there are many books, such as the ones by Giancoli, Knight,
Serway and Jewett, Tipler and Mosca, Young and Freedman, and Halliday, Resnick, and Walker.
They all cover the same material, with nearly identical tables of contents, and they’re all suitable
for AP Physics C. Most of them have titles like “University Physics” or “Physics for Scientists
and Engineers”. They’re polished and equally good, so just use whichever you can easily get.
• For more advanced calculus-based physics, I strongly recommend Physics (5th edition) by Halli-
day, Resnick, and Krane. This book is used in college honors courses, and has significantly more
challenging problems, which were edited by a past director of the USAPhO. The explanations
are very clear, and I know many people who have succeeded using it.
Like most physical things in America, introductory physics textbooks date back to the heady
days of the 1950s. After Sputnik, concerned scientists and policymakers made a societal push for
STEM education. This gave rise to many great books, such as the original Halliday and Resnick,
and the Feynman lectures. Halliday and Resnick was so successful that all the other calculus-based
textbooks listed above are just watered down descendants of it (i.e. taking topics out, but never
adding any new topics in), which explains why they’re so similar. For example, Halliday and Resnick
itself split into two versions, Physics and Fundamentals of Physics, by Halliday, Resnick, and Walker.
The latter is essentially just Physics, but with the most advanced parts of each chapter removed.
When shopping for these books, you might notice that they come in many editions, and that
the latest edition is much more expensive than the rest. For example, Serway and Jewett is on its
10th edition, while Young and Freedman is on its 15th . However, you shouldn’t worry if you can’t
afford the latest edition, or if you happen to already have an earlier edition. The core introductory
physics curriculum hasn’t changed for generations; the real purpose of the endless editions is to
keep money steadily flowing in for the publishers. To make a new edition, they randomly rearrange
the ordering of the problems and the numbers inside, add a few janky “online-only” problems3 , and
lobby massive university systems to make their instructors require their students to get the newest
edition. These changes are intended solely to prevent students from buying cheap used copies.
Of course, most of the time, you should prefer the latest edition of a book. They tend to have
fewer mistakes, and sometimes better content; for example, in the case of Halliday, Resnick, and
Krane, the latest (5th ) edition has many very useful multiple choice questions, and some extra tricky
problems. But these benefits never apply for textbooks with over 10 editions, which just tend to
get more bloated with plug-and-chug problems over time. This is all just to say that while physics
is certainly real, there are a lot of things about the physics education system that aren’t. You don’t
need to use all its hyper-monetized features, and if something seems fake to you, it probably is.
On that subject, there are many supplemental books made for test preparation, such as Schaum’s
outlines, and the Princeton Review, Barron’s, and 5 Steps to a 5 series. I generally don’t recommend
them. They tend to have much higher average review ratings than real textbooks, but that’s because
the reviews are left by students who want to cram to pass, not learn. They are designed to get you
through the simplest possible questions with the least possible mental effort, and as such, don’t
really explain how or why anything works. Not only does this suck all the joy out of learning, it’ll
leave you unable to answer any question deeper than a one-step plug-and-chug. They may be okay
for a very quick first exposure, but you’ll want to upgrade to something better quite soon.
3
This is what the back covers of these books mean when they say they use the latest educational innovations. In
reality, it just means you type the answers to plug-and-chug problems into their system, rather than writing
√ them
down
√ on paper. The difference is that the automated system will sometimes mark you wrong for typing 1/ 2 instead
of 2/2, or 5.0 instead of 5.00. The actual purpose of the system to ensure that even once you have the textbook,
you can be charged again for the homework. It’s like loot boxes and DLC, but for physics.
5
knzhou.github.io
How much time will it take to qualify for USAPhO/qualify for USAPhO camp/win an
IPhO gold medal?
This varies depending on the person and their motivation, but here’s my timeline.
• 9th grade: I took a standard pre-calculus course in school and didn’t know or learn any physics.
• 9th grade summer: I don’t recall learning anything. I grinded a lot on RuneScape, with
occasional breaks to practice for math competitions.
• 10th grade: I took standard calculus and algebra-based introductory physics courses in school,
with great teachers in both. I didn’t prep for competitions, but I asked a lot of questions in
class, thought carefully about the intuition behind the equations, and occasionally skimmed the
mediocre Holt Physics book given. I just barely qualified for the USAPhO, and scored almost
zero on it. I found that experience really motivating, since it showed me that physics was full
of cool problems, which took a lot more than just plugging numbers into a formula sheet.
• 10th grade spring/summer: I self-studied calculus-based physics by reading the honestly terrible
Barron’s AP Physics C prep book and randomly googling whenever I got confused. This took
roughly 150 hours of work. Some of this was done while avoiding MOP homework.
• 11th grade: I read the awesome Halliday, Resnick, and Krane throughout the year, mixing in
past F = ma exams in January and past USAPhOs in the spring. I worked roughly 10 hours a
week, for about 250 hours in total. That year I qualified for camp and got an IPhO gold medal.
The point is that you don’t need a decade of study or a ton of prep programs to succeed. You just
need to get the basics down, and spend about one year learning on top of that. And this isn’t just
my experience. When we ask students who qualify for camp to describe their journey, they usually
say something very similar. They learn physics for a year, or maybe two if they have a lot of other
things going on. Prep courses are common, but most just take only one such course, or read just one
good textbook. Some don’t even prepare at all; they build their skills by following their curiosity.
What makes a competition prep program effective?

The main thing that makes a prep program effective is the student.4 The simple fact is that if a
student isn’t engaged, then prep programs are useless at best. This is obvious if you just look at
the numbers. Suppose an unmotivated student is dragged to a 1.5 hour class every week for eight
weeks, then grudgingly spends an hour a week on the homework. That only adds up to 20 hours
of experience, and not very high-quality ones at that. If practice stops entirely once the class ends,
most of that knowledge will be quickly forgotten.
Compare this to what I listed above: 400 hours accumulated over a year. Objectively, that isn’t
a lot of time; people could easily spend longer than that on a single high-school course if it’s loaded
with busywork. But these hours were focused ones, and they were spaced out regularly. I didn’t
need to cram, because I’d been immersed in physics the whole time.
4
By the way, one thing that can certainly never make a prep program effective is the parent. I sometimes see
parents spend more energy dragging their kid through prep classes and books than their kids spend actually thinking
about physics. Sometimes parents even solve the problems for their kids! Parental involvement is like salt. A pinch
can enhance a dish, but too much overwhelms the taste, and adding more makes it inedible.
6
knzhou.github.io
You might think prep programs can cut down the hours needed because they “teach to the test”.
This is a myth. Even the F = ma exam requires a broad understanding of mechanics. It’s certainly
possible to characterize the solutions to individual F = ma problems as “tricks”, but if you don’t
have a foundation, there will be an overwhelmingly large number of tricks for you to memorize, and
they’ll be ten times as hard to remember because you won’t know where they come from.
If that doesn’t convince you, think about learning an instrument, playing a sport, or learning a
language. Do football players cram in eight hours of practice the day before a big match? Have you
ever seen a pianist who got anywhere on an hour a week of practice? Of course not, and learning
physics (yet another language) is no different. There is no secret. You just have to engage.
Is prep program X, book Y, or course Z enough for USAPhO?

Any decent calculus-based physics course, book, or prep program is “enough”, in the sense that
they’ll all cover everything you need. But it’s up to you to turn that coverage into understanding!
Do I really have to learn X if I want to win competitions?

For almost any value of X, the answer is “probably not”, but if you ask this kind of question
constantly, you won’t do well anyway. Stop and find a different extracurricular, one where you’re
excited to do more rather than bargaining to do less.
Jeez, okay, but can I qualify for USAPhO without knowing calculus?
Every problem on the F = ma exam can technically be solved without calculus, but most students
who pass the exam know calculus-based physics. The reason is that it’s hard to derive most equations
in physics without using calculus. And if you don’t know how the equations are derived, you might
only see them as a disconnected pile of results instead of an interconnected web of ideas. This
penalizes you on the F = ma, where many questions require the test taker to think carefully about
which equations apply and why. It’s certainly not impossible to pass without calculus, but you’re
going to have to put in the time to build a solid conceptual understanding either way. In fact,
this might end up taking longer if you try to do it without calculus. If you’re the kind of student
interested in physics competitions, you would almost certainly enjoy learning calculus anyway, so
you should go ahead and do so!
And what about those weird things I learned in middle school?

The standard American public school physics curriculum has a lot of things that don’t really make
sense. For example, you are told to remember that there are precisely 3 kinds of lever, 4 states of
matter (or was it 5?), 6 kinds of simple machine, and 7 steps in the official Scientific Method. Or
that when you round numbers, you should round to the closest digit, unless you’re rounding a 5, in
which case you should round to an even digit, unless the number was negative. In some schools, you
must remember that the pound is really a unit of mass; the unit of weight is called the pound-force.
In other schools, you must remember that the pound is really a unit of weight; the unit of mass
is called the pound-mass. In some schools, you must do multiplication and division from left to
right, so that 1/2 × 3 = 3/2. In other schools, you must do multiplication before division, so that
1/2 × 3 = 1/6. In yet others, division comes first.
When I was in elementary and middle school, I thought this minutia was incredibly boring. It
turned me off science, which seemed to boil down to the drawing of arbitrary distinctions and the
7
knzhou.github.io
memorization of arbitrary rules, occasionally punctuated by cutesy crafts5 . Thankfully, none of

this matters for the Olympiad, or physics in general. It is only repeated in school out of habit and
circumstance.
The problem for the teacher is that solving real, interesting problems takes a fair amount of
dedication and background on the part of the student. Covering minutia is a convenient alternative,
because most students can be trained to do it, and an infinite number of quiz problems on it can
be easily generated and graded. That’s why, when the physics education researcher Edward Redish
once asked his students what the most important equation in mechanics was, the most common
response was d = at2 /2. These days, teachers can use programs to automatically generate hundreds
of uniform acceleration problems.
Often, the rules you’re supposed to memorize don’t even agree from school to school, and the
reason is that they truly don’t matter. No puzzle in physics has ever hinged on whether the One
True Order of Operations was PEMDAS or PEDMAS, even though people never seem to tire of
debating it on social media. If you’re like I was as a kid, you’ll want to ignore this noise altogether,
but unfortunately grades6 are still quite important at this stage in your life. My advice is to grit
your teeth, learn it just well enough to maintain decent grades, and immediately forget it.
Of course, some of the arbitrary-looking stuff you learn in school actually does turn out to be
important. For example, you’ll probably spend a lot of time manipulating matrices, in what seems
to just be a complicated way to rewrite basic algebra. Most school teachers can’t tell you why this
is worthwhile, but matrices turn out to be extremely important in more advanced physics. So how
can you tell what you need to know? In general, you can avoid this problem by sticking to good
books. They’ll contain exactly what actually matters.
How should I prepare for the F = ma exam?

The main ingredient for success is a solid understanding of mechanics, which you should get from a
book like Halliday, Resnick, and Krane. You should also prepare for the format and quirks of the
F = ma exam, but don’t get the priority flipped: specific preparation should take a couple dozen
hours at most, while learning the foundations takes hundreds of hours.
Anyway, the F = ma exam throws tricky multiple choice questions at you under extreme time
pressure, and the best way to prepare for that is to train on similar problems under timed conditions.
5
If you didn’t go to an average American public school, examples include drawing hand turkeys, decorating cupcakes,
sculpting mitochondria, and making collages. In a typical week, I would make a burger-shaped book report, a Lego
model of Lithium, and a mosaic of Manitoba. I’d also have to bug my Chinese-educated parents to buy construction
paper, not regular paper, leaving them wondering why I needed scissors, glue, posterboard, and 5 colors of paper just
to learn long division. Indeed, most non-Americans are surprised by our emphasis of crafts over actual information,
which stems from a combination of underfunding and modern educational philosophies. These philosophies say that
it’s a sin for a teacher to simply tell students what’s true; they should construct it from themselves. In practice, what
this meant is that we’d receive about two sentences of information, then get assigned some random topic, like boron
or quasars. Then we would spend hours copy-pasting from Wikipedia, with the teacher occasionally swinging by to
remind us to “use critical thinking”, which was kind of hard when nobody knew what the hell was going on.
6
Grades can be a decent indicator of learning if you have good teachers. But if you have bad teachers, they just
indicate obedience: whether you were able to parrot back dubious information, quickly and reliably, with a smile.
Given the well-known problems of grades, one might think it would be better to measure students in a way that has
been carefully developed, refined, and standardized by a competent outside party, such as an exam of some sort...
alas, such an approach is deeply out of fashion in the United States. That’s why competitions are so important.
They are even more controversial among educators than standardized tests, since many educators view any kind of
competition as immoral, but the truth is that the competitive aspect is irrelevant. The real point of competitions
is that they’re one of the few places left you can put your skills to work on nontrivial problems, to see if you truly
understand something. And they’re the definitely the only place you can do that with no budget or outside help.
8
knzhou.github.io
There are over 20 past F = ma exams publicly available, which gradually increase in difficulty over
time. After completing past F = ma exams, you should immediately check against the answer
key, and understand how to solve any question you missed. (Earlier F = ma exams don’t come
with detailed solutions. For 2011 through 2019, you can use the solutions in the book by Kisacanin
and Zhang, which are distributed for free on the AAPT website. From 2018 onward, there are
also detailed official solutions.) Another excellent resource is Morin’s Problems and Solutions in
Introductory Mechanics, which contains a lot of multiple choice questions, with explanations, at
about the right level.
If you run out of problems, you could also try past PhysicsBowl questions, the CAP prize exam,
the first round of the British Physics Olympiad, or the Hong Kong Physics Olympiad. There are
also old F = ma exams going back to 1997 available for purchase on the AAPT website. However,
all these competitions are significantly more straightforward, and some contain non-mechanics
questions. I think it’s best to just make the most of the F = ma exams.
Sir, what are your tips to crack JEE?

I get a lot of questions from Indian students asking about the INPhO, JEE Mains, or JEE Advanced,
but I really can’t help, because Indian competitions differ from American ones in many ways. First,
for historical reasons, India’s physics curriculum is a lot closer to that of the former Soviet Union.
That means there’s a lot more emphasis on elementary mechanics, optics, circuits, and nuclear
physics. Second, because the exams have almost a million participants, they are very competitive. I
don’t think their questions are harder than those of American or East Asian competitions, but they
demand extreme precision, since the time limits are short and partial credit is almost nonexistent.
In my opinion, this degree of competition harms actual learning. For example, students often
speak of neglecting “theory” to cram in more practice with “problem solving”. When I first heard
this, I was very confused. How can you have one without the other? Any decent textbook should
explain the theory and then show example problems. But apparently, it’s common to study problem
solving without theory, in the sense that you just memorize a lot of procedures for solving standard
problems without asking why they work. Maybe that’s what the average student has to do to
succeed in such exams, but it’s not ideal. The obsession with shallow JEE prep is probably the
reason India punches below its weight in Olympiads.
What’s the best way to spend my time learning?

There are a few basic principles that almost everyone, from teachers to education researchers to
bloggers, agree on.
• Don’t passively consume content. When you read about a new physical idea, turn it over in
your head. Ask yourself where you’ve seen the idea at work in the real world. Look at the
logical development of the idea – what assumptions do you need to get from one equation to
another? Take limiting cases of the relevant equations, relating them to ideas you already know,
or try to go beyond, seeing where the idea might fail. Try to reconstruct the idea, in a way
that makes it feel intuitive. Do practice problems, or invent your own.
• All that matters is that you properly chew and digest the ideas. Everybody has their favorite
way of doing this. Some people swear that you have to handwrite your notes, not type them;
some old folks might tell you the only real way to learn is to write cursive with a fountain pen.
I type my notes in bulleted lists, but others prefer web-like structures such as mind maps, and
9
knzhou.github.io
yet others never take notes at all. Some people swear by books and others swear by lectures.7
Some people keep their books pristine and others highlight every word. I love explaining things
verbally, while others prefer visualization. None of these details really matter. Use whatever
method you like best, and it’ll work as long as it keeps you engaged with the ideas.
• The best way to remember something long-term is spaced repetition: apply the idea the
moment you learn it, then reencounter and reuse it regularly. Good physics books and courses
will automatically make you do this, as long as you work steadily and linearly through them.
• Do practice problems that are at or just above your current level. They should be hard enough
to require your full attention, but not so hard that you spend long stretches of time making no
progress. Don’t peek at solutions until you give each problem a good try. (If you need to peek
at the solutions for more than half of the problems you’re attempting, they’re too hard.) When
you finish doing a practice problem, reflect on what went well or poorly, and if you weren’t able
to do it, figure out the crucial steps you were missing.
• Make sure your studying is healthy. Long cram sessions aren’t effective. Take regular breaks
and use them to stretch your legs. Sleep at least 7 or 8 hours a day, drink water, eat food, and
generally obey common sense. Studying when your brain or body is tired is only useful for
mindless tasks like cramming things into short-term memory, the opposite of what you need.
How should I self-study from Halliday, Resnick, and Krane?

Halliday, Resnick, and Krane has a total of 52 chapters (though the last 6 are on advanced topics),
and each chapter comes with multiple choice questions, conceptual questions, exercises, and problems.
A good pace would be an average of one chapter per week, or three chapters per two weeks.
If you’re self-studying, it’s essential to continuously test your knowledge, to avoid gaps in under-
standing. While reading a chapter, you should spend at least as long thinking about its contents as
you do physically reading the words. Afterward, I recommend spending at least a moment thinking
about every conceptual question, as many of them help you connect the theory to the real world,
and some are surprisingly deep. I also recommend doing all of the multiple choice questions, since
they are excellent preparation for the F = ma exam.
On the other hand, it probably isn’t worthwhile to do all of the exercises, since many reduce
to plugging numbers into standard formulas. I recommend skimming to see if you know how to
do them, and perhaps doing a small sample to check. On the other hand, the problems are more
subtle, and form a very useful bridge from “plug and chug” problems to competition problems. I
recommend reading all of the problems carefully, and doing at least half of them, depending on
which strike your interest.
Answers to odd-numbered exercises and problems are at the end of the book, and a detailed
instructor’s solution manual to all exercises and problems can be found online. There are no answers
to the multiple choice questions, but you can find my answers for the first 17 chapters here.
7
This is a perennial controversy in the education literature. The average book is probably clearer than the average
lecture, since books often emerge from refinements of lecture notes. But lectures can be more engaging, because the
instructor presents live, in the flesh, in a room full of other students paying attention. Books can drown students in
too much detail, but lectures can keep students from thinking if they’re too busy taking notes. What about video
lectures? You can rewind, pause, or speed up a video lecture, unlike a real lecture, but in the absence of a definite
time slot you might never get around to watching it at all. The real answer is that all the methods can be good or
bad; it depends on how you, personally, best absorb information. Though we can all agree that Zoom school sucks.
10
knzhou.github.io
What are some important traps to avoid?

If you’re at a “top” high school, the biggest trap is confusing your schooling with your education.
A mild case of this looks like signing up for every AP class your school offers, and spending all
your nights and weekends grinding out busywork you don’t really care about. More advanced
symptoms include spending hours reading rubrics like a lawyer to argue a grade of 97% up to 98%,
and jockeying for “leadership” positions in a huge array of fake clubs that don’t ever do anything.
In the terminal stages of this disease, you could end up founding a fake nonprofit in junior year
because twenty of your classmates did. People act in this undignified way because they think it’ll
get them ahead in our broken system, but that’s really not how it works. Even admissions officers
can see through this most of the time, and even if they couldn’t, it still wouldn’t be worth doing,
because it leaves no room for real education!
Another common trap is overweighting the qualitative or the quantitative side. People in the
first category often say things like, “I don’t need to understand how to do mindless calculations like
everybody else, because the intuition is what matters.” People in the second category will say, “Who
cares why that works – I got the right answer this time, didn’t I?” Of course, both are misguided,
because to solve nontrivial problems you’ll need to be comfortable with both sides.8
Overplanning is another common trap. A lot of people get caught up on finding the optimal
books and the optimal practice problems, and never actually starting to do either. Some people
even make a detailed, multi-year study plan for getting an IPhO gold medal set before they learn
Newton’s laws, which is both a waste of time, and seriously demotivating once they realize that
making a plan is much easier than doing it.
Again, sports are a good analogy. Consider somebody who made their country’s youth soccer
team. They probably started by playing casual games with their friends, perhaps on their school’s
soccer team, gradually building up their skills while having fun. As they got better, the stakes were
gradually raised, until they ended up doing daily, carefully designed practice with a coach. But it
wouldn’t have made sense to go looking for that coach before even learning the rules of soccer!
Long-term motivation comes from small, consistent wins, not distant goals. After an hour of
learning, it is much more motivating to think “now I know why sunsets are red” or “now I know why
violins have those f -shaped holes” than “now I am 0.1% closer to an IPhO gold medal”. Excessive
planning gives you a false sense of a distant goal moving closer, which can be exciting, but ultimately
isn’t good for anything. The trap is to get addicted to that feeling of progress, to the point that
you want it more than actual progress. If you want to learn physics, the most important thing is
to just do physics.
Do I have enough talent to succeed?

In response to this difficult question, many well-intentioned adults assert that talent does not exist,
or that anybody can do anything if they really try. These sentiments come from a good place, but
they’re rarely satisfying to their recipients because they’re clearly not true. Talent does exist. It’s
the reason that wealthy families can spend tens of thousands of dollars propping up or outright
falsifying their children’s SAT scores, to get outscored by less advantaged kids using only the $20
Blue Book. More dramatically, it’s the reason that Ramanujan went from being the son of an Indian
8
At the frontier, it’s definitely true that there are leading researchers that lean one way or the other. But make no
mistake: all of them knew both sides of the fundamentals extremely well. You might see pop science portray Newton
and Einstein as daydreaming visionaries, but their breakthroughs were enabled by years of grinding out concrete
calculations with their immense technical skills, as you can see from their notebooks.
11
knzhou.github.io
clerk, doing mathematics alone in near starvation, to the apex of mathematics in Cambridge.
The more nuanced story is this: in legitimate systems, success comes from ability, and ability
comes from dedicated, effective practice. Dedicated practice comes from interest, and interest is
mediated by a combination of talent and socioeconomic factors.
To illustrate this point, consider the extreme case of child prodigies. Prodigies exist in chess,
music, math, and programming, but not in law, medicine, history, or literature, because the former
allow rapid learning and feedback, starting from minimal background knowledge. Children naturally
learn quickly, and a child knows immediately whether they’ve won or lost at chess, and when they’ve
made a clever move. Talent determines how often these exciting wins happen, and if there are
enough, a child can take a liking to chess, and begin a phase of rapid improvement.
Of course, socioeconomic factors play a role. Chess prodigies need someone to introduce them to
the game in the first place. They need stable homes and supportive parents, so that they have space
to focus on learning. They benefit from chess-playing adults they can look up to, a community
to help them learn faster, and a system of competitions to help them set goals and measure their
progress. That’s of course why chess prodigies appear in the West, Go prodigies appear in Asia,
and neither appear in bad times.
What does this have to do with physics? When I was a kid, I had a naive view of physics based
on talent. I thought every “level” of physics required some minimum bar of talent, and that people
just kept climbing until they hit a wall, a level of abstraction they were simply unable to grasp.
After all, that’s how adults talked about it. They’d say things like, “math stopped making sense for
me at trigonometry”, or “I couldn’t make it past differential equations.” So when things got hard,
such as when I started quantum field theory, I had a sinking feeling that I was “hitting the wall.”
But in reality, the cognitive load of learning stays relatively constant. With modern resources, the
difficulty of learning quantum mechanics is about the same as learning introductory physics, provided
you have equal mastery of the prerequisites. The reason people hit walls is largely not because the
material gets inherently harder, but because they suddenly fall through the massive holes in their
foundations. For example, algorithmically, differentiating functions is not more complex than doing
long division: the number of things to keep track of, and new rules to apply, is comparable. But
people get stuck at the former because it tends to expose all the misunderstandings they’ve ever had
about basic things, like simplifying fractions. That’s the problem; it can’t be the raw complexity of
manipulating the symbols, because all of us can follow a much larger set of rules for manipulating
a much larger set of symbols, whenever we assemble letters into words and sentences.
Incidentally, while I brought up child prodigies to illustrate a point, they shouldn’t worry you
in physics, even though they certainly exist. Sometimes people give up because they know people
younger than them who are “ahead”. This makes as little sense as worrying about all the people
older than them who know more. How are you ever going to catch up to the people who are already
in graduate school? The question doesn’t make sense, because success doesn’t come from “catching
up” to people. Success in physics requires accumulating a body of knowledge, which takes on the
order of one year for high school physics competitions, and ten for physics research. People who get
to that point earlier in life just get more time to use it; they don’t stop you from doing the same.9
Indeed, you shouldn’t ever worry if you, or others, are ahead or behind of the “usual” track.
There is nothing inherently natural about learning algebra at age 12, introductory physics at age
16, quantum mechanics at age 20, and quantum field theory at age 23. Those numbers are solely a
9
Of course, the same applies if you’re in college and never heard of Olympiads in high school. People seem to get
insecure about this for some reason, but the point of the Olympiad is just to spark interest in physics, through some
interesting elementary problems. If you’re already learning advanced physics, you’re not missing out!
12
knzhou.github.io
product of history and circumstance.10 (Plus, tons of kids could zoom right past them if they got to
spend their early years actually learning, instead of making cutesy crafts.) What you should learn
next is determined by your goals and prior knowledge, not your age.
So let’s say that you’re interested in learning more physics. Let’s say that you had a good
foundation in math, and when you learned basic elements of physics, things clicked for you. You
saw the world in a different way, and it felt good.11 Then I can assure you that if you continue
learning physics, it will keep paying off. You’ll continue to get “aha!” moments. You’ll continue to
be able to piece together, with concentrated effort, new ways of looking at the world. Of course,
the rate at which you do this is partially determined by talent. But if you’ve made physical insights
before, you will continue to make them in the future, provided your foundations are good. There is
no wall; how far you go is up to you.
10
And they get changed all the time. Under the new California state standards, tracking is removed and all students
must take introductory algebra (“solve 2x + 3 = 5”, “plot y = mx + b”) together in 9th grade, to ensure that no
students get ahead of others. What society views as the “right” age for algebra is a matter of politics – a case where
people fight over who gets what in a zero-sum game. People argue all day about how these changes will impact
statistics, from our achievement gap to our PISA ranking, but nobody seems to care about what the students actually
think. If you’re a bright young student, the best response is to secede. Ignore this debate, sit at the back of the class,
and teach yourself whatever you want.
11
This is really the only important point in this rambling answer. The problem with most general advice out there
is that the right advice depends on the person, but each chunk of advice is written with a specific kind of person
in mind. That leads to extremely one-sided treatments, with half the articles saying that anybody can do anything,
to encourage smart people full of self-doubt, and the other half playing up how arduous the road is, to discourage
naive optimists. How can you tell which side is meant for you? What matters is how learning physics makes you
feel, personally. If you don’t enjoy learning physics in the way described in this paragraph, the path forward will be
almost impossible. But if you do, many of the difficulties will take care of themselves.
13
Kevin Zhou Physics Olympiad Handouts
Syllabus
1 Advice for Handouts
Each problem set has about 30 problems, ranging in difficulty up to IPhO and beyond. For back-
ground, you should know calculus-based physics at the level of Halliday and Resnick. Each problem
set will also come with recommended reading from some of the textbooks listed below. If you’re
already comfortable with solving the problems, doing the reading is optional, and can be skipped to
save time. If the subject is new to you, I recommend spending substantial time doing background
reading before trying the problem set, doing extra problems from the textbooks if necessary.
The problems are chosen so that all of them demonstrate different ideas, so you’ll get more out
of the handouts the more you do. That said, it certainly isn’t necessary to do every problem. Every
problem will have a point value from 1 to 5, and each problem set comes with a cutoff which is
roughly 70% of the point total. If you reach this cutoff, you’ll have a good understanding of the
material at the USAPhO level. Students aiming at IPhO gold medals should try almost everything.
Problems marked with [A] are “advanced”. This doesn’t mean that they’re trickier, but rather
that they require more sophisticated mathematical techniques. These problems are less relevant to
Olympiad physics but are chosen to demonstrate interesting things. Sometimes, multi-part problems
will have one subpart that is significantly harder than the rest; these will be marked with stars (?).
If you’re interested in USAPhO preparation, you should attempt all of the USAPhO problems,
while if you’re interested in the IPhO, you should attempt the international-level (IPhO, APhO,
WoPhO, GPhO, EuPhO) problems. However, it’s valuable to try problems from all competitions.
The average IPhO question does require more insights than a USAPhO problem, but the time limits
are also longer, making them equally approachable. Conversely, newer USAPhO problems often
introduce new ideas quickly, and can be valuable practice even for IPhO contestants outside the US.
Some problems will be marked with a clock. For the best training results, they should be done
under realistic conditions, which means you should use only pencil, paper and a scientific calculator.
During this time you should write a solution by hand, with the same level of detail you would for
a real Olympiad. If you run out of time but you’re still making progress, feel free to continue, but
draw a line on your solution indicating when time ran out. Common time limits will be
01W – 22.5 minutes, 01m – 45 minutes, 01h – 100 minutes.
Usually, the time limit will match that in the actual competition (i.e. 10 minutes per point for an
international-level competition), but I’ll sometimes set a lower time limit if the question is relatively
short, or if it revolves around a single key insight that students would get hopelessly stuck without.
After finishing, check your answers and, if your solution was not complete, figure out what you
missed. It’s most efficient to do this check immediately, since the problem will be fresh in your mind.
For all problems, I strongly encourage you to write up a solution of your own if possible. It
doesn’t matter how neat your solution is (you certainly don’t have to type anything up; handwriting
is probably better since it more closely resembles exam conditions), but it should be a legitimate
full solution, with a fixed final answer. Otherwise, it’s easier to get lazy and waste lots of problems
by peeking at the official solutions too early.
1
2 Textbooks and Resources

We’ll be using a wide variety of textbooks and resources. A comprehensive list of relevant introductory
books is given in my second advice file, and relevant chapters will be mentioned in the suggested
readings. The following books are essential, and I recommend getting all of them immediately.
• Halliday, Resnick, and Krane, Physics. This book contains the foundational material required;
you should know it forwards and backwards. Even today, a solid understanding of it is enough
to get a gold medal at the IPhO, though of course more knowledge always helps. The 5th edition
is more expensive, but worth it for the extra challenging problems included.
• Kleppner and Kolenkow, An Introduction to Mechanics. Used at MIT, written more like a
physics book. Has good problems with a practical emphasis. I recommend getting the 1st
edition, not the 2nd, because the 1st edition has harder problems.
• Morin, Mechanics. Used at Harvard, written more like a math book. Has a large stock of
elegant and tricky, if sometimes contrived mechanics problems. Also contains an excellent,
careful introduction to special relativity.
• Purcell and Morin, Electricity and Magnetism. Does electromagnetism with vector calculus
and relativity baked in. Famous for using relativity to motivate magnetism, rather than just
postulating it. Has well-written problems that provide insight. The 3rd edition is a substantial
improvement on the 2nd, with SI units adopted throughout and more challenging problems.
• The Feynman Lectures on Physics. A wonderful source of physical insight. Most problem sets
will have some chapters assigned for entertainment and enrichment.
• Wang and Ricardo, Competitive Physics, used by the Singapore physics team. This book contains
clear, detailed explanations of the theory needed to bridge the gap from an introductory textbook
to the IPhO, especially for thermodynamics and waves. The main drawback is that many of its
problems are mathematically complex but straightforward; use another book for problems.
Besides past Olympiads and textbooks, problems are also sourced from the following books.
• ? 200 Puzzling Physics Problems and 200 More Puzzling Physics Problems. Tricky questions
written in Eastern European style. The first book is highly recommended; the second book is
at times too mathematically contrived to be too relevant to Olympiads, but still lots of fun.
• ? Handouts by Jaan Kalda. These handouts and formula sheets provide excellent training for
Eastern European style Olympiads. Good solutions written by students are available here. If
you like the style of the EFPhO/NBPhO, you can find more questions from earlier rounds here.
• Cahn and Nadgorny, A Guide to Physics Problems. A thorough collection of graduate school
qualification exam problems. Many great classic problems were given on these exams, though
most problems in the book are too technical to be useful for Olympiad preparation. The massive
series entitled “Major American Universities Ph.D. Qualifying Questions and Solutions”, edited
by Yung-Kuo Lim, has more questions of this type, though they’re easier on average.
• Here are some cute, ∼ 100 page books which you might enjoy spending an afternoon with.
2
– Mahajan, Order of Magnitude Physics. A nice book about dimensional analysis and esti-
mation. The Art of Insight is a longer work by the same author on the same themes.
– Lemons, A Student’s Guide to Dimensional Analysis. Shows a good variety of examples
throughout physics, with interesting historical asides.
– Levi, The Mathematical Mechanic. Uses mechanical setups to find slick solutions for calculus
and geometry problems.
• There are a number of compilations of Russian physics problems.
– Irodov, Problems in General Physics. This is a massive list of 2,000 practice problems.
Personally, I don’t recommend using it for the USAPhO or IPhO, because it’s simply too
long, with many tedious filler problems. On the other hand, it’s still a standard and useful
reference for the Indian Physics Olympiad.
– Krotov, Problems in Physics. A collection of old Russian Olympiad problems, trickier than
the average Irodov problem.
– Savqenko, Zadaqi po Fizike . This is an updated and upgraded version of Irodov, with
many tough problems. It is commonly used for training in Eastern Europe, but it has not
been translated from Russian. Many problems in Kalda’s handouts are drawn from it.
– Kiselev and Slobodyanin, Russian Physics Olympiads. This book is the modern analogue
of Krotov, containing problems from 2005 to 2017, but it’s hard to find a copy. Most of the
tricky homework problems in PhysicsWOOT are drawn from it. If you’re preparing for this
kind of exam and have already gone through the EFPhO/NBPhO problems, you should do
your best to find this book.
There’s also a great tradition of tough Russian problems at the university level, though unfortu-
nately many have not been translated from Russian. For some problems accessible to first year
students, see here. The Landau and Lifshitz series is also great for further study.
• Western Europe does not have a strong tradition of Olympiad physics, but there are some nice
compilations of problems at the university level which can be relevant.
– Povey, Professor Povey’s Perplexing Problems. A collection of simple but tricky Oxford
admissions interview questions with neat historical anecdotes.
– Cavendish Problems in Classical Physics. Some classic problems used for second year exams
in Cambridge, back when things were more hardcore.
– Thomas and Raine, Physics to a Degree. A collection of well-motivated questions used for
undergraduate physics training, with many real-world applications.
• Olympiad material from Asian countries is often exceptional in its depth, but unfortunately,
most such resources have not been translated to English.
– Physics Olympiad – Basic to Advanced Exercises, training material used by the Japanese
physics team. This book contains clear explanations of basic theory, along with a useful
introduction to experiments.
– H.C. Verma, Concepts of Physics. This well-written book covers the material in Halliday,
Resnick, and Krane, with many worked examples and problems. It is commonly used to
prepare for the Physics Olympiad in India. (However, I would strongly advise against using
3
any source used to cram for the JEE, such as Pathfinder or the Cengage series. I would
be surprised if anybody could learn from the extremely brief “coverage” of theory in these
books. The solved problems might be alright, but just about every page of exposition has
something wrong with it, and some of the “concepts” taught are misleading or meaningless.)
– 舒幼生, 物理学难题集萃. A massive, classic compilation of problems from the former head
coach of the Chinese physics team. Despite being almost 30 years old, it is still the most
comprehensive book in this whole syllabus, by a wide margin. Unfortunately, I haven’t
been able to include any questions from this book, or its successors, since I can’t read the
language. They would be a great source of fresh problems if you finish my problem sets.
To learn more advanced physics, see my second advice file for book recommendations. More special-
ized books will also be mentioned in the suggested readings for each problem set.
3 Olympiad Problems
You can access most of the Olympiad problems in the handouts using the following links.
• Recent USAPhO exams can be accessed here. We will also occasionally use older quarterfinals
and semifinals (and their solutions).
• You can access past IPhO exams here and past APhO exams here.
• We’ll also draw problems from the EuPhO, GPhO, EFPhO/NBPhO, BAUPC, BPhO, JPhO,
AuPhO, CPhO, IZhO, INPhO, and PPRDPhO. (For some other Olympiads, see here.)
EFPhO/NBPhO problems will not be timed, but if you’d like to compare yourself against the
competitors, this competition allows about 8 minutes per point (in contrast to the 10 minutes per
point in the IPhO and APhO). If you run into issues with math rendering, try downloading a local
copy and opening it with a dedicated PDF viewer.
4 Curriculum
An outline of the full curriculum is shown below. In all cases, the relevant material in Halliday,
Resnick, and Krane is a prerequisite, and most problem sets require all of the previous ones in that
topic. The 24 units most relevant to USAPhO preparation are underlined. Prior exposure to vector
calculus is useful, especially for thermodynamics and electromagnetism, but not necessary.
• 2 units of problem solving.
– P1: dimensional analysis, limiting cases, series expansions, differentials, iterative solutions.
– P2: probability, error analysis, data analysis, estimation, experimental technique.
• 8 units of mechanics.
– M1: kinematics. Solving F = ma, projectile motion, optimal launching. (P1 helpful)
– M2: statics. Force and torque balance, extended bodies, pressure and surface tension.
– M3: dynamics. Momentum, energy and center-of-mass energy, collisions.
4
– M4: oscillations. Damped/driven oscillators, normal modes, small oscillations, adiabaticity.

– M5: rotation. Angular kinematics, angular impulse, physical pendulums. (P2 helpful)
– M6: gravity. Kepler’s laws, rocket science, non-inertial frames, tides.
– M7: fluids. Buoyancy, Bernoulli’s principle, viscosity and surface tension. (M2 helpful)
– M8: synthesis. 3D rotation, precession, and tricky problems.
• 3 units of thermodynamics.
– T1: ideal gases, statistical mechanics, kinetic theory, the atmosphere. (M7 required)
– T2: laws of thermodynamics, quantum statistical mechanics, radiation, conduction.
– T3: surface tension, real fluids, phase transitions, compressible flow.
• 8 units of electromagnetism.
– E1: electrostatics. Coulomb’s law, Gauss’s law, potentials, conductors.

– E2: electricity. Images, capacitors, conduction, DC circuits.
– E3: magnetostatics. More circuits, Biot–Savart law, Ampere’s law, dipoles and solenoids.
– E4: Lorentz force. Dynamic charges, permanent magnets, solid state physics. (M4 helpful)
– E5: induction. Faraday’s law, inductors, dynamos, superconductors.
– E6: circuits. RLC circuits, filters, normal modes, diodes. (M4 required)
– E7: electrodynamics. More circuits, displacement current, radiation, field energy-momentum.
– E8: synthesis. Electromagnetic fields in matter, and tricky problems.
• 3 units of relativity.
– R1: kinematics. Lorentz transformations, Doppler effect, acceleration, classic paradoxes.

– R2: dynamics. Momentum, energy, four-vectors, forces, relativistic strings. (E4 helpful)
– R3: fields. Electromagnetic field transformations, the equivalence principle. (E7 required)
• 3 units of waves.
– W1: wave equation, standing waves, music, interferometry. (M4 required)

– W2: interference and diffraction, crystallography, real world examples. (E7 required)
– W3: sound waves, water waves, polarization, geometrical optics. (M7 required)
• 3 units of modern physics.
– X1: semiclassical quantum mechanics, bosons and fermions. (M4, T2, W1 required)
– X2: nuclear, particle, and atomic physics. (R2 required)
– X3: condensed matter, astrophysics, and cosmology. (W3 helpful)
My recommended path through the core 24 problem sets is P1, P2, M1–4, E1–4, M5–7, T1–3,
E5–7, R1, R2, W1, W2, X1. (This reflects the technical sophistication of the problems, and also
splits up the long topics.) There are six further advanced units which are more relevant for IPhO
preparation, as well as three review problem sets and eleven practice USAPhOs.
5
For USAPhO preparation, doing a fair amount of all 24 core problem sets is better than doing
a smaller number very thoroughly, because prior exposure to a wide range of ideas is useful when
encountering new questions. If you start in September, a good pace is one problem set per week;
if you start in mid-summer, a good pace is one problem set per 1.5 weeks. Note that problem sets
tend to require more background reading and have more complex questions as the course goes on.
Topics fluctuate significantly from year to year, but I would estimate that on average, a bit more
than half of the points on the USAPhO and IPhO are devoted to mechanics and electromagnetism,
with the rest roughly evenly divided between relativity, thermodynamics, waves, and modern physics.
Of these four special topics, thermodynamics and relativity tend to be a bit more common at the
USAPhO. At the IPhO, problems about specific advances in modern physics are common, though
many of the parts in such problems boil down to mechanics or electromagnetism.
6
Frequently Asked Questions

What are these handouts?
After graduating college, I tutored a few high school students per year as a side job. Over the first
two years, I developed a set of handouts to assign for homework. Then I spent a few more years
polishing them up, and writing detailed solutions with the help of two members of the US IPhO
team, Gopal Goel and Sean Chen, who both won gold medals after going through the handouts.
These handouts are the most comprehensive resource for Olympiad physics that I know of. They
total over 1,000 pages in length, containing over 250 worked examples and 1,000 challenging problems,
along with 7 practice USAPhOs. Most USAPhO, IPhO, APhO, and EuPhO problems are included,
though I’ve also taken problems from more obscure competitions, recent research, the Victorian-era
Cambridge Tripos, graduate qualifying exams, and more. I’ve also written some problems myself,
to illustrate ideas that common textbooks don’t explain well, added subparts to classic problems
to bring out subtle points, corrected competition questions when needed, and included in-depth
discussions of the context for problems that originate from history or more advanced physics.
The core USAPhO material is covered in 24 handouts. There are also 3 review handouts, and
6 more advanced handouts. Since Olympiad physics ultimately is just real physics with trickier
problems and elementary math, you could probably also use the handouts to learn university physics;
they cover approximately the first two years of an undergraduate degree.
Who has used the handouts?

Excluding 2020, when almost all competitions were cancelled, I’ve tutored 10 students using these
handouts. Of these, 9 have won USAPhO gold medals or equivalent, 5 have qualified for USAPhO
camp, and 3 have won IPhO gold medals. In 2021, I managed the training of the US IPhO team,
basing it primarily on these handouts. That year, they achieved a perfect 5 gold medal finish for
the first time in history.
What’s the catch?

There is no catch. All of the materials are completely free.
Why are you giving everything away for free?

When I was in high school, I heard about the Physics Olympiad from my physics teacher, spent
an afternoon searching online for a book recommendation, and bugged my parents to buy Halliday,
Resnick, and Krane. They got me a $20 used copy with the spine falling off, and I spent an enjoyable
year puzzling over it by myself, reading it on the weekends and during chemistry class. When I
went to the IPhO, I still only had that one book, held together with layers of tape.
This shows that Olympiads are one of the most accessible high school activities in the world.
There’s just about no other activity out there where you can make it that far with no adult help, and
almost zero budget.1 On the other hand, while there was a path, it wasn’t well-paved. I spent hours
at a time stuck on little things, because of missing steps in solutions, ambiguous statements, or
outright errors. I didn’t know much beyond the textbook, which meant I had almost no knowledge
1
Of course, the same goes for the SAT, for which the best prep resource is still the $20 Blue Book. Tests are always
cheap; it’s the inspirational, classy stuff made for college essays that costs money.
1
of subjects like fluid dynamics and circuits, and online explanations were of dubious quality. I didn’t
have a broad range of sources for training problems, or any idea of how to compare their difficulty,
so I was blindsided by IPhO questions that were different from what I’d seen before. Worst of all, I
didn’t get the big picture, the deep links between all the fields of physics I was learning. I managed
to piece some of it together myself, but the process was unreliable and slow.
My goal with these handouts is to shorten the way for future students, both within the United
States and around the world. In the original spirit of the Olympiads, I hope they spark interest in
the subject for years to come.
Do you offer tutoring for the problem sets?

No, unfortunately I’m too busy to tutor individual students. However, the problems and solutions
are designed so that students can follow them on their own, or with friends.
What is the schedule for releasing the handouts?

Because polishing solutions takes a great deal of time, I’ll release about one new problem set per
month during the academic year. Everything else will come out in the summer of 2022.
How do I know if I’m ready to start using the handouts?

Before starting, you should have completed most of a standard calculus-based introductory physics
book, such as Halliday, Resnick, and Krane. (For more guidance, see my advice on learning
introductory physics. You should also have some problem solving experience from entry-level
competitions. To judge this, try the basic problems in the preliminary problem set. For the
handouts to be most useful, you should be able to solve over 75% of them.
How long should it take to go through them?

It varies a lot. On average, students took about 10 months to go through the core 24 problem sets.
The amount of time they worked varied widely, from 8 to 15 hours per week. It also varied between
problem sets: sometimes a student would blaze through an early problem set in 2 days, but then
take 2 weeks on a later one because more background reading was required, or because they wanted
to stop and think about the material more deeply. I wouldn’t worry about how much time things
are taking you, as long as you’re learning new things and enjoying yourself.
How did you choose the problems?

I’ve tried to choose each problem to teach a distinct physical lesson, which means:
• Few “cookie-cutter” questions. I try to give one or two problems or examples for each new
technique you should know, but that’s enough. If you want more routine practice problems,
you can easily find them in standard books.
• No “hidden math” questions, i.e. questions where the physics is just used as a thin shell for a
math problem. For example, ideal gas heat engine problems are pretty routine, so some exam
writers try to spice them up by adding a mathematical twist. They might give you a standard
P V diagram where “the axes have been lost”, and you somehow have to construct them with
ruler and compass given some contrived partial information. Or, pulley problems are pretty
2
easy, so why not make them harder by having infinitely many pulleys stacked on top of each
other, so that students have to sum infinite series? The problem is that none of these things
resemble anything a physicist would ever encounter. There actually are plenty of interesting
questions involving ideal gases and pulleys, but you can’t get them by starting with trivial
questions and stacking on mathematical complications.
• Few “one weird trick” questions. A lot of easier Olympiad problems are just cookie cutter
problems that also require a single key idea. As a very simple example, in many problems that
involve objects initially arranged at the vertices of a regular polygon, you can use symmetry to
conclude that this always remains true. These ideas can be quite beautiful, and they’re often
important in more advanced physics. But I think it’s sufficient to see each one just once or
twice, in the simplest possible context.
• No “two weird trick” questions. Sometimes people construct tough problems by pasting together
simpler ones, but I don’t think that’s a lot of fun, because the whole doesn’t end up greater than
the sum of the parts. A great tough problem should contain an irreducible nugget of insight: a
completely new idea that is already as simple as it can possibly be, yet still difficult to find.
You might be wondering: if you can’t make great problems by tweaking existing setups, adding
tricks to them, or combining existing tricks, then how are they created at all? Usually, they are
produced by subtraction, not addition. A physicist spots something interesting, such as a setup
commonly used in their research area, an insight from an advanced course, a recent paper, a piece
of technology, or even a children’s toy. Then they think hard about how to present it in the simplest
possible way, so that it can be solved using high school knowledge alone. The benefit of this kind
of question is that when you solve them, you’re learning something real, whether it’s a preview of
advanced material, some practical know-how, or an insight into the physics of everyday life.
Can I ask you about the problems?

I don’t give hints for problems, since there are already detailed solutions. If you’re stuck on a tough
problem, I recommend reading a line of the solution and trying to continue from there. If you’re
confused about a basic problem, I recommend reviewing the background reading. I’m happy to
give occasional feedback if you have deeper conceptual questions, but please first make sure your
question is not already answered in the solutions.
What if I find an error in the solutions?

For IPhO and APhO problems, the official solutions often do have errors. For those with major
errors, I usually either avoid assigning the problem entirely, or add a warning in the problem set.
However, I can’t do much about minor errors, such as factors of 2 or minus signs. Please tell me
about IPhO or APhO errors only if they’re major.
For USAPhO problems, the situation is different. When I was in high school, the official solutions
were in pretty bad shape: about half of the solutions had some error in them, and many of the
explanations were incomplete. A few years ago, I rewrote all of the official solutions from 2008 to
2019, and I believe they’re currently almost error-free and much easier to understand. If you’ve
spotted a error I missed, I would be very interested in hearing about it! The same goes for any
error in the handout solutions, even if it’s just a minor typo. Learning new things is hard enough
without typos, and you’ll be helping clear the road for future students.
3
Preliminary Problems
These basic problems should be approachable if you understand the core material in Halliday,
Resnick, and Krane. If you can solve at least 75% of these completely and correctly, you’re ready to
start the main problem sets. Work carefully: many of the problems are more subtle than they look.
Answers are provided for most questions, so you can check your work; solutions are deliberately not
provided, so that you have the chance to work them out for yourself.
1 Mechanics
These problems can be solved using the material in chapters 1 through 17 of Halliday and Resnick.
[2] Problem 1. As a warmup, take the Force Concept Inventory test. This should take no longer
than 30 minutes. You do not have to justify your answers; just list them.
[1] Problem 2. A sewage worker is using a ladder inside a large, frictionless, horizontal circular
aqueduct. The ladder is of the same length as the diameter of the aqueduct.
(a) First the ladder is placed perfectly vertically and the worker climbs to the midpoint. Draw a
free body diagram indicating all forces on the ladder, and their names. Do the forces balance?
(b) Now suppose the ladder is placed perfectly horizontally and the worker hangs statically from
the midpoint. Draw a free body diagram indicating all forces on the ladder, and their names.
Do the forces balance? If so, show this explicitly. If not, what happens next?
[1] Problem 3. Consider the two following setups involving pulleys and spring scales. Treat the ropes,
spring scales, and pulleys as massless and the pulleys as frictionless. Model the pulleys as uniform
discs with masses of 5 kg which are fixed by a rigid support.
(a) Draw a free-body diagram for the second setup, showing all external forces on the spring scale.
(b) What are the readings on the two spring scales?
(c) Draw a free-body diagram for the first setup, showing and naming all external forces on the
pulley. In particular, what is the magnitude of the force between the pulley and the rope?
(d) What is the magnitude of the force that must be provided by the support?
[1] Problem 4. A projectile is thrown upward and passes a point A and a point B a height h above.
Let TA be the time interval between the two times the projectile passes point A, and define TB
similarly.
1
(a) Show that g can be measured as

8h
g= .
TA2 − TB2
(b) This procedure probably looks a little contrived. Why is it better than doing something
simpler, such as just dropping the ball and using ∆y = gt2 /2?
[1] Problem 5. Consider the following system of massless pulleys and string, called a “fool’s tackle”.
If the load L has mass m, find the force F needed to keep the system static.
[2] Problem 6. Consider a projectile launched with speed v at an angle θ from the horizontal on a
flat plane.
(a) Find y(x) and the ratio of the range to the maximum height.
(b) What is the maximum θ for which the projectile is always moving away from the thrower?
[2] Problem 7. A wooden isosceles right triangle with uniform mass density is placed on a table, and
a force is applied as shown.
The force is gradually increased until the triangle begins to tip over without sliding. The force is
then removed. Next, the surface is inclined with angle θ. For what range of θ can you be certain
the triangle will not slide down the incline?
[1] Problem 8. A painter of mass M stands on a platform of mass m as shown.
2
He pulls each rope down with force F , and accelerates upward with acceleration a. Find a.
[3] Problem 9. A small block lies at the bottom of a spherical bowl of radius R.
(a) Find the period of small oscillations, assuming no friction. Can you give an intuitive explana-
tion of the simplicity of your answer?
(b) What is the period if the block is replaced with a small uniform ball, with sufficient friction
to roll without slipping?
(c) Now suppose the block moves in a circle, staying a constant height h R above the bottom
of the bowl. Find the period of the motion, assuming no friction.
(d) Again, find the period if the block is replaced with a small uniform ball, with sufficient friction
to roll without slipping.
(e) Now consider part (c) again. Suppose that at some moment, the speed of the block is
instantaneously increased by a small amount. Qualitatively describe the subsequent motion,
e.g. sketch what a top-down view would look like. What if h is not small?
[2] Problem 10. A car accelerates uniformly from rest. Initially, its door is slightly ajar. Calculate
how far the car travels before the door slams shut. Assume the door has a frictionless hinge, a
uniform mass distribution, and a length L from front to back.
[2] Problem 11. Two diametrically opposite points on a ring of mass M and radius R are marked
out. The ring is placed at rest on a frictionless floor. An ant of mass m starts at one point, then
walks horizontally along the ring to the other. Through what total angle does the ring turn?
[2] Problem 12. A baseball player holds a bat, modeled as a uniform rigid rod, horizontally at one
of its ends. Usually, when the baseball hits the bat, the player will feel a sharp jolt in their hands
as the bat recoils. This can be avoided if the baseball hits the “sweet spot”. Where is it?
[2] Problem 13. Maxwell’s wheel is a toy which demonstrates the principle of conservation of energy.
It consists of a uniform disc of mass M and radius R, with a massless axle of radius r. If the wheel
is released from rest, it falls downward, in such a way so that the strings supporting it are always
perfectly vertical. Find the acceleration.
3
[3] Problem 14. A cue ball is a uniform sphere of radius R.
(a) Find the height at which the cue ball must be hit horizontally so that it immediately begins
rolling without slipping.
(b) Skillful players can hit the cue ball so that it begins moving forwards, but then ends up moving
backwards. Model the hit as an instantaneous impulse applied at an arbitrary point on the
back half of the cue ball, in an arbitrary direction. For what impulses will this trick work?
You can treat the situation as two-dimensional; justify your answer carefully.
[4] Problem 15. Six identical uniform rods, fastened at their ends by frictionless pivots, form a regular
hexagon and lie on a frictionless surface.
A blow is given at a right angle to the midpoint of the bottom rod. Immediately afterward, the
bottom rod has velocity u, as shown. Find the speed of the opposite rod at this moment.
[2] Problem 16. Several possible elliptical orbits of a satellite are shown below.
(a) Which orbit has the most angular momentum?
(b) Which orbit has the highest total energy?
(c) On which orbit is the largest speed acquired?
In all cases, justify your answer carefully.
[2] Problem 17. A comet passes by Sun as shown, in a parabolic path.
4
How long, in years, does the comet take to get from point A to point B? (Hint: if you apply Kepler’s
laws and properties of conics, this problem can be done with almost no computation.)
[2] Problem 18. Because of the rotation of the Earth, the line of a plumb bob will not align with
the local gravitational field. Find the (small) angle of deviation between them as a function of the
latitude θ, the gravitational acceleration g, the radius R of the Earth, and its angular velocity ω.
[2] Problem 19. You should be comfortable with setting up multiple integrals. Consider a cylindrical
shell whose axis of symmetry is the z-axis. It has non-uniform mass per unit area σ(φ, z) in
cylindrical coordinates, and the shell has radius a and height h, with the bottom edge at z = 0.
(a) Write down an integral that gives the total mass of the shell.
(b) Write down an integral that gives the moment of inertia of the shell about the z-axis.
(c) How would these results change if we had a solid cylinder with mass per unit volume ρ(φ, z, r)?
[2] Problem 20. An entrepreneur proposes to propel the Earth through space by attaching many
balloons to one side of it with ropes. The balloons will experience a buoyant force, which will
create a tension in the ropes. Now consider the forces on the solid Earth. Because the atmospheric
pressure on the surface is uniform, the only net force on the Earth is from the tension, so the Earth
will get propelled.
Is this correct? If you think it is, explain why momentum conservation isn’t violated. If you
think it isn’t, identify the specific force acting on the solid Earth that cancels the tension force.
2 Problem Solving Skills

You should be able to start these questions without any background reading. However, for some
fun background on estimation, see Guesstimation: Solving the World’s Problems on the Back of
a Napkin. For practical tips for real experiments, see chapter 7 of Physics Olympiad: Basic to
Advanced Exercises.
[2] Problem 21. Argon atoms are special because they stay in the atmosphere for a very long time.
They are not recycled like oxygen and nitrogen. An average breath inhales around 0.5 L of air and
people breath on average around once every five seconds. Air is about 1% argon and has density
1.2 kg/m3 . Assume all air particles have a mass of approximately 5 × 10−26 kg. Take the atmosphere
to have constant density and be around 20 km thick. The radius of the Earth is 6.4 × 106 m.
(a) Estimate the total number of distinct argon atoms inhaled by Galileo throughout their life.
5
(b) Assuming the atmosphere has been uniform mixed since then, estimate the number of argon
atoms in each of your breaths that were once in Galileo’s lungs.
[3] Problem 22. The acceleration due to gravity can be measured by measuring the time period of a
simple pendulum. However, it can be challenging to get an accurate result.
(a) Suppose you constructed a pendulum using regular household materials. Name at least five
sources of possible experimental error in your calculated value of g. How would you make the
pendulum and perform the measurements to minimize these sources of error?
(b) Make an actual pendulum yourself and carry out the measurement. Describe your experimental
procedure and show your data. Estimate as many of the sources of experimental error identified
in part (a) as you can, and using them, give a value of g with a reasonable uncertainty.
(c) If you had $1,000 and a week to do plenty of measurements, how would you go about it?
How precise a result do you think you could get? What would be the dominant sources of
uncertainty remaining?
[2] Problem 23. Blackbody radiation is an electromagnetic phenomenon, so the radiation intensity
depends on the speed of light c. It is also a thermal phenomenon, so it depends on the thermal
energy kB T , where T is the object’s temperature and kB is Boltzmann’s constant. And it is a
quantum phenomenon, so it depends on Planck’s constant h.
(a) Using the relation E = hf , find the dimensions of h.
(b) Using dimensional analysis, show that the power emitted by the blackbody per unit area,
called the radiation intensity I, obeys I ∝ T 4 , and find the constant of proportionality up to
a dimensionless constant.
(c) How would the result change in a world with d spatial dimensions?
(d) The result you derived in part (b) is known as the Stefan–Boltzmann law. But if it can be
derived with pure dimensional analysis, without needing any detailed calculations or experi-
mental data at all, then why is it considered a law at all? Isn’t it obvious?
3 Electromagnetism
[2] Problem 24. As a warmup, take the Conceptual Survey of Electricity and Magnetism test. This
should take no longer than 1 hour. You do not have to justify your answers; just list them.
[2] Problem 25. A half-infinite line has linear charge density λ.
(a) Find the electric field at a point that is “even” with the end, a distance ` from it, as shown.
6
(b) You should find the direction of the field is independent of `. Explain why.
(c) Sketch the electric field lines everywhere.
[2] Problem 26. Some basic tasks involving intuition for vector fields.
(a) Consider the vector field

v = 2x̂ + xŷ.
Sketch some field vectors at regular points. Then, on a separate sketch, draw some field lines.
(b) Some electric field vectors in a certain situation are shown below.
Sketch a corresponding field line diagram. Then, give a mathematical expression that could
describe this field, and a physical situation which could produce it.
(c) The electric field lines in another situation are shown below.
Sketch a corresponding set of field vectors at regular points. Then, give a mathematical
expression that could describe this field, and a physical situation which could produce it.
[3] Problem 27. A parallel plate capacitor of capacitance C is placed in a region of zero electric field.
The first plate is given total charge Q1 and the second plate is given total charge Q2 . Let the plates
have area A and separation d, where A d2 .
7
(a) Each of the plates has an inner and outer surface. Using Gauss’s law, find the total charge
on each of these four surfaces.
(b) Find the potential difference between the plates.
(c) Find the force between the plates.
(d) In addition to the force between the plates found in part (c), there is a contribution to the
internal stress (force per unit area) within each plate due to the charges on its two surfaces.
Find this part of the stress for each plate, and assuming Q1 > Q2 > 0, indicate whether it is
tension or compression.
[3] Problem 28. A battery is connected to an RC circuit as shown.
The switch is initially open, and the charge on the capacitor is initially zero. The switch is closed
at t = 0.
(a) Solve for the charge on the capacitor as a function of time.
(b) Solve for the power dissipated in the resistor as a function of time.
(c) What is the total energy dissipated in the resistor over all time? Can you find a simple way
to derive this result?
(d) Suppose that the emf E(t) supplied by the battery can be adjusted freely over time, and the
capacitor must be given a total charge Q by the time t = T . Sketch the profile E(t) that
maximizes the efficiency of this process, i.e. the ratio of the energy stored in the capacitor to
the energy output by the battery, and find this efficiency.
[2] Problem 29. In this problem we estimate the maximum firing speed of a human neuron. Model
a human cell simply as a sphere of radius 10−6 m.
(a) It has been measured that 1 cm2 of cell membrane has a resistance of 1000 Ω. Estimate the
resistance of a single human cell.
(b) Estimate the capacitance of a single human cell, treating the two sides of the membrane as
capacitor plates. You will have to estimate the thickness of the cell membrane.
(c) By modeling the cell as an RC circuit, estimate the maximum firing speed of a human neuron.
Is this a reasonable result? If yes, how do you know? If not, how could this model be refined?
[2] Problem 30. An infinite solenoid with radius b has n turns per unit length. The current varies
in time according to I(t) = I0 cos ωt. A ring with radius r < b and resistance R is centered on the
solenoid’s axis, with its plane perpendicular to the axis.
8
(a) What is the induced current in the ring?
(b) A given little piece of the ring will feel a magnetic force. For what values of t is this force
maximum? At this moment, sketch the electric field everywhere.
(c) What is the effect of the force on the ring? That is, does the force cause the ring to translate,
spin, etc.?
(d) If the current is driven by the AC power from a wall outlet, which has frequency 60 Hz in
America, the ring will emit a humming sound. What is the frequency of this sound?
[2] Problem 31. A long, insulating cylinder with radius r and uniform surface charge density σ on
its outer surface rotates about its symmetry axis with angular velocity ω.
(a) Find the magnetic field everywhere.
(b) A wire is connected to a point on the cylinder, with the other end on the axis of rotation.
The wire rotates along with the cylinder. Show that the emf across the wire does not depend
on the shape of the wire’s path, and find this emf.
[2] Problem 32. A rectangular loop of wire with dimensions a and b is placed with one side parallel
to a long, straight wire carrying current I0 , at a distance l.
The resistance of the loop is R. The current in the long wire is quickly switched off.
(a) What is the net momentum p acquired by the loop?
(b) How is momentum conserved in this setup?
[1] Problem 33. A 120 V rms, 60 Hz line provides power to a 40 W light bulb. By what factor will
the brightness change if a 10 µF capacitor is connected in series with the light bulb?
4 Thermodynamics
[2] Problem 34. Two moles of a monatomic ideal gas are taken through the following cycle.
• The gas begins at point A with pressure P0 and volume V0 .
• The gas is heated at constant volume until it doubles its pressure, reaching point B.
9
• The gas is expanded at constant pressure until it doubles its volume, reaching point C.
• The gas is cooled at constant volume until it halves its pressure, reaching point D.
• The gas is compressed at constant pressure until it halves its volume, returning to point A.
Assume that all processes are quasistatic and reversible.
(a) Draw the process on a P V diagram.
(b) Calculate the net work done by the gas during the cycle.
(c) Calculate the efficiency of the cycle.
(d) Calculate the change in entropy of the gas as the system goes from state A to state D.
[3] Problem 35. Deriving some basic results in thermodynamics.
(a) Starting from the first law of thermodynamics, derive the fact that P V γ is constant in an
adiabatic process.
(b) Using the ideal gas law, derive the total work done by a gas as it expands at constant
temperature from volume V1 to V2 , in terms of n, R, T , V1 , and V2 .
(c) Show that if a general gas, not necessarily ideal, satisfies the equation P V = kU , where U is
the total internal energy, then P V n is constant in an adiabatic process for some power n, and
find n in terms of k.
(d) Does a monatomic ideal gas satisfy P V = kU ? If so, what is k?
(e) Two Carnot engines operate with the same minimum and maximum pressures, temperatures,
and volumes. One uses helium as its working substance, and the other uses air. (At the
relevant temperatures, helium behaves like a monatomic gas.) Which one performs more work
per cycle, and by what factor?
[1] Problem 36. A monatomic ideal gas is adiabatically compressed to 1/8 of its original volume. For
each of the following quantities, indicate by what factor they change.
(a) The rms velocity vrms .
(b) The mean free path λ.
(c) The average time between collisions τ for each gas molecule.
(d) The molar heat capacity Cv .
[2] Problem 37. A simple heat engine consists of a movable piston in a cylinder filled with an ideal
monatomic gas. Initially the gas in the cylinder is at a pressure P0 and volume V0 . The gas is slowly
heated at constant volume until the pressure is 32P0 . The gas is then adiabatically expanded until
its pressure is P0 again. Finally, the gas is cooled at constant pressure until its volume is V0 again.
Find the efficiency of the cycle.
10
[2] Problem 38. The total mass of a hot-air balloon (envelope, basket, and load) is 320 kg. Initially
the air pressure inside and outside the envelope is 1.01 × 105 Pa and its density is 1.29 kg/m3 . In
order to raise the hot-air balloon, a gas burner is used to heat the air inside the balloon. The
volume of the envelope filled with hot air is 650 m3 . The molar mass of air is 29 g/mol. Treat the
temperature of the air in the balloon as uniform.
(a) The balloon can either be tightly sealed, so that none of its air mixed with the outside air,
or have a hole, so that its pressure equalizes with that of the outside air. For the purpose of
generating lift, which is better?
(b) Assuming the better option has been taken, to what temperature must the air inside the
balloon be heated to make the balloon begin to rise?
[3] Problem 39. Water is heated in an electric kettle. At a certain moment of time, a piece of ice at
temperature T0 = 0 ◦ C was put in the kettle. The figure below shows the water temperature as a
function of time.
Find the mass of the ice if the heating power of the kettle is P = 1 kW. The latent heat of melting
for ice is L = 335 kJ/kg, the heat capacity of water is c = 4.2 kJ/kg K, and the temperature of the
room is T1 = 20 ◦ C. (Hint: it’s very easy to get an answer that’s off by up to 50% if you’re careless.)
5 Relativity and Waves

These problems can be solved using the material in chapters 18 through 20, and 39 through 44 of
Halliday and Resnick.
[2] Problem 40. Two bombs lie on a train platform, a distance L apart. As a train passes by at
speed v, the bombs explode simultaneously (in the platform frame) and leave marks on the train.
Due to the length contraction of the train, we know that the marks on the train are a distance γL
apart when viewed in the train’s frame, because this distance is what is length-contracted down
to the given distance L in the platform frame. How would someone on the train quantitatively
explain why the marks are a distance γL apart, considering that the bombs are a distance of only
L/γ apart in the train frame?
[1] Problem 41. An atom at rest with rest mass m radiates a photon with frequency ω. What is the
rest mass of the atom afterward?
[3] Problem 42. A rope of linear mass density σ is hung between two poles, a distance L apart, and
the middle of the wire sags a distance d L below the ends. Find the approximate frequencies
11
of standing waves on the rope. (Hint: you do not have to solve for the shape of the rope. As an
intermediate step, it will be useful to consider torque balance on half of the rope.)
[3] Problem 43. A uniform string of length L and linear density ρ is stretched between two fixed
supports. The tension in the string is T .
(a) Find the standing wave solutions and angular frequencies for the given boundary conditions.
(b) A very small mass m is now placed a distance ` from one end of the string. Find the change
in the angular frequencies to first order in m, by using the fact that the average potential
energy and average kinetic energy for each standing wave should remain equal.
(c) How would you go about finding the exact standing wave frequencies for this setup?
[2] Problem 44. A perfectly flat piece of glass is placed over a perfectly flat piece of black plastic.
They touch at point A. Light of wavelength 600 nm is incident normally from above. The location
of the dark fringes in the reflected light is shown above.
(a) How thick is the space between the glass and plastic at B?
(b) Water with n = 1.33 seeps into the region between the glass and the plastic. How many dark
fringes are seen when all of the air has been displaced by water? The straightness and equal
spacing of the fringes is an accurate test of the flatness of the glass.
(c) A setup like this one was used by the physicist Otto Wiener to measure the wavelength of
light. When setting up the experiment, one must decide on the angle between the two objects.
What are the advantages and disadvantages of making this angle smaller, for the purposes of
measuring the wavelength?
[2] Problem 45. A point source of light L emitting a single wavelength λ is situated a distance d
from an ideal mirror, at z = 0. A screen stands at the end of the mirror at distance D d from L.
Find the relative intensity of light on the screen as a function of z. (This setup is known as Lloyd’s
mirror.)
12
Problem Solving I: Mathematical Techniques

For the basics of dimensional analysis and limiting cases, see chapter 1 of Morin or chapter 2 of Order
of Magnitude Physics. Many more examples are featured in The Art of Insight; some particularly
relevant sections are 2.1, 5.5, 6.3, 8.2, and 8.3. Other sections will be mentioned throughout the
course. There is a total of 81 points.
1 Dimensional Analysis
Idea 1
The dimensions of all physical equations should match on both sides. Sometimes, this
constraint alone can determine the answers to physical questions.
Even a statement such as “the speed v is slow” does not make sense, since we need to compare
it to something else with dimensions of speed. A more meaningful statement would be “v is slow
compared to the speed of light”, which would correspond to v/c 1.
Example 1: F = ma 2018 B11
A circle of rope is spinning in outer space with an angular velocity ω0 . Transverse waves on
the rope have speed v0 , as measured in a rotating reference frame where the rope is at rest.
If the angular velocity of the rope is doubled, what is the new speed of transverse waves?
Solution
To solve this problem by dimensional analysis, we reason about what could possibly affect the
speed of transverse waves. The result could definitely depend on the rope’s length L, mass
per length λ, and angular velocity ω0 . It could also depend on the tension, but since this
tension balances the centrifugal force, it is determined by all of the other quantities. Thus
the quantities we have are
[L] = m, [λ] = kg/m, [ω0 ] = 1/s.
Since λ is the only thing with dimensions of mass, it can’t affect the speed, because there is
nothing that could cancel out the mass dimension. So the only possible answer is
v0 ∼ Lω0
where the ∼ indicates equality up to a dimensionless constant, which cannot be found by

dimensional analysis alone. In practice, the constant usually won’t be too big or too small,
so Lω0 is a decent estimate of v0 . But even if it isn’t, the dimensional analysis tells us the
scaling: if ω0 is doubled, the new speed is 2v0 .
Example 2
Find the dimensions of the magnetic field.
1
Solution
To do this, we just think of some simple equation involving B, then solve for its dimensions.
For example, we know that F = q(v × B), so
[F ] kg · m 1 1 kg
[B] = = 2
= .
[q][v] s C m/s C·s
[2] Problem 1. Find the dimensions of power, the gravitational constant G, the permittivity of free
space 0 , and the ideal gas constant R.
[1] Problem 2. Derive Kepler’s third law for circular orbits, using only dimensional analysis. (Why
do you think people didn’t figure out this argument 2000 years ago?)
[2] Problem 3. Some questions about vibrations.
(a) The typical frequency f of a vibrating star depends only on its radius R, density ρ, and
the gravitational constant G. Use dimensional analysis to find an expression for f , up to a
dimensionless constant. Then estimate f for the Sun, looking up any numbers you need.
(b) The typical frequency f of a small water droplet freely vibrating in zero gravity could depends
on its radius R, density ρ, surface tension γ, and the gravitational constant G. Argue that at
least one of these parameters doesn’t matter, and find an expression for f up to a dimensionless
constant.
[3] Problem 4. Some questions about the speed of waves. For all estimates, you can look up any
numbers you need.
(a) The speed of sound in an ideal gas depends on its pressure p and density ρ. Explain why we
don’t have to use the temperature T or ideal gas constant R in the dimensional analysis, and
then estimate the speed of sound in air.
(b) The speed of sound in a fluid depends only on its density ρ and bulk modulus B = −V dP/dV .
Estimate the speed of sound in water, which has B = 2.1 GPa.
The speed of waves on top of the surface of water can depend on the water depth h, the wavelength
λ, the density ρ, the surface tension γ, and the gravitational acceleration g.
(c) Find the speed of capillary waves, i.e. water waves of very short wavelength, up to a dimen-
sionless constant.
(d) Find the speed of long-wavelength waves in very deep water, up to a dimensionless constant.
[3] Problem 5 (Morin 1.5). A particle with mass m and initial speed v is subject to a velocity-
dependent damping force of the form bv n .
(a) For n = 0, 1, 2, . . ., determine how the stopping time and stopping distance depend on m, v,
and b.
(b) Check that these results actually make sense as m, v, and b are changed, for a few values of n.
You should find something puzzling going on. (Hint: to resolve the problem, it may be useful
to find the stopping time explicitly in a few examples.)
2
Idea 2
Dimensional analysis applies everywhere. The argument of any function that is not a mono-
mial, such as sin x, must have no dimensions. The derivative d/dx has the opposite dimensions
to x, and the dx in an integral has the same dimensions as x.
Example 3
We are given the integral Z ∞ √

2
e−x dx = π.
−∞
Find the value of the integral Z ∞
2
e−ax dx.
−∞
Solution
In the first equation, x must be dimensionless, so both sides are dimensionless. The second
equation would also be consistent if both x and a were dimensionless, but we can do better.
Suppose we arbitrarily assign x dimensions of length, [x] = m. Then to make the argument
of the exponential dimensionless, a must have dimensions [a] = m−2 . The dimensions of the
left-hand side are [dx] = m. In order to make the dimensions work out on the right-hand
side, we must have Z ∞
2 1
e−ax dx ∝ √ .
−∞ a
To find the value of the constant, treat x and a as dimensionless again. Then we know the
√
answer must reduce to π when a = 1, so
Z ∞ r
−ax2 π
e dx = .
−∞ a
This procedure is completely equivalent to using u-substitution to nondimensionalize every-

thing, but may be faster to see. In general, for all integrals except for the simplest ones,
you should use either dimensional analysis or u-substitution to reduce the integral to a
dimensionless one.
Remark
Consider the value of the definite integral
Z y
2
e−x dx.
−∞
You can try all day to compute the value of this integral, using all the integration tricks
2
you know, but nothing will work. The function e−x simply doesn’t have an antiderivative
in terms of the functions you already know, i.e. in terms of polynomials, exponents and
logarithms, and trigonometric functions (for more discussion, see here).
3
If you ask a computer algebra system like Mathematica, it’ll spit out something like
2
erf(x), which is defined by being an antiderivative of e−x . But is this really an
2
“analytic” solution? Isn’t that just saying “the integral of e−x is equal to the integral
2
of e−x ”? Well, like many things in math, it depends on what the meaning of the word “is” is.
The fact is, the set of functions we regard as “elementary” is arbitrary; we just choose a set
that’s big enough to solve most of the problems we want, and small enough to attain fluency
with. (Back in the days before calculators, it just meant all the functions whose values were
tabulated in the references on hand.) If you’re uncomfortable with erf(x), note that a similar
thing would happen if a middle schooler asked you what the ratio of the opposite to adjacent
sides of a right triangle is. You’d say tan(x), but they could say it’s tautological, because the
only way to define tan(x) at the middle school level is the ratio of opposite to adjacent sides.
Similarly, 1/x has no elementary antiderivative – unless you count log(x) as elementary, but
log(x) is simply defined to be such an antiderivative. It’s all tautology, but it’s still useful.
[1] Problem 6. Evaluate the integral Z

dx
x2 +1
using a trigonometric substitution. Using dimensional analysis, find the integral
Z
dx
.
x + a2
2
[2] Problem 7. In particle physics, it is conventional to set ~ = c = 1, resulting in equations with

seemingly incorrect units. For example, it is said that the mass of the Higgs boson is about 125 GeV,
where 1 eV is the energy gained by an electron accelerated through a voltage difference of 1 V.
Energy doesn’t have the same units as mass, but the units can be fixed by adding appropriate
01W
factors of ~ and c. Do this, and find the mass of the Higgs boson in kilograms.
[3] Problem 8. USAPhO 2002, problem A3.
Example 4
The Schrodinger equation for an electron in the electric field of a proton is
~2 2 e2
− ∇ ψ− ψ = Eψ.
2m 4π0 r
Estimate the size of the hydrogen atom.
Solution
This is yet another dimensional analysis problem: there is only one way to form a length
using the quantities given above. We have
[m] = kg, [~] = J · s = kg m2 s−1 , [e2 /4π0 ] = J · m = kg m3 s−2 .
4
Doing dimensional analysis, the only length scale is the Bohr radius,
4π~2 0
a0 = ∼ 10−10 m.
me2
I’ve thrown in a 4π above because 0 always appears in the equations as 4π0 . The
dimensional analysis would be valid without this factor, but as you’ll see in problem 11, if
you don’t include it then annoying compensating factors of 4π will appear elsewhere.
Classically (i.e. without ~), there is no way to form a length, and hence there should be
no classically stable radius for the atom. (This was one of the arguments used by Bohr to
motivate quantum mechanics; it appears in the beginning of his paper introducing the Bohr
model.) Once we introduce ~, there are three dimensionful parameters in the problem, as
listed above. And there are exactly three fundamental dimensions. So there is only one way
to create a length, which we found above, one way to create a time, one way to create an
energy, and so on. This means that the solutions to the Schrodinger equation above look
qualitatively the same no matter what these parameters are; all that changes are the overall
length, time, and energy scales. In problem 11, you’ll investigate how this conclusion changes
when we add more dimensionful parameters.
Dimensional analysis is especially helpful with scaling relations. For example, a question might ask
you how the radius of the hydrogen atom would change in a world where the electron mass was
twice as large. You would solve this problem in the exact same way as the example above, using
dimensional analysis to show that a0 ∝ 1/m.
[3] Problem 9. In this problem we’ll continue the dimensional analysis of the Schrodinger equation.
(a) Estimate the typical energy scale of quantum states of the hydrogen atom, as well as the
typical “velocity” of the electron, using dimensional analysis.
(b) Do the same for one-electron helium, the system consisting of a helium nucleus (containing
two protons) and one electron.
(c) Estimate the electric field needed to rip the electron off the hydrogen atom.
Idea 3: Buckingham Pi Theorem
Dimensional analysis can’t always pin down the form of the answer. If one has N quanti-
ties with D independent dimensions, then one can form N − D independent dimensionless
quantities. Dimensional analysis can’t say how the answer depends on them.
A familiar but somewhat trivial example is the pendulum: its period depends on L, g, and the
amplitude θ0 , three quantities which contain two dimensions (length and time). Hence we canpform
one dimensionless group, which is clearly just θ0 itself. The period of a pendulum is T = f (θ0 ) L/g.
5
Example 5: F = ma 2014 12
A paper helicopter with rotor radius r and weight W is dropped from a height h in air with
a density of ρ. Assuming the helicopter quickly reaches terminal velocity, use dimensional
analysis to analyze the total flight time T .
Solution
The answer can only depend on the parameters r, W , h, and ρ. There are four quantities in
total, but three dimensions (mass, length, and time), so by the Buckingham Pi theorem we
can form one independent dimensionless quantity. In this case, it’s clearly r/h. Continuing
with routine dimensional analysis, we find
r
2 ρ
T = f (r/h) h .
W
The form of this expression is a bit arbitrary; for instance, we could also have written
f (r/h)r2 in front, or even f (r/h)r37 h−35 . These adjustments just correspond to pulling
factors of r/h out of f , not to changing the actual result.
This is as far as we can get with dimensional analysis alone, but we can go further using
physical reasoning. If the helicopter quickly reaches terminal velocity, then it travels at a
constant speed. So we must have T ∝ h, which means that f (x) ∝ x, and
r
ρ
T ∝ rh .
W
Example 6
An hourglass is constructed with sand of density ρ and an orifice of diameter d. When the
sand level above the orifice is h, what is the mass flow rate µ?
Solution
The answer can only depend on ρ, d, h, and g. The Buckingham Pi theorem gives
p
µ = f (h/d)ρ gd5 .
That’s as far as we can get with dimensional analysis; to go further we need to know more
√
about sand. If we were dealing with an ideal fluid, then the flow speed would
√ be v = 2gh by
√
Torricelli’s law, which means the flow rate has to be proportional to h. Then f (x) ∝ x,
√
giving the result µ ∝ ρd2 gh. This is a good estimate as long as the orifice isn’t so small
that viscosity starts to dominate.
But this isn’t how sand works. Sand is a granular material, whose motion is dominated by
the friction between sand grains. So the higher pressure doesn’t actually propagate to the
orifice, and the flow rate is independent of p
h, which is apparent from watching an hourglass
run. Then f (x) is a constant, giving µ ∝ ρ gd5 , which has been experimentally verified.
6
Remark
One has to be a little careful with the Buckingham Pi theorem. For example, if all we had
were 3 speeds vi , we can form two dimensionless quantities: v1 /v2 and v1 /v3 . (The quantity
v2 /v3 is not independent, since it is the quotient of these two.) But there are 3 quantities
with 2 dimensions (length and time), so we expect only 1 dimensionless quantity.
The problem is that the two dimensions really aren’t independent: for any quantity built
from the vi , a power of length always comes with an inverse power of time, so there’s only
one independent dimension. These considerations can be put on a more rigorous footing in
linear algebra, where the Buckingham Pi theorem is merely a special case of the rank-nullity
theorem. If you’re ever in doubt, you can just forget about the theorem and play with the
equations directly.
Remark
Dimensional analysis is an incredibly common tool in Olympiad physics because it lets you
say a lot even without much advanced knowledge. If a problem ever says to find some
quantity “up to a constant/dimensionless factor”, or how that quantity scales as another
quantity changes, or what that quantity is proportional to, it’s almost certainly asking you
to do dimensional analysis. Another giveaway is if the problem looks extremely technical and
advanced, because they can’t actually be.
[3] Problem 10 (Insight). In this problem we’ll do one of the most famous dimensional analyses of
all time: estimating the yield of the first atomic bomb blast. Such a blast will create a shockwave
of air, which reaches a radius R at time t after the blast. The air density is ρ, and we want to
estimate the blast energy E.
(a) Declassified photographs of the blast indicate that R ≈ 100 m at time t ≈ 15 ms. The density
of air is ρ ≈ 1 kg/m3 . Estimate the blast energy E.
(b) How much mass-energy (in grams) was used up in this blast?
(c) If we measure the entire function R(t), what general form would we expect it to have, if this
dimensional analysis argument is correct?
Remark
The British physicist G. I. Taylor performed the dimensional analysis in problem 10 upon
seeing a picture of the first atomic blast in a magazine. The result was so good that the
physicists at the Manhattan project thought their security had been breached!
During World War II, the exact value of the critical mass needed to set off a nuclear explosion
was important and nontrivial information. The Nazi effort to make a bomb had been stopped
by Werner Heisenberg’s huge overestimation of this quantity, and after the war, the specific
value was kept a closely guarded secret. That is, it was until 1947, when a Chinese physicist
got the answer using a rough estimate that took four lines of algebra.
7
[5] Problem 11. We now consider the Schrodinger equation for the hydrogen atom in greater depth.
We begin by switching to dimensionless variables, which is useful for the same reason that writing
integrals in terms of dimensionless variables is: it highlights what is independent of unit choices.
(a) Define a dimensionless length variable r̃ = r/a0 , where a0 is the length scale found in example 4.
The ∇2 term in the Schrodinger equation is a second derivative, the 3D generalization of
d2 /dx2 . Using the chain rule, argue that
˜ 2 = a2 ∇2
∇ 0
˜ is the gradient with respect to r̃.

where ∇
(b) Similarly define a dimensionless energy Ẽ = E/E0 , using the energy scale E0 found in
problem 9. Show that the Schrodinger equation can be written in a form like
˜ 2 ψ − 1 ψ = Ẽψ
−∇
r̃
Here I’ve suppressed all dimensionless constants, like factors of 2, because they depend on
how you choose to define E0 and don’t really matter at this level of precision.
The result of this part confirms what we concluded above: solutions to the Schrodinger
equation don’t qualitatively depend on the values of the parameters, because they all come
from scaling a solution to this one dimensionless equation appropriately.
(c) This is no longer true in relativity, where the total energy is

p
E = p2 c2 + m2 c4 .
Assuming p mc, perform a Taylor expansion to show that the next term is Ap4 , and find
the coefficient A. (If you don’t know how to do this, work through the next section first.)
(d) In quantum mechanics, the momentum is represented by a gradient, p → −i~∇. (We will see
why in X1.) Show that the Schrodinger equation with the first relativistic correction is
~2 2 e2
− ∇ ψ− ψ + ~4 A∇4 ψ = Eψ.
2m 4π0 r
(e) Since there is now one more dimensionful quantity in the game, it is possible to combine the
quantities to form a dimensionless one. Create a dimensionless quantity α that is proportional
to e2 /4π, then numerically evaluate it. This is called the fine structure constant. It serves as
an objective measure of the strength of the electromagnetic force, because it is dimensionless,
and hence its value doesn’t depend on an arbitrary unit system.
(f) As the number of protons in the nucleus increases, the relativistic correction becomes more
important. Estimate the atomic number Z where the correction becomes very important.
You probably won’t see any differential equations as complex as the ones in the above problem
anywhere in Olympiad physics, but the key idea of using dimensionless quantities to simplify and
01h
clarify the physics can be used everywhere.
[5] Problem 12. IPhO 2007, problem “blue”. This problem applies thermodynamics and dimen-
sional analysis in some exotic contexts.
8
Example 7
Estimate the Young’s modulus for a material with interatomic separation a and typical atomic
bond energy Eb . Use this to estimate the spring constant of a rod of area A and length L,
as well as the speed of sound, if each atom has mass m.
Solution
This example is to get you comfortable with the Young’s modulus Y , which occasionally
comes up. It is defined in terms of how much a material stretches as it is pulled apart,
stress restoring force/cross-sectional area
Y = = .
strain change in length/length
The Young’s modulus is a useful way to characterize materials, because unlike the spring
constant, it doesn’t depend on the shape of the material. For example, putting two identical
springs side-by-side doubles the spring constant, because they both contribute to the force.
But since the stress is the force per area, it’s unchanged. Similarly, putting two identical
springs end-to-end halves the spring constants, because they both stretch, but since the
strain is change in length per length, it’s unchanged. So you would quote a material’s
Young’s modulus instead of its spring constant, for the same reason you would quote a
material’s resistivity instead of its resistance.
We note that Y has the dimensions of energy per length cubed, so

Eb
Y ∼
a3
solely by dimensional analysis. (Of course, for this dimensional analysis to work, one
has to understand why Eb and a are the only relevant quantities. It’s because Y , or
equivalently the spring constant k, determines the energy stored in a stretched spring.
But microscopically this comes from the energy stored in interatomic bonds when
they’re stretched. So the relevant energy scale is the bond energy Eb , and the relevant
distance scale is a, because that determines how many bonds get stretched, and by how much.)
To relate Y to the spring constant of a rod, note that

F/A L F L
Y = = =k
∆L/L A ∆L A
for a rod, giving the estimate k ∼ AEb /La3 . This is correct to within an order of magnitude!
To relate Y to the speed of sound, note that the sound speed, like most wave speeds, depends
on the material’s inertia and its restoring force against distortions. Since the speed of
sound doesn’t depend on the extrinsic features of a metal object, such as a length, both of
these should be measured intrinsically. The intrinsic measure of inertia is the mass density
ρ ∼ m/a3 , while the intrinsic measure of restoring force is just Y . By dimensional analysis,
s s r
Y Eb /a3 Eb
v∼ ∼ 3
∼ .
ρ m/a m
9
This is also reasonably accurate. For example, in diamond, Eb ∼ 1 eV (a typical atomic energy
scale), while a carbon nucleus contains 12 nucleons, so to the nearest order of magnitude,
m ∼ 10mp , where a useful fact is mp ∼ 1 GeV/c2 . Thus,
r
1 eV
v∼ c ∼ 10−5 c ∼ 3 km/s
1010 eV
which is roughly right. (The true answer is 12 km/s.)
Amazingly, we can get an even rougher estimate of v for any solid in terms of nothing besides
fundamental constants. To be very rough, the binding energy is on the order of that of
hydrogen. As you found in problem 9, this is, by dimensional analysis,
2
1 e2 e2

Eb ∼ ∼ me .
4π0 a0 4π0 ~
We take the nuclear mass to be very roughly the proton mass mp , which gives
s 2
e2

v me me
r
∼ ∼α
c mp 4π0 ~c mp
where α is as found in problem 11. This expresses the speed of sound in terms of the
dimensionless strength of electromagnetism α, the electron to proton mass ratio, and the
speed of light. Of course, the approximations we have made here have been so rough that
now the answer is off by two orders of magnitude, but now we know how the answer would
change if the fundamental constants did.
Estimates as simple as these can be surprising to even seasoned physicists: in 2020, the
simple estimate above was rediscovered and published in one of the top journals in science.
If you want to learn how to do more of these estimates, this paper is a good starting point.
Remark
A warning: from these examples, you could get the idea that dimensional analysis gives
you nearly godlike powers, and the ability to write down the answer to most physics
problems instantly. In reality, it only works if you’re pretty sure your physical system
depends on only about 3 or 4 variables – and the hard part is often finding which
variables matter. For example, as we saw above, you can’t get Kepler’s third law for free
because that requires knowing the dimensions of G, which require knowing that gravity
is an inverse square law in the first place, a luxury Kepler didn’t have. And as another
example, we couldn’t have figured out E = mc2 long before Einstein, as who would
have thought that the speed of light had anything to do with the energy of a lump of
matter? Without the framework of relativity, it seems as irrelevant as the speed of sound or
the speed of water waves. To illustrate this point, we consider two contrasting examples below.
In reality, dimensional analysis is best for problems where it’s easy to see which variables
10
matter, problems where you’re explicitly told what variables matter, and for checking your
work, which you should get in the habit of doing all the time!
Example 8
Cutting-edge archeological research has found that the famed T. Rex was essentially a gigantic
chicken. Suppose a T. Rex is about N = 20 times larger in scale than a chicken. How much
larger is its weight, cross-sectional area of bone, walking speed, and maximum jump height?
Solution
These kinds of biological scaling arguments are fun to think about, though the reliability of
the results is somewhat questionable – if any given scaling law doesn’t quite match data, you
can always think a bit more, and come up with a new argument yielding a different scaling.
But here are a few simple example:
• Since the densities should match, the weight should scale with the volume, so as N 3 .
• Since the maximum compressive pressure that bone can take should be the same, the
bone area should scale with the weight, so also as N 3 . That is, the width of the bones
scales as N 3/2 , while their length L scales only as N . This is one of the reasons it’s
impractical to have huge land animals; the biggest animals now are all whales.
• As a very crude model of walking, we can think of the legs as swinging like a free
pendulum. The √length of one step is proportional to L,√while √
the period of the steps is
proportional to L. Thus, the walking speed scales as L ∝ N .
• The energy stored in the muscle cells scales with the volume, but the mass also scales
with the volume. Since the jump height satisfies E = mgh where E is the energy stored
by the muscles, h ∝ N 0 . So a dinosaur can’t jump much higher than a human – and
indeed, we can’t jump much higher than fleas can!
There’s an entire literature on these arguments. For instance, this delightful paper argues that
the frequency a furry mammal will shake to dry itself off scales with its mass as f ∝ m−3/16 .
Example 9
A person with density ρ and total energy E stored in their muscles can jump to a height h
in gravity g. How high would they be able to jump in gravity 10g?
Solution
By dimensional analysis, the only possible answer is
1/4
E
h∝
ρg
which means that in gravity 10g, they can jump to height h/101/4 .
11
But this is completely wrong! In gravity 10g, a person wouldn’t be able to jump at all;
they’d be so crushed by their own weight that they wouldn’t even be able to stand. The
actual answer depends on details of the biomechanics of muscles and bone, which involve
more dimensionful quantities than just the total energy E. So, as remarked above, you can’t
solve literally any problem by just listing a few relevant quantities and doing dimensional
analysis – you need to make sure those are the only relevant quantities.
2 Approximations
Idea 4: Taylor Series
For small x, a function f (x) may be approximated as
x2 00 xn (n)
f (x) = f (0) + xf 0 (0) + f (0) + . . . + f (0) + O(xn+1 )
2 n!
where O(xn+1 ) stands for an error term which grows at most as fast as xn+1 .
There are a few Taylor series that are essential to know. The most important are
x2 x3 x2 x3
exp(x) = 1 + x + + + O(x4 ), log(1 + x) = x − + − O(x4 )
2 6 2 3
and the small angle approximations
x3 x2
sin x = x − + O(x5 ), cos x = 1 − + O(x4 ).
6 2
Another Taylor series you learned long before calculus class is
1
= 1 + x + x2 + x3 + O(x4 ).
1−x
Usually you’ll only need the first one or two terms, but for practice we’ll do examples with
more. If any of these results aren’t familiar, you should rederive them!
Example 10
Find the Taylor series for tan x up to, and including the fourth order term.
Solution
By the fourth order term, we mean the term proportional to x4 . (Not the fourth nonzero
term, which would be O(x7 ).) Of course, tan x is an odd function, so the O(x4 ) term is
zero, which means we only need to expand up to O(x3 ). That means we can neglect O(x4 )
terms and higher everywhere in the computation, subject to some caveats we’ll point out later.
12
By definition, we have
sin x x − x3 /6 + O(x5 )
tan x = = .
cos x 1 − x2 /2 + O(x4 )
However, it’s a little tricky because we have a Taylor series in a denominator. There are
two ways to deal with this. We could multiply both sides by cos x, and expand tan x in
a Taylor series with unknown coefficients. Then we would get a system of equations that
will allow us to solve for the coefficients recursively, a technique known as “reversion of series”.
A faster method is to use the Taylor series for 1/(1 − x). We have
1
= 1 + u + O(u2 )
1−u
and substituting u = x2 /2 − O(x4 ) gives
1 x2
=1+ + O(x4 ).
cos x 2
Therefore, we conclude
tan x = (x − x3 /6 + O(x5 ))(1 + x2 /2 + O(x4 )) = x + x3 /3 + O(x5 ).
Here I was fairly careful with writing out all the error terms and intermediate steps, but as
you get better at this process, you’ll be able to do it faster. (Of course, one could also have
done this example by just directly computing the Taylor series of tan x from its derivatives.
This is possible, but for more complicated situations it’s generally not a good idea, because
computing high derivatives of a complex expression tends to get very messy. It’s better to
just Taylor expand the individual pieces and combine the results, as we did here.)
Remark
Finding series up to a given order can be subtle. For example, if you want to compute an
O(x4 ) term, it is not always enough to expand everything up to O(x4 ), because powers of x
might cancel. To illustrate this, the last step here is wrong:
x3 sin x x4 + O(x6 )
tan x = = 6= x + O(x5 ).
x3 cos x x3 + O(x5 )
[2] Problem 13. Find the Taylor series for 1/ cos x up to and including the fourth order (O(x4 )) term.
[2] Problem 14. Extend the computation above to get the x5 term in the Taylor series for tan x.
[3] Problem 15. For small x, approximate the quantity
x2 e x
−1
(ex − 1)2
to lowest order. That is, find the first nonzero term in the Taylor series. If you don’t take enough
terms in the Taylor series to begin with, you’ll get an answer of zero, indicating you approximated
13
too loosely. But if you take too many, the computation will get extremely messy. (Hint: your
final answer should contain a factor of −1/12. It’s actually the same factor as in the classic result
1 + 2 + 3 + . . . = −1/12. The reason will be explained much later, in an example in X1.)
[2] Problem 16. The function cos−1 (1 − x) does not have a Taylor series about x = 0. However, it
does have a series expansion about x = 0 in a different variable. What is this variable, and what’s
the first term in the series? (Optionally, can you find higher order terms in the series?)
Idea 5: Binomial Theorem

When the quantity xn is small, it is useful to use the binomial theorem,
(1 + x)n = 1 + xn + O(x2 n2 ).
It applies even when n is not an integer. In particular, n can be very large, very small, or
even negative. The extra terms will be small as long as xn is small. If desired, one can find
higher terms using binomial coefficients,
∞
n
X n m
(1 + x) = x
m
m=0
where the definition of the binomial coefficient is formally extended to arbitrary real n.
The binomial theorem is one of the most common approximations in physics. It’s really just taking
the first two terms in the Taylor series of (1 + x)n , but we give it a name because it’s so useful.
[1] Problem 17. Suppose the period of a pendulum is one second, and recall that
s
L
T = 2π .
g
If the length is increased by 3% and g is increased by 1%, use the binomial theorem to estimate
how much the period changes. This kind of thinking is extremely useful when doing experimental
physics, and you should be able to do it in your head.
[1] Problem 18. Consider an electric charge q placed at x = 0 and a charge −q placed at x = d. The
electric field along the x axis is then

q 1 1
E(x) = − .
4π0 x2 (x − d)2
For large x, use the binomial theorem to approximate the field.
[3] Problem 19. Some exercises involving square roots.

√
(a) Manually find the Taylor series for 1 + x up to second order, and verify they agree with the
binomial theorem.
√
(b) Approximate 1 + 2x + x2 for small x using the binomial theorem. Does the result match
what you expect? If not, how can you correct it?
14
Example 11: Birthday Paradox
If you have n people in a room, around how large does n have to be for there to be at least
a 50% chance of two people sharing the same birthday?
Solution
Imagine adding people one at a time. The second person has a 1/365 chance of sharing a
birthday with the first. If they don’t share a birthday, the third person has a 2/365 chance
of sharing a birthday with either, and so on. So a decent estimate for n is the n where

1 2 n 1
1− 1− ... 1 − ≈ .
365 365 365 2
The surprising point of the birthday paradox is that n 365. So we can use the binomial
theorem in reverse, approximating the left-hand side as
2
1 2 1 n 1 n /2

1
1− 1− ... 1 − ≈ 1−
365 365 365 365
which is valid since n/365 is small. It’s tempting to use the binomial theorem again to write
2
1 n /2 n2

1
1− ≈1− =
365 2 · 365 2
which gives n = 19. However, this is a bad approximation, because the binomial theorem only
works if (n2 /2)(1/365) is very small, but here we’ve set it to 1/2, which isn’t particularly small.
Since the series expansion variable is 1/2, each term in the series expansion is roughly 1/2 as
big as the last (ignoring numerical coefficients), so we expect to be off by about (1/2)2 = 25%.
The binomial theorem is an expansion for (1 + x)y which works when xy is small. Here xy
isn’t small, and we instead want an approximation that works when only x is small. One
trick to dealing with an annoying exponent is to take the logarithm, since that just turns it
into a multiplicative factor. Note that
log((1 + x)y ) = y log(1 + x) ≈ yx
by Taylor series, which implies that
(1 + x)y ≈ eyx
when x is small, an important fact which you should remember. So we have

2
1 n /2

2 1
1− ≈ e−n /2(365) =
365 2
and solving gives n = 22.5. We should round up since n is actually an integer, giving n = 23,
which is indeed the exact answer.
15
Remark
Precisely how accurate is the approximation (1 + x)y ≈ eyx ? Note that the only approximate
step used to derive it was taking log(1 + x) ≈ x, which means we can get the corrections by
expanding to higher order. If we take the next term, log(1 + x) ≈ x − x2 /2, then we find
2 y/2
(1 + x)y ≈ eyx e−x .
Note that because we are approximating the logarithm of the quantity we want, the next
correction is multiplicative rather than additive; we’ll see a similar situation with Stirling’s
approximation in T2. Our approximation has good fractional precision as long as x2 y 1.
In the previous example, x2 y/2 = (22.5/365)2 /4 = 0.1%, so our answer was quite accurate.
[2] Problem 20. Find a series approximation for xy , given that y is small and x is neither small nor
exponentially huge. (Hint: to check if you have it right, you can try concrete numbers, such as
y = 0.01 and x = 10. The series expansion variable may look a bit unusual.)
Remark
If these questions seem complicated, rest assured that 90% of approximations on the USAPhO
and IPhO boil down to using
sin x ≈ x, cos x ≈ 1 − x2 /2, (1 + x)n ≈ 1 + xn, ex ≈ 1 + x, log(1 + x) ≈ x.
I’ve given you a lot of subtle situations above, but it’s these that you have to know by heart.
Almost all situations where you will use these will look like problem 17 or problem 18.
3 Solving Equations
Idea 6
In Olympiads, you may have to find numeric solutions for equations that can’t be solved
analytically. A simple but reliable method is to “guess and check”, starting with a reasonable
first guess (e.g. derived by solving an approximated version of the equation, or sketching the
graphs of both sides), plugging it into both sides, then proceeding with binary search.
[3] Problem 21. Sometimes, you can get an accurate numeric answer very quickly on a basic calculator
by using the method of iteration, which solves equations of the form x = f (x).
(a) Take a scientific calculator (in radians), put in any number, and press the “cos” button many
times. Convince yourself that the final number you get is the unique solution to x = cos x.
(b) What are the key features of the graphs of x and cos x that made this work? For example, why
doesn’t pressing cos−1 repeatedly give the same result? As another example, since x = sin x
has a unique solution, why does repeatedly pressing sin not work so well?
16
(c) Find a nonzero solution for x = tan(x/2).
(d) Find a nonzero solution for ex − 1 = 2x.
(e) Find a positive solution for xx = e.
[2] Problem 22. [A] Newton’s method is a more sophisticated method for solving equations, which
converges substantially faster than binary search. Suppose we want to solve the equation f (x) = 0.
Starting with a nearby guess x0 , we evaluate f (x0 ) and f 0 (x0 ), then find our next guess by applying
the tangent line approximation at this point,
f (x0 )
x1 = x0 − .
f 0 (x0 )
The process repeats until we get a suitably accurate answer.
(a) Use Newton’s method to solve x = cos x.
(b) Newton’s method converges quadratically, in the sense that for typical functions, if your
current guess is away from the answer, the next guess will be O(2 ) away. (This implies that
the number of correct digits in the answer roughly doubles with each iteration!) Explain why,
and then find an example where Newton’s method doesn’t converge this fast.
Newton’s method is very important in general, but it’s not that useful on Olympiads, because it
takes a while to set up, especially if the derivative f 0 is complicated, and you usually don’t need
that many significant figures in your answer anyway. (There are alternatives to Newton’s method,
such as Halley’s method, that converge even faster, but the tradeoff is the same: each iteration
takes more effort to calculate, as higher derivatives of f must be computed.)
Remark
You’ve seen several approximation methods above, and going forward, you should feel free
to use whichever looks best in each situation. However, if you’re solving problems using
the same calculator you use for schoolwork, you should make sure to not rely on its more
advanced features. In Olympiads, you’re generally only allowed to use an extremely basic
scientific calculator, with a tiny display and no memory except for the “Ans” key.
Example 12
In units where c = 1, the Lorentz factor is defined as

1
γ=√ .
1 − v2
Suppose that a particle traveling very close to the speed of light has γ = 108 . Find the
difference ∆v between its speed and the speed of light.
17
Solution
This problem looks easy; by some trivial algebra we find
p
∆v = 1 − 1 − 1/γ 2 .
But when you plug this into a cheap scientific calculator, you get zero, or something that’s
quite far from the right result. The problem is that we are trying to find a small quantity ∆v
by subtracting two nearby, much larger quantities. But the calculator has limited precision,
and it ends up rounding 1 − 1/γ 2 = 1 − 10−16 a bit, giving a completely wrong answer!
Instead, we can apply the binomial theorem to find

1
∆v ≈ + O(1/γ 4 ).
2γ 2
This is no longer the exact answer, but it’s a great approximation, because the error term is
around 1/γ 2 ∼ 10−16 times as small as the answer, and it’s easy for a calculator to evaluate.
The lesson is that it’s better to be accurate in practice than to be precise in theory.
[1] Problem 23. Find the solutions of the equation x2 − 1020 x + 1 = 0 to reasonable accuracy.
[4] Problem 24. [A] Consider the equation x3 − x2 + 1 = 0, where is small. Find approximate
expressions for all three roots of this equation, up to and including terms of order .
4 Limiting Cases
Idea 7
Limiting cases can be used to infer how the answer to a physical problem depends on its
parameters. It is primarily useful for remembering the forms of formulas, but can also be
powerful enough to solve multiple choice questions by itself.
Example 13
What is the horizontal range of a rock thrown with speed v at an angle θ to the horizontal?
Solution
This result is easy to derive, but dimensional analysis and extreme cases can be used to
recover the result too. The answer can only depend on v, g, and θ, so by dimensional analysis
it is proportional to v 2 /g. This is sensible, since the range increases with v and decreases
with g. Now, the range is zero in the extreme cases θ = 0 and θ = π/2, but not anywhere in
between, so if we remember the range contains a simple trigonometric function, it must be
sin(2θ), so
v2
R∝ sin(2θ).
g
18
We can also get the prefactor by a simple limiting case, the case θ 1. In this case, by the
small angle approximation,
vx ≈ v, vy ≈ vθ.
The time taken is t = 2vy /g, so the range is
2v 2
R ≈ vx t = θ.
g
Thus there is no proportionality constant; the answer is
v2
R= sin(2θ).
g
In reality, it’s probably faster to go through the full derivation than all of this reasoning, but
if you’re just not sure about whether it’s a sine or a cosine, or what the prefactor is, then
limiting cases can be quickly used to recover that piece. Also note that the approximations
we used above are frequently useful for evaluating limiting cases.
Example 14
Consider an Atwood’s machine with masses m and M , and a massless pulley. Find the
tension in the string.
Solution
Since the equations involved are all linear equations, we expect the answer should also
be simple. It can only depend on g, m, and M , so by dimensional analysis, it must be
proportional to g. By dimensional analysis, this must be multiplied by something with one
net power of mass. Since the answer remains the same if we switch the masses, it should be
symmetric in m and M .
Given all of this, the simplest possible answer would be

T ∝ g(m + M ).
To test this, we consider some limiting cases. If M m, the mass M is essentially in freefall,
so the mass m accelerates upward with acceleration g. Then the tension is approximately
2mg. Similarly, in the case M m, the tension is approximately 2M g. These can’t be
satisfied by the form above.
The next simplest option is a quadratic divided by a linear expression. Both of these must
be symmetric, so the most general possibility is
A(m2 + M 2 ) + BmM
T =g .
m+M
Then the limiting cases can be satisfied if A = 0 and B = 2, giving
2gmM
T = .
m+M
19
[1] Problem 25. Find the perimeter of a regular N -gon, if L is the distance from the center to any
of the sides. By considering a limiting case, use this to derive the circumference of a circle.
[1] Problem 26. Use similar reasoning to find the acceleration of the Atwood’s machine. (We will
show an even easier way to do this, using “generalized coordinates”, in M4.)
[2] Problem 27 (Morin 1.6). A person throws a ball (at an angle of her choosing, to achieve the
maximum distance) with speed v from the edge of a cliff of height h. Which of the below could be
an expression for the maximal range?
s r
gh2 v2 v2h v2 v2 v 2 /g

2gh 2gh
, , , 1 + , 1 + , .
v2 g g g v2 g v2 1 − 2gh/v 2
If desired, try Morin problems 1.13, 1.14, and 1.15 for additional practice.
[2] Problem 28. Consider a triangle with side lengths a, b, and c. It turns out the area of its incircle
can be expressed purely by multiplying and dividing combinations of these lengths. Moreover,
the answer is the simplest possible one consistent with limiting cases, dimensional analysis, and
symmetry. Guess it!
While we won’t have more questions that are explicitly about dimensional analysis or limiting
cases, these are not techniques but ways of life. For all future problems you solve, you should be
constantly checking the dimensions and limiting cases to make sure everything makes sense.
5 Manipulating Differentials
You might have been taught in math class that manipulating differentials like they’re just small,
finite quantities, and treating derivatives like fractions is “illegal”. But it’s also very useful.
Idea 8
Derivatives can be treated like fractions, if all functions have a single argument.
The reason is simply the chain rule. The motion of a single particle only depends on a single
parameter, so the chain rule is just the same as fraction cancellation. For example,
dv d dv dx
= v(x(t)) =
dt dt dx dt
which show that “canceling a dx” is valid. Similarly, you can show that
dy dx
=1
dx dy
by considering the derivative with respect to x of the function x(y(x)) = x.
As a warning, for functions of multiple arguments, the idea above breaks down. For example,
for a function f (x(t), y(t)), the chain rule says
df ∂f dx ∂f dy
= +
dt ∂x dt ∂y dt
20
where there are two terms, representing the change in f from changes only in x, and only
in y. Therefore, when we start studying thermodynamics, where multivariable functions are
common, we will treat differentials more carefully. But for now the basic rules will do.
Remark
Math students tend to get very upset about the above idea: they say we shouldn’t use
convenient notation if it hides what’s “really” going on. And they’re right, if your goal is
to put calculus on a rigorous footing. But in physics we have no time to luxuriate in such
rigor, because we want to figure out how specific things work. The point of notation is to
help us do that by suppressing mathematical clutter. A good notation suppresses as much
as possible while still giving correct results in the context it’s used.
To illustrate the point, note that elementary school arithmetic is itself an “unrigorous” nota-
tion that hides implementation details. If we wanted to be rigorous about, say, defining the
number 2, we would write it as S(1) where S is the successor function, obeying properties
specified by the Peano axioms. And 4 is just a shorthand for S(S(S(1))), so 2 + 2 = 4 means
S(1) + S(1) = S(S(S(1))).
Even this is not “rigorous”, because the Peano axioms don’t specify how the numbers or
the successor function are defined, just what properties they have to obey. To go deeper,
we could define the integers as sets, and operations like + in terms of set operations. For
example, in one formulation,
S(S(S(1))) = {∅, {∅}, {∅, {∅}}, {∅, {∅}, {∅, {∅}}}.
People have seriously advocated for 1st grade math to be taught this way, which has always
struck me as insane. You can always add more arbitrary layers of structure underneath the
current foundation, so such layers should only be added when absolutely necessary.
As another example, in physics we often work with v(t) in one part of a problem, and then
v(x) in another. Mathematicians hate this, because we are denoting two different functions
with the same letter, but we do it because in both cases v is standing for the same physical
quantity, which doesn’t care how we parametrize it. (And if we didn’t do it this way, even a
simple projectile motion problem can end up taking half the alphabet!) As a converse example,
I always keep track of dimensions, but in math textbooks you can often see jarring equations
like “g = −32.” They suppress dimensions, even though all physical quantities have them,
because they’re not relevant to their goals, and that’s completely legitimate. Everybody
always has to forget about something, or else nobody would ever be able to do anything.
Example 15
Derive the work-kinetic energy theorem, dW = F dx.
21
Solution
Canceling the mass from both sides, we wish to show
1
d(v 2 ) = a dx.
2
To do this, note that
1 dx dv
d(v 2 ) = v dv = dv = dx = a dx
2 dt dt
as desired. If you’re not satisfied with this derivation, because of the bare differentials floating
around, we can equivalently prove that F = dW/dx, by noting
dW dv dv dt dv
= mv = mv =m = F.
dx dx dt dx dt
[2] Problem 29. Some more about power.
(a) Use similar reasoning to derive P = F v.
(b) An electric train has a power line that can deliver power P (x), where x is the distance along
the track. If the train starts at rest at x = 0, find its speed at point x0 in terms of an integral
of P (x). (Hint: try to get rid of the dt’s to avoid having to think about the time-dependence.)
[2] Problem 30 (Kalda). The deceleration of a boat in water due to drag is given by a function a(v).
Given an initial velocity v0 , write the total distance the boat travels as a single integral.
[5] Problem 31. A particle in a potential well.
(a) Consider a particle of mass m and energy E with potential energy V (x), which performs
periodic motion. Write the period of the motion in terms of a single integral over x.
(b) Suppose the potential well has the form V (x) = V0 (x/a)n for n > 0. If the period of the motion
is T0 when it has amplitude A0 , find the period when the amplitude is A, by considering how
the integral you found in part (a) scales with A.
(c) Find a special case where you can check your answer to part (b). (In fact, there are two more
special cases you can check, one which requires negative n and negative V0 , and one which
requires V (x) to be replaced with its absolute value.)
(d) Using a similar method to part (a), write down an integral over θ giving the period of a
pendulum with length L in gravity g, without the small angle approximation. Using this,
compute the period of the pendulum with amplitude θ0 , up to order θ02 . (This result was first
published by Bernoulli, in 1749.)
(e) ? Part (d) is the kind of involved computation you might see in a graduate mechanics course.
But if you think you’re really tough, you can go one step further. Consider a mass m oscillating
on a spring of spring constant k with amplitude A. Calculate its period of oscillation up to
order A2 , accounting for special relativity. (Concretely, assume that the spring force doesn’t
change the rest mass m, and has a potential U = kx2 /2. In relativity, the
p force F = −dU/dx
2
still obeys F = dp/dt, but now E = γmc and p = γmv, where γ = 1/ 1 − v 2 /c2 .)
22
6 Multiple Integrals
It’s also useful to know how to set up multiple integrals. This is fairly straightforward, though
technically an “advanced” topic, so we’ll demonstrate it by example. For further examples, see
chapter 2 of Wang and Ricardo, volume 1, or MIT OCW 18.02, lectures 16, 17, 25, and 26.
Idea 9
In most Olympiad problems, multiple integrals can be reduced to single integrals by symmetry.
Example 16
Calculate the area of a circle of radius R.
Solution
The area A is the integral of dA, i.e. the sum of the infinitesimal areas of pieces we break the
circle into. As a first example, let’s consider using Cartesian coordinates. Then the pieces
will be the rectangular regions centered at (x, y) with sides (dx, dy), which have area dx dy.
The area is thus Z Z Z
A = dA = dx dy.
The only tricky thing about setting up the integral is writing down the bounds. The inner
integral is done first, so its bounds depend
√ on the value of x. Since the boundary of the circle
is x2 + y 2 = R2 , the bounds are y = ± R2 − x2 . Thus we have
Z R Z √R2 −x2
A= dx √ dy.
−R − R2 −x2
We then just do the integrals one at a time, from the inside out, like regular integrals,
Z R p Z 1p Z π/2
2 2
A= 2 2
2 R − x dx = 2R 2
1 − u du = 2R cos2 θ dθ = πR2
−R −1 −π/2
where we nondimensionalized the integral by letting u = x/R, and then did the trigonometric
substitution u = sin θ. (To do the final integral trivially, notice that the average value of
cos2 θ along any of its periods is 1/2.)
We can also use polar coordinates. We break the circle into regions bounded by radii r and
r + dr, and angles θ and θ + dθ. These regions are rectangular, with side lengths of dr and
r dθ, so the area element is dA = r dr dθ. Then we have
Z R Z 2π Z R
A= r dr dθ = 2π r dr = πR2
0 0 0
which is quite a bit easier. In fact, it’s so much easier that we didn’t even need to use double
integrals at all. We could have decomposed the circle into a bunch of thin circular shells,
argued that each shell contributed area (2πr) dr, then integrated over them,
Z R
A= 2πr dr = πR2 .
0
23
In Olympiad physics, there’s usually a method like this, that allows you to get the answer
without explicitly writing down any multiple integrals.
Example 17
Calculate the moment of inertia of the circle above, about the y axis, if it has total mass M
and uniform density.
Solution
The moment of inertia of a small piece of the circle is
x2 M
dI = x2 dm = x2 σ dA = dA
πR2
where x2 appears because x is the distance to the rotation axis, and σ is the mass density
per unit area. Using Cartesian coordinates, we have
√
Z R Z R2 −x2
M
I= dx √ x2 dy.
πR2 −R − R2 −x2
The inner integral is still trivial; the x2 doesn’t change anything, because from the perspective
of the dy integral, x is just some constant. However, the remaining integral becomes a bit
nasty. In general, when this happens, we can try flipping the order of integration, giving
Z R Z √R2 −y2
M
I= dy √ x2 dx.
πR2 −R 2
− R −y 2
Unfortunately, this is equally difficult. Both of these integrals can be done with trigonometric
substitutions, as you’ll check below, but there’s also a clever symmetry argument.
Notice that I is also equal to the moment of inertia about the x axis, by symmetry. So if we
add them together, we get
Z Z
2I = (x + y ) dm = r2 dm.
2 2
The r2 factor has no dependence on θ at all, so the angular integral in polar coordinates is
trivial. We end up with Z R
M 1
2I = 2
2πr r2 dr = M R2
πR 0 2
which gives an answer of I = M R2 /4, as expected.
[2] Problem 32. Calculate I in the previous example by explicitly performing either Cartesian integral.
[3] Problem 33. In this problem we’ll generalize some of the ideas above to three dimensions, where
we need triple integrals. Consider a ball of radius R.
24
(a) In Cartesian coordinates, the volume element is dV = dx dy dz. Set up an appropriate triple
integral for the volume.
(b) The inner two integrals might look a bit nasty, but we already have essentially done them.
Using the result we already know, perform the inner two integrals in a single step, and then
perform the remaining integral to derive the volume of a sphere.
(c) In cylindrical coordinates, the volume element is dV = r dr dθ dz. Set up a triple integral for
the volume, and perform it. (Hint: this can either be hard, or a trivial extension of part (b),
depending on what order of integration you choose.)
(d) In spherical coordinates, the volume element is dV = r2 dr sin φ dφ dθ. Set up a triple integral
for the volume, and perform it.
(e) Let the ball have uniform density and total mass M . Compute its moment of inertia about
the z-axis. (Hint: this can be reduced to a single integral if you use an appropriate trick.)
[2] Problem 34. Consider a spherical cap that is formed by slicing a sphere of radius R by a plane,
so that the altitude from the vertex to the base is h. Find the area of its curved surface using an
appropriate integral.
Remark
You might be wondering how good you have to be at integration to do Olympiad physics.
The answer is: not at all! You need to understand how to set up integrals, but you almost
never have to perform a nontrivial integral. There will almost always be a way to solve the
problem without doing explicit integration at all, or an approximation you can do to render
the integral trivial, or the integral will be given to you in the problem statement. The Asian
Physics Olympiad takes this really far: despite R having some of the hardest problems ever
written, they often provide information like “ x dx = xn+1 /(n + 1) + C” as a hint! This
n
is because physics competitions are generally written to make students think hard about
physical systems, and the integrals are just viewed as baggage.
In fact, plain old AP Calculus probably has harder integrals than Olympiad physics. For
example, in those classes everybody has to learn the integral
Z
sec θ dθ = log |sec θ + tan θ| + C
which has a long history. When I was in high school, I was shocked by how the trick for doing
this integral came out of nowhere; it seemed miles harder than anything else taught in the
class. And it is! Historically, it arose in 1569 from Mercator’s projection, where it gives the
vertical distance on the map from the equator to a given latitude. For decades, cartographers
simply looked up the numeric value of the integral in tables, where the Riemann sums had
been done by hand. (They had no chance of solving it analytically anyway, since Napier only
invented logarithms in 1614.) Gradually, tabulated values of the logarithms of trigonometric
functions became available, and in 1645, Bond conjectured the correct result by noticing the
close agreement of tabulated values of each side of the equation. Finally, Gregory proved the
result in 1668, using what Halley called “a long train of Consequences and Complications of
Proportions.” So it took almost a hundred years for this integral to be sorted out! (Though to
25
their credit, they had the handicap of not knowing about differentiation or the fundamental
theorem of calculus; they were finding the area under the curve with just Euclidean geometry.)
Even though Olympiad physics tries to avoid tough integrals, doing more advanced physics
tends to produce them, so physicists often get quite good at integration. By contrast,
Spivak’s calculus textbook for math majors only covers integration techniques in a single
chapter towards the end of the book. He justifies the inclusion of this material by saying:
Every once in a while you might actually need to evaluate an integral [...] For
example, you might take a physics course in which you are expected to be able
to integrate. [...] Even if you intend to forget how to integrate (and you probably
will forget some details the first time through), you must never forget the basic
methods.
That attitude is why physics students frequently win the MIT Integration Bee.
26
Problem Solving II: Experimental Methods

For more about the theory of uncertainty analysis, see this handout, or for a more introductory
take, this handout and this comic. For practical tips for real experiments, especially at the IPhO,
see chapter 7 of Physics Olympiad: Basic to Advanced Exercises. For some entertaining general
discussion, see chapters I-5 and I-6 of the Feynman lectures. There is a total of 81 points.
1 Basic Statistics
Idea 1
If a quantity X has the probability distribution p(x), that means
Z b
the probability that a ≤ X ≤ b is p(x) dx.
a
In particular, the total probability has to sum to one, so

Z ∞
p(x) dx = 1.
−∞
Using the probability distribution, we can calculate expectation values, i.e. averages. For
example, the expectation value of X, also called the mean, is
Z ∞
hXi = xp(x) dx
−∞
while the expectation value of an arbitrary function of X is

Z ∞
hf (X)i = f (x)p(x) dx.
−∞
One especially important quantity is the variance of X, defined as
var X = hX 2 i − hXi2 .
√
The standard deviation is defined by σX = var X. It describes how “spread out” the
distribution of X is, and it will play an important role in uncertainty analysis.
[1] Problem 1. Suppose that x is a length. What are the dimensions of p(x), hXi, var X, and σ?
Example 1
Trains arrive at a train station every 10 minutes. If I arrive at a random time, and X is the
number of minutes I have to wait, what is the standard deviation of X?
1
Solution
We see that X can be anywhere between 0 and 10, with all possibilities equally likely, so
(
1/10 min−1 0 ≤ x ≤ 10,
p(x) =
0 otherwise
where the 1/10 guarantees the total probability is 1. For the rest of this example, we’ll
suppress the units. We have
Z ∞ Z 10
x
hXi = xp(x) dx = dx = 5
−∞ 0 10
which makes sense, as I should have to wait half the maximum time on average, and
Z ∞ Z 10 2
x 100
hX 2 i = x2 p(x) dx = dx = .
−∞ 0 10 3
Then the standard deviation is

p 5
σX = hX 2 i − hXi2 = √ min.
3
[3] Problem 2. Consider an exponentially distributed quantity,

(
ae−ax x ≥ 0,
p(x) =
0 otherwise.
Verify that the total probability is 1, and compute the mean and standard deviation. To perform
the integrals, you will have to integrate by parts.
[2] Problem 3. The purpose of subtracting hXi2 in the variance is to make sure it doesn’t change
when a constant is added to x, since shifting something left or right on the number line shouldn’t
change its spread. Verify that for any constant c, var X = var(X + c).
[3] Problem 4. We say X is normally distributed if

2
p(x) ∝ e−a(x−b) .
For simplicity, let’s shift X so that it’s centered about x = 0, so

2
p(x) ∝ e−ax .
You may use the result given in P1,

Z ∞ √
2
e−x dx = π.
−∞
Find the constant of proportionality in p(x), the mean, and the standard deviation.
2
[2] Problem 5. If two random variables X1 and X2 are independent, then
hX1 X2 i = hX1 ihX2 i.
Use this result to show that
var(X1 + X2 ) = var(X1 ) + var(X2 )
which implies that the standard deviation “adds in quadrature”,

q
σX1 +X2 = σX 2 + σ2 .
1 X2
This is an important result we’ll use many times below.
2 Uncertainty Analysis
Idea 2
When a physical quantity is measured in an experiment and reported as x±∆x, it is uncertain
what the true value of the quantity is. If the quantity has a probability distribution p(x),
then the reported uncertainty ∆x is essentially the standard deviation of p(x).
Remark
In practice, you’ll have to use intuition and experience to assign uncertainties for real mea-
surements. For example, if you’re using a clock that times only to the nearest second, you
might take ∆t = 0.5 s. If you’re using a good ruler, which has millimeter markings, you might
take ∆x = 0.5 mm, though you can actually do a bit better if you look carefully. Of course,
the ultimate test is the results: if you assigned the uncertainties right, your final uncertainty
should encompass the true result most (but not all) of the time.
[2] Problem 6. Suppose x has uncertainty ∆x and y has uncertainty ∆y, where x and y are indepen-
dent. Explain why the uncertainty of x + y is
p
∆(x + y) = (∆x)2 + (∆y)2 .
This is called “addition in quadrature”. What is the uncertainty of x − y? How about x + x?
Remark
Note how this differs from “high school” uncertainty analysis. In school, you might be told
to show uncertainty using significant figures, and when adding two things, to keep only the
figures that are significant in both of them. That corresponds to
∆(x + y) = max(∆x, ∆y)
which is an underestimate. Or, you might be told that the uncertainty needs to encapsulate
3
all the possible values, which implies that
∆(x + y) = ∆x + ∆y
which is an overestimate, since the uncertainties could cancel.
Three students make measurements of the length of a 1.50 m rod. Each student reports an
uncertainty estimate representing an independent random error applicable to the measure-
ment.
• Alice performs a single measurement using a 2.0 m tape measure, to within 2 mm.
• Bob performs two measurements using a wooden meter stick, each to within 2 mm, which
he adds together.
• Christina performs two measurements using a machinist’s meter rule, each to within
1 mm, which she adds together.
Rank the measurements in order of their uncertainty.
Solution
√
The uncertainty in Alice’s measurement is 2 mm. The √ uncertainty in Bob’s is 2 2 mm by
quadrature, while the uncertainty in Christina’s is 2 mm by quadrature. So the lowest
uncertainty is Christina’s, followed by Alice’s, followed by Bob’s.
[2] Problem 7. Given N independent measurements of the same quantity with the same uncertainty, √
xi ± ∆x, find the uncertainty of their sum. Hence show the uncertainty of their average is ∆x/ N .
This result is extremely important, since repeating trials is one of the main ways to reduce
uncertainty. But it’s important to remember that the results derived above hold only for independent
measurements. For example, taking a single measurement, then averaging that single number with
itself 100 times certainly wouldn’t reduce the uncertainty at all!
Idea 3
If x has uncertainty ∆x, and f (x) can be approximated by its tangent line, f (x0 ) ≈ f (x) +
(x0 − x)f 0 (x) within the region x ± ∆x, then f (x) has approximate uncertainty f 0 (x) ∆x.
√
[2] Problem 8. If x has uncertainty ∆x, find the uncertainties of x2 , x, 1/x, 1/x4 , log x, and ex .
[2] Problem 9. The tangent line approximation doesn’t always make sense. For example, suppose that
x is measured to be zero, up to uncertainty ∆x. Show that the above results for the uncertainties
√
of x2 and x give nonsensical results. What would be a more reasonable uncertainty to report?
[2] Problem 10. Consider two quantities with independent uncertainties, x ± ∆x and y ± ∆y.
4
(a) Show that the uncertainty of xy is

s 2 2
∆x ∆y
∆(xy) = xy + .
x y
To do this, start by writing xy as exp(log x + log y).
(b) If we set x = y, then we find

s
∆x 2 √

2 2
∆(x ) = x 2 = 2x∆x.
x
On the other hand, in a previous problem we found ∆(x2 ) = 2x∆x. Which result is correct?
(c) Find the uncertainty of x/y.
[2] Problem 11. A student launches a projectile with speed v = 5 ± 0.1 m/s in gravitational accelera-
tion g = 9.81 ± 0.01 m/s2 . The resulting range is d = 1.5 ± 0.02 m. Given that the launch angle was
less than 45◦ , find the launch angle, with uncertainty, assuming all uncertainties are independent.
[2] Problem 12. Two physical quantities are related by y = xex .
(a) If x is measured to be 1.0 ± 0.1, find the resulting value of y, with uncertainty.
(b) If y is measured to be 2.0 ± 0.1, find the resulting value of x, with uncertainty.
Idea 4
For practical computations, it is often useful to use relative uncertainties. The relative
uncertainty of x is ∆x/x, and can be expressed as a percentage.
[1] Problem 13. Some basic relative uncertainty results.
(a) Show that the relative uncertainty of the product or quotient of two quantities with independent
uncertainties is the square root of the sum of the squares of their relative uncertainties.
(b) Show that averaging

√ N independent trials as in problem 7 reduces the relative uncertainty by
a factor of N .
Example 3
The side lengths of a rectangular prism are measured to be 1.0 ± 0.1 cm, 5.0 ± 0.1 cm, and
10.0 ± 0.1 cm. What is the relative uncertainty of the volume?
Solution
We note that the relative uncertainties of the side lengths are 10%, 2%, and 1%. Hence the
5
relative uncertainty of the volume is

∆V p
= (10%)2 + (2%)2 + (1%)2 ≈ 10%.
V
Since one of the uncertainties is so much larger than the others, it is basically the only one
which matters. This is also a common situation when doing real experiments. Thus, in many
situations it’s possible to compute the uncertainty in your head.
[2] Problem 14. Suppose the goal of an experiment is to measure the ratio T1 /T2 of the durations of
two physical processes, where T1 is about 15 seconds, and T2 is about 3 seconds. Also suppose your
stopwatch is only accurate to the nearest second. You have two minutes to perform measurements.
Assume each measurement is independent.
(a) Using your instinct, figure out whether it’s better to spend more total time measuring T1 ,
more total time measuring T2 , or an equal amount of time on both.
(b) To confirm this, qualitatively sketch the relative uncertainty of T1 /T2 as a function of the
fraction of time x spent measuring T1 , using explicit numeric examples if necessary.
Calculations of this sort are common when doing Olympiad experimental physics. You should be
01^
able to do them instinctively, getting the ballpark right answer without explicit calculation.
[3] Problem 15. Solve F = ma 2018 problems A12, A25, B19, and B25, and F = ma 2019
problems A16, B18, and B25. Make sure to strictly adhere to the total time. Since these are
F = ma problems, you don’t have to produce a writeup. If you find these questions difficult to
finish in the allotted time, go back and review the earlier material!
[3] Problem 16. In the preliminary problem set, you measured g using a pendulum. If you didn’t do
uncertainty analysis for it, as we covered above, then you should go back and estimate uncertainties
more precisely. In this problem you’ll do a different experiment: you will estimate g by finding the
time needed for an object to roll down a ramp, with everything again made of household materials.
(a) Before starting, think about what the dominant sources of uncertainty will be, and how you
can design the experiment to minimize them. In particular, do you think the result will be
more or less precise than your pendulum experiment?
(b) Perform the experiment, taking at least ten independent measurements, and report the data
and results with uncertainty.
[3] Problem 17. At any moment, a Geiger counter can click, indicating that it has detected a particle
of radiation. Suppose that there is an independent probability α dt of clicking at each infinitesimal
time interval dt. Let the number of clicks observed in a total time T be X.
(a) Find the expected value and standard deviation of X, and thereby compute its relative
uncertainty. (Hint: split the total time into many tiny time intervals, and let Xi be the
P
number of clicks in interval i, so X = i Xi .)
(b) Using a Geiger counter on a sample, you hear 197 clicks in 5 minutes of operation. Estimate
the activity α of the sample (i.e. the expected clicks per second), with uncertainty. If you
measure for longer, how does the uncertainty reduce over time?
6
(c) Now suppose that for a different sample, N = 0 after 5 minutes. Estimate the activity α of
the sample (i.e. the expected clicks per second), with a reasonable uncertainty. If you measure
for longer, and continue to hear no clicks, how does the uncertainty reduce over time?
Remark
In this problem set, we have given rules for calculating the mean and standard deviation
of derived quantities. But in general, probability distributions can have all kinds of weird
features, which aren’t captured by those two numbers. The reason we focus on them anyway
is because of the central limit theorem, which roughly states that if we have many independent
random variables, the distribution of the sum will approach a normal distribution. As you
saw in problem 4, normal distributions are characterized entirely by their mean and standard
deviation, so we don’t lose any information by reporting only those two quantities.
[4] Problem 18. [A] This problem extends problem 17 to derive some canonical results.
(a) Let λ = αT . Find the probability p(X = k) of hearing exactly k clicks in terms of λ and k.
(b) To check your result, show that the sum of the p(X = k) is equal to one.
(c) ? In the limit λ 1, show that the probabilities p(X = k) approach that of a normal
distribution with the mean and standard deviation calculated in problem 17, thereby providing
an example of the central limit theorem at work. This is a rather involved calculation, which
will use many of the techniques from P1. It will also require Stirling’s approximation,
√ n n
n! ≈ 2πn
e
for n 1, which will be important in T2. (Hint: because the relative uncertainty falls as λ
increases, start by writing k = λ(1 + δ) for |δ| 1, and expand in powers of δ. Be careful not
to drop too many terms, as δ is small, but λδ isn’t.)
[3] Problem 19. [A] Consider N independent measurements of the same quantity, with results xi ±∆xi .
They can be combined into a single result by taking a weighted average. What is the optimal weighted
average, which minimizes the uncertainty?
Remark
There are many situations where the rules above can’t be used. For example, consider the
uncertainty of x + y 2 /x, where x and y have independent uncertainties. You can calculate the
uncertainty of either term with the standard rules, but you can’t calculate the uncertainty
of their sum, because the terms are not independent (both contain x).
In these cases, you can use the multivariable equivalent of the tangent line approximation,
∂f ∂f
f (x0 , y 0 ) ≈ f (x, y) + (x0 − x) + (y 0 − y) .
∂x ∂y
7
Adding the two contributions to the uncertainty in quadrature gives

s 2 2
∂f ∂f
∆f = ∆x + ∆y .
∂x ∂y
This is the general rule that includes the rules you derived above as special cases. However, it
shouldn’t be necessary in Olympiad problems. For example, if you run into such situations in
an experiment, often one of the uncertainties is much smaller, and can be neglected entirely.
Remark
As you saw in problem 9, the tangent line approximation can sometimes fail. The proper way
to handle situations like these would be to find the full probability distribution of the desired
quantity, rather than just describing it crudely with its standard deviation. However, this
can’t be done analytically except in the simplest of cases. So when professional physicists run
into situations like these, which are quite common, they often just numerically compute a few
million or billion values, starting with randomly drawn inputs each time, and use that to infer
the probability distribution. This technique is called Monte Carlo. It’s very powerful, but
certainly not needed for Olympiads! On Olympiads, you should just fall back to something
reasonable, such as taking the minimum and maximum possible values.
3 Data Analysis
Idea 5
All graphical data analysis for the USAPhO and IPhO can be performed by drawing a line
and measuring its slope and intercept. This is a bit artificial, but it’s necessary because of
the limited calculation equipment you have during these exams. Despite this, drawing lines
can be surprisingly powerful.
Example 4
The activity of a radioactive substance obeys A(t) = A0 e−t/τ . Using measurements of t and
A(t), plot a line to find A0 and τ .
Solution
To handle exponential relationships, take the logarithm of both sides for
log A(t) = log A0 − t/τ.
Then a plot of log A(t) vs. t has slope −1/τ and y-intercept log A0 .
[1] Problem 20. For a power law y = αxn where y and x are measured, what line can be plotted to
find α and n?
8
[2] Problem 21. The rate R of electron emission from a solid in an electric field E is
R = βe−E/E0
for some constants β and E0 . The particular form is because the effect is due to quantum tunneling,
and you will derive it in X2.
(a) If E and R are measured, what line can be plotted to find β and E0 ?
(b) Your answer for part (a) should have formally incorrect dimensions, by the standards of P1.
This often happens when one takes logarithms. What’s going on? If the dimensions are wrong,
how can the result be right?
(c) Suppose both β and E0 have 1% uncertainty. For small E, which is more important for the
uncertainty of R? What about for large E? Around where is the crossover point?
Example 5
Suppose that y and x are related nonlinearly, as
y = bx + ax2 .
For example, this could model the force due to a non-Hookean spring. Using measurements
of x and y, plot a line to find a and b.
Solution
If we divide by x, we find
y
= ax + b.
x
Therefore, we can plot y/x versus x, which gives a line with slope a and intercept b. More
generally, we can plot a line whenever we can rearrange a given relation into the form
(known) = (unknown)(known) + (unknown)
where all four terms can be arbitrarily complicated. In this way, it is possible to turn a lot
of very nonlinear relations into lines.
[3] Problem 22. Some more examples of finding lines to plot.

(a) Suppose that you are given points (x, y) that lie on a circle centered at (a, 0) with radius r.
What line can be plotted to find a and r?
(b) Consider an Atwood’s machine with masses m and M > m. The acceleration of the machine
is measured as a function of M . However, since the pulley has mass, it slows the acceleration
of the Atwood’s machine, so that
M −m
a= g.
M + m + δm
Find a line that can be plotted to find g and δm, assuming m, M , and a are known. This is
an example of how plotting a line can separate out a systematic error, i.e. the value of δm,
which would be impossible if only one value of M were used.
9
(c) Suppose an object is undergoing simple harmonic motion with amplitude A and angular
frequency ω. Given measurements of the position x and velocity v, what line can be plotted
to find A and ω?
Remark
When performing data analysis in practice, you should neatly organize your work. Always
make a data table that explicitly shows what you’re calculating, and make a neat graph with
a ruler and graph paper. Set the axis scale so that the graphed data points cover almost
the entire page, and let the x-axis include x = 0 if you need to find a y-intercept. The
computation of the slope should be explicitly shown. For each line you should use at least
about five points; you don’t have to use them all. If you have a calculator that can find best-fit
slopes for you, don’t use it, as these features are generally not allowed on real Olympiads.
[3] Problem 23. 01W USAPhO 2012, problem A2. (This one requires basic thermodynamics.)
[3] Problem 24. 01W USAPhO 2011, problem A2.
[3] Problem 25. 01c INPhO 2018, problem 7. (This one requires basic fluid dynamics.)
Idea 6
Historically, uncertainty analysis has only appeared on the F = ma, and data analysis has
only appeared on the USAPhO, but the two appear together in the IPhO.
To perform uncertainty analysis for best fit lines, plot the uncertainties of the data points
as error bars. Then draw the steepest and shallowest lines that still pass through most of
the error bars. These will give you the bounds on your slope and intercept. We’ll see some
examples of this procedure in later problem sets. It isn’t the most mathematically rigorous
method, but it gives decent results.
4 Estimation
Estimation is a useful skill for checking the answers to real-world problems.
Example 6
Estimate the circumference of the Earth.
Solution
If you know that the United States is 3,000 miles wide, and there is a time zone difference of
three hours between California and New York, then a reasonable estimate is 24,000 miles.
Or, if you know the factoid that light can go about seven times around the Earth in a second,
then a reasonable estimate is (3/7) × 108 m ≈ 4 × 107 m.
10
Let’s check these results are compatible. There are about 5 miles in 8 kilometers, a fact
you can get by remembering how your car’s speedometer looks, or by noting that 3 feet are
about 1 meter. Then 4 × 104 km ≈ (5/8) × 4 × 104 mi = 2.5 × 104 mi, so the two results are
compatible. There are probably at least a hundred more ways to perform this estimation.
Example 7
Estimate the density of air, and compare this to the density of water.
Solution
We can directly use the ideal gas law, P V = nRT . The density is ρ = µn/V where µ is the
mass of one mole of air, so
µP
ρ= .
RT
Atmospheric pressure is about 105 Pa, typical temperatures are about 300 K, and air is mostly
N2 , which has a molar mass of µ = 28 g/mol, so
(0.028)(105 ) kg kg
ρ= ≈ 1 3.
(8.3)(300) m3 m
The density of water is, almost by definition,

kg
ρw ≈ 103 .
m3
Most liquids and solids have densities within an order of magnitude of this, since in all
cases the atoms are packed close together. Evidently, air molecules are about a factor of
(103 )1/3 = 10 times further apart than typical water molecules.
Example 8
Estimate how much useful power you can produce in a short burst.
Solution
This is a bit tricky to test, because most exercises just burn energy against air resistance or
friction, which is hard to estimate. However, a task that directly performs work is useful. I
weigh about 75 kg and can run up a 3 m high staircase in around 3 s, so
P = mgv = (75)(10)(3/3) W ≈ 750 W.
This is a typical max power output, while typical steady state power outputs are about a
factor of 3 or 4 smaller.
For the below questions, feel free to look up specific numbers if you’re stuck. In all cases, an answer
to the nearest order of magnitude is good enough.
11
[3] Problem 26. Some questions about light energy.
(a) Estimate the number of photons emitted per second by a standard light bulb. (The energy of
a photon is E = hf , and the frequency of a photon is related to the wavelength by c = f λ.)
(b) The Sun supplies power of intensity 1400 W/m2 to the Earth. The nearest star is about 4
light years away. Assuming this star is similar to the Sun, about how many of its photons hit
your eye per second?
[2] Problem 27. Estimate the radius of the largest asteroid you could jump off of, and never return.
[4] Problem 28. Some questions about energy.
(a) Estimate the digestible energy content of a stick of butter. (A calorie is about 4000 J, and is
also the energy needed to raise the temperature of a kilogram of water by 1 K.)
(b) Estimate the rate at which your body burns energy when at rest.
(c) Estimate the rate at which a human being radiates energy. (The Stefan–Boltzmann law states
that the radiation power per unit area from a blackbody is σT 4 , where σ = 5.7×10−8 W/m2 K4 .)
Is radiation a significant source of energy loss for a human being, or is it negligible?
(d) A human being develops hypothermia, with their core body temperature dropping by 5 ◦ F.
Neglecting any heat transfer with the environment, estimate the number of calories required
to raise their temperature back to normal.
Now let’s verify the energy content of the butter microscopically. This will be a very rough estimate,
so expect answers to be only within one or two orders of magnitude.
(e) A chemical bond typically involves two electrons, and a characteristic atomic separation
distance of one angstrom, r ∼ 10−10 m. Estimate the binding energy of one chemical bond.
(f) The fats in butter are digested by inputting energy to break the bonds in the molecules, then
harvesting energy by combining the atoms into CO2 and H2 O, which have somewhat more
stable bonds.
Estimate the energy content of a kilogram of butter. How close is this to the true result?
[2] Problem 29 (Povey). When human beings lose weight, most of it is by exhalation of carbon.
About 20% of the air in the atmosphere is oxygen. When we breathe in and then out, about 25%
of the oxygen is converted to carbon dioxide.
(a) Estimate the mass of air contained in a single breath.
(b) Estimate the amount of weight we lose every day by breathing alone.
[2] Problem 30 (Insight). How long a line can you write with a pencil?
12
5 Experimental Technique
At both the national and international Olympiad level, it’s important to have practical know-how
in order to make experiments work. It’s very hard to train this with only theoretical problems.
However, the Australian Physics Olympiad has some useful problems in this direction, since it has
01^
a strong emphasis on real-world physics.
[3] Problem 31. AuPhO 2010, problem 12.
[3] Problem 32. 01^ AuPhO 2012, problem 14.
[3] Problem 33.

answer sheets.
01Y AuPhO 2016, problem 14. You will need to print out pages 8 and 9 of the
You can look at other AuPhO questions for further practice, but as you can see here, many AuPhO
questions are confusing, misleading, or even wrong, which is an unfortunate consequence of the
innovative nature of the contest. I’ll only assign you the best ones.
13
Mechanics I: Kinematics
See chapters 3 and 4 of Morin for material on solving differential equations. For general review on
kinematics, see chapter 1 of Kleppner and Kolenkow. For fun, see chapters I-1 through I-8 of the
Feynman lectures. There is a total of 87 points.
1 Motion in One Dimension

Example 1
When a projectile moves slowly through air, the drag is linear in the velocity, F = −αmv.
Find the velocity v(t) of a projectile thrown upward at time t = 0 with speed v0 .
Solution
We write Newton’s second law as
dv
= −g − αv
dt
and multiply through by dt. Integrating both sides from the initial condition to time tf gives
Z v(tf ) Z tf
dv
=− dt.
v0 g + αv 0
Performing the integrals gives

v(tf )
1
log(g + αv) = −tf .
α v0
Renaming tf to t and solving for v yields

g −αt
v(t) = e−αt v0 + (e − 1).
α
This renaming is necessary because we don’t want to confuse t, the dummy variable that we
integrating over, with tf , the time at which we want to evaluate the velocity; t ranges from
zero to tf . Unfortunately, often people just call both of these t, so you need to watch out.
[2] Problem 1. Investigating some features of this solution.
(a) By using results from P1, verify that v(t) makes sense for both small times and large times.
(b) If the projectile is then caught at the launch point, did it spend more time going up or down?
(c) Do you think the total time is longer or shorter than for a projectile without drag?
[3] Problem 2. Now assume quadratic drag, F = −αmv 2 , which applies for fast-moving projectiles.
(a) Integrate Newton’s second law to get an implicit equation for v(t) with the same initial
conditions as above. That is, you don’t need to solve for v(t), as it’ll just make things messy.
1
(b) Your equation will only be valid when the projectile is going up; explain why.
(c) Find v(t) for an object released from rest at time t = 0. (Hint: if needed, look up some standard
integrals involving hyperbolic trigonometric functions. But don’t worry about memorizing
the results, since in competitions, any nontrivial integral needed will usually be given to you.)
Some people only call this quadratic case drag; they call the linear case viscous resistance. This is
because they behave fundamentally differently at the microscopic level, as we will explore in M7.
[3] Problem 3. A projectile of mass m is dropped from a height h above the ground. It falls and
bounces elastically, experiencing the same quadratic drag as in problem 2. Find the maximum
height to which it subsequently rises. (Hint: don’t try to use your results from problem 2.)
Example 2
Find how the speed of a rowing boat depends on the number of rowers N .
Solution
A fast-moving boat experiences quadratic friction, so a drag force
F ∝ v2A
where A is the submerged cross-sectional area of the boat. Since the submerged volume
scales as V ∝ N in hydrostatic equilibrium, we have A ∝ N 2/3 . (This is the sketchy step
of the analysis, since the scaling of A depends on how we adjust the shape of the boat as
N increases.) Thus, the power the rowers need to provide scales as P = F v ∝ v 3 N 2/3 , but
we also have P ∝ N . Combining gives the exceptionally weak dependence v ∝ N 1/9 , which
agrees decently with Olympic rowing times.
Idea 1
An ordinary differential equation is any equation involving a quantity x(t) and its derivatives.
In introductory physics, we are usually concerned with a few very simple differential equations,
with the following nice properties.
• The differential equation is at most second order, meaning that it can contain x, ẋ = v,
and ẍ = a, but no higher derivatives. This implies that the solution can be determined by
an initial position and initial velocity. (We’ll focus on second order differential equations
for the rest of this section; most first order differential equations can simply be solved
by separation and integration, as we’ve already seen above.)
• The differential equation is linear, meaning that terms don’t contain products of x, ẋ,
and ẍ. For example, a damped driven harmonic oscillator with time-dependent drag,
mẍ = −b(t)ẋ − kx + f (t)
is a second order linear differential equation. Solutions to such differential equations obey
the superposition principle: if x1 (t) and x2 (t) are both solutions, so is c1 x1 (t) + c2 x2 (t).
2
• The differential equation is homogeneous, meaning that each term is proportional to ex-
actly one power of x or its derivatives. The above differential equation is not homogeneous,
but it would be if we removed the driving f (t).
• The differential equation is time-translation invariant, meaning that no functions of time

appear except for x and its derivatives. The above equation isn’t, but it would be if we
set f (t) and b(t) to constants.
Idea 2
Linear, homogeneous, time-translation invariant differential equations are very special, and
they can all be solved by the exact same method. First, note that we can promote x(t) to
a complex variable x̃(t) and solve the differential equation over the complex numbers. As
long as we have a complex solution, we can recover a real solution by taking the real part.
Now, the method of solution, which works for almost all equations of this form, is to guess a
complex exponential solution
x̃(t) = eiωt .
Plugging this into the differential equation will yield the allowed values of ω, and the general
solution can be found by superposing the complex exponentials.
Example 3
Solve the simple harmonic oscillator, mẍ + kx = 0, using the above principles.
Solution
First, we pass to a complex differential equation,
¨ + kx̃ = 0.
mx̃
We guess x̃(t) = eiωt . Plugging this in and using the chain rule gives
m(iω)2 eiωt + keiωt = 0
and canceling eiωt and solving gives two solutions,

p
ω = ±ω0 , ω0 = k/m.
Since this a second-order linear differential equation, the general solution is given by the
superposition of these two complex exponentials,
x̃(t) = Aeiω0 t + Be−iω0 t
where A and B are general complex numbers. The real part of x̃(t) satisfies the original real
differential equation ma + kx = 0, and is
Re x(t) = C cos(ω0 t) + D sin(ω0 t)
where C and D are real numbers.
3
[1] Problem 4. To make sure you know how to go from the complex solution to the real one, write
C and D in terms of A and B.
[2] Problem 5. Now introduce a damping force and solve the differential equation for the damped
harmonic oscillator, mẍ + bẋ + kx = 0, using the same procedure, assuming b is small. (See section
01m
4.3 of Morin if you have trouble with this. We’ll consider this system in more detail in M4.)
[3] Problem 6. USAPhO 2012, problem B1.
[3] Problem 7. Above, we mentioned that guessing an exponential works almost all the time. The
reason is because at the end of the day, the exponential cancels out and we’re left with a polynomial
in ω, which has just the right number of roots. But if there are repeated roots, there are fewer
distinct solutions for ω, and hence not enough solutions.
(a) Consider a second order differential equation with a double root ω. What is the other solution,
besides eiωt ? (Hint: to help find a good guess, consider the simple case ma = 0, where ω = 0
is the double root. Then generalize your guess to nonzero ω and check that it works.)
(b) This should be setting off alarm bells: the form of the solutions to the equation changes when
the two roots are exactly equal, while it’s just exponentials/sinusoids if the roots are different,
no matter how small the difference is. Since no two roots are ever exactly equal in practice, it
seems the behavior of part (a) can never actually happen in the real world. But it gets taught
in applied differential equations courses. Why?
(c) [A] Consider the most general nth order, linear homogeneous time-translation invariant differ-
ential equation
dn dn−1

d
an n + an−1 n−1 + . . . + a1 + a0 x = 0.
dt dt dt
What does the general solution look like?
Remark
You might be wondering how to solve more general differential equations. In M4, we will
consider three extensions of the above techniques. We’ll use the idea of normal modes
to solve systems of such differential equations, add driving forces to make the equations
inhomogeneous, and use the adiabatic theorem to approximately solve non time-translation
invariant equations where the coefficients change slowly in time.
Of course, this just scratches the surface of the subject, and solving more general differential
equations can be orders of magnitude harder. We won’t try to solve nonlinear differential
equations, as there is no general technique for doing so, and the answer is often an obscure
special function. (However, such equations will occasionally appear in later problems.) On the
other hand, linear differential equations with general time-dependence are more approachable,
and the following problem illustrates the most basic method for solving them.
[3] Problem 8. [A] Some linear, homogeneous, non time-translation invariant differential equations
can be solved by simply guessing a power series. For this problem, don’t worry about dimensional
analysis; assume all variables have already been redefined to be dimensionless.
4
(a) As a warmup, consider the differential equation ẋ = kx for constant k, which we already know
how to solve. By plugging in the ansatz
∞
X
x(t) = an tn
n=0
find the solution with x(0) = 1.
(b) Now consider the non time-translation invariant differential equation
t2 ẍ + tẋ + t2 x = 0
which is called Bessel’s differential equation of order zero. By using the same ansatz, find the
unique solution with x(0) = 1 and ẋ(0) = 0.
2 Tricks
In this section we’ll consider some kinematics problems that require cleverness, not computation.
Idea 3
Many problems can be solved by a clever choice of reference frame. It is often useful to go to
the frame moving with one of the objects in the problem, or to go into a frame that makes
the motion in the problem more symmetric. For the purposes of kinematics it can even be
useful to use noninertial reference frames, such as a falling frame where projectiles don’t
accelerate, or a rotating frame, though this will introduce fictitious forces into the dynamics.
It is also useful to tilt the coordinate axes to be parallel to various objects.
[1] Problem 9 (KoMaL 2019). A cannon A is at the edge of a cliff with a 800 m drop. Cannon B is
on the ground below the cliff and 600 m horizontally away from it. Cannon A shoots a cannonball
directly towards cannon B at 60 m/s. Cannon B shoots a cannonball directly towards cannon A at
40 m/s. Will the two cannonballs hit each other in midair?
[2] Problem 10 (Wang). Two particles are released in gravitational acceleration g with leftward and
rightward speeds v1 and v2 . Find the distance between them when their velocities are perpendicular.
[3] Problem 11 (Kalda). Two intersecting circles of radius r have centers a distance a apart. If one
circle moves towards the other with speed v, what is the speed of one of the points of intersection?
[2] Problem 12 (Kalda). A mirror rotates about its center with angular speed ω. A stationary point
source of light sits at a distance a from the rotation axis. What is the speed of its mirror image?
[2] Problem 13 (Kalda). Two circles of radius r intersect at the point O. One of the circles rotates
about the point O with constant angular speed ω. The other point of intersection O0 is originally a
distance d from O. Find the speed of O0 as a function of time.
5
Idea 4
To find the minimum value of some quantity, it’s often useful to think about all possible
values of that quantity. This can reveal a solution using geometry or symmetry.
[2] Problem 14 (PPP 3). A boat can travel a speed of 3 m/s on still water. A boatman wants to
cross a river while covering the shortest possible distance.
(a) In what direction should he row if the speed of the water is 2 m/s?
(b) How about if it is 4 m/s?
Idea 5
In problems with friction, the best reference frame to use is almost always the frame of
whatever is causing the friction.
[2] Problem 15 (Kalda). A block is pushed onto a conveyor belt. The belt is moving with speed
1 m/s, and the block’s initial speed is 2 m/s, with initial velocity perpendicular to that of the belt.
During the subsequent motion, what is the minimum speed of the block with respect to the ground?
Idea 6
For a variety of kinematics problems, it can be useful to think about the motion from a
different perspective. For example, if your problem involves complicated accelerations, it
can be useful to think in “velocity space”, i.e. directly think about how the velocity vector
evolves over time, and deal with the position later. Or, if your problem involves complicated
processes occurring in time, it can be useful to think in “spacetime”, meaning to visualize
the process on a space where time is one of the axes. It can also be useful to parametrize
motion in terms of quantities other than the usual Cartesian coordinates.
[2] Problem 16 (Kalda). A boy enters a patch of ice with a coefficient of friction µ with speed v.
By running on the ice, the boy turns his velocity vector by 90◦ in the minimum possible time, so
that his final speed is also v. What is the minimum possible time, and what kind of curve is the
trajectory? Assume the normal force with the ice is constant.
[2] Problem 17 (PPP 5). Four snails travel in uniform, rectilinear motion on a plane. The velocities
are chosen so that three snails never meet at once, and no two of the velocities are equal. Since
4

time t = −∞, five of the 2 possible encounters have already occurred. Must the sixth also occur?
[2] Problem 18. Six bugs are placed at the vertices of a regular hexagon with side length s. At time
t = 0 each bug starts moving directly towards the next with speed v. At what time do they collide?
[3] Problem 19. A rabbit begins at the origin, and the fox begins at the point (0, −a). The rabbit
begins running with a constant speed vx̂. At the same time, the fox begins chasing the rabbit,
always moving towards it with speed v.
(a) Sketch the subsequent trajectory of the rabbit and fox.
6
(b) Let the displacement between the rabbit and fox be
r(t) = (x(t), y(t)).
Show that r + x is conserved.
(c) Find the distance between the rabbit and fox after a long time.
(d) Now suppose the fox has speed u > v. How long does it take to catch the rabbit?
[2] Problem 20 (PPP 85). A child is standing on an icy hill, which may be modeled as an inclined
plane.
The coefficient of friction µk = µs is small enough so that, if the child gets the tiniest push, she will
begin sliding down the plane. Now suppose the child gets a horizontal push, with initial speed v0 .
What is the child’s final speed?
3 Motion in Two Dimensions

Idea 7
Often, motion in two dimensions can be treated as two independent one-dimensional problems.
A change of reference frame may be necessary first.
Idea 8
In problems involving an inclined plane, always set the angle θ to be much closer to either
0◦ or 90◦ than to 45◦ . This reduces mistakes, because almost every angle will be either θ or
90◦ − θ, and you can identify which by sight.
Example 4
Consider projectile motion where wind provides a constant horizontal force F . At what angle
should a projectile of mass m be launched in order to return to the thrower?
7
Solution
The key idea is to use tilted coordinate systems. Clearly, when the only force is downward,
the projectile must be launched straight upward. Now, the horizontal force acts like an
effective horizontal gravitational acceleration of F/m, so that gravity is effectively tilted an
angle tan−1 (F/mg) away from the vertical. One must launch the projectile directly “upward”
with respect to this effective gravitational field, so the launch angle is an angle tan−1 (F/mg)
from the vertical. (For a related problem, see the infamous F = ma 2014 problem 19.)
[1] Problem 21 (Quarterfinal 2002). A cart is rigged with a vertical cannon so that, when the cart is
stationary on a horizontal track, the cannonball is fired straight up and lands back in the cannon.
In each of the following situations, does the cannonball land back in the cannon, in front of it, or
behind it?
(a) The cart is moving on a frictionless horizontal track with speed v.
(b) The cart is accelerating down a frictionless inclined track with angle θ.
(c) The cart is accelerating down an inclined track with angle θ, and friction slows it down.
[2] Problem 22 (Kalda). Two balls at points A and B are released from rest at the same moment,
from the locations shown below. All surfaces are frictionless.
If it takes time tA and tB for the balls to hit the ground, at what time was the distance between
the balls the smallest?
[2] Problem 23 (Kalda). Two planar frictionless walls are placed at right angles, where wall A makes
an angle α to the horizontal. A perfectly elastic ball is released from rest at a point a distance a
from wall A and b from wall B.
After a long time, what is the ratio of the number of times the ball has bounced against wall B to
01W
the number of times it has bounced against wall A?
[3] Problem 25 (EFPhO 2010). A sprinkler can be modeled as a small hemisphere on the ground.
Water shoots out from the hemisphere in all directions, with speed v perpendicular to the hemisphere.
8
(a) Find the total surface area of ground watered by the sprinkler.
(b) At what distance from the sprinkler does the ground get the wettest?
Example 5
A bug flies towards a light with constant speed v, always making an angle α with the radial
direction. If the initial distance to the lamp is L and the radius of the lamp is R, through
what total angle does it turn before hitting the lamp?
Solution
In this case we can’t avoid solving differential equations, but they’re not too hard. It’s easiest
to work in polar coordinates, with the center of the lamp at the origin. By decomposing the
velocity into radial and tangential components, we have
dr dθ
= −v cos α, r = v sin α.
dt dt
We only care about the path, not the time-dependence, so we divide these equations to get
dr r
=−
dθ tan α
where we manipulated differentials as in P1. Separating and integrating,
Z R
dr ∆θ
− =
L r tan α
which tells us that

L
∆θ = (tan α) log .
R
The shape traced out is a logarithmic spiral.
[2] Problem 26. The pilot of a supersonic jet airplane wishes to make a big noise at the origin by
flying around it in a path such that all of the noise he makes is heard simultaneously at the origin.
The jet travels with Mach number M , meaning that its speed is M times the speed of sound. If the
pilot starts at (r, θ) = (a, 0), find the pilot’s path r(θ).
[4] Problem 27. Consider a mass m on a table attached to a spring at the origin with zero relaxed
length, which exerts the force
F = −kr
on the mass. We will find the general solution for r(t) = (x(t), y(t)) in two different ways.
(a) Directly write down the answer, using the fact that the x and y coordinates are independent.
(b) Sketch a representative sample of solutions. What kind of curve does the trajectory follow?
(c) ? Here’s a more unusual way to arrive at the same answer. Go to a noninertial reference frame
rotating with angular velocity ω0 about the origin, so that the centrifugal force cancels out
9
the spring force. In this frame, the only relevant force is the Coriolis force −2mω0 × v. Find
the general solution in this frame, then transform back to the original frame and show that
you get the same answer as in part (a). (This can get a bit messy; the easiest way is to treat
the plane as the complex plane, i.e. work in terms of the variable r = x + iy.)
4 Optimal Launching
Finally, we’ll consider projectile motion questions that involve optimization. These are rare on the
USAPhO, but they are quite fun problems, with occasionally very slick solutions.
Example 6
A bug wishes to jump over a cylindrical log of radius R lying on the ground, so that it just
grazes the top of the log horizontally as it passes by. What is the minimum launch speed v
required to do this?
Solution
Let P be the point at the top of the log. For the bug to be moving horizontally at P , energy
conservation applied to the vertical motion gives an initial vy obeying
1 p
mv 2 = 2mgR, vy = 2 gR.
2 y
Thus, we need to find the minimum vx for the motion to be possible. If vx is too low, the
hypothetical trajectory of the bug will instead pass through the log. At the lowest possible
vx , the bug’s trajectory is not just tangent to the log at point P , but also has the same radius
of curvature (i.e. the trajectory and the log’s shape have the same first and second derivatives).
For uniform motion in a circle of radius r, the acceleration is a = v 2 /r. Conversely, when an
object follows a trajectory of instantaneous radius of curvature r, its acceleration component
normal to the path must be a = v 2 /r. So applying this to the bug at P gives
vx2 p
g= , vx = gR.
R
Thus, the minimum initial speed is
q √
v= vx2 + vy2 = 5 gR.
This radius of curvature trick doesn’t come up often, but it’s cool when it does.
[2] Problem 28. NBPhO 2020, problem 3. A nice warmup for the problems below.
[3] Problem 29. An object is launched from the top of a hill, where the ground lies an angle φ below
the horizontal. Show that the range of a projectile is maximized if it is launched along the angle
bisector of the vertical and the ground.
10
[3] Problem 30 (PPP 35). A point P is located above an inclined plane with angle α. It is possible
to reach the plane by sliding under gravity down a straight frictionless wire, joining P to some point
P 0 on the plane. Geometrically, how should P 0 be chosen so as to minimize the time taken? (Hint:
think about the set of points that can be reached for all possible angles of the wire, after time t.)
Idea 9
Since mechanics is time-reversible, and the speed of a projectile only depends on its height
and not the path taken, finding the way to reach point B from point A with the lowest
possible initial speed is the same as finding the way to reach point A from point B with the
lowest possible initial speed.
[4] Problem 31. Two fences of heights h1 and h2 are erected on a horizontal plain, so that the tops
of the fences are separated byp a distance d. Show that the minimum speed needed to throw a
projectile over both fences is g(h1 + h2 + d).
[4] Problem 32. It’s possible to solve problems 29 and 31 using pure geometry, with no computation.
One can show that the set of points a projectile can reach with a fixed initial speed v is a parabola
with a focus at the launching point. A parabola is defined as the set of points whose distance to
the focus equals the distance to a line, called the directrix.
(a) Show that trajectories that touch the parabola must be tangent to it.
(b) Show that if a point is hit with the smallest possible initial speed, then the initial velocity
must be perpendicular to the final velocity.
(c) Using the geometric definition of a parabola, recover the answers to problems 29 and 31.
[3] Problem 33. 01m IPhO 2012, problem 1A.
5 Reading Graphs
In some kinematics problems, you’ll have to infer what’s going on from a diagram. To make progress,
you’ll have to print out the diagram to make measurements directly on it.
[3] Problem 34. EFPhO 2015, problem 6.
Remark
For a ridiculously hard problem from the same genre, see EuPhO 2019, problem 3. Almost
all competitors received zero points on it; you can try it for entertainment if you’ve finished
everything else and really like kinematics. The official solutions are here.
11
Mechanics II: Statics

For review, read chapter 2 of Morin or chapter 2 of Kleppner and Kolenkow. Statics is covered in
more detail in chapter 7 of Wang and Ricardo, volume 1. Surface tension is covered in detail in
chapter 5 of Physics of Continuous Matter by Lautrup, which is an upper-division level introduction
to fluids in general. There is a total of 79 points.
1 Balancing Forces
Idea 1
In principle, you can always solve every statics problem by balancing forces on every individual
particle in the setup, but often you can save on effort by considering appropriate systems.
Idea 2
Any problem where everything has a uniform velocity is equivalent to a statics problem,
by going to the reference frame moving with that velocity. Any problem where everything
has a uniform acceleration a is also about statics, by going to the noninertial frame with
acceleration a, where there is an extra effective gravitational acceleration −a. The same
principle applies to uniform rotation, where a centrifugal force appears in the rotating frame,
acting like an effective gravitational acceleration ω 2 r.
Example 1
Six blocks are attached in a horizontal line with rigid rods, and placed on a table with
coefficient of friction µ. The blocks have mass m and the leftmost block is pulled with a force
F so the blocks slide to the left. Find the tension force in the rod in the middle.
Solution
There are six objects here and five rods, each with a different tension, so a direct analysis
would involve solving a system of six equations. Instead, first consider the entire set of six
blocks as one object; we can do this because the rigid rods force them to move as one. The
total mass is 6m, and applying Newton’s second law gives
F
F − 6mgµ = 6ma, a= − µg.
6m
Next, consider the rightmost three blocks as one object. Their total mass is 3m, and their
acceleration is the same acceleration a we computed above. This system experiences two
horizontal force: tension and friction. Newton’s second law gives
T − 3mgµ = 3ma
and solving for T gives

F
T = .
2
1
This is intuitive, because the differences of any two adjacent tension forces are the same;
that’s the amount of tension that needs to be spent to accelerate each block. So the middle
rod, which has to accelerate only half the blocks, has half the tension.
The reason we could ignore the tension forces in the other four rods is that the only thing
they do is ensure the blocks move with the same acceleration. Once we assume this is the
case, the specific values of the tensions don’t matter; we can just zoom out and forget them.
It’s just like how within each block there is also an internal tension which keeps it together,
but we rarely need to worry about its details.
Idea 3
To handle a problem where something is just about to slip on something else, set the frictional
force to the maximal value µN and assume slipping is not yet occurring, so the two objects
move as one. The same idea holds for problems which ask for the minimal force needed to
make something move, or the minimal force needed to keep something from moving.
[1] Problem 1 (KK 2.7). A block of mass M1 sits on a block of mass M2 on a frictionless table. The
coefficient of friction between the blocks is µ. Find the maximum horizontal force that can be
applied to (a) block 1 or (b) block 2 so that the blocks will not slip on each other.
[2] Problem 2 (KK 2.28). An automobile enters a turn of radius R.
The road is banked at angle θ, and the coefficient of friction between the wheels and road is µ. Find
the maximum and minimum speeds for the car to stay on the road without skidding sideways.
[2] Problem 3 (KK 2.19). A “pedagogical machine” is illustrated in the sketch below.
01W
All surfaces are frictionless. What force F must be applied to M1 to keep M3 from rising or falling?
2
2 Balancing Torques
Idea 4
A static rigid body will remain static as long as the total force on it vanishes, and the total
torque vanishes, where the torque about the origin is
X
τ= r i × Fi
i
where ri is the point of application of force Fi . If the total force vanishes, the total torque
doesn’t depend on where the origin is, because shifting the origin by a changes the torque by
!
X X
∆τ = a × Fi = a × Fi = 0.
i i
The origin should usually be chosen to set as many torques as possible to zero.
[1] Problem 5. The line of a force is defined to the line passing through its point of application
parallel to its direction; then the torque of the force about any point on that line vanishes. Suppose
a body is static and has three forces acting on it. Show that in two dimensions, the lines of these
forces must either be parallel or concurrent. This will be useful for several problems later.
Idea 5
The center of mass rCM of a set of masses mi at locations ri with total mass M satisfies
X
M rCM = mi ri .
i
If a system experiences no external forces, its center of mass moves at constant velocity.
Idea 6
A uniform gravitational field exerts no torque about the center of mass. Thus, for the
purposes of applying torque balance on an entire object, the gravitational force M g can
be taken to act entirely at its center of mass. (This is a formal substitution; of course, the
actual gravitational force remains distributed throughout the object.)
Torque balance works in noninertial frames, as long as one accounts for the torques due to
fictitious forces. Thus, for an accelerating frame, the −M a fictitious force can be taken to act
at the center of mass. In a uniformly rotating frame, the total centrifugal force is M ω 2 rcm ,
and for the purposes of balancing torques, can be taken to act entirely at the center of mass.
Example 2
Show that the tension in a completely flexible static rope, massive or massless, points along
the rope everywhere in the rope.
3
Solution
Consider a tiny segment d` of the rope. Since the rope is static, the tension forces on
both ends balance, so they are opposite. Let them both be at an angle θ to the rope
direction. Then the net torque on the segment is (T d`) sin θ. Since this must vanish for
static equilibrium, we must have θ = 0 and hence the tension is along the rope. In other
words, flexible ropes can transmit force, but they can’t transmit torque.
It’s important to note that the argument above doesn’t work for a rigid rod, because the
internal forces in a rigid object can look like the picture above. In other words, there can be
extra shear forces from the adjacent pieces of the rod that provide the compensating torque.
If one tried to set up forces like this in a rope, it would flex instead.
In general, the force distribution within a massless rigid rod can be quite complicated, but if
we zoom out, we can replace it with a single tension which does not necessarily point along
the rod. This transmits both a force and a torque through the rod, in the sense that a torque
is eventually exerted by whatever holds the end of the rod in place. Note that if the rod’s
supports are free to rotate, then they can’t absorb torque, so the rod acts just like a rope,
with tension always along it.
Remark
Sometimes, problem writers will intentionally not introduce any variables that are irrelevant
to the answer. This can occur in two ways. First, the variables might just cancel out, as
one can often see by dimensional analysis. Second, the specific values of the variables might
not matter in the limit when they are very large or small. For instance, if a problem simply
states a mass is “very heavy” but doesn’t give it a name like m, it is asking for the answer
in the limit m → ∞.
Idea 7
To handle problems where an object is just about to tip over, note that at this moment, the
entire normal force will often be concentrated at a point. (For example, when you’re about
to fall forward, all your weight goes on your toes.) That often means it’s a good idea to take
torques about this point.
Example 3: Povey 5.6
Suppose that on level ground, a car has a distance d between its left and right tires, and its
center of mass is a height h above the ground. Now suppose the car turns as in problem 2,
but in the extreme case θ = 90◦ , with speed v. For what v is this motion possible?
4
Solution
Again working in the noninertial frame of the car, force balance gives
mv 2
ffric = mg, N=
R
where ffric and N are the total friction and normal forces on the four tires. Since ffric /N ≤ µ,
p
v ≥ gR/µ
which matches the general solution to problem 2. But in that problem, we only considered
force balance. In this extreme situation, we also have to consider torque balance, i.e. the
possibility that the car might topple over. When the car is about to topple over, all the
normal and friction force is on the bottom tires. About this point, we have only torques from
gravity and the centrifugal force, giving
mv 2 d
mgh =
R 2
p
and solving for v gives v = 2gRh/d. Toppling is less likely the higher v is, so the answer is
p √ p
v ≥ gR max(1/ µ, 2h/d).
[2] Problem 6 (Quarterfinal 2004.3). A uniform board of length L is placed on the back of a truck.
There is no friction between the top of the board and the vertical surface of the truck. The coefficient
of static friction between the bottom of the board and the horizontal surface of the truck is µs = 0.5.
The truck always moves in the forward direction.
(a) What is the maximum starting acceleration the truck can have if the board is not to slip or
fall over?
(b) What is the maximum stopping acceleration the truck can have if the board is not to slip or
fall over?
(c) For what value of stopping acceleration is the static frictional force equal to zero?
[2] Problem 7 (Kalda). Three identical rods are connected by freely rotating hinges.
5
The rods are arranged so that CD is parallel to AB, and AB = 2 CD. A mass m is hung on hinge
C. What is the minimum force that must be exerted at hinge D to keep the system stationary?
Idea 8
An extended object supported at a point may be static if its center of mass lies directly above
or below that point. More generally, if the object is supported at a set of points, it can be
static if its center of mass lies above the convex hull of the points.
[2] Problem 8. N identical uniform bricks of length L are stacked on top of each other on the edge
of a table. What is the maximum possible length the top brick can protrude over the edge of the
table? What is the limit of this length as N goes to infinity?
[2] Problem 9 (Kalda). A cylinder with mass M is placed on an inclined slope with angle α so that
its axis is horizontal. A small block of mass m is placed inside it.
The coefficient of friction between the block and cylinder is µ. Find the maximum α so that the
cylinder can stay at rest, assuming that the coefficient of friction between the cylinder and slope is
high enough to keep the cylinder from slipping.
[2] Problem 10 (PPP 11). A sphere is made of two homogeneous hemispheres stuck together, with
different densities. Is it possible to choose the densities so that the sphere can be placed on an
inclined plane with incline 30◦ and remain in equilibrium? Assume the coefficient of friction is
sufficiently high so that the sphere cannot slip.
[3] Problem 11. An object of mass m lies on a uniform floor, with coefficient of static friction µ.
(a) First, suppose the object is a point mass. What is the minimum force required to make the
object start moving, if you can apply the force in any direction?
(b) Now suppose the object is a thin, uniform bar. What is the minimum force required to make
the object start moving, if the force can only be applied horizontally? Assume the normal
pressure on the floor remains uniform.
[3] Problem 12 (Morin 2.17). A spool consists of an axle of radius r and an outside circle of radius
R which rolls on the ground.
6
A thread is wrapped around the axle and is pulled with tension T at an angle θ with the horizontal.
(a) Which way does the spool move if it is pulled with θ = 0?
(b) Given R and r, what should θ be so that the spool doesn’t move? Assume that the friction
between the spool and the ground is large enough so that the spool doesn’t slip.
(c) Given R, r, and the coefficient of friction µ between the spool and the ground, what is the
largest value of T for which the spool remains at rest?
(d) Given R and µ, what should r be so that you can make the spool slip from the static position
with as small a T as possible? That is, what should r be so that the upper bound on T in
part (c) is as small as possible? What is the resulting value of T ?
[3] Problem 13 (PPP 44). A plate, bent at right angles along its center line, is placed on a horizontal
fixed cylinder of radius R as shown.
How large does the coefficient of static friction between the cylinder and plate need to be if the
plate is not to slip off the cylinder?
3 Trickier Torques
Idea 9
Sometimes, a clever use of torque balance can be used to remove any need to have explicit
force equations at all. Rarely, the same situation can occur in reverse.
Example 4: EFPhO 2010.4
A spherical ball of mass M is rolled up along a vertical wall, by exerting a force F to some
point P on the ball. The coefficient of friction is µ. What is the minimum possible force F ,
and in this case, where is the point P ?
7
Solution
Following the logic of idea 3, when the minimum possible force is used, the frictional
force with the wall must be maximal, f = µN , and directed upward. (If friction weren’t
pushing the ball up as hard as possible, we could get by using a smaller force F .)
So even though we don’t know the magnitude of the normal or the frictional force, we
know the direction of the sum of these two forces, so we’ll consider them as one combined force.
This reduces the number of independent forces in the problem to three: gravity (acting at
the center of mass), the force F (acting at P ), and the combined normal and friction forces
(acting at the point of contact C with the wall). Therefore, by the result of problem 5, the
lines of these forces must all intersect at some point A, as shown.
This ensures that the torques will balance, when taken about point A.
Next, we need to incorporate the information from force balance. Doing this directly will
lead us to some nasty trigonometry, but there’s a better way. There are in principle two
force balance equations, for horizontal and vertical forces. However, one of these equations is
just going to tell us the magnitude of the normal/frictional force, which we don’t care about.
So in reality, we just need one equation, which preferably doesn’t involve that force.
The trick is to use torque balance again, about the point C, which says that the torques due
to gravity and F must cancel. Now you might ask, didn’t we already use torque balance?
We did, but recall from idea 4 that taking the torque about a different point can give you a
different equation if the forces don’t balance. So by demanding the torque vanish about two
different points, we actually are using force balance! (Specifically, we are using the linear
combination of the horizontal and vertical force balance equations that doesn’t involve the
normal/friction force, which we don’t need to find anyway.)
When taking the torque about C, we see that F is minimized if P is chosen to maximize
p the
lever arm of the force. This occurs when CA ⊥ P A, in which case the lever arm is R 1 + µ2 ,
where R is the radius of the ball. So we have
p Mg
M gR = F R 1 + µ2 , F =p
1 + µ2
and P is determined as described above.
8
[2] Problem 14. NBPhO 2020, problem 4, parts (i) and (ii).
[3] Problem 15. EFPhO 2012, problem 3. The problem statement is missing some information: both
the bars and rod have diameter d.
[3] Problem 16. EFPhO 2006, problem 6. You will need to print out the problem to make measure-
ments on the provided figure.
[4] Problem 17 (Physics Cup 2012). A thin rod of mass m is placed in a corner so that the rod forms
an angle α with the floor. The gravitational acceleration is g, and the coefficient of friction with
the wall and floor is µs = tan β, which is not large enough to keep the rod from slipping.
What is the minimum additional force F needed to keep the rod static?
We’ve now covered some really mathematically elegant problems, but it’s important to remember
the real-world limitations of this kind of analysis. We discuss two examples below.
Example 5
A uniform bar with mass m and length ` hangs on four equally spaced identical light wires.
Initially, all four wires have tension mg/4.
Find the tensions after the leftmost wire is cut.
Solution
This illustrates a common issue with setups involving rigid supports: there are often more
normal forces than independent equations, so there is not a unique solution. In the real
world, the result is determined by imperfect characteristics of the wires. For example, if one
of the wires was slightly longer than the others, it would go slack, reducing the number of
normal forces by one and yielding a solution.
A reasonable assumption, if you aren’t given any further information, is to assume that the
supports are identical, very stiff springs. In equilibrium, the bar will tilt a tiny bit, so that
the length of the middle wire will be the average of the lengths of the other two. By Hooke’s
law, the force in that wire will than be the average of the other two, so the tensions are
9
mg/3 − x, mg/3, and mg/3 + x. Applying torque balance yields 7mg/12, mg/3, and mg/12.
Remark: Subtleties of Friction

Statics problems involving friction can also get quite elegant, but it’s important to remember
that at the end of the day, they’re just a decent approximation for the real world. In reality,
laws like Ffric = −µk N are only approximately true for some regimes of behavior of some
materials. In fact, the particular law Ffric = −µk N can sometimes yield equations with no
solutions, a phenomenon called the Painleve paradox!
The source of the paradox is the discontinuous dependence of the direction of the friction
force on the direction of the velocity. If you assume an object slips left, it is possible that
after you do all the calculations (assuming a rightward friction force) you instead find the
object slips right. And the reverse happens if you assume the object slips right, which makes
the actual slipping direction completely indeterminate.
You won’t have to worry about this paradox for Olympiads, which never contain setups that
trigger it. But it’s worth keeping in mind that friction is not just a simple law, but the
subject of an entire field of study called tribology, which is essential for engineering. In these
paradoxical cases, you would need to use a more refined model of friction to figure out what
actually happens.
[2] Problem 18 (Kalda). A rod with length ` is hinged to a ceiling with height h < `.
Underneath, a board is being dragged on the floor. The coefficient of (static and kinetic) friction
is µ1 between the board and rod, and µ2 between the board and floor. The rod is meant to stop
the board from being dragged to the right, no matter how hard or how quickly it is pulled. Is this
possible? If so, what are the conditions on the parameters that allow this to occur?
We conclude with some questions that train three-dimensional thinking.
[2] Problem 19 (PPP 10). In Victor Hugo’s novel les Miserables, the main character Jean Valjean, an
escaped prisoner, was noted for his ability to climb up the corner formed by the intersection of two
vertical perpendicular walls. Suppose for simplicity that Jean has no feet. Let µ be the coefficient
of static friction between his hands and the walls. What is the minimum force that Jean had to
exert on each hand to climb up the wall? Also, for what values of µ is this feat possible at all?
[3] Problem 20 (PPP 69). A homogeneous triangular plate has threads of length h1 , h2 , and h3
fastened to its vertices. The other ends of the string are fastened to a common point on the ceiling.
10
Show that the tension in each thread is proportional to its length. (Hint: with the origin at the
point on the ceiling, let the vertices be at positions ri and express everything in vector form.)
[4] Problem 21 (KoMaL 2019, BAUPC 1998). Two identical uniform solid cylinders are placed on a
level tabletop next to each other, so that they are touching. A third identical cylinder is placed on
top of the other two.
(a) Find the smallest possible values of the coefficients of static friction between the cylinders,
and between a cylinder and the table, so that the arrangement can stay at rest.
(b) Repeat part (a) for spheres. That is, put three uniform solid spheres next to each other, with
their centers forming an equilateral triangle, and put a fourth sphere on top.
(c) Now return to part (a), and suppose the setup is frictionless. A force is applied directly to
the right on the leftmost cylinder, causing the entire setup to accelerate. Find the minimum
and maximum accelerations so that all three cylinders remain in contact with each other.
Parts (a) and (b) demonstrate an interesting point: it is possible for a collection of objects to resist
some force, even though a single one of those objects would begin moving even with an infinitesimal
applied force! This is a simple example of how granular materials, like sand, can give rise to emergent
phenomena that are hard to predict from analyzing individual grains alone. Understanding these
materials is a whole field of applied research.
4 Extended Bodies
Idea 10
To deal with extended bodies, one can consider an infinitesimal piece of the body, or consider
a well-chosen symmetrical piece of it, or consider the whole body as a system. Often, multiple
approaches are needed.
Example 6
Find the tension in a circular rope of radius R spinning with angular velocity ω and mass
per length λ.
Solution
Consider an infinitesimal segment of the rope, spanning an angle dθ.
11
The mass of this segment is dm = Rλ dθ. The total force is downward, with magnitude
dθ
dF = 2T sin ≈ T dθ
2
where we used the small angle approximation. This is the centripetal force, so
dF = (dm) ω 2 R.
Combining these results yields T = R2 ω 2 λ.
Example 7
Find the distance d of the center of mass of a uniform semicircle of radius R to its center.
(Note that a semicircle is half of a circle, not half of a disc.)
Solution
This can be done by taking the setup of the previous problem, and taking a subsystem
comprising exactly half of the rope. In this case the net tension force is simply
F = 2T.
The total mass is m = πRλ, and the force must provide the centripetal force, so
F = (πRλ)(ω 2 d)
But we also know that T = R2 ω 2 λ as before, so plugging this in gives

2
d= R.
π
Alternatively, we could have worked in the frame rotating with the rope. The equations
would be the same, but instead we would say the tension balances the centrifugal force.
[1] Problem 22 (KK 2.22). A uniform rope of weight W hangs between two trees. The ends of the
rope are the same height, and they each make angle θ with the trees.
Find the tension at either end of the rope, and the tension at the middle of the rope.
[3] Problem 23 (KK 2.24). A device called a capstan is used aboard ships to control a rope which is
under great tension.
12
The rope is wrapped around a fixed drum with coefficient of friction µ, usually for several turns.
The load on the rope pulls it with a force TA . Ignore gravity.
(a) Show that the minimum force TB needed to hold the other end of the rope in place is TA e−µθ ,
an exponential decrease.
(b) How does this result depend on the shape of the capstan, if we fix the angle θ between the
initial and final tension forces? Would the answer be the same for an oval, or a square?
(c) If θ = π, explain why the total normal and friction force of the rope on the drum is TA + TB .
[2] Problem 24 (F = ma 2018 B20). A massive, uniform, flexible string of length L is placed on
a horizontal table of length L/3 that has a coefficient of friction µs = 1/7, so equal lengths L/3
of string hang freely from both sides of the table. The string passes over the edges of the table,
which are smooth frictionless curves, of size much less than L. Now suppose that one of the hanging
ends of the string is pulled a distance x downward, then released at rest. Neither end of the string
touches the ground.
(a) Find the maximum value of x so that the string does not slip off of the table.
(b) For the case x = 0, draw a free body diagram for the string, indicating only the external
forces on the entire string. Do the forces balance?
(c) Would the answer change significantly if the table’s small edges had friction as well?
[3] Problem 25 (Morin 2.25). A rope rests on two platforms that are both inclined at an angle θ.
The rope has uniform mass density, and the coefficient of friction between it and the platforms is 1.
The system has left-right symmetry. What is the largest possible fraction of the rope that does not
touch the platforms? What angle θ allows this maximum fraction?
Example 8
A chain is suspended from two points on the ceiling a distance d apart. The chain has a
uniform mass density λ, and cannot stretch. Find the shape of the chain.
13
Solution
First, we note that the horizontal component of the tension Tx is constant throughout the
chain; this just follows from balancing horizontal forces on any piece of it. Moreover, by
similar triangles, we have Ty = Tx y 0 everywhere.
Now considerp a small segment of chain with horizontal projection ∆x. The length of the
piece is ∆x 1 + y 02 which determines its weight, and this be balanced by the difference in
vertical tensions. Thus p
∆Ty = λg 1 + y 02 ∆x.
For infinitesimal ∆x, we have ∆Ty = Tx d(y 0 ) = Tx y 00 dx, so we get the differential equation
λg p
y 00 = 1 + y 02 .
Tx
Usually nonlinear differential equations with second derivatives are very hard to solve, but
this one isn’t because there is no direct dependence on y, just its derivatives. That means
we can treat y 0 as the independent variable first, and the equation is effectively first order in y 0 .
Writing y 00 = d(y 0 )/dx and separating, we have
dy 0
Z Z
λg
p = dx.
1+y 02 Tx
Integrating both sides gives

λgx
sinh−1 (y 0 ) = + C.
Tx
Choosing x = 0 to be the lowest point of the chain, the constant C is zero, and

0 λgx
y = sinh .
Tx
Integrating both sides again gives the solution for y,

Tx λgx
y= cosh
λg Tx
where we suppressed another constant of integration. This curve is called a catenary.
[4] Problem 26 (MPPP). A slinky is a uniform spring with negligible relaxed length, with mass m
and spring constant k.
(a) Find the shape of a slinky hung from two points on the ceiling separated by distance d. (Hint:
to begin, consider the mass and tension of a small piece of the spring with horizontal and
vertical extent dx and dy. Don’t forget that the slinky’s density won’t be uniform.)
(b) Suppose a slinky’s two ends are fixed, separated by distance d, and rotating uniformly with
angular frequency ω like a jump rope in zero gravity. Find the values of ω for which this
motion is possible, and the shape of the slinky in this case.
14
Example 9
A uniform spring of spring constant k, mass m, and relaxed length L is hung from the ceiling.
Find its length in equilibrium, as well as its center of mass.
Solution
Problems like this contain subtleties in notation. For example, if you talk about “the piece
of the slinky at z”, this could either mean the piece that’s actually at this position in
equilibrium, or the piece that was originally at this place in the absence of gravity. Talking
about it the first way automatically tells you where the piece is now, but talking about it
the second way makes it easier to keep track of, because then the z of a specific piece of the
spring stays the same no matter what happens.
In fluid dynamics, these are known as the Eulerian and Lagrangian approaches, respectively.
If you don’t use one consistently, you’ll get nonsensical results, and it’s easy to mix them up.
There are many ways to solve this problem, but I’ll give one that reliably works for me.
We’re going to use the Lagrangian approach, and avoid confusion with the Eulerian approach
by breaking the spring into discrete pieces. Let the spring consist of N 1 pieces, of masses
m/N , spring constants N k, and relaxed lengths L/N .
The ith spring from the bottom has tension (i/N )mg, and thus is stretched by
1 i mg
∆Li = mg = i.
kN N kN 2
The total stretch is
N Z N
X mg mg
∆Li = i di = .
kN 2 0 2k
i=1
This makes sense, since the average tension is mg/2. To find the center of mass, note that
the j th spring is displaced downward by a distance
N
j2

X mg
∆yj = ∆Li = 1− 2
2k N
i=j
downward from its position in the absence of gravity. The center of mass displacement is
N N Z N
j2

1 X 1 X 1 2
∆yCM = ∆yj ∝ 1− 2 = 3 N 2 − j 2 dj =
N N N N 0 3
j=1 j=1
so restoring the proportionality constant gives

mg
∆yCM = .
3k
If you want to test your understanding of slinkies, you can also try doing this problem with
the Eulerian approach. The first steps would be finding a relation between the density ρ(z)
and tension T (z) from Hooke’s law, and finding out how to write down local force balance.
15
Remark
In this problem set, we’ve given some examples involving static, continuous, one-dimensional
objects such as strings and ropes. The general three-dimensional theory of elasticity is
mathematically quite complicated, but extremely important in engineering. For more about
this subject, which requires comfort with tensors, see chapters 6 through 11 of Lautrup. It
is also covered in chapters II-31, II-38, and II-39 of the Feynman lectures.
Remark: Why Use Torque?
Here’s a seemingly naive question. Why is the idea of torque so incredibly useful in physics
problems, even though in principle, everything can be derived from F = ma alone? Why is
it almost impossible to solve any nontrivial problem without referring to torques, and how
would a student who’s never heard of torque come up with it in the first place?
We don’t need torque to analyze the statics of a single, featureless point particle. Torque
only became useful in this problem set when we started analyzing rigid bodies with spatial
extent. The reason we couldn’t reduce torque balance to force balance easily is because the
internal forces in these bodies, which maintain their rigidity, are generally very complicated.
It’s possible to derive torque balance from force balance in simple cases, but it’s subtle. For
example, suppose we try to make a rigid body by attaching two ideal springs of infinite spring
constant to masses m and a pivot, as shown.
If forces F1 and F2 are applied, we would like to show that the setup if static only if
the torques are balanced, r1 F1 = r2 F2 , using only force balance on the masses. But our
derivation fails at the very first step: the forces of ideal springs only point along the direction
of the spring. Therefore, they can’t balance the forces at all!
The problem is that this setup isn’t actually rigid; it will bend at the first mass. We could
make the setup rigid by replacing the spring with a rod, because, as we saw in example 2,
a rigid rod can exert downward shear forces on the first mass and upward shear forces on
the second. But in that case, it becomes nontrivial to decide when the rod can remain in
equilibrium. You would need to decompose it microscopically, to see what is responsible for
the shear forces in the first place, and balance forces on all those microscopic pieces. We
tried to make a simple rigid body out of two masses, but now we need to keep track of the
rod’s infinitely many degrees of freedom!
Luckily, there’s a simple modification of our original setup that works. Consider a triangle
whose sides are made of ideal massless springs, pivoted at one vertex, as shown.
16
This actually is a rigid body, and the internal forces are simple: each spring just carries some
tension. Let T be the tension in the spring between the two masses, opposite the pivot. The
forces on each mass can be decomposed into a component parallel to the sides r1 and r2 ,
and a component perpendicular to those sides. The former components can be balanced by
adjusting the tensions in the other two springs. The perpendicular components can only be
balanced by the tension T , from which we conclude
F⊥,1 = T sin θ1 , F⊥,2 = T sin θ2 .
Upon eliminating T and using the law of sines, we conclude that force balance on the masses
is only possible if r2 F⊥,1 = r1 F⊥,2 , which is precisely the statement of torque balance.
Tricky, right? This is the simplest example of a rigid body I know of, and in general, the
internal forces can be much more complicated. The miracle of torque is that it automatically
takes care of all those details for us, no matter what they are.
5 Pressure and Surface Tension

Example 10
A sphere of radius R contains a gas with a uniform pressure P . Find the total force exerted
by the gas on one hemisphere.
Solution
The pressure provides a force per unit area orthogonal to the sphere’s surface, so the
straightforward way to do this is to integrate the vertical component of the pressure force
over a hemisphere. However, there’s a neat shortcut in this case.
Momentarily forget about the sphere and just imagine we have a sealed hemisphere of gas at
pressure P . The net force of the gas on the hemisphere must be zero, or else it would just
begin shooting off in some direction, violating conservation of momentum. So the force on
17
the curved face must balance the force on the flat face, which is πR2 P . The same logic must
hold for the sphere, since the forces on the curved face are the same, so the answer is πR2 P .
This trick works whenever one has a uniform outward pressure on a surface, and it’ll come in
handy for several future problems. For example, it’s the quick way to do F = ma 2018 B24.
Idea 11
The surface of a fluid carries a surface tension γ. If one imagines dividing the surface into two
halves, then γ is the tension force of one half on the other per length of the cut. Specifically,
dF = γ ds × n̂
which means the tension acts along the surface and perpendicular to the cut.
Example 11
A soap bubble of radius R and surface tension γ is in air with pressure P , and contains air
with pressure P + ∆P . Compute ∆P .
Solution
We use the result of the previous problem to conclude that the force of one hemisphere
on another is πR2 ∆P . This must be balanced by the surface tension force. By imagining
cutting the surface of the bubble in half, the surface tension force is γL where L is the total
length of the surface connecting the hemispheres.
At this point, we can write L = 2πR, giving

2γ
∆P = .
R
This is called the Young–Laplace equation. However, in this particular case, this is not the
right answer. The reason is that we should actually take L = 4πR because the surface tension
is exerted at both the inside and outside surfaces of the bubble wall, and thus the answer is
4γ
∆P = .
R
The increased pressure inside balances the surface tension, which wants to collapse the bubble.
If you’re confused about why L = 4πR, you can also think about it in terms of energy. Surface
tension arises from the fact that it costs energy to take soapy water and stretch it out into a
surface, because this breaks some of the attractive intermolecular bonds. The Young–Laplace
equation would give the correct answer for a ball of soapy water. But for a bubble of soapy
water, twice as much soapy water/air surface is created. So the energy cost is double, and
the force is double.
18
[2] Problem 27. One can also derive the Young–Laplace equation by just considering energy. Suppose
the bubble radius changes by dr. The energy of the bubble changes for two reasons: first, there is
net ∆P dV work from the two pressure forces, and there is the γ dA surface tension energy cost. At
equilibrium, the energy must be at a minimum, so it should not change at all under an infinitesimal
displacement. Using this idea, rederive the Young–Laplace equation.
Remark
The general idea used in the above problem, of thinking about how energy would change
in order to find an unknown force, is known as the principle of virtual work. The principle
works for all kinds of forces. For example, if the bubble was charged, it would grow due to
electrostatic repulsion, and the new equilibrium radius could be found using virtual work.
Remark
Note that the Young–Laplace equation we gave above only holds for spherical surfaces. More
generally, a surface has two principle radii of curvature R1 and R2 at each point. These are
both equal to R for a sphere of radius R, but for, e.g. a cylinder of radius R, one is equal to
R and the other is infinity. For general surfaces, the Young–Laplace equation is

1 1
∆P = γ +
R1 R2
where the Ri can each be positive or negative, depending on the direction of curvature.
[2] Problem 28 (Kalda). Consider two soap bubbles which have stuck together. The part of the soap
film that separates the interior of the first bubble from the outside air has radius of curvature R.
The part that separates the interior of the second bubble from the outside air has radius of curvature
2R. What is the radius of curvature of the part which separates the bubbles from each other?
[3] Problem 29 (MPPP 67). When a long, straight sausage is cooked, it always splits “lengthwise”
and never “across”. The two modes of splitting are shown as dotted lines below.
Explain this observation, assuming the thickness of the sausage skin is uniform, and hence can
support a constant surface tension before breaking. (Hint: model the sausage as a cylinder of length
L capped by hemispheres of radius R L, and consider the surface tension needed to prevent the
two modes of splitting mentioned, once an excess pressure P builds up inside the sausage.)
[4] Problem 30. Two coaxial rings of radius R are placed a distance L apart from each other in
vacuum. A soap film with surface tension γ connects the two rings.
(a) Derive a differential equation for the shape r(z) of the film, and solve it.
19
(b) Show that for sufficiently large L, there are no solutions. If L is increased to this value, what
happens to the film?
(c) Using a computer or calculator, find the largest possible value of L.
We’ll consider surface tension in more detail in T3.
Example 12
A solid ball of radius R, density ρ, and Young’s modulus Y rests on a hard table. Because
of its weight, it deforms slightly, so that the area in contact with the table is a circle of radius r.
Estimate r, assuming that it is much smaller than R.
Solution
Recall from P1 that the Young’s modulus is defined by
Y = =
and has dimensions of pressure. By dimensional analysis, you can show that
r = R f (ρgR/Y )
but dimensional analysis alone can’t tell us anything more about f . Moreover, an exact
analysis using forces would be very difficult, because different parts of the ball are compressed
in different amounts, and in different directions; there’s little symmetry here. Instead, we
resort to an order of magnitude approach. To balance gravity, we need a typical stress of
stress ∼ M g/r2 ∼ ρR3 g/r2
in the deformed region. This deformed region has typical length r. To estimate the strain,
note that if the ball were not deformed at all, it would be standing a height δ ∼ r2 /R taller,
as you can show by the Pythagorean theorem. Thus,
strain ∼ δ/r ∼ r/R.
Using the definition of the Young’s modulus, we conclude
ρgR 1/3

r∝R .
Y
By the way, there’s a whole field of study devoted to figuring out how the normal and
other forces behave for realistic, deformable solids, known as contact mechanics. For an
authoritative reference, see Contact Mechanics and Friction by Popov.
20
[4] Problem 31. EFPhO 2006, problem 5. A tough problem on a deforming object.
21
Mechanics III: Dynamics

Chapters 3 and 5 of Morin cover dynamics, energy, and momentum. Alternatively, see chapters 2
and 3 of Kleppner and Kolenkow, or chapters 4 and 6 of Wang and Ricardo, volume 1. For fun, see
chapters I-9 through I-14 of the Feynman lectures. There is a total of 82 points.
1 Blocks, Pulleys, and Ramps

Idea 1
To solve dynamics problems with constraints, it’s easiest to first write the constraint in
terms of coordinates (e.g. “conservation of string” for pulleys, or stationarity of the CM for
an isolated system), then differentiate to get constraints on the velocity and acceleration.
Questions of this type are generally straightforward, as long as you write down the correct
equations. The trickiest part is often solving the equations, which can get messy.
Example 1: Morin 3.30
Find the acceleration of the masses in the Atwood’s machine shown below.
Neglect friction, and treat all pulleys are massless.
Solution
Let x and x0 be the amounts by which the left and right mass have moved down, and number
the pulleys 1 through 4 from left to right, and the strings 1 through 3 from left to right.
Pulley 4 is stationary, so conservation of string 3 means that pulley 3 moves up by x0 /2. Next,
conservation of string 2 means that pulley 2 moves up by x0 /4. Finally, conservation of string
1 implies that pulley 1 moves up by x0 /8, so our final conservation of string constraint is
x0
x=−
8
which upon applying the derivative twice gives
a0
a=− .
8
1
Now consider the tensions Ti in the strings. We know that

2T1 2T3
a=g− , a0 = g − .
m m
Since pulley 3 is massless, the forces on it must balance, so T2 = 2T3 . Similarly T1 = 2T2 ,
so T1 = 4T3 . We hence have a system of three equations in three unknowns (T1 , a, and a0 ),
which can be solved straightforwardly to give
56 7
a0 = g, a=− g.
65 65
[2] Problem 1 (Morin 3.2). Consider the double Atwood’s machine shown below.
Assuming all pulleys are massless, and neglecting friction, find the acceleration of the mass m1 .
[2] Problem 2 (KK 2.15). Consider the system of massless pulleys shown below.
The coefficient of friction between the masses and the horizontal surfaces is µ. Show that the tension
in the rope is
(µ + 1)g
T = .
2/M3 + 1/2M1 + 1/2M2
[2] Problem 3 (KK 2.20). Consider the machine shown below, which we encountered in M2.
2
Show that the acceleration of M1 when the external force F is zero is

M 2 M3 g
a=− .
M1 M2 + M1 M3 + 2M2 M3 + M32
[3] Problem 4. A block of mass m is held motionless on a frictionless plane of mass M and angle of
inclination θ. The plane rests on a frictionless horizontal table.
(a) When the block is released, what is the horizontal acceleration of the plane?
(b) Assume the block starts a distance d above the table. Using results from part (a), what is the
horizontal velocity of the block just before it reaches the floor?
(c) Find the speed of the block after it reaches the floor by applying energy and momentum
conservation to the entire process.
(d) Your results for parts (b) and (c) should not match. What’s going on?
2 Momentum
Idea 2
The momentum of a system is
X
P= mi vi = M vCM .
i
In particular, the total external force on the system is M aCM , and if there are no external
forces, the center of mass moves at constant velocity.
Example 2
A massless rope passes over a frictionless pulley. A monkey hangs on one side, while a bunch
of bananas with exactly the same weight hangs from the other side. When the monkey tries
to climb up the rope, what happens?
Solution
Remarkably, the answer doesn’t depend on how the monkey climbs, whether slowly or
quickly, or symmetrically or not! The total vertical force on the monkey is T − mg, so the
acceleration of the center of mass of the monkey is T /m − g. But since the tension is uniform
through a massless rope, the acceleration of the bananas is also T /m − g. Therefore, the
monkey and bananas rise at the same rate, and meet each other at the pulley.
Now here’s a question for you: compared to climbing up a rope fixed to the ceiling, climbing
up to the pulley takes twice as much work, because the bananas are raised too. But in both
cases, isn’t the monkey applying the same force through the same distance? Where does the
extra work come from? (The answer involves the ideas at the end of this problem set.)
3
Example 3: KK 3.14 / INPhO 2014.5
Two men, each with mass m, stand on a railway flatcar of mass M initially at rest. They
jump off one end of the flatcar with velocity u relative to the car. The car rolls in the opposite
direction without friction. Find the final velocities of the flatcar if they jump off at the same
time, and if they jump off one at a time. Generalize to the case of N 1 men, with a total
mass of mtot .
Solution
In the first case, by conservation of momentum, we have
M v + 2m(v − u) = 0
where v is the final velocity of the flatcar, so

2mu
v= .
M + 2m
In the second case, by a similar argument, we find that after the first man jumps,
mu
v1 = .
M + 2m
Now transform to the frame moving with the flatcar. When the second man jumps, he
imparts a further velocity v2 = mu/(M + m) to the flatcar by another similar argument. The
final velocity of the flatcar relative to the ground is then

1 1
v = v1 + v2 = mu + .
M + 2m M + m
It might be a bit disturbing that the final speeds and hence energies of the flatcar are
different, even though the men are doing the same thing (i.e. expending the same amount of
energy in their legs to jump) in both cases.
The reason for the difference is that in the second case, the second man to jump ends up
with less energy, since the velocity he gets from jumping is partially cancelled by the existing
velocity v1 . So the extra energy that goes into the flatcar corresponds to less kinetic energy
in the men after jumping, which would ultimately have ended up as heat after they slid to a
stop. Accounting properly for the kinetic energy of everything in the system solves a lot of
paradoxes involving energy, as we’ll see below.
In the case of many men, by similar reasoning we have

mtot
v= u
M + mtot
in the first case, while in the second case the answer is the sum
N
X mtot u 1
v= .
N M + (i/N )mtot
i=1
4
This can be converted into an integral, by letting x = i/N , in which case ∆x = 1/N and
Z 1
X mtot u mtot u M + mtot
v= ∆x ≈ dx = log u.
M + xmtot 0 M + xmtot M
i
Note that this is essentially the rocket equation, which we’ll derive in a different way in M6.
[2] Problem 5 (KK 4.11). A flexible chain of mass M and length ` is suspended vertically with its
lowest end touching a scale. The chain is released and falls onto the scale. Find the reading on the
scale when a length of chain x has fallen.
[3] Problem 6. Some qualitative questions about momentum.
(a) A box containing a vacuum is placed on a frictionless surface. The box is punctured on its
right side. How does it move immediately afterward?
(b) You are riding forward on a sled across frictionless ice. Snow falls vertically (in the frame of
the ice) on the sled. Which of the following makes the sled go the fastest or the slowest?
1. You sweep the snow off the sled, directly to the left and right in your frame.
2. You sweep the snow off the sled, directly to the left and right in the ice frame.
01^
3. You do nothing.
[3] Problem 8 (Kalda). A block is on a ramp with angle α and coefficient of friction µ > tan α. The
ramp is rapidly driven back and forth so that its velocity vector u is parallel to both the slope and
the horizontal and has constant modulus v.
The direction of u reverses abruptly after each time interval τ , where gτ v. Find the average
velocity w of the block. (Hint: as mentioned in M1, it’s best to work in the frame of the ramp,
because it causes the friction, even though this introduces fictitious forces.)
[4] Problem 9 (Morin 5.21). A sheet of mass M moves with speed V through a region of space that
contains particles of mass m and speed v. There are n of these particles per unit volume. The
sheet moves in the direction of its normal. Assume m M , and assume that the particles do not
interact with each other.
(a) If v V , what is the drag force per unit area on the sheet?
(b) If v V , what is the drag force per unit area on the sheet? Assume for simplicity that the
component of every particle’s velocity in the direction of the sheet’s motion is exactly ±v/2.
5
(c) Now suppose a cylinder of mass M , radius R, and length L moves through the same region
of space with speed V , and assume v = 0 and m M . The cylinder moves in a direction
perpendicular to its axis. What is the drag force on the cylinder?
Parts (a) and (b) are a toy model for the two regimes of drag, mentioned in M1. However, it
shouldn’t be taken too seriously, because as we’ll see in M7, the typical velocity that separates
the two types of behavior doesn’t have to be of order v. Instead, it depends on how strongly the
particles interact with each other.
3 Energy
Idea 3
The work done on a point particle is
Z
W = F · dx
and is equal to the change in kinetic energy, as you showed in P1.
Remark: Dot Products

The dot product of two vectors is defined in components as
v · w = vx wx + vy wy + vz wz
and is equal to |v| |w| cos θ where θ is the angle between them. For example, if A and B are
the sides of a triangle,
|A + B|2 = (A + B) · (A + B) = A2 + B 2 + 2AB cos θ.
Since the left-hand side is the length squared of the third side of the triangle, we’ve proven
the law of cosines. (Or, if you accept the law of cosines, you could regard this as a proof
that the dot product depends on cos θ as claimed.)
Like the ordinary product, the dot product obeys the product rule. For example,
d
(v · w) = v̇ · w + v · ẇ.
dt
Using this, it’s easy to generalize the derivation of the work-kinetic energy theorem in P1 to
three dimensions; we have
1 1 dx dv
d(v 2 ) = d(v · v) = v · dv = · dv = · dx = a · dx
2 2 dt dt
and this is equivalent to the desired theorem. As you can see, it’s all basically the same, since
the product and chain rule manipulations work the same way for vectors and scalars.
6
Example 4: IPhO 1996 1(b)
A skier starts from rest at point A and slowly slides down a hill with coefficient of friction µ,
without turning or braking, and stops at point B. At this point, his horizontal displacement
is s. What is the height difference h between points A and B?
Solution
Since the skier begins and ends at rest, the change in height is the total energy lost to friction,
Z
mgh = ffric ds
where the integral over ds goes over the skier’s path. Since the skier is always moving
slowly, the normal force is approximately mg cos θ. (More generally, there would be another
contribution to provide the centripetal acceleration.) But then
Z Z Z
ffric ds = µmg cos θ ds = µmg dx = µmgs
which gives an answer of h = µs. (If the skier’s path turned around, then this would still
hold as long as s denotes the total horizontal distance traveled.)
[3] Problem 10 (MPPP 16). On a windless day, a cyclist going “flat out” can ride uphill at a speed
of v1 = 12 km/h and downhill at v2 = 36 km/h on the same inclined road. We wish to find the
cyclist’s top speed on a flat road if their maximal effort is independent of the speed at which the
bike is traveling. Note that in this regime, the air drag force is quadratic in the speed.
(a) Solve the problem assuming that “maximal effort” refers to the force exerted on the pedals
by the rider, and that the rider never changes gears.
01m
(b) Solve the problem assuming that “maximal effort” refers to the rider’s power.

[2] Problem 12. Alice steps on the gas pedal on her car. Bob, who is standing on the sidewalk, sees
Alice’s car accelerate from rest to 10 mph. Charlie, who is passing by in another car, sees Alice’s car
accelerate from 10 mph to 20 mph. Hence Charlie sees the kinetic energy of Alice’s car increase by
three times as much. How is this compatible with energy conservation, given that the same amount
of gas was burned in both frames?
[3] Problem 13 (KK 4.8). A block of mass M is attached to a spring of spring constant k. It is pulled
a distance L from its equilibrium position and released from rest. The block has a small coefficient
of friction µ with the ground. Find the number of cycles the mass oscillates before coming to rest.
[3] Problem 14 (Morin 5.4). A massless string of length 2` connects two hockey pucks that lie on
frictionless ice. A constant horizontal force F is applied to the midpoint of the string, perpendicular
to it. The pucks eventually collide and stick together. How much kinetic energy is lost in the
collision?
7
Idea 4
If a problem can be solved using either momentum conservation or energy conservation alone,
it usually means one of the two isn’t actually conserved. In particular, many processes
are inherently inelastic and inevitably dissipate energy. For more about inherently inelastic
processes, see section 5.8 of Morin.
[2] Problem 15 (KK 4.20). Sand falls slowly at a constant rate dm/dt onto a horizontal belt driven
at constant speed v.
(a) Find the power P needed to drive the belt.
(b) Show that the rate of increase of the kinetic energy of the sand is only P/2.
(c) We can explain this discrepancy exactly. Argue that in the reference frame of the belt, the
rate of heat dissipation is P/2. Since temperature is the same in all frames, the rate of heat
dissipation is P/2 in the original frame as well, accounting for the missing energy.
Example 5: PPP 108
A fire hose of mass M and length L is coiled into a roll of radius R. The hose is sent rolling
√
along level ground, with its center of mass given initial speed v0 gR. The free end of
the hose is held fixed.
The hose unrolls and becomes straight. How long does this process take to complete?
Solution
First, we need to find what is conserved. The horizontal momentum is not conserved,
because there is an external horizontal force needed to keep the end of the hose in place.
On the other hand, the energy is conserved, even though this process looks inelastic. The
hose “sticks” to the floor as it unrolls, but this process dissipates no energy because the cir-
cular part of the hose rolls without slipping, so the bottom of this part always has zero velocity.
Once we figure out energy is conserved, the problem is straightforward. The assumption
√
v0 gR means we can neglect the change in gravitational potential energy as the hose
unrolls. After the hose travels a distance x,

1 1 2 1 1
1+ M v0 = 1+ mv 2
2 2 2 2
8
where the 1/2 terms are from rotational kinetic energy. Since m(x) = M (1 − x/L), we have
v0
v(x) = p
1 − x/L
which gives a total time

Z L Z 1√
dx L 2L
T = = 1 − u du = .
0 v(x) v0 0 3v0
Evidently, the hose accelerates as it unrolls.
[4] Problem 16. Consider the following related problems; in all parts, neglect friction.
(a) A uniform rope of length ` lies stretched out flat on a table, with a tiny portion `0 ` hanging
through a hole. The rope is released from rest, and all points on the rope begin to move with
the same speed. Since this motion is smooth, energy is conserved. Find the speed of the rope
when the end goes through the hole.
(b) ? For practice, repeat part (a) by solving for x(t) explicitly. (Hint: this is best done using the
generalized coordinate techniques of M4.)
(c) Now suppose a flexible uniform chain of length ` is placed loosely coiled close to the hole.
Again, a tiny portion `0 ` hangs through the hole, and the chain is released from rest. In
this case, the unraveling of the chain is an inherently inelastic process, because each link of
the chain sits still until it is suddenly jerked into motion. Find the speed of the chain when
the last link goes through the hole. (Hint: you should get a nonlinear differential equation,
which can be solved by guessing x(t) = Atn .)
[3] Problem 17 (PPP 95). A long slipway, inclined at an angle α to the horizontal, is fitted with
many identical rollers, consecutive ones being a distance d apart. The rollers have horizontal axles
and consist of rubber-covered solid steel cylinders each of mass m and radius r. A plank of mass
M , and length much greater than d, is released at the top of the slipway.
Find the terminal speed vmax of the plank. Ignore air drag and friction at the pivots of the rollers.
9
4 Elastic Collisions
Idea 5
Any temporary interaction between two objects that conserves energy and momentum is a
perfectly elastic collision.
Example 6
Two masses are constrained to a line. The mass m1 moves with velocity v1 , and the mass m2
moves with velocity v2 . The masses collide perfectly elastically. Find their speeds afterward.
Solution
The usual method is to directly invoke conservation of energy and momentum, which leads
to a quadratic equation. A slicker method is to work in the center of mass frame instead.
(This is useful for collision problems in general, and it’ll become even more useful for the
relativistic collisions covered in R2.)
The center of mass of the system has speed

m 1 v1 + m 2 v2
vCM = .
m1 + m2
Moreover, by momentum conservation, the center of mass never accelerates. Now we boost
into the frame moving with the center of mass. Since the total momentum is by definition
zero in the center of mass frame, the momenta of the particles cancel out. The only way for
this to remain true after the collision is if we multiply their velocities by the same number.
Energy is only conserved if this number is ±1, with the latter representing no collision at all.
Therefore, during an elastic collision, the velocities in the center of mass frame simply reverse.
The initial velocities in that frame are
v1,CM = v1 − vCM , v2,CM = v2 − vCM .
The final velocities in that frame are
0 0
v1,CM = −v1 + vCM , v2,CM = −v2 + vCM .
Finally, going back to the original frame gives the final velocities
v10 = −v1 + 2vCM , v20 = −v2 + 2vCM .
There are many special cases we can check. For example, if m1 = m2 , then the two masses
simply swap their velocities, as if they just passed through each other. As another check,
consider the case where the second mass is initially at rest, v2 = 0. Then
m1 − m2 2m1
v10 = v1 , v20 = v1 .
m1 + m2 m1 + m2
When m1 = m2 , the first mass gives all its velocity to the second. When m2 is large, the first
mass just rebounds off with velocity −v1 . When m1 is large, the first mass keeps on going
and the second mass picks up velocity 2v1 . Finally, when m1 = m2 /3, then the final speeds
are v10 = −v1 /2 and v20 = v1 /2, a nice result which is worth committing to memory.
10
Idea 6
For problems involving many collisions, a nice way to keep track of everything is to make a
2D plot of x(t) for all the masses.
[2] Problem 18 (Morin 5.23). A tennis ball with mass m2 sits on top of a basketball with a mass
m1 m2 . The bottom of the basketball is a height h above the ground. When the balls are
dropped, how high does the tennis ball bounce?
[3] Problem 19 (PPP 46). A Newton’s cradle consists of three suspended steel balls of masses m1 ,
m2 , and m3 arranged in that order with their centers in a horizontal line. The ball of mass m1 is
drawn aside in their common plane until its center has been raised by h and is then released. If
all collisions are elastic, how much m2 be chosen so that the ball of mass m3 rises to the greatest
possible height, and what is this height? (Neglect all but the first two collisions.)
[3] Problem 20. Here’s a variety problem involving some “clean” mathematical results. All three
parts can be solved without lengthy calculation.
(a) Consider n identical balls confined to a line. Assuming all collisions are perfectly elastic, what
is the maximum number of collisions that could happen? Assume no triple collisions happen.
(b) A billiard ball hits an identical billiard ball initially at rest in a perfectly elastic collision.
Show that the balls exit at a right angle to each other.
(c) A mass M collides elastically with a stationary mass m. If M > m, show that the maximum
possible angle of deflection of M is sin−1 (m/M ).
[3] Problem 21 (PPP 72). Beads of equal mass m are strung at equal distances d along a long,
horizontal, infinite wire. The beads are initially at rest but can move without friction. The first
bead is continuously accelerated towards the right by a constant force F . After some time, a “shock
wave” of moving beads will propagate towards the right.
(a) Find the speed of the shock wave, assuming all collisions are completely inelastic.
(b) Do the same, assuming all collisions are completely elastic. What is the average speed of the
01^
accelerated bead in this case?
[3] Problem 23. 01m USAPhO 2009, problem B1.
Example 7: MPPP 42
There are N identical tiny discs lying on a table, equally spaced along a semicircle, with total
mass M . Another disc D of mass m is very precisely aimed to bounce off all of the discs in
turn, then exit opposite the direction it came.
11
In the limit N → ∞, what is the minimal value of M/m for this to be possible? Given this
value, what is the ratio of the final and initial speeds of the disc?
Solution
The reason that there is a lower bound on M is that, by the result of part (c) of problem 20,
there is a maximal angle that each tiny disc can deflect the disc D. For large N , the deflection
is π/N for each disc, so
π M/N M
= sin−1 ≈
N m Nm
which implies that M/m ≥ π.
To see how much energy is lost in each collision, work in the center of mass frame and consider
the first collision. In this frame, the disc D is initially approximately still, and the tiny disc
comes in horizontally with speed v. To maximize the deflection angle in the table’s frame, the
tiny disc should rebound vertically, as this provides the maximal vertical impulse to the disc D.
Thus, going√back to the√table’s frame, where the disc D has speed v, the tiny disc scatters
with speed v 2 + v 2 = 2v. By conservation of energy,
1M √ 2

1 2
∆ mv = − ( 2v) .
2 2N
This simplifies to
∆v π
=−
v N
which means that after N collisions, we have
vf π N
= 1− ≈ e−π
vi N
where in the last step we used a result from P1.
12
In this question, we consider a simplified model of how an elastic collision actually happens.
Consider a spherical volleyball inflated with excess pressure ∆P , radius r, and mass m. If it
hits the ground with a large, but not huge, speed v, estimate how long the subsequent elastic
collision takes.
Solution
When the volleyball hits the ground, it will keep going, deforming the part that touches the
ground into a flat circular face. Specifically, when the ball has moved a distance y into the
ground, the flat face has area
p 2
A=π r2 − (r − y)2 = πy(2r − y) ≈ 2πry
where we assumed that y r at all times, which is reasonable as long as the initial speed is
not huge. As a result, the pressure of the volleyball exerts a force
F = 2πr ∆P y
on the ground. This assumes the pressure inside the volleyball remains uniform, and that
the rest of the volleyball stays approximately spherical, which is again reasonable as long as
the initial speed is not huge.
Assuming the initial velocity is not too small, gravity is negligible during the collision, so
during the collision the force on the volleyball is effectively that of an ideal spring. The
collision lasts for half a period, giving
r r
m πm
τ =π = .
keff 2r ∆P
If we plug in realistic numbers, the result is of order 10 ms, which is plausible.
5 Continuous Systems
Example 9
As shown in M2, a hanging chain takes the form of a catenary. Suppose you pull the chain
down in the middle. How does the center of mass of the chain move? Does the answer depend
on how hard you pull?
Solution
No matter how hard you pull, or in what direction, the height of the center of mass always
goes up! This is because this quantity measures the total gravitational potential energy of
the chain. If you pull a chain in equilibrium, in any direction whatsoever, you will do work
13
on it. So this raises its potential energy, and hence the center of mass.
Another way of saying this is that the equilibrium position, without the extra pull you supply,
is already in the lowest energy state, and hence already has the lowest possible center of mass.
Changing this shape in any way raises the center of mass.
[2] Problem 24. A uniform half-disc of radius R is nailed to a wall at the center of its circle and
allowed to come to equilibrium. The half-disc is then rotated by an angle dθ. By calculating the
energy needed to do this in two different ways, find the distance from the pivot point to the center
of mass.
[4] Problem 25 (Morin 5.31). Assume that a cloud consists of tiny water droplets suspended (uni-
formly distributed, and at rest) in air, and consider a raindrop falling through them. Assume the
raindrop is initially of negligible size, remains spherical at all times, and collides perfectly inelasti-
cally with the droplets. It turns out that the raindrop accelerates uniformly; assuming this, find
the acceleration.
[3] Problem 26 (Kvant). Half of a flexible pearl necklace lies on a horizontal frictionless table, while
the other half hangs down vertically at the edge. If the necklace is released from rest, it will slide off
the table. At some point, the hanging part of the necklace will begin to whip back and forth. What
fraction of the necklace is on the table when this begins? (Hint: we are considering a pearl necklace
with no empty string between adjacent pearls; as a result, all the pearls accelerate smoothly. To
solve the problem, think about the vertical forces. There is an important related problem in M2.)
[4] Problem 27 (BAUPC 2002). A small ball is attached to a massless string of length L, the other
end of which is attached to a very thin pole. The ball is thrown so that it initially travels in a
horizontal circle, with the string making an angle θ0 with the vertical. As time goes on, the string
wraps itself around the pole. Assume that (1) the pole is thin enough so that the length of string in
the air decreases very slowly, and (2) the pole has enough friction so that the string does not slide
on the pole, once it touches it. Show that the ratio of the ball’s final speed (right before it hits the
pole) to initial speed is sin θ0 .
When dealing with an extended system whose parts all move in different ways, conservation of
energy is occasionally useless. However, the somewhat obscure idea of “center of mass energy” may
become useful instead. For more about this concept, see section 13.5 of Halliday and Resnick.
Idea 7: Center of Mass Energy
The work done on a part of a system is
dW = F dx
where F is the force on that specific part of the system, and dx is its displacement. Then
dW = dE where E is the total energy of the system.
Similarly, the “center of mass work” done on a system is
dWcm = F dxcm
14
where F is the total force on the system and dxcm is the displacement of the center of mass.
Then dWcm = dEcm where the “center of mass energy” is defined as Ecm = M vcm 2 /2.
It should be noted that, like regular energy and work, center of mass energy and work depend
on the reference frame you’re using.
Example 10
Consider a cyclist who pedals their bike to accelerate. The wheels roll without slipping on
the ground. The cyclist moves a distance d, with the bike experiencing a constant friction
force f from the ground. Analyze the situation using both energy and center of mass energy.
Solution
Since the wheels roll without slipping, their contact point with the ground is always zero,
so the friction force does exactly zero work. Thus the net energy of the cyclist/bike system
is conserved. The additional kinetic energy of the cyclist/bike comes from the chemical
energy of the cyclist, which ultimately came from what they ate. So conservation of energy
is correct, but it doesn’t tell us anything useful at all.
Now consider center of mass energy. Considering the cyclist/bike system, the center of mass
2 /2. This allows us to compute the change in velocity
work is f d, which is the change in M vcm
of the cyclist/bike.
Example 11
Consider the same setup as in the previous example, but now the cyclist brakes hard. The
wheels slip on the ground, and experience a friction force −f while the cyclist moves a
distance d. Analyze the situation using both energy and center of mass energy.
Solution
The center of mass work equation tells us about the overall deceleration of the cyclist/bike,
just as in the previous example.
On the other hand, the work done by the friction force is indeterminate! It can be any
quantity between zero and −f d. When it is 0, the total energy of the cyclist/bike system
is again conserved, which means all the kinetic energy lost is dissipated as heat inside the
bike itself. When it is −f d, all the kinetic energy lost is dissipated as heat in the ground,
and hence energy is removed from the cyclist/bike system. In general, the work will be
an intermediate value, meaning that both the ground and the bike heat up, but we can’t
calculate what it is without a microscopic model of how the friction works. It depends on,
e.g. how easily the ground and bike tire surface deform.
15
[1] Problem 28. Alice and Bob stand facing each other with their arms bent and hands touching on
an ice skating rink. Bob has his back against a wall.
(a) Suppose Bob extends his arms, pushing Alice through a distance d with a force F . Analyze
what happens to Alice in terms of both work and center of mass work.
(b) Suppose Alice extends her arms, pushing herself through a distance d with a force F . Repeat
the analysis; what is different and what is the same?
(c) Suppose a spherical balloon is compressed uniformly from all sides. Is there work done on the
01m
balloon? How about center of mass work?
[4] Problem 29. USAPhO 2013, problem B1. This problem is quite tricky! Once you’re done,
carefully read the official solution, which describes how center of mass work is applied.
16
Mechanics IV: Oscillations

Chapter 4 of Morin covers oscillations, as does chapter 10 of Kleppner and Kolenkow, and chapter
10 of Wang and Ricardo, volume 1. For a deeper treatment that covers normal modes in more detail,
see chapters 1 through 6 of French. Jaan Kalda also has short articles on using Lagrangian-like
techniques and the adiabatic theorem. For some fun discussion, see chapters I-21 through I-25,
II-19, and II-38 of the Feynman lectures. There is a total of 85 points.
1 Small Oscillations
Idea 1
If an object obeys a linear force law, then its motion is simple harmonic. To compute the
frequency, one must the restoring force per unit displacement. More generally, if the force an
object experiences can be expanded in a Taylor series with a nonzero linear restoring term,
the motion is approximately simple harmonic for small displacements. (However, don’t forget
that there are also situations where oscillations are not even approximately simple harmonic,
no matter how small the displacements are.)
Example 1: KK 4.13
The Lennard–Jones potential

r0 12 r 6
0
U (r) = −2
r r
is commonly used to describe the interaction between two atoms. Find the equilibrium radius
and the frequency of small oscillations about this point for two identical atoms of mass m
bound to each other by the Lennard–Jones interaction.
Solution
To keep the notation simple, we’ll set = r0 = 1 and restore them later. The equilibrium
radius is the radius where the derivative of the potential vanishes, and
U 0 (r) = −12r−13 + 12r−7 = 0
implies that that the equilibrium radius is r = r0 . Because the force accelerates both of the
atoms, the angular frequency is s
U 00 (r)
ω=
m/2
where m/2 is the so-called reduced mass. At the equilibrium point, we have
U 00 (r0 ) = (12)(13)r0−14 − (12)(7)r0−8 = 72.
Restoring the dimensionful factors, we have U 00 (r0 ) = 72/r02 , so
r
12
ω= .
r0 m
1
[3] Problem 1 (Morin 5.13). A hole of radius R is cut out from an infinite flat sheet with mass per
unit area σ. Let L be the line that is perpendicular to the sheet and that passes through the center
of the hole.
(a) What is the force on a mass m that is located on L, a distance x from the center of the hole?
(Hint: consider the plane to consist of many concentric rings.)
(b) Now suppose the particle is released from rest at this position. If x R, find the approximate
angular frequency of the subsequent oscillations.
(c) Now suppose that x R instead. Find the period of the resulting oscillations.
(d) Now suppose the mass begins at rest on the plane, but slightly displaced from the center. Do
oscillations occur? If so, what is the approximate frequency?
[2] Problem 2. Some small oscillations questions about the buoyant force.
(a) A cubical glacier of side length L has density ρi and floats in water with density ρw . Find
the frequency of small oscillations, assuming that a face of the glacier always remains parallel
to the water surface, and that the force of the water on the glacier is always given by the
hydrostatic buoyant force.
(b) A ball of radius R floats in water with half its volume submerged. Find the frequency of small
oscillations, making the same assumption.
(c) There are important effects that both of the previous parts neglect. What are some of them?
01W
Is the true oscillation frequency higher or lower than the one found here?
[3] Problem 3. USAPhO 1998, problem A2. To avoid some confusion, skip part (a), since there
01W
actually isn’t a nice closed-form expression for it.
01m
Idea 2
A useful generalization of Newton’s second law is given by generalized coordinates. Let q be
any number that describes the state of the system, not necessarily a Cartesian coordinate.
Suppose the energy of a system can be decomposed into two parts, a potential energy that
depends only on q and a kinetic energy that depends only on q̇,
K = K(q̇), V = V (q).
Then since energy is conserved, d(K + V )/dt = 0, the chain rule gives
d ∂K ∂V
=− .
dt ∂ q̇ ∂q
We call the left-hand side the rate of change of a “generalized momentum”, and the right-hand
side a “generalized force”. When q is a Cartesian coordinate, this recovers the usual F = ma.
2
Remark
The result above is a special case of the Euler–Lagrange equation in Lagrangian mechanics,
which states that if a system is described by a Lagrangian L, then
d ∂L ∂L
= .
dt ∂ q̇ ∂q
When L = K(q̇) − V (q), we recover the previous result. But more generally, it might not be
possible to meaningfully decompose L into a “kinetic” and “potential” piece at all! We won’t
use this more general form below. While it is more powerful, it is also more complicated, and
if you find yourself using it for an Olympiad problem, there’s probably an easier way.
Example 2
Find the acceleration of an Atwood’s machine with masses m and M and a massless pulley
and string.
Solution
The standard way to do this is to let a1 and a2 be the accelerations of the masses, let T be
the unknown tension in the string, solve for T by setting a1 and a2 to have equal magnitudes,
then plug T back in to find the common acceleration. The reason this procedure is so
complicated is that we are using two coordinates when the string really ensures the system
has only a single degree of freedom.
Instead, let q be a generalized coordinate that describes “how much the string has moved”.
In other words, q = 0 initially, and for some q > 0, the mass M has moved down by q and
the mass m has moved up by q. Then
1
K = (m + M )q̇ 2 , V = qg(m − M )
2
and applying the idea above gives
M −m
q̈ = g.
M +m
Another way of saying this is that, from the standpoint of this generalized coordinate, the
“total force” is (M − m)g, and the “total inertia” is M + m.
[1] Problem 6. A rope is nestled inside a curved frictionless tube. The rope has a total length ` and
uniform mass per length λ. The shape of the tube can be arbitrarily complicated, but the left end
of the rope is higher than the right end by a height h. If the rope is released from rest, find its
acceleration. (For a related question, see F = ma 2019 B24.)
3
Idea 3
Generalized coordinates are really useful for problems that involve complicated objects but
only have one relevant degree of freedom, which is especially true for oscillations problems.
For instance, if the kinetic and potential energy have the form
1 1
K = meff q̇ 2 , V = keff q 2
2 2
then the oscillation frequency is always
p
ω= keff /meff .
Note that q need not have units of position, meff need not have units of mass, and so
on. When V (q) is a more general function, we can expand it about a minimum qmin , so
that keff = V 00 (qmin ). This technique allows us to avoid dealing with possibly complicated
constraint forces.
[3] Problem 7. Suppose a particle is constrained to move on a curve y(x) with a minimum at x = 0.
We know that if y(x) is a circular arc, then the motion is not exactly simple harmonic, for the same
reason that pendulum motion is not. Find a differential equation relating y 0 and y, so that the
motion is exactly simple harmonic for arbitrary amplitudes; you don’t have to solve it. (Hint: work
in terms of the coordinate s, the arc length along the curve.)
[3] Problem 8 (Grad). A particle of mass M is constrained to move on a horizontal plane. A second
particle of mass m is constrained to a vertical line. The two particles are connected by a massless
string which passes through a hole in the plane.
The motion is frictionless. Show that the motion is stable with respect to small changes in the
01@
height of m, and find the frequency of small oscillations.
[4] Problem 9. IPhO 1984, problem 2. If you use the energy methods above, you won’t actually
need to know anything about fluid mechanics to do this nice, short problem!
2 Springs and Pendulums

Now we’ll consider more general problems involving springs and pendulums, two very common
components in mechanics questions. As a first example, we’ll use the fictitious forces met in M2.
4
Example 3: PPP 79
A pendulum of length ` and mass m initially hangs straight downward in a train. The
train begins to move with uniform acceleration a. If a is small, what is the period of small
oscillations? If a can be large, is it possible for the pendulum to loop over its pivot?
Solution
The fictitious force in the train’s frame due to the acceleration is equivalent to an additional,
horizontal gravitational field, so the effective gravity is
geff = −ax̂ − gŷ.

p
For small oscillations, we know the period is 2π L/g in ordinary circumstances. By precisely
the same logic, it must be replaced with
s √
L L
T = 2π = 2π 2 .
geff (g + a2 )1/4
As a gets larger, the effective gravity points closer to the horizontal. In the limit g/a → 0,
the effective gravity is just horizontal, so the pendulum oscillates about the horizontal. Its
endpoints are the downward and upward directions, so it never can get past the pivot.
Here’s a follow-up question: if the train can decelerate quickly, how should you stop it so that
the pendulum doesn’t end up swinging at the end? The most efficient way is to first quickly
decelerate to half speed, which, in the frame p of the train, provides a horizontal impulse
to the pendulum. Then wait a half-period π L/g, so that the pendulum’s momentum
turns around, and then quickly stop, providing a second impulse that precisely cancels the
pendulum’s horizontal motion. This cute maneuver is useful for crane operators.
Example 4
If a spring with spring constant k1 and relaxed length `1 is combined with a spring with
spring constant k2 and relaxed length `2 , find the spring constant and relaxed length of the
combined spring, if the combination is in series or in parallel.
Solution
For the series combination, the new relaxed length is clearly ` = `1 + `2 . Suppose the first
spring is stretched by x1 and the second by x2 . The tensions in the springs must balance,
F = k 1 x1 = k 2 x2 .
Thus, the new spring constant is
F k2 x2 k1 k2
k= = = .
x1 + x2 x2 (k2 /k1 + 1) k1 + k2
For example, if the spring is cut in half, the pieces have spring constant 2k.
5
Now consider the parallel combination. In this case it’s clear that the new spring constant
is k = k1 + k2 , since the tensions of the springs add. The new relaxed length ` is when the
forces in the springs cancel out, so
k1 (` − `1 ) + k2 (` − `2 ) = 0
which implies
k1 `1 + k2 `2
`= .
k1 + k2
[2] Problem 10 (Morin 4.20). A mass m is attached to n springs with relaxed lengths of zero. The
spring constants are k1 , k2 , . . . , kn . The mass initially sits at its equilibrium position and then is
given a kick in an arbitrary direction. Describe the resulting motion.
[3] Problem 11 (Morin 4.22). A spring with relaxed length zero and spring constant k is attached
to the ground. A projectile of mass m is attached to the other end of the spring. The projectile is
then picked up and thrown with velocity v at an angle θ to the horizontal.
(a) Geometrically, what kind of curve is the resulting trajectory?
(b) Find the value of v so that the projectile hits the ground traveling straight downward.
[5] Problem 12. A uniform spring of spring constant k and total mass m is attached to the wall, and
the other end is attached to a mass M .
(a) Show that when m M , the oscillation frequency is approximately

s
k
ω= .
M + m/3
(b) [A] ? Generalize part (a) to arbitrary values of m/M . (Hint: to begin, approximate the
massive spring as a finite combination of smaller massless springs and point masses, as in
the example in M2. It will not be possible to solve for ω in closed form, but you can get a
compact implicit expression for it. Check that it reduces to the result of part (a) for small
m/M , and interpret the results for large m/M . This is a challenging problem that requires
almost all the techniques we’ve seen so far, so feel free to ask for more hints.)
[2] Problem 13 (PPP 77). A small bob of mass m is attached to two light, unstretched, identical
springs. The springs are anchored at their far ends and arranged along a straight line. If the bob is
displaced in a direction perpendicular to the line of the springs by a small length `, the period of
01W
oscillation of the bob is T . Find the period if the bob is displaced by length 2`.
6
3 Damped and Driven Oscillations

We now review damped oscillators, which we saw in M1, and consider driven oscillators. For more
guidance, see sections 4.3 and 4.4 of Morin.
[2] Problem 16. Consider a damped harmonic oscillator, which experiences force F = −bv − kx.
(a) As in M1, show that the general solution for x(t) is
x(t) = A+ eiω+ t + A− e−iω− t
and solve for the ω± .
(b) For sufficiently small b, the roots are complex. In this limit, show that by taking the real
part, one finds an exponentially damped sinusoidal oscillation. Roughly how many oscillation
cycles happen when the amplitude damps by a factor of e?
(c) For large b, the roots are pure imaginary, the position simply decays exponentially, and we
say the system is overdamped. Find the condition for the system to be overdamped.
[4] Problem 17. Analyzing a damped and driven harmonic oscillator.
(a) Consider a damped harmonic oscillator which experiences a driving force F = F0 cos(ωt).
Passing to complex variables, Newton’s second law is
mẍ + bẋ + kx = F0 eiωt .
If x(t) is a complex exponential, then we know that the left-hand side is still a complex
exponential, with the same frequency. This motivates us to guess x(t) = A0 eiωt . Show that
this solves the equation for some A0 .
(b) Of course, the general solution needs to be described by two free parameters, to match the
initial position and velocity. Argue that it takes the form
x(t) = A0 eiωt + A+ eiω+ t + A− e−iω− t
where the ω± are the ones you found in problem 16.
(c) After a long time, the “transient” A± terms will decay away, leaving the steady state solution
x(t) ≈ A0 eiωt
which oscillates at the same frequency as the driving. The actual position is the real part,
x(t) ≈ |A0 | cos(ωt − φ)
where A0 = |A0 |e−iφ . Evaluate |A0 | and φ.
(d) Sketch the amplitude |A0 | and phase shiftpφ as a function of ω. Can you intuitively see they
take the values they do, for ω small, ω ≈ k/m, and ω large?
7
(e) There are several distinct things people mean when they speak of “resonant frequencies”.
Find the driving frequency ω that maximizes (i) the amplitude |A0 |, (ii) the amplitude of the
velocity, and (iii) the average power absorbed from the driving force. (As you’ll see, these
are all about the same when the damping is weak, so the distinction between these isn’t so
important in practice.)
p
[3] Problem 18. The quality factor of a damped oscillator is defined as Q = mω0 /b, where ω0 = k/m.
It measures both how weak the damping is, and how sharp the resonance is.
(a) Show that for a lightly damped oscillator,

average energy stored in the oscillator
Q≈ .
average energy dissipated per radian
Then estimate Q for a guitar string.
(b) Show that for a lightly damped oscillator,

resonance frequency
Q≈
width of resonance curve
where the width of the resonance √curve is defined to be the range of driving frequencies for
which the amplitude is at least 1/ 2 the maximum.
For more about Q, see pages 424 through 428 of Kleppner and Kolenkow.
The next two problems explore other ways of driving harmonic oscillators.
[2] Problem 19. Consider a pendulum which can perform small-angle oscillations in a plane with
natural frequency f . The pendulum bob is attached to a string, and you hold the other end of the
string in your hand. There are three simple ways to drive the pendulum:
(a) Move the end of the string horizontally with sinusoidal frequency f 0 .
(b) Move the end of the string vertically with sinusoidal frequency f 0 .
(c) Apply a quick rightward impulse to the bob with frequency f 0 .
In each case, for what value(s) of f 0 can the amplitude become large? (This question should be
01r
done purely conceptually; don’t write any equations, just think!)
[5] Problem 20. GPhO 2016, problem 1. Record your answers on the official answer sheet.
4 Normal Modes
Idea 4: Normal Modes
A system with N degrees of freedom has N normal modes when displaced from equilibrium.
In a normal mode, the positions of the particles are of the form xi (t) = Ai cos(ωt + φi ).
That is, all particles oscillate with the same frequency. Normal modes can be either guessed
physically, or found using linear algebra as explained in section 4.5 of Morin.
8
The general motion of the system is a superposition of these normal modes. So to compute
the time evolution of the system, it’s useful to decompose the initial conditions into normal
modes, because they all evolve independently by linearity.
Example 5
Two blocks of mass m are connected with a spring of spring constant k and relaxed length
L. Initially, the blocks are at rest at positions x1 (0) = 0 and x2 (0) = L. At time t = 0, the
block on the right is hit, giving it a velocity v0 . Find x1 (t) and x2 (t).
Solution
The equations of motion are
mx¨1 = k(x2 − x1 − L)
mx¨2 = k(x1 + L − x2 ).
The system must have two normal modes. The obvious one is when the two masses oscillate
oppositely, x1 = −x2 . The other one is when the two masses move parallel to each other,
x1 = x2 , and this normal mode formally has zero frequency. The initial condition is the
superposition of these two modes.
We can show this a bit more formally. Define the normal mode amplitudes u and v as
u−v u+v
x1 = , x2 = .
2 2
Solving for u and v, we find
u = x1 + x2 , v = x2 − x1 .
Using the equations of motion for x1 and x2 , we have the equations of motion
ü = 0, mv̈ = −2k(v − L)
which
p just verifies that the normal modes are independent, with frequency zero and ω =
2k/m respectively. We can fit the initial condition if
u(0) = L, v(0) = L, u̇(0) = v0 , v̇(0) = v0 .
The normal mode amplitudes are then
v0
u(t) = L + v0 t, v(t) = L + sin ωt.
ω
Plugging this back in gives
v0 t v0 v0 t v0
x1 (t) = − sin ωt, x2 (t) = L + + sin ωt.
2 2ω 2 2ω
Each mass is momentarily stationary at time intervals of 2π/ω, though neither mass ever
moves backwards. If you didn’t know about normal modes, you could also arrive at this
conclusion by playing around with the equations; you could see that they decouple when you
add and subtract them, for instance.
9
[3] Problem 21 (Morin 4.10). Three springs and two equal masses lie between two walls, as shown.
The spring constant k of the two outside springs is much larger than the spring constant κ k of
the middle spring. Let x1 and x2 be the positions of the left and right masses, respectively, relative
to their equilibrium positions. If the initial positions are given by x1 (0) = a and x2 (0) = 0, and if
both masses are released from rest, show that
x1 (t) ≈ a cos((ω + )t) cos(t), x2 (t) ≈ a sin((ω + )t) sin(t)

p
where ω = k/m and = (κ/2k)ω. Explain qualitatively what the motion looks like. This is an
example of beats, which result from superposition two oscillations of nearly equal frequencies; we
will see more about them in W1.
[3] Problem 22 (KK 10.11). Two identical particles are hung between three identical springs.
Neglect gravity. The masses are connected as shown to a dashpot which exerts a force bv, where v
is the relative velocity of its two ends, which opposes the motion.
(a) Find the equations of motion for x1 and x2 .
(b) Show that the equations of motion can be solved in terms of the variables y1 = x1 + x2 and
y2 = x1 − x2 .
(c) Show that if the masses are initially at rest and mass 1 is given an initial velocity v0 , the
motion of the masses after a sufficiently long time is
v0
x1 (t) = x2 (t) = sin ωt
2ω
and evaluate ω.
Example 6
Three identical masses are connected by three identical springs, forming an equilateral triangle
in equilibrium. Describe the normal modes of the system.
10
Solution
Let the system be confined to the xy plane. Then there are three masses that each can move
in two dimensions, giving six degrees of freedom. Since we must be able to construct the
general solution by superposing normal modes, there should be six normal modes. They are:
• Uniform translation in the x or y directions. This yields two normal modes, both formally
with zero frequency, since sin(ωt) ∝ t in the limit ω → 0.
• Uniform rotation about the axis of symmetry.
• A “breathing” motion where the whole triangle expands and contracts.
• A “scissoring” motion where one mass moves outward and the other two move inward.
You might think there are three scissoring normal modes, which would give us one too many.
However, they are redundant: just like how the three sides of the equilateral triangle lie in a
plane, these three normal modes formally lie in a plane, in the sense that you can superpose
any two of them to get the third. Thus we have six normal modes, as expected. If the system
can move in three-dimension space, we need three more; they are uniform translation in the
x direction, and rotation about the x and y axes.
[5] Problem 23 (Morin 4.12, IPhO 1986). N identical masses m are constrained to move on a
horizontal circular hoop connected by N identical springs with spring constant k. The setup for
N = 3 is shown below.
(a) Find the normal modes and their frequencies for N = 2.
(b) Do the same for N = 3.
(c) [A] ? Do the same for general N . (Hint: consider the normal modes found in (a) and (b),
arranged so that in each normal mode, each mass oscillates with unit amplitude but a different
phase. Look at the phases and guess a pattern.)
(d) If one of the masses is replaced with a mass m0 m, qualitatively describe how the set of
frequencies changes.
(e) Now suppose the masses alternate between m and m0 m. Qualitatively describe the set of
frequencies.
Part (c) will be useful in X1, where we will quantize the normal modes found here.
[4] Problem 24. [A] In this problem, you will analyze the normal modes of the double pendulum,
which consists of a pendulum of length ` and mass m attached to the bottom of another pendulum,
11
of length ` and mass m. To solve this problem directly, one has to compute the tension forces in
the two strings, which are quite complicated. A much easier method is to use energy.
(a) Parametrize the position of the pendulum in terms of the angle θ1 the top string makes with
the vertical, and the angle θ2 the bottom string makes with the vertical. Write out the kinetic
energy K and the potential energy V to second order in the θi and θ˙i .
(b) The Euler–Lagrange equations for the system are
d ∂K ∂V
=− .
˙
dt ∂ θi ∂θi
Using the results of part (a), write these equations in the form
θ¨1

g θ1
= − A
θ¨2 L θ2
where A is a 2 × 2 matrix. This is a generalization of θ̈ = −gθ/L for a single pendulum.
(c) Find the normal modes and their frequencies, using the general method in section 4.5 of Morin.
5 [A] Adiabatic Change

Idea 5
When a problem contains two widely separate timescales, such as a fast oscillation superposed
on a slow overall motion, one can solve for the fast motion while neglecting the slow motion,
then solve for the slow motion by replacing the fast motion with an appropriate average.
Example 7: MPPP 21
A small smooth pearl is threaded onto a rigid, smooth, vertical rod, which is pivoted at
its base. Initially, the pearl rests on a small circular disc that is concentric with the rod,
and attached to it a distance d from the rotational axis. The rod starts executing simple
harmonic motion around its original position with small angular amplitude θ0 .
What frequency of oscillation is required for the pearl to leave the rod?
12
Solution
The reason the pearl leaves the rod is that the normal force rapidly varies in direction, with
an average upward component. If this average upward force is greater than gravity, the pearl
accelerates upward and leaves the rod.
In this case, the fast motion is the oscillation of the rod, while the slow motion is the rate of
change of the pearl’s distance from the pivot, which can be neglected during one oscillation.
The pearl has horizontal displacement and acceleration
x(t) = −d sin θ ≈ −dθ(t) = −θ0 d sin ωt, ax (t) = θω 2 d sin ωt.
This is supplied by the horizontal component of the normal force. The vertical component is
Ny = Nx tan θ(t) ≈ max (t)θ(t) = mθ02 ω 2 d sin2 ωt.
Now we average over the fast motion to understand the slow motion. Since the average value
of sin2 (ωt) is 1/2, the condition for the pearl to go up is
1 2 2
mθ ω d > mg
2 0
which gives r
1 2g
ω> .
θ0 d
Example 8
A mass m oscillates on a spring with spring constant k0 with amplitude A0 . Over a very
long period of time, the spring smoothly and continuously weakens until its spring constant
is k0 /2. Find the new amplitude of oscillation.
Solution
In this case the fast motion is the oscillation of the mass, while the slow motion is the
weakening of the spring. We can solve the problem by considering how the energy changes
in each oscillation, due to the slight decrease in k.
Suppose that the spring constant drops in one instant by a factor of 1 − . Then the kinetic
energy stays the same, while the potential energy drops by a factor of 1 − . Since the kinetic
and potential energy are equal on average, this means that if the spring constant gradually
decreases by a factor of 1 − x over a full cycle, with x 1, then the energy decreases by a
factor of 1 − x/2.
The process finishes after N oscillations, where (1 − x)N ≈ √e−N x = 1/2. At this point, the
energy has dropped by a √ N
factor of (1 − x/2) ≈ e −N x/2 = 1/ 2. But the energy is also kA2 ,
4
so the new amplitude is 2A0 .
13
Amazingly, the question can also be solved in one step using a subtle conserved quantity.
Solution
Sinusoidal motion is just a projection of circular motion. In particular, it’s equivalent
to think of the mass as being tied to a spring of zero rest length attached to the origin,
and performing a circular orbit about the origin, with the “actual” oscillation being the x
component. (This is special to zero-length springs obeying Hooke’s law, and occurs because
the spring force −kx = −k(x, y) has its x-component independent of y, and vice versa.)
Since the spring constant is changed gradually, the orbit has to remain circular. Then angular
momentum is conserved, and we have
√
L ∝ vr = ωA2 ∝ kA2 .
√
Then the final amplitude is 4 2A0 as before.
Both of these approaches are tricky. The energy argument is very easy to get wrong, while the
angular momentum argument seems to come out of nowhere and is inapplicable to other situations.
But angular momentum turns out to be a special case of a more general conserved quantity, which
is useful in a wide range of similar problems.
Idea 6: Adiabatic Theorem
If a particle performs a periodic motion in one dimension in a potential that changes very
slowly, then the “adiabatic invariant”
I
I = p dx
is conserved. This is the area of the orbit in phase space, an abstract space whose axes are
position and momentum.
Solution
Using conservation of energy,
p2 1
E= + kx2 .
2m 2
Therefore, the curve p(x) over one√oscillation p
cycle traces out an ellipse in phase space, with
semimajor and semiminor axes of 2mE and 2E/k. The area of this ellipse is the adiabatic
invariant,
√ √
I r
p m
I = p dx = π 2mE 2E/k = 2πE ∝ A2 km.
k
Thus, A ∝ k −1/4 in an adiabatic change of k, recovering the answer found earlier.
14
Remark
The existence of the adiabatic invariant is hard to see in pure Newtonian mechanics, but
it falls naturally out of the framework of Hamiltonian mechanics, which works with phase
space. In fact, Hamiltonian mechanics makes a lot of theoretically useful facts easier to see.
For example, as you will see in X1 using quantum statistical mechanics, the conservation of
the adiabatic invariant for a single classical particle implies the conservation of the entropy
for an adiabatic process in thermodynamics! The two meanings of “adiabatic” are actually
one and the same. If you’d like to learn more about Hamiltonian mechanics, see David Tong’s
lecture notes or chapter 15 of Morin.
[3] Problem 25. Consider a pendulum whose length adiabatically changes from L to L/2.
(a) If the initial (small) amplitude was θ0 , find the final amplitude using the adiabatic theorem.
(b) Give a physical interpretation of the adiabatic invariant.
(c) When quantum mechanics was being invented, it was proposed that the energy in a pendulum’s
oscillation was always a multiple of ~ω, where ω is the frequency. At the first Solvay conference
of 1911, Lorentz asked whether this condition would be preserved upon slow changes in the
length of the pendulum, and Einstein replied in the affirmative. Reproduce Einstein’s analysis.
[4] Problem 26. A block of mass M and velocity v0 to the right approaches a stationary puck of
mass m M . There is a wall a distance L to the right of the puck.
(a) Assuming all collisions are elastic, find the minimum distance between the block and the wall
by explicitly analyzing each collision. (Note that it does not suffice to just use the adiabatic
theorem, because it applies to slow change, while the collisions are sharp. Nonetheless, you
should find a quantity that is approximately conserved after many collisions have occurred.)
(b) Approximately how many collisions occur before the block reaches this minimum distance?
(c) The adiabatic index γ is defined so that P V γ is conserved during an adiabatic process. In one
dimension, the volume V is simply the length, and P is the average force. Using the adiabatic
theorem, infer the value of γ for a one-dimensional monatomic gas.
[4] Problem 27 (F = ma, BAUPC). Two particles of mass m are connected by pulleys as shown.
The mass on the left is given a small horizontal velocity v, and oscillates back and forth.
(a) Without doing any calculation, which mass is higher after a long time?
15
(b) Verify your answer is right by computing the average tension in the leftward string, in the
case where the other end of the string is fixed, for amplitude θ0 1.
(c) Let the masses begin a distance L from the pulleys. Find the speed of the mass which
eventually hits the pulley, at the moment it does, in terms of L and the initial amplitude θ0 .
16
Electromagnetism I: Electrostatics
The material here is covered at the right level in chapters 1–3 of Purcell. For a separate introduction
to vector calculus, see the resources mentioned in the syllabus, or chapter 1 of Griffiths. Electrostatics
is covered in more mathematical detail in chapter 2 of Griffiths. For interesting general discussion,
see chapters II-1 through II-5 of the Feynman lectures. There is a total of 80 points.
1 Coulomb’s Law and Gauss’s Law

We’ll begin with some basic problems which can be solved with symmetry arguments.
Idea 1
Gauss’s law is written in integral form as
I
Q
E · dS = .
0
In practice, you will only apply this form to situations with high symmetry, where

2
Q/4π0 r spherical symmetry,

E = λ/2π0 r cylindrical symmetry,

σ/20 infinite plane.

Example 1
Consider a spherical shell of uniform surface charge density σ. A small hole is cut out of the
surface of the shell. What is the electric field at the center of this hole?
Solution
We use the principle of superposition. First, consider the entire spherical shell, without a hole.
By Gauss’s law and spherical symmetry, the radial electric field at a point P infinitesimally
outside the sphere is σ/0 , while the electric field at a nearby point P 0 infinitesimally inside
is zero.
This field is the superposition of the fields of the charges near P and P 0 , and charges from
the entire rest of the sphere. Consider the effect of a small piece of the surface, near P and
P 0 . From the perspective of these points, this piece looks like an infinite plane, so its radial
electric field is σ/20 at P , and −σ/20 at P 0 . Therefore, the entire rest of the sphere must
contribute a radial electric field of σ/20 , at both P and P 0 . Therefore, when one cuts out a
hole, this is the only contribution that remains, so the field is just σ/20 .
[2] Problem 1 (Griffiths 2.18). Some questions about uniformly charged spheres.
(a) Consider a sphere of radius R and uniform charge density ρ. Find the electric field everywhere.
1
(b) Now two spheres, each of radius R and carrying uniform charge densities ρ and −ρ, are placed
so that they partially overlap. Call the vector from the positive center to the negative center
d. Find the electric field in the overlap region.
[2] Problem 2. A charge q sits just inside a cube, next to one of the corners. What is the flux through
each face of the cube? (More precisely, if a corner is at the origin, and the sides are parallel to the
x, y, and z axes, let the charge be at coordinates (, , ) for tiny .)
[2] Problem 3 (BAUPC). In both parts below, take the potential to be zero at infinity.
(a) Consider a solid sphere of uniform charge density. Find the ratio of the electrostatic potential
at the surface to that at the center.
(b) Consider a solid cube of uniform charge density. Find the ratio of the electrostatic potential
at a corner to that at the center. (Hint: use symmetry.)
Idea 2
If you follow an electric field line, the potential monotonically decreases along it.
[2] Problem 4. Two questions about electrostatic equilibrium.
(a) Prove that when a system of point charges is in equilibrium (i.e. the net force on each of the
charges due to the others vanishes), the total potential energy of the system is zero.
(b) Show that for a positive point charge in the electric fields of fixed, positive point charges,
there is a path along which the charge can be moved to infinity without ever needing positive
external work, i.e. a path along which the potential only decreases.
Idea 3
Gauss’s law is written in differential form as
ρ
∇·E= .
0
The divergence of a vector field F = Fx x̂ + Fy ŷ + Fz ẑ is
∇ · F = ∂x Fx + ∂y Fy + ∂z Fz
in Cartesian coordinates, where ∂x stands for ∂/∂x, and so on.
Example 2
Show that the two forms of Gauss’s law are equivalent.
2
Solution
To do this, we need to establish the geometric meaning of the divergence. For simplicity we
consider two dimensions; the proof for three dimensions is similar. Consider a small rectangle
prism with one corner at the origin, with axes aligned with the Cartesian coordinate axes
and side lengths ∆x and ∆y. To apply Gauss’s law in integral form, we need to compute the
flux through each side. The flux going out the top side is
Z ∆x
Ey (x, ∆y) dx
0
while the flux going out the bottom side is

Z ∆x
− Ey (x, 0) dx.
0
The sum of these two terms is

Z ∆x Z ∆x
(Ey (x, ∆y) − Ey (x, 0)) dx ≈ ∆y (∂y Ey )|(x,0) dx
0 0
where we applied a tangent line approximation, and the subscript indicates where the
function ∂y Ey is evaluated. Higher-order terms in the Taylor series would be proportional to
higher powers of ∆y, which is small, so we can ignore them.
The integrand is still a function of x, but we can Taylor expand it about the origin as
(∂y Ey )|(x,0) = (∂y Ey )|(0,0,0) + ∆x(. . .) + . . . .
These extra terms are again higher-order in ∆x and ∆y, so we ignore them. The net flux
through the top and bottom faces is hence, to lowest order,
Z ∆x
∆y (∂y Ey )|(0,0) dx = ∆x∆y (∂y Ey )|(0,0) .
0
By similar reasoning, pairing up the left and right faces gives
flux = ∆x∆y (∂x Ex + ∂y Ey )|(0,0) = ∆x∆y (∇ · E)|(0,0) .
Thus the divergence is the outgoing flux per unit area, or volume in three dimensions.
This shows us why the two forms of Gauss’s law area equivalent. For example, starting from
the differential form, the left-hand side is the flux per volume, while the right-hand side is
the charge per volume, divided by 0 . Integrating both sides over some volume relates the
total flux to the total charge divided by 0 , which is Gauss’s law in integral form.
If the above derivation was a bit abstract, we can also show the idea using specific examples.
3
Example 3
Suppose the region 0 < x < d has charge density −ρ, and the region −d < x < 0 has charge
density ρ. Find the electric field everywhere.
Solution
By translational symmetry, the field always points along x̂ and only depends on x, E(r) =
E(x) x̂. By applying the integral form of Gauss’s law to a rectangular prism, with one side
at xl and another at xr , we have
1 xr 1 x
Z Z
E(xr ) − E(xl ) = ρ(x) dx, E(x) = ρ(x) dx + E0 .
0 x l 0 0
Since the divergence of E(r) is just ∂E(x)/∂x, this clearly satisfies the differential form of
Gauss’s law. To fix the undetermined constant E0 , we could demand the field be zero on
both sides of the charge distribution, motivated by symmetry. Then we have

d − x 0 < x < d,
ρ 
E(x) = × d + x −d < x < 0,
0 
0 elsewhere.

Example 4
Find the electric field of a spherically symmetric charge density ρ(r).
Solution
By spherical symmetry, the field always points radially and only depends on r, E(r) = E(r) r̂.
By applying the integral form of Gauss’s law to a sphere of radius r,
1 r 0 1 1 r 0 02 0
Z Z
2 02 0
4πr E(r) = dr 4πr ρ(r ), E(r) = dr r ρ(r ).
0 0 0 r 2 0
Let’s check that this indeed satisfies the differential form of Gauss’s law, using the divergence
in spherical coordinates. For any vector field F = Fr r̂ + Fθ θ̂ + Fϕ ϕ̂, the divergence is
1 ∂(r2 Fr ) 1 ∂ 1 ∂Fϕ
∇·F= 2
+ (Fθ sin θ) + .
r ∂r r sin θ ∂θ r sin θ ∂ϕ
This looks complicated, but things turn out simple because E only has a radial component,
Er = E(r), which gives
Z r
1 ∂(r2 E(r)) 1 ∂ r2 ρ(r) ρ(r)
∇·E= 2 = 2 dr0 r02 ρ(r0 ) = 2 =
r ∂r r 0 ∂r 0 r 0 0
just as desired.
4
[3] Problem 5. Consider a vector field expressed in polar coordinates, F = Fr r̂ + Fθ θ̂ where r̂ and θ̂
are unit vectors in the radial and tangential directions. Gauss’s law in differential form still works
in these coordinates, but the form of the divergence is different.
By considering the flux per unit area out of a small region bounded by r and r + dr, and θ and
θ +dθ, and applying Gauss’s law in integral form, find what the divergence in polar coordinates must
be for Gauss’s law in differential form to hold. (Optional: try generalizing to spherical coordinates.)
[4] Problem 6. This problem is quite subtle, but will enhance your understanding of electromagnetism.
Suppose that all of space is filled with uniform charge density ρ.
(a) Show that E = (ρ/0 )xx̂ obeys the differential form of Gauss’s law.
(b) Show that E = (ρ/30 )rr̂ also obeys Gauss’s law.
(c) Argue that by symmetry, E = 0. Show that this does not obey Gauss’s law.
(d) ? What’s going on? Which, if any, is the actual field? If you think there’s more than one
possible field, how could that be consistent with Coulomb’s law, which gives the answer
explicitly? For that matter, what does Coulomb’s law say about this setup, anyway?
Idea 4
A tricky, occasionally useful idea is to use Newton’s third law: it may be easier to calculate
the force of A on B than the force of B on A.
Example 5: Purcell 1.28
Consider a point charge q. Draw any imaginary sphere of radius R around the charge. Show
that the average of the electric field over the surface of the sphere is zero.
Solution
Imagine placing a uniform surface charge σ on the sphere. Then the average of the point
charge’s electric field over the sphere times 4πR2 σ is the total force of the point charge on
the charged sphere. But this is equal in magnitude to the force of the charged sphere on
the point charge, which must be zero by the shell theorem. Thus the average field over the
sphere has to vanish.
[3] Problem 7 (Purcell 1.28). Some extensions of the previous example.
(a) Show that if the charge q is instead outside the sphere, a distance r > R from its center, the
average electric field over the surface of the sphere is the same as the electric field at the center
of the sphere.
(b) Show that for any overall neutral charge distribution contained within a sphere of radius R,
the average electric field over the interior of the sphere is −p/4π0 R3 where p is the total
dipole moment.
5
[3] Problem 8. There are two point charges, q1 > 0 and q2 < 0, in empty space. An electric field line
leaves q1 at an angle α from the line connecting the two charges. Determine whether this field line
hits q2 , and if so, at what angle β from the line connecting the two charges. (Hint: this can be done
without solving any differential equations.)
Idea 5
R
The integral dS over a surface with a fixed boundary is independent of the surface.
We proved this in a mechanical way in M2. If you want to see a proof using vector calculus,
see problem 1.62 of Griffiths.
[3] Problem 9. A hemispherical shell of radius R has uniform charge density σ and is centered at the
origin. Find the electric field at the origin. (Hint: combine the previous two ideas.)
[3] Problem 10. A point charge q is placed a distance a/2 above the center of a square of charge
density σ and side length a. Find the force of the square on the point charge.
[4] Problem 11 (Griffiths 2.47, PPP 113, MPPP 140). Consider a uniformly charged spherical shell
of radius R and total charge Q.
(a) Find the net electrostatic force that the southern hemisphere exerts on the northern hemi-
sphere.
(b) Generalize part (a) to the case where the sphere is split into two parts by a plane whose
minimum distance to the sphere’s center is h.
(c) Generalize part (a) to the case where the two hemispherical shells have uniform charge density,
opposite orientation, and the same center, but have different total charges q and Q, and
different radii r and R, where r < R.
Hint: see example 9, and use superposition and symmetry when applicable.
Example 6: IdPhO 2020.1A
A point charge of mass m and charge −q is placed at the center of a cube with side length a,
whose volume has uniform charge density ρ. The point charge is allowed to slide along a
straight line, which has an arbitrary orientation, so that the distance along the line from the
center to one of the cube’s faces is L.
6
Find the frequency of small oscillations.
Solution
The official solution goes as follows: consider displacing the point charge by some small
amount ∆r. The cube of charge can then be decomposed into (1) a slightly smaller cube of
charge centered around the point charge’s new position, and (2) three thin plates of charge
on the faces opposite to the charge’s motion. By symmetry, (1) contributes nothing, and we
know what (2) contributes from the answer to problem 10. The result is a restoring force
proportional to −∆r, whose magnitude has no dependence on the orientation of ∆r, so the
oscillation frequency doesn’t depend on L. Once you know this, you can orient the line any
way you want, so the problem is simple to finish.
Personally, I don’t like this problem because the intended solution requires knowing the
answer to problem 10, which itself is pretty tricky. That is, the difficulty of the problem
depends mostly on whether you’ve seen that tough, but standard problem elsewhere.
However, I’m including it as an example because there’s another way to solve it, which is
more advanced, but quite illustrative.
Since this is a question about small oscillations, it suffices to expand the potential energy to
second order about the center of the cube. The most general possible exprssion is
V (x, y, z) = a + b1 x + b2 y + b3 z + c1 x2 + c2 y 2 + c3 z 2 + c4 xy + c5 yz + c6 xz + O(r3 ).
The constant a doesn’t matter, so we can just ignore it. And since E vanishes at the center,
the linear terms bi are all zero as well. Because the x, y, and z axes are all equivalent by
cubical symmetry (e.g. we can rotate them into each other, while keeping the cube the same),
c = c1 = c2 = c3 , c0 = c4 = c5 = c6 .
Thus, our complicated Taylor series boils all the way down to
V (x, y, z) = c(x2 + y 2 + z 2 ) + c0 (xy + yz + xz) + O(r3 )
without even having to do any work! Finally, notice that the cube is symmetric under
reflections x → −x, y → −y, or z → −z. These reflections keep the c term the same, but flip
the c0 term. Therefore, we must have c0 = 0, so
V (r) = cr2 + O(r3 )
which is remarkably simple. The potential near the origin is spherically symmetric,
even though the setup as a whole isn’t! It’s not automatic: it wouldn’t have been this
simple if we had had a slightly more complex shape. This “accidental” spherical symme-
try is a consequence of the combination of cubical symmetry and the simplicity of Taylor series.
Therefore, to finish the problem we only need to find the coefficient c. While there are simpler
ways to do this, I’ll do it in a way that introduces some useful facts. Combining the definition
7
of V and Gauss’s law, we have

ρ
∇ · (∇V ) = −∇ · E = − .
0
This is a standard and fundamental result in electrostatics, called Poisson’s equation, which
we will see again later. The divergence of a gradient is also called a Laplacian, and written as
∂2V ∂2V ∂2V ρ

∇2 V = 2
+ 2
+ 2
=− .
∂x ∂y ∂z 0
Using this, we can easily compute the value of c, giving
ρr2
V (r) = − + O(r3 ).
60
Therefore, for a displacement ∆r in any direction, the restoring force is ρqr/30 in the
opposite direction, which means r
ρq
ω=
30 m
independent of the orientation of the line.
Remark
Accidental symmetry is important in modern physics. For example, protons are stable because
of an accidental symmetry in the Standard Model, which ensures that baryon number is
conserved. That explains why we often expect proton decay to occur in extensions of the
Standard Model, such as grand unified theories, as explained in this nice article.
2 Continuous Charge Distributions

Idea 6
In almost all cases in Olympiad physics, there will be sufficient symmetry to reduce any
multiple integral to a single integral. Remember that when using Gauss’s law, the Gaussian
surface may be freely deformed as long as it doesn’t pass through any charges.
[2] Problem 12 (Purcell 1.15). A point charge q is located at the origin. Compute the electric flux
that passes through a circle a distance ` from q, subtending an angle 2θ as shown below.
8
[3] Problem 13 (Purcell 1.8). A ring with radius R has uniform positive charge density λ. A particle
with positive charge q and mass m is initially located in the center of the ring and given a tiny kick.
If the particle is constrained to move in the plane of the ring, show that it exhibits simple harmonic
motion and find the frequency.
[3] Problem 14 (Purcell 1.12). Consider the setup of problem 9. If the hemisphere is centered at the
origin and lies entirely above the xy plane, find the electric field at an arbitrary point on the z-axis.
(This is a bit complicated, and is representative of the most difficult kinds of integrals you might
have to set up in an Olympiad. For a useful table of integrals, see Appendix K of Purcell.)
[3] Problem 15. 01^ USAPhO 2018, problem B1.
Idea 7: Electric Dipoles
The dipole moment of two charges q and −q separated by d is p = qd. More generally, the
dipole moment of a charge configuration is defined as
Z
p = ρ(r)r dr.
For an overall neutral charge configuration, the leading contribution to its electric potential
far away is the dipole potential,
p cos θ
φ(r, θ) =
4π0 r2
where θ is the angle of r to p.
Remark
Here’s a trick to remember the dipole potential. Let φ0 (r) = k/r be the potential for a unit
charge at the origin. An ideal point dipole of dipole moment p consists of charges ±p/d
separated by d, in the limit d → 0. So the potential is
φ0 (r) − φ0 (r + d)
p lim .
d→0 d
But this is precisely the (negative) derivative, so you can get the dipole potential by differen-
tiating the ordinary potential! Indeed, for a dipole aligned along the ẑ axis,
d kp kp dr kp z kp cos θ
− = 2 = 2 =
dz r r dz r r r2
which matches the above result. You can use the same trick for quadrupoles and higher
multipoles, which we’ll see in E8.
[3] Problem 16. In this problem we’ll derive essential results about dipoles, which will be used later.
(a) Using the binomial theorem, derive the dipole potential given above, for a dipole made of a
pair of point charges ±q separated by distance d, oriented along the z-axis.
9
(b) Differentiate this result to find the dipole field,

p
E(r) = (2 cos θ r̂ + sin θ θ̂)
4π0 r3
where the expression above is in spherical coordinates. (Hint: feel free to use the expression
for the gradient in spherical coordinates.)
(c) Show that this may also be written as

1
E(r) = (3(p · r̂)r̂ − p).
4π0 r3
You don’t need to memorize these expressions, but it’s useful to remember what a dipole field
looks like, the fact that its magnitude is roughly p/4π0 r3 , and the fact that the numeric
01m
prefactor is 2 along the dipole’s axis and 1 perpendicular to it.
[3] Problem 18. 01m

USAPhO 2009, problem B2. This essential problem introduces useful facts
about dipole-dipole interactions.
Idea 8
The potential energy of a set of point charges is
1 X qi qj 1X
U= = qi V (ri ).
4π0 |ri − rj | 2
i6=j i
We sum over i 6= j to avoid computing the energy of a single point charge due to its interaction
with itself, which would be infinite. For a continuous distribution of charge, we don’t have
this problem, and instead find
Z Z
1 0
U= ρ(r)V (r) dr = |E(r)|2 dr.
2 2
Unlike the other quantities we’ve considered, energy doesn’t obey the superposition principle.
[3] Problem 19. In this problem we’ll apply the above results to balls of charge.
(a) Compute the potential energy of a uniformly charged ball of total charge Q and radius R.
(b) Show that the potential energy of two point charges of charge Q/2 separated by radius R is
lower than the result of part (a).
(c) Hence it appears that it is energetically favorable to compress a ball of charge into two point
charges. Is this correct?
[3] Problem 20. An insulating circular disc of radius R has uniform surface charge density σ.
(a) Find the electric potential on the rim of the disc.
10
(b) Find the total electric potential energy stored in the disc.
[3] Problem 21. Consider a uniformly charged ball of total charge Q and radius R. Decompose this
ball into two parts, A and B, where B is a ball of radius R/2 whose center is a distance R/2 of the
ball’s center, and A is everything else. Find the potential energy due to the interaction of A and B,
i.e. the work necessary to move B to infinity.
[2] Problem 22 (PPP 149). A distant planet is at a very high electric potential compared with Earth,
say 106 V higher. A metal space ship is sent from Earth for the purpose of making a landing on the
planet. Is the mission dangerous? What happens when the astronauts open the door on the space
ship and step onto the surface of the planet?
Example 7
Since Newton’s law of gravity is so similar to Coulomb’s law, the results we’ve seen so far
should have analogues in Newtonian gravity. What are they? For example, what’s the
gravitational Gauss’s law?
Solution
The fundamental results to compare are
Gm1 m2 q1 q2
F =− , F =
r2 4π0 r2
where the minus sign indicates that the gravitational force is attractive, while the electrostatic
force between like charges is repulsive. Then we can transform a question involving (only
positive) electric charges to one involving masses if we map
1
q → m, → −G, E → g.
4π0
Thus, while electrostatics is described by
I
ρ Q
∇ × E = 0, ∇·E= , E · dS = ,
0 0
the gravitational field is described by
I
∇ × g = 0, ∇ · g = −4πGρm , g · dS = −4πGM
where ρm is the mass density. Similarly, the potential energy can be written in two ways,
Z Z
1 1
U= ρm (r)φ(r) dr = − |g(r)|2 dr
2 8πG
where φ(x) is the gravitational potential. This result was first written down by Maxwell.
11
Remark
Here’s a philosophical question: is potential energy “real”? It’s not as obvious as you might
think! In the 1700s, there was a lively debate over whether the ideas of kinetic energy and
momentum, which at the time were given various other names, were worthwhile. Which one
of the two was the true measure of motion? In our modern language, proponents of energy
pointed out that the momentum always vanished in the center of mass frame, which made it
trivial, while supporters of momentum replied that kinetic energy was clearly not conserved
in even the simplest of cases, like inelastic collisions.
In the 1800s, the theory of thermodynamics was developed, allowing the energy seemingly
lost in inelastic collisions to be accounted for as internal energy. But there still remained
the problem that kinetic energy was lost in simple situations, such as when balls are thrown
upward. By the mid-1800s, the modern language that “kinetic energy is converted to potential
energy” was finally standardized, but it was still common to read in textbooks that potential
energy was fake, a mathematical trick used to patch up energy conservation. After all,
potential energy has some suspicious qualities. If a ball has lots of potential energy, you can’t
see or feel it, or even know it’s there by considering the ball alone. It doesn’t seem to be
located anywhere in space, and its amount is arbitrary, as a constant can always be added.
In the late 1800s, a revolution on physics answered some of these questions. Maxwell and his
successors recast electromagnetism as a theory of fields, and showed that the dynamics of
charges and currents were best understood by allowing the fields themselves to carry energy
and momentum. We’ll cover this in detail in E7, but for now, it implies that electrostatic
potential energy is fundamentally stored in the field, with a density of 0 E 2 /2. This implies
that its location and total amount are directly measurable.
Maxwell believed that the dynamics of fields emerged from the microscopic motions
and elastic deformations of an all-pervading ether, in the same way that, say, a fluid’s
velocity field emerges from the average motion of fluid molecules. This makes it manifestly
positive, so he was disturbed to find that the energy density of a gravitational field is negative!
A few decades later, the arrival of special relativity answered some questions and reopened
others. On one hand, it demolished Maxwell’s vision of the ether. On the other hand, it finally
answered the question of whether all kinds of potential energy are “real”, and it got rid of
the freedom to add arbitrary constants. That’s because in special relativity, the total energy
of a system at rest is related to its mass by E = mc2 , and the mass is directly measurable.
This finally puts thermal energy, elastic potential energy, and field energy on an equal footing.
Here’s the most modern view of energy conservation. All particles and their interactions are
fundamentally described by relativistic quantum fields. A famous result called Noether’s
theorem implies that whenever such a theory is time-translationally symmetric, there
is a conserved quantity which we call the energy. (The distinction between kinetic and
potential energy becomes irrelevant; it’s all just energy.) The density of energy in space
can be computed from the state of the fields, but it doesn’t need to be explained, as
Maxwell imagined, by the internal motion of whatever the fields are made of. The fields are
fundamental: they aren’t made of anything; instead, they make up everything!
12
What happens when we throw gravity into the mix? As we’ll discuss further in R3, it turns
out that at nonrelativistic velocities, the dynamics of gravitating particles can be described
by “gravitoelectromagnetism”, a theory closely analogous to electromagnetism, where moving
masses also source “gravitomagnetic” fields Bg , which result in mv × Bg forces. But the
situation gets much more subtle when we upgrade to full general relativity. Here, the notion
of a gravitational field disappears completely, and is replaced by the curvature of spacetime,
making it hard to define an energy density for it at all. For an accessible overview of the
debate, see this paper. Ultimately, though, it doesn’t matter that much, since it doesn’t
impair our ability to use either Newtonian gravity or general relativity.
Example 8
For an infinite line of linear charge density λ, find the potential V (r) by dimensional analysis.
Solution
This example illustrates a famous subtlety of dimensional analysis. The only quantities in
the problem with dimensions are λ, 0 , and r. To get the electrical units to balance, we have
λ
V (r) = f (r)
2π0
where f (r) is a dimensionless function. But there are no nontrivial dimensionless functions
of a dimensionful quantity r. The only possibilities are that f (r) is a dimensionless constant,
or that f (r) is infinite. In the first case, the electric field would vanish, which can’t be right.
In the second case, it is unclear how to calculate the electric field at all.
In fact, the electric potential is infinite, if you insist on the usual convention of setting
V (∞) = 0. In that case, we have
Z ∞
λ dr
V (r) = =∞
r 2π0 r
independent of r. But this is useless; to get a finite result we can actually work with, we
need to subtract off an infinite constant from the potential. Equivalently, we need to set the
potential to be zero at some finite distance r = r0 . This process is known as renormalization,
and it is extremely important in modern physics. After renormalization, we have
λ r0
V (r) = log
2π0 r
which is perfectly consistent with dimensional analysis.
Notice that in the process of renormalization, a new dimensionful quantity r0 appeared out
of nowhere. This phenomenon is known as dimensional transmutation. Of course, physical
predictions don’t depend on this new scale (e.g. the electric field is independent of r0 ), but
you can’t write down quantities like the potential without it.
13
3 Conductors
Idea 9
In electrostatic conditions, E = 0 inside a conductor, which implies the conductor has constant
electric potential V . This further implies that E is always perpendicular to a conductor’s
surface. By Gauss’s law, the conductor has ρ = 0 everywhere inside, so all charge resides on
the surface.
Example 9
Consider a point on the surface of a conductor with surface charge density σ. Show that the
outward pressure on the charges at this point is σ 2 /20 .
Solution
Gauss’s law tells us that the difference of the electric fields right inside and outside the
conductor at this point is
σ
Eout − Ein =
0
by drawing a pillbox-shaped Gaussian surface. But we also know that Ein = 0 since we’re
dealing with a conductor, so Eout = σ/0 .
Let’s think about how this electric field is made. If there were no charges around except for
the ones at this surface, then the interior and exterior fields would have been ±σ/20 . This
means that all of the other charges, that lie elsewhere on the surface of the conductor, must
provide a field σ/20 here, so that Ein cancels out.
The pressure on the charges at this point on the surface is equal to the product of the surface
charge density with the field due to the rest of the charges, since the charges at this point
can’t exert an overall force on themselves, so
σ2

σ
P =σ =
20 20
2 /2.
as required. Equivalently, we can conclude that P = 0 Eout
Example 10
Is the charge density at the surface of a charged conductor usually greater at regions of higher
or lower curvature?
Solution
Of course, we can’t answer this question directly, because it is essentially impossible to
find the charge distribution of an irregularly shaped conductor. However, we can get some
insight by considering the limiting case of a conductor made of two spheres of radii R1 and
14
R2 , connected by a very long rod.
For the potential to be the same at both spheres, we must have Q1 /R1 = Q2 /R2 , so the
charge is proportional to the radius, and the charge density is inversely proportional to the
radius. Thus, there’s generally higher charge density at sharper points of the conductor.
[1] Problem 23. Show that any surface of charge density σ with electric fields E1 and E2 immediately
on its two sides experiences a force σ(E1 + E2 )/2 per unit area. (This is a generalization of the
example above, where one side was inside a conductor.)
[2] Problem 24. Is it possible for a connected, completely isolated conductor with a positive charge
to have a negative surface charge density at any point? If not, prove it. If so, sketch an example.
Idea 10: Existence and Uniqueness
In a system of conductors where the total charge or potential of each conductor is specified,
there exists a unique charge configuration that satisfies those boundary conditions.
This is very useful because in many cases, it is difficult to directly derive the charge distribu-
tions or fields. Instead, sometimes one can simply insightfully guess an answer; then it must
be the correct answer by uniqueness. For further discussion, see section 2.5 of Griffiths.
Example 11
Consider a conductor with nonzero net charge, and an empty cavity inside. Show that the
electric field is zero in the cavity.
Solution
Let’s consider a second conductor with the same net charge and the same shape, but
without the cavity. By the existence and uniqueness theorem, we know there exists
some charge configuration on the second conductor’s surface which satisfies the boundary
conditions, namely that the electric field vanishes everywhere inside the conductor. In par-
ticular, that means the field is zero where the cavity of the original conductor would have been.
Now consider the original conductor again. If we give this conductor precisely the same
surface charge distribution, then this will again solve the boundary conditions, and it’ll have
no field in the cavity. But by the existence and uniqueness theorem, the charge distribution
is unique, so this is the only possible answer: the field must be zero in the cavity.
If this is your first time seeing this, it can sound like a fast-talking swindle (which is why I
made it an example rather than a problem!). It looks like we used no effort and got a strong
conclusion out. Of course, that’s because all the work is done by the uniqueness theorem.
[1] Problem 25. Consider a spherical conducting shell with an arbitrary charge distribution inside,
with net charge Q. Find the electric field outside the shell.
15
[2] Problem 26 (Purcell 3.33). The shaded regions represent two neutral conducting spherical shells.
01W
Carefully sketch the electric field. What changes if the two shells are connected by a wire?
[4] Problem 28 (MPPP 150). A solid metal sphere of radius R is divided into two parts by a planar
cut, so that the surface area of the curved part of the smaller piece is πR2 . The cut surfaces are
coated with a negligibly thin insulating layer, and the two parts are put together again, so that the
original shape of the sphere is restored. Initially the sphere is electrically neutral.
The smaller part of the sphere is now given a small positive electric charge Q, while the larger
part of the sphere remains neutral. Find the charge distribution throughout the sphere, and the
electrostatic interaction force between the two pieces of the sphere.
[3] Problem 29. In this problem we’ll work through a heuristic proof of a version of the uniqueness
theorem. In particular, we will show that for a system of conductors in empty space, specifying the
total charge on each conductor alone specifies the entire surface charge distribution.
(a) Suppose for the sake of contradiction that two different charge distributions can exist, and
consider their difference, which has zero total charge on each conductor. Argue that at least
one conductor must have electric field lines both originating from and terminating on it.
(b) Show that at least one of these field lines must originate from or terminate on another one of
the conductors.
(c) By generalizing this reasoning, prove the desired result. (Hint: consider the conductors with
the highest and lowest potentials.)
Example 12: Griffiths 7.6
A wire loop of height h and resistance R has one end placed inside a parallel plate capacitor
with electric field E, as shown.
16
The other end of the loop is far away, where the field is negligible. Find the emf in the loop.
Solution
Of course, this is a trick question; if the answer were nonzero, the current would run forever,
yielding a perpetual motion machine. Electrostatic fields always produce zero total emf along
any loop. The σh/0 voltage drop inside the capacitor is canceled out by the voltage drop
due to the fringe fields, which are small, but accumulate over a long distance. The point of
this example is that, while we can ignore fringe fields for some calculations, they are often
essential to get a consistent overall picture. We’ll consider the subtleties of fringe fields in
much more detail in E2.
[2] Problem 30 (Purcell 3.2). Spheres A and B are connected by a wire; the total charge is zero. Two
oppositely charged spheres C and D are brought nearby, as shown.
The spheres C and D induce charges of opposite sign on A and B. Now suppose C and D are
connected by a wire. Then the charge distribution should not change, because the charges on C
and D are being held in place by the attraction of the opposite charge density. Is this correct?
17
Electromagnetism II: Electricity

Chapters 3 and 4 of Purcell cover the material presented here, as does chapter 6 of Wang and
Ricardo, volume 2. Image charges are covered in more detail in section 3.2 of Griffiths. For an array
of interesting physical examples, see chapters II-6 through II-9 of the Feynman lectures. There is a
total of 80 points.
1 The Method of Images

Idea 1
The method of images can be used in some highly symmetric situations to compute the
electric field in the vicinity of a conductor. Specifically, consider any configuration of static
charges and take any equipotential surface containing some of the charges. Then the resulting
field configuration outside that surface is the field configuration we would have if that surface
bounded a conductor. This is simply because it has constant potential on the conductor
surface, so it must be the right answer by the uniqueness theorem.
[4] Problem 1. The simplest application of the method of images is the case of a charge q a distance a
from an infinite grounded conducting plane. This problem explores some of its subtleties, assuming
you’ve already read the basic treatment in section 3.4 of Purcell.
(a) Find the force on the charge.
(b) Find the work needed to move the charge out to infinity. (Answer: q 2 /16π0 a.)
(c) Find the total potential energy of the charges on the conducting plane, i.e. the potential energy
associated only with their interaction with each other. (Answer: q 2 /16π0 a.)
(d) Now suppose there is another parallel grounded conducting plane on the other end of the
charge, a distance b away. How many image charges are needed now? Draw some of them.
(e) A conducting plane forces the electric field to be perpendicular to it. Suppose we modified the
plane so that the electric field was instead always parallel to it. Find the force on the charge.
[4] Problem 2. In this problem you’ll develop the method of images for spheres.
(a) A point charge −q is located at x = a and a point charge Q is located at x = A. Show that
the locus of points with φ = 0 is a circle in the xy plane, and hence a spherical shell in space.
(b) What relations must hold among q, Q, a, and A so the center of the sphere is at the origin?
In this case, what is the sphere’s radius?
(c) Now suppose a point charge Q is a distance b from the center of a spherical grounded conducting
shell of radius r. Find the force on the charge, considering both the cases b < r and b > r.
(d) The case b < r is a bit confusing. On one hand, argue that the total charge is Q plus the image
charge, and hence nonzero. On the other hand, argue that the total charge must be zero, by
considering an appropriate Gaussian surface. One of these arguments is wrong – which one?
1
As an aside, the fundamental reason the method of images works for spheres is that electromagnetism
has conformal symmetry, a symmetry under any local rescaling of space which preserves angles.
(One example of a conformal transformation is inversion in Euclidean geometry.) The setup here is
related to the conducting plane by such a transformation.
[2] Problem 3 (Purcell 3.50). A point charge q is located a distance b > r from the center of a
nongrounded conducting spherical shell of radius r, which also has charge q. When b is close to r,
the charge is attracted to the shell because it induces negative charge; when b is large the charge is
clearly repelled. Find the value of b so that the point charge is in equilibrium. (Hint: you should
have to solve a difficult polynomial equation. You can either use a computer or calculator, or use
the fact that it contains a factor of 1 − x − x2 .)
[2] Problem 4. Consider a spherical conductor of radius R placed in a uniform external field E0 . By
using the method of images, argue that the field created by the sphere is the same as the field of
an infinitesimal dipole at the center of the sphere, and find its dipole moment. (Hint: suppose the
external field is sourced by two point charges that are very far away.)
[3] Problem 5. Two grounded conducting half-planes intersect, so that in cylindrical coordinates, the
equations describing the planes are θ = 0 and θ = 2π/3.
(a) A charge q is placed a distance r from the origin at θ = π/3. Can the method of images be
used to find the force on the charge? If so, do it; if not, explain why not.
(b) Do the same for θ = π/2.
(c) In general, given that the second plane is at θ = θp and the charge is at θ = θq , when does
the method of images work?
[5] Problem 6 (Purcell 3.45). [A] Consider a point charge q located between two parallel infinite
grounded conducting planes. The planes are a distance ` apart, and the point charge is a distance
b from the left plane. The goal of this problem is to find the total charge induced on each plane.
(a) Argue that the total charge on each plane would not change if we replaced the point charge q
with two point charges q/2, both a distance b from the left plane. By iterating this process,
convert the point charge into a uniformly charged plane, and use this to get the answer.
(b) Alternatively, using image charges, show that the electric field on the inside surface of the
left plane, perpendicular to the plane, at a point a distance r from the axis containing all the
image charges, satisfies
∞
X 2q(2n` + b)
4π0 E⊥ = .
n=−∞
((2n` + b)2 + r2 )3/2
(c) Since σ = −0 E⊥ , we can integrate both sides to find the total charge on the left plane.
However, the integral of each term by itself is simply q, so the series doesn’t converge. To get
the result, do the following steps in this specific order: group the terms ±n together, then
integrate them only out to a distance R b, then sum over the values of |n|, then take the
limit R → ∞. You should get a finite result that matches that of part (a). As you’ll probably
see in the process, if you do the steps in any other order, you’ll get a nonsensical answer.
2
Those concerned with mathematical rigor might be bothered by the many choices made in part (c).
You might ask, couldn’t we have gotten a different result by changing how we did the computation?
In fact, by the Riemann rearrangement theorem, we could have gotten almost any result. But the
way we did it is the physically correct way. It roughly sums the terms “in to out”, which respects
the fact that real plates are finite. Closely related ideas are used to “cancel infinities” in quantum
field theory, in a process known as renormalization.
2 Capacitors
Idea 2
For a single, isolated conductor with charge Q, the self-capacitance is defined as
Q = Cφ
where φ is the potential difference between the conductor and infinity. For a set of two
isolated conductors with charges ±Q, the capacitance is defined as
Q = Cφ
where φ is the potential difference between the two conductors.
Idea 3
The definitions of C above are only useful when you have only one or two conductors in the
problem, respectively. In a situation with more than two, it’s very tricky to use the above
definitions, because all the conductors will affect each other; even a neutral conductor will
have an effect since there will be induced charges on its surface.
Instead, it’s better to revert to more general principles. The underlying principle behind
capacitance is linearity: the charges are linearly related to the fields. For multiple capacitors,
the most general possible linear relation is
X
Qi = Cij φj
j
where conductor i has charge Qi and potential φi , the potential is taken to be zero at infinity,
and the Cij are called general coefficients of capacitance, or in electrical engineering, the
Maxwell capacitance matrix. Similarly, inverting this relation,
X
φi = pij Qj
j
where the pij are called coefficients of potential.
In Olympiad physics, you’ll almost never want to compute the Cij or pij explicitly. Instead,
the point here is that if you’re given the charges and want the potentials, or vice versa, you
can build up the answer you want using the principle of superposition, computing all the
fields you need, e.g. using Gauss’s law.
3
Remark
General capacitance coefficients are discussed further in section 3.6 of Purcell. One nontrivial
fact is that Cij = Cji , which is proven by energy conservation in problem 3.64 of Purcell.
Capacitance coefficients can be clunky to work with. For example, suppose you want to
compute the familiar capacitance of a system of two conductors. By definition, we have
Q1 = C11 φ1 + C12 φ2 , Q2 = C21 φ1 + C22 φ2 .
An ordinary two-plate capacitor corresponds to the special case of opposite charges on the
plates, so we write Q = Q1 = −Q2 . There is a potential difference V across the plates, so
φ1 = φ2 + V , and plugging this in gives
Q = (C11 + C12 )φ2 + C1 V, −Q = (C22 + C21 )φ2 + C 0 V.
Eliminating φ2 from the system of equations above, we find the familiar capacitance
Q 2
C11 C22 − C12
C= =
V C11 + C22 + 2C12
where we used C12 = C21 . This is quite an inconvenient formula, so as a result we won’t
consider general capacitance coefficients any further, except briefly for practice in problem 8.
[2] Problem 7 (Purcell 3.21). Consider a capacitor made of four parallel plates with large area A,
evenly spaced with small separation s. The first and third are connected by a wire, as are the
second and fourth. What is the capacitance of the system?
[3] Problem 8. Consider two concentric spherical metal shells, with radii a < b.
(a) Compute their capacitance using Gauss’s law.
(b) Compute their capacitance by computing the four capacitance coefficients, verifying that
C12 = C21 along the way, and using the result for C above.
[3] Problem 9 (MPPP 152). Four identical non-touching metal spheres are positioned at the vertices
of a regular tetrahedron, as shown.
A charge 4q given to sphere A raises it to a potential V . Sphere A can also be raised to potential V
if it and one of the other spheres are each charged with 3q. What must be the size of equal charges
given to A and two other spheres for the potential of A to again be raised to V ? What if all four
spheres are used?
4
Idea 4
The energy stored in a capacitor is
1 1
U = QV = CV 2
2 2
where V is the voltage difference and C is the (mutual) capacitance. The mutual capacitance
C adds in parallel, while 1/C adds in series.
[2] Problem 11 (Purcell 3.24). Some estimates involving capacitance.

(a) Estimate the capacitance of the Earth.
(b) Make a rough estimate of the capacitance of the human body.
(c) By shuffling over a nylon rug on a dry winter day, you can easily charge yourself up to a
couple of kilovolts, as shown by the length of the spark when your hand comes too close to a
grounded conductor. How much energy would be dissipated in such a spark?
[2] Problem 12. The total energy can also be found by integrating the electric field energy,
Z
0
U= E 2 dV.
2
(a) Show that this agrees with U = CV 2 /2 for a parallel plate capacitor.
(b) Show that this agrees with U = CV 2 /2 for a capacitor made of concentric spheres.
The general proof is more advanced, but if you’re interested, one slick method is given in problem
1.33 of Purcell.
[3] Problem 13 (Purcell 3.26). A parallel-plate capacitor consists of a fixed plate and a movable plate
that is allowed to slide in the direction parallel to the plates. Let x be the distance of overlap.
The separation between the plates is fixed. Let C(x) be the capacitance.
(a) Assume the plates are electrically isolated, so that their charges ±Q are constant. By differ-
entiating the energy, find the leftward force on the movable plate in terms of Q and C(x).
(b) Now assume the plates are connected to a battery, so that their potential difference φ is held
constant. Find the leftward force on the movable plate, in terms of φ and C(x).
(c) If the movable plate is held in place, the two answers above should be equal because nothing
is moving. Verify that this is the case, being careful with signs.
(d) In terms of electric fields, why is there a force on the movable plate? Does the effect invoked
in the answer to this part change the conclusion of parts (a) through (c) at all?
5
Idea 5: Dielectrics
While we’re on the subject of capacitors, it’s useful to introduce dielectrics, which will be
important later. All electrical insulators are dielectrics. In the presence of an electric field, a
dielectric will polarize, with positive charges displaced slightly along the field. The resulting
electric dipoles distributed throughout the material in turn create a field that tends to
weaken the original applied field.
Each part of a dielectric polarizes based on the local electric field, but that electric field
depends on the applied field, and the polarization of every other piece of the dielectric. Thus,
solving for the electric field for a general dielectric geometry is very difficult, and usually
not possible in closed form, just like how it’s usually not possible to solve for the field of a
charged conductor. In Olympiad physics, you will almost always consider highly symmetric
situations, where a dielectric simply reduces the applied electric field everywhere by a factor
of κ, called the dielectric constant. (We’ll consider some trickier situations in E8.)
Consider a parallel plate capacitor with charge ±Q on each plate. If a dielectric is inserted
with the charge kept the same, then the field inside is reduced by a factor of κ. Thus, the
capacitance C = Q/V increases by a factor of κ. Dielectrics may increase the amount of
energy that can be stored in a capacitor, which is typically limited by the voltage V0 where
electrical breakdown occurs. So if V0 stays the same, the maximal stored energy U = CV02 /2
goes up by a factor of κ.
Plugging in the definition of C, this result implies that the energy density in the capacitor is
κ0 E 2 . But we showed in E1 that the energy density of the electric field is only 0 E 2 . The
extra energy is stored in the dielectric material itself: it takes energy to separate positive
and negative charges within the dielectric, as if we were stretching many microscopic springs.
This potential energy is released when the capacitor is discharged.
3 Tricky Problems
Example 1: PPP 151
A closed body with conducting surface F has self-capacitance C. The surface is now dented
so that the new surface F ∗ is entirely inside F . Prove that the capacitance has decreased.
Solution
The energy stored in the capacitor is U = Q2 /2C. Therefore, if we give the capacitor a fixed
charge Q, proving that F ∗ has lower C is equivalent to showing that it takes positive work
to dent the foil from F to F ∗ . It’s cleaner to show the other direction, i.e. that starting from
F ∗ , we can get to F while only lowering the energy.
Suppose without loss of generality that F is infinitesimally larger than F ∗ . (We can break
any finite change into infinitesimal stages and repeat this argument.) We can go from
6
F ∗ to F by just taking each charge on the surface and moving it outward until it hits F .
This lowers the energy because the electric field is always directed outward, as we proved in E1.
At this point, the charges lie on F , but they don’t have the right distribution, i.e. F is not
an equipotential. Now we let the charges spontaneously redistribute themselves so that F is
again an equipotential. This again lowers the energy, proving the desired result.
Example 2
Are there charge distributions that aren’t spherically symmetric, but which produce an exact
r̂/r2 field outside of them?
Solution
If you know a bit about the multipole expansion, this might seem like a daunting question.
To make the field exactly r̂/r2 , you need to make sure the charge distribution has no dipole
moment, no quadrupole moment, no octupole moment, and so on to infinity, and it seems
impossible to satisfy all of these constraints without spherical symmetry. But we have
already seen an example of such a charge distribution earlier in the problem set!
Recall that when we treated the method of images for spheres, we found that in some
situations, the complicated charge densities on conducting spheres were exactly the same as
those produced by a fictitious image charge inside the sphere, and generally away from its
center. If we place the origin at that image charge, then we have an example of a charge
distribution that is perfectly r̂/r2 far away, but which isn’t spherically symmetric. (The
general solution is given here.)
[2] Problem 14. Consider a set of n conducting, very large parallel plates, placed in zero external
electric field. The plates are given charges Qi . If the left ends of the plates are at locations xi , and
the plates have thickness di , what is the total charge on the left end of the leftmost plate, and the
right end of the rightmost plate?
[2] Problem 15 (Purcell 3.9). A conducting spherical shell has charge Q and radius R1 . A larger
concentric conducting spherical shell has charge −Q and radius R2 .
(a) If the outer shell is grounded, explain why nothing happens to the charge on it.
(b) If instead the inner shell is grounded, e.g. by connecting it to ground by a very thin wire that
passes through a very small hole in the outer shell, find its final charge.
(c) It’s not so clear why charge would leave the inner shell in part (b), thinking in terms of forces.
A small bit of positive charge will certainly want to hop on the wire and follow the electric
field across the gap to the larger shell. But when it gets to the larger shell, it seems like it
has no reason to keep going to infinity, because the field is zero outside. And, even worse, the
field will point inward once some positive charge has moved away from the shells. So it seems
like the field will drag back any positive charge that has left. Does charge actually leave the
inner shell? If so, what’s wrong with the above reasoning?
7
[2] Problem 16. The usual expression for the capacitance of a parallel-plate capacitor is A0 /d.
However, in reality the field within the capacitor is not perfectly uniform, and there are fringe fields
outside. Taking these effects into account, is the true capacitance higher or lower than A0 /d?
R
[2] Problem 17 (Purcell 4.16). In a parallel plate capacitor, the quantity E · ds should be equal to
V for any path that connects the two plates.
A charged capacitor can be discharged R by attaching a wire to the external surfaces of the plates.
No matter how one attaches the wire, E · ds along the wire should be equal to V . And as we’ve
argued in E2, this is sufficient to cause charges to move along the wire, even if the electric field
points in the “wrong” direction at some points along the wire, because the wire has negligible
capacitance: charges within it move rigidly, each pushing the next one and pulling the previous one.
But it’s puzzling how this works for a capacitor, because the electric field is supposed to be
essentially zero just outside it. Consider two possible limiting cases for the wire’s shape.
R
In each case, explain qualitatively how E · ds can be equal to V . In particular, how large are the
contributions from the distinct segments of the wire (the horizontal and vertical parts in the first
case, and the straight and curved parts in the second)?
[3] Problem 18. Consider two conducting spheres of radius r separated by a distance a r. The
spheres can be thought of as the two plates of a parallel plate capacitor.
(a) By assuming the surface charge density on each sphere is uniform, estimate the capacitance.
(b) In reality, the surface charges on each sphere will be distorted by the other sphere. The surface
charges assumed in part (a) will induce changes in the surface charges, represented by an
image charge for each sphere. But these image charges will induce further image charges, and
so on, yielding an infinite series for the charge distribution and hence the capacitance. Find
the first two terms in the series for the capacitance.
(c) If a/r = 10, roughly estimate the error in neglecting the other terms.
Example 3
Find the interaction force between a dipole of dipole moment p and a conducting sphere of
radius r, separated by a distance R r.
8
Solution
There are several ways to do this problem using the method of images; here’s one. Consider
the image charges that the two charges in the dipole make. These produce an image dipole
right at the center of the sphere, and using the results of problem 2, the dipole moment is
p0 = p(r/R)3 . Now, the force on the original dipole is
d 2kp0

d
F = p E(z) =p
dz z=R dz z 3 z=R
where E is the field made by the image dipole, and we used the results of E1. Thus,
6kp2 r3

2 3 d 2k

F = p (r/R) 3
=− .
dz z z=R
R7
As usual for charge induction effects, this force is attractive, and falls off quickly with distance.
Example 4
Estimate the interaction force between a point charge q and a thin conducting rod of length `,
which is a distance L ` from the charge and oriented along the separation between them.
Solution
The interaction occurs because the point charge induces negative charges on the near end of
the rod, and positive charges on the far end. These charges are then acted on by the electric
field of the point charge, causing a force.
To get a very crude estimate, let’s just suppose that charge Q appears on the far end and
charge −Q appears on the near end. The resulting field produced in the middle is
kQ
E∼ .
`2
On the other hand, this needs to cancel a field from the point charge of
kq
E∼
L2
which tells us that Q ∼ (`/L)2 q. The force on the induced charges is
kqQ`2 kq 2 `4

1 1
F ∼ kqQ − ∼ − ∼ − .
L2 + `2 L2 L4 L6
Again, the force is attractive, and falls off quickly with distance.
[3] Problem 19 (Physics Cup 2017). Estimate the interaction force between a point charge q and an
infinitely thin circular neutral conducting disc of radius r if the charge is at the axis of the disc,
and the distance between the disc and the charge is L r.
9
[5] Problem 20. 01TIPhO 2012, problem 2. A challenging electrostatics/fluids problem; some prior
exposure to surface tension is helpful. (For more about the kinds of bubbles encountered in this
problem, see section 5.9 of Physics of Continuous Matter by Lautrup.)
Example 5
Find the charge distribution on a conducting disc of radius R and total charge Q.
Solution
In general, there are very few situations where the charge distribution on a conductor can
be found explicitly. As you’ve seen, some of the simplest examples can be solved with image
charges. Some more complex, two-dimensional examples can be solved with a mathematical
technique called conformal mapping. And this special example can be solved with a neat trick.
Consider a uniformly charged spherical shell centered on the origin, and consider a point P
inside the shell, on the xy plane. The electric field at point P is zero, by the shell theorem.
Recall that in the usual proof of the shell theorem, one draws two cones opening out of
P in opposite directions. The charges contained in each cone produce canceling electric fields.
Now imagine shrinking the spherical shell towards the xy plane, so it becomes elliptical. The
crucial insight is that the shell theorem argument above still works, for points on the xy
plane. When we squash the shell all the way down to the xy plane, it becomes a disc, with
zero electric field on it. This is thus a valid charge distribution for a disc-shaped conductor,
and by the uniqueness theorem, it’s the only one.
By√keeping track of how much charge gets squashed to radius [r, r + dr], we find σ(r) ∝
R/ R2 − r2 , and fixing the proportionality constant gives
Q
σ(r) = √ .
4πR R2 − r2
4 Electrical Conduction
We now leave the world of electrostatics and consider magnetostatics, the study of steady currents.
Idea 6
In a conductor with conductivity σ, the current density is
J = σE.
Alternatively, E = ρJ where ρ is the resistivity. The current and charge density satisfy
∂ρ
∇·J=− .
∂t
10
The current passing through a surface S at a given time is

Z
I= J · dS.
S
Since J ∝ E, we have Ohm’s law V = IR, where V is the voltage drop across the resistor.
The power dissipated in a resistor is P = IV . The resistance R adds in series, while 1/R
adds in parallel.
[2] Problem 21 (HRK). A battery causes a current to run through a loop of wire.
(a) Suppose the wire makes a sharp corner. How do the charges know to turn around there?
(b) A copper wire with conductivity σ is joined to an iron wire with conductivity σ 0 < σ. For
the current in both sections to be the same, the electric field in the iron wire must be higher.
How does that happen?
In general, the surface charge distribution in a DC circuit can be quite complex; the aspects shown
in these two short questions are just the beginning. For some more about this, see this paper.
[2] Problem 22 (PPP 22). Two students, living in neighboring rooms, decided to economize by
connecting their ceiling lights in series. They agreed they would each install a 100 W bulb in their
own rooms and that they would pay equal shares of the electricity bill. However, both tries to get
better lighting at the other’s expense. The first student installed a 200 W bulb, while the second
student installed a 50 W bulb. Which student subsequently failed the end-of-term examinations?
To warm up for DC circuits, we’ll consider some resistor network problems.

Idea 7
If any two points in a resistor network are at the same potential, nothing will change if the
two points are connected together and treated as one. More generally, the resistance of any
resistor directly connecting the two points may be changed freely.
Example 6
Consider the 3 × 3 grid below, where every edge is a resistor R.
11
Find the equivalent resistance between nodes 1 and 16.
Solution
By the above idea, we can short together nodes 2/3, and 14/15, by the diagonal symmetry
of the network. Next, we can break nodes 8 and 9 into two pieces.
This is valid because the separated nodes 8a/8b and 9a/9b still have the same potential in
the new network, by the diagonal symmetry. (This is using the above idea in reverse.) Now,
the circuit has been reduced to combinations of series and parallel resistors. The resistance
between 1 and 2/3 is R/2. The resistance between 2/3 and 14/15 is the combination of three
networks in parallel, and finally the resistance of 14/15 and 16 is R/2. Thus,
!
1 1 1 −1 1

1 13
Req = + + + + R = R.
2 3 3 2 2 7
You won’t see any resistor problems as complicated as this one for the rest of the training,
because they’re kind of contrived; the point of this example was just to show multiple uses
of symmetry techniques.
Example 7: PPP 23
A black box contains a resistor network and has two output terminals.
If a battery of voltage V is connected across the first terminal, the voltage across the second
terminal is V /2. If a battery of voltage V is connected across the second terminal, the voltage
across the first terminal is V . Find one possible configuration of the resistors inside the box.
12
Solution
A simple configuration with two equal resistors works.
When a battery is connected across II, the horizontal resistor doesn’t do anything. When a
battery is connected across I, the two resistors comprise a voltage divider.
[2] Problem 24 (IPhO 1996). Consider the following resistor network.
Find the equivalent resistance between A and B.
[3] Problem 25. Consider a cube of side length L whose edges are resistors of resistance R.
√
(a) Compute the resistance between two vertices a distance 3L apart.
√
(b) Compute the resistance between two vertices a distance 2L apart.
(c) Compute the resistance between two vertices a distance L apart.

√
(d) Generalize to vertices nL apart on an n-dimensional cube. (Give your answer in the form
of a summation.)
[2] Problem 26 (PPP 158). Consider the circuit below, where every resistor is 1 Ω.
(a) Find the equivalence resistance between the input terminals.
(b) Do the same in the case where the chain is infinitely long.
[3] Problem 27 (PPP 159-161). Superposition can be a useful trick to analyze circuit networks.
(a) Consider an infinite two-dimensional grid of identical resistors R.
13
Find the equivalent resistance between two neighboring points by considering the superposition
of 1 A flowing into one point, and 1 A flowing out the other.
(b) What would the equivalent resistance be if the resistor directly connecting the two neighboring
points was removed?
(c) Now consider an icosahedron of identical resistors R. By superposing appropriate current

distributions, find the equivalent resistance between two neighboring vertices.
Idea 8
In a circuit of resistors and batteries, Kirchoff’s loop rule states that the sum of the voltage
drops around a loop is zero. Kirchoff’s junction rule states that the net current flowing into a
vertex is zero. (This is technically nonzero, because of the effect of problem 21, but negligible
because wires have tiny capacitance.)
Remark
If the sum of the voltage drops around a loop is zero, then why would current ever want to
flow? After all, if you had a circular tube of water, the water would never flow, because the
net drop in height along the circle is zero. The reason current flows in circuits with batteries
is that within the battery, charges are moved from lower to higher electric potential energy,
just like how a pump could be used to move water upward to start a liquid circuit, by an
“electromotive force”.
But this immediately raises the question: what is this specific force? It can’t be the electric
force, because we just established that it’s pointing the wrong way. It’s not a magnetic
effect. For some setups, it is literally a mechanical force like a pump: in the Van der
Graaff generator, a motor drives the charges on a statically charged conveyor belt to higher
potential. But that’s not how batteries work.
In a battery, there is no specific force pushing charges from low to high electric potential.
Instead, the charges just jiggle around randomly, and the result emerges from the effects of
their many collisions. To understand this, consider a gravitational analogy.
14
Consider an ideal gas at temperature T released in the trough shown above. The gas
molecules will randomly collide, sometimes being propelled upward by chance. Sometimes, a
gas molecule will climb the hill and fall into the deep hole, at which point it is unlikely to
come out again. Thus, if the hole begins empty, it is energetically favorable for gas molecules
to fill it. But there is no attractive force pulling molecules up along the slope! Gravity
always points down; molecules go up the slope when they are randomly bounced that way.
This is essentially how the potential in an initially neutral battery is set up. The hole
corresponds to the lower energy state an electron can reach inside the anode, but there is no
long-range force pushing it there, just the average effect of random collisions.
[2] Problem 28 (Purcell 4.10). The basic ingredient in older voltmeters and ammeters is the gal-
vanometer, a device to measure very small currents. (It works via magnetic effects, but the exact
mechanism isn’t important here.) Inherent in any galvanometer is some resistance Rg , so a physical
galvanometer can be represented by the system shown below.
Consider a circuit such as the one shown, with all quantities unknown. We want to measure
the current flowing across point A and the voltage difference between points B and C. Given a
galvanometer with known Rg , and also a supply of known resistors (ranging from much smaller to
much larger than Rg ), how can you accomplish these two tasks? Explain how to construct your
two devices (called an ammeter and voltmeter), and also how you should insert them in the given
circuit. You will need to make sure that you (a) affect the given circuit as little as possible, and (b)
don’t destroy your galvanometer by passing more current through it than it can handle.
15
[2] Problem 29. 01^ USAPhO Quarterfinal 2009, problems 3 and 4.
[3] Problem 30. INPhO 2021, problem 1. A nice problem on practical circuit measurements. Note
that the question statement is a bit vague. You are supposed to keep track of quantities of order
RA /R and R/RV , but you are allowed to neglect quantities as small as RA /RV .
16
Electromagnetism III: Magnetostatics

Chapters 4 and 6 of Purcell cover DC circuits and magnetostatics, as does chapter 5 of Griffiths. For
advanced circuits techniques, see chapter 9 of Wang and Ricardo, volume 2. Chapter 5 of Purcell
famously derives magnetic forces from Coulomb’s law and relativity. It’s beautiful, but not required
to understand chapter 6; we will cover relativistic electromagnetism in depth in R3. For further
discussion, see chapters II-12 through II-15 of the Feynman lectures. There is a total of 79 points.
1 Static DC Circuits
We continue with DC circuits, in more complex setups than in E2.
Idea 1
When analyzing circuits, it is sometimes useful to parametrize the currents in the circuits
in terms of the current in each independent loop. This is typically more efficient, because it
enforces Kirchoff’s junction rule automatically, leading to fewer equations.
Example 1: Wheatstone Bridge
Find the current through the following circuit, if the battery has voltage V .
Solution
This circuit can’t be simplified using series and parallel combinations, so instead we
use Kirchoff’s rules directly. From the diagram, we see the circuit has three loops.
Let I1 be the clockwise current on the left loop, I2 be the clockwise current through
the top-right loop, and I3 be the clockwise current through the bottom-right loop. For
instance, this means that the current flowing downward through the top-left resistor is I1 −I2 .
The three Kirchoff’s loop rule equations are

3I1 R − I2 R − 2I3 R = V,
4I2 R − I1 R − I3 R = 0,
4I3 R − 2I1 R − I2 R = 0.
Adding the last two equations shows that
I1 = I2 + I3
1
and plugging this back in shows that 3I2 = 2I3 , so we have

2 3
I2 = I1 , I3 = I1 .
5 5
Since the answer to the question is just I1 , we can now plug this back into the first equation,

V 2 6 7
= 3I1 − I2 − 2I3 = 3 − − I1 = I1 .
R 5 5 5
This gives the answer, 5V /7R.
Incidentally, the circuit above is also called a Wheatstone bridge. We note that the current
through the middle resistor is zero when the ratios between the top and bottom resistances
match on both sides of it. Hence if three of these outer resistances are known, we can adjust
one of them until the current through the middle resistor vanishes, thereby measuring the
fourth resistor.
Idea 2
Since Kirchoff’s loop equations are linear, currents and voltages in a DC circuit with multiple
batteries can be found by superposing the currents and voltages due to each battery alone.
Idea 3: Thevenin’s Theorem and Norton’s Theorem

Consider any system of batteries and resistors, with two external terminals A and B. Suppose
that when a current I is sent into A and out of B, then a voltage difference V = VA − VB
appears. From an external standpoint, the function V (I) is all we can measure.
Now, by the linearity of Kirchoff’s rules, V (I) is a linear function, so we can write
V (I) = Veq + IReq .
In other words, V (I) is exactly the same as if the entire system were a resistor Req in series
with a battery with emf Veq (with the positive end pointing towards A). This generalizes
the idea of replacing a system of resistors with an equivalent resistance, and is known as
Thevenin’s theorem.
We can also flip this around. Note that I(V ) must also be a linear function, and we can write
V
I(V ) = Ieq + .
Req
This is precisely the I(V ) of an ideal current source Ieq (sending current towards B) in
parallel with a resistor Req . (An ideal current source makes a fixed current flow through it,
just like a battery creates a fixed voltage across it.) This is known as Norton’s theorem.
Since these functions are inverses of each other, you can see that the Req ’s in both equations
above are the same (both are equal to the ordinary equivalent resistance), and Veq = −Ieq Req .
2
Example 2
Consider some batteries connected in parallel, with emfs Ei and internal resistances Ri . What
is the Thevenin equivalent of this circuit?
Solution
The equivalent resistance is simply
!−1
X 1
Req = .
Ri
i
To infer Veq , we just need one more V (I) value. The most convenient is to set V = 0, shorting
all of the batteries. Each battery alone would produce a current of Ei /Ri , so
!
X Ei
0 = Veq + Req .
Ri
i
Thus, we have
! −1
X Ei X 1
Veq =   .
Ri Rj
i j
Remark
With ideal batteries, it’s easy to set up circuits that don’t make any sense.
For example, in the above circuit, Kirchoff’s rules don’t determine the currents; they only
say that i1 + i2 = 1 A. If the emfs of the batteries were different, the situation would be
even worse: the equations would be contradictory, with no solution at all! In real life, this is
avoided because all batteries have some internal resistance. Adding such a resistance to each
battery, no matter how small, resolves the problem and gives a unique solution.
[2] Problem 1 (Purcell 4.12). Consider the circuit below.
3
(a) Find the potential difference between points a and b.
(b) Find the equivalent Thevenin resistance and emf between points a and b.
[2] Problem 2 (Wang). A circuit containing batteries and resistors has two terminals. When an ideal
ammeter is connected between them, the reading is I1 . When a resistor R is connected between
them, the current through the resistor is I2 , in the same direction. What would be the reading V
01W
of an ideal voltmeter connected between them?
Now we give a few problems on current flow through continuous objects. Fundamentally, all one
needs for these problems is the definition J = σE, and superposition.
Example 3
Consider two long, concentric cylindrical shells of radii a < b and length L. The volume
between the two shells is filled with material with conductivity σ(r) = k/r. What is the
resistance between the shells, and the charge density?
Solution
To find the resistance, we compute the current I when a voltage V is applied between the
shells. By symmetry, in the steady state the current density must be
I
J(r) = r̂.
2πrL
On the other hand, we also know that
b
I(b − a)
Z Z
I
V = E · dr = dr =
a 2πrLσ 2πkL
from which we conclude

b−a
R= .
2πkL
4
Note that the radial electric field between the shells is constant, so
V
E(r) = r̂.
b−a
This means that in the steady state, there must be a nonzero charge density between the
shells. (If there weren’t, then we would have E(r) ∝ 1/r, rather than a constant.)
To find the charge density explicitly, it’s easiest to use Gauss’s law in differential form in
cylindrical coordinates. We use the form of the divergence derived in E1,
1 ∂(rEr ) 1 V ρ
∇·E= + (other terms) = =
r ∂r rb−a 0
thus showing that the charge density is proportional to 1/r. Of course, we could also get this
result by applying Gauss’s law in integral form, to concentric spheres.
[2] Problem 4 (Grad). A washer is made of a material of resistivity ρ. It has a square cross section
of length a on a side, and its outer radius is 2a. A small slit is made on one side and wires are
connected to the faces exposed.
Find the resistance of the washer. (Hint: first argue that no current flows radially.)
[3] Problem 5 (BAUPC 1995). An electrical signal can be transferred between two metallic objects
buried in the ground, where the current passes through the Earth itself. Assume that these objects
are spheres of radius r, separated by a horizontal distance L r, and suppose both objects are
buried a depth much greater than L in the ground. If the Earth has uniform resistivity ρ, find the
approximate resistance between the terminals. (Hint: consider the superposition principle.)
[3] Problem 6 (PPP 162). A plane divides space into two halves. One half is filled with a homogeneous
conducting medium, and physicists work in the other. They mark the outline of a square of side
a on the plane and let a current I0 in and out at two of its neighboring corners. Meanwhile, the
measure the potential difference V between the two other corners.
5
Find the resistivity ρ of the medium.
[3] Problem 7 (MPPP 174). We aim to measure the resistivity of the material of a large, thin,
homogeneous square metal plate, of which only one corner is accessible. To do this, we chose points
A, B, C and D on the side edges of the plate that form the corner.
Points A and B are both 2d from the corner, whereas C and D are each a distance d from it. The
length of the plate’s sides is much greater than d, which, in turn, is much greater than the thickness
t of the plate. If a current I enters the plate at point A, and leaves it at B, then the reading on a
voltmeter connected between C and D is V . Find the resistivity ρ of the plate material.
Remark
Setups like those in the previous two problems are commonly used to measure resistivities,
but why do they use a complicated “four terminal” setup? Wouldn’t it have been easier to
just attach two terminals, send a current I through them, and measure the voltage drop
V ? The problem with this is that it also picks up the resistance R of the contacts between
the terminals and the material, along with the resistances of the wires. By having a pair of
terminals measure voltage alone, drawing negligible current, we avoid this problem.
[4] Problem 8. [A] This problem is just for fun; the techniques used here are too advanced to appear on
Olympiads. We will prove Rayleigh’s monotonicity law, which states that increasing the resistance
of any part of a resistor network increases the equivalent resistance between any two points. This
may seem obvious, but it’s actually tricky to prove. The following is the slickest way.
(a) Consider a graph of resistors, where a battery is attached across two of the vertices, fixing
their voltages. Write an expression for the total power dissipated, assuming the voltages at
each vertex are Vi and the resistances are Rij .
6
(b) The voltages Vi at all the other vertices are determined by Kirchoff’s rules. But suppose you
didn’t know that, or didn’t want to set up those equations. Remarkably, it turns out that
you can derive the exact same results by simply treating the voltages Vi as free to vary, and
setting them to minimize the total power dissipated! Show this result. (This is an example of
a variational principle, like the principle of least action in mechanics.)
(c) For any network of resistors, show that P = V 2 /R when V is the battery voltage applied
across two vertices, R is the equivalent resistance between them, and P is the total power
dissipated in the resistors. (This is intuitive, but it’s worth showing in detail to assist with
the next part.)
(d) By combining all of these results, prove Rayleigh’s monotonicity law.
(e) We can use Rayleigh’s monotonicity law to prove some mathematical results. Consider the
resistor network shown below, where the variables label the resistances.
By considering the resistances before and after closing the switch P Q, show that the arithmetic
mean of two numbers is at least the geometric mean.
(f) Consider the resistor network shown below.
By closing all the switches, show that the arithmetic mean of n numbers is at least the
harmonic mean.
Remark
You might think that Rayleigh’s monotonicity law is too obvious to require a proof; if you
decrease a resistance, how could the net resistance possibly go up? In fact, this kind of
non-monotonicity occurs very often! For example, Braess’s paradox is that fact that adding
more roads can slow down traffic, even when the total number of cars stays the same. A U.S.
Physics Team coach has argued that allowing more team strategies can make a basketball
team score less. For more on this subject, see the paper Paradoxical behaviour of mechanical
and electrical networks or this video.
7
Remark
Circuit questions can get absurdly hard, but at some point they start being more about
mathematical tricks than physics. As a result, I haven’t included any such problems here;
they tend not to appear on the USAPhO or IPhO, or in college physics, or in real life, or really
anywhere besides a few competitions. On the other hand, you might find such questions fun!
For some examples, see the Physics Cup problems 2013.6, 2017.2, 2018.1, and 2019.4.
2 RC Circuits
Next we’ll briefly cover RC circuits, our first exposure to a situation genuinely changing in time.
Example 4: CPhO
The capacitors in the circuit shown below were initially neutral. Then, the circuit is allowed
to reach the steady state.
After a long time, what is the charge stored on the 10 mF capacitor?
Solution
After a long time, no current flows through the capacitors, so there is effectively a single loop
in the circuit. It has a total resistance 60 Ω and a total emf 6 V, so the current is I = 0.1 A.
Using this, we can straightforwardly label the voltages everywhere on the outer loop.
8
To finish the problem, we need to know the voltage V0 of the central node, so we need one
more equation. That equation is charge conservation: the fact that the central part of the
circuit, containing the inner plates of the three capacitors, begins and remains uncharged.
Suppressing units, this means
66
20(26 − V0 ) + 20(7 − V0 ) + 10(0 − V0 ) = 0, V0 = V
5
from which we read off the answer,
Q = CV = 0.132 C.
[3] Problem 10 (Purcell 4.18). Consider the two RC circuits below.
(a) The circuit shown below contains two identical capacitors and two identical resistors, with
initial charges as shown above at left. If the switch is closed at t = 0, find the charges on the
capacitors as functions of time.
(b) Now consider the same setup with an extra resistor, as shown above at right. Find the
maximum charge that the right capacitor achieves. (Hint: the methods of M4 can be useful.)
[3] Problem 12 (Kalda). Three identical capacitors are placed in series and charged with a battery
of emf E. Once they are fully charged, the battery is removed, and simultaneously two resistors are
connected as shown.
Find the heat dissipated on each of the resistors after a long time.
[3] Problem 13 (Kalda). Find the time constant of the RC circuit shown below.
9
[3] Problem 14 (MPPP 175/176). A metal sphere of radius R has charge Q and hangs on an insulating
cord. It slowly loses charge because air has a conductivity σ. In all cases, neglect any magnetic or
radiation effects.
(a) Find the time for the charge to halve.
(b) You should have found that the time is independent of the radius R of the sphere, which follows
directly from dimensional analysis. Can you show that, in fact, it is completely independent
of the shape? (This doesn’t just follow from dimensional analysis, because the shape might
be described by dimensionless numbers, such as the eccentricity of an ellipsoid.)
(c) Air has a conductivity of σ ∼ 10−13 Ω−1 m−1 , while water has a conductivity of σ ∼ 10−2 Ω−1 m−1 .
About how long does the charge on an object last, if it is in air or water?
01h
This problem generalizes USAPhO 2010, problem A2, which you can compare.
01h
[5] Problem 15. IPhO 1993, problem 1. A really neat question with real-world relevance.
[5] Problem 16. IPhO 2007, problem “orange”. A combination of mechanics and RC circuits.
3 Computing Magnetic Fields

Idea 4
The Biot–Savart law is
ds × r
I
µ0 I
B= .
4π r3
As a consequence, we have Ampere’s law,
I
B · ds = µ0 I, ∇ × B = µ0 J
as well as Gauss’s law for magnetism,

I
B · dS = 0, ∇ · B = 0.
Idea 5
The force on a stationary wire carrying current I in a magnetic field B is
Z
F = I ds × B.
10
The energy of a magnetic field is

Z
1
U= B 2 dV.
2µ0
The magnetic dipole moment of a planar current loop of area A and current I is m = IA,
with m directed perpendicular to the loop by the right-hand rule.
You should have already seen basic examples of using the Biot–Savart law in Halliday and Resnick,
such as the field of a circular ring of current on its axis. We’ll start with some problems that are
similarly straightforward, but more technically complex.
[3] Problem 17 (Purcell 6.11). A spherical shell with radius R and uniform surface charge density σ
spins with angular frequency ω about a diameter.
(a) Find the magnetic field at the center.
(b) Find the magnetic dipole moment of the sphere.
(c) Sketch the magnetic field.
[2] Problem 18 (Purcell 6.12). A ring with radius R carries a current I. Show that the magnetic
field due to the ring, at a point in the plane of the ring, a distance r from the center, is given by
µ0 I π (R − r cos θ)R dθ
Z
B= .
2π 0 (r + R2 − 2rR cos θ)3/2
2
In the r R limit, show that

µ0 m
B≈
4π r3
where m = IA is the magnetic dipole moment of the ring.
[3] Problem 19 (Purcell 6.14). Consider a square loop with current I and side length a centered at
the origin. Show that the magnetic field at r, where r a and r is parallel to one of the square’s
sides, is B ≈ (µ0 /4π)(m/r3 ) as in the previous problem. Be careful with factors of 2!
Idea 6
The results you have found above, for the fields far from currents, are special cases of the
general magnetic dipole field: far from a magnetic dipole with magnetic moment m, its
magnetic field is just the same as the electric field of an electric dipole,
µ0 m µ0
B(r) = 3
(2 cos θ r̂ + sin θ θ̂) = (3(m · r̂)r̂ − m).
4πr 4πr3
As with the electric dipole field, you don’t need to memorize this result, but you should
remember that it’s proportional to the dipole moment, falls off as 1/r3 , and be able to sketch
it. Of course, static electric and magnetic fields behave differently; when you get inside an
electric dipole the field reverses direction, but this isn’t true for a magnetic dipole.

We now give a few arguments for computing fields using symmetry.
11
Example 5: PPP 31
An electrically charged conducting sphere “pulses” radially, i.e. its radius changes periodically
with a fixed amplitude. What is the net pattern of radiation from the sphere?
Solution
There is no radiation. By spherical symmetry, the magnetic field can only point radially.
But then this would produce a magnetic flux through a Gaussian sphere centered around
the pulsing sphere, which would violate Gauss’s law for magnetism. So there is no magnetic
field at all, and since radiation always needs both electric and magnetic fields (as you’ll see
in E7), there is no radiation at all. In fact, outside the sphere the electric field is always
exactly equal to Q/4π0 r2 , in accordance with Coulomb’s law.
Example 6
Find the magnetic field of a very long cylindrical solenoid, of radius R and n turns per unit
length, carrying current I.
Solution
Orient the solenoid along the vertical direction and use cylindrical coordinates. By symmetry,
the field must be independent of z. Now consider the radial component of the magnetic field
Br . Turning the solenoid upside-down is equivalent to reversing the current. But the former
does not flip Br while the latter does, so we must have Br = 0.
Now, by rotational symmetry, the tangential component Bφ must be uniform. But then
Ampere’s law on any circular loop gives Bφ (2πr) = 0, so we must have Bφ = 0 as well.
The only thing left to consider is Bz . By applying Ampere’s law to small vertical rectangles, we
see that Bz is constant unless that rectangle crosses the surface of the solenoid. Furthermore,
Bz must be zero far from the solenoid, so it must be zero everywhere outside the solenoid.
Now, for a rectangle of height h that does cross the surface, Ampere’s law gives
I
B · ds = Bzin h = µ0 Ienc = µ0 nIh
which tells us that Bzin = µ0 nI.
Remark
The above analysis is not quite how real solenoids behave for several reasons. First, we
didn’t account for the discreteness of the wires. We just treated them as forming a uniform
current per length K = nI, which is how we wrote Ienc = nIh. This is valid when you don’t
care about looking too close to the wires themselves.
The fact that solenoids are made by winding real wires means there is another contribution
12
to the current, even in the limit n → ∞. The wires are wound with a small slope, since a net
current I still has to move along the solenoid. Another way of saying this is that the current
per length along the solenoid surface is K = nI θ̂ + (I/2πR)ẑ. This causes a tangential
magnetic field Bφ = µ0 I/2πr outside the solenoid.
Another factor is that a real solenoid isn’t infinitely long, and end effects are important.
You’ll analyze these in a slick way in problem 22.
[2] Problem 21. A toroidal solenoid is created by wrapping N turns of wire around a torus with a
rectangular cross section. The height of the torus is h, and the inner and outer radii are a and b.
(a) In the ideal case, the magnetic field vanishes everywhere outside the toroid, and is purely
tangential inside the toroid. Find the magnetic field inside the toroid.
(b) There is another small contribution to the magnetic field due to the winding effect mentioned
above. Roughly what does the resulting extra magnetic field look like? If you didn’t want
this additional field, how would you design the solenoid to get rid of it?
[3] Problem 22 (Purcell 6.63). A number of simple facts about the fields of solenoids can be found
by using superposition. The idea is that two solenoids of the same diameter, and length L, if joined
end to end, make a solenoid of length 2L. Two semi-infinite solenoids butted together make an
infinite solenoid, and so on.
Prove the following facts.

(a) In the finite-length solenoid shown at left above, the magnetic field on the axis at the point
P2 at one end is approximately half the field at the point P1 in the center. (Is it slightly more
than half, or slightly less than half?)
13
(b) In the semi-infinite solenoid shown at right above, the field line FGH, which passes through
the very end of the winding, is a straight line from G out to infinity.
(c) The flux through the end face of the semi-infinite solenoid is half the flux through the coil at
a large distance back in the interior.
(d) Any field line that is a distance r0 from
√ the axis far back
√ in the interior of the coil exits from
the end of the coil at a radius r1 = 2r0 , assuming 2r0 is less than the solenoid radius.
Part (c) tells us that even in an ideal solenoid, we always lose at least half the flux to fringing!
[3] Problem 23 (MPPP 160). Two infinite parallel wires, a distance d apart, carry electric currents
with equal magnitudes but opposite directions. In this problem, we find the shape of the magnetic
field lines using a neat trick.
(a) Argue that if we rotated B by 90◦ at each point, it would produce a valid electrostatic field
E. (Hint: consider what happens when we rotate the B field of a single wire, first.)
(b) Argue that the field lines of B are the same as the equipotentials of this artificial E, and use
this to find the field lines.
This is a common trick used when working with vortices in two dimensions, where it converts
vortices to sources and sinks.
[2] Problem 24 (IPhO 1996). Two straight, long conductors C+ and C− , insulated from each other,
carry current I in the positive and the negative ẑ direction respectively. The cross sections of the
conductors are circles of diameter D in the xy plane, with a distance D/2 between the centers.
The current in each conductor is uniformly distributed. Find the magnetic field in the space between
the conductors.
[3] Problem 25 (MPPP 157). A regular tetrahedron is made of a wire with constant resistance per
unit length. A current I is sent into one vertex and removed from another vertex, as shown.
01h
Find the magnetic field at the center of the tetrahedron.
[5] Problem 26. APhO 2013, problem 1. A neat question on a cylindrical RC circuit that uses
many of the techniques we’ve covered so far.
14
Electromagnetism IV: Lorentz Force

The problems here mostly use material covered in previous problem sets, though chapter 5 of Purcell
covers relativistic field transformations. For further interesting physical examples, see chapter II-29
of the Feynman lectures. There is a total of 84 points.
1 Electrostatic Forces
Idea 1: Lorentz Force
A charge q in an electromagnetic field experiences the force
F = q(E + v × B).
In particular, a stationary wire carrying current I in a magnetic field experiences the force
Z
F = I ds × B.
Example 1: PPP 183
A small charged bead can slide on a circular, frictionless insulating ring. A point-like electric
dipole is fixed at the center of the circle with the dipole’s axis lying in the plane of the circle.
Initially the bead is in the plane of symmetry of the dipole, as shown.
Ignoring gravity, how does the bead move after it is released? How would the bead move if
the ring weren’t there?
Solution
Set up spherical coordinates so that the dipole is in the ẑ direction. Then
kp cos θ
V (r, θ) = .
r2
Since the ring fixes r, the potential on the ring is just proportional to cos θ, which is in turn
proportional to z. But a potential linear in z is equivalent to a uniform downward field, so
the bead oscillates like the mass of a pendulum, with amplitude π/2.
1
The answer remains the same when the ring is removed! Conservation of energy states that
kqp cos θ 1
+ mv 2 = 0.
r2 2
Let N be the normal force. Then accounting for radial forces gives
∂V v2
N +q = .
∂r r
However, plugging in our conservation of energy result for v 2 shows that N = 0, so the ring
doesn’t actually do anything, and it may be removed without effect.
Example 2
A parallel plate capacitor with separation d and area A is attached to a battery of voltage
V . Suppose the plates are moved together with speed v. Verify that energy is conserved.
Solution
The capacitance is C = A0 /d. The power supplied by the battery is
dQ dC
Pbatt = IV = V =V2 .
dt dt
On the other hand, the rate of change of the energy stored in the battery is

d 1 2 1 dC
Pcap = CV = V2 .
dt 2 2 dt
At first glance, there seems to be a problem. But then we remember that there is an attractive
force between the plates, so the plates do work on whatever is moving them together,
QE QV 1 v 1 dC
Pmech = F v = v= v = CV 2 = V 2 .
2 2d 2 d 2 dt
where E is the electric field inside the capacitor. Thus, Pbatt = Pcap + Pmech as required.
[2] Problem 1 (PPP 193). Two positrons are at opposite corners of a square of side a. The other two
corners of the square are occupied by protons. All particles have charge q, and the proton mass M
is much larger than the positron mass m. Find the approximate speeds of the particles much later.
[3] Problem 2 (PPP 114). A small positively charged ball of mass m is suspended by an insulating
thread of negligible mass. Another positively charged small ball is moved very slowly from a large
distance until it is in the original position of the first ball. As a result, the first ball rises by h.
2
How much work has been done?
[3] Problem 3 (PPP 71). Two small beads slide without friction, one on each of two long horizontal
parallel fixed rods a distance d apart.
The masses of the beads are m and M and they carry charges q and Q. Initially, the larger mass
M is at rest and the other one is far away approaching it at a speed v0 . For what values of v0 does
the smaller bead ever get to the right of the larger bead?
[2] Problem 4 (PPP 192). Classically, a conductor is made of nuclei of positive charge fixed in place,
and electrons that are free to move.
(a) Consider a solid conductor in a gravitational field g. Argue that the electric field inside the
conductor is not zero; find out what it is.
(b) Now suppose a positron is placed at the center of a hollow spherical conductor in a gravitational
01m
field g. Find its initial acceleration.
[3] Problem 5. USAPhO 2008, problem B2. You may ignore part (c), which was removed in the
01^
final version of the exam, though you can also do it for extra practice.
[5] Problem 7. 01h IPhO 2004, problem 1. A nice question on the dynamics of a multi-part system.
2 The Lorentz Force

Idea 2
The Lorentz force
dp
F= = q(E + v × B).
dt
In some problems below, you might worry that this relation is modified due to relativistic
effects, but it isn’t. The Lorentz force expression is always correct, and it’s all you need for
almost all the problems in this section.
3
What does change in relativity is the relation between p and v,

1
p = γmv, γ=p .
1 − v 2 /c2
The relativistic energy is also modified to

1
E = γmc2 = mc2 + mv 2 + . . . .
2
We will return to this subject in more detail in R2.
Example 3: Kalda 163
A beam of electrons, of mass m and charge q, is emitted with a speed v almost parallel to
a uniform magnetic field B. The initial velocities of the electrons have an angular spread
of α 1, but after a distance L the electrons converge again. Neglecting the interaction
between the electrons, what is L?
Solution
Consider an electron initially traveling at an angle α to the magnetic field. This electron has
a speed vk = v cos α ≈ v parallel to the field, and a speed v⊥ v sin α ≈ vα perpendicular to
the field. The component vk always stays the same, while v⊥ rotates, so the electron spirals
along the field lines.
The acceleration of the electron is

F qv⊥ B
a⊥ = = .
m m
The perpendicular velocity component rotates through a circle in velocity space of circumfer-
ence 2πv⊥ . After one such circle, the total perpendicular displacement is zero, so the beam
refocuses. Thus we have
2πv⊥ 2πmv
L= vk ≈ .
a qB
In other words, this setup acts like a magnetic “lens”.
[3] Problem 8 (Griffiths 5.17). A large parallel plate capacitor with uniform surface charge σ on the
upper plate and −σ on the lower is moving with a constant speed v as shown.
(a) Find the magnetic field between the plates and also above and below them.
(b) Find the magnetic force per unit area on the upper plate, including its direction.
4
(c) What happens to the net force between the plates in the limit v → c? Explain your result
using some basic ideas from special relativity.
[3] Problem 9. EFPhO 2012, problem 7. An elegant Lorentz force problem with wires. (If you enjoy
this problem, consider looking at IdPhO 2020, problem 1B, which has a similar setup but requires
three-dimensional reasoning. The official solutions are here.)
[4] Problem 10 (Purcell 6.35/INPhO 2008.6). Consider the arrangement shown below.
The force between capacitor plates is balanced against the force between parallel wires carrying
current in the same direction. A voltage alternating sinusoidally with angular frequency ω is applied
to the parallel-plate capacitor C1 and also to the capacitor C2 , and the current is equal to the
current through the rings. Assume that s a and h b.
Suppose the weights of both sides are adjusted to balance without any applied voltage, and C2
is adjusted so that the time-averaged downward forces on both sides are equal. Show that
r
1 √ b C2
√ = 2π aω .
µ 0 0 h C1
The left-hand side is equal to c, as we’ll show in E7, so this setup measures the speed of light.
[3] Problem 11. An electron beam is accelerated from rest by applying an electric field E for a time
t, and subsequently guided by magnetic fields. These magnetic fields are produced with a series of
coils, which carry currents Ii .
Now suppose the apparatus is repurposed to shoot proton beams. Suppose a proton beam is
accelerated from rest by applying an electric field E for a time t (in the opposite direction). Let an
electron have mass m and a proton have mass M .
(a) Find the currents Ii needed so that the proton follows the same trajectory the electron did,
assuming V is small enough that both the electron and proton are nonrelativistic.
(b) How does the answer change if relativistic corrections are accounted for?
5
[5] Problem 12. 01h

IPhO 2000, problem 2. A solid question on the Lorentz force with real-world
01T
relevance. Requires a little relativity, namely the expressions for relativistic momentum/energy.
[4] Problem 13. IPhO 1996, problem 2. An elegant problem on particles in a magnetic field.
3 Permanent Magnets
[3] Problem 14. Consider a current loop I in the xy plane in a constant magnetic field B.
(a) Show that the net force on the loop is zero.
(b) Show that the torque is

τ =m×B
where the magnetic moment is
m = IAẑ
where A is the area of the loop. For simplicity, you can show this in the case where the current
loop is a square of side length L, whose sides are aligned with the x and y axes. (The proof
for a general loop shape requires some vector calculus, but you can attempt it for a challenge.
You’ll need the double cross product identity, a × (b × c) + b × (c × a) + c × (a × b) = 0.)
Idea 3
The force on a small current loop is
F = (m · ∇)B
but this requires some tricky vector calculus to derive, shown here. This expression, and the
torque expression found in problem 14, can be found by differentiating the potential energy
U = −m · B.
All of these results also hold for electric dipoles, if we replace m with p and B with E.
Remark
The expression for the potential energy above is notoriously subtle. Here’s the problem: we
know the Lorentz force on a charge is qv × B, which means magnetic fields never do work.
So how can they be associated with a nonzero potential energy?
There are two levels of explanation. First, suppose the magnetic dipole is made of charges
moving in a loop. When such a current loop is placed in a magnetic field, and moved
or rotated, mechanical work can be done on the loop. But at the same time, there will
be an induced emf in the loop, which speeds up or slows down the current. The work
done by these two effects perfectly cancels, so that the energy of the loop stays constant.
For this kind of dipole, the expression for U doesn’t indicate the total energy, but only
the “mechanical” potential energy. However, differentiating it still gives the right forces
and torques. (Some further discussion of this point is in chapter II-15 of the Feynman lectures.)
6
On the other hand, the magnetic dipole moment of a common bar magnet doesn’t come from
charges moving in a loop! Instead, it comes from the intrinsic magnetic dipole moments of
the unpaired electrons in the magnet. These kinds of dipole moments aren’t composed of
any moving subcomponents; they are an elementary and immutable property of the electron,
like its mass or charge. In these cases, U = −m · B really is the total energy, and the
magnetic field can do work. You won’t hear much about these elementary dipole moments in
introductory books, because they can only be properly understood by combining relativity
and quantum mechanics, but they’re responsible for most magnetic phenomena.
Example 4
If a magnet is held over a table, it can pick up a paper clip. If the paper clip is removed, it
can pick up another paper clip just as well, and this process can seemingly continue forever
without any effect on the magnet. Since the magnet does work on each paper clip, doesn’t
this mean a permanent magnet is an infinite energy source?
Solution
This is the kind of question that makes magnets feel so mysterious. They’re basically the
only everyday example of a long range force besides gravity (in fact, Kepler once thought
the Sun acted on the planets like a giant magnet), and as such they’ve inspired countless
attempts at perpetual motion machines. Many people have spent years of their lives trying
to get elaborations of this example to work.
To see why this doesn’t work for a bar magnet, just replace the word “magnet” with “charge”.
It’s true that a positive charge can attract a negative charge to it. And if the negative
charge is then removed, the positive charge can then attract another negative charge to
it. But conservation of energy isn’t violated, because the force from the positive charge
is conservative: the work it does on the negative charge to draw it close is precisely the
opposite of the work an external agent needs to do to pull it away. The force of a magnet on
a paper clip is also conservative.
It’s also interesting to consider a slightly different case. Unlike a bar magnet, an electromagnet
(i.e. a magnet created by moving current in a loop) can be turned on and off with the flick
of a switch. Therefore, we might suspect that the following is a perpetual motion machine:
1. Turn on the electromagnet, which costs energy E0 .
2. Use it to lift a paper clip, increasing its potential energy by mgh.
3. Turn off the electromagnet, which costs energy E0 , while holding the paper clip.
4. Move the paper clip away; we’ve managed to raise it higher for free.
To see the problem, note that the attractive force between the magnet and paper clip arises
because the magnet induces a magnetic dipole moment in the paper clip, leading to a (m·∇)B
7
force. As the paper clip moves toward the magnet, its own dipole moment causes a changing
magnetic flux through the electromagnet, and thus an emf against the current. Therefore, it
costs extra energy to keep the current in the electromagnet steady. Since the qv × B Lorentz
force doesn’t do work, that energy must be precisely mgh, so nothing comes for free.
Remark
A compass needle is essentially a small magnetic dipole, whose dipole moment points towards
the end painted red. We can also approximate the Earth’s magnetic field as a dipole field.
Since the tangential component of this dipole field points north, the red end of the compass
points towards the geographic north pole, which is the Earth’s magnetic south pole.
By the way, a cheap compass calibrated to work in America or Europe won’t work well in
Australia. The reason is that the Earth’s magnetic field also has a radial component, which
acts to tip the compass needle up or down. The needle needs to be appropriately weighted
to stay horizontal, so that it can freely rotate, but the side that needs to be weighted differs
between the hemispheres.
[3] Problem 15 (Griffiths 6.23). A familiar toy consists of donut-shaped permanent magnets which
slide frictionlessly on a vertical rod.
Treat the magnets as dipoles with mass md and dipole moment m, with directions as shown above.
(a) If you put two back-to-back magnets on the rod, the upper one will “float”. At what height
z does it float?
8
(b) If you now add a third magnet parallel to the bottom one as shown, find the ratio x/y of the
two heights, using only a scientific calculator. (Answer: 0.85.)
[3] Problem 16. AuPhO 2019, problem 13. An elegant series of visual exercises on permanent
magnets, with practical applications. It will be useful to consult the answer sheet.
[3] Problem 17 (PPP 89). Two identical small bar magnets are placed on opposite ends of a rod of
length L as shown.
(a) Show that the torques the magnets exert on each other are not equal and opposite.
(b) Suppose the rod is pivoted at its center, and the magnets are attached to the rod so that
they can spin about their centers. If the magnets are released, the result of part (a) implies
that they will begin spinning. Explain how this can be consistent with energy and angular
momentum conservation, treating the latter quantitatively.
4 Point Charges
In this section we’ll give a sampling of classic problems involving just point charges in fields; these
will be a bit more mathematically advanced than the others in this problem set.
[3] Problem 18. A point charge q is released from rest a distance d from a grounded conducting
plane. Find the time until the point charge hits the plane. (Hint: consider Kepler’s laws.)
[3] Problem 19. A point charge of mass m and charge q is released from rest at the origin in the
fields E = E x̂, B = B ŷ. Find its position as a function of time by solving the differential equations
given by Newton’s second law, F = ma.
[3] Problem 20 (Wang). Two identical particles of mass m and charge q are placed in the xy plane
with a uniform magnetic field Bẑ. The particles have paths r1 (t) and r2 (t). Neglect relativistic
effects, but account for the interaction between the charges.
(a) Write down a differential equation describing the evolution of the separation r = r1 − r2 .
(b) Suppose that the initial conditions have been set up so that the particles orbit each other in
a circle in the xy plane, with constant separation d. What is the smallest d for which this
motion is possible?
[4] Problem 21. [A] Consider a point charge of mass m and charge q in the field of a magnetic
monopole at the origin,
g
B = 2 r̂.
r
In this problem we’ll investigate the strange motion that results.
(a) Argue that the speed v is constant.
9
(b) Show that the angular momentum L of the charge is not conserved, but that
V = L − qgr̂
is. The second term is the angular momentum stored in the fields of the charge and monopole.
(c) Show that the charge moves on the surface of a cone. (Hint: in spherical coordinates where
the z-axis is parallel to V, consider V · φ̂.) Sketch some typical trajectories.
One can do problem 19 slickly using field transformations, an advanced subject we will cover in R3.
Idea 4: Field Transformations
If the electromagnetic field is (E, B) in one reference frame, then in a reference frame moving
with velocity v with respect to this frame, the components of the field parallel to v are
Ek0 = Ek , Bk0 = Bk
while the components perpendicular are

v
E0⊥ = γ(E⊥ + v × B), B0⊥ = γ B⊥ − 2 × E .
c
Remark
The nonrelativistic limit of the field transformation is useful, but one has to be careful in
deriving it. You might think, what’s the need for care? Can’t we just send c → ∞, Taylor
expand the above expressions, and call it a day? The problem with this reasoning is that
there’s no such thing as setting c → ∞. You can’t change a fundamental constant, and
moreover this statement isn’t even dimensionally correct, as noted in P1. What we really
mean by the nonrelativistic limit is restricting our attention to some subset of possible
situations, within which relativistic effects don’t matter.
For example, if we have a bunch of point charges with typical speed v, then the nonrelativistic
limit is considering only situations where v/c is small. In other words, we are taking v/c → 0,
not c → ∞. Since the magnetic field of a point charge is v/c2 times the electric field, the
magnetic field ends up small. Now if we also consider boosts with small speeds v, then
expanding the field transformations to lowest order in v/c gives
v
E0 = E, B0 = B − × E.
c2
This is the nonrelativistic limit for situations where E/B c, also called the electric limit.
However, there’s another possibility. Suppose that we have a bunch of neutral wires. In this
case, it’s the electric fields that are small, E/B c. Using this in the transformations above,
we arrive at the distinct result
B0 = B, E0 = E + v × B
which apply for situations where E/B c, also called the magnetic limit.
10
You might think we could improve the approximation by combining the two,
v
E0 = E + v × B, B0 = B − ×E
c2
but this isn’t self-consistent. For example, if you apply a Galilean boost with speed v, and
then a boost with speed −v, you don’t get back the same fields you started with! A sensible
Galilean limit is only possible if E/B c or E/B c, which are called the electric and
magnetic limits, discussed further in this classic paper. It’s only in relativity that E and B
can be treated on an equal footing.
[3] Problem 22. Using the Galilean field transformations to solve problem 19.
(a) In the magnetic limit, show that the Lorentz force stays the same between frames, as it should.
Then use the field transformations to find an appropriate reference frame where the problem
becomes easy.
(b) In the electric limit, show that the Lorentz force stays the same up to terms that are order
(v/c)2 smaller, assuming B/E ∼ v/c2 . (This is fine, since we’re taking the limit v/c → 0
anyway.) Then use the field transformations to find an appropriate reference frame where the
problem becomes easy.
(c) The solutions you found in parts (a) and (b) should look very different, even though you
should have found only one type of behavior in problem 19. In fact, there is a critical value
of E/B separating the two kinds of behavior. What is this critical value, and why didn’t you
run into it when solving problem 19?
Example 5: Drude Theory
Model a conductor as a set of electrons, of charge q, mass m, and number density n, which are
completely free. Assume that in every small time interval dt, each electron has a probability
dt/τ of hitting a lattice ion, which randomizes the direction of its velocity. Under these
assumptions, compute the resistivity of the material.
Solution
First, suppose the electrons have some average momentum hpi each. Because the collisions
randomize the velocity, the average momentum falls exponentially with timescale τ ,
dhpi hpi
=− .
dt τ
On the other hand, if there is an applied field, a force term appears on the right,
dhpi hpi
=− + qE
dt τ
11
since F = dp/dt for each individual electron. In the steady state,
hpi = qEτ.
The current density is

nqhpi nq 2 τ
J = nqhvi = = E.
m m
Thus, the resistivity in the Drude model is
m
ρ= .
nq 2 τ
We can also compute the typical drift velocity,
qEτ
v= .
m
For values of m that give reasonable ρ, the value of v is a literal snail’s pace, which is why
people say that the electrons themselves move very slowly through a circuit. (Of course, a
current can get started in a circuit much faster, because when a battery is attached, each
moving electron pushes on the next one along the wire, and this wave of motion travels much
faster than the electrons themselves.)
Remark
Above we assumed there was a given probability of collision per unit time, but usually when
a particle flies through a medium, there is a given probability of collision per unit length it
travels. If we had made the latter assumption, then the typical energy of the electrons would
match the work done on them in the typical length
√ ` they travel, mv 2 /2 ∼ qE`, but that
implies √an average velocity that scales as v ∝ E. The analogue of Ohm’s law would then
be I ∝ V , completely contrary to observation!
A given probability of collision per unit time would be what we expect if the electrons had
constant speed, but how is this possible? The answer, as shown in X1, involves quantum
mechanics. The idea is that, due to the Pauli exclusion principle, the electrons in the
conductor have to occupy different quantum states, and the high density of electrons requires
most of them to have extremely high velocities, on the order of 1% of the speed of light! The
drift velocity is merely the tiny amount by which these velocities are shifted on average, so
their speeds are almost constant between collisions.
[2] Problem 23. Consider Drude theory again, but now suppose there is also a fixed magnetic field
Bẑ. In this case, J is not necessarily parallel to E, but the relation between the two can be described
by the “tensor of resistivity”. That is, the components are related by
X
Ei = ρij Jj .
j∈{x,y,z}
12
Calculate the coefficients ρij . Express your answers in terms of the quantities
m qB
ρ0 = , ω0 =
nq 2 τ m
as well as the parameter τ .
Since parallel currents attract, the currents within a single wire should contract. To estimate
this, consider a long wire of radius r. Suppose the atomic nuclei are fixed and have uniform
density, while the electrons move along the wire with speed v. Furthermore, assume that the
electrons contract, filling a cylinder of radius r0 < r with uniform negative charge density,
and that the wire is overall neutral. Find r0 .
Solution
The contraction of the electrons produces an overall inward electric field, and hence an
outward electric force on each electron, which balances the radially inward magnetic force.
Specifically, equilibrium occurs when E = vB.
Let the charge densities of the nuclei and electrons be ρ+ and ρ− . The magnetic field at
radius r is found by Ampere’s law, which gives
µ0 ρ− vr
(2πr)B = µ0 (ρ− v)(πr2 ), B= .
2
The electric field at radius r is found by Gauss’s law, which gives
1 1
(2πr)E = (ρ+ + ρ− )πr2 , E= (ρ+ + ρ− )r.
0 20
Note that both E and B are proportional to r. Then E = vB can be satisfied at all r simul-
taneously, which confirms that our assumption that ρ+ and ρ− were uniform is self-consistent.
Plugging these results into E = vB yields
v2
ρ+ + ρ− = ρ− (0 µ0 v 2 ) = ρ− .
c2
This can be written in terms of the Lorentz factor of special relativity,
1
ρ− = −γ 2 ρ+ , γ=p .
1 − v 2 /c2
Since the wire is overall neutral, ρ− r02 + ρ+ r2 = 0, so

r
r0 = .
γ
For nonrelativistic motion, the contraction is extremely small. (However, in plasmas, where
the positive charges are also free to move, this so-called pinch effect can be very significant.)
13
[2] Problem 24 (Griffiths 5.41). A current I flows to the right through a rectangular bar of conducting
material, in the presence of a uniform magnetic field B pointing out of the page, as shown.
(a) If the moving charges are positive, in what direction are they deflected by the magnetic field?
This deflection results in an accumulation of charge on the upper and lower surfaces of the
bar, which in turn produces an electric force to counteract the magnetic one. Equilibrium
occurs when the two exactly cancel. (This phenomenon is known as the Hall effect.)
(b) Find the resulting potential difference, called the Hall voltage, between the top and bottom
of the bar, in terms of B, the speed v of the charges, and the dimensions of the bar.
(c) How would the answer change if the moving charges were negative?
When measurements were performed in the early 20th century, some metals were found to have
positive moving charges! This “anomalous Hall effect” was solved by the quantum theory of solids,
as you can learn in any solid state physics textbook. Today, extensions of the Hall effect, such as
the integer and fractional quantum Hall effects, remain active areas of research, and could be used
01m
to build quantum computers. We’ll return to these effects in X3.
[3] Problem 26. 01m USAPhO 1997, problem B1. A nice problem on the dynamics of a plasma.
[3] Problem 27. 01^ USAPhO 2019, problem A3. This is a tough but useful problem. The first half
derives the so-called Child–Langmuir law, covered in problem 2.53 of Griffiths.
14
Preliminary Problems
These basic problems should be approachable if you understand the core material in Halliday,
Resnick, and Krane. If you can solve at least 75% of these completely and correctly, you’re ready to
start the main problem sets. Work carefully: many of the problems are more subtle than they look.
Answers are provided for most questions, so you can check your work; solutions are deliberately not
provided, so that you have the chance to work them out for yourself.
1 Mechanics
[2] Problem 1. As a warmup, take the Force Concept Inventory test. This should take no longer
than 30 minutes. You do not have to justify your answers; just list them.
Answer. 31352 22251 42441 12254 52213 53523
[1] Problem 2. A sewage worker is using a ladder inside a large, frictionless, horizontal circular
aqueduct. The ladder is of the same length as the diameter of the aqueduct.
(a) First the ladder is placed perfectly vertically and the worker climbs to the midpoint. Draw a
free body diagram indicating all forces on the ladder, and their names. Do the forces balance?
(b) Now suppose the ladder is placed perfectly horizontally and the worker hangs statically from
the midpoint. Draw a free body diagram indicating all forces on the ladder, and their names.
Do the forces balance? If so, show this explicitly. If not, what happens next?
[1] Problem 3. Consider the two following setups involving pulleys and spring scales. Treat the ropes,
spring scales, and pulleys as massless and the pulleys as frictionless. Model the pulleys as uniform
discs with masses of 5 kg which are fixed by a rigid support.
(a) Draw a free-body diagram for the second setup, showing all external forces on the spring scale.
(b) What are the readings on the two spring scales?
(c) Draw a free-body diagram for the first setup, showing and naming all external forces on the
pulley. In particular, what is the magnitude of the force between the pulley and the rope?
(d) What is the magnitude of the force that must be provided by the support?
√
Answer. (c) normal force from the rope, weight, and the force from the support, (d) 50 5 N
1
[1] Problem 4. A projectile is thrown upward and passes a point A and a point B a height h above.
Let TA be the time interval between the two times the projectile passes point A, and define TB
similarly.
(a) Show that g can be measured as

8h
g= .
TA2 − TB2
(b) This procedure probably looks a little contrived. Why is it better than doing something
simpler, such as just dropping the ball and using ∆y = gt2 /2?
[1] Problem 5. Consider the following system of massless pulleys and string, called a “fool’s tackle”.
If the load L has mass m, find the force F needed to keep the system static.
Answer. Static equilibrium is impossible.
[2] Problem 6. Consider a projectile launched with speed v at an angle θ from the horizontal on a
flat plane.
(a) Find y(x) and the ratio of the range to the maximum height.
(b) What is the maximum θ for which the projectile is always moving away from the thrower?
√
Answer. (b) sin−1 (2 2/3)
[2] Problem 7. A wooden isosceles right triangle with uniform mass density is placed on a table, and
a force is applied as shown.
The force is gradually increased until the triangle begins to tip over without sliding. The force is
then removed. Next, the surface is inclined with angle θ. For what range of θ can you be certain
the triangle will not slide down the incline?
Answer. θ < tan−1 (1/3)
2
[1] Problem 8. A painter of mass M stands on a platform of mass m as shown.
He pulls each rope down with force F , and accelerates upward with acceleration a. Find a.
Answer. −g + 4T /(M + m)
[3] Problem 9. A small block lies at the bottom of a spherical bowl of radius R.
(a) Find the period of small oscillations, assuming no friction. Can you give an intuitive explana-
tion of the simplicity of your answer?
(b) What is the period if the block is replaced with a small uniform ball, with sufficient friction
to roll without slipping?
(c) Now suppose the block moves in a circle, staying a constant height h R above the bottom
of the bowl. Find the period of the motion, assuming no friction.
(d) Again, find the period if the block is replaced with a small uniform ball, with sufficient friction
to roll without slipping.
(e) Now consider part (c) again. Suppose that at some moment, the speed of the block is
instantaneously increased by a small amount. Qualitatively describe the subsequent motion,
e.g. sketch what a top-down view would look like. What if h is not small?
p p
Answer. (a, c) 2π R/g, (b, d) 2π 7R/5g
[2] Problem 10. A car accelerates uniformly from rest. Initially, its door is slightly ajar. Calculate
how far the car travels before the door slams shut. Assume the door has a frictionless hinge, a
uniform mass distribution, and a length L from front to back.
Answer. π 2 L/12
[2] Problem 11. Two diametrically opposite points on a ring of mass M and radius R are marked
out. The ring is placed at rest on a frictionless floor. An ant of mass m starts at one point, then
walks horizontally along the ring to the other. Through what total angle does the ring turn?
Answer. πm/(M + m)
[2] Problem 12. A baseball player holds a bat, modeled as a uniform rigid rod, horizontally at one
of its ends. Usually, when the baseball hits the bat, the player will feel a sharp jolt in their hands
as the bat recoils. This can be avoided if the baseball hits the “sweet spot”. Where is it?
Answer. If the rod has length L, the center of percussion/sweet spot is 2L/3 from the held end.
3
[2] Problem 13. Maxwell’s wheel is a toy which demonstrates the principle of conservation of energy.
It consists of a uniform disc of mass M and radius R, with a massless axle of radius r. If the wheel
is released from rest, it falls downward, in such a way so that the strings supporting it are always
perfectly vertical. Find the acceleration.
Answer. g/(1 + R2 /2r2 )
[3] Problem 14. A cue ball is a uniform sphere of radius R.
(a) Find the height at which the cue ball must be hit horizontally so that it immediately begins
rolling without slipping.
(b) Skillful players can hit the cue ball so that it begins moving forwards, but then ends up moving
backwards. Model the hit as an instantaneous impulse applied at an arbitrary point on the
back half of the cue ball, in an arbitrary direction. For what impulses will this trick work?
You can treat the situation as two-dimensional; justify your answer carefully.
Answer. (a) 2R/5, (b) the line of the impulse passes under the contact point with the ground
[4] Problem 15. Six identical uniform rods, fastened at their ends by frictionless pivots, form a regular
hexagon and lie on a frictionless surface.
A blow is given at a right angle to the midpoint of the bottom rod. Immediately afterward, the
bottom rod has velocity u, as shown. Find the speed of the opposite rod at this moment.
Answer. u/10
[2] Problem 16. Several possible elliptical orbits of a satellite are shown below.
4
(a) Which orbit has the most angular momentum?
(b) Which orbit has the highest total energy?
(c) On which orbit is the largest speed acquired?
In all cases, justify your answer carefully.
Answer. (a) A, (b) A, (c) B
[2] Problem 17. A comet passes by Sun as shown, in a parabolic path.
How long, in years, does the comet take to get from point A to point B? (Hint: if you apply Kepler’s
laws and properties of conics, this problem can be done with almost no computation.)
√
Answer. 2 2/3π years
[2] Problem 18. Because of the rotation of the Earth, the line of a plumb bob will not align with
the local gravitational field. Find the (small) angle of deviation between them as a function of the
latitude θ, the gravitational acceleration g, the radius R of the Earth, and its angular velocity ω.
[2] Problem 19. You should be comfortable with setting up multiple integrals. Consider a cylindrical
shell whose axis of symmetry is the z-axis. It has non-uniform mass per unit area σ(φ, z) in
cylindrical coordinates, and the shell has radius a and height h, with the bottom edge at z = 0.
(a) Write down an integral that gives the total mass of the shell.
(b) Write down an integral that gives the moment of inertia of the shell about the z-axis.
(c) How would these results change if we had a solid cylinder with mass per unit volume ρ(φ, z, r)?
5
[2] Problem 20. An entrepreneur proposes to propel the Earth through space by attaching many
balloons to one side of it with ropes. The balloons will experience a buoyant force, which will
create a tension in the ropes. Now consider the forces on the solid Earth. Because the atmospheric
pressure on the surface is uniform, the only net force on the Earth is from the tension, so the Earth
will get propelled.
Is this correct? If you think it is, explain why momentum conservation isn’t violated. If you
think it isn’t, identify the specific force acting on the solid Earth that cancels the tension force.
Answer. The force is balanced by the gravitational force of the atmosphere on the Earth.
2 Problem Solving Skills

You should be able to start these questions without any background reading. However, for some
fun background on estimation, see Guesstimation: Solving the World’s Problems on the Back of
a Napkin. For practical tips for real experiments, see chapter 7 of Physics Olympiad: Basic to
Advanced Exercises.
[2] Problem 21. Argon atoms are special because they stay in the atmosphere for a very long time.
They are not recycled like oxygen and nitrogen. An average breath inhales around 0.5 L of air and
people breath on average around once every five seconds. Air is about 1% argon and has density
1.2 kg/m3 . Assume all air particles have a mass of approximately 5 × 10−26 kg. Take the atmosphere
to have constant density and be around 20 km thick. The radius of the Earth is 6.4 × 106 m.
(a) Estimate the total number of distinct argon atoms inhaled by Galileo throughout their life.
(b) Assuming the atmosphere has been uniform mixed since then, estimate the number of argon
atoms in each of your breaths that were once in Galileo’s lungs.
Answer. (a) about 1029 , (b) about 106
[3] Problem 22. The acceleration due to gravity can be measured by measuring the time period of a
simple pendulum. However, it can be challenging to get an accurate result.
(a) Suppose you constructed a pendulum using regular household materials. Name at least five
sources of possible experimental error in your calculated value of g. How would you make the
pendulum and perform the measurements to minimize these sources of error?
(b) Make an actual pendulum yourself and carry out the measurement. Describe your experimental
procedure and show your data. Estimate as many of the sources of experimental error identified
in part (a) as you can, and using them, give a value of g with a reasonable uncertainty.
(c) If you had $1,000 and a week to do plenty of measurements, how would you go about it?
How precise a result do you think you could get? What would be the dominant sources of
uncertainty remaining?
[2] Problem 23. Blackbody radiation is an electromagnetic phenomenon, so the radiation intensity
depends on the speed of light c. It is also a thermal phenomenon, so it depends on the thermal
energy kB T , where T is the object’s temperature and kB is Boltzmann’s constant. And it is a
quantum phenomenon, so it depends on Planck’s constant h.
(a) Using the relation E = hf , find the dimensions of h.
6
(b) Using dimensional analysis, show that the power emitted by the blackbody per unit area,
called the radiation intensity I, obeys I ∝ T 4 , and find the constant of proportionality up to
a dimensionless constant.
(c) How would the result change in a world with d spatial dimensions?
(d) The result you derived in part (b) is known as the Stefan–Boltzmann law. But if it can be
derived with pure dimensional analysis, without needing any detailed calculations or experi-
mental data at all, then why is it considered a law at all? Isn’t it obvious?
Answer. (c) I ∝ T d+1
3 Electromagnetism
[2] Problem 24. As a warmup, take the Conceptual Survey of Electricity and Magnetism test. This
should take no longer than 1 hour. You do not have to justify your answers; just list them.
Answer. BABBC EBBBC EDEDA EEDAD EDACD AECCA ED
[2] Problem 25. A half-infinite line has linear charge density λ.
(a) Find the electric field at a point that is “even” with the end, a distance ` from it, as shown.
(b) You should find the direction of the field is independent of `. Explain why.
(c) Sketch the electric field lines everywhere.
Answer. (b) scaling symmetry
[2] Problem 26. Some basic tasks involving intuition for vector fields.
(a) Consider the vector field

v = 2x̂ + xŷ.
Sketch some field vectors at regular points. Then, on a separate sketch, draw some field lines.
(b) Some electric field vectors in a certain situation are shown below.
7
Sketch a corresponding field line diagram. Then, give a mathematical expression that could
describe this field, and a physical situation which could produce it.
(c) The electric field lines in another situation are shown below.
Sketch a corresponding set of field vectors at regular points. Then, give a mathematical
expression that could describe this field, and a physical situation which could produce it.
[3] Problem 27. A parallel plate capacitor of capacitance C is placed in a region of zero electric field.
The first plate is given total charge Q1 and the second plate is given total charge Q2 . Let the plates
have area A and separation d, where A d2 .
(a) Each of the plates has an inner and outer surface. Using Gauss’s law, find the total charge
on each of these four surfaces.
(b) Find the potential difference between the plates.
(c) Find the force between the plates.
(d) In addition to the force between the plates found in part (c), there is a contribution to the
internal stress (force per unit area) within each plate due to the charges on its two surfaces.
Find this part of the stress for each plate, and assuming Q1 > Q2 > 0, indicate whether it is
tension or compression.
8
Answer. (c) Q1 Q2 /2A0 , (d) (Q21 − Q22 )/8A2 0 for both, plate 1 under tension, plate 2 under
compression
[3] Problem 28. A battery is connected to an RC circuit as shown.
The switch is initially open, and the charge on the capacitor is initially zero. The switch is closed
at t = 0.
(a) Solve for the charge on the capacitor as a function of time.
(b) Solve for the power dissipated in the resistor as a function of time.
(c) What is the total energy dissipated in the resistor over all time? Can you find a simple way
to derive this result?
(d) Suppose that the emf E(t) supplied by the battery can be adjusted freely over time, and the
capacitor must be given a total charge Q by the time t = T . Sketch the profile E(t) that
maximizes the efficiency of this process, i.e. the ratio of the energy stored in the capacitor to
the energy output by the battery, and find this efficiency.
Answer. (d) E(t) is linear, efficiency 1/(1 + (2RC/T )).
[2] Problem 29. In this problem we estimate the maximum firing speed of a human neuron. Model
a human cell simply as a sphere of radius 10−6 m.
(a) It has been measured that 1 cm2 of cell membrane has a resistance of 1000 Ω. Estimate the
resistance of a single human cell.
(b) Estimate the capacitance of a single human cell, treating the two sides of the membrane as
capacitor plates. You will have to estimate the thickness of the cell membrane.
(c) By modeling the cell as an RC circuit, estimate the maximum firing speed of a human neuron.
Is this a reasonable result? If yes, how do you know? If not, how could this model be refined?
Answer. (c) The RC timescale is on the order of 10 ms, which is reasonable.
[2] Problem 30. An infinite solenoid with radius b has n turns per unit length. The current varies
in time according to I(t) = I0 cos ωt. A ring with radius r < b and resistance R is centered on the
solenoid’s axis, with its plane perpendicular to the axis.
(a) What is the induced current in the ring?
(b) A given little piece of the ring will feel a magnetic force. For what values of t is this force
maximum? At this moment, sketch the electric field everywhere.
9
(c) What is the effect of the force on the ring? That is, does the force cause the ring to translate,
spin, etc.?
(d) If the current is driven by the AC power from a wall outlet, which has frequency 60 Hz in
America, the ring will emit a humming sound. What is the frequency of this sound?
Answer. (c) grows and shrinks, (d) 120 Hz
[2] Problem 31. A long, insulating cylinder with radius r and uniform surface charge density σ on
its outer surface rotates about its symmetry axis with angular velocity ω.
(a) Find the magnetic field everywhere.
(b) A wire is connected to a point on the cylinder, with the other end on the axis of rotation.
The wire rotates along with the cylinder. Show that the emf across the wire does not depend
on the shape of the wire’s path, and find this emf.
Answer. (b) µ0 σω 2 r3 /2
[2] Problem 32. A rectangular loop of wire with dimensions a and b is placed with one side parallel
to a long, straight wire carrying current I0 , at a distance l.
The resistance of the loop is R. The current in the long wire is quickly switched off.
(a) What is the net momentum p acquired by the loop?
(b) How is momentum conserved in this setup?
Answer.
1 (bµ0 I0 )2

a+l 1 1
(a) log −
8π 2 R l l a+l
[1] Problem 33. A 120 V rms, 60 Hz line provides power to a 40 W light bulb. By what factor will
the brightness change if a 10 µF capacitor is connected in series with the light bulb?
Answer. 65%
4 Thermodynamics
[2] Problem 34. Two moles of a monatomic ideal gas are taken through the following cycle.
10
• The gas begins at point A with pressure P0 and volume V0 .
• The gas is heated at constant volume until it doubles its pressure, reaching point B.
• The gas is expanded at constant pressure until it doubles its volume, reaching point C.
• The gas is cooled at constant volume until it halves its pressure, reaching point D.
• The gas is compressed at constant pressure until it halves its volume, returning to point A.
Assume that all processes are quasistatic and reversible.
(a) Draw the process on a P V diagram.
(b) Calculate the net work done by the gas during the cycle.
(c) Calculate the efficiency of the cycle.
(d) Calculate the change in entropy of the gas as the system goes from state A to state D.
Answer. (b) P0 V0 , (c) 2/13, (d) 5R log 2
[3] Problem 35. Deriving some basic results in thermodynamics.
(a) Starting from the first law of thermodynamics, derive the fact that P V γ is constant in an
adiabatic process.
(b) Using the ideal gas law, derive the total work done by a gas as it expands at constant
temperature from volume V1 to V2 , in terms of n, R, T , V1 , and V2 .
(c) Show that if a general gas, not necessarily ideal, satisfies the equation P V = kU , where U is
the total internal energy, then P V n is constant in an adiabatic process for some power n, and
find n in terms of k.
(d) Does a monatomic ideal gas satisfy P V = kU ? If so, what is k?
(e) Two Carnot engines operate with the same minimum and maximum pressures, temperatures,
and volumes. One uses helium as its working substance, and the other uses air. (At the
relevant temperatures, helium behaves like a monatomic gas.) Which one performs more work
per cycle, and by what factor?
Answer. (c) n = k + 1, (d) k = γ − 1, (e) air, by a factor of 5/3
[1] Problem 36. A monatomic ideal gas is adiabatically compressed to 1/8 of its original volume. For
each of the following quantities, indicate by what factor they change.
(a) The rms velocity vrms .
(b) The mean free path λ.
(c) The average time between collisions τ for each gas molecule.
(d) The molar heat capacity Cv .
Answer. (a) 2 times larger, (b) 8 times smaller, (c) 16 times smaller, (d) same
11
[2] Problem 37. A simple heat engine consists of a movable piston in a cylinder filled with an ideal
monatomic gas. Initially the gas in the cylinder is at a pressure P0 and volume V0 . The gas is slowly
heated at constant volume until the pressure is 32P0 . The gas is then adiabatically expanded until
its pressure is P0 again. Finally, the gas is cooled at constant pressure until its volume is V0 again.
Find the efficiency of the cycle.
Answer. 58/93
[2] Problem 38. The total mass of a hot-air balloon (envelope, basket, and load) is 320 kg. Initially
the air pressure inside and outside the envelope is 1.01 × 105 Pa and its density is 1.29 kg/m3 . In
order to raise the hot-air balloon, a gas burner is used to heat the air inside the balloon. The
volume of the envelope filled with hot air is 650 m3 . The molar mass of air is 29 g/mol. Treat the
temperature of the air in the balloon as uniform.
(a) The balloon can either be tightly sealed, so that none of its air mixed with the outside air,
or have a hole, so that its pressure equalizes with that of the outside air. For the purpose of
generating lift, which is better?
(b) Assuming the better option has been taken, to what temperature must the air inside the
balloon be heated to make the balloon begin to rise?
Answer. (a) hole, (b) 440 K
[3] Problem 39. Water is heated in an electric kettle. At a certain moment of time, a piece of ice at
temperature T0 = 0 ◦ C was put in the kettle. The figure below shows the water temperature as a
function of time.
Find the mass of the ice if the heating power of the kettle is P = 1 kW. The latent heat of melting
for ice is L = 335 kJ/kg, the heat capacity of water is c = 4.2 kJ/kg K, and the temperature of the
room is T1 = 20 ◦ C. (Hint: it’s very easy to get an answer that’s off by up to 50% if you’re careless.)
Answer. 28 g
5 Relativity and Waves

These problems can be solved using the material in chapters 18 through 20, and 39 through 44 of
Halliday and Resnick.
[2] Problem 40. Two bombs lie on a train platform, a distance L apart. As a train passes by at
speed v, the bombs explode simultaneously (in the platform frame) and leave marks on the train.
12
Due to the length contraction of the train, we know that the marks on the train are a distance γL
apart when viewed in the train’s frame, because this distance is what is length-contracted down
to the given distance L in the platform frame. How would someone on the train quantitatively
explain why the marks are a distance γL apart, considering that the bombs are a distance of only
L/γ apart in the train frame?
[1] Problem 41. An atom at rest with rest mass m radiates a photon with frequency ω. What is the
rest mass of the atom afterward?
p
Answer. m(m − 2~ω/c2 )
[3] Problem 42. A rope of linear mass density σ is hung between two poles, a distance L apart, and
the middle of the wire sags a distance d L below the ends. Find the approximate frequencies
of standing waves on the rope. (Hint: you do not have to solve for the shape of the rope. As an
intermediate step, it will be useful to consider torque balance on half of the rope.)
Answer. r
n g
fn = √
4 2 d
[3] Problem 43. A uniform string of length L and linear density ρ is stretched between two fixed
supports. The tension in the string is T .
(a) Find the standing wave solutions and angular frequencies for the given boundary conditions.
(b) A very small mass m is now placed a distance ` from one end of the string. Find the change
in the angular frequencies to first order in m, by using the fact that the average potential
energy and average kinetic energy for each standing wave should remain equal.
(c) How would you go about finding the exact standing wave frequencies for this setup?
p
Answer. (a) ωn = nπv/L where v = T /ρ, (b) ∆ωn ≈ (−m/ρL) ωn sin2 (nπ`/L)
[2] Problem 44. A perfectly flat piece of glass is placed over a perfectly flat piece of black plastic.
They touch at point A. Light of wavelength 600 nm is incident normally from above. The location
of the dark fringes in the reflected light is shown above.
(a) How thick is the space between the glass and plastic at B?
(b) Water with n = 1.33 seeps into the region between the glass and the plastic. How many dark
fringes are seen when all of the air has been displaced by water? The straightness and equal
spacing of the fringes is an accurate test of the flatness of the glass.
13
(c) A setup like this one was used by the physicist Otto Wiener to measure the wavelength of
light. When setting up the experiment, one must decide on the angle between the two objects.
What are the advantages and disadvantages of making this angle smaller, for the purposes of
measuring the wavelength?
Answer. (b) 10 dark fringes
[2] Problem 45. A point source of light L emitting a single wavelength λ is situated a distance d
from an ideal mirror, at z = 0. A screen stands at the end of the mirror at distance D d from L.
Find the relative intensity of light on the screen as a function of z. (This setup is known as Lloyd’s
mirror.)
Answer. I(z) ∝ sin2 (2πdz/Dλ)
14
Problem Solving I: Mathematical Techniques

For the basics of dimensional analysis and limiting cases, see chapter 1 of Morin or chapter 2 of Order
of Magnitude Physics. Many more examples are featured in The Art of Insight; some particularly
relevant sections are 2.1, 5.5, 6.3, 8.2, and 8.3. Other sections will be mentioned throughout the
course. There is a total of 81 points.
1 Dimensional Analysis
Idea 1
The dimensions of all physical equations should match on both sides. Sometimes, this
constraint alone can determine the answers to physical questions.
Even a statement such as “the speed v is slow” does not make sense, since we need to compare
it to something else with dimensions of speed. A more meaningful statement would be “v is slow
compared to the speed of light”, which would correspond to v/c 1.
Example 1: F = ma 2018 B11
A circle of rope is spinning in outer space with an angular velocity ω0 . Transverse waves on
the rope have speed v0 , as measured in a rotating reference frame where the rope is at rest.
If the angular velocity of the rope is doubled, what is the new speed of transverse waves?
Solution
To solve this problem by dimensional analysis, we reason about what could possibly affect the
speed of transverse waves. The result could definitely depend on the rope’s length L, mass
per length λ, and angular velocity ω0 . It could also depend on the tension, but since this
tension balances the centrifugal force, it is determined by all of the other quantities. Thus
the quantities we have are
[L] = m, [λ] = kg/m, [ω0 ] = 1/s.
Since λ is the only thing with dimensions of mass, it can’t affect the speed, because there is
nothing that could cancel out the mass dimension. So the only possible answer is
v0 ∼ Lω0
where the ∼ indicates equality up to a dimensionless constant, which cannot be found by

dimensional analysis alone. In practice, the constant usually won’t be too big or too small,
so Lω0 is a decent estimate of v0 . But even if it isn’t, the dimensional analysis tells us the
scaling: if ω0 is doubled, the new speed is 2v0 .
Example 2
Find the dimensions of the magnetic field.
1
Solution
To do this, we just think of some simple equation involving B, then solve for its dimensions.
For example, we know that F = q(v × B), so
[F ] kg · m 1 1 kg
[B] = = 2
= .
[q][v] s C m/s C·s
[2] Problem 1. Find the dimensions of power, the gravitational constant G, the permittivity of free
space 0 , and the ideal gas constant R.
Solution. The units are

kg m2 m3 C2 s2 J kg m2
[P ] = , [G] = , [0 ] = , [R] = = .
s3 kg s2 kg m3 mol K mol K s2
The easiest way to get these results is to use formulas containing the desired quantity, such as
P = F v, F = GM m/r2 , F = q 2 /(4π0 r2 ), and P V = nRT , where the units of the other quantities
are already known.
[1] Problem 2. Derive Kepler’s third law for circular orbits, using only dimensional analysis. (Why
do you think people didn’t figure out this argument 2000 years ago?)
Solution. The answer should only depend on G, M , and the radius r. By dimensional analysis,
we have the equality of units
[r] = [(GM )1/3 T 2/3 ]
which implies we must have T 2 ∝ r3 . Of course, this doesn’t mean Kepler’s third law is trivial.
The dimensions of G follow from the inverse square law for gravity, and you need to know which
quantities are allowed in the dimensional analysis in the first place. In other words, you need the
whole structure of Newtonian mechanics to be set up already to run this argument.
[2] Problem 3. Some questions about vibrations.
(a) The typical frequency f of a vibrating star depends only on its radius R, density ρ, and
the gravitational constant G. Use dimensional analysis to find an expression for f , up to a
dimensionless constant. Then estimate f for the Sun, looking up any numbers you need.
(b) The typical frequency f of a small water droplet freely vibrating in zero gravity could depends
on its radius R, density ρ, surface tension γ, and the gravitational constant G. Argue that at
least one of these parameters doesn’t matter, and find an expression for f up to a dimensionless
constant.
Solution. (a) We just do the usual dimensional analysis,
m3
[f ] = s−1 [R] = m [ρ] = kg/m3 [G] =
kg · s2
To cancel out kg, multiplying G and ρ will yield [ρG] = s−2 . Then to get [f ] = s−1 ,
f ∼ Gρ ∼ 3 × 10−3 Hz
p
2
which is in the right range. These oscillations are measured in the field of helioseismology.
Another application of this result is that the time needed for a ball of gas of density ρ to
√
collapse is of order 1/ Gρ, called the free fall time. This timescale plays an important role
in structure formation in the early universe.
(b) In this case the surface tension force dominates; the gravitational forces of the droplet on itself
are negligible, so we can drop G. Performing dimensional analysis with R, ρ, and γ gives
r
γ
f∼ .
ρR3
Of course, part (a) is equivalent to starting with the same set of four parameters and dropping
γ, which makes sense since the objects considered are huge.
[3] Problem 4. Some questions about the speed of waves. For all estimates, you can look up any
numbers you need.
(a) The speed of sound in an ideal gas depends on its pressure p and density ρ. Explain why we
don’t have to use the temperature T or ideal gas constant R in the dimensional analysis, and
then estimate the speed of sound in air.
(b) The speed of sound in a fluid depends only on its density ρ and bulk modulus B = −V dP/dV .
Estimate the speed of sound in water, which has B = 2.1 GPa.
The speed of waves on top of the surface of water can depend on the water depth h, the wavelength
λ, the density ρ, the surface tension γ, and the gravitational acceleration g.
(c) Find the speed of capillary waves, i.e. water waves of very short wavelength, up to a dimen-
sionless constant.
(d) Find the speed of long-wavelength waves in very deep water, up to a dimensionless constant.
Solution. (a) We don’t have to use R or T because all that matters is the restoring force,
determined by p, and the inertia, determined by ρ. So we have
kg s2 kg
[p] = , [ρ] =
m m3
and a routine dimensional analysis gives
s
105 Pa
r
p
v∼ ∼ ∼ 300 m/s
ρ 1 kg/m3
p
which is reasonably close. (Actually, the exact answer is v = γp/ρ, as we’ll derive in T3
and W3, so thermodynamics actually does play a role through the dimensionless constant.)
(b) We have
kg kg
[B] = [ρ] = .
m · s2 m3
A routine dimensional analysis gives
s
B
v∼ ∼ 1500 m/s.
ρ
This is actually very close to the true answer; here there is no dimensionless constant.
3
(c) In this case, the surface tension force dominates, just as it did for a small water droplet in a
previous problem, which also means that g doesn’t matter. The wavelength is so short that
the waves can’t “see” the depth of the water, so h doesn’t matter. Doing dimensional analysis
with the remaining three parameters gives
r
γ
v∼ .
ρλ
(d) In this case, the wave is big enough for surface tension not to matter; the restoring force is
gravity, so we keep g and toss out γ. Since the water is even deeper than the wavelength, we
again toss out h. Doing dimensional analysis with the remaining parameters gives
p
v ∼ gλ.
We will derive this in W3. The fact that ρ also dropped out makes sense: when gravity is
the only force, ρ usually doesn’t matter because scaling it up scales all the forces and all the
masses up the same way, keeping accelerations the same.
[3] Problem 5 (Morin 1.5). A particle with mass m and initial speed v is subject to a velocity-
dependent damping force of the form bv n .
(a) For n = 0, 1, 2, . . ., determine how the stopping time and stopping distance depend on m, v,
and b.
(b) Check that these results actually make sense as m, v, and b are changed, for a few values of n.
You should find something puzzling going on. (Hint: to resolve the problem, it may be useful
to find the stopping time explicitly in a few examples.)
Solution. (a) The dimensions of b can be found with [b] = [F/v n ] = kg · m1−n · s−2+n . To get a
stopping time or distance, the mass term must be canceled out. So we’re working with

b m
= m1−n s−2+n [v] =
m s
The stopping time t can be found by canceling out the length dimension. If t ∝ (b/m)α v β ,
then:
α(1 − n) + β = 0 α(−2 + n) − β = 1
Solving yields
mv 1−n
α = −1 β = 1 − n, t∝ .
b
The distance x traveled has dimensions of vt, so
mv 2−n
x∝ .
b
(b) The results don’t seem to make sense. At n = 1, it appears that the time it takes to stop
no longer depends on v, which doesn’t seem correct since the stopping time should always
increase with velocity. And for n > 1, the stopping time decreases with velocity, which is even
worse. Similar issues happen for the stopping distance for n ≥ 2.
The resolution is that in these cases, the stopping time/distance are actually infinite, as you
can check explicitly. In other words, dimensional analysis worked, but the hidden dimensionless
prefactor was infinity.
4
Idea 2
Dimensional analysis applies everywhere. The argument of any function that is not a mono-
mial, such as sin x, must have no dimensions. The derivative d/dx has the opposite dimensions
to x, and the dx in an integral has the same dimensions as x.
Example 3
We are given the integral Z ∞ √

2
e−x dx = π.
−∞
Find the value of the integral Z ∞
2
e−ax dx.
−∞
Solution
In the first equation, x must be dimensionless, so both sides are dimensionless. The second
equation would also be consistent if both x and a were dimensionless, but we can do better.
Suppose we arbitrarily assign x dimensions of length, [x] = m. Then to make the argument
of the exponential dimensionless, a must have dimensions [a] = m−2 . The dimensions of the
left-hand side are [dx] = m. In order to make the dimensions work out on the right-hand
side, we must have Z ∞
2 1
e−ax dx ∝ √ .
−∞ a
To find the value of the constant, treat x and a as dimensionless again. Then we know the
√
answer must reduce to π when a = 1, so
Z ∞ r
−ax2 π
e dx = .
−∞ a
This procedure is completely equivalent to using u-substitution to nondimensionalize every-

thing, but may be faster to see. In general, for all integrals except for the simplest ones,
you should use either dimensional analysis or u-substitution to reduce the integral to a
dimensionless one.
Remark
Consider the value of the definite integral
Z y
2
e−x dx.
−∞
You can try all day to compute the value of this integral, using all the integration tricks
2
you know, but nothing will work. The function e−x simply doesn’t have an antiderivative
in terms of the functions you already know, i.e. in terms of polynomials, exponents and
logarithms, and trigonometric functions (for more discussion, see here).
5
If you ask a computer algebra system like Mathematica, it’ll spit out something like
2
erf(x), which is defined by being an antiderivative of e−x . But is this really an
2
“analytic” solution? Isn’t that just saying “the integral of e−x is equal to the integral
2
of e−x ”? Well, like many things in math, it depends on what the meaning of the word “is” is.
The fact is, the set of functions we regard as “elementary” is arbitrary; we just choose a set
that’s big enough to solve most of the problems we want, and small enough to attain fluency
with. (Back in the days before calculators, it just meant all the functions whose values were
tabulated in the references on hand.) If you’re uncomfortable with erf(x), note that a similar
thing would happen if a middle schooler asked you what the ratio of the opposite to adjacent
sides of a right triangle is. You’d say tan(x), but they could say it’s tautological, because the
only way to define tan(x) at the middle school level is the ratio of opposite to adjacent sides.
Similarly, 1/x has no elementary antiderivative – unless you count log(x) as elementary, but
log(x) is simply defined to be such an antiderivative. It’s all tautology, but it’s still useful.
[1] Problem 6. Evaluate the integral Z

dx
x2+1
using a trigonometric substitution. Using dimensional analysis, find the integral
Z
dx
.
x + a2
2
Solution. We have
sec2 θdθ
Z Z
dx
= = arctan(x) + C.
2
x +1 1 + tan2 θ
Since x2 is added to a2 , they must have the same dimensions, so the integral has dimensions of 1/a.
The argument of arctan must be dimensionless, so we conclude
Z
dx 1 x
= arctan .
x2 + a2 a a
[2] Problem 7. In particle physics, it is conventional to set ~ = c = 1, resulting in equations with

seemingly incorrect units. For example, it is said that the mass of the Higgs boson is about 125 GeV,
where 1 eV is the energy gained by an electron accelerated through a voltage difference of 1 V.
Energy doesn’t have the same units as mass, but the units can be fixed by adding appropriate
factors of ~ and c. Do this, and find the mass of the Higgs boson in kilograms.
Solution. One easy way to start out dimensional analysis is with famous equations: E = mc2 , or
E = 12 mv 2 to get m ∼ E/c2 . Thus the mass of the Higgs boson is m = 125 GeV/c2 = 2.22×10−25 kg.
6
Example 4
The Schrodinger equation for an electron in the electric field of a proton is
~2 2 e2
− ∇ ψ− ψ = Eψ.
2m 4π0 r
Estimate the size of the hydrogen atom.
Solution
This is yet another dimensional analysis problem: there is only one way to form a length
using the quantities given above. We have
[m] = kg, [~] = J · s = kg m2 s−1 , [e2 /4π0 ] = J · m = kg m3 s−2 .
Doing dimensional analysis, the only length scale is the Bohr radius,
4π~2 0
a0 = ∼ 10−10 m.
me2
I’ve thrown in a 4π above because 0 always appears in the equations as 4π0 . The
dimensional analysis would be valid without this factor, but as you’ll see in problem 11, if
you don’t include it then annoying compensating factors of 4π will appear elsewhere.
Classically (i.e. without ~), there is no way to form a length, and hence there should be
no classically stable radius for the atom. (This was one of the arguments used by Bohr to
motivate quantum mechanics; it appears in the beginning of his paper introducing the Bohr
model.) Once we introduce ~, there are three dimensionful parameters in the problem, as
listed above. And there are exactly three fundamental dimensions. So there is only one way
to create a length, which we found above, one way to create a time, one way to create an
energy, and so on. This means that the solutions to the Schrodinger equation above look
qualitatively the same no matter what these parameters are; all that changes are the overall
length, time, and energy scales. In problem 11, you’ll investigate how this conclusion changes
when we add more dimensionful parameters.
Dimensional analysis is especially helpful with scaling relations. For example, a question might ask
you how the radius of the hydrogen atom would change in a world where the electron mass was
twice as large. You would solve this problem in the exact same way as the example above, using
dimensional analysis to show that a0 ∝ 1/m.
[3] Problem 9. In this problem we’ll continue the dimensional analysis of the Schrodinger equation.
(a) Estimate the typical energy scale of quantum states of the hydrogen atom, as well as the
typical “velocity” of the electron, using dimensional analysis.
(b) Do the same for one-electron helium, the system consisting of a helium nucleus (containing
two protons) and one electron.
(c) Estimate the electric field needed to rip the electron off the hydrogen atom.
7
Solution. (a) Recall the electrostatic potential energy formula, E = kq 2 /r. We have a length
scale, a0 to replace r. For velocity, we use E ∼ mv 2 , giving
me4 e2
E∼ , v∼ .
(4π0 )2 ~2 4π0 ~
In fact, the binding energy of the hydrogen atom in its ground state is
me4
E= = 13.6 eV
2(4π0 )2 ~2
which is a constant known as the Rydberg. So the dimensional argument (keeping the factors
of 4π) gets the answer right to a factor of 2.
(b) Adding the second proton would double the charge inside the nucleus, so the expressions for
energy and velocity should stay the same except e2 would be replaced with 2e2 (not 4, since
the electron charge stays the same) and thus the energy would be 4e4 . In general, with Z as
the atomic number,
mZ 2 e4 Ze2
E∼ , v ∼ .
(4π0 )2 ~2 4π0 ~
(c) Physically, the work the electric field does by moving the electron across the radius of its orbit
should be enough to overcome its binding energy to the proton. This also tells us how to set
up the dimensional analysis; we have electric field
E m2e e5
|E| ∼ ∼ ∼ 1012 V/m.
ea0 (4π0 )3 ~4
This is a tremendously large electric field!

All of the results above are not that accurate, but they become much more accurate if we re-
place 0 with 4π0 . That in turn makes sense because these factors always appear together in
electromagnetism.
Idea 3: Buckingham Pi Theorem
Dimensional analysis can’t always pin down the form of the answer. If one has N quanti-
ties with D independent dimensions, then one can form N − D independent dimensionless
quantities. Dimensional analysis can’t say how the answer depends on them.
A familiar but somewhat trivial example is the pendulum: its period depends on L, g, and the
amplitude θ0 , three quantities which contain two dimensions (length and time). Hence we canpform
one dimensionless group, which is clearly just θ0 itself. The period of a pendulum is T = f (θ0 ) L/g.
A paper helicopter with rotor radius r and weight W is dropped from a height h in air with
a density of ρ. Assuming the helicopter quickly reaches terminal velocity, use dimensional
analysis to analyze the total flight time T .
8
Solution
The answer can only depend on the parameters r, W , h, and ρ. There are four quantities in
total, but three dimensions (mass, length, and time), so by the Buckingham Pi theorem we
can form one independent dimensionless quantity. In this case, it’s clearly r/h. Continuing
with routine dimensional analysis, we find
r
2 ρ
T = f (r/h) h .
W
The form of this expression is a bit arbitrary; for instance, we could also have written
f (r/h)r2 in front, or even f (r/h)r37 h−35 . These adjustments just correspond to pulling
factors of r/h out of f , not to changing the actual result.
This is as far as we can get with dimensional analysis alone, but we can go further using
physical reasoning. If the helicopter quickly reaches terminal velocity, then it travels at a
constant speed. So we must have T ∝ h, which means that f (x) ∝ x, and
r
ρ
T ∝ rh .
W
Example 6
An hourglass is constructed with sand of density ρ and an orifice of diameter d. When the
sand level above the orifice is h, what is the mass flow rate µ?
Solution
The answer can only depend on ρ, d, h, and g. The Buckingham Pi theorem gives
p
µ = f (h/d)ρ gd5 .
That’s as far as we can get with dimensional analysis; to go further we need to know more
√
about sand. If we were dealing with an ideal fluid, then the flow speed would
√ be v = 2gh by
√
Torricelli’s law, which means the flow rate has to be proportional to h. Then f (x) ∝ x,
√
giving the result µ ∝ ρd2 gh. This is a good estimate as long as the orifice isn’t so small
that viscosity starts to dominate.
But this isn’t how sand works. Sand is a granular material, whose motion is dominated by
the friction between sand grains. So the higher pressure doesn’t actually propagate to the
orifice, and the flow rate is independent of p
h, which is apparent from watching an hourglass
run. Then f (x) is a constant, giving µ ∝ ρ gd5 , which has been experimentally verified.
Remark
One has to be a little careful with the Buckingham Pi theorem. For example, if all we had
were 3 speeds vi , we can form two dimensionless quantities: v1 /v2 and v1 /v3 . (The quantity
9
v2 /v3 is not independent, since it is the quotient of these two.) But there are 3 quantities
with 2 dimensions (length and time), so we expect only 1 dimensionless quantity.
The problem is that the two dimensions really aren’t independent: for any quantity built
from the vi , a power of length always comes with an inverse power of time, so there’s only
one independent dimension. These considerations can be put on a more rigorous footing in
linear algebra, where the Buckingham Pi theorem is merely a special case of the rank-nullity
theorem. If you’re ever in doubt, you can just forget about the theorem and play with the
equations directly.
Remark
Dimensional analysis is an incredibly common tool in Olympiad physics because it lets you
say a lot even without much advanced knowledge. If a problem ever says to find some
quantity “up to a constant/dimensionless factor”, or how that quantity scales as another
quantity changes, or what that quantity is proportional to, it’s almost certainly asking you
to do dimensional analysis. Another giveaway is if the problem looks extremely technical and
advanced, because they can’t actually be.
[3] Problem 10 (Insight). In this problem we’ll do one of the most famous dimensional analyses of
all time: estimating the yield of the first atomic bomb blast. Such a blast will create a shockwave
of air, which reaches a radius R at time t after the blast. The air density is ρ, and we want to
estimate the blast energy E.
(a) Declassified photographs of the blast indicate that R ≈ 100 m at time t ≈ 15 ms. The density
of air is ρ ≈ 1 kg/m3 . Estimate the blast energy E.
(b) How much mass-energy (in grams) was used up in this blast?
(c) If we measure the entire function R(t), what general form would we expect it to have, if this
dimensional analysis argument is correct?
Solution. (a) The only way to write an expression with the right dimensions is
R5 ρ
E∼ .
t2
Plugging in the numbers gives E ∼ 4 × 1013 J.
(b) The mass-energy equivalent is m = E/c2 ∼ 0.5 g. This is quite reasonable, as fission can only
release a small fraction of the mass-energy (about 0.1%) of a sample, and the critical mass is
typically on the order of a few pounds.
(c) Let’s do the dimensional analysis in reverse: we know E is fixed, so the only way to write an
expression with the right dimensions for R is
R ∼ (Et2 /ρ)1/5 ∼ t2/5 .
So R(t) must have this power-law dependence. If it doesn’t, then it means some other quantity
with dimensions is intervening, so our dimensional analysis is suspect. Luckily, around this
range of time the relation above is true, and indeed the answer of part (a) is pretty close.
10
Remark
The British physicist G. I. Taylor performed the dimensional analysis in problem 10 upon
seeing a picture of the first atomic blast in a magazine. The result was so good that the
physicists at the Manhattan project thought their security had been breached!
During World War II, the exact value of the critical mass needed to set off a nuclear explosion
was important and nontrivial information. The Nazi effort to make a bomb had been stopped
by Werner Heisenberg’s huge overestimation of this quantity, and after the war, the specific
value was kept a closely guarded secret. That is, it was until 1947, when a Chinese physicist
got the answer using a rough estimate that took four lines of algebra.
[5] Problem 11. We now consider the Schrodinger equation for the hydrogen atom in greater depth.
We begin by switching to dimensionless variables, which is useful for the same reason that writing
integrals in terms of dimensionless variables is: it highlights what is independent of unit choices.
(a) Define a dimensionless length variable r̃ = r/a0 , where a0 is the length scale found in example 4.
The ∇2 term in the Schrodinger equation is a second derivative, the 3D generalization of
d2 /dx2 . Using the chain rule, argue that
˜ 2 = a20 ∇2
∇
˜ is the gradient with respect to r̃.

where ∇
(b) Similarly define a dimensionless energy Ẽ = E/E0 , using the energy scale E0 found in
problem 9. Show that the Schrodinger equation can be written in a form like
˜ 2 ψ − 1 ψ = Ẽψ
−∇
r̃
Here I’ve suppressed all dimensionless constants, like factors of 2, because they depend on
how you choose to define E0 and don’t really matter at this level of precision.
The result of this part confirms what we concluded above: solutions to the Schrodinger
equation don’t qualitatively depend on the values of the parameters, because they all come
from scaling a solution to this one dimensionless equation appropriately.
(c) This is no longer true in relativity, where the total energy is

p
E = p2 c2 + m2 c4 .
Assuming p mc, perform a Taylor expansion to show that the next term is Ap4 , and find
the coefficient A. (If you don’t know how to do this, work through the next section first.)
(d) In quantum mechanics, the momentum is represented by a gradient, p → −i~∇. (We will see
why in X1.) Show that the Schrodinger equation with the first relativistic correction is
~2 2 e2
− ∇ ψ− ψ + ~4 A∇4 ψ = Eψ.
2m 4π0 r
11
(e) Since there is now one more dimensionful quantity in the game, it is possible to combine the
quantities to form a dimensionless one. Create a dimensionless quantity α that is proportional
to e2 /4π, then numerically evaluate it. This is called the fine structure constant. It serves as
an objective measure of the strength of the electromagnetic force, because it is dimensionless,
and hence its value doesn’t depend on an arbitrary unit system.
(f) As the number of protons in the nucleus increases, the relativistic correction becomes more
important. Estimate the atomic number Z where the correction becomes very important.
Solution. (a) For the first derivative,
dψ dψ dx
= .
drx dx drx
With the length scale, dx/drx = a0 which is a constant. The second derivative does the same,
which gives two factors of a0 . This holds true for all the other dimensions, so
˜ 2 = a20 ∇2 .
∇
(b) Ignoring all numerical factors and dividing by E0 = e2 /0 a0 , we get
~2 0 a0 1 ˜ 2

a0
− 2 2 ∇ ψ − ψ = (E/E0 )ψ
me a0 r
which simplifies to
˜ 2 ψ − 1 ψ = Ẽψ.
−∇
r̃
√
(c) Since 1 + x ≈ 1 + x/2 + (1/2)(−1/4)x2 ,
r
p2 c2 p2 1 p4 c4
E = mc2 1 + 2 4 ≈ mc2 + −
m c 2m 8 m3 c6
which implies
1
A=− .
8m3 c2
(d) With p4 = ~4 ∇4 , this is simply added to the left hand side of the equation as a correction of
the first order momentum term p2 /2m = −~2 /2m,
~2 2 e2
− ∇ ψ− ψ + ~4 A∇4 ψ = Eψ.
2m 4π0 r
(e) Just like in part (b), divide both sides by E0 . The dimensionless quantity in the added term
should be
~4 0 a0 e4
= .
m3 c2 a40 e2 ~2 c2 20
To make it proportional to e2 , take the square root to get
e2 1
α∼ ≈ .
4π0 ~c 137
12
(f) The relativistic correction is important when the above term is of order 1, and since there’s
an electron charge e and a nucleus with charge +Ze, replace e2 with Ze2 . It’s order one when
Zα ≈ 1.
So the atomic number when the correction becomes very important is around 137. Actually,
even for moderately heavy elements, the corrections are already noticeable and must be
accounted for. As a concrete example, if you don’t account for relativistic effects, you would
predict the color of gold to be silver instead. For more about the relativistic chemistry of gold,
see this paper.
You probably won’t see any differential equations as complex as the ones in the above problem
anywhere in Olympiad physics, but the key idea of using dimensionless quantities to simplify and
01h
clarify the physics can be used everywhere.
[5] Problem 12. IPhO 2007, problem “blue”. This problem applies thermodynamics and dimen-
sional analysis in some exotic contexts.
Example 7
Estimate the Young’s modulus for a material with interatomic separation a and typical atomic
bond energy Eb . Use this to estimate the spring constant of a rod of area A and length L,
as well as the speed of sound, if each atom has mass m.
Solution
This example is to get you comfortable with the Young’s modulus Y , which occasionally
comes up. It is defined in terms of how much a material stretches as it is pulled apart,
Y = = .
The Young’s modulus is a useful way to characterize materials, because unlike the spring
constant, it doesn’t depend on the shape of the material. For example, putting two identical
springs side-by-side doubles the spring constant, because they both contribute to the force.
But since the stress is the force per area, it’s unchanged. Similarly, putting two identical
springs end-to-end halves the spring constants, because they both stretch, but since the
strain is change in length per length, it’s unchanged. So you would quote a material’s
Young’s modulus instead of its spring constant, for the same reason you would quote a
material’s resistivity instead of its resistance.
We note that Y has the dimensions of energy per length cubed, so

Eb
Y ∼
a3
solely by dimensional analysis. (Of course, for this dimensional analysis to work, one
has to understand why Eb and a are the only relevant quantities. It’s because Y , or
equivalently the spring constant k, determines the energy stored in a stretched spring.
But microscopically this comes from the energy stored in interatomic bonds when
13
they’re stretched. So the relevant energy scale is the bond energy Eb , and the relevant
distance scale is a, because that determines how many bonds get stretched, and by how much.)
To relate Y to the spring constant of a rod, note that

F/A L F L
Y = = =k
∆L/L A ∆L A
for a rod, giving the estimate k ∼ AEb /La3 . This is correct to within an order of magnitude!
To relate Y to the speed of sound, note that the sound speed, like most wave speeds, depends
on the material’s inertia and its restoring force against distortions. Since the speed of
sound doesn’t depend on the extrinsic features of a metal object, such as a length, both of
these should be measured intrinsically. The intrinsic measure of inertia is the mass density
ρ ∼ m/a3 , while the intrinsic measure of restoring force is just Y . By dimensional analysis,
s s r
Y Eb /a3 Eb
v∼ ∼ 3
∼ .
ρ m/a m
This is also reasonably accurate. For example, in diamond, Eb ∼ 1 eV (a typical atomic energy
scale), while a carbon nucleus contains 12 nucleons, so to the nearest order of magnitude,
m ∼ 10mp , where a useful fact is mp ∼ 1 GeV/c2 . Thus,
r
1 eV
v∼ c ∼ 10−5 c ∼ 3 km/s
1010 eV
which is roughly right. (The true answer is 12 km/s.)
Amazingly, we can get an even rougher estimate of v for any solid in terms of nothing besides
fundamental constants. To be very rough, the binding energy is on the order of that of
hydrogen. As you found in problem 9, this is, by dimensional analysis,
2 2
1 e2 e
Eb ∼ ∼ me .
4π0 a0 4π0 ~
We take the nuclear mass to be very roughly the proton mass mp , which gives
s 2
e2

v me me
r
∼ ∼α
c mp 4π0 ~c mp
where α is as found in problem 11. This expresses the speed of sound in terms of the
dimensionless strength of electromagnetism α, the electron to proton mass ratio, and the
speed of light. Of course, the approximations we have made here have been so rough that
now the answer is off by two orders of magnitude, but now we know how the answer would
change if the fundamental constants did.
Estimates as simple as these can be surprising to even seasoned physicists: in 2020, the
simple estimate above was rediscovered and published in one of the top journals in science.
If you want to learn how to do more of these estimates, this paper is a good starting point.
14
Remark
A warning: from these examples, you could get the idea that dimensional analysis gives
you nearly godlike powers, and the ability to write down the answer to most physics
problems instantly. In reality, it only works if you’re pretty sure your physical system
depends on only about 3 or 4 variables – and the hard part is often finding which
variables matter. For example, as we saw above, you can’t get Kepler’s third law for free
because that requires knowing the dimensions of G, which require knowing that gravity
is an inverse square law in the first place, a luxury Kepler didn’t have. And as another
example, we couldn’t have figured out E = mc2 long before Einstein, as who would
have thought that the speed of light had anything to do with the energy of a lump of
matter? Without the framework of relativity, it seems as irrelevant as the speed of sound or
the speed of water waves. To illustrate this point, we consider two contrasting examples below.
In reality, dimensional analysis is best for problems where it’s easy to see which variables
matter, problems where you’re explicitly told what variables matter, and for checking your
work, which you should get in the habit of doing all the time!
Example 8
Cutting-edge archeological research has found that the famed T. Rex was essentially a gigantic
chicken. Suppose a T. Rex is about N = 20 times larger in scale than a chicken. How much
larger is its weight, cross-sectional area of bone, walking speed, and maximum jump height?
Solution
These kinds of biological scaling arguments are fun to think about, though the reliability of
the results is somewhat questionable – if any given scaling law doesn’t quite match data, you
can always think a bit more, and come up with a new argument yielding a different scaling.
But here are a few simple example:
• Since the densities should match, the weight should scale with the volume, so as N 3 .
• Since the maximum compressive pressure that bone can take should be the same, the
bone area should scale with the weight, so also as N 3 . That is, the width of the bones
scales as N 3/2 , while their length L scales only as N . This is one of the reasons it’s
impractical to have huge land animals; the biggest animals now are all whales.
• As a very crude model of walking, we can think of the legs as swinging like a free
pendulum. The √length of one step is proportional to L,√while √
the period of the steps is
proportional to L. Thus, the walking speed scales as L ∝ N .
• The energy stored in the muscle cells scales with the volume, but the mass also scales
with the volume. Since the jump height satisfies E = mgh where E is the energy stored
by the muscles, h ∝ N 0 . So a dinosaur can’t jump much higher than a human – and
indeed, we can’t jump much higher than fleas can!
There’s an entire literature on these arguments. For instance, this delightful paper argues that
the frequency a furry mammal will shake to dry itself off scales with its mass as f ∝ m−3/16 .
15
Example 9
A person with density ρ and total energy E stored in their muscles can jump to a height h
in gravity g. How high would they be able to jump in gravity 10g?
Solution
By dimensional analysis, the only possible answer is
1/4
E
h∝
ρg
which means that in gravity 10g, they can jump to height h/101/4 .
But this is completely wrong! In gravity 10g, a person wouldn’t be able to jump at all;
they’d be so crushed by their own weight that they wouldn’t even be able to stand. The
actual answer depends on details of the biomechanics of muscles and bone, which involve
more dimensionful quantities than just the total energy E. So, as remarked above, you can’t
solve literally any problem by just listing a few relevant quantities and doing dimensional
analysis – you need to make sure those are the only relevant quantities.
2 Approximations
16
Idea 4: Taylor Series
For small x, a function f (x) may be approximated as
x2 00 xn (n)
f (x) = f (0) + xf 0 (0) + f (0) + . . . + f (0) + O(xn+1 )
2 n!
where O(xn+1 ) stands for an error term which grows at most as fast as xn+1 .
There are a few Taylor series that are essential to know. The most important are
x2 x3 x2 x3
exp(x) = 1 + x + + + O(x4 ), log(1 + x) = x − + − O(x4 )
2 6 2 3
and the small angle approximations
x3 x2
sin x = x − + O(x5 ), cos x = 1 − + O(x4 ).
6 2
Another Taylor series you learned long before calculus class is
1
= 1 + x + x2 + x3 + O(x4 ).
1−x
Usually you’ll only need the first one or two terms, but for practice we’ll do examples with
more. If any of these results aren’t familiar, you should rederive them!
Example 10
Find the Taylor series for tan x up to, and including the fourth order term.
Solution
By the fourth order term, we mean the term proportional to x4 . (Not the fourth nonzero
term, which would be O(x7 ).) Of course, tan x is an odd function, so the O(x4 ) term is
zero, which means we only need to expand up to O(x3 ). That means we can neglect O(x4 )
terms and higher everywhere in the computation, subject to some caveats we’ll point out later.
By definition, we have
sin x x − x3 /6 + O(x5 )
tan x = = .
cos x 1 − x2 /2 + O(x4 )
However, it’s a little tricky because we have a Taylor series in a denominator. There are
two ways to deal with this. We could multiply both sides by cos x, and expand tan x in
a Taylor series with unknown coefficients. Then we would get a system of equations that
will allow us to solve for the coefficients recursively, a technique known as “reversion of series”.
A faster method is to use the Taylor series for 1/(1 − x). We have
1
= 1 + u + O(u2 )
1−u
17
and substituting u = x2 /2 − O(x4 ) gives
1 x2
=1+ + O(x4 ).
cos x 2
Therefore, we conclude
tan x = (x − x3 /6 + O(x5 ))(1 + x2 /2 + O(x4 )) = x + x3 /3 + O(x5 ).
Here I was fairly careful with writing out all the error terms and intermediate steps, but as
you get better at this process, you’ll be able to do it faster. (Of course, one could also have
done this example by just directly computing the Taylor series of tan x from its derivatives.
This is possible, but for more complicated situations it’s generally not a good idea, because
computing high derivatives of a complex expression tends to get very messy. It’s better to
just Taylor expand the individual pieces and combine the results, as we did here.)
Remark
Finding series up to a given order can be subtle. For example, if you want to compute an
O(x4 ) term, it is not always enough to expand everything up to O(x4 ), because powers of x
might cancel. To illustrate this, the last step here is wrong:
x3 sin x x4 + O(x6 )
tan x = = 6= x + O(x5 ).
x3 cos x x3 + O(x5 )
[2] Problem 13. Find the Taylor series for 1/ cos x up to and including the fourth order (O(x4 )) term.
Solution. The derivatives of cos(x) at x = 0 are 0, −1, 0, 1, so
x2 x4
cos x = 1 − + + O(x6 ).
2 4!
To expand the inverse, note that
1
= 1 + u + u2 + O(u3 )
1−u
where in our case, u = x2 /2 − x4 /24. Plugging this in gives
1 5x4
= 1 + (x2 /2 − x4 /24) + (x2 /2 − x4 /24)2 + O(x6 ) = 1 + x2 /2 + .
cos x 24
[2] Problem 14. Extend the computation above to get the x5 term in the Taylor series for tan x.
Solution. From this point on we will start omitting the explicit O(xn ) error terms. We have
(x − x3 /6 + x5 /120)(1 + x2 /2 + 5x4 /24) = x + x3 /2 + 5x5 /24 − x3 /6 − x5 /12 + x5 /120
giving the answer,

x3 2x5
tan(x) = x + + .
3 15
18
[3] Problem 15. For small x, approximate the quantity
x2 e x
−1
(ex − 1)2
to lowest order. That is, find the first nonzero term in the Taylor series. If you don’t take enough
terms in the Taylor series to begin with, you’ll get an answer of zero, indicating you approximated
too loosely. But if you take too many, the computation will get extremely messy. (Hint: your
final answer should contain a factor of −1/12. It’s actually the same factor as in the classic result
1 + 2 + 3 + . . . = −1/12. The reason will be explained much later, in an example in X1.)
Solution. Again suppressing the error terms, we have
x2 (1 + x + x2 /2) 1 + x + x2 /2
− 1 = −1
(x + x2 /2 + x3 /6)2 1 + x + 7x2 /12
= (1 + x + x2 /2)(1 − x + 5x2 /12) − 1
x2
=− .
12
Note that we have been careful to keep the manipulations as simple as possible, e.g. by canceling
factors of x/x as early as possible. If you don’t do this, everything gets very messy and it’s unclear
what is contributing at what order, because of the subtlety pointed out in the above remark.
[2] Problem 16. The function cos−1 (1 − x) does not have a Taylor series about x = 0. However, it
does have a series expansion about x = 0 in a different variable. What is this variable, and what’s
the first term in the series? (Optionally, can you find higher order terms in the series?)
Solution. We have
d −1
arccos(1 − x) = p
dx 1 − (1 − x)2
which is unfortunately undefined at x = 0, so there is no Taylor series. But note that if we let
y = cos−1 (1 − x) and take the cosine of both sides, we have
cos y = 1 − x.
Now y does have a good Taylor series near y = 0, which corresponds to where x = 0. At lowest
order, we have
1 − y 2 /2 ≈ 1 − x
which implies that √
y≈ 2x.
√
More generally, the answer is a series in x. Since cosine is even, the next term is O(x3/2 ).
In order to get higher order terms, we can write
cos−1 (1 − x) = cos−1 (1 − u2 )
√
where u = x, and directly compute a Taylor series in u, using the usual rule for a derivative of an
inverse function. There is also an alternative route that uses the Taylor series for cosine directly.
Let’s suppose we just want the O(x3/2 ) term, so let
√
y = 2x1/2 + Ax3/2 + O(x5/2 ).
19
We also know that

y2 y4
1 − x = cos y = 1 − + + O(y 6 ).
2 24
Now, the lowest order term in the series found above is what matches to 1 − x. The next term
in the series can be found by demanding that the right-hand side contain x2 with zero coefficient.
Thus, we are only interested in expanding up to O(x2 ), and since y 6 = O(x3 ) we can drop it, so
1 √ 1/2 2 1 √ 1/2 4
1−x=1− 2x + Ax3/2 + 2x + Ax3/2 + O(x3 )
2 24
1 √ 2 1 √ 1/2 4
= 1 − (2x + 2 2Ax ) + 2x + O(x3 )
2 24
√ x2
= 1 − x − 2Ax2 + + O(x3 )
6
√
from which we conclude A = 1/6 2. This is an example of the technique of reversion of series.
Idea 5: Binomial Theorem

When the quantity xn is small, it is useful to use the binomial theorem,
(1 + x)n = 1 + xn + O(x2 n2 ).
It applies even when n is not an integer. In particular, n can be very large, very small, or
even negative. The extra terms will be small as long as xn is small. If desired, one can find
higher terms using binomial coefficients,
∞
n
X n m
(1 + x) = x
m
m=0
where the definition of the binomial coefficient is formally extended to arbitrary real n.
The binomial theorem is one of the most common approximations in physics. It’s really just taking
the first two terms in the Taylor series of (1 + x)n , but we give it a name because it’s so useful.
[1] Problem 17. Suppose the period of a pendulum is one second, and recall that
s
L
T = 2π .
g
If the length is increased by 3% and g is increased by 1%, use the binomial theorem to estimate
how much the period changes. This kind of thinking is extremely useful when doing experimental
physics, and you should be able to do it in your head.
Solution. L is raised to the power of 1/2, and for g it is −1/2. Thus T = T0 (1 + δL/2L)(1 − δg/2g).
Putting in the numbers yield T ≈ 1.01T0 so the period should be 1% longer.
[1] Problem 18. Consider an electric charge q placed at x = 0 and a charge −q placed at x = d. The
electric field along the x axis is then

q 1 1
E(x) = − .
4π0 x2 (x − d)2
For large x, use the binomial theorem to approximate the field.
20
Solution. Use the binomial theorem with d/x 1 to get

1 1 2d
= 1 + .
(x − d)2 x2 x
Then
2qd qd
E(x) = − 3
=− .
4π0 x 2π0 x3
This is the field of an electric dipole.
[3] Problem 19. Some exercises involving square roots.

√
(a) Manually find the Taylor series for 1 + x up to second order, and verify they agree with the
binomial theorem.
√
(b) Approximate 1 + 2x + x2 for small x using the binomial theorem. Does the result match
what you expect? If not, how can you correct it?
√
Solution. (a) The binomial theorem gives 1 + x/2. By differentiating, we get 1/(2 1 + x) and
−1/(4(1 + x)3/2 ). Then
√ 1 1
1 + x = 1 + x − x2 + O(x3 ).
2 8
The first two terms agree with the usual form of the binomial theorem. For the third term,
note that the coefficient should be

1/2 (1/2)(−1/2) 1
= =−
2 2 8
which is indeed what we find.
(b) Of course, the result is 1 + x, so we want the O(x2 ) term to vanish. On the other hand,
applying the binomial theorem gives
p 1 x2
1 + 2x + x2 ≈ 1 + (2x + x2 ) = 1 + x +
2 2
which is wrong! The reason is that the first order binomial theorem isn’t good enough, because
the second order term in the binomial theorem will also contribute a second order term to the
answer. Using the result of part (a),
p 1 1
1 + 2x + x2 = 1 + (2x + x2 ) − (2x + x2 )2 + O((2x + x2 )3 )
2 8
x2 1
=1+x+ − (2x + x2 )2 + O(x3 )
2 8
x2 1
=1+x+ − (2x)2 + O(x3 )
2 8
= 1 + x + O(x3 )
as desired.
21
Example 11: Birthday Paradox
If you have n people in a room, around how large does n have to be for there to be at least
a 50% chance of two people sharing the same birthday?
Solution
Imagine adding people one at a time. The second person has a 1/365 chance of sharing a
birthday with the first. If they don’t share a birthday, the third person has a 2/365 chance
of sharing a birthday with either, and so on. So a decent estimate for n is the n where

1 2 n 1
1− 1− ... 1 − ≈ .
365 365 365 2
The surprising point of the birthday paradox is that n 365. So we can use the binomial
theorem in reverse, approximating the left-hand side as
2
1 2 1 n 1 n /2

1
1− 1− ... 1 − ≈ 1−
365 365 365 365
which is valid since n/365 is small. It’s tempting to use the binomial theorem again to write
2
1 n /2 n2

1
1− ≈1− =
365 2 · 365 2
which gives n = 19. However, this is a bad approximation, because the binomial theorem only
works if (n2 /2)(1/365) is very small, but here we’ve set it to 1/2, which isn’t particularly small.
Since the series expansion variable is 1/2, each term in the series expansion is roughly 1/2 as
big as the last (ignoring numerical coefficients), so we expect to be off by about (1/2)2 = 25%.
The binomial theorem is an expansion for (1 + x)y which works when xy is small. Here xy
isn’t small, and we instead want an approximation that works when only x is small. One
trick to dealing with an annoying exponent is to take the logarithm, since that just turns it
into a multiplicative factor. Note that
log((1 + x)y ) = y log(1 + x) ≈ yx
by Taylor series, which implies that
(1 + x)y ≈ eyx
when x is small, an important fact which you should remember. So we have

2
1 n /2

2 1
1− ≈ e−n /2(365) =
365 2
and solving gives n = 22.5. We should round up since n is actually an integer, giving n = 23,
which is indeed the exact answer.
22
Remark
Precisely how accurate is the approximation (1 + x)y ≈ eyx ? Note that the only approximate
step used to derive it was taking log(1 + x) ≈ x, which means we can get the corrections by
expanding to higher order. If we take the next term, log(1 + x) ≈ x − x2 /2, then we find
2 y/2
(1 + x)y ≈ eyx e−x .
Note that because we are approximating the logarithm of the quantity we want, the next
correction is multiplicative rather than additive; we’ll see a similar situation with Stirling’s
approximation in T2. Our approximation has good fractional precision as long as x2 y 1.
In the previous example, x2 y/2 = (22.5/365)2 /4 = 0.1%, so our answer was quite accurate.
[2] Problem 20. Find a series approximation for xy , given that y is small and x is neither small nor
exponentially huge. (Hint: to check if you have it right, you can try concrete numbers, such as
y = 0.01 and x = 10. The series expansion variable may look a bit unusual.)
Solution. We have
xy = ey log(x) .
If y is small, then for any reasonable x (i.e. x not exponentially huge), y log(x) is also small. So we
can use the Taylor series for the exponential to get
xy ≈ 1 + y log(x) + O((y log(x))2 )
with further terms easily computed.
Remark
If these questions seem complicated, rest assured that 90% of approximations on the USAPhO
and IPhO boil down to using
sin x ≈ x, cos x ≈ 1 − x2 /2, (1 + x)n ≈ 1 + xn, ex ≈ 1 + x, log(1 + x) ≈ x.
I’ve given you a lot of subtle situations above, but it’s these that you have to know by heart.
Almost all situations where you will use these will look like problem 17 or problem 18.
3 Solving Equations
Idea 6
In Olympiads, you may have to find numeric solutions for equations that can’t be solved
analytically. A simple but reliable method is to “guess and check”, starting with a reasonable
first guess (e.g. derived by solving an approximated version of the equation, or sketching the
graphs of both sides), plugging it into both sides, then proceeding with binary search.
23
[3] Problem 21. Sometimes, you can get an accurate numeric answer very quickly on a basic calculator
by using the method of iteration, which solves equations of the form x = f (x).
(a) Take a scientific calculator (in radians), put in any number, and press the “cos” button many
times. Convince yourself that the final number you get is the unique solution to x = cos x.
(b) What are the key features of the graphs of x and cos x that made this work? For example, why
doesn’t pressing cos−1 repeatedly give the same result? As another example, since x = sin x
has a unique solution, why does repeatedly pressing sin not work so well?
(c) Find a nonzero solution for x = tan(x/2).
(d) Find a nonzero solution for ex − 1 = 2x.
(e) Find a positive solution for xx = e.
Solution. (a) Well, just try it!
(b) What makes cos x work and arccos x fail is that near the Dottie number (the solution to
x = cos x), the slope of cos(x) (call it m1 ) is greater than −1, and the slope of arccos(x) (m2 )
is less than −1. This means that when one starts with x + , to the first order cos(x) will map
it to x + m1 , and for arccos x it will be x + m2 . The repeated factors of m1 with absolute
value less than 1 will make iterations of cos x converge to the Dottie number exponentially,
but with |m2 | > 1, the iterations will result in numbers getting farther away from the Dottie
number.
Another, more global reason that cos x works so well is that it’s bounded. So whatever your
initial guess is, at the next stage it’ll be mapped to within [−1, 1], and from then on it’ll
close into the answer. For general functions, you usually have to choose the initial guess more
carefully, or else you’ll get the wrong solution, or diverge to infinity.
The solution to sin x = x is x = 0, but near zero, the slope of sine is gets closer and closer to
1. This makes convergence excruciatingly slow! If you play around a bit with series, you can
√
show that after n iterations, your answer starts shrinking as 1/ n, which is much worse than
the exponential convergence. This is a pretty weird case though; you probably won’t see it in
practice.
In general, iteration can “go wrong” in far weirder ways. For example, suppose you tried to
iterate x → rx(1 − x) for a constant r. This is called the logistic map, and it turns out that
if r is in the right range, the result is chaotic! The result bounces around in an unpredictable
way, never repeating itself, and you get a completely different result after a few iterations if
you start with a very slightly different number.
(c) Note that iterating tan(x/2) will lead to x = 0. In this case, the solution x = 0 is stable,
while the solution we actually want is unstable. Thus to get the other solution, try the inverse:
x = 2 arctan(x). So type in a guess in your calculator like 3, and then enter 2 arctan(Ans),
and keep pressing ”=”. Eventually you’ll get x = 2.331. There’s also a negative solution,
x = −2.331, and which one you get depends on your initial guess.
(d) Iterating (eAns − 1)/2 will also yield x = 0, so iterate x = ln(1 + 2x). Type in a guess like 2,
and type in ln(1 + 2Ans). Eventually you’ll get to x = 1.256.
24
(e) Taking the log of both sides gives x log x = 1. The iteration from x = 1/ log(x) is unstable, so
instead iterate x = e1/x . That is, type in a guess close to 1.8 or so, and iterate e1/Ans . You’ll
get x = 1.7632.
[2] Problem 22. [A] Newton’s method is a more sophisticated method for solving equations, which
converges substantially faster than binary search. Suppose we want to solve the equation f (x) = 0.
Starting with a nearby guess x0 , we evaluate f (x0 ) and f 0 (x0 ), then find our next guess by applying
the tangent line approximation at this point,
f (x0 )
x1 = x0 − .
f 0 (x0 )
The process repeats until we get a suitably accurate answer.
(a) Use Newton’s method to solve x = cos x.
(b) Newton’s method converges quadratically, in the sense that for typical functions, if your
current guess is away from the answer, the next guess will be O(2 ) away. (This implies that
the number of correct digits in the answer roughly doubles with each iteration!) Explain why,
and then find an example where Newton’s method doesn’t converge this fast.
Newton’s method is very important in general, but it’s not that useful on Olympiads, because it
takes a while to set up, especially if the derivative f 0 is complicated, and you usually don’t need
that many significant figures in your answer anyway. (There are alternatives to Newton’s method,
such as Halley’s method, that converge even faster, but the tradeoff is the same: each iteration
takes more effort to calculate, as higher derivatives of f must be computed.)
Solution. (a) We want to solve f (x) = cos x − x = 0, which means we iterate

cos x − x
x− .
sin x + 1
Starting from a reasonable guess x0 = 0.5, we find
x1 = 0.755222, x2 = 0.739142, x3 = 0.739085.
The next iteration gives the same thing for the first six decimal places, so after just three
iterations, we already have six significant digits in the answer.
(b) If the tangent line approximation was exact, then Newton’s method would converge to the
answer in one iteration, f (x1 ) = 0. So if you’re already close to the answer, the leading
source of inaccuracy is the second-order term in the Taylor expansion of f , giving f (x1 ) ≈
2 f 00 (x0 )/2. Applying the tangent line approximation again, this implies we are roughly a
distance 2 f 00 (x0 )/2f 0 (x1 ) ∝ 2 from the answer.
Convergence will be slower if f 0 (x1 ) happens to be small. For example, for finding roots of
polynomials, this will occur for double roots, as the first derivative vanishes at the root itself.
In this case f 0 (x1 ) ∝ , so the error after an iteration is still order , not 2 .
The simplest example where this happens is f (x) = x2 , where
x20 x0
x1 = x0 − = .
2x0 2
25
This is no longer quadratically convergent; instead the error goes down by the same factor in
each iteration, so the number of significant figures correct goes up linearly.
It’s interesting to compare this to iteration. When the method of iteration works, we typically
have exponential convergence, which means the number of significant figures goes up linearly.
However, in cases like f (x) = x2 where f 0 (x) vanishes at the solution, the error is squared in
each iteration, so the method of iteration instead converges quadratically! In other words, for
these exceptional cases, the convergence rates of iteration and Newton’s method swap.
Remark
You’ve seen several approximation methods above, and going forward, you should feel free
to use whichever looks best in each situation. However, if you’re solving problems using
the same calculator you use for schoolwork, you should make sure to not rely on its more
advanced features. In Olympiads, you’re generally only allowed to use an extremely basic
scientific calculator, with a tiny display and no memory except for the “Ans” key.
Example 12
In units where c = 1, the Lorentz factor is defined as

1
γ=√ .
1 − v2
Suppose that a particle traveling very close to the speed of light has γ = 108 . Find the
difference ∆v between its speed and the speed of light.
Solution
This problem looks easy; by some trivial algebra we find
p
∆v = 1 − 1 − 1/γ 2 .
But when you plug this into a cheap scientific calculator, you get zero, or something that’s
quite far from the right result. The problem is that we are trying to find a small quantity ∆v
by subtracting two nearby, much larger quantities. But the calculator has limited precision,
and it ends up rounding 1 − 1/γ 2 = 1 − 10−16 a bit, giving a completely wrong answer!
Instead, we can apply the binomial theorem to find

1
∆v ≈ + O(1/γ 4 ).
2γ 2
This is no longer the exact answer, but it’s a great approximation, because the error term is
around 1/γ 2 ∼ 10−16 times as small as the answer, and it’s easy for a calculator to evaluate.
The lesson is that it’s better to be accurate in practice than to be precise in theory.
[1] Problem 23. Find the solutions of the equation x2 − 1020 x + 1 = 0 to reasonable accuracy.
26
Solution. Applying the quadratic formula, the solutions are

√
1020 ± 1040 − 4
x= .
2
Of course you can’t just plug this into a calculator and expect a reasonable result. Instead, we need
to approximate. For the larger root, an excellent approximation is
√
1020 + 1040
x≈ = 1020 .
2
Then by Vieta’s formula, an excellent approximation for the other root is 10−20 .
[4] Problem 24. [A] Consider the equation x3 − x2 + 1 = 0, where is small. Find approximate
expressions for all three roots of this equation, up to and including terms of order .
Solution. If we set = 0, then the roots of the resulting quadratic equation are ±1. Thus, two
of the roots should be near ±1. To calculate the O() correction, let x = 1 + A + O(2 ). Then
plugging this into the equation gives
(1 + A)3 − (1 + A)2 + 1 = − 2A + O()2 = 0.
Thus, we find A = 1/2. A similar calculation can be done for the root near x = −1, giving roots

x=1+ + O(2 ), x = −1 + + O(2 ).
2 2
However, the third root is nowhere to be found in this analysis, because the quadratic only has two
roots. Upon graphing the function, you can see that the third root is at very large x, once the cubic
term catches up in size to the quadratic term. This happens when x ≈ 1/. This appearance of an
inverse power of makes this a “singular perturbation series”.
Here’s a general way to conceptualize what’s going on here. The equation in this problem has
three terms, and it’s easy to find a root if any one of the terms is negligible compared to the others.
For example, for the first two roots, we assumed the x3 term was negligible, and then found x = ±1.
Then, adding on the x3 term produces O() and higher corrections to the left-hand side, which
can be used to compute O() and higher corrections to the root itself. Now, this third root we’ve
just found occurs when the 1 term is negligible. In this case, both of the first two terms are of order
1/2 , and the 1 creates small corrections to the root (relative to its huge size).
Since 1 is two orders in smaller than 1/2 , we expect these terms only appear two orders down
in the root. That is, we expect the root has the form
1
1 + A2 + O(3 )

x=

with no O() term in parentheses. (If you don’t believe this, check this term vanishes for yourself!)
Plugging this into the equation gives
1 1
2
(1 + A2 + O(3 ))3 − 2 (1 + A2 + O(3 ))2 + 1 = 0

which is equivalent to
3A − 2A + 1 + O() = 0
27
from which we conclude A = −1, and hence the third root is

1
x= − + O(2 ).

Finally, you might be wondering what happens if the x2 term is the negligible one. However, this
never happens. If we assume it’s negligible, then we need x ≈ −−1/3 , so that both the other terms
are about 1. But then the x2 term is 1/2/3 1. So we can’t assume the x2 term is negligible
self-consistently, so it doesn’t give any new roots. The idea used above, of supposing two of the
terms are large, using that to solve a simpler equation, and then checking for consistency, is known
as the method of dominant balance.
4 Limiting Cases
Idea 7
Limiting cases can be used to infer how the answer to a physical problem depends on its
parameters. It is primarily useful for remembering the forms of formulas, but can also be
powerful enough to solve multiple choice questions by itself.
Example 13
What is the horizontal range of a rock thrown with speed v at an angle θ to the horizontal?
Solution
This result is easy to derive, but dimensional analysis and extreme cases can be used to
recover the result too. The answer can only depend on v, g, and θ, so by dimensional analysis
it is proportional to v 2 /g. This is sensible, since the range increases with v and decreases
with g. Now, the range is zero in the extreme cases θ = 0 and θ = π/2, but not anywhere in
between, so if we remember the range contains a simple trigonometric function, it must be
sin(2θ), so
v2
R∝ sin(2θ).
g
We can also get the prefactor by a simple limiting case, the case θ 1. In this case, by the
small angle approximation,
vx ≈ v, vy ≈ vθ.
The time taken is t = 2vy /g, so the range is
2v 2
R ≈ vx t = θ.
g
Thus there is no proportionality constant; the answer is
v2
R= sin(2θ).
g
In reality, it’s probably faster to go through the full derivation than all of this reasoning, but
if you’re just not sure about whether it’s a sine or a cosine, or what the prefactor is, then
28
limiting cases can be quickly used to recover that piece. Also note that the approximations
we used above are frequently useful for evaluating limiting cases.
Example 14
Consider an Atwood’s machine with masses m and M , and a massless pulley. Find the
tension in the string.
Solution
Since the equations involved are all linear equations, we expect the answer should also
be simple. It can only depend on g, m, and M , so by dimensional analysis, it must be
proportional to g. By dimensional analysis, this must be multiplied by something with one
net power of mass. Since the answer remains the same if we switch the masses, it should be
symmetric in m and M .
Given all of this, the simplest possible answer would be
T ∝ g(m + M ).
To test this, we consider some limiting cases. If M m, the mass M is essentially in freefall,
so the mass m accelerates upward with acceleration g. Then the tension is approximately
2mg. Similarly, in the case M m, the tension is approximately 2M g. These can’t be
satisfied by the form above.
The next simplest option is a quadratic divided by a linear expression. Both of these must
be symmetric, so the most general possibility is
A(m2 + M 2 ) + BmM
T =g .
m+M
Then the limiting cases can be satisfied if A = 0 and B = 2, giving
2gmM
T = .
m+M
[1] Problem 25. Find the perimeter of a regular N -gon, if L is the distance from the center to any
of the sides. By considering a limiting case, use this to derive the circumference of a circle.
Solution. By basic trigonometry, the perimeter is 2N L sin(π/N ). Then the circumference of a

circle is
π
lim 2N L sin(π/N ) = 2N L = 2πL
N →∞ N
as expected. We can see that the limit of N sin(π/N ) is π through the small angle approximation.
If you want more rigor, you could also say that this is an indeterminate form ∞ × 0, and use
l’Hospital’s rule.
29
[1] Problem 26. Use similar reasoning to find the acceleration of the Atwood’s machine. (We will
show an even easier way to do this, using “generalized coordinates”, in M4.)
Solution. We know from dimensional analysis that the acceleration is gf (m, M ) where f (m, M )
is dimensionless. Thus it should be a fraction.
If either of the masses is much more massive than the other mass, then the acceleration should
be g. Thus the coefficients of m, M should be ±1. If the masses are equal, then the acceleration
should be 0. This leads to a M − m term in the numerator. Since the denominator should be
different but still have factors of ±1, a reasonable answer is
m−M
a= g.
m+M
which is indeed the real answer.
[2] Problem 27 (Morin 1.6). A person throws a ball (at an angle of her choosing, to achieve the
maximum distance) with speed v from the edge of a cliff of height h. Which of the below could be
an expression for the maximal range?
s r
gh2 v2 v2h v2 v2 v 2 /g

2gh 2gh
, , , 1 + , 1 + , .
v2 g g g v2 g v2 1 − 2gh/v 2
If desired, try Morin problems 1.13, 1.14, and 1.15 for additional practice.
Solution. First check if they’re all dimensionally correct (they are). When h = 0, the maximum
range as found above with sin(2θ) = 1 is v 2 /g. Also the maximum range obviously depends on the
height of the edge of the cliff, and there shouldn’t be a case of a finite height or velocity where the
range becomes infinite. This leaves 2 options:
r
v2 v2

2gh 2gh
1+ 2 , 1+ 2
g v g v
When h is small, the extra distance at the end of the trajectory from dipping down a vertical
distance h can be found with binomial theorem: h, and 2h respectively. Since the trajectory is
symmetric, when h ≈ 0 (to be more concise, h v 2 /g) the optimal launch angle is 45 deg, so by
geometry the extra distance should also be h. Thus the correct formula is
r
v2 2gh
1+ 2 .
g v
[2] Problem 28. Consider a triangle with side lengths a, b, and c. It turns out the area of its incircle
can be expressed purely by multiplying and dividing combinations of these lengths. Moreover,
the answer is the simplest possible one consistent with limiting cases, dimensional analysis, and
symmetry. Guess it!
Solution. In the limiting case a = b + c, the triangle collapses and the area must be zero, which
means the answer must be proportional to b + c − a. But the answer should also be symmetric
between exchanging a, b, and c, so it must be proportional to (b + c − a)(c + a − b)(a + b − c). The
dimension of this quantity is one too high, so we need to divide by a length, and the only possibility
30
consistent with symmetry is a + b + c. Finally, the overall constant can be fixed using the special
case of an equilateral triangle, giving the result
π (a + b − c)(b + c − a)(c + a − b)
A= .
4 a+b+c
Incidentally, the area of the excircle is π(abc)2 /((a + b + c)(a + b − c)(b + c − a)(c + a − b)). While
most of the denominator makes sense from limiting cases, the overall expression is certainly harder
to guess, since powers of abc and a + b + c could cancel while preserving all the limiting cases and
symmetry. That just goes to show that limiting cases can only get you so far. In some sense, “real”
math starts once all the easy information accessible to methods like these has been accounted for.
While we won’t have more questions that are explicitly about dimensional analysis or limiting
cases, these are not techniques but ways of life. For all future problems you solve, you should be
constantly checking the dimensions and limiting cases to make sure everything makes sense.
5 Manipulating Differentials
You might have been taught in math class that manipulating differentials like they’re just small,
finite quantities, and treating derivatives like fractions is “illegal”. But it’s also very useful.
Idea 8
Derivatives can be treated like fractions, if all functions have a single argument.
The reason is simply the chain rule. The motion of a single particle only depends on a single
parameter, so the chain rule is just the same as fraction cancellation. For example,
dv d dv dx
= v(x(t)) =
dt dt dx dt
which show that “canceling a dx” is valid. Similarly, you can show that
dy dx
=1
dx dy
by considering the derivative with respect to x of the function x(y(x)) = x.
As a warning, for functions of multiple arguments, the idea above breaks down. For example,
for a function f (x(t), y(t)), the chain rule says
df ∂f dx ∂f dy
= +
dt ∂x dt ∂y dt
where there are two terms, representing the change in f from changes only in x, and only
in y. Therefore, when we start studying thermodynamics, where multivariable functions are
common, we will treat differentials more carefully. But for now the basic rules will do.
31
Remark
Math students tend to get very upset about the above idea: they say we shouldn’t use
convenient notation if it hides what’s “really” going on. And they’re right, if your goal is
to put calculus on a rigorous footing. But in physics we have no time to luxuriate in such
rigor, because we want to figure out how specific things work. The point of notation is to
help us do that by suppressing mathematical clutter. A good notation suppresses as much
as possible while still giving correct results in the context it’s used.
To illustrate the point, note that elementary school arithmetic is itself an “unrigorous” nota-
tion that hides implementation details. If we wanted to be rigorous about, say, defining the
number 2, we would write it as S(1) where S is the successor function, obeying properties
specified by the Peano axioms. And 4 is just a shorthand for S(S(S(1))), so 2 + 2 = 4 means
S(1) + S(1) = S(S(S(1))).
Even this is not “rigorous”, because the Peano axioms don’t specify how the numbers or
the successor function are defined, just what properties they have to obey. To go deeper,
we could define the integers as sets, and operations like + in terms of set operations. For
example, in one formulation,
S(S(S(1))) = {∅, {∅}, {∅, {∅}}, {∅, {∅}, {∅, {∅}}}.
People have seriously advocated for 1st grade math to be taught this way, which has always
struck me as insane. You can always add more arbitrary layers of structure underneath the
current foundation, so such layers should only be added when absolutely necessary.
As another example, in physics we often work with v(t) in one part of a problem, and then
v(x) in another. Mathematicians hate this, because we are denoting two different functions
with the same letter, but we do it because in both cases v is standing for the same physical
quantity, which doesn’t care how we parametrize it. (And if we didn’t do it this way, even a
simple projectile motion problem can end up taking half the alphabet!) As a converse example,
I always keep track of dimensions, but in math textbooks you can often see jarring equations
like “g = −32.” They suppress dimensions, even though all physical quantities have them,
because they’re not relevant to their goals, and that’s completely legitimate. Everybody
always has to forget about something, or else nobody would ever be able to do anything.
Example 15
Derive the work-kinetic energy theorem, dW = F dx.
Solution
Canceling the mass from both sides, we wish to show
1
d(v 2 ) = a dx.
2
32
To do this, note that

1 dx dv
d(v 2 ) = v dv = dv = dx = a dx
2 dt dt
as desired. If you’re not satisfied with this derivation, because of the bare differentials floating
around, we can equivalently prove that F = dW/dx, by noting
dW dv dv dt dv
= mv = mv =m = F.
dx dx dt dx dt
[2] Problem 29. Some more about power.
(a) Use similar reasoning to derive P = F v.
(b) An electric train has a power line that can deliver power P (x), where x is the distance along
the track. If the train starts at rest at x = 0, find its speed at point x0 in terms of an integral
of P (x). (Hint: try to get rid of the dt’s to avoid having to think about the time-dependence.)
Solution. (a) First, let’s use differentials. Since P = dW/dt, we have
dW = F v dt.
Using the same reasoning as before, dW = md(v 2 )/2 = mv dv, so
mv dv = mav dt.
Canceling on both sides, this simplifies to dv = a dt, which is clearly true. Alternatively, we
can use derivatives directly. We have
dW dv
P = = mv = mva = F v
dt dt
as desired.
(b) We note that

dW = mv dv
but we also have
dt P
dW = P dt = P dx = dx
dx v
where we introduced the power of v to convert dt (which we don’t want to deal with) to dx.
Doing some rearrangement, Z Z
mv 2 dv = P dx.
Performing the integral, we have

Z 1/3
3
v= P dx .
m
[2] Problem 30 (Kalda). The deceleration of a boat in water due to drag is given by a function a(v).
Given an initial velocity v0 , write the total distance the boat travels as a single integral.
33
Solution. We have Z Z Z Z
dx dx dt v dv
dx = dv = dv =
dv dt dv a(v)
which is a single integral in terms of the function a(v), as desired. Putting the bounds in, the total
distance is Z 0
v dv
∆x = .
v0 a(v)
The signs are correct here, since both dv and a(v) are negative.
[5] Problem 31. A particle in a potential well.
(a) Consider a particle of mass m and energy E with potential energy V (x), which performs
periodic motion. Write the period of the motion in terms of a single integral over x.
(b) Suppose the potential well has the form V (x) = V0 (x/a)n for n > 0. If the period of the motion
is T0 when it has amplitude A0 , find the period when the amplitude is A, by considering how
the integral you found in part (a) scales with A.
(c) Find a special case where you can check your answer to part (b). (In fact, there are two more
special cases you can check, one which requires negative n and negative V0 , and one which
requires V (x) to be replaced with its absolute value.)
(d) Using a similar method to part (a), write down an integral over θ giving the period of a
pendulum with length L in gravity g, without the small angle approximation. Using this,
compute the period of the pendulum with amplitude θ0 , up to order θ02 . (This result was first
published by Bernoulli, in 1749.)
(e) ? Part (d) is the kind of involved computation you might see in a graduate mechanics course.
But if you think you’re really tough, you can go one step further. Consider a mass m oscillating
on a spring of spring constant k with amplitude A. Calculate its period of oscillation up to
order A2 , accounting for special relativity. (Concretely, assume that the spring force doesn’t
change the rest mass m, and has a potential U = kx2 /2. In relativity, the
p force F = −dU/dx
still obeys F = dp/dt, but now E = γmc2 and p = γmv, where γ = 1/ 1 − v 2 /c2 .)
Solution. (a) The statement of conservation of energy is

r
1 2 2(E − V (x))
E = mv + V (x), v = .
2 m
Therefore, the period is
Z Z Z
dt dx
T = dt = dx = p .
dx 2(E − V (x))/m
To be more precise, we should put the bounds of integration back in. If the lowest and highest
values of x are xmin and xmax , then
Z xmax
dx
T =2 p
xmin 2(E − V (x))/m
where the factor of two is because this is just half of the oscillation.
34
(b) The particle can perform periodic motion if at x = ±A, v = 0 so V0 (A/a)n = E. Thus
Z A Z A
dx dx
T =2 p ∝ √
−A 2(V0 (A/a)n − V0 (x/a)n ))/m −A An − x n
By dimensional analysis, the integral (a function of A) is proportional to A1−n/2 , so

1−n/2
A
T = T0
A0
Incidentally, you can also do this problem by dimensional analysis directly on the parameters.
At first glance, this is impossible because there are too many dimensionful quantities: E, m, a,
V0 , and T , which permit 5 − 3 = 2 dimensionless groups. (Recall from an earlier problem that
one can usually get a scaling relation only if there’s only 1 dimensionless group.) However, the
situation can be saved by noting that V0 and a only ever appear together in the combination
V0 /an . So there are only 4 independent dimensionful parameters, and a standard dimensional
analysis yields the same result.
(c) The three analytically tractable example are:
• For n = 2 we have simple harmonic motion, and indeed here the period is independent
of amplitude. (Incidentally, can you think of any potentials that aren’t simple harmonic,
but also have this property?)
• For n = −1 we have an inverse square force and T ∝ A3/2 . This makes sense, because it
matches the form of Kepler’s third law, which gives the general scaling of orbits in inverse
square forces. (Here we’re considering the degenerate case of a straight-line orbit.)
• For n = 1 we have a constant force, which doesn’t yield oscillations. But the scaling
argument of part (b) would still work if we used√the potential V (x) = V0 |x/a|, which does
have oscillations. In this case we predict T ∝ A, which makes sense; it corresponds to
the usual time-dependence ∆x = gt2 /2 ∝ t2 of uniformly accelerated motion.
That’s basically as far as you can go with the functions you learn in high school and college.
There are analytic solutions for other n, but they tend to be in terms of exotic special functions.
For instance, for n = 4 the solutions can be written in terms of Jacobi elliptic functions, as
you can see here. Of course, since we’re not living in the 19th century, you don’t need to know
about them to do Olympiads, or even most fields of physics research.
(d) Conservation of energy states

1 2
Iω = mgL(cos θ − cos θ0 ), I = mL2
2
which means Z θ0 Z θ0
dθ dθ
T =4 =4 p .
0 ω 0 (2g/L)(cos θ − cos θ0 )
Now the trick is figuring out how far to expand the cosines. To zeroth order in θ and θ0 , the
denominator is simply zero, but this is too weak of an approximation. If we expand to second
order, however, we get s Z s
L θ0 dθ L
T =4 p = 2π
g 0 θ02 − θ2 g
35
after using a trigonometric substitution, which is the leading order result for the period. This
illustrates the point that it can be nontrivial which order of approximation in one quantity
(the denominator) corresponds to the same order of approximation in another (the period).
The O(θ0 ) correction to the period simply vanishes, because the period should not depend
on the sign of θ0 . Thus, we only need to compute the O(θ02 ) correction, which corresponds to
expanding the cosine to O(θ04 ). This gives
s Z
L θ0 dθ
T =4 p .
g 0 θ0 − θ2 + (θ4 − θ04 )/12
2
The scaling is clearer if we substitute x = θ/θ0 , giving

s Z s Z
L 1 dx L 1 dx 1
T =4 p = 4 √ p .
g 0 2
1 − x2 + θ0 (x4 − 1)/12 g 0 2
1 − x2 1 − θ0 (x2 + 1)/12
This form makes it clear that to get the period to O(θ02 ), we just need to expand the square
root with the binomial theorem, giving
s Z
L 1 θ02

dx 2
T =4 √ 1 + (1 + x ) .
g 0 1 − x2 24
The new term can also be integrated straightforwardly with a trigonometric substitution
u = sin−1 (x), giving the final result
s
θ02

L 4
T = 2π 1+ + O(θ0 ) .
g 16
(e) This is a taste of the kind of problem you’ll see in R2. It can get quite messy, but it’s not
too bad if you work in the right variables. First, note that since F = −dU/dx, we still have
energy conservation, but with the relativistic energy expression,
1 1
γmc2 + mω 2 x2 = mc2 + mω 2 A2
2 2
where ω 2 = k/m as usual. Solving for γ, we find
ω2 2
γ =1+ (A − x2 ).
2c2
Next, using the definition of γ, we have
Z A Z A
dx 4 γ
T =4 = p dx.
0 v c 0 γ2 −1
At this point we can perform a quick check to make sure we’re on the right track. Note that
in the ultrarelativistic limit, where the spring is so strong that the mass is always moving
at nearly the speed of light, we have γ → ∞, so that the integrand just reduces to 1. Then
T ≈ 4A/c, which is exactly as expected.
36
Anyway, we’re interested in the case where the relativistic corrections are small, γ − 1 1.
The easiest way to make this manifest is to eliminate γ in favor of A, using our result above.
There we found that γ − 1 = O((ωA/c)2 ), so we can expand in the small quantity ωA/c, giving
4 A c
Z
1 3 ωp 2
T = √ + A − x2 + O((ωA/c)4 ) dx.
c 0 ω A2 − x 2 8 c
The first term simply recovers the nonrelativistic result T = 2π/ω, and the second term is
straightforward to integrate, yielding
3 ω 2 A2

2π 4
T = 1+ + O((ωA/c) ) .
ω 16 c2
Since the peak speed v0 is approximately ωA in the nonrelativistic limit, this result is therefore
accurate up to corrections of order (v0 /c)4 .
6 Multiple Integrals
It’s also useful to know how to set up multiple integrals. This is fairly straightforward, though
technically an “advanced” topic, so we’ll demonstrate it by example. For further examples, see
chapter 2 of Wang and Ricardo, volume 1, or MIT OCW 18.02, lectures 16, 17, 25, and 26.
Idea 9
In most Olympiad problems, multiple integrals can be reduced to single integrals by symmetry.
Example 16
Calculate the area of a circle of radius R.
Solution
The area A is the integral of dA, i.e. the sum of the infinitesimal areas of pieces we break the
circle into. As a first example, let’s consider using Cartesian coordinates. Then the pieces
will be the rectangular regions centered at (x, y) with sides (dx, dy), which have area dx dy.
The area is thus Z Z Z
A = dA = dx dy.
The only tricky thing about setting up the integral is writing down the bounds. The inner
integral is done first, so its bounds depend
√ on the value of x. Since the boundary of the circle
2 2 2
is x + y = R , the bounds are y = ± R2 − x2 . Thus we have
√
Z R Z R2 −x2
A= dx √ dy.
−R − R2 −x2
We then just do the integrals one at a time, from the inside out, like regular integrals,
Z R p Z 1 p Z π/2
2 2
A= 2 2
2 R − x dx = 2R 2
1 − u du = 2R cos2 θ dθ = πR2
−R −1 −π/2
37
where we nondimensionalized the integral by letting u = x/R, and then did the trigonometric
substitution u = sin θ. (To do the final integral trivially, notice that the average value of
cos2 θ along any of its periods is 1/2.)
We can also use polar coordinates. We break the circle into regions bounded by radii r and
r + dr, and angles θ and θ + dθ. These regions are rectangular, with side lengths of dr and
r dθ, so the area element is dA = r dr dθ. Then we have
Z R Z 2π Z R
A= r dr dθ = 2π r dr = πR2
0 0 0
which is quite a bit easier. In fact, it’s so much easier that we didn’t even need to use double
integrals at all. We could have decomposed the circle into a bunch of thin circular shells,
argued that each shell contributed area (2πr) dr, then integrated over them,
Z R
A= 2πr dr = πR2 .
0
In Olympiad physics, there’s usually a method like this, that allows you to get the answer
without explicitly writing down any multiple integrals.
Example 17
Calculate the moment of inertia of the circle above, about the y axis, if it has total mass M
and uniform density.
Solution
The moment of inertia of a small piece of the circle is
x2 M
dI = x2 dm = x2 σ dA = dA
πR2
where x2 appears because x is the distance to the rotation axis, and σ is the mass density
per unit area. Using Cartesian coordinates, we have
√
Z R Z R2 −x2
M
I= dx √ x2 dy.
πR2 −R − R2 −x2
The inner integral is still trivial; the x2 doesn’t change anything, because from the perspective
of the dy integral, x is just some constant. However, the remaining integral becomes a bit
nasty. In general, when this happens, we can try flipping the order of integration, giving
Z R Z √R2 −y2
M
I= dy √ x2 dx.
πR2 −R − R2 −y 2
Unfortunately, this is equally difficult. Both of these integrals can be done with trigonometric
substitutions, as you’ll check below, but there’s also a clever symmetry argument.
38
Notice that I is also equal to the moment of inertia about the x axis, by symmetry. So if we
add them together, we get
Z Z
2I = (x + y ) dm = r2 dm.
2 2
The r2 factor has no dependence on θ at all, so the angular integral in polar coordinates is
trivial. We end up with Z R
M 1
2I = 2
2πr r2 dr = M R2
πR 0 2
which gives an answer of I = M R2 /4, as expected.
[2] Problem 32. Calculate I in the previous example by explicitly performing either Cartesian integral.
Solution. Starting from the second expression in the example,
Z R Z √R2 −y2 Z R
M 2 M
I= dy √ x dx = 2(R2 − y 2 )3/2 dy.
πR2 −R − R2 −y 2 3πR2 −R
Let y = R sin θ. Then we have

π/2
2M R2
Z
I= cos4 θ dθ.
3π −π/2
This integral can be done by repeatedly using the double angle formula,
1 + cos(2θ) 2
Z π/2 Z Z π/2
1 1 1 3π
cos4 θ dθ = dθ = + cos(2θ) + + cos(4θ) dθ = .
−π/2 2 −π/2 4 8 8 8
Personally, I can never remember all the trigonometric formulas, and I usually just expand everything
in complex exponentials. Here that method gives a slick solution, as
Z π/2 Z π/2
1
4
cos θ dθ = (eiθ + e−iθ )4 dθ.
−π/2 16 −π/2
Now note that expanding with the binomial theorem gives terms of the form e2inθ for integers n,
which integrate to zero unless n = 0. So the only term that matters gives
Z π/2 Z π/2
4 1 4 3π
cos θ dθ = dθ = .
−π/2 16 −π/2 2 8
Whichever method you used, we conclude the answer is I = M R2 /4, as expected.

[3] Problem 33. In this problem we’ll generalize some of the ideas above to three dimensions, where
we need triple integrals. Consider a ball of radius R.
(a) In Cartesian coordinates, the volume element is dV = dx dy dz. Set up an appropriate triple
integral for the volume.
(b) The inner two integrals might look a bit nasty, but we already have essentially done them.
Using the result we already know, perform the inner two integrals in a single step, and then
perform the remaining integral to derive the volume of a sphere.
39
(c) In cylindrical coordinates, the volume element is dV = r dr dθ dz. Set up a triple integral for
the volume, and perform it. (Hint: this can either be hard, or a trivial extension of part (b),
depending on what order of integration you choose.)
(d) In spherical coordinates, the volume element is dV = r2 dr sin φ dφ dθ. Set up a triple integral
for the volume, and perform it.
(e) Let the ball have uniform density and total mass M . Compute its moment of inertia about
the z-axis. (Hint: this can be reduced to a single integral if you use an appropriate trick.)
Solution. (a) By analogy to the two-dimensional case,

Z R Z √
R2 −x2 Z √R2 −x2 −y2
V = dx √ dy √ dz.
−R − R2 −x2 − R2 −x2 −y 2
(b) The inner two integrals just represent the area of a circle, formed by slicing the ball along a
plane of constant x. Thus, the answer has to be πr2 where r is the radius of that circle (as
we derived explicitly in the example), and in this case r2 = R2 − x2 . Thus, we have
Z R Z 1
4
V = π(R2 − x2 ) dx = πR3 1 − x2 dx = πR3 .
−R −1 3
(c) By analogy to the two-dimensional case,

√
Z R Z R2 −z 2 Z 2π
V = dz r dr dθ.
−R 0 0
Again, the inner two integrals look a bit nasty, but they represent nothing more than the area
of a circle of radius r, leaving Z R
V = π(R2 − x2 ) dx
−R
upon which the solution continues just as in part (b).
(d) The first task is to decide what order the integrals appear in. It’s probably best to have the dr
integral be the outermost one, because surfaces of constant dr are spheres, which are simple;
thus the final integral is just an integral over spherical slices, which we know are simple. By
comparison, if the last integral were dθ we would have hemispherical slices, while if it were
dφ we would have slices with a really weird shape. We thus have
Z R Z π Z 2π
2
V = r dr sin φ dφ dθ.
0 0 0
The inner two integrals can be done easily, giving

Z R
4
V = 4π r2 dr = πR3 .
0 3
(e) We are looking for Z

I= x2 + y 2 dm.
40
By spherical symmetry, the integrals of x2 dm, y 2 dm, and z 2 dm are all equal. Thus,
Z
2
I= x2 + y 2 + z 2 dm
3
but this integral is now easy to do because it has spherical symmetry. We have
Z R
2 M 2
I= 4 3 4πr2 r2 dr = M R2
3 3 πR 0 5
as expected. The same trick can be used to show that the moment of inertia of a spherical
shell is (2/3)M R2 .
[2] Problem 34. Consider a spherical cap that is formed by slicing a sphere of radius R by a plane,
so that the altitude from the vertex to the base is h. Find the area of its curved surface using an
appropriate integral.
Solution. This is a double integral, where it’s best to use spherical coordinates. Recall that the
volume element in spherical coordinates was dV = r2 dr sin φ dφ dθ. Thus, the area element for a
part of this sphere is dA = R2 sin φ dφ dθ. The area integral is
Z cos−1 ((R−h)/R) Z 2π
A = R2 sin φ dφ dθ = 2πhR.
0 0
After doing the trivial inner integral, this approach is just slicing the surface by dφ. You can also
equivalently solve it by slicing it in dz. In that case the integrand is a bit more complicated, but
the bounds are simpler.
Remark
You might be wondering how good you have to be at integration to do Olympiad physics.
The answer is: not at all! You need to understand how to set up integrals, but you almost
never have to perform a nontrivial integral. There will almost always be a way to solve the
problem without doing explicit integration at all, or an approximation you can do to render
the integral trivial, or the integral will be given to you in the problem statement. The Asian
Physics Olympiad takes this really far: despite R having some of the hardest problems ever
written, they often provide information like “ xn dx = xn+1 /(n + 1) + C” as a hint! This
is because physics competitions are generally written to make students think hard about
physical systems, and the integrals are just viewed as baggage.
In fact, plain old AP Calculus probably has harder integrals than Olympiad physics. For
example, in those classes everybody has to learn the integral
Z
sec θ dθ = log |sec θ + tan θ| + C
which has a long history. When I was in high school, I was shocked by how the trick for doing
this integral came out of nowhere; it seemed miles harder than anything else taught in the
class. And it is! Historically, it arose in 1569 from Mercator’s projection, where it gives the
vertical distance on the map from the equator to a given latitude. For decades, cartographers
simply looked up the numeric value of the integral in tables, where the Riemann sums had
41
been done by hand. (They had no chance of solving it analytically anyway, since Napier only
invented logarithms in 1614.) Gradually, tabulated values of the logarithms of trigonometric
functions became available, and in 1645, Bond conjectured the correct result by noticing the
close agreement of tabulated values of each side of the equation. Finally, Gregory proved the
result in 1668, using what Halley called “a long train of Consequences and Complications of
Proportions.” So it took almost a hundred years for this integral to be sorted out! (Though to
their credit, they had the handicap of not knowing about differentiation or the fundamental
theorem of calculus; they were finding the area under the curve with just Euclidean geometry.)
Even though Olympiad physics tries to avoid tough integrals, doing more advanced physics
tends to produce them, so physicists often get quite good at integration. By contrast,
Spivak’s calculus textbook for math majors only covers integration techniques in a single
chapter towards the end of the book. He justifies the inclusion of this material by saying:
Every once in a while you might actually need to evaluate an integral [...] For
example, you might take a physics course in which you are expected to be able
to integrate. [...] Even if you intend to forget how to integrate (and you probably
will forget some details the first time through), you must never forget the basic
methods.
That attitude is why physics students frequently win the MIT Integration Bee.
42
Problem Solving II: Experimental Methods

For more about the theory of uncertainty analysis, see this handout, or for a more introductory
take, this handout and this comic. For practical tips for real experiments, especially at the IPhO,
see chapter 7 of Physics Olympiad: Basic to Advanced Exercises. For some entertaining general
discussion, see chapters I-5 and I-6 of the Feynman lectures. There is a total of 81 points.
1 Basic Statistics
Idea 1
If a quantity X has the probability distribution p(x), that means
Z b
the probability that a ≤ X ≤ b is p(x) dx.
a
In particular, the total probability has to sum to one, so

Z ∞
p(x) dx = 1.
−∞
Using the probability distribution, we can calculate expectation values, i.e. averages. For
example, the expectation value of X, also called the mean, is
Z ∞
hXi = xp(x) dx
−∞
while the expectation value of an arbitrary function of X is

Z ∞
hf (X)i = f (x)p(x) dx.
−∞
One especially important quantity is the variance of X, defined as
var X = hX 2 i − hXi2 .
√
The standard deviation is defined by σX = var X. It describes how “spread out” the
distribution of X is, and it will play an important role in uncertainty analysis.
[1] Problem 1. Suppose that x is a length. What are the dimensions of p(x), hXi, var X, and σ?
Solution. Since p(x) dx is dimensionless, we have
[p(x)] = L−1
where L denotes length. Similarly,
[hXi] = L, [var X] = L2 , [σX ] = L.
1
Example 1
Trains arrive at a train station every 10 minutes. If I arrive at a random time, and X is the
number of minutes I have to wait, what is the standard deviation of X?
Solution
We see that X can be anywhere between 0 and 10, with all possibilities equally likely, so
(
1/10 min−1 0 ≤ x ≤ 10,
p(x) =
0 otherwise
where the 1/10 guarantees the total probability is 1. For the rest of this example, we’ll
suppress the units. We have
Z ∞ Z 10
x
hXi = xp(x) dx = dx = 5
−∞ 0 10
which makes sense, as I should have to wait half the maximum time on average, and
Z ∞ Z 10 2
2 2 x 100
hX i = x p(x) dx = dx = .
−∞ 0 10 3
Then the standard deviation is

p 5
σX = hX 2 i − hXi2 = √ min.
3
[3] Problem 2. Consider an exponentially distributed quantity,

(
ae−ax x ≥ 0,
p(x) =
0 otherwise.
Verify that the total probability is 1, and compute the mean and standard deviation. To perform
the integrals, you will have to integrate by parts.
Solution. First, to check normalization,
Z ∞ Z ∞ Z ∞
p(x)dx = ae−ax dx = e−u du = 1 − 0 = 1.
−∞ 0 0
Now, the mean can be evaluated using integration by parts,

Z ∞ ∞ Z ∞ ∞
1 1
hxi = xae−ax dx = −xe−ax + e−ax dx = 0 − e−ax = .

0 0 0 a 0 a
To calculate the standard deviation, we must evaluate
Z ∞ Z ∞
2 −ax 2 2
2
hx i = x ae dx = 0 + (2x)e−ax dx = hxi = 2 .
0 0 a a
We thus conclude r
p 2 1 1
σX = hX 2 i − hXi2 = 2
− 2 = .
a a a
2
[2] Problem 3. The purpose of subtracting hXi2 in the variance is to make sure it doesn’t change
when a constant is added to x, since shifting something left or right on the number line shouldn’t
change its spread. Verify that for any constant c, var X = var(X + c).
Solution. We have
var(X + c) = h(X + c)2 i − hX + ci2 .
By the definition of the expectation value, we have
hA + Bi = hAi + hBi, hcAi = chAi
for any quantities A and B and any constant c. Thus,
var(X + c) = hX 2 i + h2Xci + hc2 i − hXi2 − 2hXihci − hci2 = var X
as desired.
[3] Problem 4. We say X is normally distributed if

2
p(x) ∝ e−a(x−b) .
For simplicity, let’s shift X so that it’s centered about x = 0, so

2
p(x) ∝ e−ax .
You may use the result given in P1,

Z ∞ √
2
e−x dx = π.
−∞
Find the constant of proportionality in p(x), the mean, and the standard deviation.
2
Solution. Let p(x) = ke−ax . We fix the constant k by demanding normalization,
Z ∞ Z ∞
−ax2 k 2
ke dx = √ e−u du = 1.
−∞ −∞ a
Using the provided integral, we conclude

r
a
k= .
π
The mean is clearly zero, since the distribution is symmetric about that point. Thus, we have
r Z ∞ Z ∞
a 2 −ax2 1 2
2
var X = hX i = x e dx = √ u2 e−u du.
π −∞ a π −∞
This remaining integral can be evaluated using integration by parts,

√
1 −u2 ∞
Z Z
2 −u2 1 −u2 π
u e du = − ue + e du = 0 +
2 −∞ 2 2
from which we conclude √
1 π 1
var X = √ , σ=√ .
a π 2 2a
3
[2] Problem 5. If two random variables X1 and X2 are independent, then
hX1 X2 i = hX1 ihX2 i.
Use this result to show that
var(X1 + X2 ) = var(X1 ) + var(X2 )
which implies that the standard deviation “adds in quadrature”,

q
σX1 +X2 = σX 2 + σ2 .
1 X2
This is an important result we’ll use many times below.

Solution. By definition, we have
var(X1 + X2 ) = h(X1 + X2 )2 i − hX1 + X2 i2
Using the properties listed in problem 3,
var(X1 + X2 ) = hX12 i + 2hX1 X2 i + hX2 i2 − hX1 i2 − 2hX1 ihX2 i − hX2 i2

= var(X1 ) + var(X2 ) + 2(hX1 X2 i − hX1 ihX2 i)
When X1 and X2 are independent, the last term vanishes, giving
var(X1 + X2 ) = var(X1 ) + var(X2 ).
2 Uncertainty Analysis
Idea 2
When a physical quantity is measured in an experiment and reported as x±∆x, it is uncertain
what the true value of the quantity is. If the quantity has a probability distribution p(x),
then the reported uncertainty ∆x is essentially the standard deviation of p(x).
Remark
In practice, you’ll have to use intuition and experience to assign uncertainties for real mea-
surements. For example, if you’re using a clock that times only to the nearest second, you
might take ∆t = 0.5 s. If you’re using a good ruler, which has millimeter markings, you might
take ∆x = 0.5 mm, though you can actually do a bit better if you look carefully. Of course,
the ultimate test is the results: if you assigned the uncertainties right, your final uncertainty
should encompass the true result most (but not all) of the time.
[2] Problem 6. Suppose x has uncertainty ∆x and y has uncertainty ∆y, where x and y are indepen-
dent. Explain why the uncertainty of x + y is
p
∆(x + y) = (∆x)2 + (∆y)2 .
This is called “addition in quadrature”. What is the uncertainty of x − y? How about x + x?
4
Solution. For independent variables, var(X

p 1 + X2 ) = var(X1 ) + var(X2 ). Since our uncertainties
represent the standard deviation, σX = var(X), we have
p
∆(x + y) = (∆x)2 + (∆y)2 .
For x − y = x + (−y), and ∆(−y) = ∆y, we get that ∆(x − y) = ∆(x + y). Finally, by linearity we
clearly have ∆(x + x) = 2∆x. (The formula above doesn’t apply, because x isn’t independent of x.)
Remark
Note how this differs from “high school” uncertainty analysis. In school, you might be told
to show uncertainty using significant figures, and when adding two things, to keep only the
figures that are significant in both of them. That corresponds to
∆(x + y) = max(∆x, ∆y)
which is an underestimate. Or, you might be told that the uncertainty needs to encapsulate
all the possible values, which implies that
∆(x + y) = ∆x + ∆y
which is an overestimate, since the uncertainties could cancel.
Three students make measurements of the length of a 1.50 m rod. Each student reports an
uncertainty estimate representing an independent random error applicable to the measure-
ment.
• Alice performs a single measurement using a 2.0 m tape measure, to within 2 mm.
• Bob performs two measurements using a wooden meter stick, each to within 2 mm, which
he adds together.
• Christina performs two measurements using a machinist’s meter rule, each to within
1 mm, which she adds together.
Rank the measurements in order of their uncertainty.
Solution
√
The uncertainty in Alice’s measurement is 2 mm. The √ uncertainty in Bob’s is 2 2 mm by
quadrature, while the uncertainty in Christina’s is 2 mm by quadrature. So the lowest
uncertainty is Christina’s, followed by Alice’s, followed by Bob’s.
[2] Problem 7. Given N independent measurements of the same quantity with the same uncertainty, √
xi ± ∆x, find the uncertainty of their sum. Hence show the uncertainty of their average is ∆x/ N .
This result is extremely important, since repeating trials is one of the main ways to reduce
uncertainty. But it’s important to remember that the results derived above hold only for independent
5
measurements. For example, taking a single measurement, then averaging that single number with
itself 100 times certainly wouldn’t reduce the uncertainty at all!
Solution. The uncertainty of their sum ∆X can be found by adding in quadrature,
X
∆X 2 = ∆x2 = N ∆x2
i
√
which implies ∆X = ∆x/ N .
Idea 3
If x has uncertainty ∆x, and f (x) can be approximated by its tangent line, f (x0 ) ≈ f (x) +
(x0 − x)f 0 (x) within the region x ± ∆x, then f (x) has approximate uncertainty f 0 (x) ∆x.
√
[2] Problem 8. If x has uncertainty ∆x, find the uncertainties of x2 , x, 1/x, 1/x4 , log x, and ex .
Solution. Differentiate the functions and multiply by ∆x to find the uncertainties. The sign isn’t
important, since uncertainties are always positive. The results are:
√ ∆x ∆x
∆(x2 ) = 2x∆x ∆( x) = √ ∆(1/x) =
2 x x2
4∆x ∆x
∆(1/x4 ) = ∆(log(x)) = ∆(ex ) = ex ∆x
x5 x
[2] Problem 9. The tangent line approximation doesn’t always make sense. For example, suppose that
x is measured to be zero, up to uncertainty ∆x. Show that the above results for the uncertainties
√
of x2 and x give nonsensical results. What would be a more reasonable uncertainty to report?
√
Solution. The above uncertainties give 0, ∞ for the uncertainties of x2 and x respectively. Since
x, 2 2
the uncertainties were found
2 2
√ with ∆x √ √ now with x ∆x, we can get (x + ∆x) − x =2
2x∆x √+ ∆x ≈ ∆x and x + ∆x − x ≈ ∆x. Thus, more reasonable uncertainties are (∆x)
and ∆x. There are numerical factors of order 1 because the shapes of the probability distributions
will be distorted, but we won’t worry about those, because we’re just looking to get a reasonable
result. (Of course, a professional would keep track of all these details.)
[2] Problem 10. Consider two quantities with independent uncertainties, x ± ∆x and y ± ∆y.
(a) Show that the uncertainty of xy is
s 2 2
∆x ∆y
∆(xy) = xy + .
x y
To do this, start by writing xy as exp(log x + log y).
(b) If we set x = y, then we find

s
∆x 2 √

2 2
∆(x ) = x 2 = 2x∆x.
x
On the other hand, in a previous problem we found ∆(x2 ) = 2x∆x. Which result is correct?
6
(c) Find the uncertainty of x/y.
Solution. (a) We can write

xy = exp(log x + log y)
which implies
s 2 2
∆x ∆y
∆(xy) = exp(log x + log y)∆(log x + log y) = xy + .
x y
(b) The result that ∆(x2 ) = 2x∆x is correct, since the formula for ∆(xy) assumes x, y are
independent, which fails when we set y = x.
(c) We have
x
= exp(log x − log y)
y
and by a very similar calculation to part (a), we conclude
s
∆x 2 ∆y 2

x
∆(x/y) = + .
y x y
[2] Problem 11. A student launches a projectile with speed v = 5 ± 0.1 m/s in gravitational accelera-
tion g = 9.81 ± 0.01 m/s2 . The resulting range is d = 1.5 ± 0.02 m. Given that the launch angle was
less than 45◦ , find the launch angle, with uncertainty, assuming all uncertainties are independent.
Solution. From the projectile range equation d = v 2 sin(2θ)/g, we get

1 dg
θ = arcsin = 18.03◦ .
2 v2
To find the uncertainty, we write sin(2θ) = gd/v 2 . The left-hand side is
2 cos(2θ)∆θ
by the tangent line approximation. By the result of problem 10, the right-hand side is
s
∆d 2 ∆g 2 2∆v 2

dg
+ + = 0.0248
v2 d g v
Combining the results, we have

∆θ = 0.015 rad = 0.9◦
which means the final result should be written as
θ = 18.0◦ ± 0.9◦
where we removed a superfluous significant figure.
[2] Problem 12. Two physical quantities are related by y = xex .
(a) If x is measured to be 1.0 ± 0.1, find the resulting value of y, with uncertainty.
7
(b) If y is measured to be 2.0 ± 0.1, find the resulting value of x, with uncertainty.
Solution. (a) To find the central value of y, we plug in to get y = e = 2.7183. To find the error,
we use the tangent line approximation,
dy
= ex (x + 1)
dx
which gives us
∆y ≈ ex (x + 1)∆x = 0.54.
Thus, rounding to a reasonable number of significant figures, we have
y = 2.7 ± 0.5.
Note that it would be incorrect to apply the “addition in quadrature” rule for products,
s
∆x 2 ∆(ex ) 2

∆y = xex +
x ex
because x and ex aren’t independent.
(b) To find the central value of x, we solve the equation 2 = xe−x numerically. This can be done
using the method of iteration introduced in P1. That is, we have x = 2e−x , so by repeatedly
plugging 2e−Ans into the calculator, we get x = 0.8526.
Under the tangent line approximation,
∆y
∆x ≈ = 0.023.
ex (x + 1)
Rounding to a reasonable number of significant figures, we conclude
x = 0.85 ± 0.02.
Idea 4
For practical computations, it is often useful to use relative uncertainties. The relative
uncertainty of x is ∆x/x, and can be expressed as a percentage.
[1] Problem 13. Some basic relative uncertainty results.
(a) Show that the relative uncertainty of the product or quotient of two quantities with independent
uncertainties is the square root of the sum of the squares of their relative uncertainties.
(b) Show that averaging

√ N independent trials as in problem 7 reduces the relative uncertainty by
a factor of N .
Solution. (a) Above we found that

s 2 2
∆x ∆y
∆(xy) = xy +
x y
8
Dividing both sides by xy gives

s 2 2
∆(xy) ∆x ∆y
= +
xy x y
which is the desired result.
√ √
(b) We have ∆xN = ∆(Σx)/N = ∆x N /N = ∆x/ N . Then
∆xN ∆x 1
= √
x x N
as expected.
Example 3
The side lengths of a rectangular prism are measured to be 1.0 ± 0.1 cm, 5.0 ± 0.1 cm, and
10.0 ± 0.1 cm. What is the relative uncertainty of the volume?
Solution
We note that the relative uncertainties of the side lengths are 10%, 2%, and 1%. Hence the
relative uncertainty of the volume is
∆V p
= (10%)2 + (2%)2 + (1%)2 ≈ 10%.
V
Since one of the uncertainties is so much larger than the others, it is basically the only one
which matters. This is also a common situation when doing real experiments. Thus, in many
situations it’s possible to compute the uncertainty in your head.
[2] Problem 14. Suppose the goal of an experiment is to measure the ratio T1 /T2 of the durations of
two physical processes, where T1 is about 15 seconds, and T2 is about 3 seconds. Also suppose your
stopwatch is only accurate to the nearest second. You have two minutes to perform measurements.
Assume each measurement is independent.
(a) Using your instinct, figure out whether it’s better to spend more total time measuring T1 ,
more total time measuring T2 , or an equal amount of time on both.
(b) To confirm this, qualitatively sketch the relative uncertainty of T1 /T2 as a function of the
fraction of time x spent measuring T1 , using explicit numeric examples if necessary.
Calculations of this sort are common when doing Olympiad experimental physics. You should be
able to do them instinctively, getting the ballpark right answer without explicit calculation.
Solution. (a) Since T2 is smaller, a single measurement of T2 has a much higher relative error.
Furthermore, T2 takes less time to measure. This means we definitely want more distinct
measurements of T2 than of T1 . As for how we split up the time, this is a bit harder to judge,
but intuitively because error adds in quadrature, taking a single measurement of each makes
T2 ’s uncertainty not 5 times as bad, but 25 times as bad. So T2 really completely dominates
the error here, and we should spent most of our time getting its error down.
9
√
(b) We have ∆T ≈ 1 s and ∆Ti = ∆T / Ni , giving
s 2 2
T1 ∆T ∆T
∆(T1 /T2 ) = √ + √
T2 T1 N1 T2 N2
The total time is constant, N1 T1 + N2 T2 = Tt where N1 T1 /Tt = x. We want to minimize

1 1
f (x) = + .
T12 x (1 − x)T22
The derivative is
1 1
f 0 (x) = − +
T12 x2 T22 (q − x)2
and setting this to zero gives
x2 (1 − T12 /T22 ) − 2x + 1 = 0.
The smaller root is the desired one since x < 1, giving
1 − (T1 /T2 ) 1
x= 2 2 =
1 − (T1 /T2 ) 1 + (T1 /T2 )
so we should spend 1/6 of our time measuring T1 . The graph of the uncertainty as a function
01^
of x looks like a U shape due to the two asymptotes at x = 0 and x = 1.
[3] Problem 15. Solve F = ma 2018 problems A12, A25, B19, and B25, and F = ma 2019
problems A16, B18, and B25. Make sure to strictly adhere to the total time. Since these are
F = ma problems, you don’t have to produce a writeup. If you find these questions difficult to
finish in the allotted time, go back and review the earlier material!
[3] Problem 16. In the preliminary problem set, you measured g using a pendulum. If you didn’t do
uncertainty analysis for it, as we covered above, then you should go back and estimate uncertainties
more precisely. In this problem you’ll do a different experiment: you will estimate g by finding the
time needed for an object to roll down a ramp, with everything again made of household materials.
(a) Before starting, think about what the dominant sources of uncertainty will be, and how you
can design the experiment to minimize them. In particular, do you think the result will be
more or less precise than your pendulum experiment?
(b) Perform the experiment, taking at least ten independent measurements, and report the data
and results with uncertainty.
Solution. Our formula for g is

2`(1 + β)
g=
t2 sin(θ)
where β = I/M R2 of the rolling object, and `, t are the distance and time for the path. Thus,
s
2∆t 2
2
cos(θ)∆θ 2

∆`
∆g = g + + .
t ` sin(θ)
10
To prevent slipping and to make t smaller, θ should be small for gentle rolling so tan θ ≈ θ. We
can estimate ∆θ ≈ 1◦ , θ around 5-10◦ , ∆t ≈ 0.1 s, t from 2-10 seconds, ∆` ≈ 1 mm, ` around 1
meter, and if we find a nice enough object ∆β ≈ 0. We can see that most of the error will probably
come from angle and time. Increasing angle will decrease time, and increasing time will need the
angle to decrease given the same length of ramp/rolling object available. To minimize these errors,
repeated measurements should be made with varying angles and recorded times.
The pendulum would probably have less error, since ∆t will be about the same and many
consecutive periods can be taken (resulting in ∆t ∼ 0.1s/N ) if the pendulum isn’t damped by a lot,
and the ∆θ term won’t exist (however the length will be much shorter).
[3] Problem 17. At any moment, a Geiger counter can click, indicating that it has detected a particle
of radiation. Suppose that there is an independent probability α dt of clicking at each infinitesimal
time interval dt. Let the number of clicks observed in a total time T be X.
(a) Find the expected value and standard deviation of X, and thereby compute its relative
uncertainty. (Hint: split the total time into many tiny time intervals, and let Xi be the
P
number of clicks in interval i, so X = i Xi .)
(b) Using a Geiger counter on a sample, you hear 197 clicks in 5 minutes of operation. Estimate
the activity α of the sample (i.e. the expected clicks per second), with uncertainty. If you
measure for longer, how does the uncertainty reduce over time?
(c) Now suppose that for a different sample, N = 0 after 5 minutes. Estimate the activity α of
the sample (i.e. the expected clicks per second), with a reasonable uncertainty. If you measure
for longer, and continue to hear no clicks, how does the uncertainty reduce over time?
Solution. (a) There are N = T /dt time intervals. Using the hint and applying linearity of
expectation, X
hXi = hXi i = N (α dt) = αT.
i
Since the Xi are independent, their variances add. The variance of Xi is
hXi2 i − hXi i2 = α dt − (α dt)2 = α dt.
Thus, by adding the variances, we have
var X = αT
√ √
so the standard deviation is ∆X = αT . The relative uncertainty is ∆X/hXi = 1/ αT .
(b) Applying the formulas above, we estimate

197
α= = 0.66 s−1
T
with an uncertainty of r
α α
∆α = √ = = 0.05 s−1 .
αT T
√
The uncertainty falls as 1/ T . Note that this is very similar to previous results we’ve found,
√
where the uncertainty falls as 1/ n where n is the number of trials. In some sense, each
instant of time we wait is another trial here.
11
(c) Of course, we estimate α = 0, but then the formulas above imply ∆α = 0 and hence that we
are absolutely certain α = 0, which is absurd. (If you don’t think that’s absurd, note that
the same result would have occurred if we had heard zero clicks in an arbitrarily short time
interval, such as a nanosecond.)
This is a case where the basic rules of error propagation break down, and we need to think.
The point of giving an uncertainty is to indicate the range of parameter values compatible
with the data we observed. Now, the probability of having no clicks in time T is e−αT . If
αT 1, then it would be very unlikely to have no clicks, so we can rule out α 1/T . But
if αT . 1, this isn’t unlikely at all. Thus, your uncertainty window should be α ∈ [0, c/T ]
where c is an order-one number, whose value depends √ on the specific statistical procedure you
use. (Note that the upper bound falls as 1/T , not 1/ T .)
Remark
In this problem set, we have given rules for calculating the mean and standard deviation
of derived quantities. But in general, probability distributions can have all kinds of weird
features, which aren’t captured by those two numbers. The reason we focus on them anyway
is because of the central limit theorem, which roughly states that if we have many independent
random variables, the distribution of the sum will approach a normal distribution. As you
saw in problem 4, normal distributions are characterized entirely by their mean and standard
deviation, so we don’t lose any information by reporting only those two quantities.
[4] Problem 18. [A] This problem extends problem 17 to derive some canonical results.
(a) Let λ = αT . Find the probability p(X = k) of hearing exactly k clicks in terms of λ and k.
(b) To check your result, show that the sum of the p(X = k) is equal to one.
(c) ? In the limit λ 1, show that the probabilities p(X = k) approach that of a normal
distribution with the mean and standard deviation calculated in problem 17, thereby providing
an example of the central limit theorem at work. This is a rather involved calculation, which
will use many of the techniques from P1. It will also require Stirling’s approximation,
√ n n
n! ≈ 2πn
e
for n 1, which will be important in T2. (Hint: because the relative uncertainty falls as λ
increases, start by writing k = λ(1 + δ) for |δ| 1, and expand in powers of δ. Be careful not
to drop too many terms, as δ is small, but λδ isn’t.)
P
Solution. (a) Following the notation of problem 17, we have X = i Xi , and we get k clicks if
precisely k of the Xi are equal to 1. Thus,
Nk λk −λ

N
p(X = k) = (α dt)k (1 − α dt)N −k ≈ (α dt)k (1 − α dt)N = e .
k k! k!
This is known as the Poisson distribution.
(b) This follows from the Taylor series of the exponential,
∞ ∞
X
−λ
X λk
p(X = k) = e = 1.
k!
k=0 k=0
12
(c) Using Stirling’s approximation, we have

k
1 λe
p(X = k) = √ e−λ
2πk k
λ(1+δ)
1 e
=p e−λ
2πλ(1 + δ) 1+δ
1
≈√ eδλ (1 + δ)−λ(1+δ)
2πλ
where we used the fact that δ 1.
Now we need to use a technique from P1. Letting the final term be equal to 1/y, we have
δ2 δ2λ

3
log y = λ(1 + δ) log(1 + δ) = λ(1 + δ) δ − + O(δ ) = δλ + + O(δ 3 λ).
2 2
In P1, we only expanded up to the first term, but here we need to keep the order δ 2 term.
The reason is we want an approximation that works for √the whole peak of the probability
distribution,
√ and we know it has relative uncertainty 1/ λ, which means we need to take
δ ∼ 1/ λ. That implies that δ 2 λ is of order one and cannot be dropped, but δ 3 λ is small and
can be dropped. Anyway, plugging this in, we find
1 2 1 2
p(X = k) ≈ √ e−δ λ/2 = √ e−(k−λ) /2λ
2πλ 2πλ
which is precisely a normal distribution with the appropriate mean and standard deviation.
[3] Problem 19. [A] Consider N independent measurements of the same quantity, with results xi ±∆xi .
They can be combined into a single result by taking a weighted average. What is the optimal weighted
average, which minimizes the uncertainty?
Solution. Let the weights be wi , so we report the value

X
x= wi x i .
i
The uncertainty obeys X

(∆x)2 = wi2 (∆xi )2 .
i
A tempting but incorrect way to minimize this quantity to set the derivative with respect to wi
equal to zero. This doesn’t work because the solution is just w1 = . . . = wN = 0, which isn’t a
weighted average at all. To actually have a weighted average, we need the weights to sum to one,
X
wi = 1.
i
This is an optimization problem with a constraint, which can be solved with Lagrange multipliers.
However, for this particular problem, the constraint is simple enough to handle manually. Because
of the constraint, if one increases some weight, then one must decrease others. At the minimum,
the effect of increasing any weight infinitesimally and decreasing another the same amount must be
13
zero, as if it weren’t, we could just adjust those two weights to get a lower uncertainty. Setting the
change in the uncertainty due to adjusting wi and wj in this way to zero gives
0 = d(wi2 )(∆xi )2 + d(wj2 )(∆xj )2 = (2 dw)(−wi (∆xi )2 + wj (∆xj )2 ).
This tells us that wi ∝ 1/∆x2i , which means
1/(∆xi )2
wi = P 2
.
j 1/(∆xj )
Note that all measurements are included in the optimal average, no matter how bad they may be.
Remark
There are many situations where the rules above can’t be used. For example, consider the
uncertainty of x + y 2 /x, where x and y have independent uncertainties. You can calculate the
uncertainty of either term with the standard rules, but you can’t calculate the uncertainty
of their sum, because the terms are not independent (both contain x).
In these cases, you can use the multivariable equivalent of the tangent line approximation,
∂f ∂f
f (x0 , y 0 ) ≈ f (x, y) + (x0 − x) + (y 0 − y) .
∂x ∂y
Adding the two contributions to the uncertainty in quadrature gives
s 2 2
∂f ∂f
∆f = ∆x + ∆y .
∂x ∂y
This is the general rule that includes the rules you derived above as special cases. However, it
shouldn’t be necessary in Olympiad problems. For example, if you run into such situations in
an experiment, often one of the uncertainties is much smaller, and can be neglected entirely.
Remark
As you saw in problem 9, the tangent line approximation can sometimes fail. The proper way
to handle situations like these would be to find the full probability distribution of the desired
quantity, rather than just describing it crudely with its standard deviation. However, this
can’t be done analytically except in the simplest of cases. So when professional physicists run
into situations like these, which are quite common, they often just numerically compute a few
million or billion values, starting with randomly drawn inputs each time, and use that to infer
the probability distribution. This technique is called Monte Carlo. It’s very powerful, but
certainly not needed for Olympiads! On Olympiads, you should just fall back to something
reasonable, such as taking the minimum and maximum possible values.
14
3 Data Analysis
Idea 5
All graphical data analysis for the USAPhO and IPhO can be performed by drawing a line
and measuring its slope and intercept. This is a bit artificial, but it’s necessary because of
the limited calculation equipment you have during these exams. Despite this, drawing lines
can be surprisingly powerful.
Example 4
The activity of a radioactive substance obeys A(t) = A0 e−t/τ . Using measurements of t and
A(t), plot a line to find A0 and τ .
Solution
To handle exponential relationships, take the logarithm of both sides for
log A(t) = log A0 − t/τ.
Then a plot of log A(t) vs. t has slope −1/τ and y-intercept log A0 .
[1] Problem 20. For a power law y = αxn where y and x are measured, what line can be plotted to
find α and n?
Solution. We have
log(y) = log(αxn ) = log(α) + n log(x).
Thus, if we plot log y against log x, the slope will be n and the y-intercept will be log(α).
[2] Problem 21. The rate R of electron emission from a solid in an electric field E is
R = βe−E/E0
for some constants β and E0 . The particular form is because the effect is due to quantum tunneling,
and you will derive it in X2.
(a) If E and R are measured, what line can be plotted to find β and E0 ?
(b) Your answer for part (a) should have formally incorrect dimensions, by the standards of P1.
This often happens when one takes logarithms. What’s going on? If the dimensions are wrong,
how can the result be right?
(c) Suppose both β and E0 have 1% uncertainty. For small E, which is more important for the
uncertainty of R? What about for large E? Around where is the crossover point?
Solution. (a) Take the natural log of the equation to get
E
log R = − + log β.
E0
Plotting E on the x-axis and log R on the y-axis will give a line with slope − E10 and y-intercept
log β.
15
(b) This gets into the details of what it even means to plot data. As a simpler case, consider the
relationship y = kx where y and x both have units of energy. We can plot y versus x to find
the slope k, but in reality, you can’t actually plot a dimensionful quantity: what would it
even mean to move your pencil a distance of “3.7 J” on a page? Instead, we write y and x as
dimensionless multiples of a standard unit of energy. That is, we are actually plotting
y x
=k
E0 E0
where E0 is some unit of energy, which is typically 1 J. But we don’t bother to write E0
explicitly because this step is kind of obvious.
Exactly the same thing is going on in this problem, but it looks strange because logarithms
have the property log(xy) = log x + log y. Both R and β have units of rate, so define a unit
of rate R0 and subtract log R0 from both sides to get an equation with correct dimensions,
R E β
log =− + log .
R0 E0 R0
This reflects what we actually do when constructing a log plot, though it is usually left implicit.
(c) The uncertainty in β alone always gives a 1% uncertainty in R. But the uncertainty in R
due to the uncertainty in E0 depends on the value of E. For E E0 , we can expand the
exponential as (1 − E/E0 ), and in this case the uncertainty in E0 does almost nothing at all,
so the uncertainty in β dominates. For E E0 , the reverse is true. By dimensional analysis,
the crossover must be around E ∼ E0 .
Example 5
Suppose that y and x are related nonlinearly, as
y = bx + ax2 .
For example, this could model the force due to a non-Hookean spring. Using measurements
of x and y, plot a line to find a and b.
Solution
If we divide by x, we find
y
= ax + b.
x
Therefore, we can plot y/x versus x, which gives a line with slope a and intercept b. More
generally, we can plot a line whenever we can rearrange a given relation into the form
(known) = (unknown)(known) + (unknown)
where all four terms can be arbitrarily complicated. In this way, it is possible to turn a lot
of very nonlinear relations into lines.
[3] Problem 22. Some more examples of finding lines to plot.
16
(a) Suppose that you are given points (x, y) that lie on a circle centered at (a, 0) with radius r.
What line can be plotted to find a and r?
(b) Consider an Atwood’s machine with masses m and M > m. The acceleration of the machine
is measured as a function of M . However, since the pulley has mass, it slows the acceleration
of the Atwood’s machine, so that
M −m
a= g.
M + m + δm
Find a line that can be plotted to find g and δm, assuming m, M , and a are known. This is
an example of how plotting a line can separate out a systematic error, i.e. the value of δm,
which would be impossible if only one value of M were used.
(c) Suppose an object is undergoing simple harmonic motion with amplitude A and angular
frequency ω. Given measurements of the position x and velocity v, what line can be plotted
to find A and ω?
Solution. (a) The equation of the circle is
(x − a)2 + y 2 = r2 , y 2 + x2 = 2ax + r2 − a2
Plotting y 2 + x2 vs. x will give a slope of 2a and a y-intercept of r2 − a2 . Combining the two
pieces of information yields a and r.
(b) The equation can be slightly rearranged to give
M −m M + m δm
= + .
a g g
Therefore, a plot of (M − m)/a vs. M + m has slope 1/g and y-intercept δm/g.
(c) By conservation of energy, A2 = x2 + v 2 /ω 2 , so
x2 = A2 − v 2 /ω 2 .
Thus, a plot of x2 vs. v 2 has y-intercept A2 and slope −1/ω 2 .
Remark
When performing data analysis in practice, you should neatly organize your work. Always
make a data table that explicitly shows what you’re calculating, and make a neat graph with
a ruler and graph paper. Set the axis scale so that the graphed data points cover almost
the entire page, and let the x-axis include x = 0 if you need to find a y-intercept. The
computation of the slope should be explicitly shown. For each line you should use at least
about five points; you don’t have to use them all. If you have a calculator that can find best-fit
slopes for you, don’t use it, as these features are generally not allowed on real Olympiads.
01W
01W
[3] Problem 23. USAPhO 2012, problem A2. (This one requires basic thermodynamics.)
01c
[3] Problem 25. INPhO 2018, problem 7. (This one requires basic fluid dynamics.)
Solution. See the official solutions here.
17
Idea 6
Historically, uncertainty analysis has only appeared on the F = ma, and data analysis has
only appeared on the USAPhO, but the two appear together in the IPhO.
To perform uncertainty analysis for best fit lines, plot the uncertainties of the data points
as error bars. Then draw the steepest and shallowest lines that still pass through most of
the error bars. These will give you the bounds on your slope and intercept. We’ll see some
examples of this procedure in later problem sets. It isn’t the most mathematically rigorous
method, but it gives decent results.
4 Estimation
Estimation is a useful skill for checking the answers to real-world problems.
Example 6
Estimate the circumference of the Earth.
Solution
If you know that the United States is 3,000 miles wide, and there is a time zone difference of
three hours between California and New York, then a reasonable estimate is 24,000 miles.
Or, if you know the factoid that light can go about seven times around the Earth in a second,
then a reasonable estimate is (3/7) × 108 m ≈ 4 × 107 m.
Let’s check these results are compatible. There are about 5 miles in 8 kilometers, a fact
you can get by remembering how your car’s speedometer looks, or by noting that 3 feet are
about 1 meter. Then 4 × 104 km ≈ (5/8) × 4 × 104 mi = 2.5 × 104 mi, so the two results are
compatible. There are probably at least a hundred more ways to perform this estimation.
Example 7
Estimate the density of air, and compare this to the density of water.
Solution
We can directly use the ideal gas law, P V = nRT . The density is ρ = µn/V where µ is the
mass of one mole of air, so
µP
ρ= .
RT
Atmospheric pressure is about 105 Pa, typical temperatures are about 300 K, and air is mostly
N2 , which has a molar mass of µ = 28 g/mol, so
(0.028)(105 ) kg kg
ρ= 3
≈ 1 3.
(8.3)(300) m m
18
The density of water is, almost by definition,

kg
ρw ≈ 103 .
m3
Most liquids and solids have densities within an order of magnitude of this, since in all
cases the atoms are packed close together. Evidently, air molecules are about a factor of
(103 )1/3 = 10 times further apart than typical water molecules.
Example 8
Estimate how much useful power you can produce in a short burst.
Solution
This is a bit tricky to test, because most exercises just burn energy against air resistance or
friction, which is hard to estimate. However, a task that directly performs work is useful. I
weigh about 75 kg and can run up a 3 m high staircase in around 3 s, so
P = mgv = (75)(10)(3/3) W ≈ 750 W.
This is a typical max power output, while typical steady state power outputs are about a
factor of 3 or 4 smaller.
For the below questions, feel free to look up specific numbers if you’re stuck. In all cases, an answer
to the nearest order of magnitude is good enough.
[3] Problem 26. Some questions about light energy.
(a) Estimate the number of photons emitted per second by a standard light bulb. (The energy of
a photon is E = hf , and the frequency of a photon is related to the wavelength by c = f λ.)
(b) The Sun supplies power of intensity 1400 W/m2 to the Earth. The nearest star is about 4
light years away. Assuming this star is similar to the Sun, about how many of its photons hit
your eye per second?
Solution. Before we continue, it’s important to note that for estimation questions, one should only
expect an answer to within an order of magnitude. Some teachers tweak their example calculations
until they give almost exactly the right answer. This makes them look brilliant, but it’s deceptive,
because then when the student tries to do the same, their results will be much further off. So to
combat this, in all solutions here, we’ve just presented our very first, simplest guesses. They can be
up to an order of magnitude off from the real numbers, so if your numbers are within two orders of
magnitude of ours, you’re fine!
(a) We can estimate a standard light bulb to have around 50 W of power. The power P = N E
where N is the number of photons emitted per second, and the wavelength of visible light is
from 400 − 700 nm (we can use 500). Then
Pλ
N= ≈ 1020 photons/s
hc
19
(b) 1 AU is about 1.5 × 1011 m. (If you forget, you can use something like GMS /r2 = (2π/T )2 r,
where T is one year and MS ≈ 2 × 1030 kg). 1 light year is c × 1 year ≈ 9.5 × 1015 m. Then
the intensity from the star is reduced by a factor of (1 AU/4 ly)2 due to the inverse square
law, so I ≈ 3.5 × 10−7 W/m2 . The radius of our eyes is about 1 cm, so the area is about 12 πr2
since our eyes aren’t that open. Then the number of photons that hit per second is P λ/hc,
which gives N ≈ 108 photons/s.
[2] Problem 27. Estimate the radius of the largest asteroid you could jump off of, and never return.
p
Solution. The escape velocity is v = 2GM/R, and we will assume a uniform spherical asteroid
with density ρ. Rock is probably a few times denser than water, so ρ ≈ 3 × 103 kg/m3 and
√
M ≈ 43 πρR3 . Humans can jump around half a meter, which determines v = 2gh. Thus
2G 4
2gh = πρR3 .
R 3
Since g ≈ π 2 in SI units, this simplifies to
s
3πh
R≈ ≈ 2 km.
4Gρ
[4] Problem 28. Some questions about energy.
(a) Estimate the digestible energy content of a stick of butter. (A calorie is about 4000 J, and is
also the energy needed to raise the temperature of a kilogram of water by 1 K.)
(b) Estimate the rate at which your body burns energy when at rest.
(c) Estimate the rate at which a human being radiates energy. (The Stefan–Boltzmann law states
that the radiation power per unit area from a blackbody is σT 4 , where σ = 5.7×10−8 W/m2 K4 .)
Is radiation a significant source of energy loss for a human being, or is it negligible?
(d) A human being develops hypothermia, with their core body temperature dropping by 5 ◦ F.
Neglecting any heat transfer with the environment, estimate the number of calories required
to raise their temperature back to normal.
Now let’s verify the energy content of the butter microscopically. This will be a very rough estimate,
so expect answers to be only within one or two orders of magnitude.
(e) A chemical bond typically involves two electrons, and a characteristic atomic separation
distance of one angstrom, r ∼ 10−10 m. Estimate the binding energy of one chemical bond.
(f) The fats in butter are digested by inputting energy to break the bonds in the molecules, then
harvesting energy by combining the atoms into CO2 and H2 O, which have somewhat more
stable bonds.
Estimate the energy content of a kilogram of butter. How close is this to the true result?
20
Solution. (a) Recall the usual “2000 calories per day diet” you see on the nutrition facts for
food. Note that those calories are referring to kilocalories (∼ 4000 J). Eating a few sticks of
butter will probably make me feel quite full for a day (and disgusted), so a stick of butter
probably has around 500-1000 kcal of digestive energy content (let’s use 800, which is close
to the actual value). Then E = 800 kcal × 4000 J/kcal ≈ 3 × 106 J.
(b) Again, we will use what we see on the nutrition facts: 2000 cal/day≈ 100 W. This energy is
used to maintain homeostasis in your body, and it eventually gets exhausted as heat.
(c) First we approximate the surface area of a human, then assume a spherical human that’s a
perfect blackbody. Our height is about 1.7 m, and our width is around 0.25 m and negligible
thickness. Then the surface area is around 2 ∗ 1.7 ∗ 0.25 ≈ 1 m2 (rounding up makes more
sense for thickness and limbs). Then using the Stefan–Boltzmann law, P = AσT 4 . Humans
skin is on the order of 300 K, so P ≈ 500 W.
This is much too high, as it can’t possible be higher than (b). The main difference is that the
radiation output by the human body is almost completely cancelled by the radiation input by
the environment, which is at almost the same temperature (in absolute terms). For example, in
typical room-temperature conditions, the environment is at 70◦ F and human skin is at 90◦ F,
for a difference of about 10 K. So the power is smaller by a factor of 1 − (290/300)4 = 0.13,
giving a reasonable 65 W. It’s still a significant contribution, but not unreasonably large. Of
course, in colder environments one can reduce this contribution by, e.g. wearing clothes.
(d) 5◦ F is 10/9◦ C ≈ 1◦ C. Now we use Q = mcT , and since humans are mostly water, we’ll
approximate the specific heat to be the same as water. The mass of humans is usually around
60 kg. Since the ”food calorie” is a kilocalorie (amount of energy needed to raise 1 kg by 1◦
C), we need 60 food calories to raise our temperature back to normal.
(e) A basic estimate for the binding energy is
e2
E∼ ∼ 2 × 10−18 J.
4π0 r
As a check, this is about 10 eV, and the binding energy of hydrogen is about 13.6 eV (one of
those classic numbers you should remember), so this is in the right ballpark. Of course, the
energy is actually negative, even though electrons repel, because it’s due to how the electrons
are attracted to the nuclei. We can, however, very roughly estimate this negative energy using
the positive energy of repulsion e2 /4π0 r because all energy scales in the problem should be
roughly similar.
Actually, in reality the answer should be about an order of magnitude lower, for two reasons.
The first is simply that atomic separations are a bit bigger, but this is cancelled by the fact
that the nuclei have charge Zi > 1. The main issue is that covalent bonds are a bit more
subtle.
Naively, you could say that a covalent bond is attractive because the electrons in one atom are
attracted to the nuclei of the other. But this is too naive, because at least parametrically, it’s
cancelled out by the repulsion of the nuclei with each other, and the repulsion of the electrons
with each other, as all four of these terms are of order ±e2 /4π0 r. Covalent bonds are stable
because the electron orbitals can deform a bit, so that the negative contributions end up a
bit bigger than the positive ones. So e2 /4π0 r isn’t really an estimate for the binding energy,
21
but for the sizes of terms which mostly cancel out to give the binding energy, which is why
the real answer is about 10 times smaller.
(f) Fats are mostly carbon. As a very rough estimate let’s say that the carbon atoms end up in
bonds that are twice as stable as before, so the energy released per carbon atom is on the
order of magnitude of what we found in part (e). Then
−1 −1
energy energy C atoms kilograms −18 12 g
= ∼ (2 × 10 J) NA = 108 J/kg.
kilogram C atom mole mole mole
For comparison, the energy of one gram of fat is 9 calories, so the true answer is
J
(9)(4000)(1000) = 3.6 × 107 J/kg
kg
which is not too far off!
[2] Problem 29 (Povey). When human beings lose weight, most of it is by exhalation of carbon.
About 20% of the air in the atmosphere is oxygen. When we breathe in and then out, about 25%
of the oxygen is converted to carbon dioxide.
(a) Estimate the mass of air contained in a single breath.
(b) Estimate the amount of weight we lose every day by breathing alone.
Solution. (a) If I don’t take a deep breath, I can barely blow up a crushed plastic water bottle
(holds half a liter of volume), so I would estimate the volume in a single breath to be around
0.5 L. From chemistry class (or ideal gas law: n = P V /RT ), we know that mole of gas takes
up 22.4 liters of volume at STP (our body temperature, 310 K isn’t that much more than 273
K but we can just use 22.4 × 310/273 ≈ 25 L). Most of the air is nitrogen (N2 ) with molecular
mass 28 g/mol (oxygen, O2 , is 32 which is pretty close). Then one breath should have a mass
of 0.5 L/25 L/mol × 28 g/mol ≈ 0.6 g.
(b) By counting, we can estimate humans to breathe around 10 − 15 times a minute (let’s use
12.5), so around 20,000 breaths in a day. In each breath, 25% of the oxygen (which is 20%
of the air) will be around 5% of the air to converted from oxygen to carbon dioxide. Carbon
dioxide (CO2 ) has a molecular mass of 48 g/mol, and oxygen is 32. Thus we lose a proportion
of (48/32 − 1) × 0.05 = 0.5 = 0.025 of the mass of the air we breathe in every day, which is
about 0.3 kg. Most of the mass of the food we eat leaves this way.
[2] Problem 30 (Insight). How long a line can you write with a pencil?
Solution. Graphene, a layer of carbons arranged in a hexagonal way, famously can be made from
using scotch tape to extract a few layers of graphite from pencil markings. It’ll take plenty of tries
to erase pencil from paper with tape probably (but progress is definitely noticeable), so we can
estimate there to be around 100 layers of the hexagonal carbon from graphite.
We will assume that the line is drawn with the pencil perfectly vertical and the lead not sharpened.
The diameter of the lead is around 2 mm, and the mass of a pencil should be around 2-10 grams,
so the mass of the lead is on the order of 1 g. Assuming that the lead is almost all made out of
carbon, we can estimate how many carbon atoms it has, and the surface density of carbon atoms.
22
The carbons are spread apart in a hexagonal fashion with a characteristic distance of r ≈ 10−10
m, and the centers of 3 adjacent hexagons will have a carbon atom at its center, giving a spacing
of 1 carbon atom every r2 square meters. There should be
1
6.022 × 1023 atoms/mole × 1 g × ≈ 5 × 1022 atoms of C
12 g/mol
Thus that gives around 500 m2 of a single layer of carbon, so around 5 m2 of lead usage. The line
will be approximately a rectangle with area d`, where d is the diameter of 2 mm.
Thus the pencil line should be around 2.5 km long. One can find other estimates of the order
50 km, i.e. a spread of an order of magnitude. The precise result within this order of magnitude of
course depends on the details of the pencil.
5 Experimental Technique
At both the national and international Olympiad level, it’s important to have practical know-how
in order to make experiments work. It’s very hard to train this with only theoretical problems.
However, the Australian Physics Olympiad has some useful problems in this direction, since it has
01^
a strong emphasis on real-world physics.
01^
01Y
[3] Problem 33. AuPhO 2016, problem 14. You will need to print out pages 8 and 9 of the
answer sheets.
You can look at other AuPhO questions for further practice, but as you can see here, many AuPhO
questions are confusing, misleading, or even wrong, which is an unfortunate consequence of the
innovative nature of the contest. I’ll only assign you the best ones.
23
Mechanics I: Kinematics
See chapters 3 and 4 of Morin for material on solving differential equations. For general review on
kinematics, see chapter 1 of Kleppner and Kolenkow. For fun, see chapters I-1 through I-8 of the
Feynman lectures. There is a total of 87 points.
1 Motion in One Dimension

Example 1
When a projectile moves slowly through air, the drag is linear in the velocity, F = −αmv.
Find the velocity v(t) of a projectile thrown upward at time t = 0 with speed v0 .
Solution
We write Newton’s second law as
dv
= −g − αv
dt
and multiply through by dt. Integrating both sides from the initial condition to time tf gives
Z v(tf ) Z tf
dv
=− dt.
v0 g + αv 0
Performing the integrals gives

v(tf )
1
log(g + αv) = −tf .
α v0
Renaming tf to t and solving for v yields

g −αt
v(t) = e−αt v0 + (e − 1).
α
This renaming is necessary because we don’t want to confuse t, the dummy variable that we
integrating over, with tf , the time at which we want to evaluate the velocity; t ranges from
zero to tf . Unfortunately, often people just call both of these t, so you need to watch out.
[2] Problem 1. Investigating some features of this solution.
(a) By using results from P1, verify that v(t) makes sense for both small times and large times.
(b) If the projectile is then caught at the launch point, did it spend more time going up or down?
(c) Do you think the total time is longer or shorter than for a projectile without drag?
Solution. (a) For small times (αt 1), we have

g
v(t) ≈ (1 − αt)v0 + (−αt) = v0 − (g + αv0 )t
α
1
which makes sense, since it’s just the result of uniform acceleration g + αv0 , under the initial
net force. For large times (αt 1), the exponentials decay away and we get v(t) ≈ −g/α,
which is the terminal velocity.
(b) Let Pu be the path going up, and Pd be the path going down. Additionally, imagine the
video of Pu playing in reverse. By conservation of energy, the magnitude of the velocity of
the projectile in Pu will be greater than that of Pd at a given height y. Since dt = dy/v, and
at every value of y, |vu | ≥ |vd | by conservation of energy (and the bounds of integration are
the same), then the ball will spend less time going up.
(c) I’m just asking this so you can use your intuition. Intuitively, it feels like it should take longer
with drag, and for the linear drag this is indeed the case. More generally, if the drag force
is proportional to v n , then it turns out that the trajectory with drag always takes longer
for n ≥ 1, but for n < 1 it depends on the initial speed. You can find a proof of all these
statements here.
[3] Problem 2. Now assume quadratic drag, F = −αmv 2 , which applies for fast-moving projectiles.
(a) Integrate Newton’s second law to get an implicit equation for v(t) with the same initial
conditions as above. That is, you don’t need to solve for v(t), as it’ll just make things messy.
(b) Your equation will only be valid when the projectile is going up; explain why.
(c) Find v(t) for an object released from rest at time t = 0. (Hint: if needed, look up some standard
integrals involving hyperbolic trigonometric functions. But don’t worry about memorizing
the results, since in competitions, any nontrivial integral needed will usually be given to you.)
Some people only call this quadratic case drag; they call the linear case viscous resistance. This is
because they behave fundamentally differently at the microscopic level, as we will explore in M7.
Solution. (a) Newton’s second law is
dv
= −g − αv 2 .
dt
By the same reasoning as before, we find
dv 0
Z v(t) Z t
02
=− dt0 = −t.
v0 g + αv 0
By nondimensionalizing the integral as described in P1, the left-hand side is

Z v(t)√α/g r r
1 dx 1 −1 α −1 α
−t = √ √ 2
=√ tan v(t) − tan v0
αg v0 α/g 1 + x αg g g
√
where I pulled out a factor of 1/ αg to get the right overall dimensions, then used dimensional
analysis again to convert the integration bounds to dimensionless numbers. (You can also do
this by ordinary u-substitution if you prefer.) This is essentially the final result. It can be
solved for v(t), but that just makes it look worse.
(b) The reason the equation only makes sense when the projectile is going up is that the force
should always opposite the direction of motion, so we really wanted to solve F = −mα|v|v.
Equivalently, the sign of α changes when the direction of the velocity changes. This means
our solution really should have two separate cases.
2
(c) By the same reasoning, we have

v(t)
dv 0
Z
= −t
0 g − αv 02
where the changes are the initial condition and the sign of α. The left-hand side is
Z v(t)√α/g r
1 dx 1 −1 α
√ 2
=√ tanh v(t) .
αg 0 1−x αg g
If you don’t know this hyperbolic trig integral, you could also derive it by expanding 1/(1 − x2 )
in partial fractions and integrating each term. You will get a bunch of logarithms, which is
equivalent to the hyperbolic tangent. However, if you don’t know what the hyperbolic tangent
is, you should look it up now, because such functions will be useful later!
Because of the simpler initial condition, we can get an explicit solution,
r
g √
v(t) = − tanh( αg t).
α
p √
The speed approaches g/α with a timescale 1/ αg, a fact we could also have deduced by
physical intuition and dimensional analysis. Actually, another way to arrive at this result is
by just substituting α → −α in the answer for part (a)! This will produce the tangent of an
imaginary number, which is in fact how the hyperbolic tangent is defined.
[3] Problem 3. A projectile of mass m is dropped from a height h above the ground. It falls and
bounces elastically, experiencing the same quadratic drag as in problem 2. Find the maximum
height to which it subsequently rises. (Hint: don’t try to use your results from problem 2.)
Solution. The reason you shouldn’t try to use the results from problem 2 is that they are in terms
of time. Given how complicated the implicit expressions for v(t) are, the expressions for x(t) would
be extremely clunky. And they’re not necessary, because in this problem we don’t care about the
time-dependence at all; we just want to know the final height.
Another way to say this is that we aren’t interested in v(t), we’re interested in v(x). While
the projectile is moving downward, we can integrate dv/dx to find the speed v0 at the moment it
hits the ground. Then, when it’s moving upward, we integrate dv/dx until it has zero speed again,
which is its final height. This will be a lot simpler than integrating dv/dt.
For the upward and downward trajectories, Newton’s second law says
dv
= −g ± αv 2
dt
and multiplying both sides by dt/dx gives
dv g
= − ± αv.
dx v
Separating and integrating, on the way down we have
Z −v0
1 −v0 v dv
Z 0 Z
dv
dx = = .
h 0 αv − g/v α 0 v 2 − g/α
3
Carrying out the integral and simplifying,

1
h=− log(1 − αv02 /g).
2α
Now, on the way up, we have
Z h0 Z 0 Z v0
dv 1 v dv
dx = =
0 v0 −g/v − αv α 0 v2 + g/α
and carrying out the integral gives

1
h0 = log(1 + αv02 /g).
2α
Combining the two equations gives
1
h0 = log(2 − e−2αh )
2α
which you can check has the right limits. Also note that g drops out, as required by dimensional
analysis.
Example 2
Find how the speed of a rowing boat depends on the number of rowers N .
Solution
A fast-moving boat experiences quadratic friction, so a drag force
F ∝ v2A
where A is the submerged cross-sectional area of the boat. Since the submerged volume
scales as V ∝ N in hydrostatic equilibrium, we have A ∝ N 2/3 . (This is the sketchy step
of the analysis, since the scaling of A depends on how we adjust the shape of the boat as
N increases.) Thus, the power the rowers need to provide scales as P = F v ∝ v 3 N 2/3 , but
we also have P ∝ N . Combining gives the exceptionally weak dependence v ∝ N 1/9 , which
agrees decently with Olympic rowing times.
Idea 1
An ordinary differential equation is any equation involving a quantity x(t) and its derivatives.
In introductory physics, we are usually concerned with a few very simple differential equations,
with the following nice properties.
• The differential equation is at most second order, meaning that it can contain x, ẋ = v,
and ẍ = a, but no higher derivatives. This implies that the solution can be determined by
an initial position and initial velocity. (We’ll focus on second order differential equations
for the rest of this section; most first order differential equations can simply be solved
by separation and integration, as we’ve already seen above.)
4
• The differential equation is linear, meaning that terms don’t contain products of x, ẋ,
and ẍ. For example, a damped driven harmonic oscillator with time-dependent drag,
mẍ = −b(t)ẋ − kx + f (t)
is a second order linear differential equation. Solutions to such differential equations obey
the superposition principle: if x1 (t) and x2 (t) are both solutions, so is c1 x1 (t) + c2 x2 (t).
• The differential equation is homogeneous, meaning that each term is proportional to ex-
actly one power of x or its derivatives. The above differential equation is not homogeneous,
but it would be if we removed the driving f (t).
• The differential equation is time-translation invariant, meaning that no functions of time

appear except for x and its derivatives. The above equation isn’t, but it would be if we
set f (t) and b(t) to constants.
Idea 2
Linear, homogeneous, time-translation invariant differential equations are very special, and
they can all be solved by the exact same method. First, note that we can promote x(t) to
a complex variable x̃(t) and solve the differential equation over the complex numbers. As
long as we have a complex solution, we can recover a real solution by taking the real part.
Now, the method of solution, which works for almost all equations of this form, is to guess a
complex exponential solution
x̃(t) = eiωt .
Plugging this into the differential equation will yield the allowed values of ω, and the general
solution can be found by superposing the complex exponentials.
Example 3
Solve the simple harmonic oscillator, mẍ + kx = 0, using the above principles.
Solution
First, we pass to a complex differential equation,
¨ + kx̃ = 0.
mx̃
We guess x̃(t) = eiωt . Plugging this in and using the chain rule gives
m(iω)2 eiωt + keiωt = 0
and canceling eiωt and solving gives two solutions,
p
ω = ±ω0 , ω0 = k/m.
Since this a second-order linear differential equation, the general solution is given by the
superposition of these two complex exponentials,
x̃(t) = Aeiω0 t + Be−iω0 t
5
where A and B are general complex numbers. The real part of x̃(t) satisfies the original real
differential equation ma + kx = 0, and is
Re x(t) = C cos(ω0 t) + D sin(ω0 t)
where C and D are real numbers.
[1] Problem 4. To make sure you know how to go from the complex solution to the real one, write
C and D in terms of A and B.
Solution. Let A = aA + bA i and B = aB + bB i where ai , bi are real. Applying Euler’s formula,
Re x(t) = (aA + aB ) cos(ω0 t) + (−bA + bB ) sin(ω0 t)
from which we read off

C = Re(A + B), D = Im(B − A).
[2] Problem 5. Now introduce a damping force and solve the differential equation for the damped
harmonic oscillator, mẍ + bẋ + kx = 0, using the same procedure, assuming b is small. (See section
4.3 of Morin if you have trouble with this. We’ll consider this system in more detail in M4.)
Solution. Guessing an exponential, every time derivative yields a factor of iω, so
m(iω)2 + b(iω) + k = 0.
Using the quadratic formula, √

−ib ± 4km − b2
ω= .
−2m
In other words, we have
p ib
ω = ± k/m − b2 /4m2 + .
2m
The oscillation is slightly slowed down, as you might expect, and the frequency has an imaginary
part. This corresponds to exponential decay of the solution, by ei(ib)t/2m = e−bt/2m .

[3] Problem 7. Above, we mentioned that guessing an exponential works almost all the time. The
reason is because at the end of the day, the exponential cancels out and we’re left with a polynomial
in ω, which has just the right number of roots. But if there are repeated roots, there are fewer
distinct solutions for ω, and hence not enough solutions.
(a) Consider a second order differential equation with a double root ω. What is the other solution,
besides eiωt ? (Hint: to help find a good guess, consider the simple case ma = 0, where ω = 0
is the double root. Then generalize your guess to nonzero ω and check that it works.)
(b) This should be setting off alarm bells: the form of the solutions to the equation changes when
the two roots are exactly equal, while it’s just exponentials/sinusoids if the roots are different,
no matter how small the difference is. Since no two roots are ever exactly equal in practice, it
seems the behavior of part (a) can never actually happen in the real world. But it gets taught
in applied differential equations courses. Why?
6
(c) [A] Consider the most general nth order, linear homogeneous time-translation invariant differ-
ential equation
dn dn−1

d
an n + an−1 n−1 + . . . + a1 + a0 x = 0.
dt dt dt
What does the general solution look like?
Solution. (a) In the case of a double root ω = 0, the differential equation is ẍ = 0. The solution
we get by guessing an exponential is x(t) = ei(0)t = 1, which is a constant. The other solution
is linear, x(t) = tei(0)t = t. This leads us to guess that for a double root ω, the two independent
solutions are eiωt and teiωt .
(b) As two roots get closer and closer together, we can get solutions that look more and more like
(A+Bt)eiωt , which is what we would get if they were exactly the same. Of course, it’s intuitive
that we can get eiωt . To get a solution that looks like teiωt , note that for roots ω ± ∆ω,
ei(ω+∆ω)t − ei(ω−∆ω)t = eiωt (2i sin(∆ω t)) ∝ sin(∆ω t)eiωt .
This is an eiωt oscillation with a slowly varying envelope sin(∆ω t). For small times, t 1/∆ω,
the envelope is just proportional to t. As the roots get closer and closer together, this linear
behavior persists for longer and longer time, but nothing is ever discontinuous. One can see
this kind of envelope behavior in two weakly coupled pendulums, a system which has two
nearby oscillation frequencies. You’ll investigate this kind of thing in more detail in M4.
So the point is that when the roots are to each other, the solutions look like (A + Bt)eiωt for
times shorter than 1/∆ω. This is more direct and intuitive than superposing two sinusoids
with almost equal frequencies, so we use it in practice.
(c) Guessing eiωt gives

an (iω)n + an−1 (iω)n−1 + . . . + a0 = 0.
In the case where the roots are distinct, there are n possible values for ω, and hence n
parameters in our trial solution,
Xn
x(t) = Ai eiωi t .
i=1
Since the differential equation has order n, there are n parameters needed to specify the
solution, so this is the general solution. If ωi is a double root, then both eiωi t and teiωi t are
solutions. For a triple root, t2 eiωi t is also a solution, and so on.
Remark
You might be wondering how to solve more general differential equations. In M4, we will
consider three extensions of the above techniques. We’ll use the idea of normal modes
to solve systems of such differential equations, add driving forces to make the equations
inhomogeneous, and use the adiabatic theorem to approximately solve non time-translation
invariant equations where the coefficients change slowly in time.
Of course, this just scratches the surface of the subject, and solving more general differential
equations can be orders of magnitude harder. We won’t try to solve nonlinear differential
7
equations, as there is no general technique for doing so, and the answer is often an obscure
special function. (However, such equations will occasionally appear in later problems.) On the
other hand, linear differential equations with general time-dependence are more approachable,
and the following problem illustrates the most basic method for solving them.
[3] Problem 8. [A] Some linear, homogeneous, non time-translation invariant differential equations
can be solved by simply guessing a power series. For this problem, don’t worry about dimensional
analysis; assume all variables have already been redefined to be dimensionless.
(a) As a warmup, consider the differential equation ẋ = kx for constant k, which we already know
how to solve. By plugging in the ansatz
∞
X
x(t) = an tn
n=0
find the solution with x(0) = 1.
(b) Now consider the non time-translation invariant differential equation
t2 ẍ + tẋ + t2 x = 0
which is called Bessel’s differential equation of order zero. By using the same ansatz, find the
unique solution with x(0) = 1 and ẋ(0) = 0.
Solution. (a) Plugging the ansatz in gives

∞
X ∞
X
nan tn−1 = k an tn .
n=0 n=0
Shifting the sum on the left-hand side, we have

∞
X
(kan − (n + 1)an+1 )tn = 0.
n=0
For this quantity to be zero for all t, each term in the sum must individually be zero, so
k
an+1 = an .
n+1
The initial condition x(0) = 1 tells us that a0 = 1, from which we conclude
k2 k3
a1 = k, a2 = , a3 = ,...
2 6
or more generally,
∞
X kn
x(t) = tn = ekt
n!
n=0
which is just as expected.
8
(b) Plugging the ansatz in gives

∞
X
n(n − 1)an tn + nan tn + an tn+2 = 0.
n=0
Simplifying and shifting the sum as in part (a) gives

∞
X
(n2 an + an−2 )tn = 0.
n=0
We therefore have the recursion relation an = −an−2 /n2 . The initial conditions give a0 = 1
and a1 = 0, from which we conclude the a2n+1 are all zero. We then have
1 1 1
a2 = − , a4 = , a6 = − ,...
22 22 42 22 42 62
∞
(−1)m t 2m
X
x(t) = .
(m!)2 2
m=0
This function is known as the Bessel function of the first kind, of zeroth order, J0 (t).
2 Tricks
In this section we’ll consider some kinematics problems that require cleverness, not computation.
Idea 3
Many problems can be solved by a clever choice of reference frame. It is often useful to go to
the frame moving with one of the objects in the problem, or to go into a frame that makes
the motion in the problem more symmetric. For the purposes of kinematics it can even be
useful to use noninertial reference frames, such as a falling frame where projectiles don’t
accelerate, or a rotating frame, though this will introduce fictitious forces into the dynamics.
It is also useful to tilt the coordinate axes to be parallel to various objects.
[1] Problem 9 (KoMaL 2019). A cannon A is at the edge of a cliff with a 800 m drop. Cannon B is
on the ground below the cliff and 600 m horizontally away from it. Cannon A shoots a cannonball
directly towards cannon B at 60 m/s. Cannon B shoots a cannonball directly towards cannon A at
40 m/s. Will the two cannonballs hit each other in midair?
Solution. Work in the frame freely falling with the cannonballs. In this case, the balls have a
relative velocity of 100 m/s and initial separation of 1000 m, so it takes 10 s to collide. If there were
no gravity, this collision would occur at a point (2/5)(800 m) = 320 m above the ground. However,
because of gravity both balls have fallen by an extra gt2 /2 = 500 m by this time. Hence the balls
hit the ground before they can hit each other in midair.
[2] Problem 10 (Wang). Two particles are released in gravitational acceleration g with leftward and
rightward speeds v1 and v2 . Find the distance between them when their velocities are perpendicular.
9
Solution. After time t, the velocity vectors are (−v1 , −gt) and (v2 , −gt). These are perpendicular
when the dot product is zero, so v1 v2 = (gt)2 , which you can also show with basic geometry. Thus,
√
v1 v2
t= .
g
To compute the distance, we can just work in the frame falling with the masses. Then it’s clear
that the acceleration g doesn’t matter, and the distance is just
√
(v1 + v2 ) v1 v2
d = (v1 + v2 )t = .
g
[3] Problem 11 (Kalda). Two intersecting circles of radius r have centers a distance a apart. If one
circle moves towards the other with speed v, what is the speed of one of the points of intersection?
Solution. Work in the frame where the circles are moving towards each other with speed v/2.
Then by the Pythagorean theorem, the speed of the point of intersection is
dp 2 av
r − (a/2)2 = p
dt 4 r2 − a2 /4
where we used da/dt = v. However, we’re not done yet, because the speed of the point of intersection
depends on the frame; we need to go back to the original frame. Using the Pythagorean theorem
again, the answer is v
u !2
u av v 2 v 1
t p + = p .
2 2
4 r − a /4 2 2 1 − (a/2r)2
[2] Problem 12 (Kalda). A mirror rotates about its center with angular speed ω. A stationary point
source of light sits at a distance a from the rotation axis. What is the speed of its mirror image?
Solution. Work in the frame rotating with the mirror. Because the image is always flipped across
the mirror with respect to the source, since the source rotates with angular velocity −ω, the image
rotates with angular velocity ω. Then the relative angular velocity of the source and image is 2ω,
which holds in all frames. Thus, in the original frame the image has angular velocity 2ω and speed
2ωa.
[2] Problem 13 (Kalda). Two circles of radius r intersect at the point O. One of the circles rotates
about the point O with constant angular speed ω. The other point of intersection O0 is originally a
distance d from O. Find the speed of O0 as a function of time.
Solution. Remarkably, the answer does not depend on the time! Let d be the distance between
the points of intersection, and work in the rotating frame where the circles rotate with angular
velocities ω/2 and −ω/2 about O.
10
Since θ̇ = ω/2 and cos θ = d/2r, we have
ω d˙
− sin θ = , d˙ = −rω sin θ.
2 2r
This is the vertical velocity of O0 . Now we need to go back to the original frame, which involves
rotating with angular velocity ω/2 about O. Then O0 picks up a horizontal velocity of (2r cos θ)(ω/2)
for a total speed of p
v = r2 ω 2 sin2 θ + r2 ω 2 cos2 θ = rω
which is constant. The geometrical reason is that the second intersection point rotates around the
nonrotating circle with uniform angular velocity ω, as you can show by some angle chasing.
Idea 4
To find the minimum value of some quantity, it’s often useful to think about all possible
values of that quantity. This can reveal a solution using geometry or symmetry.
[2] Problem 14 (PPP 3). A boat can travel a speed of 3 m/s on still water. A boatman wants to
cross a river while covering the shortest possible distance.
(a) In what direction should he row if the speed of the water is 2 m/s?
(b) How about if it is 4 m/s?
Solution. (a) The boatman can completely cancel out the horizontal velocity of the water. He
should row an angle cos−1 (2/3) from the upstream direction, so that the boat moves directly
across the river.
(b) The boatman cannot cancel out the horizontal velocity. Instead, the set of possible velocities
forms a circle in velocity space, as shown.
By taking the velocity with the angle closest to directly across the river, we see the boatman
should row an angle cos−1 (3/4) from the upstream direction.
Idea 5
In problems with friction, the best reference frame to use is almost always the frame of
whatever is causing the friction.
11
[2] Problem 15 (Kalda). A block is pushed onto a conveyor belt. The belt is moving with speed
1 m/s, and the block’s initial speed is 2 m/s, with initial velocity perpendicular to that of the belt.
During the subsequent motion, what is the minimum speed of the block with respect to the ground?
Solution. If the belt were not moving, the block would just decelerate in the direction of its speed,
so that’s what happens in the reference frame of the belt. The possible block velocities are shown
in this frame.
√
The minimum relative speed with the ground is shown by the altitude, which has length 2/ 5 m/s
by similar triangles.
Idea 6
For a variety of kinematics problems, it can be useful to think about the motion from a
different perspective. For example, if your problem involves complicated accelerations, it
can be useful to think in “velocity space”, i.e. directly think about how the velocity vector
evolves over time, and deal with the position later. Or, if your problem involves complicated
processes occurring in time, it can be useful to think in “spacetime”, meaning to visualize
the process on a space where time is one of the axes. It can also be useful to parametrize
motion in terms of quantities other than the usual Cartesian coordinates.
[2] Problem 16 (Kalda). A boy enters a patch of ice with a coefficient of friction µ with speed v.
By running on the ice, the boy turns his velocity vector by 90◦ in the minimum possible time, so
that his final speed is also v. What is the minimum possible time, and what kind of curve is the
trajectory? Assume the normal force with the ice is constant.
Solution. The acceleration
√ always has magnitude
√ µg. The velocity needs to change by v(x̂ − ŷ) if
v
it starts at vŷ, so v 2 = µgt. Thus, t = µg 2. The acceleration is constant, so the trajectory is a
parabola.
[2] Problem 17 (PPP 5). Four snails travel in uniform, rectilinear motion on a plane. The velocities
are chosen so that three snails never meet at once, and no two of the velocities are equal. Since
4

time t = −∞, five of the 2 possible encounters have already occurred. Must the sixth also occur?
Solution. It’s hard to visualize what’s going on in the plane; instead think about what’s going
on in spacetime. The spacetime here is three-dimensional, and the paths of the worms are lines
through it, called worldlines; two worms will encounter each other if their worldlines intersect. For
some set of three of the snails, all possible encounters occur, so their worldlines lie in a plane in
spacetime. (This means that in space, these three snails move on the same line.)
If the fourth snail’s worldline lies in this plane, then it must intersect all three others. If it
doesn’t, it can intersect at most one. Hence if five encounters have already occurred, the sixth must
also occur.
12
[2] Problem 18. Six bugs are placed at the vertices of a regular hexagon with side length s. At time
t = 0 each bug starts moving directly towards the next with speed v. At what time do they collide?
Solution. By symmetry, the bugs always remain in a hexagon shape, but this hexagon rotates and
shrinks. We want to know the time when it collapses completely.
We can first do this by considering how the distance between adjacent bugs changes in an
infinitesimal
√ time dt. The first bug moves a distance v dt towards the second. The second moves a
distance ( 3/2)v dt to the side, and a distance (v/2) dt directly away from the first. The side-to-side
motion doesn’t contribute to the change in distance (one can use the Pythagorean theorem and
binomial theorem to show it is second order, and hence negligible for infinitesimals), so we ignore
it. Then the rate of change of distance between the bugs is just v − v/2 = v/2, so the bugs meet at
t = 2s/v.
Another method is to note that all the bugs meet in the center of the original hexagon, so we
can consider the component of velocity for each bug directed towards the center. This is always
v/2 by the hexagonal symmetry, and the original distance from the center is s, so the bugs again
meet in time t = 2s/v.
[3] Problem 19. A rabbit begins at the origin, and the fox begins at the point (0, −a). The rabbit
begins running with a constant speed vx̂. At the same time, the fox begins chasing the rabbit,
always moving towards it with speed v.
(a) Sketch the subsequent trajectory of the rabbit and fox.
(b) Let the displacement between the rabbit and fox be
r(t) = (x(t), y(t)).
Show that r + x is conserved.
(c) Find the distance between the rabbit and fox after a long time.
(d) Now suppose the fox has speed u > v. How long does it take to catch the rabbit?
Solution. (a) Initially the fox moves up while the rabbit moves to the right. After a while, the
two simply follow each other, with a constant distance between them, along the x axis.
(b) Let θ be the angle between the velocity vectors. Then
dr
= −v + v cos θ
dt
because of the fox and rabbit, and
dx
= v − v cos θ
dt
because of the rabbit and fox. Then r + x is constant, as desired.
(c) Initially, r + x = a + 0 = a. After a long time, r = x, so r = x = a/2.
13
(d) Solving for the trajectory of the fox is extremely difficult, but we can use an extension of the
idea of part (b). Now the equations of motion are
dr dx
= −u + v cos θ, = v − u cos θ.
dt dt
Combining these equations, we can cancel out θ to get
dr dx
u +v = v 2 − u2 .
dt dt
This can now easily be integrated from between the initial and final time. During this time,
the change in r is −a, while the change in x is zero, so
ua
−au = (v 2 − u2 )t, t = 2 .
u − v2
If you’re curious what the full trajectory looks like, you can find it in this paper, which was
written by a past coach of the U.S. Physics Team.
[2] Problem 20 (PPP 85). A child is standing on an icy hill, which may be modeled as an inclined
plane.
The coefficient of friction µk = µs is small enough so that, if the child gets the tiniest push, she will
begin sliding down the plane. Now suppose the child gets a horizontal push, with initial speed v0 .
What is the child’s final speed?
Solution. This is easy because you’ve already solved the problem; it’s just the same thing as
problem 19. Specifically, the displacement between the rabbit and fox there corresponds to the
velocity of the child here. At every increment of time, the velocity changes in two ways: it shrinks
along its direction by friction (corresponding to the fox) and it has a constant added by gravity
(corresponding to the rabbit). Hence the answer is v0 /2.
You should definitely not try to solve for the trajectory exactly, since it’s very messy, but you
can find the gory result in this paper.
3 Motion in Two Dimensions

Idea 7
Often, motion in two dimensions can be treated as two independent one-dimensional problems.
A change of reference frame may be necessary first.
14
Idea 8
In problems involving an inclined plane, always set the angle θ to be much closer to either
0◦ or 90◦ than to 45◦ . This reduces mistakes, because almost every angle will be either θ or
90◦ − θ, and you can identify which by sight.
Example 4
Consider projectile motion where wind provides a constant horizontal force F . At what angle
should a projectile of mass m be launched in order to return to the thrower?
Solution
The key idea is to use tilted coordinate systems. Clearly, when the only force is downward,
the projectile must be launched straight upward. Now, the horizontal force acts like an
effective horizontal gravitational acceleration of F/m, so that gravity is effectively tilted an
angle tan−1 (F/mg) away from the vertical. One must launch the projectile directly “upward”
with respect to this effective gravitational field, so the launch angle is an angle tan−1 (F/mg)
from the vertical. (For a related problem, see the infamous F = ma 2014 problem 19.)
[1] Problem 21 (Quarterfinal 2002). A cart is rigged with a vertical cannon so that, when the cart is
stationary on a horizontal track, the cannonball is fired straight up and lands back in the cannon.
In each of the following situations, does the cannonball land back in the cannon, in front of it, or
behind it?
(a) The cart is moving on a frictionless horizontal track with speed v.
(b) The cart is accelerating down a frictionless inclined track with angle θ.
(c) The cart is accelerating down an inclined track with angle θ, and friction slows it down.
Solution. (a) The motion in the x and y directions is independent. In the x direction, both the
cannonball and cart just continue moving with speed v, so the cannonball lands right back
into the cannon.
(b) Work in the tilted frame where the x axis is parallel to the track. In the x direction, both the
cannonball and cart start with the same speed v and accelerate with the same acceleration
g sin θ, so the cannonball lands right back into the cannon, again.
(c) In this case the cart accelerates less, so the cannonball lands in front.
[2] Problem 22 (Kalda). Two balls at points A and B are released from rest at the same moment,
from the locations shown below. All surfaces are frictionless.
15
If it takes time tA and tB for the balls to hit the ground, at what time was the distance between
the balls the smallest?
Solution. Both balls have a downward acceleration of g sin α, and they have leftward and rightward
accelerations of g 0 = g cos α. Since the balls always have the same vertical speed, we can ignore
the vertical motion entirely. The distance between the balls is thus smallest when their horizontal
separation is zero.
Let the total horizontal distances the balls travel be dA and dB . Then
1 1
dA = g 0 t2A , dB = g 0 t2B
2 2
and we are looking for the time t where
dA − dB 1
= g 0 t2 .
2 2
Solving these equations for t gives r
t2A − t2B
t= .
2
[2] Problem 23 (Kalda). Two planar frictionless walls are placed at right angles, where wall A makes
an angle α to the horizontal. A perfectly elastic ball is released from rest at a point a distance a
from wall A and b from wall B.
After a long time, what is the ratio of the number of times the ball has bounced against wall B to
the number of times it has bounced against wall A?
Solution. In the coordinate system tilted by angle α, the motions in the x and y directions are
independent, because collisions with wall A leave vx unchanged and vice versa. In the y direction,
the ball simply bounces up and down with uniform acceleration g cos α and bounce height a, so
r
2a
∆tA = 2 .
g cos α
By similar reasoning, in the x direction
s
2b
∆tB = 2 .
g sin α
Thus the answer is r

∆tA a sin α
= .
∆tB b cos α
When this ratio is a rational number, the ball eventually returns to its starting point. If it isn’t, it
never does; instead it eventually explores all of the space permitted by energy conservation, i.e. it
eventually passes arbitrarily close to any point whose height is at most the height of the starting
point.
16

[3] Problem 25 (EFPhO 2010). A sprinkler can be modeled as a small hemisphere on the ground.
Water shoots out from the hemisphere in all directions, with speed v perpendicular to the hemisphere.
(a) Find the total surface area of ground watered by the sprinkler.
(b) At what distance from the sprinkler does the ground get the wettest?
Solution. (a) The range of the sprinkler is maximized at 45◦ and is equal to v 2 /g. Then the
area is π(v 2 /g)2 = πv 4 /g 2 .
(b) The outermost circle, at radius v 2 /g, gets by far the wettest. This is because a maximum
of radius is achieved here, so a large range of launch angles gets to near this radius. (It’s
the same reason that balls thrown upward spend the most time near the very top of their
trajectories.)
This idea is a little tricky, but very general; for instance, it’s the principle behind the formation
of caustics such as rainbows, as we’ll see in W3. It is also the way in which classical mechanics
emerges from quantum mechanics: classically things follow the trajectory of least action
because it’s a caustic of the quantum sum over all trajectories. So if you continue in physics,
you’ll see this beautiful little idea over and over again, in richer and richer settings! For an
Olympiad problem that gives a bit more detail about caustics in optics, see here.
Example 5
A bug flies towards a light with constant speed v, always making an angle α with the radial
direction. If the initial distance to the lamp is L and the radius of the lamp is R, through
what total angle does it turn before hitting the lamp?
Solution
In this case we can’t avoid solving differential equations, but they’re not too hard. It’s easiest
to work in polar coordinates, with the center of the lamp at the origin. By decomposing the
velocity into radial and tangential components, we have
dr dθ
= −v cos α, r = v sin α.
dt dt
We only care about the path, not the time-dependence, so we divide these equations to get
dr r
=−
dθ tan α
where we manipulated differentials as in P1. Separating and integrating,
Z R
dr ∆θ
− =
L r tan α
which tells us that
L
∆θ = (tan α) log .
R
The shape traced out is a logarithmic spiral.
17
[2] Problem 26. The pilot of a supersonic jet airplane wishes to make a big noise at the origin by
flying around it in a path such that all of the noise he makes is heard simultaneously at the origin.
The jet travels with Mach number M , meaning that its speed is M times the speed of sound. If the
pilot starts at (r, θ) = (a, 0), find the pilot’s path r(θ).
Solution. In order for the sound to reach the origin simultaneously, we must have r(t) = a − ct,
so that the sound all reaches the origin at time a/c. On the other hand, we have
(M c)2 = ṙ2 + r2 θ̇2 = c2 + r2 θ̇2 .
This is a bit messy because we have two functions of time, but we can eliminate time by using
dθ dr dθ
θ̇ = =c .
dr dt dr
Plugging this in above, we have
2
2 2 dθ
M −1=r
dr
and separating and integrating gives
Z
dr
Z
dθ √
M 2 −1
= √ , r(θ) = ae−θ/ .
r M2 − 1
[4] Problem 27. Consider a mass m on a table attached to a spring at the origin with zero relaxed
length, which exerts the force
F = −kr
on the mass. We will find the general solution for r(t) = (x(t), y(t)) in two different ways.
(a) Directly write down the answer, using the fact that the x and y coordinates are independent.
(b) Sketch a representative sample of solutions. What kind of curve does the trajectory follow?
(c) ? Here’s a more unusual way to arrive at the same answer. Go to a noninertial reference frame
rotating with angular velocity ω0 about the origin, so that the centrifugal force cancels out
the spring force. In this frame, the only relevant force is the Coriolis force −2mω0 × v. Find
the general solution in this frame, then transform back to the original frame and show that
you get the same answer as in part (a). (This can get a bit messy; the easiest way is to treat
the plane as the complex plane, i.e. work in terms of the variable r = x + iy.)
Solution. (a) We just have two separate equations for each component,
d2 x k d2 y k
= − x, = − y.
dt2 m dt2 m
p
Both describe a harmonic oscillator with frequency ω0 = k/m. Then the general solution
can be written as
x(t) = A cos(ω0 t + φ1 ), y(t) = B sin(ω0 t + φ2 ).
In general, it is very rare for the x and y coordinates to be independent. Another example of
this type is projectile motion in linear drag, F = −kv. In these cases the 2D or 3D problem
is no harder than the 1D version, but we’re rarely so lucky.
18
(b) In the case where φ1 = φ2 = 0 and A = B, the mass moves in a circle centered at the origin.
More generally, when the angles φi are unequal, the mass can move in an ellipse with center
at the origin.
p
(c) The centrifugal force is mω02 r, so to cancel the spring force we need to choose ω0 = k/m.
Now, in the rotating frame, the Coriolis force acts just like a magnetic field: it’s always
perpendicular to the motion, so the solution is circular motion. The angular frequency ωc of
that circular motion satisfies
mv 2
2mω0 v = = mωc v
r
from which we conclude ωc = 2ω0 . So in complex notation,
r(t) = r0 + r1 e2iω0 t
in the rotating frame. We can return to the original frame by simply multiplying by e−iω0 t
(the sign is important, i.e. you have to keep track of the direction of ω0 ), to give
r(t) = r0 e−iω0 t + r1 eiω0 t .
Taking real and imaginary parts and letting ri = ai + ibi ,
x(t) = (a0 + a1 ) cos(ω0 t) + (b0 − b1 ) sin(ω0 t), y(t) = (b0 + b1 ) cos(ω0 t) + (a1 − a0 ) sin(ω0 t).
This is the same as our result for part (a), after you use the sine and cosine addition formulas
and appropriately redefine the parameters. Evidently, elliptical motion is just the superposition
of two opposite circular motions! (In general, complex numbers are a useful way to deal with
magnetic or Coriolis forces for motion in a plane, where B or ω points perpendicular to the
plane. In these cases the force lies in the plane perpendicular to the velocity, so it’s just
proportional to iṙ, which is nice and simple; we’ll see this idea again later.)
4 Optimal Launching
Finally, we’ll consider projectile motion questions that involve optimization. These are rare on the
USAPhO, but they are quite fun problems, with occasionally very slick solutions.
Example 6
A bug wishes to jump over a cylindrical log of radius R lying on the ground, so that it just
grazes the top of the log horizontally as it passes by. What is the minimum launch speed v
required to do this?
Solution
Let P be the point at the top of the log. For the bug to be moving horizontally at P , energy
conservation applied to the vertical motion gives an initial vy obeying
1 p
mv 2 = 2mgR, vy = 2 gR.
2 y
19
Thus, we need to find the minimum vx for the motion to be possible. If vx is too low, the
hypothetical trajectory of the bug will instead pass through the log. At the lowest possible
vx , the bug’s trajectory is not just tangent to the log at point P , but also has the same radius
of curvature (i.e. the trajectory and the log’s shape have the same first and second derivatives).
For uniform motion in a circle of radius r, the acceleration is a = v 2 /r. Conversely, when an
object follows a trajectory of instantaneous radius of curvature r, its acceleration component
normal to the path must be a = v 2 /r. So applying this to the bug at P gives
vx2 p
g= , vx = gR.
R
Thus, the minimum initial speed is
q √
v= vx2 + vy2 = 5 gR.
This radius of curvature trick doesn’t come up often, but it’s cool when it does.
[2] Problem 28. NBPhO 2020, problem 3. A nice warmup for the problems below.
[3] Problem 29. An object is launched from the top of a hill, where the ground lies an angle φ below
the horizontal. Show that the range of a projectile is maximized if it is launched along the angle
bisector of the vertical and the ground.
Solution. This is a straightforward if messy problem; we’ll show one of many ways to set it up.
Setting the origin at the launch point and using ordinary horizontal/vertical coordinates, the object
hits the hill when tan φ = −y/x. Using results for projectile trajectories from the preliminary
problem set, we have
y gx
= − tan φ = tan θ − 2
x 2v cos2 θ
where θ is the launch angle from the horizontal. Solving for x,
2v 2 cos2 θ
x= (tan θ + tan φ) ∝ sin θ cos θ + cos2 θ tan φ.
g
To maximize the range, we want to maximize x, so setting the derivative to zero gives
0 = cos2 θ − sin2 θ − 2 sin θ cos θ tan φ
which simplifies to
1 φ + π/2
tan(2θ) = = tan(φ + π/2), θ =
tan φ 2
as desired. This famous problem was first posed by Torricelli in the 1640s, and solved by Halley in
the 1690s.
[3] Problem 30 (PPP 35). A point P is located above an inclined plane with angle α. It is possible
to reach the plane by sliding under gravity down a straight frictionless wire, joining P to some point
P 0 on the plane. Geometrically, how should P 0 be chosen so as to minimize the time taken? (Hint:
think about the set of points that can be reached for all possible angles of the wire, after time t.)
20
Solution. Suppose the wire is at angle θ with respect to the vertical. Then, the distance traveled
in time t is 12 (g cos θ)t2 . Keeping t fixed, we then see that the locus of all reached points is a circle
whose topmost point is P , and whose radius is 12 gt2 . Therefore, P 0 is the point where one of these
circles is tangent to the incline, so an α/2 angle to the vertical.
Idea 9
Since mechanics is time-reversible, and the speed of a projectile only depends on its height
and not the path taken, finding the way to reach point B from point A with the lowest
possible initial speed is the same as finding the way to reach point A from point B with the
lowest possible initial speed.
[4] Problem 31. Two fences of heights h1 and h2 are erected on a horizontal plain, so that the tops
of the fences are separated byp a distance d. Show that the minimum speed needed to throw a
projectile over both fences is g(h1 + h2 + d).
Solution. It’s very confusing to think about how to throw the projectile starting from the ground,
because you need to figure out where to launch and at what angle, under the condition that the
trajectory just touches the tops of both fences. A much better way is to imagine the projectile
starts at the top of the higher fence; the goal is then to throw it with minimal energy so that it
just touches the top of the lower fence. At some point, this projectile will then reach the ground,
though we don’t have to worry about where. Since mechanics is time-reversible, its speed at this
point (which is found easily by energy conservation) will be the minimal possible speed.
Now there are many ways to do this problem. A very slick solution, which requires no computation
at all, is presented in problem 32. However, we’ll present a more direct attack for completeness.
Note that if you want to hit the top of the lower fence with the minimum velocity, it’s equivalent
to maximizing your throwing range down an inclined plane, namely the plane that connects the
tops of the two fences. Then the optimal launch angle is along the angle bisector, as we found in
problem 29. Using the same starting point as the solution to that problem, we have
√
h g d2 − h2
−√ = tan θ −
d2 − h2 2v 2 cos2 θ
where we let h = h2 − h1 > 0. That solution gives a simple expression for tan 2θ, so we massage
this equation to
g sin 2θ h
2
=√ + 2 (1 + cos 2θ).
v 2
d −h 2 d − h2
21
We then plug in our previous results, which are

√
d2 − h2 h
sin 2θ = , cos 2θ =
d d
to get the result
v 2 = (d − h)g = (d + h1 − h2 )g.
By energy conservation, the speed at the ground is
v02 = v 2 + 2h2 g = (d + h1 + h2 )g
as desired.
[4] Problem 32. It’s possible to solve problems 29 and 31 using pure geometry, with no computation.
One can show that the set of points a projectile can reach with a fixed initial speed v is a parabola
with a focus at the launching point. A parabola is defined as the set of points whose distance to
the focus equals the distance to a line, called the directrix.
(a) Show that trajectories that touch the parabola must be tangent to it.
(b) Show that if a point is hit with the smallest possible initial speed, then the initial velocity
must be perpendicular to the final velocity.
(c) Using the geometric definition of a parabola, recover the answers to problems 29 and 31.
Solution. (a) This is just because the parabola is defined to be the set of points you can hit. If
the trajectory weren’t tangent to the parabola, you would be able to hit a point outside the
parabola by continuing it.
(b) Let vi be the initial velocity and v̂⊥ be a unit vector in the perpendicular direction. If we
replace the initial velocity by vi + v̂⊥ , where is infinitesimal, then the speed isn’t changed,
which implies that the new trajectory should remain inside the parabola. Now suppose the
original projectile’s velocity is vf when it is tangent to the parabola, at position rf . Then
at the same time, the new projectile’s position is rf + tv̂⊥ . In order to keep this inside
the parabola for all infinitesimal , both positive and negative, tv̂⊥ must be tangent to the
parabola at this point. Hence v̂⊥ is parallel to vf , so vi is perpendicular to vf , as desired.
(c) A parabola is the set of points equidistant from a focus and a line, called the directrix. Refer
to the diagram below.
The final velocity vf is tangent to the parabola. Therefore, it points along the angle bisector
between the vertical and the direction along the plane since the distance from the focus and
22
the directrix will remain equal to each other. Now, vi is perpendicular to this, which means it
is along the angle bisector between the vertical and the downward direction along the plane,
which is precisely the result we found in problem 29.
Assume h2 > h1 . Draw the parabola of the projectile range with its focus on h2 . At the
minimum launching velocity, the parabola should just touch the top of the h1 fence. The
(horizontal) directrix will be a distance 2x above h2 , where x = v22 /2g and v2 is the launching
velocity from h2 . Then use the fact that the distances from a point on a parabola to the focus
and directrix are the same.
From the picture, we see that d + h1 = 2x + h2 . Thus the launching velocity at h2 satisfies
v22 /g = d + h1 − h2 , and it has a total energy upon launching of
1 d + h1 − h2 1
E/m = v22 + gh2 = g + gh2 = mv02 .
2 2 2
This gives the answer with almost no computation,
p
v0 = g(d + h1 + h2 ).
[3] Problem 33. 01m IPhO 2012, problem 1A.
5 Reading Graphs
In some kinematics problems, you’ll have to infer what’s going on from a diagram. To make progress,
you’ll have to print out the diagram to make measurements directly on it.
Remark
For a ridiculously hard problem from the same genre, see EuPhO 2019, problem 3. Almost
all competitors received zero points on it; you can try it for entertainment if you’ve finished
everything else and really like kinematics. The official solutions are here.
23
Mechanics II: Statics

For review, read chapter 2 of Morin or chapter 2 of Kleppner and Kolenkow. Statics is covered in
more detail in chapter 7 of Wang and Ricardo, volume 1. Surface tension is covered in detail in
chapter 5 of Physics of Continuous Matter by Lautrup, which is an upper-division level introduction
to fluids in general. There is a total of 79 points.
1 Balancing Forces
Idea 1
In principle, you can always solve every statics problem by balancing forces on every individual
particle in the setup, but often you can save on effort by considering appropriate systems.
Idea 2
Any problem where everything has a uniform velocity is equivalent to a statics problem,
by going to the reference frame moving with that velocity. Any problem where everything
has a uniform acceleration a is also about statics, by going to the noninertial frame with
acceleration a, where there is an extra effective gravitational acceleration −a. The same
principle applies to uniform rotation, where a centrifugal force appears in the rotating frame,
acting like an effective gravitational acceleration ω 2 r.
Example 1
Six blocks are attached in a horizontal line with rigid rods, and placed on a table with
coefficient of friction µ. The blocks have mass m and the leftmost block is pulled with a force
F so the blocks slide to the left. Find the tension force in the rod in the middle.
Solution
There are six objects here and five rods, each with a different tension, so a direct analysis
would involve solving a system of six equations. Instead, first consider the entire set of six
blocks as one object; we can do this because the rigid rods force them to move as one. The
total mass is 6m, and applying Newton’s second law gives
F
F − 6mgµ = 6ma, a= − µg.
6m
Next, consider the rightmost three blocks as one object. Their total mass is 3m, and their
acceleration is the same acceleration a we computed above. This system experiences two
horizontal force: tension and friction. Newton’s second law gives
T − 3mgµ = 3ma
and solving for T gives

F
T = .
2
1
This is intuitive, because the differences of any two adjacent tension forces are the same;
that’s the amount of tension that needs to be spent to accelerate each block. So the middle
rod, which has to accelerate only half the blocks, has half the tension.
The reason we could ignore the tension forces in the other four rods is that the only thing
they do is ensure the blocks move with the same acceleration. Once we assume this is the
case, the specific values of the tensions don’t matter; we can just zoom out and forget them.
It’s just like how within each block there is also an internal tension which keeps it together,
but we rarely need to worry about its details.
Idea 3
To handle a problem where something is just about to slip on something else, set the frictional
force to the maximal value µN and assume slipping is not yet occurring, so the two objects
move as one. The same idea holds for problems which ask for the minimal force needed to
make something move, or the minimal force needed to keep something from moving.
[1] Problem 1 (KK 2.7). A block of mass M1 sits on a block of mass M2 on a frictionless table. The
coefficient of friction between the blocks is µ. Find the maximum horizontal force that can be
applied to (a) block 1 or (b) block 2 so that the blocks will not slip on each other.
Solution. Let the horizontal force be F . In both cases the friction is maximal, f = µM1 g, and the
blocks move together, so a = F/(M1 + M2 ).
(a) The bottom block experiences only the force f = M2 a, so
M1
µM1 g = M2 a, F = µg(M1 + M2 )
M2
(b) The top block experiences only the force f = M1 a, so
µM1 g = M1 a, F = µg(M1 + M2 ).
[2] Problem 2 (KK 2.28). An automobile enters a turn of radius R.
The road is banked at angle θ, and the coefficient of friction between the wheels and road is µ. Find
the maximum and minimum speeds for the car to stay on the road without skidding sideways.
Solution. Let N be the normal force, and let f be the friction force (with sign f > 0 if it’s pointing
up the hill). We see that N cos θ + f sin θ = mg, and N sin θ − f cos θ = mv 2 /R. Therefore,
v2 N sin θ − f cos θ
=
gR N cos θ + f sin θ
2
Since −N µ ≤ f ≤ N µ, we have
2
vmin sin θ − µ cos θ 2
vmax sin θ + µ cos θ
= , = .
gR cos θ + µ sin θ gR cos θ − µ sin θ
In addition, there are some cases where these formulas break down. If µ > tan θ, then the minimum
speed is zero. If µ > cot θ, then the maximum speed is infinity. (In these cases, the formulas
give nonsense, i.e. imaginary numbers for the speeds. That’s one way of checking, at the end of a
problem, whether there are more special cases that must be accounted for.)
[2] Problem 3 (KK 2.19). A “pedagogical machine” is illustrated in the sketch below.
All surfaces are frictionless. What force F must be applied to M1 to keep M3 from rising or falling?
F
Solution. By considering all the masses as one system, we see that a = M1 +M2 +M3 . We see that
the tension T = M3 g, and T = M2 a, so
F M3 M3
M3 g = M2 a =⇒ = g =⇒ F = (M1 + M2 + M3 ) g.
M1 + M2 + M 3 M2 M2
2 Balancing Torques
Idea 4
A static rigid body will remain static as long as the total force on it vanishes, and the total
torque vanishes, where the torque about the origin is
X
τ= r i × Fi
i
where ri is the point of application of force Fi . If the total force vanishes, the total torque
doesn’t depend on where the origin is, because shifting the origin by a changes the torque by
!
X X
∆τ = a × Fi = a × Fi = 0.
i i
The origin should usually be chosen to set as many torques as possible to zero.
[1] Problem 5. The line of a force is defined to the line passing through its point of application
parallel to its direction; then the torque of the force about any point on that line vanishes. Suppose
a body is static and has three forces acting on it. Show that in two dimensions, the lines of these
forces must either be parallel or concurrent. This will be useful for several problems later.
3
Solution. Let F1 , F2 , F3 be the forces. Suppose two are parallel, then the third must be parallel to
the first two to balance forces in the direction perpendicular to the direction of the first two. Now,
suppose they are not parallel, and let the origin be at the intersection of the lines of forces of F1
and F2 . Then, the torque due to these two is zero, so the torque due to F3 must also be zero, so
the line of action of F3 must also pass through the origin.
Idea 5
The center of mass rCM of a set of masses mi at locations ri with total mass M satisfies
X
M rCM = mi ri .
i
If a system experiences no external forces, its center of mass moves at constant velocity.
Idea 6
A uniform gravitational field exerts no torque about the center of mass. Thus, for the
purposes of applying torque balance on an entire object, the gravitational force M g can
be taken to act entirely at its center of mass. (This is a formal substitution; of course, the
actual gravitational force remains distributed throughout the object.)
Torque balance works in noninertial frames, as long as one accounts for the torques due to
fictitious forces. Thus, for an accelerating frame, the −M a fictitious force can be taken to act
at the center of mass. In a uniformly rotating frame, the total centrifugal force is M ω 2 rcm ,
and for the purposes of balancing torques, can be taken to act entirely at the center of mass.
Example 2
Show that the tension in a completely flexible static rope, massive or massless, points along
the rope everywhere in the rope.
Solution
Consider a tiny segment d` of the rope. Since the rope is static, the tension forces on
both ends balance, so they are opposite. Let them both be at an angle θ to the rope
direction. Then the net torque on the segment is (T d`) sin θ. Since this must vanish for
static equilibrium, we must have θ = 0 and hence the tension is along the rope. In other
words, flexible ropes can transmit force, but they can’t transmit torque.
It’s important to note that the argument above doesn’t work for a rigid rod, because the
internal forces in a rigid object can look like the picture above. In other words, there can be
extra shear forces from the adjacent pieces of the rod that provide the compensating torque.
If one tried to set up forces like this in a rope, it would flex instead.
4
In general, the force distribution within a massless rigid rod can be quite complicated, but if
we zoom out, we can replace it with a single tension which does not necessarily point along
the rod. This transmits both a force and a torque through the rod, in the sense that a torque
is eventually exerted by whatever holds the end of the rod in place. Note that if the rod’s
supports are free to rotate, then they can’t absorb torque, so the rod acts just like a rope,
with tension always along it.
Remark
Sometimes, problem writers will intentionally not introduce any variables that are irrelevant
to the answer. This can occur in two ways. First, the variables might just cancel out, as
one can often see by dimensional analysis. Second, the specific values of the variables might
not matter in the limit when they are very large or small. For instance, if a problem simply
states a mass is “very heavy” but doesn’t give it a name like m, it is asking for the answer
in the limit m → ∞.
Idea 7
To handle problems where an object is just about to tip over, note that at this moment, the
entire normal force will often be concentrated at a point. (For example, when you’re about
to fall forward, all your weight goes on your toes.) That often means it’s a good idea to take
torques about this point.
Example 3: Povey 5.6
Suppose that on level ground, a car has a distance d between its left and right tires, and its
center of mass is a height h above the ground. Now suppose the car turns as in problem 2,
but in the extreme case θ = 90◦ , with speed v. For what v is this motion possible?
Solution
Again working in the noninertial frame of the car, force balance gives
mv 2
ffric = mg, N=
R
where ffric and N are the total friction and normal forces on the four tires. Since ffric /N ≤ µ,
p
v ≥ gR/µ
which matches the general solution to problem 2. But in that problem, we only considered
force balance. In this extreme situation, we also have to consider torque balance, i.e. the
possibility that the car might topple over. When the car is about to topple over, all the
normal and friction force is on the bottom tires. About this point, we have only torques from
gravity and the centrifugal force, giving
mv 2 d
mgh =
R 2
5
p
and solving for v gives v = 2gRh/d. Toppling is less likely the higher v is, so the answer is
p √ p
v ≥ gR max(1/ µ, 2h/d).
[2] Problem 6 (Quarterfinal 2004.3). A uniform board of length L is placed on the back of a truck.
There is no friction between the top of the board and the vertical surface of the truck. The coefficient
of static friction between the bottom of the board and the horizontal surface of the truck is µs = 0.5.
The truck always moves in the forward direction.
(a) What is the maximum starting acceleration the truck can have if the board is not to slip or
fall over?
(b) What is the maximum stopping acceleration the truck can have if the board is not to slip or
fall over?
(c) For what value of stopping acceleration is the static frictional force equal to zero?
Solution. Let us work in the accelerating frame of the truck.
Force balance gives mg = N and N2 + ma = f , and torque balance gives

L L
−mg sin θ + ma cos θ + N2 L cos θ = 0
2 2
which implies
2N2 + ma = mg tan θ.
6
Thus,
m(g tan θ − a) m(g tan θ + a)
N2 = , f= .
2 2
Since −mgµ ≤ f ≤ mgµ, to avoid slipping we require
3 7 1
−g ≤ g tan θ + a ≤ g =⇒ −g ≤ g + a ≤ g =⇒ − g ≤ a ≤ g.
4 4 4
To avoid falling over, we need N2 > 0, which is equivalent to
3
a ≤ g tan θ = g.
4
We can now read off the answers.
(a) For starting accelerations above 3g/4 we would have falling, while for ones above g/4 we would
have slipping. So slipping kicks in first, and the answer is g/4.
(b) Here the only constraint is slipping, and the answer is 7g/4.
(c) Here 34 g + a = 0, so the truck decelerates with acceleration 0.75g.
[2] Problem 7 (Kalda). Three identical rods are connected by freely rotating hinges.
The rods are arranged so that CD is parallel to AB, and AB = 2 CD. A mass m is hung on hinge
C. What is the minimum force that must be exerted at hinge D to keep the system stationary?
Solution. Let the rods have length `. There are many ways to solve the problem, but the quickest
is to consider the torque on the system of rod CD and its hinges, about the intersection point of AC
and BD. About this point, the torque due to the weight of rod CD vanishes. Since the hinges are
freely rotating, the force of rod AC on the system is directed along AC, so it also exerts no torque,
and the same applies for the force from rod BD.
Thus, the only torque is mg`/2, from the weight of the mass. The applied force must balance
this torque, and by some elementary geometry, we find that its maximum possible lever arm is `,
when the force is perpendicular to BD. Therefore, the minimum force is mg/2 .
Idea 8
An extended object supported at a point may be static if its center of mass lies directly above
or below that point. More generally, if the object is supported at a set of points, it can be
static if its center of mass lies above the convex hull of the points.
[2] Problem 8. N identical uniform bricks of length L are stacked on top of each other on the edge
of a table. What is the maximum possible length the top brick can protrude over the edge of the
table? What is the limit of this length as N goes to infinity?
7
Solution. Suppose we begin with all N blocks stacked directly on top of each other and slide them
to the right. The maximal extension is reached when the center of mass of the top n blocks lies on
the edge of the (n + 1)th block. Let ` = L/2, and suppose we have already adjusted the top n − 1
blocks to be in the optimal position. Then the center of mass of the top n blocks is a distance `/n
from the edge of the (n + 1)th block, so the nth block and everything on top of it may be moved
`/n to the right. Hence the total distance is
L N dx
Z
L 1 1 L
1 + + ... + ≈ ≈ log N
2 2 N 2 1 x 2
which is unbounded as N → ∞.
[2] Problem 9 (Kalda). A cylinder with mass M is placed on an inclined slope with angle α so that
its axis is horizontal. A small block of mass m is placed inside it.
The coefficient of friction between the block and cylinder is µ. Find the maximum α so that the
cylinder can stay at rest, assuming that the coefficient of friction between the cylinder and slope is
high enough to keep the cylinder from slipping.
Solution.
We see that the center of mass of the cylinder-block system must be right above the contact point
A. Now, we see that the CM is at B where CB/BO = M/m. Thus, OA = OB + OC = k(m + M ),
so by the law of sines on OAB, we have
OB OA
= =⇒ sin(α + θ) = (1 + M/m) sin α.
sin α sin(α + θ)
We see that m slips when tan(α + θ) = µ, or sin(α + θ) = √ µ , so

1+µ2
!
M −1

µ
αmax = sin−1 p 1+ .
1 + µ2 m
8
[2] Problem 10 (PPP 11). A sphere is made of two homogeneous hemispheres stuck together, with
different densities. Is it possible to choose the densities so that the sphere can be placed on an
inclined plane with incline 30◦ and remain in equilibrium? Assume the coefficient of friction is
sufficiently high so that the sphere cannot slip.
Solution. By balancing torques around the point of contact, we need the center of mass to be
straight above the point of contact. Doing some geometry, we learn that the center of mass be more
than a distance R/2 away from the center of the sphere.
We now show that this is impossible. Consider a homogeneous hemisphere flat on a table. Its
center of mass must be at a height lower than R/2, since the mass above the plane z = R/2 is less
than the mass below it, and concentrated closer to the plane. Therefore, the centers of masses of
the hemispheres are each within R/2 of the center of the sphere. Since the overall center of mass is
a convex combination of the two, it is also within R/2 of the center, so the sphere cannot be stable.
[3] Problem 11. An object of mass m lies on a uniform floor, with coefficient of static friction µ.
(a) First, suppose the object is a point mass. What is the minimum force required to make the
object start moving, if you can apply the force in any direction?
(b) Now suppose the object is a thin, uniform bar. What is the minimum force required to make
the object start moving, if the force can only be applied horizontally? Assume the normal
pressure on the floor remains uniform.
Solution. (a) Just before the block slides, the friction force is N µ, so we can think of the normal
force and friction force as exerting an effective normal force with angle φ with respect to the
vertical, where tan φ = µ.
We see that the magnitude of N varies as P varies along the line ` which is the line at angle
φ to AO, and the magnitude of F is AP . AP is minimized when P is the foot of the altitude
from A to `, so we see
M gµ
Fmin = M g sin φ = p .
1 + µ2
As a sidenote, if the block were treated as an extended object, not just a point particle, one
would have to worry about whether it’s possible to do this without tipping the block over
instead. However, by choosing the point of application of the force correctly, it’s always
possible to make the block slide without tipping. Can you see why?
9
(b) Naively the answer is µmg, because that’s the maximum total friction force. However, we
know from everyday experience that it’s easier to get the object to start moving if you pull
at the edge. That’s because the friction forces distributed along the bar also need to balance
torque, which means some of them must point along the force you exert.
Just before slipping, friction has the maximum possible magnitude everywhere, and points
either directly against or directly along the force you exert. Using the variables defined in the
figure, just barely balancing forces and torques simultaneously gives

` L−` ` ` L−`L−`
F = µmg − , F ` = µmg + .
L L L2 L 2
√
Solving for ` gives ` = L/ 2, and plugging this in gives
√
F = ( 2 − 1)µmg
which is less than half the naive answer!
[3] Problem 12 (Morin 2.17). A spool consists of an axle of radius r and an outside circle of radius
R which rolls on the ground.
A thread is wrapped around the axle and is pulled with tension T at an angle θ with the horizontal.
(a) Which way does the spool move if it is pulled with θ = 0?
(b) Given R and r, what should θ be so that the spool doesn’t move? Assume that the friction
between the spool and the ground is large enough so that the spool doesn’t slip.
(c) Given R, r, and the coefficient of friction µ between the spool and the ground, what is the
largest value of T for which the spool remains at rest?
(d) Given R and µ, what should r be so that you can make the spool slip from the static position
with as small a T as possible? That is, what should r be so that the upper bound on T in
part (c) is as small as possible? What is the resulting value of T ?
Solution. (a) The torque about the contact point with the ground is clockwise, so the spool rolls
to the right. You might think it would roll to the left, by thinking about torque about the
center, but one must also account for the torque from friction with the ground; taking torques
about the contact point avoids this complication.
10
(b) Let O be the center of the spool, A the point where the thread leaves the inner circle, and B
the point of contact of the outer circle with the floor. We see that ∠BAO = θ. Considering
torques about B, we see that gravity provides 0 torque, so the tension must provide 0 torque
as well. This means BA is tangent to the inner circle. Since BAO is a right triangle with
∠BAO = 90◦ , we have that cos θ = r/R .
(c) Let f be the friction force, and N the normal force. We see that T cos θ = f and N =
M g − T sin θ. Since f ≤ µN , we see
µM g
T cos θ ≤ µ(M g − T sin θ) =⇒ T ≤ ,
cos θ + µ sin θ
where θ = cos−1 (r/R).

1
(d) We see that cos θ + µ sin θ = √ cos(θ − β) where tan β = µ. Thus,
1+µ2
µM g
T =p ,
1 + µ2 cos(θ − β)
R
so to minimize T , we want θ = β, so r = R cos β = p , and the minimum value of T
1 + µ2
µM g
is p .
1 + µ2
[3] Problem 13 (PPP 44). A plate, bent at right angles along its center line, is placed on a horizontal
fixed cylinder of radius R as shown.
How large does the coefficient of static friction between the cylinder and plate need to be if the
plate is not to slip off the cylinder?
Solution. Let the normal and friction forces at the top be Nt , ft and at the right Nr , fr , and the
static coefficient of friction be µ. Balancing forces on the plate gives
ft = Nr , Nt + fr = mg.
Now, it’s not obvious whether friction will be maximal at the top or the right contact point, or
both, so we define
ft = µt Nt , fr = µr Nr
where µt , µt ≤ µ. Eliminating the friction forces and solving the force balance equations gives
mg mgµt
Nt = , Nr = .
1 + µr µt 1 + µr µt
11
Next, consider torques on the plate about its vertex. (This is an arbitrary choice; taking torques
about either of the contact points also works about equally well.) The weight of the vertical of the
plate contributes no torque, so the torque balance equation is
Nr + mg/2 = Nt .
Plugging in our results for Nr and Nt gives
µt (2 + µr ) = 1.
To find the minimum coefficient of friction to avoid slipping, we need to find the solution to this
equation where the larger of µr and µt is as small as possible. But it’s clear now that increasing
one decreases the other, so this is achieved when the two are equal. In other words, at the limit,
slipping is just about to occur at both contact points simultaneously. Setting µr = µt = µ gives
√
µ2 + 2µ − 1 = 0, µ= 2 − 1.
Incidentally, you can also do this problem with the idea of problem 5. At the minimum µ, we
assume both friction forces are saturated. The lines of these forces must cross at a point directly
above/below the center of mass, where gravity is applied. This quickly yields the same quadratic
equation as found above. If you do it this way, though, it’s a bit harder to see why both friction
forces are saturated simultaneously at the minimum µ. It’s usually true, but not guaranteed in
general; our more explicit derivation above shows why.
3 Trickier Torques
Idea 9
Sometimes, a clever use of torque balance can be used to remove any need to have explicit
force equations at all. Rarely, the same situation can occur in reverse.
A spherical ball of mass M is rolled up along a vertical wall, by exerting a force F to some
point P on the ball. The coefficient of friction is µ. What is the minimum possible force F ,
and in this case, where is the point P ?
Solution
Following the logic of idea 3, when the minimum possible force is used, the frictional
force with the wall must be maximal, f = µN , and directed upward. (If friction weren’t
pushing the ball up as hard as possible, we could get by using a smaller force F .)
So even though we don’t know the magnitude of the normal or the frictional force, we
know the direction of the sum of these two forces, so we’ll consider them as one combined force.
This reduces the number of independent forces in the problem to three: gravity (acting at
the center of mass), the force F (acting at P ), and the combined normal and friction forces
12
(acting at the point of contact C with the wall). Therefore, by the result of problem 5, the
lines of these forces must all intersect at some point A, as shown.
This ensures that the torques will balance, when taken about point A.
Next, we need to incorporate the information from force balance. Doing this directly will
lead us to some nasty trigonometry, but there’s a better way. There are in principle two
force balance equations, for horizontal and vertical forces. However, one of these equations is
just going to tell us the magnitude of the normal/frictional force, which we don’t care about.
So in reality, we just need one equation, which preferably doesn’t involve that force.
The trick is to use torque balance again, about the point C, which says that the torques due
to gravity and F must cancel. Now you might ask, didn’t we already use torque balance?
We did, but recall from idea 4 that taking the torque about a different point can give you a
different equation if the forces don’t balance. So by demanding the torque vanish about two
different points, we actually are using force balance! (Specifically, we are using the linear
combination of the horizontal and vertical force balance equations that doesn’t involve the
normal/friction force, which we don’t need to find anyway.)
When taking the torque about C, we see that F is minimized if P is chosen to maximize
p the
lever arm of the force. This occurs when CA ⊥ P A, in which case the lever arm is R 1 + µ2 ,
where R is the radius of the ball. So we have
p Mg
M gR = F R 1 + µ2 , F =p
1 + µ2
and P is determined as described above.
[2] Problem 14. NBPhO 2020, problem 4, parts (i) and (ii).
[3] Problem 15. EFPhO 2012, problem 3. The problem statement is missing some information: both
the bars and rod have diameter d.
[3] Problem 16. EFPhO 2006, problem 6. You will need to print out the problem to make measure-
ments on the provided figure.
13
[4] Problem 17 (Physics Cup 2012). A thin rod of mass m is placed in a corner so that the rod forms
an angle α with the floor. The gravitational acceleration is g, and the coefficient of friction with
the wall and floor is µs = tan β, which is not large enough to keep the rod from slipping.
What is the minimum additional force F needed to keep the rod static?
Solution. See the solutions here.
We’ve now covered some really mathematically elegant problems, but it’s important to remember
the real-world limitations of this kind of analysis. We discuss two examples below.
Example 5
A uniform bar with mass m and length ` hangs on four equally spaced identical light wires.
Initially, all four wires have tension mg/4.
Find the tensions after the leftmost wire is cut.
Solution
This illustrates a common issue with setups involving rigid supports: there are often more
normal forces than independent equations, so there is not a unique solution. In the real
world, the result is determined by imperfect characteristics of the wires. For example, if one
of the wires was slightly longer than the others, it would go slack, reducing the number of
normal forces by one and yielding a solution.
A reasonable assumption, if you aren’t given any further information, is to assume that the
supports are identical, very stiff springs. In equilibrium, the bar will tilt a tiny bit, so that
the length of the middle wire will be the average of the lengths of the other two. By Hooke’s
law, the force in that wire will than be the average of the other two, so the tensions are
mg/3 − x, mg/3, and mg/3 + x. Applying torque balance yields 7mg/12, mg/3, and mg/12.
14
Remark: Subtleties of Friction

Statics problems involving friction can also get quite elegant, but it’s important to remember
that at the end of the day, they’re just a decent approximation for the real world. In reality,
laws like Ffric = −µk N are only approximately true for some regimes of behavior of some
materials. In fact, the particular law Ffric = −µk N can sometimes yield equations with no
solutions, a phenomenon called the Painleve paradox!
The source of the paradox is the discontinuous dependence of the direction of the friction
force on the direction of the velocity. If you assume an object slips left, it is possible that
after you do all the calculations (assuming a rightward friction force) you instead find the
object slips right. And the reverse happens if you assume the object slips right, which makes
the actual slipping direction completely indeterminate.
You won’t have to worry about this paradox for Olympiads, which never contain setups that
trigger it. But it’s worth keeping in mind that friction is not just a simple law, but the
subject of an entire field of study called tribology, which is essential for engineering. In these
paradoxical cases, you would need to use a more refined model of friction to figure out what
actually happens.
[2] Problem 18 (Kalda). A rod with length ` is hinged to a ceiling with height h < `.
Underneath, a board is being dragged on the floor. The coefficient of (static and kinetic) friction
is µ1 between the board and rod, and µ2 between the board and floor. The rod is meant to stop
the board from being dragged to the right, no matter how hard or how quickly it is pulled. Is this
possible? If so, what are the conditions on the parameters that allow this to occur?
Solution. Let the angle between the rod and the vertical be θ. Consider the torque balance about
the hinge for the rod, giving the rod a mass of m and length `, a normal force between the board
and rode N and frictional force pulling to rod to the right of f ,
`
N ` sin θ = mg sin θ + f ` cos θ.
2
When friction is maximal and the board is about to move, f = µ1 N , so
mg sin θ
N= .
2(sin θ − µ1 cos θ)
It is impossible to move the board if µ1 is large enough to make N blow up, so the board is stuck if
√
`2 − h2
µ1 ≥ tan θ = .
h
15
Physically, what’s going on is that the harder you pull, the larger the normal force becomes, and
so the larger the friction can be; that’s how things get jammed.
We conclude with some questions that train three-dimensional thinking.
[2] Problem 19 (PPP 10). In Victor Hugo’s novel les Miserables, the main character Jean Valjean, an
escaped prisoner, was noted for his ability to climb up the corner formed by the intersection of two
vertical perpendicular walls. Suppose for simplicity that Jean has no feet. Let µ be the coefficient
of static friction between his hands and the walls. What is the minimum force that Jean had to
exert on each hand to climb up the wall? Also, for what values of µ is this feat possible at all?
Solution. Jean Valjean experiences two normal forces and two friction forces, one from each hand.
Each friction force must balance the other normal force, plus half the weight, so
2
ffric = N 2 + (mg/2)2 .
Assuming the friction is maximal, ffric = µN , we have

mg
N= p
2 µ2 − 1
and the force Jean Valjean exerts with each hand is
s
mg µ2 + 1
q
F = 2
N 2 + ffric = .
2 µ2 − 1
The feat is only possible if µ > 1.

[3] Problem 20 (PPP 69). A homogeneous triangular plate has threads of length h1 , h2 , and h3
fastened to its vertices. The other ends of the string are fastened to a common point on the ceiling.
Show that the tension in each thread is proportional to its length. (Hint: with the origin at the
point on the ceiling, let the vertices be at positions ri and express everything in vector form.)
Solution. Define the origin to be the attachment point on the ceiling, and let the vertices be at
positions ri . The tensions are along the ropes, so let them be Ti = −ηi ri . Force balance says
η1 r1 + η2 r2 + η3 r3 = mg.
Torque balance tells us that the center of mass of the triangle must lie directly below the attachment
point, and the center of mass is at
1
rCM = (r1 + r2 + r2 )
3
which means that
r1 + r2 + r3 ∝ g.
Thus, we know that the sum of the ri is in the vertical direction, and also that the weighted sum of
the ηi ri is in the same vertical direction. This is only possible if all the ηi are equal to each other,
which proves the desired result.
In case you’re not convinced, we can justify this in more detail. Let r1 + r2 + r3 = αg. Then
subtracting this equation from α/m times the force balance equation gives
X α
1 − ηi ri = 0.
m
i
16
The only way a nontrivial sum of three vectors can vanish is if they lie in a plane, which isn’t true
here. So each of the coefficients must vanish, so 1 − (α/m)ηi = 0, which means all the ηi are the
same, ηi = m/α.
[4] Problem 21 (KoMaL 2019, BAUPC 1998). Two identical uniform solid cylinders are placed on a
level tabletop next to each other, so that they are touching. A third identical cylinder is placed on
top of the other two.
(a) Find the smallest possible values of the coefficients of static friction between the cylinders,
and between a cylinder and the table, so that the arrangement can stay at rest.
(b) Repeat part (a) for spheres. That is, put three uniform solid spheres next to each other, with
their centers forming an equilateral triangle, and put a fourth sphere on top.
(c) Now return to part (a), and suppose the setup is frictionless. A force is applied directly to
the right on the leftmost cylinder, causing the entire setup to accelerate. Find the minimum
and maximum accelerations so that all three cylinders remain in contact with each other.
Parts (a) and (b) demonstrate an interesting point: it is possible for a collection of objects to resist
some force, even though a single one of those objects would begin moving even with an infinitesimal
applied force! This is a simple example of how granular materials, like sand, can give rise to emergent
phenomena that are hard to predict from analyzing individual grains alone. Understanding these
materials is a whole field of applied research.
Solution. (a) Call the top cylinder A, and the bottom ones B and C. Suppose the normal force
between the top cylinder and one of the bottom cylinders is N , and the friction force is f .
When the coefficient of static friction is at its smallest between the cylinders, f = µ1 N . B
and C are being pushed apart by A, so only the ground, gravity, and A are exerting forces on
B and C.
The total torque about C is 0, so the friction force from the ground is also f (towards B).
Then for the net horizontal force on C to be 0, by drawing out the forces and their directions
we get
f 1
f + f cos(π/6) = N sin(π/6) µ1 = = √ = 0.268
N 2+ 3
The normal force between the cylinder and the ground is N2 = N cos(π/6) + f sin(π/6) + mg.
The vertical force exerted on A from C is N cos(π/6) + f sin(π/6) = mg/2. Or, you can use
the fact that N2 = 3mg/2 by symmetry. Using the latter 2 equations:

cos π/6
N2 = 3 + sin π/6 f
µ1
µ1
f= N2 = 0.0893N2
3
So the coefficient of friction between the cylinders and the ground is 0.0893.
(b) All the spheres are being pushed apart, so the analysis above is the same except now the
angle is a bit different and the bottom balls exert a vertical force of mg/3 on the top ball
since there are 3 supports now.
The lines connecting the centers of the spheres form a tetrahedron by symmetry.
17
Let the length of the sides of a tetrahedron ABCD be `, and A being the point at the top √
(center of the top sphere). Then the distance from the centroid of triangle BCD to B is `/ 3
(use the fact that medians intersect in a ratio of 2 to 1 or draw a line from the centroid to
a side). Since AB has length `, the angle between the vertical√and the lines connecting the
centers of the top sphere and a bottom sphere is α = arcsin(1/ 3).
√ p
We will replace sin(π/6) with 1/ 3 and cos with 2/3 in the previous results. Thus with
the same analysis on a bottom ball with only the top ball acting on it, the friction coefficient
between the balls is:
sin α √ √
µ1 = = 3 − 2 = 0.318
1 + cos α
And copying the above formula except N2 = 4mg/3 and N cos(π/6) + f sin(π/6) = mg/3 will
get
cos α µ1
N2 = 4 + sin α f µ2 = ≈ 0.0795.
µ1 4
(c) Call the leftmost cylinder B, the top cylinder A, and the rightmost C. At the minimum
acceleration, NBC = 0 where A pushes them apart. Let θ = π/6. The accelerate C:
NAC sin θ = ma
Accelerate A:
(NBA − NAC ) sin θ = ma
Balance gravity:
(NBA + NAC ) cos θ = mg
Solve:
mg
2NAC = 4ma = √ − 2ma
3/2
g
amin = √
3 3
For the max acceleration, A is falling off so NAC = 0. Accelerate A:
NBA sin θ = ma
Balance A:
NBA cos θ = mg
Then
g
a= √
3
4 Extended Bodies
Idea 10
To deal with extended bodies, one can consider an infinitesimal piece of the body, or consider
a well-chosen symmetrical piece of it, or consider the whole body as a system. Often, multiple
approaches are needed.
18
Example 6
Find the tension in a circular rope of radius R spinning with angular velocity ω and mass
per length λ.
Solution
Consider an infinitesimal segment of the rope, spanning an angle dθ.
The mass of this segment is dm = Rλ dθ. The total force is downward, with magnitude
dθ
dF = 2T sin ≈ T dθ
2
where we used the small angle approximation. This is the centripetal force, so
dF = (dm) ω 2 R.
Combining these results yields T = R2 ω 2 λ.
Example 7
Find the distance d of the center of mass of a uniform semicircle of radius R to its center.
(Note that a semicircle is half of a circle, not half of a disc.)
Solution
This can be done by taking the setup of the previous problem, and taking a subsystem
comprising exactly half of the rope. In this case the net tension force is simply
F = 2T.
The total mass is m = πRλ, and the force must provide the centripetal force, so
F = (πRλ)(ω 2 d)
But we also know that T = R2 ω 2 λ as before, so plugging this in gives

2
d= R.
π
Alternatively, we could have worked in the frame rotating with the rope. The equations
would be the same, but instead we would say the tension balances the centrifugal force.
19
[1] Problem 22 (KK 2.22). A uniform rope of weight W hangs between two trees. The ends of the
rope are the same height, and they each make angle θ with the trees.
Find the tension at either end of the rope, and the tension at the middle of the rope.
Solution. Let the tension at the end be T0 , and T1 at the center. Considering the entire rope as
W
one system, we see that 2T0 cos θ = W , so T0 = 2 cos θ . Considering one half of the rope as a system,
W
we see T1 = T0 sin θ = 2 tan θ.
[3] Problem 23 (KK 2.24). A device called a capstan is used aboard ships to control a rope which is
under great tension.
The rope is wrapped around a fixed drum with coefficient of friction µ, usually for several turns.
The load on the rope pulls it with a force TA . Ignore gravity.
(a) Show that the minimum force TB needed to hold the other end of the rope in place is TA e−µθ ,
an exponential decrease.
(b) How does this result depend on the shape of the capstan, if we fix the angle θ between the
initial and final tension forces? Would the answer be the same for an oval, or a square?
(c) If θ = π, explain why the total normal and friction force of the rope on the drum is TA + TB .
Solution. (a) Consider a small piece of the rope that turns through an angle dθ. Using the
small angle approximation, the normal force must be T dθ, and the friction force must be dT .
Setting f = µN gives µT dθ = dT , or dT /T = µ dθ, and integrating gives the desired result.
(b) The infinitesimal reasoning above doesn’t care about the shape as long as it’s reasonably
smooth, so the answer for an oval is the same: just break it into pieces that turn through dθ
again. On the other hand, for a square one has sharp kinks where the normal force is singular,
in which case the answer won’t be as reliable.
(c) Consider the system consisting of the curved part of the rope. This system experiences a
force TA + TB from the straight part of the rope. But it is static, which means it must also
experience an equal and opposite force from the drum, which comes from integrating the
friction and normal forces along the contact surface.
20
That’s all you have to say, but we can also show this more explicitly. For concreteness, let
both tensions be vertical. We have a normal force and difference in tension forces
dN = T dθ, dT = −dffric
on a small piece dθ of the rope. The contribution to the vertical force on the drum is
dFy = dN sin θ + dffric cos θ = T sin θ dθ − dT cos θ = −d(T cos θ)
by the product rule. So the total vertical force is
Z Z π
Fy = dFy = − d(T cos θ) = −(TA + TB )
0
as expected. A very similar manipulation shows that Fx = 0.
[2] Problem 24 (F = ma 2018 B20). A massive, uniform, flexible string of length L is placed on
a horizontal table of length L/3 that has a coefficient of friction µs = 1/7, so equal lengths L/3
of string hang freely from both sides of the table. The string passes over the edges of the table,
which are smooth frictionless curves, of size much less than L. Now suppose that one of the hanging
ends of the string is pulled a distance x downward, then released at rest. Neither end of the string
touches the ground.
(a) Find the maximum value of x so that the string does not slip off of the table.
(b) For the case x = 0, draw a free body diagram for the string, indicating only the external
forces on the entire string. Do the forces balance?
(c) Would the answer change significantly if the table’s small edges had friction as well?
Solution. (a) The difference in weights is 2(M g/L)x, and needs to be balanced by the friction
force f . At the max value of x, f = µs N = µs M g/3 (the normal force at the top only holds
up the top of the string), so x = (µs /6)L = L/42.
(b) At first, it may seem that the forces don’t balance, because the normal force from the flat
part of the table only balances the weight of the string above it, leaving nothing to balance
the weight of the vertical parts of the string. But we must recall that there is an enormous
normal pressure at the smooth corners. The total normal force there is large enough so that
its vertical component holds up all of the string underneath it.
21
(c) Yes, the answer changes sufficiently no matter how small the edges are. This is because, as we
saw in part (b), there is a sizable normal force at the edges, since they alone are responsible
for holding up a significant part of the rope. So turning on a coefficient of friction at the edges
would yield a sizable friction force. (You can calculate it using problem 23.)
[3] Problem 25 (Morin 2.25). A rope rests on two platforms that are both inclined at an angle θ.
The rope has uniform mass density, and the coefficient of friction between it and the platforms is 1.
The system has left-right symmetry. What is the largest possible fraction of the rope that does not
touch the platforms? What angle θ allows this maximum fraction?
Solution. Let η be the fraction of the rope that does not touch the platforms. Split the rope into
the 3 obvious pieces (the left touching portion, the hanging portion, the right touching portion).
Let T be the tension at the boundaries (its the same on both sides by symmetry). Balancing forces
on the middle portion tells us
ηmg
2T sin θ = ηmg =⇒ T = .
2 sin θ
1−η
We see the friction force on the left piece is f = T + 2 mg sin θ, and the normal force is N =
1−η
2 mg cos θ. We have f ≤ N µ, so
ηmg 1−η 1−η

+ mg sin θ ≤ mgµ cos θ,
2 sin θ 2 2
or
η
+ (1 − η) sin θ ≤ (1 − η) cos θ,
sin θ
so some algebra reveals
cos θ − sin θ
η≤ tan θ .
cos θ + sin θ
Doing some more algebra turns this into
sin 2θ + cos 2θ − 1
η≤ .
sin 2θ + cos 2θ + 1
To maximize η, we need to maximize sin 2θ + cos 2θ, which implies θ = π/8 . The corresponding
√
value of η is 3 − 2 2.
Example 8
A chain is suspended from two points on the ceiling a distance d apart. The chain has a
uniform mass density λ, and cannot stretch. Find the shape of the chain.
22
Solution
First, we note that the horizontal component of the tension Tx is constant throughout the
chain; this just follows from balancing horizontal forces on any piece of it. Moreover, by
similar triangles, we have Ty = Tx y 0 everywhere.
Now considerp a small segment of chain with horizontal projection ∆x. The length of the
piece is ∆x 1 + y 02 which determines its weight, and this be balanced by the difference in
vertical tensions. Thus p
∆Ty = λg 1 + y 02 ∆x.
For infinitesimal ∆x, we have ∆Ty = Tx d(y 0 ) = Tx y 00 dx, so we get the differential equation
λg p
y 00 = 1 + y 02 .
Tx
Usually nonlinear differential equations with second derivatives are very hard to solve, but
this one isn’t because there is no direct dependence on y, just its derivatives. That means
we can treat y 0 as the independent variable first, and the equation is effectively first order in y 0 .
Writing y 00 = d(y 0 )/dx and separating, we have
dy 0
Z Z
λg
p = dx.
1+y 02 Tx
Integrating both sides gives

λgx
sinh−1 (y 0 ) = + C.
Tx
Choosing x = 0 to be the lowest point of the chain, the constant C is zero, and

0 λgx
y = sinh .
Tx
Integrating both sides again gives the solution for y,

Tx λgx
y= cosh
λg Tx
where we suppressed another constant of integration. This curve is called a catenary.
[4] Problem 26 (MPPP). A slinky is a uniform spring with negligible relaxed length, with mass m
and spring constant k.
(a) Find the shape of a slinky hung from two points on the ceiling separated by distance d. (Hint:
to begin, consider the mass and tension of a small piece of the spring with horizontal and
vertical extent dx and dy. Don’t forget that the slinky’s density won’t be uniform.)
(b) Suppose a slinky’s two ends are fixed, separated by distance d, and rotating uniformly with
angular frequency ω like a jump rope in zero gravity. Find the values of ω for which this
motion is possible, and the shape of the slinky in this case.
23
Solution. (a) Consider a small piece of the spring with mass dm, and horizontal and vertical
extent dx and dy. This piece of the spring has spring constant km/dm, which means
dx dy
Tx = km , Ty = Tx .
dm dx
By horizontal force balance, Tx is a constant, which means dx/dm is a constant; the same
amount of mass is contained within each horizontal interval. Thus
dx d
= .
dm m
Balancing vertical forces on this segment gives
dTy = y 00 Tx dx = g dm
and combining this with the previous result gives

mg
y 00 = .
kd2
We thus conclude that the shape is a parabola. Centering it at x = 0, we have
mgx2
y= .
2kd2
In particular, the lowest point of the parabola is a distance y(d/2) − y(0) = mg/8k below the
supports. (This solution is very similar to that of the example;
p the only difference is that
the weight of the segment is proportional to dx instead of 1 + y 02 dx. This is because the
slinky’s mass per length is not constant, while the chain’s was.)
(b) The only difference with respect to part (a) is that now we have a radial “gravity” force of
geff = −ω 2 y, because of the centrifugal acceleration in the frame rotating with the slinky.
Therefore,
mω 2
y 00 = − 2 y
kd
The solution is a sinusoid. For concreteness, let’s suppose one endpoint is at x = 0, imposing
y(0) = 0. Then r
mω
y(x) = y0 sin y .
k d
For the other endpoint to be fixed, y(d) = 0, we must have
r
m
ω = nπ.
k
If ω satisfies this condition for some n, then the slinky can rotate with uniform angular velocity,
and its shape is a sinusoid. The value of y0 is arbitrary.
Another way to say this is that the solutions we have found here are standing waves. The valid
values of ω, given the spring parameters, are just the standing wave frequencies. The fact that
ω doesn’t depend on d follows from dimensional analysis, and reflects the fact that stretching
the string further increases the tension and decreases the density, therefore increasing the
wave speed. These two effects cancel, keeping the standing wave frequencies the same.
24
Note that so far we’ve considered three cases: a hanging rope (in the example), a hanging slinky,
and a rotating slinky.pSo what about a rotating rope? Unfortunately, the differential equation
describing it is y 00 ∝ y 1 + y 02 , since the centrifugal acceleration is proportional to y. And unlike
the example, this is a genuine nonlinear second order differential equation. Mathematica reports
that the solution is not an elementary function, but rather an inverse elliptic integral. Unfortunately,
that’s just what happens most of the time.
Example 9
A uniform spring of spring constant k, mass m, and relaxed length L is hung from the ceiling.
Find its length in equilibrium, as well as its center of mass.
Solution
Problems like this contain subtleties in notation. For example, if you talk about “the piece
of the slinky at z”, this could either mean the piece that’s actually at this position in
equilibrium, or the piece that was originally at this place in the absence of gravity. Talking
about it the first way automatically tells you where the piece is now, but talking about it
the second way makes it easier to keep track of, because then the z of a specific piece of the
spring stays the same no matter what happens.
In fluid dynamics, these are known as the Eulerian and Lagrangian approaches, respectively.
If you don’t use one consistently, you’ll get nonsensical results, and it’s easy to mix them up.
There are many ways to solve this problem, but I’ll give one that reliably works for me.
We’re going to use the Lagrangian approach, and avoid confusion with the Eulerian approach
by breaking the spring into discrete pieces. Let the spring consist of N 1 pieces, of masses
m/N , spring constants N k, and relaxed lengths L/N .
The ith spring from the bottom has tension (i/N )mg, and thus is stretched by
1 i mg
∆Li = mg = i.
kN N kN 2
The total stretch is
N Z N
X mg mg
∆Li = i di = .
kN 2 0 2k
i=1
This makes sense, since the average tension is mg/2. To find the center of mass, note that
the j th spring is displaced downward by a distance
N
j2

X mg
∆yj = ∆Li = 1− 2
2k N
i=j
downward from its position in the absence of gravity. The center of mass displacement is
N N Z N
j2

1 X 1 X 1 2
∆yCM = ∆yj ∝ 1− 2 = 3 N 2 − j 2 dj =
N N N N 0 3
j=1 j=1
25
so restoring the proportionality constant gives

mg
∆yCM = .
3k
If you want to test your understanding of slinkies, you can also try doing this problem with
the Eulerian approach. The first steps would be finding a relation between the density ρ(z)
and tension T (z) from Hooke’s law, and finding out how to write down local force balance.
Remark
In this problem set, we’ve given some examples involving static, continuous, one-dimensional
objects such as strings and ropes. The general three-dimensional theory of elasticity is
mathematically quite complicated, but extremely important in engineering. For more about
this subject, which requires comfort with tensors, see chapters 6 through 11 of Lautrup. It
is also covered in chapters II-31, II-38, and II-39 of the Feynman lectures.
Remark: Why Use Torque?
Here’s a seemingly naive question. Why is the idea of torque so incredibly useful in physics
problems, even though in principle, everything can be derived from F = ma alone? Why is
it almost impossible to solve any nontrivial problem without referring to torques, and how
would a student who’s never heard of torque come up with it in the first place?
We don’t need torque to analyze the statics of a single, featureless point particle. Torque
only became useful in this problem set when we started analyzing rigid bodies with spatial
extent. The reason we couldn’t reduce torque balance to force balance easily is because the
internal forces in these bodies, which maintain their rigidity, are generally very complicated.
It’s possible to derive torque balance from force balance in simple cases, but it’s subtle. For
example, suppose we try to make a rigid body by attaching two ideal springs of infinite spring
constant to masses m and a pivot, as shown.
If forces F1 and F2 are applied, we would like to show that the setup if static only if
the torques are balanced, r1 F1 = r2 F2 , using only force balance on the masses. But our
derivation fails at the very first step: the forces of ideal springs only point along the direction
of the spring. Therefore, they can’t balance the forces at all!
The problem is that this setup isn’t actually rigid; it will bend at the first mass. We could
make the setup rigid by replacing the spring with a rod, because, as we saw in example 2,
a rigid rod can exert downward shear forces on the first mass and upward shear forces on
the second. But in that case, it becomes nontrivial to decide when the rod can remain in
26
equilibrium. You would need to decompose it microscopically, to see what is responsible for
the shear forces in the first place, and balance forces on all those microscopic pieces. We
tried to make a simple rigid body out of two masses, but now we need to keep track of the
rod’s infinitely many degrees of freedom!
Luckily, there’s a simple modification of our original setup that works. Consider a triangle
whose sides are made of ideal massless springs, pivoted at one vertex, as shown.
This actually is a rigid body, and the internal forces are simple: each spring just carries some
tension. Let T be the tension in the spring between the two masses, opposite the pivot. The
forces on each mass can be decomposed into a component parallel to the sides r1 and r2 ,
and a component perpendicular to those sides. The former components can be balanced by
adjusting the tensions in the other two springs. The perpendicular components can only be
balanced by the tension T , from which we conclude
F⊥,1 = T sin θ1 , F⊥,2 = T sin θ2 .
Upon eliminating T and using the law of sines, we conclude that force balance on the masses
is only possible if r2 F⊥,1 = r1 F⊥,2 , which is precisely the statement of torque balance.
Tricky, right? This is the simplest example of a rigid body I know of, and in general, the
internal forces can be much more complicated. The miracle of torque is that it automatically
takes care of all those details for us, no matter what they are.
5 Pressure and Surface Tension

Example 10
A sphere of radius R contains a gas with a uniform pressure P . Find the total force exerted
by the gas on one hemisphere.
27
Solution
The pressure provides a force per unit area orthogonal to the sphere’s surface, so the
straightforward way to do this is to integrate the vertical component of the pressure force
over a hemisphere. However, there’s a neat shortcut in this case.
Momentarily forget about the sphere and just imagine we have a sealed hemisphere of gas at
pressure P . The net force of the gas on the hemisphere must be zero, or else it would just
begin shooting off in some direction, violating conservation of momentum. So the force on
the curved face must balance the force on the flat face, which is πR2 P . The same logic must
hold for the sphere, since the forces on the curved face are the same, so the answer is πR2 P .
This trick works whenever one has a uniform outward pressure on a surface, and it’ll come in
handy for several future problems. For example, it’s the quick way to do F = ma 2018 B24.
Idea 11
The surface of a fluid carries a surface tension γ. If one imagines dividing the surface into two
halves, then γ is the tension force of one half on the other per length of the cut. Specifically,
dF = γ ds × n̂
which means the tension acts along the surface and perpendicular to the cut.
Example 11
A soap bubble of radius R and surface tension γ is in air with pressure P , and contains air
with pressure P + ∆P . Compute ∆P .
Solution
We use the result of the previous problem to conclude that the force of one hemisphere
on another is πR2 ∆P . This must be balanced by the surface tension force. By imagining
cutting the surface of the bubble in half, the surface tension force is γL where L is the total
length of the surface connecting the hemispheres.
At this point, we can write L = 2πR, giving

2γ
∆P = .
R
This is called the Young–Laplace equation. However, in this particular case, this is not the
right answer. The reason is that we should actually take L = 4πR because the surface tension
is exerted at both the inside and outside surfaces of the bubble wall, and thus the answer is
4γ
∆P = .
R
The increased pressure inside balances the surface tension, which wants to collapse the bubble.
28
If you’re confused about why L = 4πR, you can also think about it in terms of energy. Surface
tension arises from the fact that it costs energy to take soapy water and stretch it out into a
surface, because this breaks some of the attractive intermolecular bonds. The Young–Laplace
equation would give the correct answer for a ball of soapy water. But for a bubble of soapy
water, twice as much soapy water/air surface is created. So the energy cost is double, and
the force is double.
[2] Problem 27. One can also derive the Young–Laplace equation by just considering energy. Suppose
the bubble radius changes by dr. The energy of the bubble changes for two reasons: first, there is
net ∆P dV work from the two pressure forces, and there is the γ dA surface tension energy cost. At
equilibrium, the energy must be at a minimum, so it should not change at all under an infinitesimal
displacement. Using this idea, rederive the Young–Laplace equation.
Solution. The work done by the surface tension should be balanced by the work done by the
pressure difference. Noting that the total surface area is 8πR2 , we have

4 3
∆P dV = ∆P d πR = ∆P (4πR2 ) dR = d(8πR2 γ) = 16πγR dR
3

4γ
∆P = .
R
Remark
The general idea used in the above problem, of thinking about how energy would change
in order to find an unknown force, is known as the principle of virtual work. The principle
works for all kinds of forces. For example, if the bubble was charged, it would grow due to
electrostatic repulsion, and the new equilibrium radius could be found using virtual work.
Remark
Note that the Young–Laplace equation we gave above only holds for spherical surfaces. More
generally, a surface has two principle radii of curvature R1 and R2 at each point. These are
both equal to R for a sphere of radius R, but for, e.g. a cylinder of radius R, one is equal to
R and the other is infinity. For general surfaces, the Young–Laplace equation is

1 1
∆P = γ +
R1 R2
where the Ri can each be positive or negative, depending on the direction of curvature.
[2] Problem 28 (Kalda). Consider two soap bubbles which have stuck together. The part of the soap
film that separates the interior of the first bubble from the outside air has radius of curvature R.
The part that separates the interior of the second bubble from the outside air has radius of curvature
2R. What is the radius of curvature of the part which separates the bubbles from each other?
29
Solution. The key is that the Young–Laplace equation should hold for every point on the surface
since the surface tension and pressure should balance for every infinitesimal surface element. The
gauge pressure (pressure minus the atmospheric pressure) inside the first bubble is P1 = 4γ/R,
and for the second P2 = 4γ/(2R). Thus the pressure difference between the two bubbles is
∆P = 2γ/R = P2 , giving a radius of curvature of 2R for the part separating the bubbles.
[3] Problem 29 (MPPP 67). When a long, straight sausage is cooked, it always splits “lengthwise”
and never “across”. The two modes of splitting are shown as dotted lines below.
Explain this observation, assuming the thickness of the sausage skin is uniform, and hence can
support a constant surface tension before breaking. (Hint: model the sausage as a cylinder of length
L capped by hemispheres of radius R L, and consider the surface tension needed to prevent the
two modes of splitting mentioned, once an excess pressure P builds up inside the sausage.)
Solution. Let the pressure differential from inside the sausage to outside be P . Cutting it across
so the cross section is a circle tells us that the surface tension γa will exert a force F = (2πr)γa on
each end since F = γ`. Using the trick from example 10, it must balance the force F = πr2 P , so
γa = P r/2.
Lengthwise, the cross section has perimeter 2L + 2πr ≈ 2L. If we apply the trick to each
half-cylinder, we find that the pressure force is F = (2rL)P , so balancing forces gives γL = P r.
Since this is a greater requirement on the surface tension, the sausage will break lengthwise, as we
regularly observe in the kitchen.
[4] Problem 30. Two coaxial rings of radius R are placed a distance L apart from each other in
vacuum. A soap film with surface tension γ connects the two rings.
(a) Derive a differential equation for the shape r(z) of the film, and solve it.
(b) Show that for sufficiently large L, there are no solutions. If L is increased to this value, what
happens to the film?
(c) Using a computer or calculator, find the largest possible value of L.
We’ll consider surface tension in more detail in T3.
Solution. (a) Consider a segment of the bubble between z and z +√ dz. The net forces exerted
by surface tension on both sides along the z-direction 02
√ are 2πrγ/ 1 + r . To balance forces
02
in the z-direction for each segment, the quantity r/ 1 + r must be independent of z, so
r2 = A2 (1 + r02 )
for some constant A. Separating and integrating, we have

Z Z
A dr
dz = √
r 2 − A2
30
and substituting r = A cosh u and integrating yields

−1 z+C
z + C = A cosh (r/A), r = A cosh
A
for another constant C. Setting the rings to be at z = ±L/2, we have C = 0. The quantity
A is the minimum radius, which occurs by symmetry at z = 0.
You may have noticed that the answer is a catenary, which is the same as the answer to
example 8. The reason is that both problems can be solved by minimizing a similar quantity.
Here, we want to find the function r(z) that minimizes the area,
Z p
A = 2πr 1 + r02 dz
where the value of r at two given values of z is fixed. In that example, we wanted to find the
shape y(x) of the chain that minimizes the gravitational potential energy,
Z p
U = λ y 1 + y 02 dx.
This function is extremely similar in form, which explains why the form of the solution is
similar. But there’s an important physical difference: the length of the chain is fixed, and
you need to specify it to determine the solution. (To see how this constraint can be imposed
with Lagrange multipliers, see here.) By contrast, the soap bubble is more free to vary. That
explains why, in part (b), you can sometimes have no solution for a soap bubble at all. In
those cases, the middle of the film can just get thinner and thinner, always decreasing the
area, until it pinches off into two separate pieces.
(b) We introduced the parameter A above, which describes the shape of the solution. It is fixed
by R and L by the requirement that the bubble fit the rings,
L
R = A cosh .
2A
Now, we wish to find the largest L so that there exists some A so that the left-hand side
can be R. This is a somewhat annoying optimization problem. It’s clearer to note that by
dimensional analysis, the only invariant thing is the single dimensionless ratio R/L. (A doesn’t
count as a dimensionful parameter, because it’s fixed by R and L.) So finding the largest L
for fixed R is equivalent to finding the smallest R for fixed L.
But this is now easy, because we already have R as a function of L, which is fixed, and A,
which can vary. By graphing the function R(A), we see it has a single minimum, so there is
indeed a minimum possible R/L and hence a maximum possible L/R.
(c) Setting the derivative dR/dA to zero, the minimum occurs when
2A L
= tanh .
L 2A
This equation cannot be solved analytically. Using a calculator and the techniques of P1, we
find the maximum possible L is about 1.33R. After this point, it’s favorable from an area
standpoint for the bubble to just pinch off, and collapse to two separate circles bounded by
the rings.
31
Example 12
A solid ball of radius R, density ρ, and Young’s modulus Y rests on a hard table. Because
of its weight, it deforms slightly, so that the area in contact with the table is a circle of radius r.
Estimate r, assuming that it is much smaller than R.
Solution
Recall from P1 that the Young’s modulus is defined by

Y = =
and has dimensions of pressure. By dimensional analysis, you can show that
r = R f (ρgR/Y )
but dimensional analysis alone can’t tell us anything more about f . Moreover, an exact
analysis using forces would be very difficult, because different parts of the ball are compressed
in different amounts, and in different directions; there’s little symmetry here. Instead, we
resort to an order of magnitude approach. To balance gravity, we need a typical stress of
stress ∼ M g/r2 ∼ ρR3 g/r2
in the deformed region. This deformed region has typical length r. To estimate the strain,
note that if the ball were not deformed at all, it would be standing a height δ ∼ r2 /R taller,
as you can show by the Pythagorean theorem. Thus,
strain ∼ δ/r ∼ r/R.
Using the definition of the Young’s modulus, we conclude

1/3
ρgR
r∝R .
Y
By the way, there’s a whole field of study devoted to figuring out how the normal and
other forces behave for realistic, deformable solids, known as contact mechanics. For an
authoritative reference, see Contact Mechanics and Friction by Popov.
[4] Problem 31. EFPhO 2006, problem 5. A tough problem on a deforming object.
32
33
Mechanics III: Dynamics

Chapters 3 and 5 of Morin cover dynamics, energy, and momentum. Alternatively, see chapters 2
and 3 of Kleppner and Kolenkow, or chapters 4 and 6 of Wang and Ricardo, volume 1. For fun, see
chapters I-9 through I-14 of the Feynman lectures. There is a total of 82 points.
1 Blocks, Pulleys, and Ramps

Idea 1
To solve dynamics problems with constraints, it’s easiest to first write the constraint in
terms of coordinates (e.g. “conservation of string” for pulleys, or stationarity of the CM for
an isolated system), then differentiate to get constraints on the velocity and acceleration.
Questions of this type are generally straightforward, as long as you write down the correct
equations. The trickiest part is often solving the equations, which can get messy.
Example 1: Morin 3.30
Find the acceleration of the masses in the Atwood’s machine shown below.
Neglect friction, and treat all pulleys are massless.
Solution
Let x and x0 be the amounts by which the left and right mass have moved down, and number
the pulleys 1 through 4 from left to right, and the strings 1 through 3 from left to right.
Pulley 4 is stationary, so conservation of string 3 means that pulley 3 moves up by x0 /2. Next,
conservation of string 2 means that pulley 2 moves up by x0 /4. Finally, conservation of string
1 implies that pulley 1 moves up by x0 /8, so our final conservation of string constraint is
x0
x=−
8
which upon applying the derivative twice gives
a0
a=− .
8
1
Now consider the tensions Ti in the strings. We know that

2T1 2T3
a=g− , a0 = g − .
m m
Since pulley 3 is massless, the forces on it must balance, so T2 = 2T3 . Similarly T1 = 2T2 ,
so T1 = 4T3 . We hence have a system of three equations in three unknowns (T1 , a, and a0 ),
which can be solved straightforwardly to give
56 7
a0 = g, a=− g.
65 65
[2] Problem 1 (Morin 3.2). Consider the double Atwood’s machine shown below.
Assuming all pulleys are massless, and neglecting friction, find the acceleration of the mass m1 .
Solution. Let the downward accelerations be a1 , a2 , and a3 . By conservation of string,
2a1 + a2 + a3 = 0.
Let T be the tension in the lower pulley. Then we have
m1 g − 2T = m1 a1 , m2 g − T = m2 a2 , m3 g − T = m3 a3 .
Therefore,
0 = 2a1 + a2 + a3 = 2(g − 2T /m1 ) + g − T /m2 + g − T /m3 = 4g − T (4/m1 + 1/m2 + 1/m3 ),
so that
g
T = .
1/m1 + 1/4m2 + 1/4m3
Thus, we conclude
2g m1 (m2 + m3 ) − 4m2 m3
a1 = g − 2T /m1 = g − = g .
1 + m1 /4m2 + m1 /4m3 4m2 m3 + m1 (m2 + m3 )
[2] Problem 2 (KK 2.15). Consider the system of massless pulleys shown below.
2
The coefficient of friction between the masses and the horizontal surfaces is µ. Show that the tension
in the rope is
(µ + 1)g
T = .
2/M3 + 1/2M1 + 1/2M2
Solution. Let the acceleration of block 1 be a1 to the right, and the acceleration of block 2 be a2
to the left, and the acceleration of block 3 be a3 down. We see that 2a3 = a1 + a2 by conservation
of string. We also see that M3 g − 2T = M3 a3 , T − M1 gµ = M1 a1 , and T − M2 gµ = M2 a2 , or
M3 g − 2T T − M1 gµ T − M2 gµ
2a3 = 2 , a1 = , a2 = .
M3 M1 M2
Thus, we have
2(g − 2T /M3 ) = (T /M1 − gµ) + (T /M2 − gµ) =⇒ 2g(1 + µ) = T (1/M1 + 1/M2 + 4/M3 ).
Solving for T gives the result.

[2] Problem 3 (KK 2.20). Consider the machine shown below, which we encountered in M2.
Show that the acceleration of M1 when the external force F is zero is

M 2 M3 g
a=− .
M1 M2 + M1 M3 + 2M2 M3 + M32
Solution. Let the acceleration of M2 with respect to M1 be w. Then, we see that T = M2 (w + a),
and M3 g − T = M3 w, so adding gives us M3 g = M2 (w + a) + M3 w. Since the acceleration of the
center of mass is zero,
M1 + M3 M1 + M2 + M3
M1 a + M2 (w + a) + M3 a = 0 =⇒ w = − a − a = −a .
M2 M2
Thus,
M1 + M3 M1 + M2 + M3
M3 g = M2 (−a) − aM3 ,
M2 M2
3
or
M 2 M3 g
a=− ,
M2 (M1 + M3 ) + M3 (M1 + M2 + M3 )
as desired.
By the way, if you tried to solve the problem by considering just forces, there’s a subtlety; it’s
easy to forget that there must be a force on M1 due to the normal force of the rope on the pulley.
(This force has to be there, or else the forces on the massless rope wouldn’t balance.) Indeed, you
learned about these forces in the preliminary problem set. The solution above implicitly took this
into account, by using the fact that the center of mass doesn’t move.
[3] Problem 4. A block of mass m is held motionless on a frictionless plane of mass M and angle of
inclination θ. The plane rests on a frictionless horizontal table.
(a) When the block is released, what is the horizontal acceleration of the plane?
(b) Assume the block starts a distance d above the table. Using results from part (a), what is the
horizontal velocity of the block just before it reaches the floor?
(c) Find the speed of the block after it reaches the floor by applying energy and momentum
conservation to the entire process.
(d) Your results for parts (b) and (c) should not match. What’s going on?
Solution. (a) Applying Newton’s second law to the plane gives
M a = N sin θ.
Next, work in the noninertial frame of the plane, where the block only moves parallel to the
plane. Balancing the forces perpendicular to the plane gives
N = mg cos θ − ma sin θ.
It’s now straightforward to eliminate N and thereby solve for a, which gives
mg sin θ cos θ
a= .
M + m sin2 θ
By the way, it’s easy to get confused on this problem if you focus too much on the block, because
its motion is somewhat confusing; you have to decompose it into motion parallel to the plane,
and motion of the plane itself. Once you do that, the problem can be solved straightforwardly.
The solution above is especially short because it never considers the acceleration of the block
parallel to the plane, which isn’t required to get the answer.
(b) Since the only horizontal forces in the problem are between the block and plane, the horizontal
acceleration of the block is
M
ab = a.
m
Thus, the relative acceleration of the two is
M +m
arel = a + ab = a.
m
The block goes off the plane
p once the two have moved a relative horizontal distance of d/ tan θ,
which takes a time t = 2d/arel tan θ. At this point the block has a horizontal velocity
p M cos θ
vb = ab t = 2gd p .
(M + m sin2 θ)(M + m)
4
(c) By momentum conservation, the final horizontal speeds obey
M vp = mvb
while by energy conservation,

1 1
M vp2 + mvb2 = mgd.
2 2
Combining the two and solving gives
r
p M
vb = 2gd .
M +m
(d) It turns out that both results are correct, but they’re the answers to different questions. Note
that at the instant the block gets to the bottom of the plane, its velocity isn’t horizontal, but
right after it’s off the plane, its velocity must be exactly horizontal. This requires a rather
large vertical impulse.
Depending on how the plane and block are constructed, there are several possibilities. If the
plane abruptly ends, and the block immediately begins moving horizontally, then we have
an inherently inelastic process. The vertical kinetic energy mvy2 /2 of the block is lost, so the
answer to part (c) doesn’t apply, and the answer to part (b) is correct. (Another possibility,
if the ball is very bouncy, is that the sign of its vy will flip, and it’ll bounce off the floor. In
this case, the answer to part (b) is still correct. Energy is conserved now, but the answer to
part (c) is still wrong because it assumes the final velocity of the block is horizontal.)
On the other hand, if the plane ends in a transition region, where θ smoothly goes to zero,
then ideally energy remains conserved, and the answer of part (c) applies. In this region, a
strong normal force reorients the velocity to be horizontal, supplying both a large horizontal
and vertical impulse. As a result, the answer to part (b) doesn’t apply.
This subtlety about how inclined planes end applies to lots of Olympiad problems. The usual
assumption seems to be that option (c) holds, but in reality it depends sensitively on how the
plane and block are made. In fact, in practice you can lose a lot of energy even if there’s a
smooth curve at the end, if that curve is not gradual enough.
2 Momentum
Idea 2
The momentum of a system is
X
P= mi vi = M vCM .
i
In particular, the total external force on the system is M aCM , and if there are no external
forces, the center of mass moves at constant velocity.
5
Example 2
A massless rope passes over a frictionless pulley. A monkey hangs on one side, while a bunch
of bananas with exactly the same weight hangs from the other side. When the monkey tries
to climb up the rope, what happens?
Solution
Remarkably, the answer doesn’t depend on how the monkey climbs, whether slowly or
quickly, or symmetrically or not! The total vertical force on the monkey is T − mg, so the
acceleration of the center of mass of the monkey is T /m − g. But since the tension is uniform
through a massless rope, the acceleration of the bananas is also T /m − g. Therefore, the
monkey and bananas rise at the same rate, and meet each other at the pulley.
Now here’s a question for you: compared to climbing up a rope fixed to the ceiling, climbing
up to the pulley takes twice as much work, because the bananas are raised too. But in both
cases, isn’t the monkey applying the same force through the same distance? Where does the
extra work come from? (The answer involves the ideas at the end of this problem set.)
Example 3: KK 3.14 / INPhO 2014.5
Two men, each with mass m, stand on a railway flatcar of mass M initially at rest. They
jump off one end of the flatcar with velocity u relative to the car. The car rolls in the opposite
direction without friction. Find the final velocities of the flatcar if they jump off at the same
time, and if they jump off one at a time. Generalize to the case of N 1 men, with a total
mass of mtot .
Solution
In the first case, by conservation of momentum, we have
M v + 2m(v − u) = 0
where v is the final velocity of the flatcar, so

2mu
v= .
M + 2m
In the second case, by a similar argument, we find that after the first man jumps,
mu
v1 = .
M + 2m
Now transform to the frame moving with the flatcar. When the second man jumps, he
imparts a further velocity v2 = mu/(M + m) to the flatcar by another similar argument. The
final velocity of the flatcar relative to the ground is then

1 1
v = v1 + v2 = mu + .
M + 2m M + m
6
It might be a bit disturbing that the final speeds and hence energies of the flatcar are
different, even though the men are doing the same thing (i.e. expending the same amount of
energy in their legs to jump) in both cases.
The reason for the difference is that in the second case, the second man to jump ends up
with less energy, since the velocity he gets from jumping is partially cancelled by the existing
velocity v1 . So the extra energy that goes into the flatcar corresponds to less kinetic energy
in the men after jumping, which would ultimately have ended up as heat after they slid to a
stop. Accounting properly for the kinetic energy of everything in the system solves a lot of
paradoxes involving energy, as we’ll see below.
In the case of many men, by similar reasoning we have

mtot
v= u
M + mtot
in the first case, while in the second case the answer is the sum
N
X mtot u 1
v= .
N M + (i/N )mtot
i=1
This can be converted into an integral, by letting x = i/N , in which case ∆x = 1/N and
Z 1
X mtot u mtot u M + mtot
v= ∆x ≈ dx = log u.
M + xmtot 0 M + xmtot M
i
Note that this is essentially the rocket equation, which we’ll derive in a different way in M6.
[2] Problem 5 (KK 4.11). A flexible chain of mass M and length ` is suspended vertically with its
lowest end touching a scale. The chain is released and falls onto the scale. Find the reading on the
scale when a length of chain x has fallen.
Solution. Because the chain is flexible, each link just crumples when it hits the ground, without
pulling the rest of the chain downward. In other words, the assumption of ideal flexibility implies
the tension in the chain vanishes, so that the vertical part of the chain is always in free fall.
√ √
Now, the lowest end of the chain is moving with velocity 2gx, so in time dt, a mass M 2gxdt/`
falls on to the scale, so the change in momentum of that piece is (2M gx/`)dt. Thus, we need a force
2M gx/` to stop the links that are falling on the scale. In addition, there must be a force M gx/` to
balance the weight of the chain that’s already lying on the scale, for a total of 3M gx/` .
This is nice and elegant, but is it true? The result is actually pretty accurate, as you can see
from experimental data here. The deviation from the expected result is because no chain is perfectly
flexible. Since the chain has to bend at the spot it hits the scale, some tension is produced, which
pulls down the rest of the chain slightly faster than free fall.
This has a connection to the “inherently inelastic” processes mentioned later in the problem set.
The fastest possible fall corresponds to the case where energy is conserved, i.e. when all the kinetic
energy of each link hitting the ground is nearly transferred through tension to the still falling part
7
of the chain. The answer we gave above corresponds to the slowest possible fall, where each link
collides perfectly inelastically with the ground. For a flexible chain, the latter is closer to reality.
[3] Problem 6. Some qualitative questions about momentum.
(a) A box containing a vacuum is placed on a frictionless surface. The box is punctured on its
right side. How does it move immediately afterward?
(b) You are riding forward on a sled across frictionless ice. Snow falls vertically (in the frame of
the ice) on the sled. Which of the following makes the sled go the fastest or the slowest?
1. You sweep the snow off the sled, directly to the left and right in your frame.
2. You sweep the snow off the sled, directly to the left and right in the ice frame.
3. You do nothing.
Solution. (a) Consider the system of the air plus the box. The air flows to the left, so to keep
momentum conserved, the box moves to the right.
(b) Case (1) is clearly the fastest, as the ice ends up with none of the sled’s horizontal momentum;
in the sled’s frame this looks like propelling the ice backwards.
Then we just have to decide between (2) and (3). In both cases, sweeping the snow off (or
doing nothing) doesn’t change the sled’s horizontal momentum. However, in case (3), as the
sled gets heavier, each bit of new snow landing on it has less effect. So (2) is the slowest, and
01^
(3) is in the middle.

[3] Problem 8 (Kalda). A block is on a ramp with angle α and coefficient of friction µ > tan α. The
ramp is rapidly driven back and forth so that its velocity vector u is parallel to both the slope and
the horizontal and has constant modulus v.
The direction of u reverses abruptly after each time interval τ , where gτ v. Find the average
velocity w of the block. (Hint: as mentioned in M1, it’s best to work in the frame of the ramp,
because it causes the friction, even though this introduces fictitious forces.)
Solution. Work in the frame of the ramp and orient the x axis along u and the y axis along the
ramp. At all times, the acceleration due to friction is µg cos α and the acceleration due to gravity
is g sin α. Every time period τ , an impulsive fictitious force changes wx by ±2v. Since gτ v, the
total acceleration during the time period τ due to the friction and gravitational forces is negligible
compared to this change. Assuming for now that wx is symmetric so that wx = 0, this means
|wx | ≈ u at all times.
Now consider wy . In the steady state, the acceleration due to friction must be balanced by the
acceleration due to gravity, so
wy
p µg cos α = g sin α
wy 2 + u2
8
which yields the answer,

u
w = wy ŷ, wy = p .
µ2 cot2 α −1
Note that this diverges when µ = tan α, because at that point the friction is not strong enough to
prevent the block from accelerating down the ramp indefinitely. For µ > tan α, we reach a steady
state where only a portion of the friction is directed vertically, due to the horizontal speed, and
that portion balances gravity.
This also allows us to argue that wx = 0. If wx is not zero, |wx | will be higher during one of the
two halves of the cycle. But during that half, a greater share of the frictional acceleration will be
directed against the wx motion, tending to move wx to zero.
This seemingly weird problem actually has real-world applications! The point here is that you
can make a block slide down a ramp even if friction would prevent it from doing so, and moreover
make it slide at a controlled speed. This technique is used in factories, in the form of vibratory
conveyors. In fact, a more complex vibration pattern can even make something slide up a ramp!
[4] Problem 9 (Morin 5.21). A sheet of mass M moves with speed V through a region of space that
contains particles of mass m and speed v. There are n of these particles per unit volume. The
sheet moves in the direction of its normal. Assume m M , and assume that the particles do not
interact with each other.
(a) If v V , what is the drag force per unit area on the sheet?
(b) If v V , what is the drag force per unit area on the sheet? Assume for simplicity that the
component of every particle’s velocity in the direction of the sheet’s motion is exactly ±v/2.
(c) Now suppose a cylinder of mass M , radius R, and length L moves through the same region
of space with speed V , and assume v = 0 and m M . The cylinder moves in a direction
perpendicular to its axis. What is the drag force on the cylinder?
Parts (a) and (b) are a toy model for the two regimes of drag, mentioned in M1. However, it
shouldn’t be taken too seriously, because as we’ll see in M7, the typical velocity that separates
the two types of behavior doesn’t have to be of order v. Instead, it depends on how strongly the
particles interact with each other.
Solution. (a) We can set v = 0. In time t, an area A hits nAV t particles, and the total change
in momentum of these particles is (nAV t)m(2V ), so the pressure is 2nmV 2 .
(b) Let’s say the sheet is moving to the right. In the frame of the sheet, the particles are moving
at velocity V ± v/2. The particles hitting the sheet from the right will have velocity v/2 + V ,
and from the left v/2 − V . From the right in time dt, there will be 12 nA(v/2 + V )dt particles
hitting the sheet (with the 12 coming from other particles moving away from the sheet), each
with impulse 2m(v/2 + V ). Thus the pressure will be nm(v/2 + V )2 from the right, and
replacing V with −V gives a pressure of nm(v/2 − V )2 from the left. Thus the total pressure
on the sheet is 2nmV v.
(c) Work in cylindrical coordinates, with θ = 0 along the direction of the cylinder’s motion. For
a segment dθ, we have
collisions particles volume swept out
= = nV LR cos θ dθ
time volume time
9
where L is the length of the cylinder. To calculate the rebound velocity, it’s best to work
in the frame of the cylinder. In this case, the particles come in with vertical velocity −V ,
and then bounce off elastically, ending up with vertical velocity V cos 2θ. So the impulse per
collision is mV (1 + cos 2θ). The drag force is
Z π/2 Z π/2
2
F = mV (1 + cos 2θ)nV LR cos θ dθ = nmV LR (1 + cos 2θ) cos θ dθ.
−π/2 −π/2
The integral can be done straightforwardly using either the cosine double angle identity, or
decomposing into complex exponentials, yielding 8/3, so

2 4
F = (2nmV )(LR) .
3
Compare this to the answer to part (a). The force is quadratic in V for the same reason, but
now the area is replaced by an effective area (4/3)LR. This is slightly less than the actual
area 2LR, since the surface is curved, and hence more aerodynamic.
You can also get a more “realistic” result by averaging over a Maxwell–Boltzmann distribution for
the molecular speeds, as introduced in T1. But this is a lot more work, and the simpler calculation
done in this problem gives all the essential insight.
3 Energy
Idea 3
The work done on a point particle is
Z
W = F · dx
and is equal to the change in kinetic energy, as you showed in P1.
Remark: Dot Products

The dot product of two vectors is defined in components as
v · w = vx wx + vy wy + vz wz
and is equal to |v| |w| cos θ where θ is the angle between them. For example, if A and B are
the sides of a triangle,
|A + B|2 = (A + B) · (A + B) = A2 + B 2 + 2AB cos θ.
Since the left-hand side is the length squared of the third side of the triangle, we’ve proven
the law of cosines. (Or, if you accept the law of cosines, you could regard this as a proof
that the dot product depends on cos θ as claimed.)
Like the ordinary product, the dot product obeys the product rule. For example,
d
(v · w) = v̇ · w + v · ẇ.
dt
10
Using this, it’s easy to generalize the derivation of the work-kinetic energy theorem in P1 to
three dimensions; we have
1 1 dx dv
d(v 2 ) = d(v · v) = v · dv = · dv = · dx = a · dx
2 2 dt dt
and this is equivalent to the desired theorem. As you can see, it’s all basically the same, since
the product and chain rule manipulations work the same way for vectors and scalars.
Example 4: IPhO 1996 1(b)
A skier starts from rest at point A and slowly slides down a hill with coefficient of friction µ,
without turning or braking, and stops at point B. At this point, his horizontal displacement
is s. What is the height difference h between points A and B?
Solution
Since the skier begins and ends at rest, the change in height is the total energy lost to friction,
Z
mgh = ffric ds
where the integral over ds goes over the skier’s path. Since the skier is always moving
slowly, the normal force is approximately mg cos θ. (More generally, there would be another
contribution to provide the centripetal acceleration.) But then
Z Z Z
ffric ds = µmg cos θ ds = µmg dx = µmgs
which gives an answer of h = µs. (If the skier’s path turned around, then this would still
hold as long as s denotes the total horizontal distance traveled.)
[3] Problem 10 (MPPP 16). On a windless day, a cyclist going “flat out” can ride uphill at a speed
of v1 = 12 km/h and downhill at v2 = 36 km/h on the same inclined road. We wish to find the
cyclist’s top speed on a flat road if their maximal effort is independent of the speed at which the
bike is traveling. Note that in this regime, the air drag force is quadratic in the speed.
(a) Solve the problem assuming that “maximal effort” refers to the force exerted on the pedals
by the rider, and that the rider never changes gears.
(b) Solve the problem assuming that “maximal effort” refers to the rider’s power.
Solution. (a) Let F0 be the force due to gravity along the hill and let kv 2 be the drag force. If
the rider exerts force F 0 on the pedals, then the wheels exert a force F on the ground, but
the ratio F/F 0 is constant if there are no gear switches. Then
F − F0 = kv12 , F + F0 = kv22 , F = kv32
where v3 is the answer. Combining these equations gives
r
v12 + v22
v3 = = 27 km/h.
2
11
(b) In this case, the equations are a bit nastier,
P/v1 − F0 = kv12 , P/v2 + F0 = kv22 , P/v3 = kv32 .
Some tedious but straightforward algebra gives

s
2 2
3 v1 v2 (v1 + v2 )
v3 = = 23.5 km/h.
v1 + v2
[2] Problem 12. Alice steps on the gas pedal on her car. Bob, who is standing on the sidewalk, sees
Alice’s car accelerate from rest to 10 mph. Charlie, who is passing by in another car, sees Alice’s car
accelerate from 10 mph to 20 mph. Hence Charlie sees the kinetic energy of Alice’s car increase by
three times as much. How is this compatible with energy conservation, given that the same amount
of gas was burned in both frames?
Solution. The difference in energy comes from the change in kinetic energy of the Earth. In Bob’s
frame, the final kinetic energy of the Earth is p2 /2M where p is the total frictional impulse, and this
is negligible since p is moderately sized, while the Earth’s mass M is huge. Another way of saying
this is that the final kinetic energy of the car is p2 /2mcar , which is much larger since mcar M .
On the other hand, in Charlie’s frame, the Earth has some initial momentum P . The change in
kinetic energy of the Earth is
(P + p)2 − P 2 Pp p2
∆K = = + .
2M M 2M
The last term is again negligible, but now we have a term that is linear in p, which isn’t negligible.
Let v0 = 10 mph. We have P/M = v0 and p = mcar v0 , so
∆K = mcar v02
which is precisely the extra energy that ends up in the car. The lesson of this problem is that when
you go into a different reference frame, kinetic energies and even changes in kinetic energy can differ
dramatically. While you can get the right answer either way, it’s generally least confusing to work
in the rest frame of the largest object in the problem.
When there are multiple large objects, you can get interesting effects. For example, naively a
gravitational slingshot can’t work, because the gravitational force is conservative. And indeed, a
rocket doing a gravitational slingshot off of Jupiter gets no additional energy, in Jupiter’s frame.
However, for rockets that far out, the most important object is the Sun, since it determines,
e.g. whether the rocket can escape the solar system. To answer that kind of question we should
work in the Sun’s frame, and in this frame the rocket does get more energy, as it harvests it from
Jupiter’s large kinetic energy. You’ll investigate this in more detail in M6.
[3] Problem 13 (KK 4.8). A block of mass M is attached to a spring of spring constant k. It is pulled
a distance L from its equilibrium position and released from rest. The block has a small coefficient
of friction µ with the ground. Find the number of cycles the mass oscillates before coming to rest.
12
Solution. First let’s present a short solution that only works for small µ. Let A be the amplitude,
so the energy is E = 12 kA2 . Hence in one cycle, the change in energy is related to the change in
amplitude by
dE = kA dA
where we can use infinitesimals for one cycle since the friction is assumed small. But the energy
loss is also 4µmgA, so plugging this in gives
4µmg
dA = − .
k
The oscillation ends when the amplitude drops to zero, so
kL
N= .
4µmg
We expect this result to be trustworthy whenever N is large, i.e. when the fractional amplitude
change during a cycle is small.
We will now show that, in fact, this result is correct even when N isn’t large. Notice that
during the left-moving part of a cycle, the friction provides a constant force of µmg to the right.
Therefore, just like how gravity shifts the equilibrium position of a vertical spring, the friction
shifts the equilibrium position to the right by µmg/k. The left-moving motion is a perfect sinusoid
centered at this position. Similarly, the right-moving part of the oscillation is a perfect sinusoid,
but instead centered at −µmg/k. The net effect of one cycle is thus to decrease the amplitude by
exactly 4µmg/k, giving the result.
[3] Problem 14 (Morin 5.4). A massless string of length 2` connects two hockey pucks that lie on
frictionless ice. A constant horizontal force F is applied to the midpoint of the string, perpendicular
to it. The pucks eventually collide and stick together. How much kinetic energy is lost in the
collision?
Solution. Suppose the bend in the rope is θ, where originally θ = 0. We see that the tension T
satisfies 2T sin θ = F , by balancing forces at the midpoint. Thus, the y-component of the force on
the top mass is T cos θ, so the total work done by tension in the y direction is
Z π/2 Z π/2
W =− 2(T cos θ) d(` cos θ) = `F cos θ dθ = F `.
0 0
This determines the vertical kinetic energy, mvy2 /2, of each puck. When the pucks collide, all of
this energy is lost, giving the answer F `.
There’s also a slick alternate solution using a noninertial reference frame. Now, in general work
depends on the reference frame, as we just saw in problem 12, since displacement does, so we always
need to be careful calculating energies in other frames. However, the amount of dissipated energy
determines how much the pucks warm up, which is independent of frame! Therefore, we are free to
use any frame we want.
In particular, consider the frame with acceleration F/2m along the force. In this frame, there
is a fictitious force −F/2 on each puck. The net force on the system is zero, so the pucks move
directly towards each other. When the pucks collide, the point of application of the force F has
traveled a distance `, doing work F `. Since the pucks are stationary after collision, all this energy
is dissipated, giving the answer F ` again.
13
Idea 4
If a problem can be solved using either momentum conservation or energy conservation alone,
it usually means one of the two isn’t actually conserved. In particular, many processes
are inherently inelastic and inevitably dissipate energy. For more about inherently inelastic
processes, see section 5.8 of Morin.
[2] Problem 15 (KK 4.20). Sand falls slowly at a constant rate dm/dt onto a horizontal belt driven
at constant speed v.
(a) Find the power P needed to drive the belt.
(b) Show that the rate of increase of the kinetic energy of the sand is only P/2.
(c) We can explain this discrepancy exactly. Argue that in the reference frame of the belt, the
rate of heat dissipation is P/2. Since temperature is the same in all frames, the rate of heat
dissipation is P/2 in the original frame as well, accounting for the missing energy.
Solution. (a) We have P = F v = (dp/dt) v = (v (dm/dt))v = v 2 dm/dt.
(b) It’s 12 (dm/dt)v 2 = P/2.
(c) In the belt’s frame, the sand comes in with a speed of v, and friction slows it down to zero
speed. Hence the sand loses all its kinetic energy to heat, at a rate 12 (dm/dt)v 2 = P/2.
Example 5: PPP 108
A fire hose of mass M and length L is coiled into a roll of radius R. The hose is sent rolling
√
along level ground, with its center of mass given initial speed v0 gR. The free end of
the hose is held fixed.
The hose unrolls and becomes straight. How long does this process take to complete?
Solution
First, we need to find what is conserved. The horizontal momentum is not conserved,
because there is an external horizontal force needed to keep the end of the hose in place.
On the other hand, the energy is conserved, even though this process looks inelastic. The
hose “sticks” to the floor as it unrolls, but this process dissipates no energy because the cir-
cular part of the hose rolls without slipping, so the bottom of this part always has zero velocity.
14
Once we figure out energy is conserved, the problem is straightforward. The assumption
√
v0 gR means we can neglect the change in gravitational potential energy as the hose
unrolls. After the hose travels a distance x,

1 1 2 1 1
1+ M v0 = 1+ mv 2
2 2 2 2
where the 1/2 terms are from rotational kinetic energy. Since m(x) = M (1 − x/L), we have
v0
v(x) = p
1 − x/L
which gives a total time

Z L Z 1√
dx L 2L
T = = 1 − u du = .
0 v(x) v0 0 3v0
Evidently, the hose accelerates as it unrolls.
[4] Problem 16. Consider the following related problems; in all parts, neglect friction.
(a) A uniform rope of length ` lies stretched out flat on a table, with a tiny portion `0 ` hanging
through a hole. The rope is released from rest, and all points on the rope begin to move with
the same speed. Since this motion is smooth, energy is conserved. Find the speed of the rope
when the end goes through the hole.
(b) ? For practice, repeat part (a) by solving for x(t) explicitly. (Hint: this is best done using the
generalized coordinate techniques of M4.)
(c) Now suppose a flexible uniform chain of length ` is placed loosely coiled close to the hole.
Again, a tiny portion `0 ` hangs through the hole, and the chain is released from rest. In
this case, the unraveling of the chain is an inherently inelastic process, because each link of
the chain sits still until it is suddenly jerked into motion. Find the speed of the chain when
the last link goes through the hole. (Hint: you should get a nonlinear differential equation,
which can be solved by guessing x(t) = Atn .)
Solution. (a) We use energy conservation. The height of the center of mass falls by `/2, so
√
`M g/2 = M v 2 /2, which gives the answer of v = `g.
(b) For convenience, we use the idea of “generalized coordinates”, which will be covered in more
detail in M4. The point is that a direct application of Newton’s second law would be very
tough, because we’d have to solve for the tension everywhere in the rope. But we can instead
treat the rope as a single object by parametrizing its motion in terms of “how far it’s gone
through the hole”. The net force “putting the rope into the hole” is just gravity acting on
the hanging part of the rope, xM g. Thus,
xM g = M a
which implies
g
ẍ = x.
`
15
Now, this is a linear differential equation which can be solved √ with the techniques of M1.
± g/` t
Guessing exponentials gives growing and decaying solutions e . It’s most convenient to
rewrite these in terms of hyperbolic trig functions,
p p
x = A cosh( g/` t) + B sinh( g/` t).
p
The initial conditions then give us x(t) = `0 cosh( g/`t), so in the limit of small `0 , the final
time obeys
`0 √
` ≈ e g/` tf .
2
The velocity is
`0 g √g/`tf
r r r
g p g p
v(tf ) = `0 sinh( g/`tf ) ≈ e ≈ ` = `g
` 2 ` `
as found by energy conservation.
(c) In this case energy conservation doesn’t work, so we need to use momentum/force ideas.
Unlike part (a), it’s best to use Newton’s second law directly, by considering the vertical
momentum of the vertical part of the chain. We didn’t do this in part (a) because we would
have to know the tension at the hole, since this provides an external vertical force, but here
it’s easy because the chain links on the table are slack, so the tension is zero. On the other
hand, the generalized coordinate approach of part (a) wouldn’t work here because the chain is
not moving as one smooth unit; there’s nasty discontinuous stuff going on at the hole, which
we don’t want to think about.
Now, let m be the time-dependent mass of the vertical part. The only external vertical force
is gravity, so applying Fy = dpy /dt gives
mg = mv̇ + ṁv = mv̇ + (m/x)v 2
which implies
ẍ = g − ẋ2 /x.
This is a nonlinear second-order differential equation. There’s no general way to solve such
equations, so we’ll resort to the hint. If we guess a pure power Atn , then all three terms are
the same power of t as long as n = 2. Plugging in x(t) = At2 gives the solution
1
x(t) = gt2
6
so there is a uniform acceleration of g/3. (The 1/6 is not an arbitrary constant, if you change
it you don’t get a solution to the differential equation at all! That’s because this equation
is nonlinear, so there’s no reason to expect that multiplying a solution by a constant gives
another solution.)
p
The amount of time it takes for last link to pass is t = 6`/g, so the speed there is
r
2`g
v = (g/3)t = .
3
This is smaller than the answer to part (a) because energy is not conserved.
16
[3] Problem 17 (PPP 95). A long slipway, inclined at an angle α to the horizontal, is fitted with
many identical rollers, consecutive ones being a distance d apart. The rollers have horizontal axles
and consist of rubber-covered solid steel cylinders each of mass m and radius r. A plank of mass
M , and length much greater than d, is released at the top of the slipway.
Find the terminal speed vmax of the plank. Ignore air drag and friction at the pivots of the rollers.
Solution. Let the terminal velocity be v, and consider the forces acting on the plank along the
plane. There is of course a constant gravitational force M g sin α. In addition, every time the plank
hits a roller, it experiences an impulse as it spin the roller up. The angular impulse on each roller
is equal to its angular momentum, so
Z
1
f (t)r dt = mr2 ω.
2
This implies the linear impulse on the plank has magnitude
Z
1 1
J = f (t) dt = mrω = mv.
2 2
This impulse must be equal to the total gravitational impulse along the plane between rollers,
1 d
mv = M g sin α
2 v
which gives the answer,
r
2M gd sin α
v= .
m
The subtle thing about this problem is that a similar argument based on energy conservation gives
the wrong answer. Equating the gravitational potential energy lost per roller to the rotational
kinetic energy given to each roller gives
1 2 1
Iω = mv 2 = M gd sin α
2 4
√
which gives an answer different by a factor of 2. The reason is that energy is also dissipated into
heat, as the plank and roller initially slip with respect to each other. By an argument extremely
similar to that of problem 15, but with angular variables instead of linear ones, you can show that
precisely half the gravitational potential energy goes into heat. Accounting for this gives exactly
the same answer as momentum conservation.
4 Elastic Collisions
17
Idea 5
Any temporary interaction between two objects that conserves energy and momentum is a
perfectly elastic collision.
Example 6
Two masses are constrained to a line. The mass m1 moves with velocity v1 , and the mass m2
moves with velocity v2 . The masses collide perfectly elastically. Find their speeds afterward.
Solution
The usual method is to directly invoke conservation of energy and momentum, which leads
to a quadratic equation. A slicker method is to work in the center of mass frame instead.
(This is useful for collision problems in general, and it’ll become even more useful for the
relativistic collisions covered in R2.)
The center of mass of the system has speed

m 1 v1 + m 2 v2
vCM = .
m1 + m2
Moreover, by momentum conservation, the center of mass never accelerates. Now we boost
into the frame moving with the center of mass. Since the total momentum is by definition
zero in the center of mass frame, the momenta of the particles cancel out. The only way for
this to remain true after the collision is if we multiply their velocities by the same number.
Energy is only conserved if this number is ±1, with the latter representing no collision at all.
Therefore, during an elastic collision, the velocities in the center of mass frame simply reverse.
The initial velocities in that frame are
v1,CM = v1 − vCM , v2,CM = v2 − vCM .
The final velocities in that frame are
0 0
v1,CM = −v1 + vCM , v2,CM = −v2 + vCM .
Finally, going back to the original frame gives the final velocities
v10 = −v1 + 2vCM , v20 = −v2 + 2vCM .
There are many special cases we can check. For example, if m1 = m2 , then the two masses
simply swap their velocities, as if they just passed through each other. As another check,
consider the case where the second mass is initially at rest, v2 = 0. Then
m1 − m2 2m1
v10 = v1 , v20 = v1 .
m1 + m2 m1 + m2
When m1 = m2 , the first mass gives all its velocity to the second. When m2 is large, the first
mass just rebounds off with velocity −v1 . When m1 is large, the first mass keeps on going
and the second mass picks up velocity 2v1 . Finally, when m1 = m2 /3, then the final speeds
are v10 = −v1 /2 and v20 = v1 /2, a nice result which is worth committing to memory.
18
Idea 6
For problems involving many collisions, a nice way to keep track of everything is to make a
2D plot of x(t) for all the masses.
[2] Problem 18 (Morin 5.23). A tennis ball with mass m2 sits on top of a basketball with a mass
m1 m2 . The bottom of the basketball is a height h above the ground. When the balls are
dropped, how high does the tennis ball bounce?
√
Solution. Let v = 2gh. Once the basketball hits the ground, it has velocity v upwards, whereas
the tennis ball has velocity v downwards. Going into CM frame, which coincides with basketball
frame since m1 m2 , we see that the new velocity of the tennis ball is 2v up, so the velocity in
the ground frame is 3v up, so the tennis ball bounces to 9h since its energy got multiplied by 9.
[3] Problem 19 (PPP 46). A Newton’s cradle consists of three suspended steel balls of masses m1 ,
m2 , and m3 arranged in that order with their centers in a horizontal line. The ball of mass m1 is
drawn aside in their common plane until its center has been raised by h and is then released. If
all collisions are elastic, how much m2 be chosen so that the ball of mass m3 rises to the greatest
possible height, and what is this height? (Neglect all but the first two collisions.)
√
Solution. The ball of mass m1 has speed v = 2gh once it hits the ball of mass m2 . By applying
the result of example 6 twice, the speed of mass m3 after the first two collisions is
2m1 2m2
v0 = v
m1 + m2 m2 + m3
which means the final height is
2
0 4m1 m2
h = h.
(m1 + m2 )(m2 + m3 )
By some basic calculus, this is maximized when

√
m2 = m1 m3
in which case !4
2
h0 = p h.
1 + m3 /m1
For a wide range of m3 /m1 , this is pretty close to perfect efficiency. (Transferring 100% of the
energy would yield h0 = (m1 /m3 )h.)
[3] Problem 20. Here’s a variety problem involving some “clean” mathematical results. All three
parts can be solved without lengthy calculation.
(a) Consider n identical balls confined to a line. Assuming all collisions are perfectly elastic, what
is the maximum number of collisions that could happen? Assume no triple collisions happen.
(b) A billiard ball hits an identical billiard ball initially at rest in a perfectly elastic collision.
Show that the balls exit at a right angle to each other.
19
(c) A mass M collides elastically with a stationary mass m. If M > m, show that the maximum
possible angle of deflection of M is sin−1 (m/M ).
Solution. (a) Note when two identical balls collide, they simply swap velocities. Therefore, we
can imagine the balls as passing through one another, and we want the maximum number of
times two balls pass through each other. All n(n − 1)/2 pairs can pass through each other,
as long as the ith ball on the right has the ith smallest rightward velocity.
(b) Let the initial velocity be v, and the final velocities be v1 and v2 . By momentum conservation,
v1 + v2 = v. By energy conservation, v12 + v22 = v 2 . By the law of cosines, this is only possible
if v1 and v2 are orthogonal.
(c) Let v be the initial speed. Recall that in the center of mass frame, the mass M can only
change the direction, but not the magnitude, of its velocity. In this frame, the speed of the
mass M is v 0 = mv/(m + M ). Thus, the possible final velocities lie on a circle of radius v 0 .
Now let’s transform back to the lab frame. In this frame, it’s still true that the possible final
velocities lie on a circle of radius v 0 , but the center of this circle is now at the center of mass
velocity vc = M v/(m + M ).
From the sketch above, the final velocity that maximizes the angle θ of deflection yields
v0 m
sin θ = =
vc M
which is the desired result.
[3] Problem 21 (PPP 72). Beads of equal mass m are strung at equal distances d along a long,
horizontal, infinite wire. The beads are initially at rest but can move without friction. The first
bead is continuously accelerated towards the right by a constant force F . After some time, a “shock
wave” of moving beads will propagate towards the right.
(a) Find the speed of the shock wave, assuming all collisions are completely inelastic.
(b) Do the same, assuming all collisions are completely elastic. What is the average speed of the
accelerated bead in this case?
Solution. (a) In the steady state, a large clump of particles will be moving towards the right. If
the steady state speed is v0 , then collisions occur at time intervals d/v0 , so the momentum of
the blob must grow at rate (mv0 )(v0 /d) as new beads join it. This must be equal to F , and
solving gives r
Fd
v0 = .
m
20
(b) By basic kinematics, the speed of the first, accelerated bead the moment before it hits the
next bead is r
2F d
v1 = .
m
At the moment of collision, the first bead loses all its velocity to the second. The second
bead moves towards the third with velocity v1 and gives its velocity to the third, and so on,
creating a shock front with velocity v1 .
In the meantime, the first bead is still accelerating. After another time interval, it hits the
second bead, which is now where the third bead originally was, and the same phenomenon
happens again, creating another bead with velocity v1 just behind the leading one. So after a
long time, we build up a shock front of beads traveling with speed v1 .
On the other hand, the first bead keeps uniformly accelerating between zero speed and v1 , so
its average speed is just v1 /2. Evidently, the shock wave separates from the first bead after
01^
time.
Example 7: MPPP 42
There are N identical tiny discs lying on a table, equally spaced along a semicircle, with total
mass M . Another disc D of mass m is very precisely aimed to bounce off all of the discs in
turn, then exit opposite the direction it came.
In the limit N → ∞, what is the minimal value of M/m for this to be possible? Given this
value, what is the ratio of the final and initial speeds of the disc?
Solution
The reason that there is a lower bound on M is that, by the result of part (c) of problem 20,
there is a maximal angle that each tiny disc can deflect the disc D. For large N , the deflection
is π/N for each disc, so
π M/N M
= sin−1 ≈
N m Nm
which implies that M/m ≥ π.
21
To see how much energy is lost in each collision, work in the center of mass frame and consider
the first collision. In this frame, the disc D is initially approximately still, and the tiny disc
comes in horizontally with speed v. To maximize the deflection angle in the table’s frame, the
tiny disc should rebound vertically, as this provides the maximal vertical impulse to the disc D.
Thus, going√back to the√table’s frame, where the disc D has speed v, the tiny disc scatters
with speed v 2 + v 2 = 2v. By conservation of energy,
1M √ 2

1
∆ mv 2 = − ( 2v) .
2 2N
This simplifies to
∆v π
=−
v N
which means that after N collisions, we have
vf π N
= 1− ≈ e−π
vi N
where in the last step we used a result from P1.
In this question, we consider a simplified model of how an elastic collision actually happens.
Consider a spherical volleyball inflated with excess pressure ∆P , radius r, and mass m. If it
hits the ground with a large, but not huge, speed v, estimate how long the subsequent elastic
collision takes.
Solution
When the volleyball hits the ground, it will keep going, deforming the part that touches the
ground into a flat circular face. Specifically, when the ball has moved a distance y into the
ground, the flat face has area
p 2
A=π r2 − (r − y)2 = πy(2r − y) ≈ 2πry
where we assumed that y r at all times, which is reasonable as long as the initial speed is
not huge. As a result, the pressure of the volleyball exerts a force
F = 2πr ∆P y
on the ground. This assumes the pressure inside the volleyball remains uniform, and that
the rest of the volleyball stays approximately spherical, which is again reasonable as long as
the initial speed is not huge.
Assuming the initial velocity is not too small, gravity is negligible during the collision, so
during the collision the force on the volleyball is effectively that of an ideal spring. The
22
collision lasts for half a period, giving

r r
m πm
τ =π = .
keff 2r ∆P
If we plug in realistic numbers, the result is of order 10 ms, which is plausible.
Example 9
As shown in M2, a hanging chain takes the form of a catenary. Suppose you pull the chain
down in the middle. How does the center of mass of the chain move? Does the answer depend
on how hard you pull?
Solution
No matter how hard you pull, or in what direction, the height of the center of mass always
goes up! This is because this quantity measures the total gravitational potential energy of
the chain. If you pull a chain in equilibrium, in any direction whatsoever, you will do work
on it. So this raises its potential energy, and hence the center of mass.
Another way of saying this is that the equilibrium position, without the extra pull you supply,
is already in the lowest energy state, and hence already has the lowest possible center of mass.
Changing this shape in any way raises the center of mass.
[2] Problem 24. A uniform half-disc of radius R is nailed to a wall at the center of its circle and
allowed to come to equilibrium. The half-disc is then rotated by an angle dθ. By calculating the
energy needed to do this in two different ways, find the distance from the pivot point to the center
of mass.
Solution.
Suppose the CM is at radius r. The energy required to turn the disc by dθ is

mgx 2
(1 − cos(dθ)) mgx = dθ .
2
23
However, when rotated, all that has changed is that there is a new sector of angle dθ above (sector
A), and one sector is now missing (sector B). The CM of a thin sector is at radius 2R/3 (it’s just
an isosceles triangle), so the total extra energy of A is (dm)g(2R/3)dθ/2, where dm/m = dθ/π, so
it’s mgR(1/3π) dθ2 , and the extra energy due to the absence of B is the same, so we have
2 mgx 2
mgR dθ2 = dθ
3π 2
which implies
4
x= R.
3π
[4] Problem 25 (Morin 5.31). Assume that a cloud consists of tiny water droplets suspended (uni-
formly distributed, and at rest) in air, and consider a raindrop falling through them. Assume the
raindrop is initially of negligible size, remains spherical at all times, and collides perfectly inelasti-
cally with the droplets. It turns out that the raindrop accelerates uniformly; assuming this, find
the acceleration.
Solution. Suppose the mass density in the cloud is λ and the mass density of the raindrop is ρ
(note ρ > λ), and suppose r is the radius of the drop, M the mass, and v the velocity. We see that
ṙ
Ṁ = 4πr2 ṙρ = 3M
r
and
Ṁ = πr2 vλ,
which combine to give
4ρ
v= ṙ.
λ
We see that M g = Ṁ v + M v̇, so Newton’s second law is

3ṙ 4ρ 4ρ
Mg = M ṙ + r̈
r λ λ
and writing everything in terms of r gives
rgλ/ρ = 12ṙ2 + 4rr̈.
This is a nonlinear second-order differential equation; there is no general method to solve these
equations. Certainly an exponential won’t work, because you won’t get the same exponential on
the left and right-hand sides. However, we can use the hint, which indicates that v is linear in time.
This implies that r is a quadratic, so guessing r = At2 gives
At2 gλ/ρ = 4A2 (12t2 + 2t2 ).
This implies that we indeed have a solution, as long as
A = (gλ/ρ)/56.
Using our relation between v and ṙ, we finally have

4ρ g
v̇ = r̈ =
λ 7
24
which is the acceleration.

As you can see from this problem and an earlier one, nonlinear second-order differential equations
are actually quite common in physics. Trying a pure power Atn is a decent first guess, because
monomials remain monomials under both differentiation and multiplication; for the same reason,
an exponential AeBt can also work. However, in practice, the vast majority of such differential
equations don’t have analytic solutions at all, or only have solutions in terms of exotic special
functions. Problems for Olympiads and undergraduate textbooks are generally chosen precisely to
avoid these complications, since they draw attention away from the essential physics.
This raindrop problem is a classic, first invented (though with a slightly different mass accretion
rule) for the infamous Cambridge Smith’s Prize Examination in 1853. Many papers have been
written about it, such as this one by K. S. Krane, who is the K in HRK.
[3] Problem 26 (Kvant). Half of a flexible pearl necklace lies on a horizontal frictionless table, while
the other half hangs down vertically at the edge. If the necklace is released from rest, it will slide off
the table. At some point, the hanging part of the necklace will begin to whip back and forth. What
fraction of the necklace is on the table when this begins? (Hint: we are considering a pearl necklace
with no empty string between adjacent pearls; as a result, all the pearls accelerate smoothly. To
solve the problem, think about the vertical forces. There is an important related problem in M2.)
Solution. Physically, what happens is that a sizable force is needed to turn the pearls around when
they reach the corner, to go from moving horizontally to vertically. At a certain critical velocity,
tension will no longer be enough to do this, and the pearl necklace will jump off the corner. This
will lead to the hanging part whipping back and forth.
To see when this happens, consider the vertical forces on the pearl necklace. Suppose the necklace
has mass M and a fraction x is hanging off the table. As we saw with a similar but static problem
in M2, the normal force on the horizontal part of the table has to cancel the force of gravity acting
on the part of the necklace on the table. This is because otherwise a piece of the horizontal part of
the necklace would have an unbalanced vertical force, and would have to go into the table or jump
off it, neither of which make sense.
Thus, considering the vertical forces just gives
dpy
= xM g − Nc,y
dt
where we take the downward direction as positive for convenience, and Nc,y is the vertical part of
the normal force acting at the corner of the table. (Recall from M2 that this is significant even
when the corner is small.) Since the pearl necklace is flexible, the process is elastic, so energy is
conserved. Our strategy is to use energy conservation to find dpy /dt and use that to find Nc,y . The
necklace jumps off the corner when Nc,y becomes zero.
Taking the necklace to have length L, energy conservation gives
1 x M gL p p
M v 2 = M xgL − =⇒ v 2 = gL(x2 − 1/4) =⇒ ẋ = g/L x2 − 1/4.
2 2 8
The acceleration is
p 1 p x p p
ẍ = g/L p · 2xẋ = g/L p g/L x2 − 1/4 = (g/L)x
2 x2 − 1/4 x2 − 1/4
which we also found using the generalized coordinate approach explained in problem 16. Thus,
dpy d
= M L (xẋ) = M L(xẍ + ẋ2 ) = M L((g/L)x2 + (g/L)(x2 − 1/4)) = M g(2x2 − 1/4).
dt dt
25
Setting Nc,y to zero gives

2x2 − 1/4 = x
√
for which the relevant root is x = (1 + 3)/4. At this point, the fraction on the table is
√
3− 3
1−x= .
4
This is barely less than 1/2, so the necklace jumps almost immediately!
In general, ideal ropes like pearl necklaces, which aren’t inherently inelastic at all, do unintuitive
things like this. An even subtler, related phenomenon is the chain fountain, where a chain of beads
can seemingly pull itself out of a jar.
[4] Problem 27 (BAUPC 2002). A small ball is attached to a massless string of length L, the other
end of which is attached to a very thin pole. The ball is thrown so that it initially travels in a
horizontal circle, with the string making an angle θ0 with the vertical. As time goes on, the string
wraps itself around the pole. Assume that (1) the pole is thin enough so that the length of string in
the air decreases very slowly, and (2) the pole has enough friction so that the string does not slide
on the pole, once it touches it. Show that the ratio of the ball’s final speed (right before it hits the
pole) to initial speed is sin θ0 .
Solution. The official solution is here, but it’s a lot longer because it also solves for the evolution
of the height of the ball; in this shorter version of the problem, we only want the final speed, which
is a lot easier to find.
In this problem, energy is conserved because the string doesn’t slide, but angular momentum isn’t
conserved. Now note that if θ were constant, the ball would move at a constant height. Changes in θ
as the string winds up change the ball’s height, which then changes its speed by energy conservation.
By doing some geometry, you can show
mv dv = −mg` sin θ dθ.
The radial F = ma equation is

mv 2
mg tan θ = .
` sin θ
Substituting this into the equation above gives
mv 2
−mg dθ = mv dv
mg tan θ
which upon separation gives the simple result
Z π/2 Z vf
dv
− cot θ dθ = .
θ0 vi v
Therefore, we have
sin θ0 vf
log = log
sin π/2 vi
which gives the desired result, vf /vi = sin θ0 .
When dealing with an extended system whose parts all move in different ways, conservation of
energy is occasionally useless. However, the somewhat obscure idea of “center of mass energy” may
become useful instead. For more about this concept, see section 13.5 of Halliday and Resnick.
26
Idea 7: Center of Mass Energy
The work done on a part of a system is
dW = F dx
where F is the force on that specific part of the system, and dx is its displacement. Then
dW = dE where E is the total energy of the system.
Similarly, the “center of mass work” done on a system is
dWcm = F dxcm
where F is the total force on the system and dxcm is the displacement of the center of mass.
Then dWcm = dEcm where the “center of mass energy” is defined as Ecm = M vcm 2 /2.
It should be noted that, like regular energy and work, center of mass energy and work depend
on the reference frame you’re using.
Example 10
Consider a cyclist who pedals their bike to accelerate. The wheels roll without slipping on
the ground. The cyclist moves a distance d, with the bike experiencing a constant friction
force f from the ground. Analyze the situation using both energy and center of mass energy.
Solution
Since the wheels roll without slipping, their contact point with the ground is always zero,
so the friction force does exactly zero work. Thus the net energy of the cyclist/bike system
is conserved. The additional kinetic energy of the cyclist/bike comes from the chemical
energy of the cyclist, which ultimately came from what they ate. So conservation of energy
is correct, but it doesn’t tell us anything useful at all.
Now consider center of mass energy. Considering the cyclist/bike system, the center of mass
2 /2. This allows us to compute the change in velocity
work is f d, which is the change in M vcm
of the cyclist/bike.
Example 11
Consider the same setup as in the previous example, but now the cyclist brakes hard. The
wheels slip on the ground, and experience a friction force −f while the cyclist moves a
distance d. Analyze the situation using both energy and center of mass energy.
27
Solution
The center of mass work equation tells us about the overall deceleration of the cyclist/bike,
just as in the previous example.
On the other hand, the work done by the friction force is indeterminate! It can be any
quantity between zero and −f d. When it is 0, the total energy of the cyclist/bike system
is again conserved, which means all the kinetic energy lost is dissipated as heat inside the
bike itself. When it is −f d, all the kinetic energy lost is dissipated as heat in the ground,
and hence energy is removed from the cyclist/bike system. In general, the work will be
an intermediate value, meaning that both the ground and the bike heat up, but we can’t
calculate what it is without a microscopic model of how the friction works. It depends on,
e.g. how easily the ground and bike tire surface deform.
[1] Problem 28. Alice and Bob stand facing each other with their arms bent and hands touching on
an ice skating rink. Bob has his back against a wall.
(a) Suppose Bob extends his arms, pushing Alice through a distance d with a force F . Analyze
what happens to Alice in terms of both work and center of mass work.
(b) Suppose Alice extends her arms, pushing herself through a distance d with a force F . Repeat
the analysis; what is different and what is the same?
(c) Suppose a spherical balloon is compressed uniformly from all sides. Is there work done on the
balloon? How about center of mass work?
Solution. (a) p The center of mass work and work done on Alice are both F d, so she moves with
speed v = 2F d/m. In this situation Alice effectively behaves like a particle, so the two
notions are the same.
(b) The center of mass work on Alice is again F d, so her final speed is the same. But the work
done on her is 0 since the contact point did not move. Accordingly, Alice’s total energy did
not change; she merely converted some of her internal energy to kinetic energy.
(c) There is no center of mass work on the balloon, but there is work done, at every point on the
01m
balloon’s surface. This work is just the P dV work in thermodynamics.
[4] Problem 29. USAPhO 2013, problem B1. This problem is quite tricky! Once you’re done,
carefully read the official solution, which describes how center of mass work is applied.
Solution. As usual, see the official solutions. Recently, this Veritasium video reignited the debate
over the Blackbird, leading to this followup video, which cites the USAPhO solution. If you’re still
confused about how the Blackbird works, I recommend watching the second video, which shows
various arguments and a mechanical model.
28
Mechanics IV: Oscillations

Chapter 4 of Morin covers oscillations, as does chapter 10 of Kleppner and Kolenkow, and chapter
10 of Wang and Ricardo, volume 1. For a deeper treatment that covers normal modes in more detail,
see chapters 1 through 6 of French. Jaan Kalda also has short articles on using Lagrangian-like
techniques and the adiabatic theorem. For some fun discussion, see chapters I-21 through I-25,
II-19, and II-38 of the Feynman lectures. There is a total of 85 points.
1 Small Oscillations
Idea 1
If an object obeys a linear force law, then its motion is simple harmonic. To compute the
frequency, one must the restoring force per unit displacement. More generally, if the force an
object experiences can be expanded in a Taylor series with a nonzero linear restoring term,
the motion is approximately simple harmonic for small displacements. (However, don’t forget
that there are also situations where oscillations are not even approximately simple harmonic,
no matter how small the displacements are.)
Example 1: KK 4.13
The Lennard–Jones potential

r0 12 r 6
0
U (r) = −2
r r
is commonly used to describe the interaction between two atoms. Find the equilibrium radius
and the frequency of small oscillations about this point for two identical atoms of mass m
bound to each other by the Lennard–Jones interaction.
Solution
To keep the notation simple, we’ll set = r0 = 1 and restore them later. The equilibrium
radius is the radius where the derivative of the potential vanishes, and
U 0 (r) = −12r−13 + 12r−7 = 0
implies that that the equilibrium radius is r = r0 . Because the force accelerates both of the
atoms, the angular frequency is s
U 00 (r)
ω=
m/2
where m/2 is the so-called reduced mass. At the equilibrium point, we have
U 00 (r0 ) = (12)(13)r0−14 − (12)(7)r0−8 = 72.
Restoring the dimensionful factors, we have U 00 (r0 ) = 72/r02 , so
r
12
ω= .
r0 m
1
[3] Problem 1 (Morin 5.13). A hole of radius R is cut out from an infinite flat sheet with mass per
unit area σ. Let L be the line that is perpendicular to the sheet and that passes through the center
of the hole.
(a) What is the force on a mass m that is located on L, a distance x from the center of the hole?
(Hint: consider the plane to consist of many concentric rings.)
(b) Now suppose the particle is released from rest at this position. If x R, find the approximate
angular frequency of the subsequent oscillations.
(c) Now suppose that x R instead. Find the period of the resulting oscillations.
(d) Now suppose the mass begins at rest on the plane, but slightly displaced from the center. Do
oscillations occur? If so, what is the approximate frequency?
Solution. (a) Consider a ring with radius r and thickness dr. It has mass dM = 2πrσ dr. By
symmetry, the net force is towards the sheet so we only take that√component. Thus we
multiply the total force from the ring dF = Gm dM/(x2 + r2 ) by x/ x2 + r2 . We integrate
from R to infinity due to the hole, giving
Z ∞ Z ∞
2πrxσ dr du x
F =− Gm 2 2 3/2
= −πGmxσ 3/2
= −2πGmσ √
R (x + r ) x2 +R2 u x + R2
2
(b) Approximating the force at small x gives

x
F = −2πGmσ p ≈ −2πGmσx/R.
R 1 + x2 /R2
This implies simple harmonic oscillations,

r
2πGσ 2πGσ
ẍ = − x = ω 2 x, ω= .
R R
(c) In this case, approximating the force at large x gives

1
F = −2πGmσ p ≈ −2πGmσ
1 + R2 /x2
with the force always directed towards

p the plane. This corresponds to a uniform acceleration
of g = 2πGσ. A ball takes time t = 2x/g to fall from a height x. The period is thus
r
x
T =4 .
πGσ
(d) This situation is unstable; the mass will just accelerate further away from the center, so no
oscillations occur. This is related to Earnshaw’s theorem, which we cover in E1, which tells
us that no gravitational field (or electrostatic field) in vacuum can have a point that is stable
in all directions.
[2] Problem 2. Some small oscillations questions about the buoyant force.
2
(a) A cubical glacier of side length L has density ρi and floats in water with density ρw . Find
the frequency of small oscillations, assuming that a face of the glacier always remains parallel
to the water surface, and that the force of the water on the glacier is always given by the
hydrostatic buoyant force.
(b) A ball of radius R floats in water with half its volume submerged. Find the frequency of small
oscillations, making the same assumption.
(c) There are important effects that both of the previous parts neglect. What are some of them?
Is the true oscillation frequency higher or lower than the one found here?
Solution. (a) Let V = xL2 be the submerged volume, and let V0 = L3 . We then have that
r
2
p
2 3
ρw g
F = −ρw V g + ρi V0 g = −ρw L gx + const. Thus, ω = ρw L g/ρi L = .
ρi L
(b) The density of the ball is half that of water, so its mass is (2π/3)ρw R3 . The “spring constant”
is πR2 ρw g. Hence s r
πR2 ρw g 3g
ω= 3
= .
(2π/3)ρw R 2R
(c) The most egregious omission is that we have completely neglected the motion of the water.
This clearly should add extra inertia, because the water has to move around to accommodate
the moving glacier or ball, and it should be a significant change since the water is more dense
than these objects. This “virtual mass” effect leads to a decrease in the oscillation frequency.
However, treating the motion of the water is quite mathematically involved; we’ll see some
examples in M7 where we partially account for the water’s motion, and even then it’s quite
tricky.
In fact, the situation is even worse. As we’ll see in M7, viscosity between the object and water
leads to a boundary layer of water carried along with the object. But since this boundary
layer builds up over time, its thickness depends on the entire history of the object’s motion!
This is called the Basset force, and it turns Newton’s second law into an “integro-differential
equation”, one where the second derivative of the position depends on an integral over all
the past positions. It has the effect of damping the oscillations. In general, nothing in fluid
01W
dynamics is easy.
[3] Problem 3. USAPhO 1998, problem A2. To avoid some confusion, skip part (a), since there
01W
actually isn’t a nice closed-form expression for it.
Idea 2
A useful generalization of Newton’s second law is given by generalized coordinates. Let q be
any number that describes the state of the system, not necessarily a Cartesian coordinate.
Suppose the energy of a system can be decomposed into two parts, a potential energy that
3
depends only on q and a kinetic energy that depends only on q̇,
K = K(q̇), V = V (q).
Then since energy is conserved, d(K + V )/dt = 0, the chain rule gives
d ∂K ∂V
=− .
dt ∂ q̇ ∂q
We call the left-hand side the rate of change of a “generalized momentum”, and the right-hand
side a “generalized force”. When q is a Cartesian coordinate, this recovers the usual F = ma.
Remark
The result above is a special case of the Euler–Lagrange equation in Lagrangian mechanics,
which states that if a system is described by a Lagrangian L, then
d ∂L ∂L
= .
dt ∂ q̇ ∂q
When L = K(q̇) − V (q), we recover the previous result. But more generally, it might not be
possible to meaningfully decompose L into a “kinetic” and “potential” piece at all! We won’t
use this more general form below. While it is more powerful, it is also more complicated, and
if you find yourself using it for an Olympiad problem, there’s probably an easier way.
Example 2
Find the acceleration of an Atwood’s machine with masses m and M and a massless pulley
and string.
Solution
The standard way to do this is to let a1 and a2 be the accelerations of the masses, let T be
the unknown tension in the string, solve for T by setting a1 and a2 to have equal magnitudes,
then plug T back in to find the common acceleration. The reason this procedure is so
complicated is that we are using two coordinates when the string really ensures the system
has only a single degree of freedom.
Instead, let q be a generalized coordinate that describes “how much the string has moved”.
In other words, q = 0 initially, and for some q > 0, the mass M has moved down by q and
the mass m has moved up by q. Then
1
K = (m + M )q̇ 2 , V = qg(m − M )
2
and applying the idea above gives
M −m
q̈ = g.
M +m
Another way of saying this is that, from the standpoint of this generalized coordinate, the
4
“total force” is (M − m)g, and the “total inertia” is M + m.
[1] Problem 6. A rope is nestled inside a curved frictionless tube. The rope has a total length ` and
uniform mass per length λ. The shape of the tube can be arbitrarily complicated, but the left end
of the rope is higher than the right end by a height h. If the rope is released from rest, find its
acceleration. (For a related question, see F = ma 2019 B24.)
Solution. Let q be the amount the rope has moved. The kinetic energy is λ`q̇ 2 /2. The “generalized
force” is the change in gravitational potential energy per change in q, so ∂V /∂q = λgh. Therefore
the acceleration is gh/`.
Idea 3
Generalized coordinates are really useful for problems that involve complicated objects but
only have one relevant degree of freedom, which is especially true for oscillations problems.
For instance, if the kinetic and potential energy have the form
1 1
K = meff q̇ 2 , V = keff q 2
2 2
then the oscillation frequency is always
p
ω= keff /meff .
Note that q need not have units of position, meff need not have units of mass, and so
on. When V (q) is a more general function, we can expand it about a minimum qmin , so
that keff = V 00 (qmin ). This technique allows us to avoid dealing with possibly complicated
constraint forces.
[3] Problem 7. Suppose a particle is constrained to move on a curve y(x) with a minimum at x = 0.
We know that if y(x) is a circular arc, then the motion is not exactly simple harmonic, for the same
reason that pendulum motion is not. Find a differential equation relating y 0 and y, so that the
motion is exactly simple harmonic for arbitrary amplitudes; you don’t have to solve it. (Hint: work
in terms of the coordinate s, the arc length along the curve.)
Solution. The reason s is useful is because

1
K = mṡ2
2
exactly. Thus, the condition for simple harmonic motion is V ∝ s2 , which means y = ks2 for some
constant k. Taking the derivative with respect to x of both sides, we have
ds
q p q
y 0 = 2ks = 2ks 1 + y 0 2 = 2 ky 1 + y 0 2 .
dx
Solving this for y 0 gives s
4ky
y0 =
1 − 4ky
5
which is the desired differential equation. Solving this is a bit nasty, but it turns out to be a
cycloid. This fact was first discovered by Huygens, who invented the cycloidal pendulum for
accurate timekeeping.
[3] Problem 8 (Grad). A particle of mass M is constrained to move on a horizontal plane. A second
particle of mass m is constrained to a vertical line. The two particles are connected by a massless
string which passes through a hole in the plane.
The motion is frictionless. Show that the motion is stable with respect to small changes in the
height of m, and find the frequency of small oscillations.
Solution. We’ll use the energy method for this one. In equilibrium we have
M v2 L2
mg = =
r M r3
where r is the radius of the circle, and L is the conserved angular momentum. If the hanging mass
goes downward, then r decreases, so the tension in the string increases providing a restoring force;
hence the orbit is stable.
To find the frequency of small oscillations, we’ll use the energy method, with generalized coordi-
nate r. The “kinetic energy”, which is the part of the energy dependent on ṙ, is
1 1
T = M ṙ2 + mṙ2 .
2 2
The “potential energy”, which is the part of the energy dependent on r, is
L2
V = mgr + .
2M r2
Note that L2 /2M r2 is treated as potential energy here even though it is associated with the motion
of the large mass. From the kinetic energy, we see the “effective mass” is meff = M + m, as one
might expect. From the potential energy, we see the “effective spring constant” is
3L2 3mg
keff = V 00 = 4
= .
Mr R
Hence the oscillation frequency is
s r r
keff 3g m
ω= = .
meff R m+M
01@
We’ll see much more of this technique in M7.
[4] Problem 9. IPhO 1984, problem 2. If you use the energy methods above, you won’t actually
need to know anything about fluid mechanics to do this nice, short problem!
6
Solution. We will find the period of oscillation, we will find expressions for the kinetic energy and
potential energy associated with the seiching.
First, to find the potential energy increase when the water is displaced by ξ, note that a triangular
prism of water has effectively been moved upward, as shown above. The centers of masses of these
triangles are ξ/3 from their bases, so the center of mass of the triangle will move up a distance of
2ξ/3. Let the width of the container by w. Then the potential energy U will be 2mgξ/3, where the
mass of the triangular prism of water is m = 12 L2 ξwρ, so
1
U = ρLwgξ 2 .
6
To find a rough estimate of the kinetic energy, consider the movement of the center of mass alone;
this won’t get all of the kinetic energy, but it’ll get enough to get a reasonable answer. By thinking
of the contribution of moving the triangle mentioned above, we have
1
m(2L/3) Lξw(2L/3) 1 Lξ
∆xcm = = 4 = .
M Lwh 6 h
m(2ξ/3) ξ2
∆ycm = = .
M 6h
We see that ∆xcm dominates since ξ is small, so we focus on it. The total mass of the water is
M = ρLwh
and our approximation for the kinetic energy is
1 1 L2 ξ˙2
K ≈ M ẋ2cm = ρLwh .
2 2 36h2
Besides the overall side-to-side center of mass motion of the water, the water also has internal
motions that can’t be described just in terms of the center of mass moving. However, our result is
good enough for the purposes of this problem.
Putting this together yields
ρwL3 ˙2 1 1 1
E≈ ξ + ρLwgξ 2 = meff ξ˙2 + keff ξ 2
72h 6 2 2
and thus a period of s
L2
T ≈ 2π .
12gh
This is compatible with the data given in the problem statement, up to order-one factors. Your
answer may look different, since we’ve made a lot of approximations throughout the problem; as
long as it agrees dimensionally, with the prefactor within an order of magnitude, you can regard it
as correct.
7
2 Springs and Pendulums

Now we’ll consider more general problems involving springs and pendulums, two very common
components in mechanics questions. As a first example, we’ll use the fictitious forces met in M2.
Example 3: PPP 79
A pendulum of length ` and mass m initially hangs straight downward in a train. The
train begins to move with uniform acceleration a. If a is small, what is the period of small
oscillations? If a can be large, is it possible for the pendulum to loop over its pivot?
Solution
The fictitious force in the train’s frame due to the acceleration is equivalent to an additional,
horizontal gravitational field, so the effective gravity is
geff = −ax̂ − gŷ.

p
For small oscillations, we know the period is 2π L/g in ordinary circumstances. By precisely
the same logic, it must be replaced with
s √
L L
T = 2π = 2π 2 .
geff (g + a2 )1/4
As a gets larger, the effective gravity points closer to the horizontal. In the limit g/a → 0,
the effective gravity is just horizontal, so the pendulum oscillates about the horizontal. Its
endpoints are the downward and upward directions, so it never can get past the pivot.
Here’s a follow-up question: if the train can decelerate quickly, how should you stop it so that
the pendulum doesn’t end up swinging at the end? The most efficient way is to first quickly
decelerate to half speed, which, in the frame p of the train, provides a horizontal impulse
to the pendulum. Then wait a half-period π L/g, so that the pendulum’s momentum
turns around, and then quickly stop, providing a second impulse that precisely cancels the
pendulum’s horizontal motion. This cute maneuver is useful for crane operators.
Example 4
If a spring with spring constant k1 and relaxed length `1 is combined with a spring with
spring constant k2 and relaxed length `2 , find the spring constant and relaxed length of the
combined spring, if the combination is in series or in parallel.
Solution
For the series combination, the new relaxed length is clearly ` = `1 + `2 . Suppose the first
spring is stretched by x1 and the second by x2 . The tensions in the springs must balance,
F = k 1 x1 = k 2 x2 .
8
Thus, the new spring constant is

F k2 x2 k1 k2
k= = = .
x1 + x2 x2 (k2 /k1 + 1) k1 + k2
For example, if the spring is cut in half, the pieces have spring constant 2k.
Now consider the parallel combination. In this case it’s clear that the new spring constant
is k = k1 + k2 , since the tensions of the springs add. The new relaxed length ` is when the
forces in the springs cancel out, so
k1 (` − `1 ) + k2 (` − `2 ) = 0
which implies
k1 `1 + k2 `2
`= .
k1 + k2
[2] Problem 10 (Morin 4.20). A mass m is attached to n springs with relaxed lengths of zero. The
spring constants are k1 , k2 , . . . , kn . The mass initially sits at its equilibrium position and then is
given a kick in an arbitrary direction. Describe the resulting motion.
Solution. Suppose the anchor of spring i is at ri . Then the force on the mass is
!
X X
F=− ki (r − ri ) = ki r − C
i i
where C is some constant

q P vector. Thus, we see that the mass undergoes simple harmonic motion
i ki
with frequency ω = m .
[3] Problem 11 (Morin 4.22). A spring with relaxed length zero and spring constant k is attached
to the ground. A projectile of mass m is attached to the other end of the spring. The projectile is
then picked up and thrown with velocity v at an angle θ to the horizontal.
(a) Geometrically, what kind of curve is the resulting trajectory?
(b) Find the value of v so that the projectile hits the ground traveling straight downward.
Solution. (a) Let the anchor of the spring be the origin. Then, the force on the particle is
−kr − mgŷ = −k(r − r0 ), so it is effectively a single spring force. The motion in 2D due to
a spring force is an ellipse (independent x and y oscillations of the same frequency), so the
shape is a portion of an ellipse, whose center is a distance mg/k directly below the launch
point.
(b) Note that the horizontal velocity takes the form vx (t) = (v cos θ) cos(ωt), because the motion
in the horizontal direction is just simple harmonic. The horizontal velocity vanishes when the
phase is π/2, a total of a quarter cycle.
At this time, the vertical displacement must vanish. Vertically, the motion is just simple
harmonic but with an equilibrium point shifted downward by mg/k. Let the vertical velocity
take the form
vy (t) = v0 cos(ωt + φ).
9
The initial phase is φ, and just before the mass hits the ground, its vertical velocity is the
opposite of the original one, so the final phase is π − φ. So hitting the ground occurs at the
same time as having a vertical velocity if the phase difference is π/2, which implies φ = π/4.
Now, by matching the initial velocity and acceleration, we know that
v0 cos φ = v sin θ, −v0 ω sin φ = −g
Dividing these equations gives

g
tan φ =
ωv sin θ
and we must have tan φ = 1, so
r
g m g
v= = .
ω sin θ k sin θ
[5] Problem 12. A uniform spring of spring constant k and total mass m is attached to the wall, and
the other end is attached to a mass M .
(a) Show that when m M , the oscillation frequency is approximately

s
k
ω= .
M + m/3
(b) [A] ? Generalize part (a) to arbitrary values of m/M . (Hint: to begin, approximate the
massive spring as a finite combination of smaller massless springs and point masses, as in
the example in M2. It will not be possible to solve for ω in closed form, but you can get a
compact implicit expression for it. Check that it reduces to the result of part (a) for small
m/M , and interpret the results for large m/M . This is a challenging problem that requires
almost all the techniques we’ve seen so far, so feel free to ask for more hints.)
Solution. In the m M case, we can assume the velocity of a piece of spring that is at position
a fraction x of the total length is proportional to x. (More precisely, accounting for nonlinear
stretching of the spring would contribute at higher order.) Therefore, the total kinetic energy of the
spring is Z L
1 1
(xv0 )2 mdx = mv02
0 2 6
where v0 is the velocity of M , and L is the current length of the spring. Therefore, the total kinetic
energy is 12 (M +m/3)v02 , so we have an effective mass of M +m/3. The spring is uniformly stretched
at this order, so the effective spring constant is still k, giving the desired result.
Part (b) is a nice exercise in dealing with continuum systems, with many possible pitfalls – it
is very easy to spend a lot of time attacking extremely complicated equations. We’ll present a
relatively clean approach.
First, as usual, we break the spring into pieces. Suppose the spring is made of N masses connected
with small springs, and let their displacements from equilibrium be xi . Each piece has mass m/N
and each small spring has spring constant kN , as established in an earlier problem. The equation
of motion for each mass is
m
ẍi = N k(xi−1 + xi+1 − 2xi ).
N
10
We define x0 = 0 and let xN be the position of the mass M . Then its equation is different,
M ẍN = N k(xN −1 − xN ).
The spring is really continuous, so we would like to take the limit N → ∞. To this end, define the
displacement function x(s, t) to be the continuous function with values
x(i/N, t) = xi (t).
The argument s ranges from 0 at the left end of the spring to 1 at the right end. We’ll suppress
the t argument for brevity. Plugging this into the second equation above gives
x(1 − 1/N ) − x(1)

M ẍ(1) = N k(x(1 − 1/N ) − x(1)) = k .
1/N
Upon taking the limit N → ∞, the fraction on the right becomes a derivative, giving
M ẍ(1) = −kx0 (1)
where a prime denotes a derivative with respect to s. Similarly, in the N → ∞ limit, the quantity
N 2 (xi−1 + xi+1 − 2xi ) becomes a second derivative (check this!), so our first equation becomes
mẍ(i/N ) = kx00 (i/N ).

p
Rearranging a bit and defining ω0 = k/M , we have shown that
m ẍ(s) ẍ(1)
= x00 (s), = x0 (1).
M ω02 ω02
Since we are looking for solutions where the whole spring oscillates uniformly with frequency ω, we
plug in x(s) = cos(ωt)f (s) for
m ω2 ω2
2 f = −f 00 , f (1) = f 0 (1).
M ω0 ω02
p
For simplicity we now define α = m/M and set ω0 = 1. Solving the first equation gives
f (s) ∝ sin(αωs)
which yields the expected nonlinear stretching of the spring. The second equation says
ω 2 sin(αω) = αω cos(αω)
or alternatively
α
tan(αω) = .
ω
There are generically infinitely many solutions for ω, which correspond to the infinitely many normal
modes of the spring. However, we’re concerned with the lowest-frequency mode. This is the unique
mode with αω < π/2 where all the pieces of the spring are going in the same direction at the same
time; it is the fundamental frequency.
11
The transcendental equation we have here has no closed form solution, but we can approximate
it. For small α, if we Taylor expand the tangent to third order we recover the answer to the previous
problem. To see this, note that
(αω)3 α
αω + =
3 ω
which can be simplified to
α2 4
ω + ω 2 − 1 = 0.
3
If we parametrize the frequency shift by ω 2 = 1 + , then plugging in gives
α2
+ + (higher order terms) = 0
3
which tells us that
α2 m
=− =−
3 3M
which is the same result found earlier, to first order.
For large α, the right-hand side is large, so the tangent must be large. The lowest frequency mode
has αω ≈ π/2. In this case it’s also useful to look at all the modes, which have αω ≈ (n + 1/2)π.
Restoring the units, this means
1 p
ω ≈ n+ π k/m.
2
To understand this, note that in this limit the mass M doesn’t matter; the spring acts as if it has
a free end. Hence we’ve just found the standing wave frequencies for longitudinal waves with one
fixed and one free end! The lowest frequency is the fundamental.
Jumping ahead a bit, we can compare this with some results from W1. The wavenumbers for
these boundary conditions are
1
kn = n + π
2
and the wave velocity is s
Y
v=
ρ
where Y is the Young’s modulus, and ρ is the mass density. (If this isn’t familiar, you can also
derive it using dimensional analysis.) But this wave velocity can also be written as
s r
kL/A k
v= =L .
m/LA m
Putting these two together using ωn = vkn recovers exactly the frequencies we p found above! So
another way of saying what we’ve done is that we’re derived the expression v = Y /ρ for the speed
of sound.
[2] Problem 13 (PPP 77). A small bob of mass m is attached to two light, unstretched, identical
springs. The springs are anchored at their far ends and arranged along a straight line. If the bob is
displaced in a direction perpendicular to the line of the springs by a small length `, the period of
oscillation of the bob is T . Find the period if the bob is displaced by length 2`.
12
Solution. Suppose it is stretched x in the perpendicular direction, and let θ be the angle of the
spring with respect to the horizontal (the original line of the springs). We see that tan θ = x/L
where L is the unstretched spring length. Now, the change in the length of one of the springs is
L(1/ cos θ − 1) ≈ Lθ2 /2. The restoring force is then proportional to θ2 sin θ ≈ θ3 ∝ x3 . Therefore,
the potential energy is of the form Cx4 for some constant C. Thus, we have
1
mẋ2 + Cx4 = CA4
2
√
where A is the amplitude of oscillation. Thus, ẋ = D A4 − x4 for some constant D, so dt ∝ √Adx
4 −x4
.
R A dx
Thus, the period is proportional to 0 √A4 −x4 , which by dimensional analysis is proportional to
1/A. Thus, by doubling the amplitude, we halve the period, so the answer is T /2 . Note that we
01W
saw a similar idea in P1.
01m
3 Damped and Driven Oscillations

We now review damped oscillators, which we saw in M1, and consider driven oscillators. For more
guidance, see sections 4.3 and 4.4 of Morin.
[2] Problem 16. Consider a damped harmonic oscillator, which experiences force F = −bv − kx.
(a) As in M1, show that the general solution for x(t) is
x(t) = A+ eiω+ t + A− e−iω− t
and solve for the ω± .
(b) For sufficiently small b, the roots are complex. In this limit, show that by taking the real
part, one finds an exponentially damped sinusoidal oscillation. Roughly how many oscillation
cycles happen when the amplitude damps by a factor of e?
(c) For large b, the roots are pure imaginary, the position simply decays exponentially, and we
say the system is overdamped. Find the condition for the system to be overdamped.
Solution. (a) By setting up and solving a quadratic equation,

√ s
−ib ± −b2 + 4mk b 2

ib k
ω± = = ± − .
−2m 2m m 2m
(b) In this limit, we have r

ib k
ω± ≈ ±
2m m
in which case we have √
eiω± t ≈ e−bt/2m ei k/m t
which is an exponentially
√ damped oscillation. The time for a damping of a factor of e is 2m/b,
which occurs after km/πb cycles.
13
(c) This occurs if

2
k b
− <0
m 2m
which implies b2 > 4mk.
[4] Problem 17. Analyzing a damped and driven harmonic oscillator.
(a) Consider a damped harmonic oscillator which experiences a driving force F = F0 cos(ωt).
Passing to complex variables, Newton’s second law is
mẍ + bẋ + kx = F0 eiωt .
If x(t) is a complex exponential, then we know that the left-hand side is still a complex
exponential, with the same frequency. This motivates us to guess x(t) = A0 eiωt . Show that
this solves the equation for some A0 .
(b) Of course, the general solution needs to be described by two free parameters, to match the
initial position and velocity. Argue that it takes the form
x(t) = A0 eiωt + A+ eiω+ t + A− e−iω− t
where the ω± are the ones you found in problem 16.
(c) After a long time, the “transient” A± terms will decay away, leaving the steady state solution
x(t) ≈ A0 eiωt
which oscillates at the same frequency as the driving. The actual position is the real part,
x(t) ≈ |A0 | cos(ωt − φ)
where A0 = |A0 |e−iφ . Evaluate |A0 | and φ.
(d) Sketch the amplitude |A0 | and phase shiftpφ as a function of ω. Can you intuitively see they
take the values they do, for ω small, ω ≈ k/m, and ω large?
(e) There are several distinct things people mean when they speak of “resonant frequencies”.
Find the driving frequency ω that maximizes (i) the amplitude |A0 |, (ii) the amplitude of the
velocity, and (iii) the average power absorbed from the driving force. (As you’ll see, these
are all about the same when the damping is weak, so the distinction between these isn’t so
important in practice.)
Solution. (a) If we plug in x = A0 eiωt , we find the differential equation is satisfied if
(−mω 2 + ibω + k)A0 = F0 ,
which yields
F0
A0 = .
(k − mω 2 ) + ibω
(b) This follows from linearity. If we plug this solution in, then the first term balances the driving
term on the right-hand side. Then the other two terms need to satisfy the damped harmonic
oscillator equation with no driving, so they’re just the same as in problem 16.
14
(c) The answers are

F0 bω
|A0 | = p , tan φ = .
(k − mω 2 )2 + (bω)2 k − mω 2
(d) The amplitude is shown below, for a few values of b (called ζ here).
The phase shift is shown below.
This all makes physical sense. For very small frequency, we are effectively stretching the spring
statically,pso the amplitude approaches a constant |A0 | = F0 /k, and the phase shift is zero.
For ω ≈ k/m, the amplitude is high because we’re driving the oscillator at the frequency it
wants to oscillate at, in the absence of driving and damping. Here, a large power is absorbed
from the driving force, and since P = F v, that means F and v must be approximately in
phase, so the phase shift between F and x is 90◦ . Finally, for high frequencies, the amplitude
goes to zero because the mass doesn’t have time to move far before the force turns around.
In this case, the driving force is always the largest force acting on the mass, so F and a are
in phase, so the phase shift between F and x is 180◦ .
(e) First, to find the maximum |A0 |, it suffices to minimize the square of its denominator. Setting
the derivative of that quantity to zero gives
2b2 ω = 2(k − mω 2 )(2mω)
which can be solved to yield p
ω= k/m − b2 /2m2 .
15
The amplitude of the velocity is

F0 ω F0
v0 = ω|A0 | = p =p
(k − mω 2 )2 + (bω)2 (k/ω − mω)2 + b2
p
which is clearly minimized when ω = k/m. Finally, the rate of power dissipation is
P = F (t)v(t) = −F0 v0 cos(ωt) sin(ωt − φ) = F0 v0 cos(ωt) cos(ωt + (π/2 − φ)).

p
As we’ve just seen, v0 is maximized at ω = k/m. In addition, the average value of the
product of cosines is maximized
p when they are in phase with each other, φ = π/2, which
also happens
p when ω = k/m. Therefore, the maximum average power dissipation occurs at
ω = k/m.
p
[3] Problem 18. The quality factor of a damped oscillator is defined as Q = mω0 /b, where ω0 = k/m.
It measures both how weak the damping is, and how sharp the resonance is.
(a) Show that for a lightly damped oscillator,

average energy stored in the oscillator
Q≈ .
average energy dissipated per radian
Then estimate Q for a guitar string.
(b) Show that for a lightly damped oscillator,

resonance frequency
Q≈
width of resonance curve
where the width of the resonance √curve is defined to be the range of driving frequencies for
which the amplitude is at least 1/ 2 the maximum.
For more about Q, see pages 424 through 428 of Kleppner and Kolenkow.
Solution. (a) Take x = A cos ω0 t. In one cycle, the energy dissipated is

Z 2π/ω0
bv · vdt = bA2 ω0 π,
0
so the average energy dissipated per radian is bA2 ω0 /2. The average energy stored is 12 mω02 A2 ,
so the ratio is mω0 /b = Q.
A typical guitar string oscillates at ∼ 100 Hz for a few seconds. Hence Q ∼ 1000.
F0
(b) We have |A0 | = √ . At the edge of the range that we call the width, we have
m2 (ω02 −ω 2 )2 +(bω)2
m2 (ω02 − ω 2 )2 + (bω)2 = 2(bω0 )2 =⇒ m(ω02 − ω 2 ) = ±bω0 ,
so m(ω0 + ω)(ω0 − ω) = ±bω0 . We have ω ≈ ω0 (to first order), so

1
2mω0 (ω0 − ω) = ±bω0 =⇒ 1 − ω/ω0 = ± .
2Q
Thus the width is approximately ω0 /Q, as desired.
16
The next two problems explore other ways of driving harmonic oscillators.
[2] Problem 19. Consider a pendulum which can perform small-angle oscillations in a plane with
natural frequency f . The pendulum bob is attached to a string, and you hold the other end of the
string in your hand. There are three simple ways to drive the pendulum:
(a) Move the end of the string horizontally with sinusoidal frequency f 0 .
(b) Move the end of the string vertically with sinusoidal frequency f 0 .
(c) Apply a quick rightward impulse to the bob with frequency f 0 .
In each case, for what value(s) of f 0 can the amplitude become large? (This question should be
done purely conceptually; don’t write any equations, just think!)
Solution. (a) In the frame of the string, this is a sinusoidal horizontal (fictitious) force, so it’s
just the same kind of sinusoidal driving we saw above. Resonance happens when f 0 ≈ f .
(b) In this case, there is a sinusoidal vertical force by the same reasoning. Resonance can happen
when f 0 ≈ 2f , in which case gravity is weaker whenever the bob is moving up and stronger
whenever it is moving down. In fact, one can show that weaker resonance occurs for f 0 ≈ 2nf .
This is called parametric resonance, and you can find more details here.
(c) This works as long as the impulse always comes when the object is moving to the right, i.e. in
the same phase of the object’s oscillation. This happens as long as the impulse’s period is an
integer multiple of the object’s period, so f 0 ≈ f /n.
[5] Problem 20. 01r GPhO 2016, problem 1. Record your answers on the official answer sheet.
4 Normal Modes
Idea 4: Normal Modes
A system with N degrees of freedom has N normal modes when displaced from equilibrium.
In a normal mode, the positions of the particles are of the form xi (t) = Ai cos(ωt + φi ).
That is, all particles oscillate with the same frequency. Normal modes can be either guessed
physically, or found using linear algebra as explained in section 4.5 of Morin.
The general motion of the system is a superposition of these normal modes. So to compute
the time evolution of the system, it’s useful to decompose the initial conditions into normal
modes, because they all evolve independently by linearity.
Example 5
Two blocks of mass m are connected with a spring of spring constant k and relaxed length
L. Initially, the blocks are at rest at positions x1 (0) = 0 and x2 (0) = L. At time t = 0, the
block on the right is hit, giving it a velocity v0 . Find x1 (t) and x2 (t).
17
Solution
The equations of motion are
mx¨1 = k(x2 − x1 − L)
mx¨2 = k(x1 + L − x2 ).
The system must have two normal modes. The obvious one is when the two masses oscillate
oppositely, x1 = −x2 . The other one is when the two masses move parallel to each other,
x1 = x2 , and this normal mode formally has zero frequency. The initial condition is the
superposition of these two modes.
We can show this a bit more formally. Define the normal mode amplitudes u and v as
u−v u+v
x1 = , x2 = .
2 2
Solving for u and v, we find
u = x1 + x2 , v = x2 − x1 .
Using the equations of motion for x1 and x2 , we have the equations of motion
ü = 0, mv̈ = −2k(v − L)
which
p just verifies that the normal modes are independent, with frequency zero and ω =
2k/m respectively. We can fit the initial condition if
u(0) = L, v(0) = L, u̇(0) = v0 , v̇(0) = v0 .
The normal mode amplitudes are then

v0
u(t) = L + v0 t, v(t) = L + sin ωt.
ω
Plugging this back in gives
v0 t v0 v0 t v0
x1 (t) = − sin ωt, x2 (t) = L + + sin ωt.
2 2ω 2 2ω
Each mass is momentarily stationary at time intervals of 2π/ω, though neither mass ever
moves backwards. If you didn’t know about normal modes, you could also arrive at this
conclusion by playing around with the equations; you could see that they decouple when you
add and subtract them, for instance.
[3] Problem 21 (Morin 4.10). Three springs and two equal masses lie between two walls, as shown.
18
The spring constant k of the two outside springs is much larger than the spring constant κ k of
the middle spring. Let x1 and x2 be the positions of the left and right masses, respectively, relative
to their equilibrium positions. If the initial positions are given by x1 (0) = a and x2 (0) = 0, and if
both masses are released from rest, show that
x1 (t) ≈ a cos((ω + )t) cos(t), x2 (t) ≈ a sin((ω + )t) sin(t)

p
where ω = k/m and = (κ/2k)ω. Explain qualitatively what the motion looks like. This is an
example of beats, which result from superposition two oscillations of nearly equal frequencies; we
will see more about them in W1.
Solution. The equations of motion are
mẍ1 = −kx1 − κ(x1 − x2 )

mẍ2 = −kx2 − κ(x2 − x1 ).
Again define y1 = x1 + x2 and y2 = x1 − x2 . Adding and subtracting the two EOMs tells us that
mÿ1 = −ky1
mÿ2 = −(k + 2κ)y2 .
The initial conditions are y1 (0) = y2 (0) = a and ẏ1 (0) = ẏ2 (0) = 0. The solution is
p
y1 (t) = a cos( k/mt)
p
y2 (t) = a cos( (k + 2κ)/mt).
Solving for x1 and x2 , we see that

p p ! p p !
k/m + (k + 2κ)/m k/m + (k + 2κ)/m
x1 (t) = a cos t cos − t
2 2
p p ! p p !
k/m + (k + 2κ)/m k/m + (k + 2κ)/m
x2 (t) = a sin t sin − t .
2 2
Note that p p
k/m + (k + 2κ)/m
≈ω
2
and p p
k/m + (k + 2κ)/m p κ/k
− ≈ k/m = ,
2 2
so the result follows. We have an envelope curve of a cos(t) and a sin(t), and a very high frequency
oscillation that matches the envelope. What this looks like is energy gradually sloshing back and
forth between the masses. If the second mass begins still, it will gradually pick up energy, until the
first mass begins still. Then the process repeats in reverse.
[3] Problem 22 (KK 10.11). Two identical particles are hung between three identical springs.
19
Neglect gravity. The masses are connected as shown to a dashpot which exerts a force bv, where v
is the relative velocity of its two ends, which opposes the motion.
(a) Find the equations of motion for x1 and x2 .
(b) Show that the equations of motion can be solved in terms of the variables y1 = x1 + x2 and
y2 = x1 − x2 .
(c) Show that if the masses are initially at rest and mass 1 is given an initial velocity v0 , the
motion of the masses after a sufficiently long time is
v0
x1 (t) = x2 (t) = sin ωt
2ω
and evaluate ω.
Solution. The equations of motion are
M ẍ1 = −kx1 − k(x1 − x2 ) − b(ẋ1 − ẋ2 )

M ẍ2 = −kx2 − k(x2 − x1 ) − b(ẋ2 − ẋ1 ).
Adding the two tells us that

M ÿ1 = −ky1
and subtracting tells us that
M ÿ2 = −3ky2 − 2bẏ2 .
Let us solve for y1 . The initial condition is y1 (0) = 0 and ẏ1 (0) = v0 . Thus,
v0
y1 (t) = sin(ω0 t)
ω0
p
where ω0 = k/M . After a very long time, the y2 terms goes to 0, since it is damped. Thus, we
have that x1 − x2 = 0, so x = x1 = x2 , and
v0
2x = sin(ω0 t)
ω0
so
v0
x1 = x2 = sin(ω0 t)
2ω0
p
where ω0 = k/M .
20
Example 6
Three identical masses are connected by three identical springs, forming an equilateral triangle
in equilibrium. Describe the normal modes of the system.
Solution
Let the system be confined to the xy plane. Then there are three masses that each can move
in two dimensions, giving six degrees of freedom. Since we must be able to construct the
general solution by superposing normal modes, there should be six normal modes. They are:
• Uniform translation in the x or y directions. This yields two normal modes, both formally
with zero frequency, since sin(ωt) ∝ t in the limit ω → 0.
• Uniform rotation about the axis of symmetry.
• A “breathing” motion where the whole triangle expands and contracts.
• A “scissoring” motion where one mass moves outward and the other two move inward.
You might think there are three scissoring normal modes, which would give us one too many.
However, they are redundant: just like how the three sides of the equilateral triangle lie in a
plane, these three normal modes formally lie in a plane, in the sense that you can superpose
any two of them to get the third. Thus we have six normal modes, as expected. If the system
can move in three-dimension space, we need three more; they are uniform translation in the
x direction, and rotation about the x and y axes.
[5] Problem 23 (Morin 4.12, IPhO 1986). N identical masses m are constrained to move on a
horizontal circular hoop connected by N identical springs with spring constant k. The setup for
N = 3 is shown below.
(a) Find the normal modes and their frequencies for N = 2.
(b) Do the same for N = 3.
(c) [A] ? Do the same for general N . (Hint: consider the normal modes found in (a) and (b),
arranged so that in each normal mode, each mass oscillates with unit amplitude but a different
phase. Look at the phases and guess a pattern.)
(d) If one of the masses is replaced with a mass m0 m, qualitatively describe how the set of
frequencies changes.
21
(e) Now suppose the masses alternate between m and m0 m. Qualitatively describe the set of
frequencies.
Part (c) will be useful in X1, where we will quantize the normal modes found here.
Solution. (a) Let the positions of the masses along the circle be x1 and x2 . Then
mẍ1 = −k(2x1 − 2x2 ),

mẍ2 = −k(2x2 − 2x1 ).
p
Adding and subtracting these equations and letting ω0 = k/m gives
ẍ1 + ẍ2 = 0, ẍ1 − ẍ2 = −4ω02 (x1 − x2 )
which tells us the normal mode frequencies are zero and 2ω0 . These correspond to the masses
uniformly rotating around the circle together, and to the two moving oppositely.
(b) Defining quantities analogously to part (a), we have
ẍ1 = −ω02 (2x1 − x2 − x3 ), ẍ2 = −ω02 (2x2 − x1 − x3 ), ẍ3 = −ω02 (2x3 − x1 − x2 ).
If we subtract the first two equations, we get
ẍ1 − ẍ2 = −3ω02 (x1 − x2 )

√
which gives a normal mode with frequency 3 ω0 , where the first two masses move oppositely
and the third doesn’t move at all. The same happens if we subtract the first and third equation,
and second and third equation. Finally, if we add all three equations, we get
ẍ1 + ẍ2 + ẍ3 = 0
which gives
√ a normal mode with frequency zero. Therefore, the normal mode frequencies are
zero and 3 ω0 .
Strangely, it seems like we have four normal modes even though there are only three masses!
The reason is that the first three we found are redundant: if you sum any two of them, you
get the third.
(c) Using the same notation as in the previous parts,
ẍj = −ω02 (2xj − xj−1 − xj+1 ), j = 1, 2, . . . N.
Solving these equations requires an inspired guess. Notice that in part (a), the zero frequency
normal mode has x1 and x2 in phase, while the other normal mode has them π out of phase.
In part (b), the zero frequency normal mode has x1 , x2 , and x3 in phase. The other two
normal modes can be expressed as having phases
       
x1 1 x1 1
x2  = e2πi/3  eiωt , x2  = e−2πi/3  eiωt .
x3 e4πi/3 x3 e−4πi/3
In other words, in each normal mode, the phase differences between adjacent masses are
uniform! Therefore, we are inspired to guess
2πn
xj = eiωt eikj , k=
N
22
for integer n. Plugging this in, the differential equations simplify to
ω 2 = ω02 (2 − e−ik − eik )
which simplifies to
k πn
ω = 2ω0 sin = 2ω0 sin .
2 N
For n = 0, . . . , N − 1, these are the normal mode frequencies.
As an aside, for n N , note that ωn is proportional to n, which is also proportional to the
wavenumber k of the wave. This indicates that waves built out of only normal modes with
n N travel with constant velocity, and hence satisfy the ideal wave equation. In general,
systems that satisfy the ideal wave equation are often found by taking the low n limit of a
system with many discrete parts. We’ll see these points in more detail in W1.
As another aside, we could have also gotten the solution without “inspiration”. As we’ve
discussed above, guessing a complex exponential eiωt is the general technique when dealing
with linear equations with time translation symmetry. Similarly, in this problem we considered
linear equations with a discrete spatial translational symmetry, i.e. the equations stay the same
upon substituting j → j + 1. Thus, the general technique is to guess a complex exponential
in j, which is precisely the eikj factor.
p
(d) When we add the one light mass, it adds a new normal mode with frequency 2k/m0 , where
the light mass oscillates back and forth and nothing else moves. The band of frequencies from
zero to 2ω0 barely changes.
(e) Naively,
p if we turn half the masses into light masses, we get N/2 modes with frequency
0
2k/m , consisting of each light mass oscillating independently. But this isn’t right, because
we must take sinusoidal combinations of these modes to get normal modes, by the same logic
as we used in thepprevious parts. This broadens the normal mode frequencies into a band
centered around 2k/m0 . Meanwhile, for the low-frequency modes, the heavy masses can’t
even see the light masses, so it’s as if every spring has been doubled
p in length. We hence have
a second band of normal modes with frequencies centered on k/2m, which is nonoverlapping
if m0 m. This idea of normal mode frequencies filling dense but separated bands is crucial
in solid state physics.
[4] Problem 24. [A] In this problem, you will analyze the normal modes of the double pendulum,
which consists of a pendulum of length ` and mass m attached to the bottom of another pendulum,
of length ` and mass m. To solve this problem directly, one has to compute the tension forces in
the two strings, which are quite complicated. A much easier method is to use energy.
(a) Parametrize the position of the pendulum in terms of the angle θ1 the top string makes with
the vertical, and the angle θ2 the bottom string makes with the vertical. Write out the kinetic
energy K and the potential energy V to second order in the θi and θ˙i .
(b) The Euler–Lagrange equations for the system are
d ∂K ∂V
=− .
dt ∂ θ˙i ∂θi
23
Using the results of part (a), write these equations in the form
θ¨1

g θ
¨ =− A 1
θ2 L θ2
where A is a 2 × 2 matrix. This is a generalization of θ̈ = −gθ/L for a single pendulum.
(c) Find the normal modes and their frequencies, using the general method in section 4.5 of Morin.
Solution. (a) To second order, the horizontal displacements of the masses are
x1 = `θ1 , x2 = `(θ1 + θ2 )
which gives a kinetic energy of

m`2 ˙ 2
K= ((θ1 ) + (θ˙1 + θ˙2 )2 ).
2
The vertical displacements are
y1 = `(1 − cos(θ1 )), y2 = `(2 − cos(θ1 ) − cos(θ2 ))
and expanding the cosines to second order gives

` `
y1 = θ12 , y2 = (θ12 + θ22 )
2 2
which gives a potential energy of
mg` 2
V = (2θ1 + θ22 ).
2
(b) The resulting Euler–Lagrange equations are

2g g
2θ¨1 + θ¨2 = − θ1 , θ¨1 + θ¨2 = − θ2 .
` `
Solving the system, we find
2 −1
A=
−2 2
straightforwardly.
(c) We must find the eigenvalues of the matrix, which obey the equation
(2 − λ)2 − 2 = 0
√
which implies λ = 2 ± 2. The normal mode amplitudes are

1
√ 1
high frequency : , low frequency : √
− 2 2
√
2 = (g/L)(2 ± 2).
and the frequencies are ω±
5 [A] Adiabatic Change
24
Idea 5
When a problem contains two widely separate timescales, such as a fast oscillation superposed
on a slow overall motion, one can solve for the fast motion while neglecting the slow motion,
then solve for the slow motion by replacing the fast motion with an appropriate average.
Example 7: MPPP 21
A small smooth pearl is threaded onto a rigid, smooth, vertical rod, which is pivoted at
its base. Initially, the pearl rests on a small circular disc that is concentric with the rod,
and attached to it a distance d from the rotational axis. The rod starts executing simple
harmonic motion around its original position with small angular amplitude θ0 .
What frequency of oscillation is required for the pearl to leave the rod?
Solution
The reason the pearl leaves the rod is that the normal force rapidly varies in direction, with
an average upward component. If this average upward force is greater than gravity, the pearl
accelerates upward and leaves the rod.
In this case, the fast motion is the oscillation of the rod, while the slow motion is the rate of
change of the pearl’s distance from the pivot, which can be neglected during one oscillation.
The pearl has horizontal displacement and acceleration
x(t) = −d sin θ ≈ −dθ(t) = −θ0 d sin ωt, ax (t) = θω 2 d sin ωt.
This is supplied by the horizontal component of the normal force. The vertical component is
Ny = Nx tan θ(t) ≈ max (t)θ(t) = mθ02 ω 2 d sin2 ωt.
Now we average over the fast motion to understand the slow motion. Since the average value
of sin2 (ωt) is 1/2, the condition for the pearl to go up is
1 2 2
mθ ω d > mg
2 0
which gives r
1 2g
ω> .
θ0 d
25
Example 8
A mass m oscillates on a spring with spring constant k0 with amplitude A0 . Over a very
long period of time, the spring smoothly and continuously weakens until its spring constant
is k0 /2. Find the new amplitude of oscillation.
Solution
In this case the fast motion is the oscillation of the mass, while the slow motion is the
weakening of the spring. We can solve the problem by considering how the energy changes
in each oscillation, due to the slight decrease in k.
Suppose that the spring constant drops in one instant by a factor of 1 − . Then the kinetic
energy stays the same, while the potential energy drops by a factor of 1 − . Since the kinetic
and potential energy are equal on average, this means that if the spring constant gradually
decreases by a factor of 1 − x over a full cycle, with x 1, then the energy decreases by a
factor of 1 − x/2.
The process finishes after N oscillations, where (1 − x)N ≈ √e−N x = 1/2. At this point, the
energy has dropped by a √ N
factor of (1 − x/2) ≈ e −N x/2 = 1/ 2. But the energy is also kA2 ,
so the new amplitude is 4 2A0 .
Amazingly, the question can also be solved in one step using a subtle conserved quantity.
Solution
Sinusoidal motion is just a projection of circular motion. In particular, it’s equivalent
to think of the mass as being tied to a spring of zero rest length attached to the origin,
and performing a circular orbit about the origin, with the “actual” oscillation being the x
component. (This is special to zero-length springs obeying Hooke’s law, and occurs because
the spring force −kx = −k(x, y) has its x-component independent of y, and vice versa.)
Since the spring constant is changed gradually, the orbit has to remain circular. Then angular
momentum is conserved, and we have
√
L ∝ vr = ωA2 ∝ kA2 .
√
Then the final amplitude is 4 2A0 as before.
Both of these approaches are tricky. The energy argument is very easy to get wrong, while the
angular momentum argument seems to come out of nowhere and is inapplicable to other situations.
But angular momentum turns out to be a special case of a more general conserved quantity, which
is useful in a wide range of similar problems.
Idea 6: Adiabatic Theorem
If a particle performs a periodic motion in one dimension in a potential that changes very
26
slowly, then the “adiabatic invariant”

I
I= p dx
is conserved. This is the area of the orbit in phase space, an abstract space whose axes are
position and momentum.
Solution
Using conservation of energy,
p2 1
E= + kx2 .
2m 2
Therefore, the curve p(x) over one√oscillation p
cycle traces out an ellipse in phase space, with
semimajor and semiminor axes of 2mE and 2E/k. The area of this ellipse is the adiabatic
invariant,
√ √
I r
p m
I = p dx = π 2mE 2E/k = 2πE ∝ A2 km.
k
Thus, A ∝ k −1/4 in an adiabatic change of k, recovering the answer found earlier.
Remark
The existence of the adiabatic invariant is hard to see in pure Newtonian mechanics, but
it falls naturally out of the framework of Hamiltonian mechanics, which works with phase
space. In fact, Hamiltonian mechanics makes a lot of theoretically useful facts easier to see.
For example, as you will see in X1 using quantum statistical mechanics, the conservation of
the adiabatic invariant for a single classical particle implies the conservation of the entropy
for an adiabatic process in thermodynamics! The two meanings of “adiabatic” are actually
one and the same. If you’d like to learn more about Hamiltonian mechanics, see David Tong’s
lecture notes or chapter 15 of Morin.
[3] Problem 25. Consider a pendulum whose length adiabatically changes from L to L/2.
(a) If the initial (small) amplitude was θ0 , find the final amplitude using the adiabatic theorem.
(b) Give a physical interpretation of the adiabatic invariant.
(c) When quantum mechanics was being invented, it was proposed that the energy in a pendulum’s
oscillation was always a multiple of ~ω, where ω is the frequency. At the first Solvay conference
of 1911, Lorentz asked whether this condition would be preserved upon slow changes in the
length of the pendulum, and Einstein replied in the affirmative. Reproduce Einstein’s analysis.
Solution. (a) Using the small angle approximation, we have

1 1
E = mv 2 + mgLθ2
2 2
27
and the adiabatic invariant is

I I I
p dx = L p dθ = mL v dθ.
On the other hand, from

p conservation
p of energy, we know that v(θ) is an ellipse with semimajor
and semiminor axes 2E/m and 2E/mgL, so
I s
p p L
p dx ∝ mL E/m E/mgL = E .
g
The total energy is E = mgLθ02 /2, so

I
p dx ∝ θ02 L3/2 g 1/2
which implies that when L halves, the amplitude becomes 23/4 θ0 . Since we kept track of
factors of g, this derivation also tells us what happens to the amplitude if g is slowly changed.
The most famous literary example of a pendulum with changing length appears in Edgar Allan
Poe’s short story, The Pit and the Pendulum. In the story, the narrator is strapped to a table,
and sees a pendulum above him slowly moving and lengthening, bringing its razor edge toward
him. Poe describes the pendulum’s amplitude as initially small, but “increasing inexorably”.
As you can see from the solution to this problem, precisely the opposite happens. When the
length goes up, the amplitude goes down, so it would just gently land on the narrator’s chest.
(b) As for the case of a mass on a spring, we can add a third dimension and let the pendulum
oscillate in a horizontal circle. Then the adiabatic invariant is simply
I
Lz dθ = 2πLz ∝ Lz
which is the angular momentum in the z-direction.
(c) Given the way we did part (a), this is immediate. The adiabatic invariant is
s
L E
E = .
g ω
Therefore, E/ω remains an integer multiple of ~ under adiabatic change.

[4] Problem 26. A block of mass M and velocity v0 to the right approaches a stationary puck of
mass m M . There is a wall a distance L to the right of the puck.
(a) Assuming all collisions are elastic, find the minimum distance between the block and the wall
by explicitly analyzing each collision. (Note that it does not suffice to just use the adiabatic
theorem, because it applies to slow change, while the collisions are sharp. Nonetheless, you
should find a quantity that is approximately conserved after many collisions have occurred.)
(b) Approximately how many collisions occur before the block reaches this minimum distance?
(c) The adiabatic index γ is defined so that P V γ is conserved during an adiabatic process. In one
dimension, the volume V is simply the length, and P is the average force. Using the adiabatic
theorem, infer the value of γ for a one-dimensional monatomic gas.
28
Solution. (a) Let the speeds of the block and puck be v and w. Every collision, w increases by 2v.
If the block is a distance x from the wall, then a collision happens in time 2x/w. Therefore,
we have
∆w 2v w
= =− .
∆x −2xv/w x
Because m M , many collisions happen. After many collisions have happened, w will be
very large, so in the next collision, ∆x will be small compared to x, and ∆w will be small
compared to w. In this case, we can approximate the finite differences with a derivative,
giving
dw w
≈− .
dx x
Separating and integrating shows that wx is conserved. We could also have arrived at this by
the adiabatic theorem, I
I= p dx = mw(2x) ∝ wx.
However, in the earlier collisions (∆w)/w and (∆x)/x aren’t small, so this reasoning is invalid.
For instance, wx is zero before the first collision and nonzero right after it. Thus, we must
treat the first few collisions manually. Right before the second collision, we have
w ≈ 2v0 , x ≈ L/3
by one-dimensional kinematics. Right before the third collision we have
w ≈ 4v0 , x ≈ L/5
where for these early few collisions we are treating v as constant since m M . It is not hard
to show that right before collision n + 1, we have w ≈ 2nv0 and x ≈ L/(2n + 1), which means
that after many (but not too many collisions) we have wx ≈ Lv0 . Then, for future collisions,
wx stays at this value.
The block turns around when the puck has all its energy, so
1 1
M v02 = mw2 .
2 2
p
Plugging in wx = Lv0 and solving for x gives the solution, x = L m/M .
(b) At each collision we have ∆w = 2v, and energy conservation gives

m 2
v2 + w = v02 .
M
Therefore, the number of collisions is approximately
r r
Z Z Z 1
dw 1 dw 1 M dx π M
n≈ = p = √ = .
2v 2 2
v0 − (m/M )w 2 2 m 1−x 2 4 m
0
Note that we didn’t need to separate out the first few collisions here, even though the approx-
imation as an integral technically doesn’t work, because they’re just that not important for
calculating the total number of collisions, which is large. The appearance of π in this result
has a nice geometric interpretation, as explained here.
29
(c) The analogue of pressure in one dimension is just force. The average force exerted by the
puck, which we now think of as a gas molecule, is
∆p 2mw mw2
F = = = .
∆t 2x/w x
Meanwhile, the analogue of volume is one dimension is simply x. Then the conservation of
wx says that F x3 is conserved, which means γ = 3. This is exactly what we would expect for
a one-dimensional gas, where Cv = kT /2 and Cp = 3kT /2.
[4] Problem 27 (F = ma, BAUPC). Two particles of mass m are connected by pulleys as shown.
The mass on the left is given a small horizontal velocity v, and oscillates back and forth.
(a) Without doing any calculation, which mass is higher after a long time?
(b) Verify your answer is right by computing the average tension in the leftward string, in the
case where the other end of the string is fixed, for amplitude θ0 1.
(c) Let the masses begin a distance L from the pulleys. Find the speed of the mass which
eventually hits the pulley, at the moment it does, in terms of L and the initial amplitude θ0 .
Solution. (a) The mass on the right will be higher. If the masses didn’t move up or down, both
would have the same average y-component of tension. But the mass on the left also has
an x-component of tension, so its average magnitude of tension would be higher. This is a
contradiction; to make the tension constant throughout the rope the mass on the right must
rise.
(b) The tension provides the centripetal force, so
mv 2
T = + mg cos θ
r
where θ is the angle from the vertical. By energy conservation, the first term is mgr(cos θ −
cos θ0 ) where θ0 is the amplitude, so
T = (3 cos θ − 2 cos θ0 )mg.
Since the amplitude is small, we Taylor expand for

2 3 2
T = 1 + θ0 − θ mg.
2
Again using the small angle approximation, so that the motion is approximately simple
harmonic, the average value of θ2 is θ02 /2, so

2 3 2
T = 1 + θ0 − θ0 mg > mg
4
as expected.
30
(c) As we’ve seen above,

1 2
T = 1 + θ mg
4
where θ is the amplitude. Let x be the distance the right mass has risen. From the standpoint
of the left mass, it is simply a pendulum whose length is being adiabatically lengthened, so
by the result of problem 25, we have
mg 2L 3/2 2 dx
Z Z
mg 2 1
T − mg dx = L θ0 −3/2 = Lθ0 1 − √ .
4 L x 2 2
This is the increase in kinetic energy of the right mass, so setting this equal to mv 2 /2 gives
s
1
v = gL 1 − √ θ0 .
2
31
Electromagnetism I: Electrostatics
The material here is covered at the right level in chapters 1–3 of Purcell. For a separate introduction
to vector calculus, see the resources mentioned in the syllabus, or chapter 1 of Griffiths. Electrostatics
is covered in more mathematical detail in chapter 2 of Griffiths. For interesting general discussion,
see chapters II-1 through II-5 of the Feynman lectures. There is a total of 80 points.
1 Coulomb’s Law and Gauss’s Law

We’ll begin with some basic problems which can be solved with symmetry arguments.
Idea 1
Gauss’s law is written in integral form as
I
Q
E · dS = .
0
In practice, you will only apply this form to situations with high symmetry, where

2
Q/4π0 r spherical symmetry,

E = λ/2π0 r cylindrical symmetry,

σ/20 infinite plane.

Example 1
Consider a spherical shell of uniform surface charge density σ. A small hole is cut out of the
surface of the shell. What is the electric field at the center of this hole?
Solution
We use the principle of superposition. First, consider the entire spherical shell, without a hole.
By Gauss’s law and spherical symmetry, the radial electric field at a point P infinitesimally
outside the sphere is σ/0 , while the electric field at a nearby point P 0 infinitesimally inside
is zero.
This field is the superposition of the fields of the charges near P and P 0 , and charges from
the entire rest of the sphere. Consider the effect of a small piece of the surface, near P and
P 0 . From the perspective of these points, this piece looks like an infinite plane, so its radial
electric field is σ/20 at P , and −σ/20 at P 0 . Therefore, the entire rest of the sphere must
contribute a radial electric field of σ/20 , at both P and P 0 . Therefore, when one cuts out a
hole, this is the only contribution that remains, so the field is just σ/20 .
[2] Problem 1 (Griffiths 2.18). Some questions about uniformly charged spheres.
(a) Consider a sphere of radius R and uniform charge density ρ. Find the electric field everywhere.
1
(b) Now two spheres, each of radius R and carrying uniform charge densities ρ and −ρ, are placed
so that they partially overlap. Call the vector from the positive center to the negative center
d. Find the electric field in the overlap region.
Solution. (a) The field inside a uniform sphere of density ρ and center a is
ρ
E= (r − a).
30
Outside the sphere, the field falls off as an inverse square,
ρ R3
E= (r − a).
30 |r − a|3
(b) If the two centers are a1 and a2 , then by superposition,

ρ ρ
E= ((r − a1 ) − (r − a2 )) = d
30 30
which is a constant.
[2] Problem 2. A charge q sits just inside a cube, next to one of the corners. What is the flux through
each face of the cube? (More precisely, if a corner is at the origin, and the sides are parallel to the
x, y, and z axes, let the charge be at coordinates (, , ) for tiny .)
Solution. There are three “opposite” faces with the same flux, and three “adjacent” faces with
the same flux. Now consider adding seven more cubes, so that the charge is now at the center of
a 2 × 2 × 2 cube. The total flux through the outer faces of the cube is q/0 , and there are 24 unit
faces, so the flux out of each “opposite” face is q/240 . Now consider the original cube. By Gauss’s
law the total flux out must be q/0 , which means the flux out of each “adjacent” face is 7q/240 .
(Note that if the charge were instead exactly at one of the corners, the fluxes through the opposite
faces would still be q/240 , while the fluxes though the adjacent faces would technically be undefined,
since the electric field blows up on the face. But roughly speaking, the flux ought to be zero. Then
the total flux out of the cube is only q/80 , and that’s because the corner cuts out one “octant” of
the point charge’s field.)
[2] Problem 3 (BAUPC). In both parts below, take the potential to be zero at infinity.
(a) Consider a solid sphere of uniform charge density. Find the ratio of the electrostatic potential
at the surface to that at the center.
(b) Consider a solid cube of uniform charge density. Find the ratio of the electrostatic potential
at a corner to that at the center. (Hint: use symmetry.)
Solution. (a) Let the uniform charge density be ρ and the sphere have radius R, so the total
charge is Q = 43 πρR3 . We can treat the field outside the sphere to be like a point charge, so
the potential at the surface relative to infinity is U0 = Q/4π0 R.
To go from the surface
R 0 to the center, we need to go against the field lines and change the
potential by ∆U = − R E(r) dr. The electric field inside the sphere can be found with Gauss’s
law inside the sphere: E = 43 πρr3 /4π0 r2 .
Z R
kQr 1
∆U = 3
dr = U0
0 R 2
Thus the potential at the center of the sphere is U0 + 12 U0 = 32 U0 , so the ratio of the potential
at the surface to that at the center is 23 .
2
(b) Being at the center of the cube is like being at the corner of 8 identical cubes with half the
length. From U ∼ kQ/r ∼ kρr2 , we see that the potential is proportional to the square of the
length scale. Let the potential at the corner be U0 . For each cube with half the length, the
potential from that cube is 14 U0 . With eight of those half cubes at the center, the potential
at the center of the cube is 2U0 . So the ratio of the potentials at corner to center is 12 .
Idea 2
If you follow an electric field line, the potential monotonically decreases along it.
[2] Problem 4. Two questions about electrostatic equilibrium.
(a) Prove that when a system of point charges is in equilibrium (i.e. the net force on each of the
charges due to the others vanishes), the total potential energy of the system is zero.
(b) Show that for a positive point charge in the electric fields of fixed, positive point charges,
there is a path along which the charge can be moved to infinity without ever needing positive
external work, i.e. a path along which the potential only decreases.
Solution. (a) Fix some point O not on any of the charges, and scale the system up about O
continuously, to send all the charges to infinity. At all points in time, there are no forces
on any of the charges, so no work is done. The final potential energy is zero, so the initial
potential energy must also have been zero.
(b) Consider the field line going through the test charge. It can’t end on a negative charge, since
there are none, so it must end at infinity. Moving the charge along this field line gives the
desired path.
Idea 3
Gauss’s law is written in differential form as
ρ
∇·E= .
0
The divergence of a vector field F = Fx x̂ + Fy ŷ + Fz ẑ is
∇ · F = ∂x Fx + ∂y Fy + ∂z Fz
in Cartesian coordinates, where ∂x stands for ∂/∂x, and so on.
Example 2
Show that the two forms of Gauss’s law are equivalent.
3
Solution
To do this, we need to establish the geometric meaning of the divergence. For simplicity we
consider two dimensions; the proof for three dimensions is similar. Consider a small rectangle
prism with one corner at the origin, with axes aligned with the Cartesian coordinate axes
and side lengths ∆x and ∆y. To apply Gauss’s law in integral form, we need to compute the
flux through each side. The flux going out the top side is
Z ∆x
Ey (x, ∆y) dx
0
while the flux going out the bottom side is

Z ∆x
− Ey (x, 0) dx.
0
The sum of these two terms is

Z ∆x Z ∆x
(Ey (x, ∆y) − Ey (x, 0)) dx ≈ ∆y (∂y Ey )|(x,0) dx
0 0
where we applied a tangent line approximation, and the subscript indicates where the
function ∂y Ey is evaluated. Higher-order terms in the Taylor series would be proportional to
higher powers of ∆y, which is small, so we can ignore them.
The integrand is still a function of x, but we can Taylor expand it about the origin as
(∂y Ey )|(x,0) = (∂y Ey )|(0,0,0) + ∆x(. . .) + . . . .
These extra terms are again higher-order in ∆x and ∆y, so we ignore them. The net flux
through the top and bottom faces is hence, to lowest order,
Z ∆x
∆y (∂y Ey )|(0,0) dx = ∆x∆y (∂y Ey )|(0,0) .
0
By similar reasoning, pairing up the left and right faces gives
flux = ∆x∆y (∂x Ex + ∂y Ey )|(0,0) = ∆x∆y (∇ · E)|(0,0) .
Thus the divergence is the outgoing flux per unit area, or volume in three dimensions.
This shows us why the two forms of Gauss’s law area equivalent. For example, starting from
the differential form, the left-hand side is the flux per volume, while the right-hand side is
the charge per volume, divided by 0 . Integrating both sides over some volume relates the
total flux to the total charge divided by 0 , which is Gauss’s law in integral form.
If the above derivation was a bit abstract, we can also show the idea using specific examples.
4
Example 3
Suppose the region 0 < x < d has charge density −ρ, and the region −d < x < 0 has charge
density ρ. Find the electric field everywhere.
Solution
By translational symmetry, the field always points along x̂ and only depends on x, E(r) =
E(x) x̂. By applying the integral form of Gauss’s law to a rectangular prism, with one side
at xl and another at xr , we have
1 xr 1 x
Z Z
E(xr ) − E(xl ) = ρ(x) dx, E(x) = ρ(x) dx + E0 .
0 x l 0 0
Since the divergence of E(r) is just ∂E(x)/∂x, this clearly satisfies the differential form of
Gauss’s law. To fix the undetermined constant E0 , we could demand the field be zero on
both sides of the charge distribution, motivated by symmetry. Then we have

d − x 0 < x < d,
ρ 
E(x) = × d + x −d < x < 0,
0 
0 elsewhere.

Example 4
Find the electric field of a spherically symmetric charge density ρ(r).
Solution
By spherical symmetry, the field always points radially and only depends on r, E(r) = E(r) r̂.
By applying the integral form of Gauss’s law to a sphere of radius r,
1 r 0 1 1 r 0 02 0
Z Z
2 02 0
4πr E(r) = dr 4πr ρ(r ), E(r) = dr r ρ(r ).
0 0 0 r 2 0
Let’s check that this indeed satisfies the differential form of Gauss’s law, using the divergence
in spherical coordinates. For any vector field F = Fr r̂ + Fθ θ̂ + Fϕ ϕ̂, the divergence is
1 ∂(r2 Fr ) 1 ∂ 1 ∂Fϕ
∇·F= 2
+ (Fθ sin θ) + .
r ∂r r sin θ ∂θ r sin θ ∂ϕ
This looks complicated, but things turn out simple because E only has a radial component,
Er = E(r), which gives
Z r
1 ∂(r2 E(r)) 1 ∂ r2 ρ(r) ρ(r)
∇·E= 2 = 2 dr0 r02 ρ(r0 ) = 2 =
r ∂r r 0 ∂r 0 r 0 0
just as desired.
5
[3] Problem 5. Consider a vector field expressed in polar coordinates, F = Fr r̂ + Fθ θ̂ where r̂ and θ̂
are unit vectors in the radial and tangential directions. Gauss’s law in differential form still works
in these coordinates, but the form of the divergence is different.
By considering the flux per unit area out of a small region bounded by r and r + dr, and θ and
θ +dθ, and applying Gauss’s law in integral form, find what the divergence in polar coordinates must
be for Gauss’s law in differential form to hold. (Optional: try generalizing to spherical coordinates.)
Solution. By summing up contributions from each of the four sides, and letting (Fr , Fθ ) be the
vector field at one of the corners, the flux through the region is
dΦ = (Fr + dFr )((r + dr)dθ) − Fr (rdθ) + (Fθ + dFθ )dr − Fθ dr.
In two dimensions, the divergence is the flux per area, dA = r dr dθ, so

dΦ 1 ∂(rFr ) 1 ∂Fθ
∇·F= = + .
dA r ∂r r ∂θ
[4] Problem 6. This problem is quite subtle, but will enhance your understanding of electromagnetism.
Suppose that all of space is filled with uniform charge density ρ.
(a) Show that E = (ρ/0 )xx̂ obeys the differential form of Gauss’s law.
(b) Show that E = (ρ/30 )rr̂ also obeys Gauss’s law.
(c) Argue that by symmetry, E = 0. Show that this does not obey Gauss’s law.
(d) ? What’s going on? Which, if any, is the actual field? If you think there’s more than one
possible field, how could that be consistent with Coulomb’s law, which gives the answer
explicitly? For that matter, what does Coulomb’s law say about this setup, anyway?
Solution. (a) We see that ∇ · E = ∂x ((ρ/0 )x) = ρ/0 , as desired.
(b) In Cartesian coordinates, this field is

ρ
E= (xx̂ + yŷ + zẑ)
30
whose divergence is ρ/0 , as desired.
(c) This has to hold by symmetry because the electric field can’t point in any particular direction,
by rotational symmetry. It also can’t just point radially, because that breaks translational
symmetry; the center is a special point. So the only option is E = 0, but ∇ · E = 0, so Gauss’s
law is not obeyed.
(d) The issue is boundary conditions. Just like any differential equation, the solution for the
electric field is not defined without boundary conditions (or initial conditions, as we called
them in mechanics). Usually, we get a unique solution by demanding the fields go to zero at
infinity. (Though in some cases, this might not be the right physical answer. For example,
the electric field of a capacitor in the lab doesn’t have to be zero outside, because it might
be inside some bigger capacitor.) However, we can’t do this here because the charge density
goes to infinity too. By taking different choices of boundary conditions, we can get (a), (b),
or many other answers. The symmetry argument in (c) fails, because any choice of boundary
conditions will break the perfect translational symmetry.
6
At first glance, it could seem that Coulomb’s law could give us a unique answer. Coulomb’s
law for a point charge is itself derived by implicitly assuming that there are no “extra” fields
flying around, just the spherically symmetric field of the point charge itself. This looks very
reasonable, so what stops us from just saying that each charge in this problem has such a field,
and then integrating over the charges? R ∞Well, if you write down the integral, you’ll find that it’s
divergent, analogous to the integral −∞ x dx. By itself, the integral is not even well-defined.
In order to get an answer, you have to “regulate” the integral (i.e. change it in a way that makes
it well-defined). One possible
RL regulator, for example, is to just chop off the limits of integration
at finite values, like −L x dx. But that particular regulator is equivalent to just replacing
the charge distribution with a finite one centered at the origin! In other words, Coulomb’s
law also fails to give a unique answer, because it requires a regulator to give a well-defined
answer, and there are many possible regulators. If you treat the charge distribution as a giant
ball with center at the origin, you get the result of part (b). If you treat it as a thick, huge
slab along the yz plane centered at the origin, you get the result of part (a). The symmetry
argument fails once again, because all the regulators break translational symmetry. This is a
simple example of an “anomalous symmetry”, an important idea in theoretical physics.
The exact same problem applies to Newtonian cosmology, where charge density is replaced
with mass density, and this problem confused Newton himself, who incorrectly thought that
g = 0 by symmetry. In this context, all regulators/boundary conditions are unsatisfactory.
Of course, we want a rotationally symmetric universe to match experiment, so we have to put
that in by hand. But then every solution has a center towards which everything collapses, so
to keep the solar system an inertial frame, we’d have to put it at the center of the universe!
Surely, this would make Copernicus roll in his grave.
Some of these problems are fixed in general relativity. You still have to postulate rotational
symmetry (again, on the basis of experimental data), but once you do that, there are no further
problems. In general relativity, acceleration is not absolute in the way it is in Newtonian
mechanics. Instead, there is no center; everything just gets closer to everything else. For
further discussion and references, see this paper.
Idea 4
A tricky, occasionally useful idea is to use Newton’s third law: it may be easier to calculate
the force of A on B than the force of B on A.
Example 5: Purcell 1.28
Consider a point charge q. Draw any imaginary sphere of radius R around the charge. Show
that the average of the electric field over the surface of the sphere is zero.
Solution
Imagine placing a uniform surface charge σ on the sphere. Then the average of the point
charge’s electric field over the sphere times 4πR2 σ is the total force of the point charge on
the charged sphere. But this is equal in magnitude to the force of the charged sphere on
the point charge, which must be zero by the shell theorem. Thus the average field over the
sphere has to vanish.
7
[3] Problem 7 (Purcell 1.28). Some extensions of the previous example.
(a) Show that if the charge q is instead outside the sphere, a distance r > R from its center, the
average electric field over the surface of the sphere is the same as the electric field at the center
of the sphere.
(b) Show that for any overall neutral charge distribution contained within a sphere of radius R,
the average electric field over the interior of the sphere is −p/4π0 R3 where p is the total
dipole moment.
Solution. The same Newton’s third law trick will work for both parts.
(a) Let the desired answer be Eavg and let the charge q be at r. Now imagine a charge Q
is uniformly distributed over the surface of the sphere. The force of the charge q on the
distributed charge Q is precisely FqQ = QEavg . But we also know that
kQq
FqQ = −FQq = − r̂
r2
by Newton’s third law and the shell theorem. Therefore we have
kq
Eavg = − r̂
r2
which is precisely the electric field at the center of the sphere due to q. (Note that r̂ points
from the center of the sphere to the charge q.)
(b) Let the desired answer be Eavg . Now imagine a charge Q is uniformly distributed over the
volume of the sphere. The force of the charge distribution (with charge density ρ(x)) on the
distributed charge Q is precisely FqQ = QEavg . But we also know that
Z
FqQ = −FQq = − ρ(r)EQ (r) dr
where EQ is the field due to Q. Now, this field is easy to find, as it is just the field of a
uniformly charged sphere, so
kQ
EQ = 3 r
R
as shown in problem 1. Putting this in the integral, we have
Z
kQ
QEavg = − 3 ρ(r)r dr
R
so by the definition of the dipole moment,
Z
k kp
Eavg =− 3 ρ(r)r dr = −
R R3
as desired.
[3] Problem 8. There are two point charges, q1 > 0 and q2 < 0, in empty space. An electric field line
leaves q1 at an angle α from the line connecting the two charges. Determine whether this field line
hits q2 , and if so, at what angle β from the line connecting the two charges. (Hint: this can be done
without solving any differential equations.)
8
Solution. Suppose the field line does hit q2 . Rotate the field line about the line connecting the two
charges, to form a Gaussian surface. Because no electric field lines go across this surface, the total
charge inside must be zero. Now, this surface envelopes “slices” of each point charge. (If you’re
not happy with “slicing a point charge”, just replace the point charges with tiny uniformly charged
spheres; everything stays the same.) The solid angle of the first point charge enveloped is
Z Z 2π Z α
dΩ = dφ sin θ dθ = 2π(1 − cos α)
0 0
so the amount of charge enclosed is

Ω 1 − cos α α
q1 = q1 = q1 sin2 .
4π 2 2
Reasoning similarly for the other surface, we have
α β
q1 sin2 = |q2 | sin2
2 2
and the field line hits q2 if there is a solution for β, i.e. when |q1 /q2 | sin2 (α/2) < 1.
Idea 5
R
The integral dS over a surface with a fixed boundary is independent of the surface.
We proved this in a mechanical way in M2. If you want to see a proof using vector calculus,
see problem 1.62 of Griffiths.
[3] Problem 9. A hemispherical shell of radius R has uniform charge density σ and is centered at the
origin. Find the electric field at the origin. (Hint: combine the previous two ideas.)
Solution. Place a point charge q at the origin. To find the magnitude of the field, we will compute
the force on the hemisphere divided by q. The force on the hemisphere is
Z Z
q qσ
σ dS = dS.
4π0 R2 4π0 R2
By idea 5, we can replace the surface of integration with a flat disk, so | dS| = πR2 . Thus, the
R
force is F = qσ/40 , so the field is

σ
E= .
40
[3] Problem 10. A point charge q is placed a distance a/2 above the center of a square of charge
density σ and side length a. Find the force of the square on the point charge.
Solution. This is a tricky problem, whose solution uses a one-time trick. It’s equivalent to find the
force of the point charge on the square. Set up coordinates so that the square is in the xy plane,
and its center is the origin. Then we have
Z
F = σ E dS
9
where the surface integral is over the square. On the other hand, we know that F is along the ẑ
direction by symmetry, so Z
F = F · ẑ = σ Ez dS.
Now, since dS is parallel to ẑ, this is in fact the same thing as

Z
F = σ E · dS
where the integral is just the electric flux through the square! By symmetry, this flux is q/60 , so
σq
F = .
60
[4] Problem 11 (Griffiths 2.47, PPP 113, MPPP 140). Consider a uniformly charged spherical shell
of radius R and total charge Q.
(a) Find the net electrostatic force that the southern hemisphere exerts on the northern hemi-
sphere.
(b) Generalize part (a) to the case where the sphere is split into two parts by a plane whose
minimum distance to the sphere’s center is h.
(c) Generalize part (a) to the case where the two hemispherical shells have uniform charge density,
opposite orientation, and the same center, but have different total charges q and Q, and
different radii r and R, where r < R.
Hint: see example 9, and use superposition and symmetry when applicable.
Solution. (a) The net force that the northern hemisphere exerts on itself is 0, so it is equivalent
to find the force on the north due to the entire sphere. The surface charge density is σ =
Q/(4πR2 ). By the result of example 9, the outward pressure on the northern hemisphere is
σ 2 /20 . Therefore, the total force is
σ2 σ2 Q2
Z
F = dS = (πR2 ) =
20 N 20 32π0 R2
where N refers to the northern hemisphere, and the surface integral was done as in problem 9.
(b) This is exactly the same as in part (a), except that now the integral over the piece is
Z
dS = π(R2 − h2 )
which gives the result
Q2
F = (σ 2 /20 )π(R2 − h2 ) = (1 − h2 /R2 ).
32π0 R2
(c) This can be solved using an ingenious superposition and symmetry argument.
10
The force we want to compute is shown in (a). Now consider superposing a uniformly negatively
charged sphere with radius just larger than R, as shown in (b). By the shell theorem, this
doesn’t change the force on the hemisphere of radius r. The result of the superposition is (c).
Flipping the charge of one of the hemispheres in (c) flips the force, leading to (d). Finally,
reflecting (d) gives (e).
This has all been preamble to the ingenious step: superpose (a) and (e) to get (f), which
involves the force on a complete sphere of radius r. Using Newton’s third law, 2F can now
be computed by finding the force on the hemisphere. But that is easy because of the shell
theorem, which tells us that F is the net force on the hemisphere shown in (g). Using the
method of problem 9 again, we conclude
q Q Qq
F = (πR2 ) =
4π0 R2 2πR2 8π0 R2
which is independent of r! (Setting r = 0 and r = R recovers the answers to two previous
problems.)
Example 6: IdPhO 2020.1A
A point charge of mass m and charge −q is placed at the center of a cube with side length a,
whose volume has uniform charge density ρ. The point charge is allowed to slide along a
straight line, which has an arbitrary orientation, so that the distance along the line from the
center to one of the cube’s faces is L.
11
Find the frequency of small oscillations.
Solution
The official solution goes as follows: consider displacing the point charge by some small
amount ∆r. The cube of charge can then be decomposed into (1) a slightly smaller cube of
charge centered around the point charge’s new position, and (2) three thin plates of charge
on the faces opposite to the charge’s motion. By symmetry, (1) contributes nothing, and we
know what (2) contributes from the answer to problem 10. The result is a restoring force
proportional to −∆r, whose magnitude has no dependence on the orientation of ∆r, so the
oscillation frequency doesn’t depend on L. Once you know this, you can orient the line any
way you want, so the problem is simple to finish.
Personally, I don’t like this problem because the intended solution requires knowing the
answer to problem 10, which itself is pretty tricky. That is, the difficulty of the problem
depends mostly on whether you’ve seen that tough, but standard problem elsewhere.
However, I’m including it as an example because there’s another way to solve it, which is
more advanced, but quite illustrative.
Since this is a question about small oscillations, it suffices to expand the potential energy to
second order about the center of the cube. The most general possible exprssion is
V (x, y, z) = a + b1 x + b2 y + b3 z + c1 x2 + c2 y 2 + c3 z 2 + c4 xy + c5 yz + c6 xz + O(r3 ).
The constant a doesn’t matter, so we can just ignore it. And since E vanishes at the center,
the linear terms bi are all zero as well. Because the x, y, and z axes are all equivalent by
cubical symmetry (e.g. we can rotate them into each other, while keeping the cube the same),
c = c1 = c2 = c3 , c0 = c4 = c5 = c6 .
Thus, our complicated Taylor series boils all the way down to
V (x, y, z) = c(x2 + y 2 + z 2 ) + c0 (xy + yz + xz) + O(r3 )
without even having to do any work! Finally, notice that the cube is symmetric under
reflections x → −x, y → −y, or z → −z. These reflections keep the c term the same, but flip
the c0 term. Therefore, we must have c0 = 0, so
V (r) = cr2 + O(r3 )
12
which is remarkably simple. The potential near the origin is spherically symmetric,
even though the setup as a whole isn’t! It’s not automatic: it wouldn’t have been this
simple if we had had a slightly more complex shape. This “accidental” spherical symme-
try is a consequence of the combination of cubical symmetry and the simplicity of Taylor series.
Therefore, to finish the problem we only need to find the coefficient c. While there are simpler
ways to do this, I’ll do it in a way that introduces some useful facts. Combining the definition
of V and Gauss’s law, we have
ρ
∇ · (∇V ) = −∇ · E = − .
0
This is a standard and fundamental result in electrostatics, called Poisson’s equation, which
we will see again later. The divergence of a gradient is also called a Laplacian, and written as
∂2V ∂2V ∂2V ρ

∇2 V = + + =− .
∂x2 ∂y 2 ∂z 2 0
Using this, we can easily compute the value of c, giving
ρr2
V (r) = − + O(r3 ).
60
Therefore, for a displacement ∆r in any direction, the restoring force is ρqr/30 in the
opposite direction, which means r
ρq
ω=
30 m
independent of the orientation of the line.
Remark
Accidental symmetry is important in modern physics. For example, protons are stable because
of an accidental symmetry in the Standard Model, which ensures that baryon number is
conserved. That explains why we often expect proton decay to occur in extensions of the
Standard Model, such as grand unified theories, as explained in this nice article.
2 Continuous Charge Distributions

Idea 6
In almost all cases in Olympiad physics, there will be sufficient symmetry to reduce any
multiple integral to a single integral. Remember that when using Gauss’s law, the Gaussian
surface may be freely deformed as long as it doesn’t pass through any charges.
[2] Problem 12 (Purcell 1.15). A point charge q is located at the origin. Compute the electric flux
that passes through a circle a distance ` from q, subtending an angle 2θ as shown below.
13
Solution. Let ` = R cos θ, and deform the disk into a spherical cap with radius R. Then the answer
is then just kq/0 , where k is the ratio of the area of the cap to the total area of the sphere. In
spherical coordinates,
Z θ
1 1 − cos θ
k= 2π sin θ dθ =
4π 0 2
so the answer is
1 − cos θ q
.
2 0
You can also show this using the original flat Gaussian surface, though that takes more work.
[3] Problem 13 (Purcell 1.8). A ring with radius R has uniform positive charge density λ. A particle
with positive charge q and mass m is initially located in the center of the ring and given a tiny kick.
If the particle is constrained to move in the plane of the ring, show that it exhibits simple harmonic
motion and find the frequency.
Solution. Suppose it is moved by r R in the x direction. Set up polar coordinates with θ = 0
being the positive x axis. By the law of cosines, we have
Z π
1 q(λRdθ)
U (r) = 2 √
0 4π0 R + r2 − 2Rr cos θ
2
Z π
qλ dθ
= p .
2π0 0 1 + (r2 /R2 ) − 2(r/R) cos θ
Next, we can expand the square root using a Taylor series. If we expand to first order in r/R, then
the result will be proportional to the integral of cos θ, which vanishes. Thus, to get the leading
contribution we must expand to second order, giving
Z π" 2 #
1 r2

qλ 3 2r
U (r) = 1− + − cos θ dθ
2π0 0 2 R2 8 R
Z π 2
qλ r
= (3 cos2 θ − 1) dθ + const
2π0 0 2R2
qλr2
= + const.
80 R2
Therefore, the effective spring constant is k = qλ/40 R2 , so
r
qλ
ω= .
4m0 R2
[3] Problem 14 (Purcell 1.12). Consider the setup of problem 9. If the hemisphere is centered at the
origin and lies entirely above the xy plane, find the electric field at an arbitrary point on the z-axis.
(This is a bit complicated, and is representative of the most difficult kinds of integrals you might
have to set up in an Olympiad. For a useful table of integrals, see Appendix K of Purcell.)
14
Solution. Set up spherical coordinates with the hemisphere being the equation of r = R and
θ ∈ [0, π/2]. Suppose our location is (0, 0, z). The hemisphere has surface charge σ. We see that
the field points in the z-direction by symmetry, so we’ll only worry about that piece. The ring at
angle θ with width dθ provides fields
√ at an angle, and some geometry shows that we have to correct
R cos θ−z
by a factor of r where r ≡ R2 + r2 − 2Rz cos θ. We then have
σ(2πR2 sin θ dθ) R cos θ − z

dEz = − · ,
4π0 r2 r
so
π/2
σR2 (R cos θ − z) sin θ dθ
Z
E(z) = − .
20 0 (R2 + r2 − 2Rz cos θ)3/2
Consulting Appendix K tells us that
!
σR2 R R−z
E(z) = √ −p .
20 z 2 2
R +z 2 (R − z)2
Taking some care with the square root, we conclude


σR 2  √ 12 2 − 1 z<R
1+z /R
E(z) = × .
20 z 2  √ 1 +1 z>R
2 2
01^
1+z /R
Idea 7: Electric Dipoles
The dipole moment of two charges q and −q separated by d is p = qd. More generally, the
dipole moment of a charge configuration is defined as
Z
p = ρ(r)r dr.
For an overall neutral charge configuration, the leading contribution to its electric potential
far away is the dipole potential,
p cos θ
φ(r, θ) =
4π0 r2
where θ is the angle of r to p.
Remark
Here’s a trick to remember the dipole potential. Let φ0 (r) = k/r be the potential for a unit
charge at the origin. An ideal point dipole of dipole moment p consists of charges ±p/d
separated by d, in the limit d → 0. So the potential is
φ0 (r) − φ0 (r + d)
p lim .
d→0 d
But this is precisely the (negative) derivative, so you can get the dipole potential by differen-
15
tiating the ordinary potential! Indeed, for a dipole aligned along the ẑ axis,
d kp kp dr kp z kp cos θ
− = 2 = 2 =
dz r r dz r r r2
which matches the above result. You can use the same trick for quadrupoles and higher
multipoles, which we’ll see in E8.
[3] Problem 16. In this problem we’ll derive essential results about dipoles, which will be used later.
(a) Using the binomial theorem, derive the dipole potential given above, for a dipole made of a
pair of point charges ±q separated by distance d, oriented along the z-axis.
(b) Differentiate this result to find the dipole field,
p
E(r) = (2 cos θ r̂ + sin θ θ̂)
4π0 r3
where the expression above is in spherical coordinates. (Hint: feel free to use the expression
for the gradient in spherical coordinates.)
(c) Show that this may also be written as
1
E(r) = (3(p · r̂)r̂ − p).
4π0 r3
You don’t need to memorize these expressions, but it’s useful to remember what a dipole field
looks like, the fact that its magnitude is roughly p/4π0 r3 , and the fact that the numeric
prefactor is 2 along the dipole’s axis and 1 perpendicular to it.
Solution. (a) Let the charges be at (0, 0, 0) and (0, 0, d). Then
!
q 1 qd cos θ
V (r, θ) = −1 + p ≈ .
4π0 r 1 − 2(d/r) cos θ + (d/r)2 4π0 r2
(b) We use the definition E = −∇V , along with the gradient in spherical coordinates. Then
∂V p
Er = − = · 2 cos θ
∂r 4π0 r3
and
1 ∂V p
Eθ = − = · sin θ,
r ∂θ 4π0 r3
as desired.
(c) We see that p · r̂ = p cos θ and p = pẑ = p(r̂ cos θ − θ̂ sin θ). Thus,
3(p · r̂) r̂ − p = 3p cos θ r̂ − p(r̂ cos θ − θ̂ sin θ) = p(2 cos θ r̂ + sin θ θ̂),
01m
as desired.
01m
[3] Problem 18. USAPhO 2009, problem B2. This essential problem introduces useful facts
about dipole-dipole interactions.
16
Idea 8
The potential energy of a set of point charges is
1 X qi qj 1X
U= = qi V (ri ).
4π0 |ri − rj | 2
i6=j i
We sum over i 6= j to avoid computing the energy of a single point charge due to its interaction
with itself, which would be infinite. For a continuous distribution of charge, we don’t have
this problem, and instead find
Z Z
1 0
U= ρ(r)V (r) dr = |E(r)|2 dr.
2 2
Unlike the other quantities we’ve considered, energy doesn’t obey the superposition principle.
[3] Problem 19. In this problem we’ll apply the above results to balls of charge.
(a) Compute the potential energy of a uniformly charged ball of total charge Q and radius R.
(b) Show that the potential energy of two point charges of charge Q/2 separated by radius R is
lower than the result of part (a).
(c) Hence it appears that it is energetically favorable to compress a ball of charge into two point
charges. Is this correct?
Solution. (a) We can find the potential by building up the ball by placing charges from infinity.
Consider a shell of charge at radius r, and let the charge density by ρ = Q/( 43 πR3 ). The
energy needed to put the shell there is dU = kQenc dQ/r, where Qenc = 43 ρπr3 is the charge
inside and dQ = 4ρπr2 dr is the charge in the shell added to the sphere. Then the energy
needed to build the ball, which is the potential energy of the ball, is
Z R
r3 3kQ2 R 4 3kQ2 3Q2
Z
2 3
Ua = kQ 3 (3Qr dr/R )/r = r d r = = .
0 R R6 0 5R 20π0 R
(b) From U = kq1 q2 /r, we find that for two point charges the potential energy is
kQ2 Q2
Ub = =
4R 16π0 R
which is less than Ua .
(c) It’s wrong because in part (b), the energy needed to create the point charges, by squeezing
the two halves of the ball down, is not included. Plugging in a radius of zero into part (a), we
see that this energy is actually infinite. (Of course, in reality it doesn’t take infinite energy
to produce electrons, which are point charges. Classical electrodynamics breaks down when
describing such a process, which can only be properly understood within relativistic quantum
field theory.)
[3] Problem 20. An insulating circular disc of radius R has uniform surface charge density σ.
(a) Find the electric potential on the rim of the disc.
17
(b) Find the total electric potential energy stored in the disc.
Solution. (a) Place the origin at a point on the rim and use polar coordinates. Because the
polar equation of a circle is r = 2R cos θ, we have
Z π/2 Z 2R cos θ Z π/2
σ σR σR
V = dθ dr = cos θ dθ = .
−π/2 0 4π0 −π/2 2π0 π0
(b) Consider building up the ring outward in radius. When we add charges to bring the radius
from r to r + dr, we do work
σr 2σ 2 r2
dW = V dq = (2πrσ dr) = dr
π0 0
which means
R
2σ 2 r2 2σ 2 R3
Z
W = dr = .
0 0 30
[3] Problem 21. Consider a uniformly charged ball of total charge Q and radius R. Decompose this
ball into two parts, A and B, where B is a ball of radius R/2 whose center is a distance R/2 of the
ball’s center, and A is everything else. Find the potential energy due to the interaction of A and B,
i.e. the work necessary to move B to infinity.
Solution. By the shell theorem, everywhere within B, the electric field due to A is constant, with
magnitude
k(Q/8) kQ
E= 2
= .
(R/2) 2R2
This is just the same reasoning as in problem 1. Now, if we take B and squish it down to its center,
this takes no net work against A’s electric field. If we then move that point a distance R/2 out to
the edge of A, then this takes work
QR kQ2
W =E = .
8 2 32R
At this point, we still need to move B out to infinity. The field that B experiences is the superposition
of a uniform ball (the original one) plus a uniform ball of negative density where B was. So again
by the shell theorem, the potential energy of B is

Q kQ k(Q/8)
U= − + .
8 R R/2
The total work needed to move B out to infinity is the work we already did, plus what we need to
move it out of this potential well,
kQ2
W −U = .
8R
Note that we haven’t counted the work harvested by squishing B, because the question is asking
about the interaction between A and B, so this contribution is irrelevant.
[2] Problem 22 (PPP 149). A distant planet is at a very high electric potential compared with Earth,
say 106 V higher. A metal space ship is sent from Earth for the purpose of making a landing on the
planet. Is the mission dangerous? What happens when the astronauts open the door on the space
ship and step onto the surface of the planet?
18
Solution. As the space ship approaches the planet, its potential gradually increases from that of
the Earth, to that of the distant planet. Meanwhile, all the astronauts inside are doing just fine
since the ship acts like a Faraday cage. Once the ship lands, it’s already at the same potential as
the planet, and when the astronauts step out, nothing happens. In other words, it’s electric field
that’s dangerous, not potential, and the electric fields in this problem are always small.
Another way to see that there’s no danger is to replace electric fields with gravitational fields,
and thus electric potential with gravitational potential. An elevator in a skyscraper takes you from
a low to a very high gravitational potential. But nothing violent happens when you get off!
Example 7
Since Newton’s law of gravity is so similar to Coulomb’s law, the results we’ve seen so far
should have analogues in Newtonian gravity. What are they? For example, what’s the
gravitational Gauss’s law?
Solution
The fundamental results to compare are
Gm1 m2 q1 q2
F =− , F =
r2 4π0 r2
where the minus sign indicates that the gravitational force is attractive, while the electrostatic
force between like charges is repulsive. Then we can transform a question involving (only
positive) electric charges to one involving masses if we map
1
q → m, → −G, E → g.
4π0
Thus, while electrostatics is described by
I
ρ Q
∇ × E = 0, ∇·E= , E · dS = ,
0 0
the gravitational field is described by
I
∇ × g = 0, ∇ · g = −4πGρm , g · dS = −4πGM
where ρm is the mass density. Similarly, the potential energy can be written in two ways,
Z Z
1 1
U= ρm (r)φ(r) dr = − |g(r)|2 dr
2 8πG
where φ(x) is the gravitational potential. This result was first written down by Maxwell.
Remark
Here’s a philosophical question: is potential energy “real”? It’s not as obvious as you might
think! In the 1700s, there was a lively debate over whether the ideas of kinetic energy and
momentum, which at the time were given various other names, were worthwhile. Which one
19
of the two was the true measure of motion? In our modern language, proponents of energy
pointed out that the momentum always vanished in the center of mass frame, which made it
trivial, while supporters of momentum replied that kinetic energy was clearly not conserved
in even the simplest of cases, like inelastic collisions.
In the 1800s, the theory of thermodynamics was developed, allowing the energy seemingly
lost in inelastic collisions to be accounted for as internal energy. But there still remained
the problem that kinetic energy was lost in simple situations, such as when balls are thrown
upward. By the mid-1800s, the modern language that “kinetic energy is converted to potential
energy” was finally standardized, but it was still common to read in textbooks that potential
energy was fake, a mathematical trick used to patch up energy conservation. After all,
potential energy has some suspicious qualities. If a ball has lots of potential energy, you can’t
see or feel it, or even know it’s there by considering the ball alone. It doesn’t seem to be
located anywhere in space, and its amount is arbitrary, as a constant can always be added.
In the late 1800s, a revolution on physics answered some of these questions. Maxwell and his
successors recast electromagnetism as a theory of fields, and showed that the dynamics of
charges and currents were best understood by allowing the fields themselves to carry energy
and momentum. We’ll cover this in detail in E7, but for now, it implies that electrostatic
potential energy is fundamentally stored in the field, with a density of 0 E 2 /2. This implies
that its location and total amount are directly measurable.
Maxwell believed that the dynamics of fields emerged from the microscopic motions
and elastic deformations of an all-pervading ether, in the same way that, say, a fluid’s
velocity field emerges from the average motion of fluid molecules. This makes it manifestly
positive, so he was disturbed to find that the energy density of a gravitational field is negative!
A few decades later, the arrival of special relativity answered some questions and reopened
others. On one hand, it demolished Maxwell’s vision of the ether. On the other hand, it finally
answered the question of whether all kinds of potential energy are “real”, and it got rid of
the freedom to add arbitrary constants. That’s because in special relativity, the total energy
of a system at rest is related to its mass by E = mc2 , and the mass is directly measurable.
This finally puts thermal energy, elastic potential energy, and field energy on an equal footing.
Here’s the most modern view of energy conservation. All particles and their interactions are
fundamentally described by relativistic quantum fields. A famous result called Noether’s
theorem implies that whenever such a theory is time-translationally symmetric, there
is a conserved quantity which we call the energy. (The distinction between kinetic and
potential energy becomes irrelevant; it’s all just energy.) The density of energy in space
can be computed from the state of the fields, but it doesn’t need to be explained, as
Maxwell imagined, by the internal motion of whatever the fields are made of. The fields are
fundamental: they aren’t made of anything; instead, they make up everything!
What happens when we throw gravity into the mix? As we’ll discuss further in R3, it turns
out that at nonrelativistic velocities, the dynamics of gravitating particles can be described
by “gravitoelectromagnetism”, a theory closely analogous to electromagnetism, where moving
20
masses also source “gravitomagnetic” fields Bg , which result in mv × Bg forces. But the
situation gets much more subtle when we upgrade to full general relativity. Here, the notion
of a gravitational field disappears completely, and is replaced by the curvature of spacetime,
making it hard to define an energy density for it at all. For an accessible overview of the
debate, see this paper. Ultimately, though, it doesn’t matter that much, since it doesn’t
impair our ability to use either Newtonian gravity or general relativity.
Example 8
For an infinite line of linear charge density λ, find the potential V (r) by dimensional analysis.
Solution
This example illustrates a famous subtlety of dimensional analysis. The only quantities in
the problem with dimensions are λ, 0 , and r. To get the electrical units to balance, we have
λ
V (r) = f (r)
2π0
where f (r) is a dimensionless function. But there are no nontrivial dimensionless functions
of a dimensionful quantity r. The only possibilities are that f (r) is a dimensionless constant,
or that f (r) is infinite. In the first case, the electric field would vanish, which can’t be right.
In the second case, it is unclear how to calculate the electric field at all.
In fact, the electric potential is infinite, if you insist on the usual convention of setting
V (∞) = 0. In that case, we have
Z ∞
λ dr
V (r) = =∞
r 2π0 r
independent of r. But this is useless; to get a finite result we can actually work with, we
need to subtract off an infinite constant from the potential. Equivalently, we need to set the
potential to be zero at some finite distance r = r0 . This process is known as renormalization,
and it is extremely important in modern physics. After renormalization, we have
λ r0
V (r) = log
2π0 r
which is perfectly consistent with dimensional analysis.
Notice that in the process of renormalization, a new dimensionful quantity r0 appeared out
of nowhere. This phenomenon is known as dimensional transmutation. Of course, physical
predictions don’t depend on this new scale (e.g. the electric field is independent of r0 ), but
you can’t write down quantities like the potential without it.
21
3 Conductors
Idea 9
In electrostatic conditions, E = 0 inside a conductor, which implies the conductor has constant
electric potential V . This further implies that E is always perpendicular to a conductor’s
surface. By Gauss’s law, the conductor has ρ = 0 everywhere inside, so all charge resides on
the surface.
Example 9
Consider a point on the surface of a conductor with surface charge density σ. Show that the
outward pressure on the charges at this point is σ 2 /20 .
Solution
Gauss’s law tells us that the difference of the electric fields right inside and outside the
conductor at this point is
σ
Eout − Ein =
0
by drawing a pillbox-shaped Gaussian surface. But we also know that Ein = 0 since we’re
dealing with a conductor, so Eout = σ/0 .
Let’s think about how this electric field is made. If there were no charges around except for
the ones at this surface, then the interior and exterior fields would have been ±σ/20 . This
means that all of the other charges, that lie elsewhere on the surface of the conductor, must
provide a field σ/20 here, so that Ein cancels out.
The pressure on the charges at this point on the surface is equal to the product of the surface
charge density with the field due to the rest of the charges, since the charges at this point
can’t exert an overall force on themselves, so
σ2

σ
P =σ =
20 20
2 /2.
as required. Equivalently, we can conclude that P = 0 Eout
Example 10
Is the charge density at the surface of a charged conductor usually greater at regions of higher
or lower curvature?
Solution
Of course, we can’t answer this question directly, because it is essentially impossible to
find the charge distribution of an irregularly shaped conductor. However, we can get some
insight by considering the limiting case of a conductor made of two spheres of radii R1 and
22
R2 , connected by a very long rod.
For the potential to be the same at both spheres, we must have Q1 /R1 = Q2 /R2 , so the
charge is proportional to the radius, and the charge density is inversely proportional to the
radius. Thus, there’s generally higher charge density at sharper points of the conductor.
[1] Problem 23. Show that any surface of charge density σ with electric fields E1 and E2 immediately
on its two sides experiences a force σ(E1 + E2 )/2 per unit area. (This is a generalization of the
example above, where one side was inside a conductor.)
Solution. Let E be the field due to all the other charges. Again, we have E1 = E + σ/20 n̂ and
E2 = E − σ/20 n̂. Thus, E = 12 (E1 + E2 ), and the force per area is σE.
[2] Problem 24. Is it possible for a connected, completely isolated conductor with a positive charge
to have a negative surface charge density at any point? If not, prove it. If so, sketch an example.
Solution. This can’t happen. Note that the surface of the conductor has a constant, positive
potential. Now suppose there was a region with negative charge on the conductor, and consider
a field line that ends on such a charge. It can’t have come from infinity, because the potential at
infinity is lower than that of the conductor. And it can’t have come from elsewhere on the conductor,
because the conductor is an equipotential. This yields a contradiction.
Idea 10: Existence and Uniqueness
In a system of conductors where the total charge or potential of each conductor is specified,
there exists a unique charge configuration that satisfies those boundary conditions.
This is very useful because in many cases, it is difficult to directly derive the charge distribu-
tions or fields. Instead, sometimes one can simply insightfully guess an answer; then it must
be the correct answer by uniqueness. For further discussion, see section 2.5 of Griffiths.
Example 11
Consider a conductor with nonzero net charge, and an empty cavity inside. Show that the
electric field is zero in the cavity.
Solution
Let’s consider a second conductor with the same net charge and the same shape, but
without the cavity. By the existence and uniqueness theorem, we know there exists
some charge configuration on the second conductor’s surface which satisfies the boundary
conditions, namely that the electric field vanishes everywhere inside the conductor. In par-
ticular, that means the field is zero where the cavity of the original conductor would have been.
Now consider the original conductor again. If we give this conductor precisely the same
surface charge distribution, then this will again solve the boundary conditions, and it’ll have
23
no field in the cavity. But by the existence and uniqueness theorem, the charge distribution
is unique, so this is the only possible answer: the field must be zero in the cavity.
If this is your first time seeing this, it can sound like a fast-talking swindle (which is why I
made it an example rather than a problem!). It looks like we used no effort and got a strong
conclusion out. Of course, that’s because all the work is done by the uniqueness theorem.
[1] Problem 25. Consider a spherical conducting shell with an arbitrary charge distribution inside,
with net charge Q. Find the electric field outside the shell.
Solution. The shell is an equipotential. The field of a point charge Q at the center of the shell
hence satisfies the boundary conditions. By the uniqueness theorem, this is the only solution.
[2] Problem 26 (Purcell 3.33). The shaded regions represent two neutral conducting spherical shells.
Carefully sketch the electric field. What changes if the two shells are connected by a wire?
Solution. The results are shown below.
Whether or not there are any field lines coming from the charge outside the shell depends on how
close that charge is to the shell. (The entire field configuration in this problem can be found exactly
using the method of images, as shown in E2.) In the case that the two spheres are connected with
01W
a wire, the field in between the two spheres disappears, but nothing else changes.
24
[4] Problem 28 (MPPP 150). A solid metal sphere of radius R is divided into two parts by a planar
cut, so that the surface area of the curved part of the smaller piece is πR2 . The cut surfaces are
coated with a negligibly thin insulating layer, and the two parts are put together again, so that the
original shape of the sphere is restored. Initially the sphere is electrically neutral.
The smaller part of the sphere is now given a small positive electric charge Q, while the larger
part of the sphere remains neutral. Find the charge distribution throughout the sphere, and the
electrostatic interaction force between the two pieces of the sphere.
Solution. We know that distributing charge uniformly on the outer surface of the entire sphere will
give a valid configuration, in the sense that the field is everywhere perpendicular to the conductors.
Similarly, distributing equal and opposite charges uniformly on the two flat faces will give a valid
configuration, since it acts like a parallel plate capacitor, making the field vanish everywhere outside.
Neither of these solutions have the right total charge on each piece, but we can fix this by
superposing the two. By solving a system of two equations, we find the charge distribution is
• total charge Q distributed uniformly on the sphere,
• charges ±(3/4)Q distributed uniformly on the flat surfaces.

The two flat faces attract each other and the two curved faces repel each other; there are no other
forces by the shell theorem. The pressure on the flat faces is σ 2 /20 . With a little trigonometry, we
find the area of the flat faces is (3/4)πR2 , giving a force
3Q2
F1 = .
8π0 R2
As for the repulsive force, using the result of problem 11 we get
3Q2 45 Q2
F2 = − , Ftot = F1 + F2 = .
128π0 R2 128 π0 R2
[3] Problem 29. In this problem we’ll work through a heuristic proof of a version of the uniqueness
theorem. In particular, we will show that for a system of conductors in empty space, specifying the
total charge on each conductor alone specifies the entire surface charge distribution.
(a) Suppose for the sake of contradiction that two different charge distributions can exist, and
consider their difference, which has zero total charge on each conductor. Argue that at least
one conductor must have electric field lines both originating from and terminating on it.
(b) Show that at least one of these field lines must originate from or terminate on another one of
the conductors.
(c) By generalizing this reasoning, prove the desired result. (Hint: consider the conductors with
the highest and lowest potentials.)
25
Solution. (a) Since the overall charge distributions are difference, at least one conductor C must
have different charge distributions in the two cases. So when we consider the difference, C
has areas of both positive and negative surface charge. Field lines come out of the latter, and
go into the former.
(b) The field lines can’t go back to C, because by following the field line, one would prove that C
has a higher potential than itself, which is impossible. They also can’t all go off to infinity,
because we can consider “infinity” to just be a big, far away neutral conductor at zero potential.
If lines both came from infinity to C and from C to infinity, then C would again have a higher
potential than itself, which is impossible. So some field line must go between C and another
conductor C 0 .
(c) Consider the conductor with the highest potential φmax which has nontrivial surface charges.
Field lines have to go into this conductor, but they can’t come from anywhere except from
infinity. Since infinity is at zero potential, we have φmax ≤ 0.
Now consider the conductor with the lowest potential φmin which has nontrivial surface charges.
Field lines have to leave this conductor, but they can’t go anywhere except for infinity. Since
infinity is at zero potential, φmin ≥ 0. This forces φmin = φmax = 0, so everything must be at
zero potential, which means there aren’t any electric field lines at all.
A wire loop of height h and resistance R has one end placed inside a parallel plate capacitor
with electric field E, as shown.
The other end of the loop is far away, where the field is negligible. Find the emf in the loop.
Solution
Of course, this is a trick question; if the answer were nonzero, the current would run forever,
yielding a perpetual motion machine. Electrostatic fields always produce zero total emf along
any loop. The σh/0 voltage drop inside the capacitor is canceled out by the voltage drop
due to the fringe fields, which are small, but accumulate over a long distance. The point of
this example is that, while we can ignore fringe fields for some calculations, they are often
essential to get a consistent overall picture. We’ll consider the subtleties of fringe fields in
much more detail in E2.
[2] Problem 30 (Purcell 3.2). Spheres A and B are connected by a wire; the total charge is zero. Two
oppositely charged spheres C and D are brought nearby, as shown.
26
The spheres C and D induce charges of opposite sign on A and B. Now suppose C and D are
connected by a wire. Then the charge distribution should not change, because the charges on C
and D are being held in place by the attraction of the opposite charge density. Is this correct?
Solution. This isn’t correct. To see this rigorously, we can use the uniqueness theorem. After
connecting the wires, we have two conductors (A/B, and C/D), each with zero net charge. One
possible solution is to have zero charge everywhere. By uniqueness, this is the only possible solution,
so anything else cannot have been in equilibrium.
That is rigorous, but it might not be intuitive; after all, it sure looks like the charges on C are
stuck where they are. However, though the charges on C are attracted towards A, they also strongly
repel each other. It’s this repulsion that causes charge on C to start flowing to D when the wire is
connected.
27
Electromagnetism II: Electricity

Chapters 3 and 4 of Purcell cover the material presented here, as does chapter 6 of Wang and
Ricardo, volume 2. Image charges are covered in more detail in section 3.2 of Griffiths. For an array
of interesting physical examples, see chapters II-6 through II-9 of the Feynman lectures. There is a
total of 80 points.
1 The Method of Images

Idea 1
The method of images can be used in some highly symmetric situations to compute the
electric field in the vicinity of a conductor. Specifically, consider any configuration of static
charges and take any equipotential surface containing some of the charges. Then the resulting
field configuration outside that surface is the field configuration we would have if that surface
bounded a conductor. This is simply because it has constant potential on the conductor
surface, so it must be the right answer by the uniqueness theorem.
[4] Problem 1. The simplest application of the method of images is the case of a charge q a distance a
from an infinite grounded conducting plane. This problem explores some of its subtleties, assuming
you’ve already read the basic treatment in section 3.4 of Purcell.
(a) Find the force on the charge.
(b) Find the work needed to move the charge out to infinity. (Answer: q 2 /16π0 a.)
(c) Find the total potential energy of the charges on the conducting plane, i.e. the potential energy
associated only with their interaction with each other. (Answer: q 2 /16π0 a.)
(d) Now suppose there is another parallel grounded conducting plane on the other end of the
charge, a distance b away. How many image charges are needed now? Draw some of them.
(e) A conducting plane forces the electric field to be perpendicular to it. Suppose we modified the
plane so that the electric field was instead always parallel to it. Find the force on the charge.
Solution. (a) The field above the plate from the screening charges on the plate can be mimicked
by placing an image charge −q below the plane. This follows because in both cases, the plane
of the plate is an equipotential of potential 0. Thus, the force on the charge is q 2 /(16π0 a2 )
pointing towards the plane.
(b) We just integrate the expression from before to get q 2 /(16π0 a).
It’s tempting to just write down q 2 /(8π0 a) because this is the energy of two charges ±q
separated by 2a, but we must remember that the image charge isn’t a real charge. It instead
describes the effects of all the screening charges on the plane. You might think it takes work
to move the screening charges too, but they can move “for free”, since the electric field is
perpendicular to the plane throughout this entire process.
1
(c) Suppose we freeze the plane’s screening charges in place, then move the point charge out to
infinity. In this case the image charge is stationary, so the work needed is q 2 /(8π0 a).
Let’s think about what this means. There are two components to the initial potential energy:
the energy U1 of the point charge interacting with the screening charges, and the energy U2
of the screening charges interacting with each other.
In part (b), we showed that it takes work q 2 /(16π0 a) to move the point charge out, after
which point all charges in the problem are widely separated. So
q2
U1 + U2 = − .
16π0 a
By the other argument we just made, if we freeze the screening charges and move the point
charge away, we are left with just the screening charges. So
q2
U1 = − .
8π0 a
The quantity we’re looking for is U2 , which is thus
q2
U2 = .
16π0 a
Of course this energy is positive, because the screening charges repel each other.
(At first glance this argument may seem to contradict the statement made in part (b), which
states that the screening charges cost zero energy to move, as we move the point charge.
However, that statement is only true if the screening charges are always allowed to move,
so that they preserve the boundary condition Ek = 0. But above we instead considered
an artificial situation where the screening charges were frozen, and this reasoning no longer
holds.)
(d) We actually need infinitely many.
This is the same reason you see infinitely many images of yourself when between two mirrors.
(e) This boundary condition can be met by placing a charge +q at z = −d. Then, the force on
the charge would be kq 2 /(4d2 ) pointing away from the plane.
[4] Problem 2. In this problem you’ll develop the method of images for spheres.
2
(a) A point charge −q is located at x = a and a point charge Q is located at x = A. Show that
the locus of points with φ = 0 is a circle in the xy plane, and hence a spherical shell in space.
(b) What relations must hold among q, Q, a, and A so the center of the sphere is at the origin?
In this case, what is the sphere’s radius?
(c) Now suppose a point charge Q is a distance b from the center of a spherical grounded conducting
shell of radius r. Find the force on the charge, considering both the cases b < r and b > r.
(d) The case b < r is a bit confusing. On one hand, argue that the total charge is Q plus the image
charge, and hence nonzero. On the other hand, argue that the total charge must be zero, by
considering an appropriate Gaussian surface. One of these arguments is wrong – which one?
As an aside, the fundamental reason the method of images works for spheres is that electromagnetism
has conformal symmetry, a symmetry under any local rescaling of space which preserves angles.
(One example of a conformal transformation is inversion in Euclidean geometry.) The setup here is
related to the conducting plane by such a transformation.
Solution. If you happen to know Euclidean geometry, you can very quickly solve parts (a) and (b)
using properties of Apollonian circles. However, for everyone else, we’ll present a straightforward
solution using coordinates.
(a) The condition for the potential to vanish is

q Q
p =p .
(x − a)2 + y2 (x − A)2 + y 2
Squaring both sides and clearing denominators, we find
(q 2 − Q2 )(x2 + y 2 ) + (q 2 A2 − Q2 a2 ) + 2x(aQ2 − Aq 2 ) = 0.
This has the form of a conic section, and since the coefficients of x2 and y 2 are equal, it’s a
circle. (Strictly speaking, it could also be the empty set, since, for example, x2 + y 2 = −1 has
no solutions. But we know there have to be places where φ = 0, because it’s positive near the
positive charge and negative near the negative charge, so it must cross zero by continuity.)
(b) The center of the sphere is at the origin if the coefficient of x vanishes,
aQ2 = Aq 2 .
Note that this forces a and A to have the same sign. Plugging this in and simplifying, we find
x2 + y 2 = Aa
√
from which we conclude the radius is r = Aa. If you know some Euclidean geometry, at
this point you can also recognize that the two point charges are inversions of each other with
respect to the sphere.
(c) We can treat both cases at once. The 0 2

0
p image charge is a distance b = r /b from the center of
0
the shell, and its charge is q = −q b /b = −qr/b. Therefore, the force on the charge is
qq 0 q 2 rb
F = 0 2
= .
4π0 (b − b ) 4π0 (b2 − r2 )2
It always attracts the point charge towards the nearest point on the surface of the sphere.
3
(d) The second argument is right. Since the conductor shields the details of the charges inside,
the field outside the sphere must be spherically symmetric. But we also know the sphere is
at zero potential, so the field outside must be exactly zero, so by using a spherical Gaussian
surface, the total charge is zero.
On the other hand, the first argument seems to work for all the other examples we’ve seen:
whenever we have an image charge Q0 , it seems to represent the effects of a total charge Q0 on
the conducting surface. But in this case, the method of images doesn’t determine the charge
distribution on the surface, because adding a constant charge density to the sphere’s surface
creates absolutely no field inside. So in reality, this constant density takes whatever value is
necessary to make the sphere overall neutral, in accordance with the second argument.
[2] Problem 3 (Purcell 3.50). A point charge q is located a distance b > r from the center of a
nongrounded conducting spherical shell of radius r, which also has charge q. When b is close to r,
the charge is attracted to the shell because it induces negative charge; when b is large the charge is
clearly repelled. Find the value of b so that the point charge is in equilibrium. (Hint: you should
have to solve a difficult polynomial equation. You can either use a computer or calculator, or use
the fact that it contains a factor of 1 − x − x2 .)
Solution. The image charge has value −q 0 = −qr/b, and is at position r2 /b. As of now, the
spherical shell has total charge −q 0 (by Gauss’s law), so to compensate, we add a charge of value
q + q 0 at the center. Thus, to balance forces we have
q0 q + q0
= .
(b − r2 /b)2 b2
Defining x = r/b, this simplifies to
x = (1 + x)3 (1 − x)2
which is a rather intimidating quintic equation. At this point you can pull out your calculator, but
amazingly, it factors as
(1 − x − x2 )(1 + x − x3 ) = 0.
√
The only root with 0 < x < 1 is from the quadratic, x = ( 5 − 1)/2, giving
√
1+ 5
b= r.
2
[2] Problem 4. Consider a spherical conductor of radius R placed in a uniform external field E0 . By
using the method of images, argue that the field created by the sphere is the same as the field of
an infinitesimal dipole at the center of the sphere, and find its dipole moment. (Hint: suppose the
external field is sourced by two point charges that are very far away.)
Solution. Suppose E0 = E0 ẑ. Let the conductor be centered at the origin. Now, consider a point
charge −Q at z = −L, and Q at z = L for L R. This clearly generates a uniform electric
field near the origin (for r L) with magnitude 2 4πQ0 L2 . So we simply choose Q = 2π0 L2 E0 .
Now, the image charges are very close to the origin and form a dipole with moment 2Q0 L0 =
2(QR/L)(R2 /L) = 4π0 R3 E0 .
[3] Problem 5. Two grounded conducting half-planes intersect, so that in cylindrical coordinates, the
equations describing the planes are θ = 0 and θ = 2π/3.
4
(a) A charge q is placed a distance r from the origin at θ = π/3. Can the method of images be
used to find the force on the charge? If so, do it; if not, explain why not.
(b) Do the same for θ = π/2.
(c) In general, given that the second plane is at θ = θp and the charge is at θ = θq , when does
the method of images work?
Solution. (a) It can’t be done. When we reflect about the θ = 0 half plane, we get a point that
is collinear with θ = 2π/3. Thus, the new image charge cannot be further reflected in the
plane θ = 2π/3 to get the whole set of image charges.
(b) It can’t be done.
Here, A is the original charge, and B, C, D, E, F are image charges with signs alternating (so
qB = −q, qC = q, and so on). We see that it’s impossible to satisfy all the image charge
constraints unless we include the image charge F . But F is in the same region as A. That isn’t
allowed for image charges: the point of image charges is to provide an easy way of calculating
the effects of screening charges on conducting surfaces on a given set of real charges, so it’s
not legal to introduce new real charges in the process.
(c) If we reflect the charge in the second plane, we get
θ → 2θp − θ.
If we reflect in the first plane again, we get
2θp − θ = θ − 2θp .
That is, in general the composition of two reflections is just a rotation by −2θp . In order to
avoid an infinite number of image charges, θp must be a rational multiple of 2π. If θp is of the
form π/n for integer n, there is no problem for any θq . But if θp is not of this form, then after
a sufficient number of these operations we will arrive at an image charge inside the physical
region, which is unacceptable. So the procedure works if and only if θp = π/n.
[5] Problem 6 (Purcell 3.45). [A] Consider a point charge q located between two parallel infinite
grounded conducting planes. The planes are a distance ` apart, and the point charge is a distance
b from the left plane. The goal of this problem is to find the total charge induced on each plane.
5
(a) Argue that the total charge on each plane would not change if we replaced the point charge q
with two point charges q/2, both a distance b from the left plane. By iterating this process,
convert the point charge into a uniformly charged plane, and use this to get the answer.
(b) Alternatively, using image charges, show that the electric field on the inside surface of the
left plane, perpendicular to the plane, at a point a distance r from the axis containing all the
image charges, satisfies
∞
X 2q(2n` + b)
4π0 E⊥ = .
n=−∞
((2n` + b)2 + r2 )3/2
(c) Since σ = −0 E⊥ , we can integrate both sides to find the total charge on the left plane.
However, the integral of each term by itself is simply q, so the series doesn’t converge. To get
the result, do the following steps in this specific order: group the terms ±n together, then
integrate them only out to a distance R b, then sum over the values of |n|, then take the
limit R → ∞. You should get a finite result that matches that of part (a). As you’ll probably
see in the process, if you do the steps in any other order, you’ll get a nonsensical answer.
Those concerned with mathematical rigor might be bothered by the many choices made in part (c).
You might ask, couldn’t we have gotten a different result by changing how we did the computation?
In fact, by the Riemann rearrangement theorem, we could have gotten almost any result. But the
way we did it is the physically correct way. It roughly sums the terms “in to out”, which respects
the fact that real plates are finite. Closely related ideas are used to “cancel infinities” in quantum
field theory, in a process known as renormalization.
Solution. (a) Consider the induced charge distribution for one point charge q. By the superpo-
sition principle, the boundary conditions in this case are also satisfied if we take half that
charge distribution, and center it about each of the charges q/2. By uniqueness, this is the
solution.
Extrapolating to infinitely many charges, we get a uniform plane of charge. For the two plates
to be at the same voltage, the electric fields above and below the plane have the ratio b/(` − b).
Then the charges have the ratio b/(` − b) and sum to −q, so the charge on the near plate is
−q(` − b)/`, while the charge on the far plate is −qb/`.
(b) We do this by summing over image charges. As we can see from the figure in the solution to
problem 1, there are infinitely many image charges. The n = 0 term in the sum corresponds to
the image charge and real charge closest to the left plane. The n = 1 term corresponds to the
two image charges a distance 2` + b from the left plane, while the n = −1 term corresponds
to the image charges a distance 2` − b from the left plane, and so on.
(c) Applying Gauss’s law, σ = 0 E⊥ , and grouping the terms ±n together, we have a total charge
on the left plane of
Z ∞
Q= (−0 E⊥ ) · 2πr dr
0
Z ∞" ∞ #
br X (2n` − b)r (2n` + b)r
=q − 2 + − dr.
0 (b + r2 )3/2 n=1 ((2n` − b)2 + r2 )3/2 ((2n` + b)2 + r2 )3/2
6
The first term integrates to 1, so we will deal with the sum. Consider one term for some given
n, and say we integrate out to some finite but large R. Note that we can only assume b R,
not n` R. We will drop all factors of b2 . The term integrates out to
2n` − b 2n` + b
−√ +√
4n2 `2 + R2 − 4n`b 4n2 `2 + R2 + 4n`b

2n` − b 1 4n`b 2n` + b 1 4n`b
≈− √ 1+ +√ 1−
R2 + 4n2 `2 2 R2 + 4n2 `2 R2 + 4n2 `2 2 R2 + 4n2 `2

1 −2n`b(2n` − b) − 2n`b(2n` + b)
=√ b − 2n` + 2n` + b +
R2 + 4n2 `2 R2 + 4n2 `2
2R2 b
= 2
(R + 4n2 `2 )3/2
2b 1
= .
R (1 + (2n`/R)2 )3/2
Since R → ∞, this sum in this limit can be written as an integral, giving
2b ∞ 2b R ∞ dx
Z Z
1 b
2 3/2
dn = 2
= .
R 0 (1 + (2n`/R) ) R 2` 0 1 + x `
Therefore, the total charge is simply −q(1 − b/`) .

If you found the analysis in part (c) quite tricky, know that you’re in good company. About 10
papers have been written about this exact system in the American Journal of Physics alone!
2 Capacitors
Idea 2
For a single, isolated conductor with charge Q, the self-capacitance is defined as
Q = Cφ
where φ is the potential difference between the conductor and infinity. For a set of two
isolated conductors with charges ±Q, the capacitance is defined as
Q = Cφ
where φ is the potential difference between the two conductors.
Idea 3
The definitions of C above are only useful when you have only one or two conductors in the
problem, respectively. In a situation with more than two, it’s very tricky to use the above
definitions, because all the conductors will affect each other; even a neutral conductor will
have an effect since there will be induced charges on its surface.
Instead, it’s better to revert to more general principles. The underlying principle behind
capacitance is linearity: the charges are linearly related to the fields. For multiple capacitors,
7
the most general possible linear relation is

X
Qi = Cij φj
j
where conductor i has charge Qi and potential φi , the potential is taken to be zero at infinity,
and the Cij are called general coefficients of capacitance, or in electrical engineering, the
Maxwell capacitance matrix. Similarly, inverting this relation,
X
φi = pij Qj
j
where the pij are called coefficients of potential.
In Olympiad physics, you’ll almost never want to compute the Cij or pij explicitly. Instead,
the point here is that if you’re given the charges and want the potentials, or vice versa, you
can build up the answer you want using the principle of superposition, computing all the
fields you need, e.g. using Gauss’s law.
Remark
General capacitance coefficients are discussed further in section 3.6 of Purcell. One nontrivial
fact is that Cij = Cji , which is proven by energy conservation in problem 3.64 of Purcell.
Capacitance coefficients can be clunky to work with. For example, suppose you want to
compute the familiar capacitance of a system of two conductors. By definition, we have
Q1 = C11 φ1 + C12 φ2 , Q2 = C21 φ1 + C22 φ2 .
An ordinary two-plate capacitor corresponds to the special case of opposite charges on the
plates, so we write Q = Q1 = −Q2 . There is a potential difference V across the plates, so
φ1 = φ2 + V , and plugging this in gives
Q = (C11 + C12 )φ2 + C1 V, −Q = (C22 + C21 )φ2 + C 0 V.
Eliminating φ2 from the system of equations above, we find the familiar capacitance
Q 2
C11 C22 − C12
C= =
V C11 + C22 + 2C12
where we used C12 = C21 . This is quite an inconvenient formula, so as a result we won’t
consider general capacitance coefficients any further, except briefly for practice in problem 8.
[2] Problem 7 (Purcell 3.21). Consider a capacitor made of four parallel plates with large area A,
evenly spaced with small separation s. The first and third are connected by a wire, as are the
second and fourth. What is the capacitance of the system?
Solution. Say we charge the conductors to equal and opposite charges. Then by symmetry, the
surface charges are
σ1 , −σ2 , σ2 , −σ1
8
reading left to right (1 to 4). The field in between the first and the second plates is σ1 /0 , and the
field between the second and the third plates is (σ1 − σ2 )/0 . Since 1 and 3 are connected, and 2
and 4 are, we must have that
σ1 s = (σ2 − σ1 )s =⇒ σ2 = 2σ1 .
Thus, the potential difference is σ1 s/0 , so the capacitance is C = Q/φ = 3σ1 A/(σ1 s/0 ) = 30 A/s.
[3] Problem 8. Consider two concentric spherical metal shells, with radii a < b.
(a) Compute their capacitance using Gauss’s law.
(b) Compute their capacitance by computing the four capacitance coefficients, verifying that
C12 = C21 along the way, and using the result for C above.
Solution. (a) Let the shells have charge ±Q. The field between the shells is (Q/4π0 r2 )r̂, so

Q 1 1
V = − .
4π0 a b
Thus the capacitance is

Q ab
C= = 4π0 .
V b−a
(b) Let the first capacitor be the inner shell. If only the outer shell is charged, with charge Q,
then φ1 = φ2 = Q/4π0 b. The general capacitance equations in this case are
0 = C11 φ1 + C12 φ2 , Q = C21 φ1 + C22 φ2
from which we see that

C11 + C12 = 0, C21 + C22 = 4π0 b.
Now suppose only the inner shell is charged, with charge Q. In this case we have φ1 = Q/4π0 a
while φ2 = Q/4π0 b, so
Q = C11 φ1 + C12 φ2 , 0 = C21 φ1 + C22 φ2
from which we see that

C11 C12 C21 C22
+ = 4π0 , + = 0.
a b a b
Solving these four equations for the capacitance coefficients gives
ab ab b2
C11 = 4π0 , C12 = C21 = −4π0 , C22 = 4π0 .
b−a b−a b−a
Plugging into the general formula, we have
4π0 (ab)b2 − (ab)2 4π0 b2 a(b − a) ab

C= = = 4π0 .
b − a ab + b2 − 2ab b − a b(b − a) b−a
This is certainly a longer route to get to the same conclusion!
9
[3] Problem 9 (MPPP 152). Four identical non-touching metal spheres are positioned at the vertices
of a regular tetrahedron, as shown.
A charge 4q given to sphere A raises it to a potential V . Sphere A can also be raised to potential V
if it and one of the other spheres are each charged with 3q. What must be the size of equal charges
given to A and two other spheres for the potential of A to again be raised to V ? What if all four
spheres are used?
Solution. Suppose sphere A has charge 4y + 3x + 3x, and spheres B and C have charges of 3x. We
can write this a superposition of three cases, (4y, 0, 0, 0) + (3x, 3x, 0, 0) + (3x, 0, 3x, 0). The potential
at A is then just (y/q)V + 2(x/q)V because of linearity (one can use the capacitance coefficients
to formalize this). Therefore, we want to set V = (y/q) + 2(x/q)V , so 2x + y = q. We also want
4y + 3x + 3x = 3x, so y = −3x/4. Thus, 2x − 3x/4 = q, so x = 4q/5 and y = −3q/5. Thus, the
charges are 12q/5 .
We know consider the case of all four spheres being used. Suppose sphere A has charge 4y +
3x + 3x + 3x, and all others have charge 3x. The potential at A is then (y/q)V + 3(x/q)V = V ,
so 3x + y = q. Also, 4y + 9x = 3x, so y = −3x/2. Thus, 3x − 3x/2 = q, so x = 2q/3 and y = −q.
Thus, the charge is 3x = 2q .
Idea 4
The energy stored in a capacitor is
1 1
U = QV = CV 2
2 2
where V is the voltage difference and C is the (mutual) capacitance. The mutual capacitance
C adds in parallel, while 1/C adds in series.
[2] Problem 11 (Purcell 3.24). Some estimates involving capacitance.
(a) Estimate the capacitance of the Earth.
(b) Make a rough estimate of the capacitance of the human body.
(c) By shuffling over a nylon rug on a dry winter day, you can easily charge yourself up to a
couple of kilovolts, as shown by the length of the spark when your hand comes too close to a
grounded conductor. How much energy would be dissipated in such a spark?
10
Solution. (a) The Earth is a sphere of radius of order 107 m, so
C = 4π0 r ∼ 10−3 F.
Notably, we can make “bigger” capacitors in the lab! Nonetheless, a huge amount of charge
can be delivered to the Earth, such as by lightning strikes. This is because the voltage of the
Earth is also huge, which is possible because its huge size means the corresponding electric
fields aren’t that big.
(b) A human is approximately a sphere of radius 0.5 m. Then, C = 4π0 r ∼ 5 × 10−11 m.
(c) We have U = 12 CV 2 ∼ 10−4 J.
[2] Problem 12. The total energy can also be found by integrating the electric field energy,
Z
0
U= E 2 dV.
2
(a) Show that this agrees with U = CV 2 /2 for a parallel plate capacitor.
(b) Show that this agrees with U = CV 2 /2 for a capacitor made of concentric spheres.
The general proof is more advanced, but if you’re interested, one slick method is given in problem
1.33 of Purcell.
Solution. (a) Let the plate area be A and the distance between them be d. Then
σ2 C σ 2 d2 CV 2
U = E 2 (Ad) = Ad = = .
2 20 2 20 2
(b) Let the radii be R1 and R2 and the charges be ±Q. The field is Q/(4π0 r2 ), so
R2
Q2 dV Q2 4πr2 dr Q2
Z Z
0 1 1
U= 2 4
= = − .
2 2
16π 0 r 32π 2 0 R1 r 4 8π0 R1 R2
On the other hand, this should be equal to U = QV /2, which follows directly from the result
of problem 8.
[3] Problem 13 (Purcell 3.26). A parallel-plate capacitor consists of a fixed plate and a movable plate
that is allowed to slide in the direction parallel to the plates. Let x be the distance of overlap.
The separation between the plates is fixed. Let C(x) be the capacitance.
(a) Assume the plates are electrically isolated, so that their charges ±Q are constant. By differ-
entiating the energy, find the leftward force on the movable plate in terms of Q and C(x).
(b) Now assume the plates are connected to a battery, so that their potential difference φ is held
constant. Find the leftward force on the movable plate, in terms of φ and C(x).
11
(c) If the movable plate is held in place, the two answers above should be equal because nothing
is moving. Verify that this is the case, being careful with signs.
(d) In terms of electric fields, why is there a force on the movable plate? Does the effect invoked
in the answer to this part change the conclusion of parts (a) through (c) at all?
Solution. (a) The energy as a function of x is
Q2
U (x) =
2C
where we understand that C is also a function of x. Thus, the force on the plate is
Q2 d Q2 dC

dU 1
F =− = − = .
dx 2 dx C 2C 2 dx
(b) Here, the energy is U (x) = 12 Cφ2 , so naively we have
dU φ2 dC
F =− =− .
dx 2 dx
This is negative, while the answer to part (a) is positive. The reason is that U should reflect
the total energy of the system – and in this case, the system must include the battery that
does work to maintain the potential difference φ.
Say x increases by dx. Let the change in capacitance be dC. Then, dQ = φ dC. Thus, the
work the battery does is
dW = φ dQ = φ2 dC.
Therefore, if F is the net force the plate feels, we have that
1 dC
dW = F dx + dU =⇒ F = φ2 .
2 dx
(c) Let FQ be the first force, and Fφ the second. We have that
Q2
FQ /Fφ = = 1.
φ2 C 2
If we didn’t account for the subtlety in part (b), we would have gotten −1 here.
(d) The force is due to the fringe fields, as the non-fringe field is perfectly vertical. Thus, we
have managed to compute a fringe field effect using energy, even though we were able to
completely ignore the fringe fields in the energy calculation! This is yet another example of
how conservation laws can hand you information that’s very hard to get otherwise.
Idea 5: Dielectrics
While we’re on the subject of capacitors, it’s useful to introduce dielectrics, which will be
important later. All electrical insulators are dielectrics. In the presence of an electric field, a
dielectric will polarize, with positive charges displaced slightly along the field. The resulting
electric dipoles distributed throughout the material in turn create a field that tends to
12
weaken the original applied field.
Each part of a dielectric polarizes based on the local electric field, but that electric field
depends on the applied field, and the polarization of every other piece of the dielectric. Thus,
solving for the electric field for a general dielectric geometry is very difficult, and usually
not possible in closed form, just like how it’s usually not possible to solve for the field of a
charged conductor. In Olympiad physics, you will almost always consider highly symmetric
situations, where a dielectric simply reduces the applied electric field everywhere by a factor
of κ, called the dielectric constant. (We’ll consider some trickier situations in E8.)
Consider a parallel plate capacitor with charge ±Q on each plate. If a dielectric is inserted
with the charge kept the same, then the field inside is reduced by a factor of κ. Thus, the
capacitance C = Q/V increases by a factor of κ. Dielectrics may increase the amount of
energy that can be stored in a capacitor, which is typically limited by the voltage V0 where
electrical breakdown occurs. So if V0 stays the same, the maximal stored energy U = CV02 /2
goes up by a factor of κ.
Plugging in the definition of C, this result implies that the energy density in the capacitor is
κ0 E 2 . But we showed in E1 that the energy density of the electric field is only 0 E 2 . The
extra energy is stored in the dielectric material itself: it takes energy to separate positive
and negative charges within the dielectric, as if we were stretching many microscopic springs.
This potential energy is released when the capacitor is discharged.
3 Tricky Problems
Example 1: PPP 151
A closed body with conducting surface F has self-capacitance C. The surface is now dented
so that the new surface F ∗ is entirely inside F . Prove that the capacitance has decreased.
Solution
The energy stored in the capacitor is U = Q2 /2C. Therefore, if we give the capacitor a fixed
charge Q, proving that F ∗ has lower C is equivalent to showing that it takes positive work
to dent the foil from F to F ∗ . It’s cleaner to show the other direction, i.e. that starting from
F ∗ , we can get to F while only lowering the energy.
Suppose without loss of generality that F is infinitesimally larger than F ∗ . (We can break
any finite change into infinitesimal stages and repeat this argument.) We can go from
F ∗ to F by just taking each charge on the surface and moving it outward until it hits F .
This lowers the energy because the electric field is always directed outward, as we proved in E1.
At this point, the charges lie on F , but they don’t have the right distribution, i.e. F is not
an equipotential. Now we let the charges spontaneously redistribute themselves so that F is
13
again an equipotential. This again lowers the energy, proving the desired result.
Example 2
Are there charge distributions that aren’t spherically symmetric, but which produce an exact
r̂/r2 field outside of them?
Solution
If you know a bit about the multipole expansion, this might seem like a daunting question.
To make the field exactly r̂/r2 , you need to make sure the charge distribution has no dipole
moment, no quadrupole moment, no octupole moment, and so on to infinity, and it seems
impossible to satisfy all of these constraints without spherical symmetry. But we have
already seen an example of such a charge distribution earlier in the problem set!
Recall that when we treated the method of images for spheres, we found that in some
situations, the complicated charge densities on conducting spheres were exactly the same as
those produced by a fictitious image charge inside the sphere, and generally away from its
center. If we place the origin at that image charge, then we have an example of a charge
distribution that is perfectly r̂/r2 far away, but which isn’t spherically symmetric. (The
general solution is given here.)
[2] Problem 14. Consider a set of n conducting, very large parallel plates, placed in zero external
electric field. The plates are given charges Qi . If the left ends of the plates are at locations xi , and
the plates have thickness di , what is the total charge on the left end of the leftmost plate, and the
right end of the rightmost plate?
Solution. If you try to solve this one directly, the calculations can get pretty messy, but there’s
a simple solution using basic facts about capacitors. First, consider the two leftmost plates. Since
they’re conductors, the electric field must vanish inside them. Now split the charge in the problem
into two parts: (1) the charge on the right side of the leftmost plate, and the left side of the second
leftmost plate, and (2) all other charge.
The electric field due to (2) has some uniform value Eext within the first and second plates.
Therefore, in order for the electric field to vanish in both of those plates, the electric field due to
(1) has to be the same in both plates. This is only possible if the two charges in (1) are opposite,
which corresponds to those sides forming a parallel plate capacitor.
We can then repeat this reasoning, which shows that the charges on the right side of plate i
must cancel with the charges on the left side of plate i + 1, producing no field anywhere except in
between plates i and i + 1. All that is left is the left end of the leftmost plate and the right end of
P
the rightmost plate, which must have total charge i Qi . Finally, in order to ensure no field in any
P
plate, these two charges must be equal, so the answer is that both have total charge ( i Qi )/2.
[2] Problem 15 (Purcell 3.9). A conducting spherical shell has charge Q and radius R1 . A larger
concentric conducting spherical shell has charge −Q and radius R2 .
(a) If the outer shell is grounded, explain why nothing happens to the charge on it.
14
(b) If instead the inner shell is grounded, e.g. by connecting it to ground by a very thin wire that
passes through a very small hole in the outer shell, find its final charge.
(c) It’s not so clear why charge would leave the inner shell in part (b), thinking in terms of forces.
A small bit of positive charge will certainly want to hop on the wire and follow the electric
field across the gap to the larger shell. But when it gets to the larger shell, it seems like it
has no reason to keep going to infinity, because the field is zero outside. And, even worse, the
field will point inward once some positive charge has moved away from the shells. So it seems
like the field will drag back any positive charge that has left. Does charge actually leave the
inner shell? If so, what’s wrong with the above reasoning?
Solution. (a) The potential at the outer shell due to itself is −Q/4π0 R2 and the potential due
to the inner shell is Q/4π0 R2 , so it is zero overall. Thus, the outer shell is already effectively
grounded.
(b) The potential at the inner shell due to itself is Q0 /4π0 R1 and the potential to the outer shell
is −Q/4π0 R2 . Since the total must be zero, Q0 = R1 Q/R2 .
(c) The key mistake is that the positive charges are repelled also by the charges behind it in the
wire. So yes, eventually the field due to the shells may even become inward, there is a whole
line of plus charge behind a given charge that force it forward.
Another way of saying this is that a wire has negligible capacitance; like a thin metal pipe of
water, it cannot store extra net charge but can only let charge move rigidly through the entire
thing. It is energetically favorable for this to happen, so even if some charges don’t want to
move forward, their neighbors will push them forward.
[2] Problem 16. The usual expression for the capacitance of a parallel-plate capacitor is A0 /d.
However, in reality the field within the capacitor is not perfectly uniform, and there are fringe fields
outside. Taking these effects into account, is the true capacitance higher or lower than A0 /d?
Solution. For concreteness, suppose the plates are circular disks, with charge ±Q, and consider
the electric field along the axis of symmetry. In the naive derivation, we assume the charge density
on the plates is uniform. Then we approximate the plates as infinite in order to use Gauss’s law to
conclude that the field inside is σ/0 = Q/A0 . This is inaccurate for two reasons:
• The plates are not actually infinite, so the field on the symmetry axis should actually be smaller.
• The charge distribution is not actually uniform. Instead, since the like charges on each plate
repel each other, some charge gets pushed outward. This further decreases the field on the
symmetry axis.
Therefore, the field on the symmetry axis is less than σ/0 , so the voltage drop is lower, which
implies the capacitance is higher.
R
[2] Problem 17 (Purcell 4.16). In a parallel plate capacitor, the quantity E · ds should be equal to
V for any path that connects the two plates.
A charged capacitor can be dischargedR by attaching a wire to the external surfaces of the plates.
No matter how one attaches the wire, E · ds along the wire should be equal to V . And as we’ve
argued in E2, this is sufficient to cause charges to move along the wire, even if the electric field
15
points in the “wrong” direction at some points along the wire, because the wire has negligible
capacitance: charges within it move rigidly, each pushing the next one and pulling the previous one.
But it’s puzzling how this works for a capacitor, because the electric field is supposed to be
essentially zero just outside it. Consider two possible limiting cases for the wire’s shape.
R
In each case, explain qualitatively how E · ds can be equal to V . In particular, how large are the
contributions from the distinct segments of the wire (the horizontal and vertical parts in the first
case, and the straight and curved parts in the second)?
Solution. In the first case, the horizontal parts of the wire contribute almost nothing. That’s
because the radial part of the electric field vanishes within the conductor plates themselves (since
they must be equipotentials), and the horizontal path is right next to the plates. Therefore, the
contribution is almost entirely from the vertical segment.
You might be wondering how this is possible, because at the edge of the plates, there seems to
be less charge nearby, so the electric field should be smaller than near the middle of the plates. The
resolution is that the surface charge density near the edge of plates is a lot higher than the surface
charge density near the middle, because like charges repel.
It’s interesting to compare this to the case of two parallel plates with uniform charge density.
In that case, the vertical segment contributes roughly V /2. To see why, consider putting a second,
identical parallel plate capacitor directly to the left of the first one. Now the vertical segment is
in the middle of a big capacitor, and has voltage drop V . So each of the two halves of that big
capacitor contributes V /2 to the vertical segment. However, the total voltage drop is still V , because
for uniform charge density there are substantial horizontal fields, so that the horizontal segments
contribute roughly V /4 each.
In the second case, the result is due to the far-field behavior. When you zoom out, the capacitor
looks like a dipole, so the field at long distances is a dipole field. Now, the dipole field falls off as
1/r3 , and the circumference of the curved part is proportional to r, so the contribution of this part
goes as r/r3 → 0 as r → ∞. So all the contribution is from the straight part.
To see how this can be the case, note that the vertical field just above the capacitor plates
is negligible; the dipole field only kicks in once we’re far enough so that the plates look small,
i.e. subtending a small angle from our perspective. If the plates are squares of side length a, this
occurs at a distance of order a. Then
Z Z ∞
p 2 p
E · ds ∼ 2 3
dr =
a 2π0 r π 0 a 2
where p is the electric dipole moment. If the plates are separated by a distance d a, then
p = Qd = σa2 d, giving
2 p 2 σd 2
2
= = V
π 0 a π 0 π
16
which is on the order of the voltage across the capacitor plates. Of course, we didn’t get precisely
V because we made a lot of approximations in the calculation, but this illustrates the conceptual
point: the full integral of E · ds can indeed be equal to V , and most of the contribution to this
integral comes from the part of the vertical wire which is a distance of order a from the capacitor.
[3] Problem 18. Consider two conducting spheres of radius r separated by a distance a r. The
spheres can be thought of as the two plates of a parallel plate capacitor.
(a) By assuming the surface charge density on each sphere is uniform, estimate the capacitance.
(b) In reality, the surface charges on each sphere will be distorted by the other sphere. The surface
charges assumed in part (a) will induce changes in the surface charges, represented by an
image charge for each sphere. But these image charges will induce further image charges, and
so on, yielding an infinite series for the charge distribution and hence the capacitance. Find
the first two terms in the series for the capacitance.
(c) If a/r = 10, roughly estimate the error in neglecting the other terms.
Solution. (a) The potential of a single sphere, relative to infinity, is V = Q/4π0 r. Hence the
two spheres, at this level of approximation, have potentials ±Q/4π0 r, and the capacitance is
Q
C= = 2π0 r.
V1 − V2
This is ignoring any interaction between the charges on different spheres.
(b) The uniform charge on one sphere looks like a point charge to the other sphere, by the shell
theorem, and this induces an image charge. The same applies for the spheres in reverse. The
image charges are smaller than the original ones by a factor of r/a, by the result of problem 2.
Now, these “first-order” image charges on each sphere in turn induce second-order image
charges in the other sphere, which are smaller by another factor of r/a, and so on. Once we
sum up an infinite series of image charges, we get the true charge configuration, from which
we can compute the exact capacitance, which is now a power series in r/a.
In this part we’re only interested in the first two terms. We found the first term in part (a),
and the next term comes from accounting for the “first-order” image charges. Let the spheres
have charges ±Q. The negative sphere induces an image charge of Qr/a within the positive
sphere, so that the negative sphere plus the image charge yield a total of zero potential on
the positive sphere. So the total potential of the positive sphere is still V = Q/4π0 r, except
that the total charge on the sphere is not Q, but Q(1 + r/a). Applying the same reasoning in
reverse to the other sphere,
Q(1 + r/a)
C= = 2π0 r(1 + r/a + O((r/a)2 )).
V1 − V2
It makes sense the true capacitance is higher than the naive guess of part (a), because in part
(a) we demanded a uniform charge density, and charges can lower their energy further if we
let them spread out.
(c) As we’ve seen, the answer is a power series in r/a. So if we take only the first order term, the
second order and higher terms are suppressed by r/a = 1%, so we expect our approximation
to be good within a few percent.
17
Example 3
Find the interaction force between a dipole of dipole moment p and a conducting sphere of
radius r, separated by a distance R r.
Solution
There are several ways to do this problem using the method of images; here’s one. Consider
the image charges that the two charges in the dipole make. These produce an image dipole
right at the center of the sphere, and using the results of problem 2, the dipole moment is
p0 = p(r/R)3 . Now, the force on the original dipole is
d 2kp0

d
F = p E(z) =p
dz z=R dz z 3 z=R
where E is the field made by the image dipole, and we used the results of E1. Thus,
6kp2 r3

2 3 d 2k

F = p (r/R) = − .
dz z 3 z=R R7
As usual for charge induction effects, this force is attractive, and falls off quickly with distance.
Example 4
Estimate the interaction force between a point charge q and a thin conducting rod of length `,
which is a distance L ` from the charge and oriented along the separation between them.
Solution
The interaction occurs because the point charge induces negative charges on the near end of
the rod, and positive charges on the far end. These charges are then acted on by the electric
field of the point charge, causing a force.
To get a very crude estimate, let’s just suppose that charge Q appears on the far end and
charge −Q appears on the near end. The resulting field produced in the middle is
kQ
E∼ .
`2
On the other hand, this needs to cancel a field from the point charge of
kq
E∼
L2
which tells us that Q ∼ (`/L)2 q. The force on the induced charges is
kqQ`2 kq 2 `4

1 1
F ∼ kqQ − ∼ − ∼ − .
L2 + `2 L2 L4 L6
Again, the force is attractive, and falls off quickly with distance.
18
[3] Problem 19 (Physics Cup 2017). Estimate the interaction force between a point charge q and an
infinitely thin circular neutral conducting disc of radius r if the charge is at the axis of the disc,
and the distance between the disc and the charge is L r.
Solution. The interaction is because charges redistribute on the disc to keep it an equipotential.
As an extremely rough approximation, suppose that charge Q appears at the center of the disc and
charge −Q appears on the rim. Then essentially by dimensional analysis, the electric field in the
disc is
kQ
E∼ 2.
r
On the other hand, the electric field due to the point charge along the disc is of order
kQ r
E∼
L2 L
where the r/L factor is from projecting the field along the disc. Then
r3
Q ∼ −q .
L3
The force is, by Coulomb’s law,
kq 2 r5

1 L
F ∼ kqQ 2
− 2 ∼ .
L (L + r2 )3/2 L7
01T
A more accurate estimate would get the numeric prefactors.
[5] Problem 20. IPhO 2012, problem 2. A challenging electrostatics/fluids problem; some prior
exposure to surface tension is helpful. (For more about the kinds of bubbles encountered in this
problem, see section 5.9 of Physics of Continuous Matter by Lautrup.)
Example 5
Find the charge distribution on a conducting disc of radius R and total charge Q.
Solution
In general, there are very few situations where the charge distribution on a conductor can
be found explicitly. As you’ve seen, some of the simplest examples can be solved with image
charges. Some more complex, two-dimensional examples can be solved with a mathematical
technique called conformal mapping. And this special example can be solved with a neat trick.
Consider a uniformly charged spherical shell centered on the origin, and consider a point P
inside the shell, on the xy plane. The electric field at point P is zero, by the shell theorem.
Recall that in the usual proof of the shell theorem, one draws two cones opening out of
P in opposite directions. The charges contained in each cone produce canceling electric fields.
Now imagine shrinking the spherical shell towards the xy plane, so it becomes elliptical. The
crucial insight is that the shell theorem argument above still works, for points on the xy
plane. When we squash the shell all the way down to the xy plane, it becomes a disc, with
19
zero electric field on it. This is thus a valid charge distribution for a disc-shaped conductor,
and by the uniqueness theorem, it’s the only one.
By√keeping track of how much charge gets squashed to radius [r, r + dr], we find σ(r) ∝
R/ R2 − r2 , and fixing the proportionality constant gives
Q
σ(r) = √ .
4πR R2 − r2
4 Electrical Conduction
We now leave the world of electrostatics and consider magnetostatics, the study of steady currents.
Idea 6
In a conductor with conductivity σ, the current density is
J = σE.
Alternatively, E = ρJ where ρ is the resistivity. The current and charge density satisfy
∂ρ
∇·J=− .
∂t
The current passing through a surface S at a given time is
Z
I= J · dS.
S
Since J ∝ E, we have Ohm’s law V = IR, where V is the voltage drop across the resistor.
The power dissipated in a resistor is P = IV . The resistance R adds in series, while 1/R
adds in parallel.
[2] Problem 21 (HRK). A battery causes a current to run through a loop of wire.
(a) Suppose the wire makes a sharp corner. How do the charges know to turn around there?
(b) A copper wire with conductivity σ is joined to an iron wire with conductivity σ 0 < σ. For
the current in both sections to be the same, the electric field in the iron wire must be higher.
How does that happen?
In general, the surface charge distribution in a DC circuit can be quite complex; the aspects shown
in these two short questions are just the beginning. For some more about this, see this paper.
Solution. (a) The first charges to make it there don’t; they just stop at the surface of the wire,
due to the attraction from the protons. Once this charge builds up at the kink, it repels
the next electrons so that they automatically turn around. This typically occurs extremely
quickly, as the relevant timescale is the RC of the wire and C is tiny. The amount of charge
required is very small, less than a few hundred electrons.
20
(b) It’s the same story as part (a). The first charges to reach the iron will start moving slower,
because the fields are the same. This then causes a buildup of charge at the interface between
them, which increases the field in the iron and decreases the field in the copper. In the steady
state, the current densities in both are equal. Again, this occurs very quickly and requires
very little charge.
[2] Problem 22 (PPP 22). Two students, living in neighboring rooms, decided to economize by
connecting their ceiling lights in series. They agreed they would each install a 100 W bulb in their
own rooms and that they would pay equal shares of the electricity bill. However, both tries to get
better lighting at the other’s expense. The first student installed a 200 W bulb, while the second
student installed a 50 W bulb. Which student subsequently failed the end-of-term examinations?
Solution. A standard bulb is designed to be hooked up in parallel with other bulbs, across some
fixed voltage V . Since P = V 2 /R, higher wattage bulbs have lower resistance.
Since the bulbs are in series, then have the same current through them. Since P = I 2 R, that
means the bulb with the higher wattage rating draws less power, so the first student fails. The
second student’s bulb burns four times brighter. (The fact that the brightness of a bulb depends
on what else is attached is also a reason why hooking lights up in series is impractical.)
To warm up for DC circuits, we’ll consider some resistor network problems.
Idea 7
If any two points in a resistor network are at the same potential, nothing will change if the
two points are connected together and treated as one. More generally, the resistance of any
resistor directly connecting the two points may be changed freely.
Example 6
Consider the 3 × 3 grid below, where every edge is a resistor R.
Find the equivalent resistance between nodes 1 and 16.
Solution
By the above idea, we can short together nodes 2/3, and 14/15, by the diagonal symmetry
of the network. Next, we can break nodes 8 and 9 into two pieces.
21
This is valid because the separated nodes 8a/8b and 9a/9b still have the same potential in
the new network, by the diagonal symmetry. (This is using the above idea in reverse.) Now,
the circuit has been reduced to combinations of series and parallel resistors. The resistance
between 1 and 2/3 is R/2. The resistance between 2/3 and 14/15 is the combination of three
networks in parallel, and finally the resistance of 14/15 and 16 is R/2. Thus,
!
1 1 1 −1 1

1 13
Req = + + + + R = R.
2 3 3 2 2 7
You won’t see any resistor problems as complicated as this one for the rest of the training,
because they’re kind of contrived; the point of this example was just to show multiple uses
of symmetry techniques.
Example 7: PPP 23
A black box contains a resistor network and has two output terminals.
If a battery of voltage V is connected across the first terminal, the voltage across the second
terminal is V /2. If a battery of voltage V is connected across the second terminal, the voltage
across the first terminal is V . Find one possible configuration of the resistors inside the box.
Solution
A simple configuration with two equal resistors works.
When a battery is connected across II, the horizontal resistor doesn’t do anything. When a
battery is connected across I, the two resistors comprise a voltage divider.
22
[2] Problem 24 (IPhO 1996). Consider the following resistor network.
Find the equivalent resistance between A and B.
[3] Problem 25. Consider a cube of side length L whose edges are resistors of resistance R.
√
(a) Compute the resistance between two vertices a distance 3L apart.
√
(b) Compute the resistance between two vertices a distance 2L apart.
(c) Compute the resistance between two vertices a distance L apart.

√
(d) Generalize to vertices nL apart on an n-dimensional cube. (Give your answer in the form
of a summation.)
Solution. (a) Let the two vertices be A and B. Let the vertices distance 1 from A be labeled a
and distance two labeled b. Note that all the vertices labeled a have the same potential by
symmetry, and same for b. Thus, we can treat all the vertices labeled the same thing as one
vertex.
We have 3 connections from A to a, 6 from a to b, and 3 from b to B. Thus, our resistance is
R/3 + R/6 + R/3 = 5R/6 .
(b) Apply potential V and −V to the vertices, and label the rest of the vertices as shown.
We claim that all the vertices labeled c have potential 0. The idea is then that negating the
potentials V and −V must negate the potentials at all vertices. But negating is equivalent to
a simple reflection that preserves the locations of the vertices labeled c. Thus the only option
is that their potential is 0. Therefore, all the cs can be treated as one vertex, and we have the
drawn equivalent circuit. This is a combination of parallel and series, and we compute the
answer to be
1
2· 1 R = 3R/4 .
1 + 1 + 1+1/2
23
(c) In a similar fashion, the points labeled the same below have the same potential, and on the
right is the equivalent circuit.
This is again just a series/parallel problem, and we compute the answer to be

1
R 1 1 = 7R/12 .
1
+ 1 1
+ 12
+ 1
2 1
1 + 1 +1+ 1
2 2 2
(d) The coordinates take the form (x1 , x2 , . . . , xn ) where xi is zero or one. We consider the vertices
(0, 0, . . . , 0) and (1, 1, . . . , 1). The first vertex is connected to all the vertices with one 1, of
which there are n. By symmetry, these are all at the same voltage. Next, these vertices are
connected to all the vertices with two 1’s, of which there are n2 , and so on.

We hence have n + 1 effective vertices of different voltages. Consider the vertex representing
n

points with k 1’s. The number of connections to points with k + 1 1’s is k (n − k). Then by
adding series and parallel resistances,
n −1
X n
Req =R (n − k) .
k
k=0
For example, this recovers the result of part (a) for n = 3.
[2] Problem 26 (PPP 158). Consider the circuit below, where every resistor is 1 Ω.
(a) Find the equivalence resistance between the input terminals.
(b) Do the same in the case where the chain is infinitely long.
Solution. (a) Suppose the rightmost resistor has unit current flowing down, and let V be the
potential difference across A and B.
Number the resistors, starting from the back, and going to the left. Let Ik be the current in
Rk . We claim that
Ik = Fk
24
where the Fk are the Fibonacci numbers F1 = F2 = 1 and Fn = Fn−1 + Fn−2 . This is true by
induction. Note that
I2k = I2k−1 + I2k−2
by junction law. By the loop law,
I2k+1 = I2k + I2k−1 .
Note that all currents are either to the right or down. Let the number of vertical resistors be
n. Then, the potential difference is
V = I2n−1 + I2n = I2n+1 = F2n+1 .
The equivalent resistance is

R = (F2n+1 /F2n ) Ω.
(b) By taking the limit n → ∞ above, we get the golden ratio,

√
1+ 5
R= Ω.
2
There’s another, slicker way to do this. In the infinite case, if we let the answer be R, then
we have
1
R=1+
1 + R1
which is equivalent to the quadratic
√
2 1± 5
R − R − 1 = 0, R= .
2
Then taking the positive root gives the answer.
This presents a question: what’s the deal with the negative root? Both the positive and
negative roots are technically equally good. In fact, there exist circuit elements with negative
resistance, as you’ll see in E6, though they need an active source of power to maintain.
However, the positive root is still better in an important way. Suppose you start out with
something on the right end, with (possibly negative) equivalent resistance R0 . If we attach
two 1 Ω resistors on the left, as in the diagram, then we get some new resistance R1 , and
attaching two more gives R2 , and so on. The two roots for R found above are the two fixed
points for this iteration. However, the positive root is more physical
√ because it’s the stable
fixed point. If you
√ start out with any resistance other than (1 − 5)/2, the recursion will
converge to (1 + 5)/2. (For example, in part (a), we showed this √ happens if you start with
1 Ω.) Only if you start out with the exact negative resistance (1 − 5)/2 will you stay there.
Therefore, even though a mathematically infinite network doesn’t have anything “on the√right
end” (since it has no end at all), it’s still meaningful to say that “the” resistance is (1 + 5)/2.
When we introduce infinite objects in physics, we usually do so just to get a mathematically
tractable approximation for a √ real, finite object. And for almost any long, but finite chain,
you’ll get an answer near (1 + 5)/2, so that’s the useful answer in the infinite case.
[3] Problem 27 (PPP 159-161). Superposition can be a useful trick to analyze circuit networks.
25
(a) Consider an infinite two-dimensional grid of identical resistors R.
Find the equivalent resistance between two neighboring points by considering the superposition
of 1 A flowing into one point, and 1 A flowing out the other.
(b) What would the equivalent resistance be if the resistor directly connecting the two neighboring
points was removed?
(c) Now consider an icosahedron of identical resistors R. By superposing appropriate current

distributions, find the equivalent resistance between two neighboring vertices.
Solution. (a) We see that if just I flowed in, then the current in the joining resistor would be
I/4, so in this case its I/2. Thus, the potential drop is IR/2 for a current of I/2, so the
resistance is R/2 .
(b) Suppose the answer is R0 . We have that R0 and R attached in parallel give R/2. Thus,
1 1 2
0
+ = ,
R R R
so R0 = R .
(c) Consider the current distribution where I flows into a vertex, and I/11 leaves from each other
vertex. We see that the current in each edge coming out from the source vertex is I/5. So
superposing a similar but with signs flipped current distribution for an adjacent vertex, we
see that (12/11)IReff = 2RI/5, so Reff = 11R/30 .
Idea 8
In a circuit of resistors and batteries, Kirchoff’s loop rule states that the sum of the voltage
drops around a loop is zero. Kirchoff’s junction rule states that the net current flowing into a
vertex is zero. (This is technically nonzero, because of the effect of problem 21, but negligible
because wires have tiny capacitance.)
Remark
If the sum of the voltage drops around a loop is zero, then why would current ever want to
flow? After all, if you had a circular tube of water, the water would never flow, because the
net drop in height along the circle is zero. The reason current flows in circuits with batteries
is that within the battery, charges are moved from lower to higher electric potential energy,
26
just like how a pump could be used to move water upward to start a liquid circuit, by an
“electromotive force”.
But this immediately raises the question: what is this specific force? It can’t be the electric
force, because we just established that it’s pointing the wrong way. It’s not a magnetic
effect. For some setups, it is literally a mechanical force like a pump: in the Van der
Graaff generator, a motor drives the charges on a statically charged conveyor belt to higher
potential. But that’s not how batteries work.
In a battery, there is no specific force pushing charges from low to high electric potential.
Instead, the charges just jiggle around randomly, and the result emerges from the effects of
their many collisions. To understand this, consider a gravitational analogy.
Consider an ideal gas at temperature T released in the trough shown above. The gas
molecules will randomly collide, sometimes being propelled upward by chance. Sometimes, a
gas molecule will climb the hill and fall into the deep hole, at which point it is unlikely to
come out again. Thus, if the hole begins empty, it is energetically favorable for gas molecules
to fill it. But there is no attractive force pulling molecules up along the slope! Gravity
always points down; molecules go up the slope when they are randomly bounced that way.
This is essentially how the potential in an initially neutral battery is set up. The hole
corresponds to the lower energy state an electron can reach inside the anode, but there is no
long-range force pushing it there, just the average effect of random collisions.
[2] Problem 28 (Purcell 4.10). The basic ingredient in older voltmeters and ammeters is the gal-
vanometer, a device to measure very small currents. (It works via magnetic effects, but the exact
mechanism isn’t important here.) Inherent in any galvanometer is some resistance Rg , so a physical
galvanometer can be represented by the system shown below.
27
Consider a circuit such as the one shown, with all quantities unknown. We want to measure
the current flowing across point A and the voltage difference between points B and C. Given a
galvanometer with known Rg , and also a supply of known resistors (ranging from much smaller to
much larger than Rg ), how can you accomplish these two tasks? Explain how to construct your
two devices (called an ammeter and voltmeter), and also how you should insert them in the given
circuit. You will need to make sure that you (a) affect the given circuit as little as possible, and (b)
don’t destroy your galvanometer by passing more current through it than it can handle.
Solution. We explain how to create an ammeter and a voltmeter. As usual, attach the ammeter
in series, and the voltmeter in parallel.
To create an ammeter, connect a resistor R Rg in parallel with the galvanometer (including
Rg obviously). Suppose a current I passes through the device. If N = Rg /R, then its not hard to
see that the current through galvanometer is about I/N , and the system acts very much like a wire.
Therefore, to get I, multiply the reading by N .
To create a voltmeter, attach a resistor R Rg in series to the galvanometer (again, attach
this contraption in parallel to the circuit). If the voltage drop we want to measure is V , then the
current will be about V /R, so we can use the current to find V .
[2] Problem 29. 01^ USAPhO Quarterfinal 2009, problems 3 and 4.
[3] Problem 30. INPhO 2021, problem 1. A nice problem on practical circuit measurements. Note
that the question statement is a bit vague. You are supposed to keep track of quantities of order
RA /R and R/RV , but you are allowed to neglect quantities as small as RA /RV .
28
Electromagnetism III: Magnetostatics

Chapters 4 and 6 of Purcell cover DC circuits and magnetostatics, as does chapter 5 of Griffiths. For
advanced circuits techniques, see chapter 9 of Wang and Ricardo, volume 2. Chapter 5 of Purcell
famously derives magnetic forces from Coulomb’s law and relativity. It’s beautiful, but not required
to understand chapter 6; we will cover relativistic electromagnetism in depth in R3. For further
discussion, see chapters II-12 through II-15 of the Feynman lectures. There is a total of 79 points.
1 Static DC Circuits
We continue with DC circuits, in more complex setups than in E2.
Idea 1
When analyzing circuits, it is sometimes useful to parametrize the currents in the circuits
in terms of the current in each independent loop. This is typically more efficient, because it
enforces Kirchoff’s junction rule automatically, leading to fewer equations.
Example 1: Wheatstone Bridge
Find the current through the following circuit, if the battery has voltage V .
Solution
This circuit can’t be simplified using series and parallel combinations, so instead we
use Kirchoff’s rules directly. From the diagram, we see the circuit has three loops.
Let I1 be the clockwise current on the left loop, I2 be the clockwise current through
the top-right loop, and I3 be the clockwise current through the bottom-right loop. For
instance, this means that the current flowing downward through the top-left resistor is I1 −I2 .
The three Kirchoff’s loop rule equations are

3I1 R − I2 R − 2I3 R = V,
4I2 R − I1 R − I3 R = 0,
4I3 R − 2I1 R − I2 R = 0.
Adding the last two equations shows that
I1 = I2 + I3
1
and plugging this back in shows that 3I2 = 2I3 , so we have

2 3
I2 = I1 , I3 = I1 .
5 5
Since the answer to the question is just I1 , we can now plug this back into the first equation,

V 2 6 7
= 3I1 − I2 − 2I3 = 3 − − I1 = I1 .
R 5 5 5
This gives the answer, 5V /7R.
Incidentally, the circuit above is also called a Wheatstone bridge. We note that the current
through the middle resistor is zero when the ratios between the top and bottom resistances
match on both sides of it. Hence if three of these outer resistances are known, we can adjust
one of them until the current through the middle resistor vanishes, thereby measuring the
fourth resistor.
Idea 2
Since Kirchoff’s loop equations are linear, currents and voltages in a DC circuit with multiple
batteries can be found by superposing the currents and voltages due to each battery alone.
Idea 3: Thevenin’s Theorem and Norton’s Theorem

Consider any system of batteries and resistors, with two external terminals A and B. Suppose
that when a current I is sent into A and out of B, then a voltage difference V = VA − VB
appears. From an external standpoint, the function V (I) is all we can measure.
Now, by the linearity of Kirchoff’s rules, V (I) is a linear function, so we can write
V (I) = Veq + IReq .
In other words, V (I) is exactly the same as if the entire system were a resistor Req in series
with a battery with emf Veq (with the positive end pointing towards A). This generalizes
the idea of replacing a system of resistors with an equivalent resistance, and is known as
Thevenin’s theorem.
We can also flip this around. Note that I(V ) must also be a linear function, and we can write
V
I(V ) = Ieq + .
Req
This is precisely the I(V ) of an ideal current source Ieq (sending current towards B) in
parallel with a resistor Req . (An ideal current source makes a fixed current flow through it,
just like a battery creates a fixed voltage across it.) This is known as Norton’s theorem.
Since these functions are inverses of each other, you can see that the Req ’s in both equations
above are the same (both are equal to the ordinary equivalent resistance), and Veq = −Ieq Req .
2
Example 2
Consider some batteries connected in parallel, with emfs Ei and internal resistances Ri . What
is the Thevenin equivalent of this circuit?
Solution
The equivalent resistance is simply
!−1
X 1
Req = .
Ri
i
To infer Veq , we just need one more V (I) value. The most convenient is to set V = 0, shorting
all of the batteries. Each battery alone would produce a current of Ei /Ri , so
!
X Ei
0 = Veq + Req .
Ri
i
Thus, we have
! −1
X Ei X 1
Veq =   .
Ri Rj
i j
Remark
With ideal batteries, it’s easy to set up circuits that don’t make any sense.
For example, in the above circuit, Kirchoff’s rules don’t determine the currents; they only
say that i1 + i2 = 1 A. If the emfs of the batteries were different, the situation would be
even worse: the equations would be contradictory, with no solution at all! In real life, this is
avoided because all batteries have some internal resistance. Adding such a resistance to each
battery, no matter how small, resolves the problem and gives a unique solution.
[2] Problem 1 (Purcell 4.12). Consider the circuit below.
3
(a) Find the potential difference between points a and b.
(b) Find the equivalent Thevenin resistance and emf between points a and b.
Solution. (a) We’ll use loop currents, with positive being clockwise. Let the loop currents be
I1 , I2 , I3 , from top to bottom. We see that
E − E − 2RI1 + RI2 = 0
E − 3RI2 + RI3 + RI1 = 0
−E − 2RI3 + RI2 = 0,
or
I2 = 2I1
3I2 − I3 − I1 = E/R
I2 − 2I3 = E/R.
This can be solved to give I1 = E/8R, I2 = E/4R, and I1 = −3E/8R. We see that Vb − Va =
(I2 − I3 )R = 5E/8 .
(b) We’ll do this in two ways for variety. First, note that we already found the voltage between a
and b in part (a), and this is precisely the Thevenin emf, Veff = 5E/8. The Thevenin resistance
is simply the equivalent resistance between a and b. By a straightforward application of the
series and parallel rules, this is Reff = 3R/8 .
Second, suppose we short points a and b with a wire. Then by Thevenin’s theorem, the current
flowing through that wire should be I = Veff /Reff . We already know Veff from part (a). To
compute the current, we just use Kirchoff’s loop rules again; these are now as follows.
E − E − 2RI1 + RI2 = 0
E − 2RI2 + RI1 = 0
−E − RI3 = 0
Solving these equations gives I1 = E/3R, I2 = 2E/3R, and I3 = −E/R. The current through
the wire is now I2 − I3 = 5E/3R. Thus, Reff = (5E/8)/(5E/3R) = 3R/8.
4
[2] Problem 2 (Wang). A circuit containing batteries and resistors has two terminals. When an ideal
ammeter is connected between them, the reading is I1 . When a resistor R is connected between
them, the current through the resistor is I2 , in the same direction. What would be the reading V
of an ideal voltmeter connected between them?
Solution. We consider the Thevenin equivalent, i.e. the function V (I). The first piece of information
tells us that when V = 0, I = I1 . The second tells us that when V = −I2 R, then I = I2 . Thus,
0 = Veq + I1 Req , −I2 R = Veq + I2 Req .
Solving this system of equations gives

I1 I2 R I2 R
Veq = , Req = .
I2 − I1 I1 − I2
When an ideal voltmeter is connected, we have I = 0, so
I1 I2 R
V = Veq = .
I2 − I1
Note that your answer may differ by a harmless sign, which ultimately depends on your sign
01W
conventions for I1 and I2 (i.e. which terminal is A and which terminal is B).

Now we give a few problems on current flow through continuous objects. Fundamentally, all one
needs for these problems is the definition J = σE, and superposition.
Example 3
Consider two long, concentric cylindrical shells of radii a < b and length L. The volume
between the two shells is filled with material with conductivity σ(r) = k/r. What is the
resistance between the shells, and the charge density?
Solution
To find the resistance, we compute the current I when a voltage V is applied between the
shells. By symmetry, in the steady state the current density must be
I
J(r) = r̂.
2πrL
On the other hand, we also know that
b
I(b − a)
Z Z
I
V = E · dr = dr =
a 2πrLσ 2πkL
b−a
R= .
2πkL
Note that the radial electric field between the shells is constant, so
V
E(r) = r̂.
b−a
5
This means that in the steady state, there must be a nonzero charge density between the
shells. (If there weren’t, then we would have E(r) ∝ 1/r, rather than a constant.)
To find the charge density explicitly, it’s easiest to use Gauss’s law in differential form in
cylindrical coordinates. We use the form of the divergence derived in E1,
1 ∂(rEr ) 1 V ρ
∇·E= + (other terms) = =
r ∂r rb−a 0
thus showing that the charge density is proportional to 1/r. Of course, we could also get this
result by applying Gauss’s law in integral form, to concentric spheres.
[2] Problem 4 (Grad). A washer is made of a material of resistivity ρ. It has a square cross section
of length a on a side, and its outer radius is 2a. A small slit is made on one side and wires are
connected to the faces exposed.
Find the resistance of the washer. (Hint: first argue that no current flows radially.)
Solution. As suggested, we’ll first show that there is no radial current. Suppose there was. Then, if
we swap the sign of the current, the direction should flip, say from outward to inward. But flipping
the sign of the current can also be achieved by simply flipping the disk over, which does not change
the direction of the radial current. So the radial current must be zero.
Now, we split the washer into a bunch of radial rings with width dr. We see that r ranges from a
to 2a. Each little ring has resistance ρ(2πr)/(adr), and they are all effectively connected in parallel.
Thus, Z 2a
1 a a
= ρ−1 dr = ρ−1 log 2,
R a 2πr 2π
giving the answer,
2π ρ
R= .
log 2 a
[3] Problem 5 (BAUPC 1995). An electrical signal can be transferred between two metallic objects
buried in the ground, where the current passes through the Earth itself. Assume that these objects
are spheres of radius r, separated by a horizontal distance L r, and suppose both objects are
buried a depth much greater than L in the ground. If the Earth has uniform resistivity ρ, find the
approximate resistance between the terminals. (Hint: consider the superposition principle.)
6
Solution. We can consider one object at a time, and then use superposition to find the combined
effect of both. Suppose that current I comes out from one of the objects. Placing this object at the
origin, we have
I ρI
J= 2
r̂, E = r̂.
4πr 4πr2
Therefore, the potential difference between this object and where the other object would be is
Z L
ρI 1 ρI
V = 2
≈
4π r r 4πr
where we used r L. Finally, the other object takes in current I, with its J, E, and V superposing
with the first object. Thus, the total potential difference is ρI/2πr, so
V ρ
R= = .
I 2πr
[3] Problem 6 (PPP 162). A plane divides space into two halves. One half is filled with a homogeneous
conducting medium, and physicists work in the other. They mark the outline of a square of side
a on the plane and let a current I0 in and out at two of its neighboring corners. Meanwhile, the
measure the potential difference V between the two other corners.
Find the resistivity ρ of the medium.
Solution. Consider the case where there is current coming in just at A. Then, J is always radial
with respect to A, and has magnitude
I0
J · 2πr2 = I0 =⇒ J = .
2πr2
Then we have that
I0
E=
r̂
2πσr2
where σ is the conductivity of the material. Then, we have that
√
Z 2a
I0 I0 √
VD − VC = 2
dr = (1 − 1/ 2).
a 2πσr 2πσa
Similarly, for the case where the current is coming out of B. Then, again in this case we also have
that
I0 √ I0 √
VD − VC = (1/ 2 − 1) = (1 − 1/ 2).
2πσa 2πσa
7
Thus, superposing these two states, we have that

I0 ρ √
∆V = (2 − 2)
2πa
where ρ = 1/σ is the resistivity. Then,
√
2πa∆V πa(2 + 2)∆V
ρ= √ =
I0 (2 − 2) I0
[3] Problem 7 (MPPP 174). We aim to measure the resistivity of the material of a large, thin,
homogeneous square metal plate, of which only one corner is accessible. To do this, we chose points
A, B, C and D on the side edges of the plate that form the corner.
Points A and B are both 2d from the corner, whereas C and D are each a distance d from it. The
length of the plate’s sides is much greater than d, which, in turn, is much greater than the thickness
t of the plate. If a current I enters the plate at point A, and leaves it at B, then the reading on a
voltmeter connected between C and D is V . Find the resistivity ρ of the plate material.
Solution. The reason this problem is a lot harder than the previous one is that there is a boundary
condition, namely that the current density at the edges of the plate is parallel to the plate. The key
insight here is that we can do an “image current” configuration as follows to automatically satisfy
the boundary condition.
Here, the plate is now infinite, and we have a new current source and sink respectively at the
2I
reflections of A and B in O. Now, if a current 2I enters the plate, the current at a distance r is 2πrt ,
ρI ρI
so the electric field at a distance r is πrt (pointing radially outward), so the potential is πt log(r/r0 )
for some arbitrary r0 . For our purposes, set the potential to be 0 at O, so r0 = AO. Now,
ρI √ ρI
VC = (log(r0 /d) + log(r0 /3d) − 2 log(r0 / 5d)) = log(5/3).
πt πt
8
By symmetry, VC = −VD . Thus,

2ρI
V = 2VC = log(5/3),
πt
πt V
so ρ = .
2 log(5/3) I
Remark
Setups like those in the previous two problems are commonly used to measure resistivities,
but why do they use a complicated “four terminal” setup? Wouldn’t it have been easier to
just attach two terminals, send a current I through them, and measure the voltage drop
V ? The problem with this is that it also picks up the resistance R of the contacts between
the terminals and the material, along with the resistances of the wires. By having a pair of
terminals measure voltage alone, drawing negligible current, we avoid this problem.
[4] Problem 8. [A] This problem is just for fun; the techniques used here are too advanced to appear on
Olympiads. We will prove Rayleigh’s monotonicity law, which states that increasing the resistance
of any part of a resistor network increases the equivalent resistance between any two points. This
may seem obvious, but it’s actually tricky to prove. The following is the slickest way.
(a) Consider a graph of resistors, where a battery is attached across two of the vertices, fixing
their voltages. Write an expression for the total power dissipated, assuming the voltages at
each vertex are Vi and the resistances are Rij .
(b) The voltages Vi at all the other vertices are determined by Kirchoff’s rules. But suppose you
didn’t know that, or didn’t want to set up those equations. Remarkably, it turns out that
you can derive the exact same results by simply treating the voltages Vi as free to vary, and
setting them to minimize the total power dissipated! Show this result. (This is an example of
a variational principle, like the principle of least action in mechanics.)
(c) For any network of resistors, show that P = V 2 /R when V is the battery voltage applied
across two vertices, R is the equivalent resistance between them, and P is the total power
dissipated in the resistors. (This is intuitive, but it’s worth showing in detail to assist with
the next part.)
(d) By combining all of these results, prove Rayleigh’s monotonicity law.
(e) We can use Rayleigh’s monotonicity law to prove some mathematical results. Consider the
resistor network shown below, where the variables label the resistances.
By considering the resistances before and after closing the switch P Q, show that the arithmetic
mean of two numbers is at least the geometric mean.
9
(f) Consider the resistor network shown below.
By closing all the switches, show that the arithmetic mean of n numbers is at least the
harmonic mean.
Solution. (a) The power is

X (Vi − Vj )2
P = .
Rij
i<j
Here the sum over i < j counts all pairs of vertices once. If there is no direct connection
between i and j, the resistance Rij is infinite.
(b) The power is minimized when its derivative is zero, and we are free to vary all voltages except
for the two points where the battery is connected. Let Vi be one of these voltages. Then
∂P X 2(Vi − Vj )
= = 0.
∂Vi Rij
j6=i
Now compare this to how we would solve the problem using Kirchoff’s laws. The fact that
the sum of the voltage drops along a loop is zero is already accounted for, because we already
have specified the voltages at each vertex. The only new equations we would write down
would be charge conservation at each vertex,
X
Iij = 0.
j6=i
However, applying Ohm’s law, we see this is precisely the equation that power minimization
has given us!
(c) By the definition of the equivalent resistance, V = IR where I is the total current going
through the circuit. By the definition of power, the power put in by the battery is P = IV ,
since any current going through the circuit must go through the battery. By conservation of
energy, the power dissipated in the circuit is equal to the power put in by the battery. So the
power dissipated is P = IV = V 2 /R.
(d) Put a battery of voltage V across the points we are considering. By part (c) Rayleigh’s
monotonicity law is equivalent to the statement that, if we increase any of the Rij , the total
power P dissipated in the resistor network goes down.
We can account for the effect of increasing one of the Rij in two steps. First, suppose we
do so while artificially keeping all the voltages Vi constant. Then by part (a), P decreases.
10
Second, in reality the voltages quickly rearrange themselves to satisfy Kirchoff’s laws, which
we saw in part (b) is equivalent to minimizing the power. So this further rearrangement can
only further decrease P . This shows the desired result.
(e) Before closing the switch, the resistance is
a+b
Ri = .
2
After closing the switch, the resistance is
1 1 −1

2ab
Rf = 2 + = .
a b a+b
Closing the switch is equivalent to decreasing RP Q from infinity to zero, so Rf ≤ Ri by

Rayleigh’s monotonicity law. This gives
√ a+b
ab ≤
2
which is the AM-GM inequality.
(f) Before closing the switches,

1X
Ri = ai
n
i
which is the arithmetic mean. After closing the switches,

!−1
X 1
Rf = n
ai
i
which is the harmonic mean. Thus, the arithmetic mean is at least the harmonic mean.
Remark
You might think that Rayleigh’s monotonicity law is too obvious to require a proof; if you
decrease a resistance, how could the net resistance possibly go up? In fact, this kind of
non-monotonicity occurs very often! For example, Braess’s paradox is that fact that adding
more roads can slow down traffic, even when the total number of cars stays the same. A U.S.
Physics Team coach has argued that allowing more team strategies can make a basketball
team score less. For more on this subject, see the paper Paradoxical behaviour of mechanical
and electrical networks or this video.
Remark
Circuit questions can get absurdly hard, but at some point they start being more about
mathematical tricks than physics. As a result, I haven’t included any such problems here;
they tend not to appear on the USAPhO or IPhO, or in college physics, or in real life, or really
anywhere besides a few competitions. On the other hand, you might find such questions fun!
For some examples, see the Physics Cup problems 2013.6, 2017.2, 2018.1, and 2019.4.
11
2 RC Circuits
Next we’ll briefly cover RC circuits, our first exposure to a situation genuinely changing in time.
Example 4: CPhO
The capacitors in the circuit shown below were initially neutral. Then, the circuit is allowed
to reach the steady state.
After a long time, what is the charge stored on the 10 mF capacitor?
Solution
After a long time, no current flows through the capacitors, so there is effectively a single loop
in the circuit. It has a total resistance 60 Ω and a total emf 6 V, so the current is I = 0.1 A.
Using this, we can straightforwardly label the voltages everywhere on the outer loop.
To finish the problem, we need to know the voltage V0 of the central node, so we need one
more equation. That equation is charge conservation: the fact that the central part of the
circuit, containing the inner plates of the three capacitors, begins and remains uncharged.
Suppressing units, this means
66
20(26 − V0 ) + 20(7 − V0 ) + 10(0 − V0 ) = 0, V0 = V
5
from which we read off the answer,
Q = CV = 0.132 C.
12
[3] Problem 10 (Purcell 4.18). Consider the two RC circuits below.
(a) The circuit shown below contains two identical capacitors and two identical resistors, with
initial charges as shown above at left. If the switch is closed at t = 0, find the charges on the
capacitors as functions of time.
(b) Now consider the same setup with an extra resistor, as shown above at right. Find the
maximum charge that the right capacitor achieves. (Hint: the methods of M4 can be useful.)
Solution. (a) Let the two loop currents be I1 and I2 , both counterclockwise. The loop equations
are Q1 /C = I1 R and Q2 /C = I2 R. We also have Ik = −Q̇k . Thus, Qk + RC Q̇k = 0 for
k = 1, 2. Based on the initial conditions, we see then that the solutions are Q1 (t) = Q0 e−t/RC
and Q2 (t) = 0. (The simple reason Q2 (t) is zero is because the middle wire effectively shorts
out the right half of the circuit.)
(b) Again, with the same setup of variables, we get that
Q1 /C − 2I1 R + I2 R = 0
Q2 /C − 2I2 R + I1 R = 0.
This is a system of two linear differential equations, which can be solved using the methods
of M4. However, in this case we can just add and subtract the equations, giving
(Q1 + Q2 )/C − (I1 + I2 )R = 0, (Q1 − Q2 )/C − 3(I1 − I2 )R = 0.
That is, the sum of the two acts like an RC circuit with time constant RC, while the difference
acts like one with time constant 3RC. (These are the “normal modes”.) By superposing these
solutions and fitting the initial conditions, we get
Q0 −t/RC Q0 −t/RC
Q1 (t) = (e + e−t/3RC ), Q2 (t) = (e − e−t/3RC ).
2 2
We want to maximize |Q2 |, so setting the derivative to zero gives t = 23 RC log(3), so
Q0
|Q2 |max = √ .
3 3
[3] Problem 12 (Kalda). Three identical capacitors are placed in series and charged with a battery
of emf E. Once they are fully charged, the battery is removed, and simultaneously two resistors are
connected as shown.
13
Find the heat dissipated on each of the resistors after a long time.
Solution. In the beginning, the charges on the plates are EC/3, −EC/3, EC/3, −EC/3, EC/3, −EC/3.
After a long time, let the charges on the plates be q1 , −q1 , q2 , −q2 , q3 , −q3 . Note that all currents are
0 now, so we may effectively ignore the resistors and treat the wires as zero resistance. Therefore, the
potential at points connected by wires is the same, so q1 = −q2 = q3 . Also, by charge conservation
on the two disjoint pieces (q1 , −q2 , q3 and −q1 , q2 , −q3 ), we see
q1 − q2 + q3 = EC/3,
P1 2
which implies |q1 | = |q2 | = |q3 | = EC/9. The energy is 2 Q /C, so the charges dropping by a
factor of 3 means we lose 8/9 of the original total energy, so each resistor dissipates 4/9 of the
original total energy. This is
4 (EC/3)2 2 2
·3· = E C.
9 2C 27
[3] Problem 13 (Kalda). Find the time constant of the RC circuit shown below.
Solution. For the purposes of computing the time constant, it is equivalent to assume the capacitor
is already charged, then take out the battery and see how it discharges. Thus all that matters is
the resistance between the capacitor plates, which is
R2 R3 R1 R2 + R1 R3 + R2 R3
R = R1 + = ,
R2 + R3 R2 + R3
R1 R2 + R1 R3 + R2 R3
so τ = C .
R2 + R3
[3] Problem 14 (MPPP 175/176). A metal sphere of radius R has charge Q and hangs on an insulating
cord. It slowly loses charge because air has a conductivity σ. In all cases, neglect any magnetic or
radiation effects.
(a) Find the time for the charge to halve.
(b) You should have found that the time is independent of the radius R of the sphere, which follows
directly from dimensional analysis. Can you show that, in fact, it is completely independent
of the shape? (This doesn’t just follow from dimensional analysis, because the shape might
be described by dimensionless numbers, such as the eccentricity of an ellipsoid.)
14
(c) Air has a conductivity of σ ∼ 10−13 Ω−1 m−1 , while water has a conductivity of σ ∼ 10−2 Ω−1 m−1 .
About how long does the charge on an object last, if it is in air or water?
This problem generalizes USAPhO 2010, problem A2, which you can compare.
Solution. (a) We can analyze this as an RC circuit. (The circuit is completed by the “sphere at
infinity”.) The capacitance is the self-capacitance of the sphere,
C = 4π0 R.
The resistance is the resistance between the sphere and infinity. The air can be thought of as
a set of resistors in series, with each resistor being a spherical shell of air. Then
1 ∞ dr
Z Z
1
Req = dR = = .
σ R 4πr2 4πσR
This gives a time constant of

0
τ = RC = .
σ
Therefore, the time is
0
t= log 2.
σ
(b) Of course, dimensional analysis doesn’t work, because there might be dimensionless parameters
describing a general shape (e.g. the eccentricity of an ellipsoid). Instead we use the following
more general argument. We note that
I I
I = J · dS, ΦE = E · dS
over any surface completely enclosing the object. The right-hand sides are related by J = σE,
and Gauss’s law gives ΦE = Q/0 . Combining these gives
σ
Q̇ = − Q
0
so the charge decreases exponentially with timescale 0 /σ, completely independently of the
shape. (Of course, the sphere is still special, because with the sphere we are guaranteed there
are no magnetism or radiation effects (why?). For a general shape, we have to assume these
effects are negligible.)
(c) The relevant timescale is 0 /σ. Thus we find t ∼ 10 s for air, and t ∼ 1 ns for water.
[5] Problem 15. 01h IPhO 1993, problem 1. A really neat question with real-world relevance.
[5] Problem 16. 01h IPhO 2007, problem “orange”. A combination of mechanics and RC circuits.
3 Computing Magnetic Fields
15
Idea 4
The Biot–Savart law is
ds × r
I
µ0 I
B= .
4π r3
As a consequence, we have Ampere’s law,
I
B · ds = µ0 I, ∇ × B = µ0 J
as well as Gauss’s law for magnetism,

I
B · dS = 0, ∇ · B = 0.
Idea 5
The force on a stationary wire carrying current I in a magnetic field B is
Z
F = I ds × B.
The energy of a magnetic field is

Z
1
U= B 2 dV.
2µ0
The magnetic dipole moment of a planar current loop of area A and current I is m = IA,
with m directed perpendicular to the loop by the right-hand rule.
You should have already seen basic examples of using the Biot–Savart law in Halliday and Resnick,
such as the field of a circular ring of current on its axis. We’ll start with some problems that are
similarly straightforward, but more technically complex.
[3] Problem 17 (Purcell 6.11). A spherical shell with radius R and uniform surface charge density σ
spins with angular frequency ω about a diameter.
(a) Find the magnetic field at the center.
(b) Find the magnetic dipole moment of the sphere.
(c) Sketch the magnetic field.

Solution. (a) First, we find the field due to a ring of counterclockwise current I with radius a in
the z = 0 plane at a point directly above the center at some height z. Using the Biot–Savart
law, we see that
a
µ0 I 2πa √a2 +z 2 µ0 I a2
B= 2 2
ẑ = ẑ.
4π a + z 2 (a2 + z 2 )3/2
Let us work in spherical coordinates with the axis of rotation being the z axis. Then, at angle
θ, we essentially have a ring of charge of radius a = R sin θ, z = −R cos θ, and
dI = σR(dθ)ω(R sin θ) = σR2 ω sin θ dθ.
16
Therefore,
µ0 σR2 ω sin θ dθ R2 sin2 θ 1
dB = 3
ẑ = µ0 σωR sin3 θ dθẑ.
2 R 2
Integrating from 0 to π to obtain the full field,
Z π
1 2
B = µ0 σωRẑ sin3 θ dθ = µ0 σωRẑ.
2 0 3
(b) The moment of a slice is

dm = ẑπ(R sin θ)2 ωσR2 sin θdθ.
Integrating this gives
4
m = πωσR4 ẑ.
3
(c) The field is as shown below.
This is an example of a magnetic dipole field. Note that it looks very similar to an electric
dipole field far away, but inside the dipole the field points along the dipole moment, rather
than against it. This is necessary to make the magnetic field lines form closed loops, and
hence obey Gauss’s law for magnetism.
One unexpected feature is that the field is perfectly uniform inside, which is very difficult to
show starting from the Biot–Savart law. It’s easier to use a tool called the vector potential
(which is the analogue of the electric potential, but for magnetic fields), solve for the vector
potential, and take its curl to get the magnetic field. Along the way, it turns out that the
vector potential integral simplifies dramatically, leading to the uniform field. Ultimately, it’s
the same simplification as occurs for the electric potential when proving the shell theorem. If
you want to see this full story, see section 5.4 of Griffiths.
[2] Problem 18 (Purcell 6.12). A ring with radius R carries a current I. Show that the magnetic
field due to the ring, at a point in the plane of the ring, a distance r from the center, is given by
µ0 I π (R − r cos θ)R dθ
Z
B= .
2π 0 (r2 + R2 − 2rR cos θ)3/2
17
In the r R limit, show that

µ0 m
B≈
4π r3
where m = IA is the magnetic dipole moment of the ring.
Solution. Let the ring be centered at the origin, and let the field point be ax̂, and say we are at
an angle θ. Then, dl = (−R sin θ) x̂ + (R cos θ) ŷ, and
r = a x̂ − R(cos θ x̂ + sin θ ŷ) = (a − R cos θ) x̂ + (−R sin θ) ŷ.

√
Note that r = |r| = a2 + R2 − 2aR cos θ. Therefore, by the Biot–Savart law,
µ0 I R(R − a cos θ)ẑ dθ

dB = ,
4π (a2 + R2 − 2aR cos θ)3/2
so integrating from 0 to 2π and noting that θ and −θ contribute the same, we arrive at the desired
result. Now, to take the r R limit cleanly and consistently, it’s best to nondimensionalize
everything. Defining x = R/r 1, we can pull dimensionful factors out of the integral to get
µ0 I rR π x − cos θ
Z
B= dθ.
2π r3 0 (1 + x2 − 2x cos θ)3/2
Now, it’s not immediately obvious to what order in x we should expand in. If we already know the
answer is proportional to 1/r3 , then we can see the answer must be first order in x. But if we didn’t
know that, we could expand to zeroth order, giving
µ0 I R π
Z
B= (− cos θ) dθ = 0.
2π r2 0
The fact that the answer vanishes means we need to go to higher order to find the true answer. At
first order, applying the binomial theorem, the integrand is
(x − cos θ)(1 + x2 − 2x cos θ)−3/2 ≈ (x − cos θ)(1 + 3x cos θ) ≈ − cos θ + x(1 − 3 cos2 θ)
where we threw away higher order terms throughout. Then
µ0 I R 2 π
Z
B= (1 − 3 cos2 θ) dθ.
2π r3 0
Using the fact that cos2 θ averages to 1/2 over a cycle, the integral is −π/2, giving
µ0 I πR2 µ0 m
B= 3
=
4π r 4π r3
as desired.
[3] Problem 19 (Purcell 6.14). Consider a square loop with current I and side length a centered at
the origin. Show that the magnetic field at r, where r a and r is parallel to one of the square’s
sides, is B ≈ (µ0 /4π)(m/r3 ) as in the previous problem. Be careful with factors of 2!
18
Solution. This calculation is a bit subtle. It is tempting to ignore the vertical portions of the
square, because the current is almost parallel to r, so ds × r is small; more precisely, it is suppressed
by a power of a/r. The horizontal portions each give much larger contributions. However, the
horizontal sides give opposite contributions and nearly cancel out, so it turns out that the vertical
portions contribute significantly to the final answer, which contains one overall factor of a/r.
First consider the vertical segments. We get a factor of (a/2)/r from the sin θ factor in the cross
product. Similarly, a appears in the Biot–Savart integral in the denominator; however, its effect
here would give higher-order terms in a/r, which we don’t want. So the vertical segments each
contribute
a/2
µ0 I a r
B=−
4π r2
and hence the total contribution from them is
µ0 I a2
B=− .
4π r3
The horizontal ones sides contribute, to the same order in a/r,
µ0 I a µ0 I a
, −
4π (r − a)2 4π (r + a)2
where here we had to include a in the denominator, because there isn’t an a/r suppression factor
in the numerator. Now adding the two, at leading order in a/r we get
µ0 I 2a2
B= .
4π r3
Adding the horizontal and vertical contributions yields the correct result. If one forgets the vertical
contributions, one gets an answer that is two times too big.
Idea 6
The results you have found above, for the fields far from currents, are special cases of the
general magnetic dipole field: far from a magnetic dipole with magnetic moment m, its
magnetic field is just the same as the electric field of an electric dipole,
µ0 m µ0
B(r) = 3
(2 cos θ r̂ + sin θ θ̂) = (3(m · r̂)r̂ − m).
4πr 4πr3
As with the electric dipole field, you don’t need to memorize this result, but you should
remember that it’s proportional to the dipole moment, falls off as 1/r3 , and be able to sketch
it. Of course, static electric and magnetic fields behave differently; when you get inside an
electric dipole the field reverses direction, but this isn’t true for a magnetic dipole.
We now give a few arguments for computing fields using symmetry.
19
Example 5: PPP 31
An electrically charged conducting sphere “pulses” radially, i.e. its radius changes periodically
with a fixed amplitude. What is the net pattern of radiation from the sphere?
Solution
There is no radiation. By spherical symmetry, the magnetic field can only point radially.
But then this would produce a magnetic flux through a Gaussian sphere centered around
the pulsing sphere, which would violate Gauss’s law for magnetism. So there is no magnetic
field at all, and since radiation always needs both electric and magnetic fields (as you’ll see
in E7), there is no radiation at all. In fact, outside the sphere the electric field is always
exactly equal to Q/4π0 r2 , in accordance with Coulomb’s law.
Example 6
Find the magnetic field of a very long cylindrical solenoid, of radius R and n turns per unit
length, carrying current I.
Solution
Orient the solenoid along the vertical direction and use cylindrical coordinates. By symmetry,
the field must be independent of z. Now consider the radial component of the magnetic field
Br . Turning the solenoid upside-down is equivalent to reversing the current. But the former
does not flip Br while the latter does, so we must have Br = 0.
Now, by rotational symmetry, the tangential component Bφ must be uniform. But then
Ampere’s law on any circular loop gives Bφ (2πr) = 0, so we must have Bφ = 0 as well.
The only thing left to consider is Bz . By applying Ampere’s law to small vertical rectangles, we
see that Bz is constant unless that rectangle crosses the surface of the solenoid. Furthermore,
Bz must be zero far from the solenoid, so it must be zero everywhere outside the solenoid.
Now, for a rectangle of height h that does cross the surface, Ampere’s law gives
I
B · ds = Bzin h = µ0 Ienc = µ0 nIh
which tells us that Bzin = µ0 nI.
Remark
The above analysis is not quite how real solenoids behave for several reasons. First, we
didn’t account for the discreteness of the wires. We just treated them as forming a uniform
current per length K = nI, which is how we wrote Ienc = nIh. This is valid when you don’t
care about looking too close to the wires themselves.
The fact that solenoids are made by winding real wires means there is another contribution
20
to the current, even in the limit n → ∞. The wires are wound with a small slope, since a net
current I still has to move along the solenoid. Another way of saying this is that the current
per length along the solenoid surface is K = nI θ̂ + (I/2πR)ẑ. This causes a tangential
magnetic field Bφ = µ0 I/2πr outside the solenoid.
Another factor is that a real solenoid isn’t infinitely long, and end effects are important.
You’ll analyze these in a slick way in problem 22.
[2] Problem 21. A toroidal solenoid is created by wrapping N turns of wire around a torus with a
rectangular cross section. The height of the torus is h, and the inner and outer radii are a and b.
(a) In the ideal case, the magnetic field vanishes everywhere outside the toroid, and is purely
tangential inside the toroid. Find the magnetic field inside the toroid.
(b) There is another small contribution to the magnetic field due to the winding effect mentioned
above. Roughly what does the resulting extra magnetic field look like? If you didn’t want
this additional field, how would you design the solenoid to get rid of it?
Solution. (a) Applying Ampere’s law on a circular loop gives B(r)(2πr) = µ0 N I, so
µ0 N I
B(r) = .
2πr
(b) Note that the twisting of the wire adds an effective small current in the tangential direction.
This looks like a current loop, so, e.g. it produces a magnetic field pointing vertically through
the toroid’s hole. We can remove it by using a bunch of current loops instead of a single
winding wire, or by using counterwinding: after winding the wire around the toroid once
clockwise, wind it around again counterclockwise.
[3] Problem 22 (Purcell 6.63). A number of simple facts about the fields of solenoids can be found
by using superposition. The idea is that two solenoids of the same diameter, and length L, if joined
end to end, make a solenoid of length 2L. Two semi-infinite solenoids butted together make an
infinite solenoid, and so on.
21
Prove the following facts.

(a) In the finite-length solenoid shown at left above, the magnetic field on the axis at the point
P2 at one end is approximately half the field at the point P1 in the center. (Is it slightly more
than half, or slightly less than half?)
(b) In the semi-infinite solenoid shown at right above, the field line FGH, which passes through
the very end of the winding, is a straight line from G out to infinity.
(c) The flux through the end face of the semi-infinite solenoid is half the flux through the coil at
a large distance back in the interior.
(d) Any field line that is a distance r0 from
√ the axis far back
√ in the interior of the coil exits from
the end of the coil at a radius r1 = 2r0 , assuming 2r0 is less than the solenoid radius.
Part (c) tells us that even in an ideal solenoid, we always lose at least half the flux to fringing!
Solution. (a) Glue two of them together. Now, 2B2 = B1 + for some small . Then, we have
that B2 = B11 + /2, so it is slightly more than half.
(b) Let G0 be the reflection of G in the axis. Say the field line GH comes out at an angle θ. Then,
at G0 , it also comes out with an angle θ. Now, making a copy and rotating 180◦ and flipping
the current direction, the field at G becomes one pointing at θ above the horizontal (coming
from G at the original), and one at angle π − θ to the horizontal (coming from G0 in the copy).
Therefore, the field there would be non-zero right outside the solenoid, unless θ = 0, in which
case the fields cancel.
(c) Do the same procedure as in (a), and the flux at the glue points gets doubled to what it was
originally. However, now we have an infinite solenoid, so double the flux through the end is
equal to the flux in the middle.
22
(d) Note that (c) holds even if we take a constant disk of radius a as our surface to take the flux
over. Note that the flux through the disk at the edge with radius r is the same as at the
middle with radius r0 (same field lines). However, if we draw a disk of radius r at the middle,
it will have twice the flux as it did at the top, or twice the flux as with r0 . However, here in
the middle, the magnetic√ field is essentially constant, so the areas must be twice each other,
so πr2 = 2πr02 , or r = 2r0 .
[3] Problem 23 (MPPP 160). Two infinite parallel wires, a distance d apart, carry electric currents
with equal magnitudes but opposite directions. In this problem, we find the shape of the magnetic
field lines using a neat trick.
(a) Argue that if we rotated B by 90◦ at each point, it would produce a valid electrostatic field
E. (Hint: consider what happens when we rotate the B field of a single wire, first.)
(b) Argue that the field lines of B are the same as the equipotentials of this artificial E, and use
this to find the field lines.
This is a common trick used when working with vortices in two dimensions, where it converts
vortices to sources and sinks.
Solution. (a) First, we can get the intuition using a single wire. In this case,
µ0 I
B= θ̂
2πr
in cylindrical coordinates. Upon a 90◦ rotation, θ̂ turns into r̂, giving

µ0 I
E= r̂.
2πr
Of course, this is a purely formal idea, since electric and magnetic fields don’t even have the
same units. Now note that this is precisely the electric field of an infinite, uniform line charge,
if it had a linear charge density λ = µ0 0 I. So it’s a valid electric field, and the same would
be true if we had started from two wires, by superposition.
(b) The field lines of B are always parallel to B. Now, this artificial E is always perpendicular to
B, and equipotentials are always perpendicular to E, so the equipotentials follow the magnetic
field lines.
On the other hand, we know precisely what the potential is in this problem. By integrating
the 1/r field, the potential is proportional to log r, so
V (r) ∝ log(r+ ) − log(r− ) = log(r+ /r− )
where r+ and r− are the distances to the two wires. So the equipotentials have constant
r+ /r− . We’ve already found, when investigating the method of images for spheres in E2, that
this implies the equipotentials are circles, specifically circles of Appolonius. So the magnetic
field lines are circles!
[2] Problem 24 (IPhO 1996). Two straight, long conductors C+ and C− , insulated from each other,
carry current I in the positive and the negative ẑ direction respectively. The cross sections of the
conductors are circles of diameter D in the xy plane, with a distance D/2 between the centers.
23
The current in each conductor is uniformly distributed. Find the magnetic field in the space between
the conductors.
Solution. See page 12 of the official solutions here.
[3] Problem 25 (MPPP 157). A regular tetrahedron is made of a wire with constant resistance per
unit length. A current I is sent into one vertex and removed from another vertex, as shown.
Find the magnetic field at the center of the tetrahedron.
Solution. By symmetry C and D are at the same potential, so IDC = 0. Then the current from A
to B just splits up into three branches, which have resistances RACB = RADB = 2RAB . Therefore,
the currents are
1 1
IAB = I, IAC = IAD = ICB = IDB = I.
2 4
−−→
The field at O due to the current along AD is directed along the vector CB. Similarly, the magnetic
−−→
field due to the current along AC is directed along BD, and so on. By repeating this reasoning for
all five contributions, we find that the magnetic field at O is proportional to
−−→ −−→ −−→ −−→ −→ −−→ −−→ −−→
2DC + BD + CB + AD + CA = 2DC + CD + CD = 0
01h
so there is no field at O.
[5] Problem 26. APhO 2013, problem 1. A neat question on a cylindrical RC circuit that uses
many of the techniques we’ve covered so far.
24
Electromagnetism IV: Lorentz Force

The problems here mostly use material covered in previous problem sets, though chapter 5 of Purcell
covers relativistic field transformations. For further interesting physical examples, see chapter II-29
of the Feynman lectures. There is a total of 84 points.
1 Electrostatic Forces
Idea 1: Lorentz Force
A charge q in an electromagnetic field experiences the force
F = q(E + v × B).
In particular, a stationary wire carrying current I in a magnetic field experiences the force
Z
F = I ds × B.
Example 1: PPP 183
A small charged bead can slide on a circular, frictionless insulating ring. A point-like electric
dipole is fixed at the center of the circle with the dipole’s axis lying in the plane of the circle.
Initially the bead is in the plane of symmetry of the dipole, as shown.
Ignoring gravity, how does the bead move after it is released? How would the bead move if
the ring weren’t there?
Solution
Set up spherical coordinates so that the dipole is in the ẑ direction. Then
kp cos θ
V (r, θ) = .
r2
Since the ring fixes r, the potential on the ring is just proportional to cos θ, which is in turn
proportional to z. But a potential linear in z is equivalent to a uniform downward field, so
the bead oscillates like the mass of a pendulum, with amplitude π/2.
1
The answer remains the same when the ring is removed! Conservation of energy states that
kqp cos θ 1
+ mv 2 = 0.
r2 2
Let N be the normal force. Then accounting for radial forces gives
∂V v2
N +q = .
∂r r
However, plugging in our conservation of energy result for v 2 shows that N = 0, so the ring
doesn’t actually do anything, and it may be removed without effect.
Example 2
A parallel plate capacitor with separation d and area A is attached to a battery of voltage
V . Suppose the plates are moved together with speed v. Verify that energy is conserved.
Solution
The capacitance is C = A0 /d. The power supplied by the battery is
dQ dC
Pbatt = IV = V =V2 .
dt dt
On the other hand, the rate of change of the energy stored in the battery is

d 1 2 1 dC
Pcap = CV = V2 .
dt 2 2 dt
At first glance, there seems to be a problem. But then we remember that there is an attractive
force between the plates, so the plates do work on whatever is moving them together,
QE QV 1 v 1 dC
Pmech = F v = v= v = CV 2 = V 2 .
2 2d 2 d 2 dt
where E is the electric field inside the capacitor. Thus, Pbatt = Pcap + Pmech as required.
[2] Problem 1 (PPP 193). Two positrons are at opposite corners of a square of side a. The other two
corners of the square are occupied by protons. All particles have charge q, and the proton mass M
is much larger than the positron mass m. Find the approximate speeds of the particles much later.
Solution. The idea is that since the positrons are so light, they will be extremely far away before
the protons hardly move. Let v1 be their final speed. Then, energy conservation tells us that
kq 2 kq 2

2 1 2
4+ √ ≈ √ +2 mv .
a 2 2a 2 1
Solving for v1 yields r
kq 2 √
v1 = (4 + 1/ 2).
am
2
The speed of the protons can be calculated by assuming that the positrons didn’t even exist, since
by the time the protons move appreciably, the positrons are long gone away to a very far distance.
Therefore, energy conservation again tells us that the final speed v2 of the protons obeys
s
2 kq 2

kq 1
√ ≈2 M v22 , v2 = √ .
2a 2 2 aM
[3] Problem 2 (PPP 114). A small positively charged ball of mass m is suspended by an insulating
thread of negligible mass. Another positively charged small ball is moved very slowly from a large
distance until it is in the original position of the first ball. As a result, the first ball rises by h.
How much work has been done?
Solution. Let r be the final separation of the balls, and let L be the length of the string. By basic
trigonometry,
h
= sin θ
r
where θ is half the angle of the string to the vertical. Letting the balls have charges q and Q and
balancing forces, we have
kqQ
cos θ = mg sin 2θ = 2mg sin θ cos θ
r2
kqQ
= 2mgr sin θ = 2mgh.
r
Furthermore, one of the balls has been raised by a height h during the process. Thus, the total work
done is 3mgh . It’s neat how almost all the dimensionful quantities drop out in the final answer!
[3] Problem 3 (PPP 71). Two small beads slide without friction, one on each of two long horizontal
parallel fixed rods a distance d apart.
The masses of the beads are m and M and they carry charges q and Q. Initially, the larger mass
M is at rest and the other one is far away approaching it at a speed v0 . For what values of v0 does
the smaller bead ever get to the right of the larger bead?
Solution. When v0 is just large enough for the small bead to get to the right of the big bead, when
both beads end up side-to-side, the small bead’s velocity should be just a bit greater than that of
3
the big bead for it to get past. This means the minimum possible value vm of v0 should be just
large enough to provide enough energy so that both beads can move together at some velocity v,
1 kqQ 1
mv 2 = + (m + M )v 2 .
2 m d 2
Since the total horizontal momentum is conserved,
mvm = (m + M )v.
Thus, we have r
m2

1 2 kqQ 2kqQ m + M
m− vm = , vm = .
2 m+M d d mM
[2] Problem 4 (PPP 192). Classically, a conductor is made of nuclei of positive charge fixed in place,
and electrons that are free to move.
(a) Consider a solid conductor in a gravitational field g. Argue that the electric field inside the
conductor is not zero; find out what it is.
(b) Now suppose a positron is placed at the center of a hollow spherical conductor in a gravitational
field g. Find its initial acceleration.
Solution. (a) We usually argue that the electric field has to vanish to keep the electrons from
accelerating. In this case, the electric field has to be nonzero, because otherwise the electrons
will fall down. Specifically, there is a downward electric field of magnitude mg/e, where e > 0
is the magnitude of the electron charge and m is the electron mass.
You might wonder how the forces on the positive ions are balanced, since they experience both
downward gravitational and electrical forces. The answer is that they’re locked into a lattice,
and held up by internal stresses within the lattice. These ultimately come from whatever is
keeping the conductor as a whole from falling, such as a normal force from the ground.
(b) The electric field found in part (a) also exists in a cavity, which one can argue by uniqueness
or by the fact that the electric field is conservative. So the positron has an initial downward
acceleration of 2g. (We had to specify the positron was at the center, or else it would have an
additional acceleration due to charge induction, which we could compute using image charges.)
[3] Problem 5. 01m USAPhO 2008, problem B2. You may ignore part (c), which was removed in the
01^
final version of the exam, though you can also do it for extra practice.
[5] Problem 7. 01h IPhO 2004, problem 1. A nice question on the dynamics of a multi-part system.
2 The Lorentz Force
4
Idea 2
The Lorentz force
dp
F= = q(E + v × B).
dt
In some problems below, you might worry that this relation is modified due to relativistic
effects, but it isn’t. The Lorentz force expression is always correct, and it’s all you need for
almost all the problems in this section.
What does change in relativity is the relation between p and v,

1
p = γmv, γ=p .
1 − v 2 /c2
The relativistic energy is also modified to

1
E = γmc2 = mc2 + mv 2 + . . . .
2
We will return to this subject in more detail in R2.
Example 3: Kalda 163
A beam of electrons, of mass m and charge q, is emitted with a speed v almost parallel to
a uniform magnetic field B. The initial velocities of the electrons have an angular spread
of α 1, but after a distance L the electrons converge again. Neglecting the interaction
between the electrons, what is L?
Solution
Consider an electron initially traveling at an angle α to the magnetic field. This electron has
a speed vk = v cos α ≈ v parallel to the field, and a speed v⊥ v sin α ≈ vα perpendicular to
the field. The component vk always stays the same, while v⊥ rotates, so the electron spirals
along the field lines.
The acceleration of the electron is

F qv⊥ B
a⊥ = = .
m m
The perpendicular velocity component rotates through a circle in velocity space of circumfer-
ence 2πv⊥ . After one such circle, the total perpendicular displacement is zero, so the beam
refocuses. Thus we have
2πv⊥ 2πmv
L= vk ≈ .
a qB
In other words, this setup acts like a magnetic “lens”.
[3] Problem 8 (Griffiths 5.17). A large parallel plate capacitor with uniform surface charge σ on the
upper plate and −σ on the lower is moving with a constant speed v as shown.
5
(a) Find the magnetic field between the plates and also above and below them.
(b) Find the magnetic force per unit area on the upper plate, including its direction.
(c) What happens to the net force between the plates in the limit v → c? Explain your result
using some basic ideas from special relativity.
Solution. (a) Let x̂ be the direction of the velocity. Let ŷ point into the page, and let ẑ point
up. The magnetic field due to the top plane is − 12 µ0 σvŷ above the above plane and 12 µ0 σvŷ
below the above plane. Similarly for the lower plane, we have 12 µ0 σvŷ above and − 12 µ0 σvŷ
below. Thus, the magnetic field is µ0 σvŷ between the plates, and zero outside.
(b) The force per unit area (i.e. pressure) is σvx̂ × 12 µ0 σvŷ = 12 µ0 σ 2 v 2 ẑ. The factor of 1/2 is there
since it only feels a force due to the contribution of the other plate; this is the essentially the
same 1/2 as we found for the pressure on a conductor in E1.
(c) The forces balance when v = c,

1 1 2
µ0 σ 2 c2 = σ
2 20
because c2 = 1/0 µ0 . Thus, as v increases to approach c, the attractive force between the
plates gets smaller and smaller. If we invoke special relativity, this makes perfect sense. In
the rest frame of the plates, there is only the attractive electrostatic force, so the plates move
together. This implies that in the lab frame, the plates also have to move together, so the
force must be attractive. But for high v, there’s a lot of time dilation, so the plates move
together more slowly. (This is partially due to the force decreasing, as derived here, and
partially due to the higher “transverse mass” due to the plates’ relativistic momentum, as
we’ll see in R2.)
[3] Problem 9. EFPhO 2012, problem 7. An elegant Lorentz force problem with wires. (If you enjoy
this problem, consider looking at IdPhO 2020, problem 1B, which has a similar setup but requires
three-dimensional reasoning. The official solutions are here.)
[4] Problem 10 (Purcell 6.35/INPhO 2008.6). Consider the arrangement shown below.
6
The force between capacitor plates is balanced against the force between parallel wires carrying
current in the same direction. A voltage alternating sinusoidally with angular frequency ω is applied
to the parallel-plate capacitor C1 and also to the capacitor C2 , and the current is equal to the
current through the rings. Assume that s a and h b.
Suppose the weights of both sides are adjusted to balance without any applied voltage, and C2
is adjusted so that the time-averaged downward forces on both sides are equal. Show that
r
1 √ b C2
√ = 2π aω .
µ 0 0 h C1
The left-hand side is equal to c, as we’ll show in E7, so this setup measures the speed of light.
Solution. It can be a little tricky to read the diagram. The key point is that the triangles are
conductors. They represent the fulcrum of a see-saw, but they also allow the voltage to be applied
across the capacitors on the left and right. The charge buildup on the capacitors on the left causes
them to attract, while the current flowing through the circular wires on the right causes them to
attract as well.
Let a current I flow on in the right-hand side. Since h b, the magnetic field created by the
bottom circular loop at a point on the top circular loop is approximately the same as that created
by an infinite wire, B = µ0 I/2πh. Thus, the force between the wires is
µ0 I µ0 bI 2
F = (2πbI) = .
2πh h
This force oscillates over time. The charge on the capacitor C2 is
Q2 (t) = C2 E0 cos(ωt)
so that
C22 E02 ω 2
hI 2 (t)i = C22 E02 ω 2 hsin2 (ωt)i = .
2
7
Thus, the average force on the right is

µ0 C22 E02 ω 2 b
hF i = .
2h
On the left-hand side, the force between the plates is, by a result in E1,
1
F = 0 E 2 (πa2 )
2
where E is the electric field inside the plates, and we have
1 E02
hE 2 (t)i = hE(t)i = .
s2 2s2
Combining these results and eliminating s, since it doesn’t appear in the final result,
C12 E02
hF i = .
4πa2 0
Equating the averaged forces gives
µ0 C22 ω 2 b C12
= ,
h 2πa2 0
which is equivalent to the desired result.
[3] Problem 11. An electron beam is accelerated from rest by applying an electric field E for a time
t, and subsequently guided by magnetic fields. These magnetic fields are produced with a series of
coils, which carry currents Ii .
Now suppose the apparatus is repurposed to shoot proton beams. Suppose a proton beam is
accelerated from rest by applying an electric field E for a time t (in the opposite direction). Let an
electron have mass m and a proton have mass M .
(a) Find the currents Ii needed so that the proton follows the same trajectory the electron did,
assuming V is small enough that both the electron and proton are nonrelativistic.
(b) How does the answer change if relativistic corrections are accounted for?
Solution. (a) The electron and proton have the same momentum p, and we have
dp
= qvB ∼ qvI
dt
since B ∝ I. Now, the magnetic field can only rotate the particle’s momentum. Suppose at
some moment it is curving in a trajectory with radius of curvature r, and speed v. Then it
has instantaneous angular velocity ω = v/r along the circle tangent to its trajectory, and
dp
= ωp
dt
in magnitude. Hence we have
v
qvI ∼ p
r
and since r is the same for both the electron and proton, we suppress it to give
p
I∼ .
q
In other words, we have I ∝ 1/q, so all that happens is that the currents should change sign
to accommodate the proton.
8
(b) Every step in part (a) still works with relativity accounted for (the change of p = mv to
p = γmv doesn’t matter, because we never used p = mv), so the answer is the same: we just
01h
flip the currents.
[5] Problem 12. IPhO 2000, problem 2. A solid question on the Lorentz force with real-world
relevance. Requires a little relativity, namely the expressions for relativistic momentum/energy.
01T
Solution. The official files are a mess; the solutions to this particular problem are here and here.
[4] Problem 13. IPhO 1996, problem 2. An elegant problem on particles in a magnetic field.
3 Permanent Magnets
[3] Problem 14. Consider a current loop I in the xy plane in a constant magnetic field B.
(a) Show that the net force on the loop is zero.
(b) Show that the torque is

τ =m×B
where the magnetic moment is
m = IAẑ
where A is the area of the loop. For simplicity, you can show this in the case where the current
loop is a square of side length L, whose sides are aligned with the x and y axes. (The proof
for a general loop shape requires some vector calculus, but you can attempt it for a challenge.
You’ll need the double cross product identity, a × (b × c) + b × (c × a) + c × (a × b) = 0.)
Solution. (a) We see that

I I
F=I ds × B = I ds × B = 0,
as desired.
(b) The magnetic moment of the square is
m = IL2 ẑ.
The torque on a side of the square is

Z Z
τ = r × dF = I s × (ds × B).
In particular, it’s useful to pair the two sides parallel to the x axis. These have opposite
currents and differ only by a translation ∆r = Lŷ, so adding their contributions gives a torque
Z L Z L
τ = −I (Lŷ) × (x̂ dx × B) = −IL(ŷ × (x̂ × B)) dx = −IL2 (ŷ × (x̂ × B)).
0 0
Similarly, the torques due to the other two sides add up to
τ = IL2 (x̂ × (ŷ × B)).
9
Manually performing the cross products, we have
−ŷ × (x̂ × B) = −By x̂, x̂ × (ŷ × B) = Bx ŷ.
Adding these together gives exactly the desired result, τ = m × B.

For completeness, we display a fully general, vector calculus solution, valid for any loop shape.
We note that along the full, closed loop, the fundamental theorem of calculus implies
I
d(s × (s × B)) = 0
simply because the closed loop integral of d(anything) is the net change in (anything) along
the loop, which is zero. Expanding with the product rule gives
I
ds × (s × B) + s × (ds × B) = 0.
Using these results and the double cross product identity, the torque is
I
τ = I s × (ds × B)
I I
= −I ds × (B × s) − I B × (s × ds)
I
= −τ − IB × s × ds .
Now, s × ds = 2 dA, because as s moves a little along the loop it sweeps out a small triangle
of area. Thus we have 2τ = 2IB × A, giving the result.
Idea 3
The force on a small current loop is
F = (m · ∇)B
but this requires some tricky vector calculus to derive, shown here. This expression, and the
torque expression found in problem 14, can be found by differentiating the potential energy
U = −m · B.
All of these results also hold for electric dipoles, if we replace m with p and B with E.
Remark
The expression for the potential energy above is notoriously subtle. Here’s the problem: we
know the Lorentz force on a charge is qv × B, which means magnetic fields never do work.
So how can they be associated with a nonzero potential energy?
There are two levels of explanation. First, suppose the magnetic dipole is made of charges
moving in a loop. When such a current loop is placed in a magnetic field, and moved
10
or rotated, mechanical work can be done on the loop. But at the same time, there will
be an induced emf in the loop, which speeds up or slows down the current. The work
done by these two effects perfectly cancels, so that the energy of the loop stays constant.
For this kind of dipole, the expression for U doesn’t indicate the total energy, but only
the “mechanical” potential energy. However, differentiating it still gives the right forces
and torques. (Some further discussion of this point is in chapter II-15 of the Feynman lectures.)
On the other hand, the magnetic dipole moment of a common bar magnet doesn’t come from
charges moving in a loop! Instead, it comes from the intrinsic magnetic dipole moments of
the unpaired electrons in the magnet. These kinds of dipole moments aren’t composed of
any moving subcomponents; they are an elementary and immutable property of the electron,
like its mass or charge. In these cases, U = −m · B really is the total energy, and the
magnetic field can do work. You won’t hear much about these elementary dipole moments in
introductory books, because they can only be properly understood by combining relativity
and quantum mechanics, but they’re responsible for most magnetic phenomena.
Example 4
If a magnet is held over a table, it can pick up a paper clip. If the paper clip is removed, it
can pick up another paper clip just as well, and this process can seemingly continue forever
without any effect on the magnet. Since the magnet does work on each paper clip, doesn’t
this mean a permanent magnet is an infinite energy source?
Solution
This is the kind of question that makes magnets feel so mysterious. They’re basically the
only everyday example of a long range force besides gravity (in fact, Kepler once thought
the Sun acted on the planets like a giant magnet), and as such they’ve inspired countless
attempts at perpetual motion machines. Many people have spent years of their lives trying
to get elaborations of this example to work.
To see why this doesn’t work for a bar magnet, just replace the word “magnet” with “charge”.
It’s true that a positive charge can attract a negative charge to it. And if the negative
charge is then removed, the positive charge can then attract another negative charge to
it. But conservation of energy isn’t violated, because the force from the positive charge
is conservative: the work it does on the negative charge to draw it close is precisely the
opposite of the work an external agent needs to do to pull it away. The force of a magnet on
a paper clip is also conservative.
It’s also interesting to consider a slightly different case. Unlike a bar magnet, an electromagnet
(i.e. a magnet created by moving current in a loop) can be turned on and off with the flick
of a switch. Therefore, we might suspect that the following is a perpetual motion machine:
1. Turn on the electromagnet, which costs energy E0 .
2. Use it to lift a paper clip, increasing its potential energy by mgh.
11
3. Turn off the electromagnet, which costs energy E0 , while holding the paper clip.
4. Move the paper clip away; we’ve managed to raise it higher for free.
To see the problem, note that the attractive force between the magnet and paper clip arises
because the magnet induces a magnetic dipole moment in the paper clip, leading to a (m·∇)B
force. As the paper clip moves toward the magnet, its own dipole moment causes a changing
magnetic flux through the electromagnet, and thus an emf against the current. Therefore, it
costs extra energy to keep the current in the electromagnet steady. Since the qv × B Lorentz
force doesn’t do work, that energy must be precisely mgh, so nothing comes for free.
Remark
A compass needle is essentially a small magnetic dipole, whose dipole moment points towards
the end painted red. We can also approximate the Earth’s magnetic field as a dipole field.
Since the tangential component of this dipole field points north, the red end of the compass
points towards the geographic north pole, which is the Earth’s magnetic south pole.
By the way, a cheap compass calibrated to work in America or Europe won’t work well in
Australia. The reason is that the Earth’s magnetic field also has a radial component, which
acts to tip the compass needle up or down. The needle needs to be appropriately weighted
to stay horizontal, so that it can freely rotate, but the side that needs to be weighted differs
between the hemispheres.
[3] Problem 15 (Griffiths 6.23). A familiar toy consists of donut-shaped permanent magnets which
slide frictionlessly on a vertical rod.
12
Treat the magnets as dipoles with mass md and dipole moment m, with directions as shown above.
(a) If you put two back-to-back magnets on the rod, the upper one will “float”. At what height
z does it float?
(b) If you now add a third magnet parallel to the bottom one as shown, find the ratio x/y of the
two heights, using only a scientific calculator. (Answer: 0.85.)
Solution. (a) We know that the field from a magnetic dipole is
µ0 m
B= 2 cos θ r̂ − sin θ θ̂ .
4πr3
Along the z-axis, this reduces to
µ0 m
Bz =
.
2πz 3
The force on the upper magnet must balance gravity, so
µ0 m2 d

1
− − md g = 0
2π dz z 3
which yields
1/4
3µ0 m2

z= .
2πmd g
(b) The net force on the middle magnet comes from the field from the top and bottom magnets,
along with gravity,
3µ0 m2 1

1
− = md g.
2π x4 y 4
Similarly, the top magnet, experiences forces from the bottom and middle magnets,
3µ0 m2 1

1
− = md g.
2π y 4 (y + x)4
Putting these two equations together yields
1 1 1 1
− = 4− .
x4 y 4 y (y + x)4
Defining α = x/y, we then need to solve
1/4
(1 + α)4

α= .
2(1 + α)4 − 1
13
We solve this using iteration, as introduced in P1. That is, we guess a reasonable value like
α = 0.5, then repeatedly plug in
1/4
(1 + Ans)4

2(1 + Ans)4 − 1
which yields x/y = 0.85.
[3] Problem 16. AuPhO 2019, problem 13. An elegant series of visual exercises on permanent
magnets, with practical applications. It will be useful to consult the answer sheet.
[3] Problem 17 (PPP 89). Two identical small bar magnets are placed on opposite ends of a rod of
length L as shown.
(a) Show that the torques the magnets exert on each other are not equal and opposite.
(b) Suppose the rod is pivoted at its center, and the magnets are attached to the rod so that
they can spin about their centers. If the magnets are released, the result of part (a) implies
that they will begin spinning. Explain how this can be consistent with energy and angular
momentum conservation, treating the latter quantitatively.
Solution. (a) Referring to the dipole fields computed in E1, the field at D due to C is twice
that at C due to D, so they can’t possibly cancel. Worse, the directions of the torques are
the same (both out of the page).
(b) Energy is conserved because there is an energy density B 2 /2µ0 stored in the magnetic field of
the two magnets. As the rotational kinetic energy of the system increases, the energy stored
in the field decreases to compensate.
Angular momentum conservation holds for a different reason. While electromagnetic fields
can store angular momentum too, they don’t in this particular case. Instead, the answer is
something more familiar. The magnets also exert forces on each other, so a force from the
rod is necessary to keep the magnets in place. This implies the magnets exert a torque on the
rod, which begins spinning in the opposite direction. Thus, angular momentum is conserved.
To show this quantitatively, set up coordinates with the origin at the center of the rod, and
the z-axis pointing out of the page. The total torque on the two magnets is
3µ0 µ2
τ0 = ẑ
4π L3
where µ is the magnetic moment of each magnet. This is the rate of change of their spin
angular momentum. Next, we consider forces. The force on magnet C due to D is
3µ0 µ2

µ0 2 d 1
FCD = µ∂x B = µ = ŷ.
4π dx x3 x=L 4π L4
14
This produces a torque on the rod, about its pivot point, of

1 3µ0 µ2
τ1 = − ẑ.
2 4π L3
The force on magnet D due to magnet C is equal and opposite, and therefore provides an
equal torque τ 2 on the rod. Therefore, the total rate of change of angular momentum is
τ 0 + τ 1 + τ 2 = 0.
In introductory textbooks, you might have read that angular momentum is conserved as
a consequence of the strong form of Newton’s third law, which is that forces are equal
and opposite, and always directed along the line separating two particles. As we’ve just
seen, this isn’t actually necessary: here we have an example of a force which isn’t directed
along the separation, but angular momentum is still conserved. In E7 we’ll see even more
exotic examples, where even the weak form of Newton’s third law (i.e. that forces are equal
and opposite) breaks down, but momentum remains conserved anyway, as a consequence of
electromagnetic fields carrying away the excess momentum. Generally speaking, the deeper
you get into physics, the less important Newton’s laws become, while conservation laws remain
as important as ever.
4 Point Charges
In this section we’ll give a sampling of classic problems involving just point charges in fields; these
will be a bit more mathematically advanced than the others in this problem set.
[3] Problem 18. A point charge q is released from rest a distance d from a grounded conducting
plane. Find the time until the point charge hits the plane. (Hint: consider Kepler’s laws.)
Solution. By using image charges, we see that the particle always experiences a force
kq 2
F =
4z 2
directly towards the plane, where z is its separation from the plane. Let the particle impact the
plane at point O.
This force has the form of an inverse-square law. In particular, we would get the exact same
result if the force were always directed towards O (rather than always directed towards the plane),
kq 2
F=− r̂.
4r2
But in this case, the problem is perfectly analogous to the central force of gravity, where O serves
as the location of the Sun, and one of the foci of the charge’s orbit. In particular, releasing the
charge from near rest and waiting for it to hit the plane corresponds to performing the first half of
an extremely eccentric elliptic orbit.
The trick is now to use Kepler’s third law. If the charge had performed a circular orbit of radius
d about O, then
kq 2 mv 2
= = mω 2 d
4d2 d
which gives a period of s
2π md3
T = = 4π .
ω kq 2
15
We can use Kepler’s third law to find the period of the eccentric elliptic orbit the charge actually
follows. This orbit has semimajor axis d/2, so it has period
T
T0 = √ .
2 2
The actual path of the charge is only the first half of this orbit, so the answer is
s
T0 T π md3
= √ =√ .
2 4 2 2 kq 2
Of course, the problem can also be solved by directly integrating the differential equation. This is
less subtle, but more complicated.
[3] Problem 19. A point charge of mass m and charge q is released from rest at the origin in the
fields E = E x̂, B = B ŷ. Find its position as a function of time by solving the differential equations
given by Newton’s second law, F = ma.
Solution. We will assume non-relativistic motion throughout. Note that the motion is solely in
the xz plane, since the electric and magnetic forces are in that plane. Newton’s second law gives
q
ẍ = (E0 − B0 ż),
m
q
z̈ = B0 ẋ.
m
Taking the time derivative of the first equation and plugging in into the second, we find
... q2B 2
x = − 20 ẋ,
m
and along with the initial condition that ẋ(0) = 0, we see that
ẋ = v0 sin(ωt)
where v0 is some yet to be determined velocity, and ω ≡ qB0 /m. Integrating, and using the initial
condition that x(0) = 0, we see that
v0
x(t) = (1 − cos(ωt)).
ω
We also have that
z̈ = ω ẋ = ωv0 sin(ωt).
Integrating twice and using the fact that z(0) = ż(0) = 0, we see that
v0
z(t) = v0 t − sin(ωt).
ω
All that is to be found now is v0 . Plugging our x and z into the first equation, we see that
q
v0 ω cos(ωt) = (E0 − B0 v0 (1 − cos(ωt))) =⇒ v0 = E0 /B0 .
m
16
Thus, our final solution is

v0
x(t) = (1 − cos(ωt)),
ω
v0
z(t) = v0 t − sin(ωt)
ω
where v0 = E0 /B0 and ω = qB0 /m.
Notice that while naively one might have thought the motion would be along E, on average the
particle actually moves along E × B. This is actually quite general. For example, it remains true
even if there’s a bit of friction; the steady state velocity turns out to be along E × B. Another
example is how weather systems work. When I was a kid, I was always confused about how entire
regions could have low or high pressure; would the wind just go along the pressure gradient to even
it out? That doesn’t happen because the Coriolis force deflects the wind sideways. In this case, the
pressure gradient is acting like E, and the Coriolis force behaves like a magnetic field B k ω k ẑ.
The net effect is that in the steady state, wind tends to move along lines of constant pressure, not
perpendicular to them. So a low pressure system stays low pressure but spins around.
[3] Problem 20 (Wang). Two identical particles of mass m and charge q are placed in the xy plane
with a uniform magnetic field Bẑ. The particles have paths r1 (t) and r2 (t). Neglect relativistic
effects, but account for the interaction between the charges.
(a) Write down a differential equation describing the evolution of the separation r = r1 − r2 .
(b) Suppose that the initial conditions have been set up so that the particles orbit each other in
a circle in the xy plane, with constant separation d. What is the smallest d for which this
motion is possible?
Solution. (a) The equations of motion for the two particles are
q2 q2
mr̈1 = r̂ + q ṙ1 × B, mr̈2 = − r̂ + q ṙ2 × B.
4π0 r3 4π0 r3
Subtracting the two, we have
q2
mr̈ = r + q ṙ × B.
2π0 r3
(b) Note that since B is along the ẑ direction, and v = ω × r where ω is also along the ẑ direction,
all three vector terms in the above equation are parallel. So we have
q2

2
+ qωB + mω r = 0
2π0 r3
and setting the term in parentheses to zero, we get

p
−(qB/m) ± (qB/m)2 − 2q 2 /πm0 d3
ω=
2
where ωc = qB/m is the usual cyclotron frequency. For this equation to have a solution, the
discriminant must be nonnegative, so
q2B 2 2q 2
≥
m2 πm0 d3
17
which gives
1/3
2m
d≥ .
π0 B 2
For smaller d, the charges will always fly apart, either due to electrostatic repulsion if they’re
slow, or the angular momentum barrier if they’re fast.
[4] Problem 21. [A] Consider a point charge of mass m and charge q in the field of a magnetic
monopole at the origin,
g
B = 2 r̂.
r
In this problem we’ll investigate the strange motion that results.
(a) Argue that the speed v is constant.
(b) Show that the angular momentum L of the charge is not conserved, but that
V = L − qgr̂
is. The second term is the angular momentum stored in the fields of the charge and monopole.
(c) Show that the charge moves on the surface of a cone. (Hint: in spherical coordinates where
the z-axis is parallel to V, consider V · φ̂.) Sketch some typical trajectories.
Solution. (a) The force is qv × B ⊥ v, so no work is done on the particle, so its speed remains
the same.
(b) Note that

r·ṙ
˙r̂ = d r = ṙr − ṙr = ṙr − r r = (r̂ · r̂) ṙ − r̂ · ṙ r̂ = r̂ × ṙ × r̂ = 1 r × (ṙ × r̂).

dt r r2 r2 r r r r2
˙ so L − qgr̂ is conserved.
We have L̇ = τ = r × (q ṙ × (g/r2 )r̂) = qg r̂,
(c) Take coordinates so that V is directed along ẑ and the particle is instantaneously in the xz
plane. Now take the y-component of the above equation, to give Ly = 0. In components, this
tells us that xpz − zpx = 0, or in other words that ẋ/ż = x/z. By drawing similar triangles,
this implies that the particle is momentarily moving so that x/z is conserved. By repeating
this argument at all times, r/z is conserved, where r is the distance to the z-axis. This defines
a cone.
In a typical trajectory, the charge spirals in towards the monopole along this cone, reaches
some minimum distance from it, then turns around and spirals out. In fact, it turns out that
if the cone is “cut and unfolded” and laid flat, the trajectory is a straight line! In other words,
it is a geodesic on the cone.
One can do problem 19 slickly using field transformations, an advanced subject we will cover in R3.
Idea 4: Field Transformations
If the electromagnetic field is (E, B) in one reference frame, then in a reference frame moving
18
with velocity v with respect to this frame, the components of the field parallel to v are
Ek0 = Ek , Bk0 = Bk
while the components perpendicular are

v
E0⊥ = γ(E⊥ + v × B), B0⊥ = γ B⊥ − 2 × E .
c
Remark
The nonrelativistic limit of the field transformation is useful, but one has to be careful in
deriving it. You might think, what’s the need for care? Can’t we just send c → ∞, Taylor
expand the above expressions, and call it a day? The problem with this reasoning is that
there’s no such thing as setting c → ∞. You can’t change a fundamental constant, and
moreover this statement isn’t even dimensionally correct, as noted in P1. What we really
mean by the nonrelativistic limit is restricting our attention to some subset of possible
situations, within which relativistic effects don’t matter.
For example, if we have a bunch of point charges with typical speed v, then the nonrelativistic
limit is considering only situations where v/c is small. In other words, we are taking v/c → 0,
not c → ∞. Since the magnetic field of a point charge is v/c2 times the electric field, the
magnetic field ends up small. Now if we also consider boosts with small speeds v, then
expanding the field transformations to lowest order in v/c gives
v
E0 = E, B0 = B − × E.
c2
This is the nonrelativistic limit for situations where E/B c, also called the electric limit.
However, there’s another possibility. Suppose that we have a bunch of neutral wires. In this
case, it’s the electric fields that are small, E/B c. Using this in the transformations above,
we arrive at the distinct result
B0 = B, E0 = E + v × B
which apply for situations where E/B c, also called the magnetic limit.
You might think we could improve the approximation by combining the two,
v
E0 = E + v × B, B0 = B − ×E
c2
but this isn’t self-consistent. For example, if you apply a Galilean boost with speed v, and
then a boost with speed −v, you don’t get back the same fields you started with! A sensible
Galilean limit is only possible if E/B c or E/B c, which are called the electric and
magnetic limits, discussed further in this classic paper. It’s only in relativity that E and B
can be treated on an equal footing.
19
[3] Problem 22. Using the Galilean field transformations to solve problem 19.
(a) In the magnetic limit, show that the Lorentz force stays the same between frames, as it should.
Then use the field transformations to find an appropriate reference frame where the problem
becomes easy.
(b) In the electric limit, show that the Lorentz force stays the same up to terms that are order
(v/c)2 smaller, assuming B/E ∼ v/c2 . (This is fine, since we’re taking the limit v/c → 0
anyway.) Then use the field transformations to find an appropriate reference frame where the
problem becomes easy.
(c) The solutions you found in parts (a) and (b) should look very different, even though you
should have found only one type of behavior in problem 19. In fact, there is a critical value
of E/B separating the two kinds of behavior. What is this critical value, and why didn’t you
run into it when solving problem 19?
Solution. (a) Suppose a particle has velocity u in the original frame, so the force there is
F = q(E + u × B). The force in the boosted frame is F0 = q(E + v × B + (u − v) × B) = F.
We can find a frame where there’s no electric field, by letting E + v × B = 0. In this frame, the
particle just orbits in a circle. (Going back to the original frame just gives back the cycloid
we found earlier.)
(b) We use the same setup as (a). The boosted force is

0 2 v×v×E u×v×E
F = q(E + (u − v) × (B − v × E/c )) = F + q −v × B + − .
c2 c2
The extra terms are all second order in v/c.

We can now find a frame where there’s no magnetic field, by letting B − v × E/c2 = 0. In
this frame, the particle just accelerates straight along E. This indicates that in the original
frame, the particle is always going along E, while getting deflected a bit to the side by the
magnetic field.
(c) As we found in problem 7, the characteristic velocity of the particle during the cycloid motion
is v0 = E/B. Hence in the electric limit v0 c, so our nonrelativistic solution must break
down. Accounting for the full relativistic dynamics, it turns out it is harder to “turn around”
a relativistic charge, so the particle is never turned around by the magnetic field.
To find the critical field value, we see the full relativistic field transformations allow us to
remove E exactly when E/B < c, and to remove B exactly when E/B > c. Hence E/B = c
separates the two behaviors. Intuitively, as E/B increases up to c, the cycloid solution is
stretched in the E direction more and more by relativistic effects, until when E/B = c it is
infinitely stretched.
20
Example 5: Drude Theory
Model a conductor as a set of electrons, of charge q, mass m, and number density n, which are
completely free. Assume that in every small time interval dt, each electron has a probability
dt/τ of hitting a lattice ion, which randomizes the direction of its velocity. Under these
assumptions, compute the resistivity of the material.
Solution
First, suppose the electrons have some average momentum hpi each. Because the collisions
randomize the velocity, the average momentum falls exponentially with timescale τ ,
dhpi hpi
=− .
dt τ
On the other hand, if there is an applied field, a force term appears on the right,
dhpi hpi
=− + qE
dt τ
since F = dp/dt for each individual electron. In the steady state,
hpi = qEτ.
The current density is

nqhpi nq 2 τ
J = nqhvi = = E.
m m
Thus, the resistivity in the Drude model is
m
ρ= .
nq 2 τ
We can also compute the typical drift velocity,
qEτ
v= .
m
For values of m that give reasonable ρ, the value of v is a literal snail’s pace, which is why
people say that the electrons themselves move very slowly through a circuit. (Of course, a
current can get started in a circuit much faster, because when a battery is attached, each
moving electron pushes on the next one along the wire, and this wave of motion travels much
faster than the electrons themselves.)
Remark
Above we assumed there was a given probability of collision per unit time, but usually when
a particle flies through a medium, there is a given probability of collision per unit length it
travels. If we had made the latter assumption, then the typical energy of the electrons would
21
match the work done on them in the typical length

√ ` they travel, mv 2 /2 ∼ qE`, but that
an average velocity that scales as v ∝ E. The analogue of Ohm’s law would then
implies √
be I ∝ V , completely contrary to observation!
A given probability of collision per unit time would be what we expect if the electrons had
constant speed, but how is this possible? The answer, as shown in X1, involves quantum
mechanics. The idea is that, due to the Pauli exclusion principle, the electrons in the
conductor have to occupy different quantum states, and the high density of electrons requires
most of them to have extremely high velocities, on the order of 1% of the speed of light! The
drift velocity is merely the tiny amount by which these velocities are shifted on average, so
their speeds are almost constant between collisions.
[2] Problem 23. Consider Drude theory again, but now suppose there is also a fixed magnetic field
Bẑ. In this case, J is not necessarily parallel to E, but the relation between the two can be described
by the “tensor of resistivity”. That is, the components are related by
X
Ei = ρij Jj .
j∈{x,y,z}
Calculate the coefficients ρij . Express your answers in terms of the quantities
m qB
ρ0 = , ω0 =
nq 2 τ m
as well as the parameter τ .
Solution. The Lorentz force expression says
dhpi hpi
=− + q(E + v × B).
dt τ
In the steady state, the left-hand side vanishes, so
hpi 1
= E + hpi × B.
qτ m
Switching from hpi to J and using the variables defined gives
E = ρ0 J − ρ0 ω0 τ J × ẑ.
From this, we can directly read off the components of the resistivity,
 
ρ0 −ρ0 ω0 τ
ρ = ρ0 ω0 τ ρ0 .
ρ0
When the electric field is in the ẑ direction, the magnetic field does nothing, which makes sense.
22
Since parallel currents attract, the currents within a single wire should contract. To estimate
this, consider a long wire of radius r. Suppose the atomic nuclei are fixed and have uniform
density, while the electrons move along the wire with speed v. Furthermore, assume that the
electrons contract, filling a cylinder of radius r0 < r with uniform negative charge density,
and that the wire is overall neutral. Find r0 .
Solution
The contraction of the electrons produces an overall inward electric field, and hence an
outward electric force on each electron, which balances the radially inward magnetic force.
Specifically, equilibrium occurs when E = vB.
Let the charge densities of the nuclei and electrons be ρ+ and ρ− . The magnetic field at
radius r is found by Ampere’s law, which gives
µ0 ρ− vr
(2πr)B = µ0 (ρ− v)(πr2 ), B= .
2
The electric field at radius r is found by Gauss’s law, which gives
1 1
(2πr)E = (ρ+ + ρ− )πr2 , E= (ρ+ + ρ− )r.
0 20
Note that both E and B are proportional to r. Then E = vB can be satisfied at all r simul-
taneously, which confirms that our assumption that ρ+ and ρ− were uniform is self-consistent.
Plugging these results into E = vB yields
v2
ρ+ + ρ− = ρ− (0 µ0 v 2 ) = ρ− .
c2
This can be written in terms of the Lorentz factor of special relativity,
1
ρ− = −γ 2 ρ+ , γ=p .
1 − v 2 /c2
Since the wire is overall neutral, ρ− r02 + ρ+ r2 = 0, so

r
r0 = .
γ
For nonrelativistic motion, the contraction is extremely small. (However, in plasmas, where
the positive charges are also free to move, this so-called pinch effect can be very significant.)
[2] Problem 24 (Griffiths 5.41). A current I flows to the right through a rectangular bar of conducting
material, in the presence of a uniform magnetic field B pointing out of the page, as shown.
23
(a) If the moving charges are positive, in what direction are they deflected by the magnetic field?
This deflection results in an accumulation of charge on the upper and lower surfaces of the
bar, which in turn produces an electric force to counteract the magnetic one. Equilibrium
occurs when the two exactly cancel. (This phenomenon is known as the Hall effect.)
(b) Find the resulting potential difference, called the Hall voltage, between the top and bottom
of the bar, in terms of B, the speed v of the charges, and the dimensions of the bar.
(c) How would the answer change if the moving charges were negative?
When measurements were performed in the early 20th century, some metals were found to have
positive moving charges! This “anomalous Hall effect” was solved by the quantum theory of solids,
as you can learn in any solid state physics textbook. Today, extensions of the Hall effect, such as
the integer and fractional quantum Hall effects, remain active areas of research, and could be used
to build quantum computers. We’ll return to these effects in X3.
Solution. (a) By using the right-hand rule twice, we find they are deflected down.
(b) The electric field is E = vB, so V = Eh = vBh where h is the thickness. Thus, in equilibrium,
the bottom is at a higher potential.
(c) If the current stays the same, the charges move the other direction. Since both the charge
and velocity flip, the Lorentz force stays the same, so the charges are still deflected down.
Thus, the sign of the charge that accumulates on the bottom is flipped, so now the top is at a
higher potential. Hence measuring the Hall voltage can be used to find the sign of the charge
01m
carriers in a material.
[3] Problem 26. 01m USAPhO 1997, problem B1. A nice problem on the dynamics of a plasma.
[3] Problem 27. 01^ USAPhO 2019, problem A3. This is a tough but useful problem. The first half
derives the so-called Child–Langmuir law, covered in problem 2.53 of Griffiths.
24
knzhou.github.io
Advice After Introductory Physics

I’ve met lots of people who have self-studied calculus-based physics, want to learn more, and are
unsure what to do next. This document lists some resources for going further. If you want advice
for how to start out learning physics or doing physics competitions, see my introductory advice file.
Deepening Core Knowledge

You’re done with the introductory material after finishing a book like Halliday, Resnick, and Krane.
If you enjoyed what you learned, there are many ways to deepen your knowledge.
• To get a deeper understanding of mechanics, try An Introduction to Mechanics by Kleppner
and Kolenkow or Introduction to Classical Mechanics by Morin. These are used for honors
introductory physics at MIT and Harvard, respectively. There’s substantial overlap between
the two, so just look at both and see which you like better. Morin has more sophisticated but
less realistic problems. Morin’s problems are about frictionless planes and massless strings,
while Kleppner’s problems are about cars and rockets.
• Halliday, Resnick, and Krane’s brief coverage of relativity probably isn’t enough to confidently
handle Olympiad problems on the subject. You can learn more about the subject in the last few
chapters of either Kleppner and Kolenkow or Morin. I think Morin’s discussion of relativistic
kinematics is excellent, carefully resolving many of the confusions a beginner might have.
• To get a deeper understanding of electromagnetism, try Electricity and Magnetism (3rd edition)
by Purcell and Morin, which is so exceptional that it’s used for honors introductory physics at
almost all top colleges. It has many insightful discussions and problems, but it’s best known
for a beautiful chapter that uses relativity to introduce magnetism. For a short and sweet
introduction to vector calculus that complements the book, try Schey’s Div, Grad, Curl, and
All That, which has good pictures; for a more thorough introduction, see MIT OCW’s 18.02.
• The waves and thermodynamics chapters in Halliday, Resnick, and Krane are enough to get
started on competitions, but other books can help deepen your understanding. For a clear
introduction intended explicitly for Olympiad preparation, try the first and second volumes of
Competitive Physics, by Wang and Ricardo, respectively.
• It’s also very rewarding to get a broad overview of physics as an intellectual tradition. I strongly
recommend the Feynman Lectures on Physics for fascinating discussion from a master of the
subject. If you’re interested in the history behind the material in an introductory physics course,
try Understanding Physics by Cassidy, Holton, and Rutherford.
• I generally recommend against reading any book with a title like “Modern Physics.” These
books have the same problem as algebra-based physics textbooks: they try to explain too much
with too little. Equations that should be derived with two lines of math end up “motivated” with
hundreds of words of vague argumentation. Since modern physics courses need to have a large
bank of problems to test students with, but they don’t teach the students to derive anything, the
textbooks end up packed with thousands of joyless, cookie-cutter questions where you simply
look up numbers in a table and plug them into a given formula. In my opinion, a standard
introductory book such as Halliday, Resnick, and Krane already gives enough background on
modern physics for the Olympiad. But if you really want to jump into these topics early, Krane’s
Modern Physics is well-written and comprehensible.
1
knzhou.github.io
There are many resources out there, but don’t get too worried about choosing one. All good books
contain the same essential core. As long as you understand the resource you’re using, and you’re
learning new things, you’re on the right track. If you’re not sure whether you can or should start a
book, just read the first chapter and see how it feels!
Competition Practice
There are many good sources of practice problems for competitions.
• Your primary source should be past USAPhO exams. I’ve rewritten the solutions to these
exams for clarity, but make sure to give each problem your best try before peeking; it’s easy to
waste one that way.
• If you run out of practice USAPhOs and really want more, AAPT sells a CD-ROM of USAPhO
exams going back to 1997. You can ask your physics teacher to purchase this, though they’re
substantially easier than current USAPhOs.
• Many problem sources are listed in the Syllabus for my handouts, but my favorite are:
– Jaan Kalda’s study guides. These are exceptional because they pair problems with the
specific ideas needed to do them, allowing for rapid progress. They leave a lot to the reader;
you won’t get far if you don’t already know calculus-based physics well. Student-written
solutions are available here.
– 200 Puzzling Physics Problems. This is a set of entertaining and devilishly tricky problems,
mostly compiled from Eastern European physics competitions. (However, I don’t recom-
mend using this as your main source of practice. A number of problems rely on obscure
tricks. It’s best when used as food for thought.)
More Advanced Physics

There’s a lot you can do if you want to move beyond competitions, towards more advanced physics.
• First, a warning. At this level, you’ll see a lot of “study guide” books marketed as “student
friendly”, “for students”, or even “for dummies”. They all have glowing reviews1 , but I strongly
recommend against using them. They don’t explain how the physics works; they just tell you
the mechanical steps you need to remember to solve the simplest exam problems.
In a course, they can be a life saver if you’re pressed for time and mostly interested in grades, but
the superficial understanding you’ll get from them will collapse by the first lecture of the next
course. It’s like building the facade of a house without the foundation. If you’re self-studying,
there’s no reason to ever use these books, because you have the time to do it right.
• If you’re following a standard college curriculum, the next thing to do is to get an understanding
of waves and oscillations. This corresponds to the MIT OCW’s 8.03, with further background
on differential equations given in 18.03. For entertaining demonstrations, see Walter Lewin’s
1
These reviews are especially annoying because they make it hard to differentiate quality. For example, A Student’s
Guide to Maxwell’s Equations is good, but A Student’s Guide to Lagrangians and Hamiltonians is atrocious: its
notation goes undefined and its derivations are just wrong. Every month I see a nonsensical StackExchange question
from a poor student who’s been terribly misled by the latter book. But both have a hundred five-star ratings online.
2
knzhou.github.io
old 8.03 lectures. For a textbook reference, Vibrations and Waves by French is short and clear,
covering the essentials of the subject, while Waves by Crawford is more thorough, with many
interesting real-world applications and home experiments. If you enjoyed Morin’s style, you
can also try his Waves book draft.
• For intermediate electromagnetism, the best book is Introduction to Electrodynamics by Griffiths,

by a big margin. The book is very clear, and its problems get surprisingly deep, due to Griffiths’
experience writing many papers on the subtleties of classical electromagnetism. This level of
electromagnetism requires deeper familiarity with vector calculus, which you can pick up from
the first few chapters. For a dedicated introduction to the math, see David Tong’s notes on
Vector Calculus for a clean approach that gets to tensors2 and introduces index notation, a
powerful tool for deriving more complicated vector calculus identities.
• After learning mechanics at the level of Morin or Kleppler and Kolenkow, you can start learning
Lagrangian and Hamiltonian mechanics, which are typically not useful for the Olympiad, but
extremely important for physics in general. For a clear and gentle introduction, see Classical
Mechanics by Taylor. For a nice supplement, see David Tong’s notes on Dynamics.
• The standard undergraduate textbooks are clear, but they focus on explaining theory, at
the expense of truly tricky problems. They can give you the mistaken impression that more
advanced physics just consists of grinding out calculations using established procedures. The
worst example of this is how most undergraduate mechanics books cover the formalism of
Hamiltonian mechanics for 100 pages, but then never use it to solve anything less trivial than
the harmonic oscillator. Of course, the giants of the 19th century loved tricky problems, and they
invented these theoretical tools precisely to solve them. To see a bit of this side of physics, try
flipping through Exploring Classical Mechanics by Kotkin and Serbo, and Static and Dynamic
Electricity by Smythe. The latter is a real classic; it’s the harder version of Jackson.
• For intermediate thermodynamics, I recommend either Concepts in Thermal Physics by Blundell

and Blundell, or Thermal Physics by Schroeder. Blundell and Blundell is more comprehensive,
with many interesting sidenotes and applications; Schroeder is short, clear, and crisp.
• All of the above is just an appetizer for what is arguably the main course of a physics major:
quantum mechanics. There are two ways to enter the subject. The traditional way is to start
with using Schrodinger’s equations to solve for wavefunctions; this lets you seamlessly transition
into it after finishing a course on waves, and supplies a lot of visual intuition. A newer approach
is to start by applying the fundamental postulates to low-dimensional systems, like the qubits
of quantum computing; this makes the subject accessible to people with no physics background
at all, but it is somewhat abstract and requires familiarity with linear algebra. Of course, if
you want to really know quantum mechanics, you’re have to understand both perspectives.
2
By the way, tensors are another one of those topics where you need an “official” source, like a book chapter or a
polished set of lecture notes. When I google “introduction to tensors”, most of the first ten results are terrible for
beginners. I recognize a few that once got me confused for days, because they were full of typos, and cross-checking
them made things worse because they used incompatible conventions. Others are correct, but way too sophisticated
to serve as a real introduction. It doesn’t help that mathematicians, computer scientists, engineers, and physicists
all mean different things when they say “tensor”! Don’t try to brute force learning this kind of thing by opening
twenty tabs. Just use one good source and stick with it. Exploring will deepen your knowledge only once you have
the foundation set.
3
knzhou.github.io
• To start the first way, read Introduction to Quantum Mechanics by Griffiths, which comes with
clear explanations and great problems. For video lectures at a similar level, try the MIT OCW
8.04 lectures by Barton Zwiebach (more clear) and Allan Adams (more energetic).
• To start the second way, read Introduction to Linear Algebra by Strang, used for MIT OCW’s
18.06, is excellent for this. (This is an excellent thing to do in general, even if you don’t stay
in physics, because linear algebra is arguably even more useful than calculus. You’ll probably
still need to use it if you go into computer science, finance, data science, or even, god forbid,
management consulting.) Then read the first half of Quantum Computation and Quantum
Information by Nielsen and Chuang, or try Umesh Vazirani’s Quantum Computation course.
• More advanced quantum mechanics courses will freely use both perspectives. For material at
this level, see Principles of Quantum Mechanics by Shankar, or the MIT OCW 8.05 lectures.
• If you want to start heading towards general relativity, the two gentlest books are Gravity by
Hartle and A First Course in General Relativity by Schutz. While both cover similar ground,
Schutz puts all the mathematical background up front (including a great introduction to four-
vectors and tensors), while Hartle starts with physical results, having you take some of the
math on faith until it’s filled in later. Both are good, so just pick whichever style you prefer.
• If you’re interested in fancy stuff like quantum field theory or string theory, I generally don’t
recommend reading anything about it at this stage. String theory in particular has an enormous
amount of prerequisites, which means that books which try to popularize it skew towards
“mindblowing” metaphysics. These books are built on layers of analogies, and you’ll naturally
want to probe deeper. But the second you try, the analogies will fall apart, because they are
merely shadows on the wall. You’ll find many lost souls online who have taken the analogies
too literally, arriving at homemade theories that have little to do with anything in physics.3
Anyway, if you’re determined to get a taste of these subjects, I recommend resources with at
least a few equations in them. For example, in ascending order of difficulty:
– Cumrun Vafa’s Puzzles to Unravel the Universe is like a popular string theory book, but
with points explained with neat mathematical puzzles. It’s meant for freshmen at Harvard,
which means you only need high school algebra to understand it.
– David Tong’s Particle Physics lecture notes cover the basics of the Standard Model and
beyond, with many references to the history and deeper theory, and only high school math.
– Leonard Susskind’s Theoretical Minimum lectures cover graduate-level topics using just
calculus. They’re not nearly detailed enough to serve as a foundation, but they do a great
job of giving the flavor of the logic.
– Griffiths’ Introduction to Elementary Particles clearly explains the basics of particle physics
and the structure of the Standard Model, along with how to do some toy calculations
in quantum field theory. It requires a good understanding of undergraduate quantum
mechanics and special relativity.
3
And it’s easy to go astray even if you have excellent reading comprehension, because English is ambiguous. For
example, if nothing is better than ice cream, and licorice is better than nothing, is licorice better than ice cream?
There’s a whole field of philosophy devoted to paradoxes like these, though to people who know math, it tends to
look like trying to bang rocks together to make fire.
4
knzhou.github.io
– Barton Zwiebach’s A First Course in String Theory is the string theory textbook with
the least prerequisites, and it actually derives many of the results it uses. It also requires
undergraduate quantum mechanics and special relativity.
Don’t get too confident if you use these – they only cover a small fraction of what typical
introductory books in these subjects do. To really get started in these fields, you should at
minimum learn graduate-level quantum mechanics first, at the level of Sakurai. For more
advanced resources for each subfield of physics, see the introductory pages of my lecture notes.
Second Opinions
The internet has some great sources for physics learning advice, and unfortunately many not-so-great
sources. There are two main reasons advice can be bad. The first is that many study guides are
written by people who have never actually learned the basics. They write guides because it takes
less effort to compile a big list of books and imagine knowing what’s in them than to open a single
book and actually learn something. The second is that most of the remaining study guides are
written by eminent professors who haven’t read an undergraduate-level book in decades. They’re
not going to start doing that again, so they tend to recommend outdated books that worked for
them forty years ago. Or, worse, they simply skim through a newer book’s table of contents and
recommend it if the chapter titles and the author’s name sound familiar; this way of doing things
keeps a lot of subpar books on the market. The basic rule is that you should only believe a book
recommendation if it was written by somebody who actually read and understood the book.
With that out of the way, here are links to some often-cited internet resources.
• Chicago Undergraduate Physics Bibliography. This is a list of books aimed towards learning
theoretical particle physics. It has solid advice, but it’s also 20 years out of date and missing
many of the canonical books.
• How to Learn Math and Physics. This list, oriented towards mathematical physics, is also 20
years out of date. It has great books on the list, but it could be difficult to use for a self-learner
because fluffy, equation-free popular books are placed right next to advanced graduate texts;
undergraduate level books are almost absent.
• How to become a GOOD Theoretical Physicist. Gerard ’t Hooft was one of the giants of
particle physics. About 30 years ago, he compiled a massive list of resources available around
the internet at the time. Unfortunately, the list was never finished or updated. Now, half
the links are broken, and the rest point to abandoned, rough drafts of lecture notes that are
honestly subpar, even compared to Wikipedia. It’s constantly recommended because its length
and author are impressive, but as far as I know, nobody has ever actually used it.4 If you ever
run into anybody who claims this is the best overall resource for learning physics, you can be
sure that they don’t know anything about physics at all.
• So You Want to Learn Physics.... Susan Rigetti’s book list is much more useful than the
others here, because she understands how it felt to be a student. This short list covers a full
undergraduate and graduate education, based on the canonical books.
4
A lot of books are like this too. For example, The Road to Reality is a wildly popular coffee-table book that
promises to cover all of physics in a couple hundred pages, but less than 1% of people who have bought it and
recommend it have read past the first chapter, and less than 1% of those people were learning anything new. If you
already know the material, it’s a brilliant and masterful review, which explains the praise from professors. But if you
don’t, there’s absolutely no way you can climb from arithmetic to quantum gravity with its couple hundred pages.
5
knzhou.github.io
• So You Want To Be A Physicist. This is a huge resource that covers everything about the
process of becoming a physicist in the US, from high school to postdoc applications. There’s
not much on the learning process itself, but it’s helpful, especially if you’re coming from outside
physics or outside the US.5
• A Physics Booklist. Once upon a time, the internet was arranged around discussion forums,
the greatest of which was sci.physics. As old-timers bitterly recount, it died a slow, agonizing
death 20 years ago as crackpots slowly outnumbered the people who actually knew things. This
unhelpfully long list is one of the relics of this lost civilization.
• Book Recommendations. I help curate this massive list of lists of book recommendations on
Physics StackExchange, which is the closest thing to a successor of sci.physics. It’s many times
larger than all the other lists here combined, but that also means it has a lot of cruft.
5
If you’re an international student, you can also find information on Academia.SE. However, I don’t recommend
taking the site too seriously, because it has a strong bias. Any post that complains about international students quickly
receives a hundred upvotes and thousands of views. But any post from an international student being mistreated is
condescendingly mocked, downvoted, closed, and deleted. I’ve heard privately from several people demoralized by this
process, and I’m sorry. It’s simply a fact that the site’s most devoted users view international students as all liars and
cheaters. Don’t worry so much about their opinions. The professors who spend all day on that site moralizing about
the “inferior culture” of other countries are not doing any important research anyway. If some of these petty people
go out of their way to be nasty to you, just remember it’s because there’s nothing of value going on in their lives.
6
knzhou.github.io
Physics Activities
Fast-Paced Competitions
Here are some fast-paced competitions you can participate in as a high schooler in the US:
• The F = ma/USAPhO exams are the premier physics competitions in the US. The F =
ma exam is a fast-paced multiple choice exam (75 minutes, 25 questions) focusing on tricky
mechanics problems, while the USAPhO exam has longer problems which require written
solutions (3 hours, 6 questions).
• The PhysicsBowl is a fast-paced multiple choice exam (45 minutes, 40 questions) at the level
of AP Physics 1 and 2.
• The Science Bowl is an exciting buzzer-based team competition covering all fields of science.
You need a team of four, preferably with everybody specialized in one or two subjects, and
qualifying for the national competition can be very competitive if you live in a large state. To
do well as the team’s physics player, you need to be able to solve AP Physics 1 and 2 questions
extremely quickly, and also have a good deal of general physics knowledge. If you have a lot of
general knowledge about science, consider being the science player on a quiz bowl team.
• Check to see if your state has local science competitions. For example, in New Jersey you can
participate in the New Jersey Science League.
• The Physics Unlimited Premier Competition is a fast-paced short answer exam (90 minutes, 4
questions) focusing on mechanics, which can also be taken internationally.
• It’s important to remember that even if you’re a hotshot at high school physics, there’s a lot
more to learn. For example, try playing arXiv vs. snarXiv to see if you can even tell the titles
of real physics papers from randomly generated ones.
Slower-Paced Competitions
Next, here are some slower-paced competitions, generally open internationally. These tend to have
less name recognition, but they’re fun and can help build long-term problem solving skills.
• KoMaL and FYKOS are Hungarian and Czech contests for high school students which issue a
problem set every month. The problems in both tend to be elegant and instructive.
• The team behind FYKOS also runs the team-based Online Physics Brawl and Fyziklani com-
petitions, which each last a few hours and have about 50 questions. Unfortunately, you’ll have
to stay up late to participate if you live in the US.
• The AAPT’s journal, The Physics Teacher , publishes a monthly problem under the column
“Physics Challenge for Teachers and Students”. The difficulty is usually between an F = ma
and a USAPhO problem.
• The Physics Unlimited Explorer Competition is a team-based, two week competition where
students explore an open-ended problem and write a research paper-style report.
1
knzhou.github.io
• The High School Mathematical Contest in Modeling and MathWorks Math Modeling Challenge
are team-based, one to two day competitions where students make are faced with an open-ended
real-world problem and write a report issuing policy recommendations.
• The Online Physics Olympiad is a team-based, several day competition written entirely by high
schoolers, founded in 2020. It is one of the largest international physics competitions, with
hundreds of teams participating. Because it’s student written, the problems can be ambiguous
or difficult to understand, but they improved by a lot in 2021.
• The Physics Cup is an extremely difficult competition with one question per month, with a
hint released per week. The problems are very instructive, and range from applied physics to
almost pure mathematics. They always are clearly posed and have unambiguous answers, but
they are sometimes so hard that almost nobody can solve them even after four weeks of hints.
In theory, all the problems can be solved with only introductory physics. A good knowledge of
Euclidean geometry tends to help.
• The Rudolf Ortvay competition is a marathon competition with about 30 questions to be solved
in 10 days, requiring undergraduate physics knowledge to solve.
• The International Theoretical Physics Olympiad is a 24-hour open-book competition for teams
of undergraduates. The problems tend to be deep and open-ended, and can require graduate
physics knowledge or use of the research literature.
Science Fairs
You can also consider entering a high school science fair, such as ISEF or Siemens, which have
great name recognition. Unfortunately, I’m not qualified to give advice on this, since I placed at
the bottom of my county’s science fair, after being dragged there by a graduation requirement. My
comparison of bean growing methods, with everything bought at Home Depot, was no match against
the genetic engineering projects other people showed up with, using equipment from university labs.
I remember spending the day forlorning glancing back and forth, from the Excel beanstalk height
bar charts on my cardboard poster, to the towering achievements all around me. Gene therapy,
antibiotic resistance, cancer immunotherapy – even their poster material was nicer! For quite some
time, this made me doubt I could ever be a scientist. I must have been lacking some inherent spark,
some essential creativity, to have come up with those projects myself.
Of course, that wasn’t right at all, and the reason is that science fairs don’t work like the
competitions I’ve listed above. In physics competitions, you can do well by learning physics on your
own, or in any decent school, and thinking hard about it. For them, I never needed anything but a
dog-eared, $20 used textbook, but in science fairs, you need to know people. The reason is that it’s
almost impossible to come up with and carry out a novel and important physics research project
sitting on your own at home, given the immense background required; even the nation’s brightest
child prodigies are barely at the level of the average new graduate student. The press releases try
to make it sound like the projects spring out of a flash of creative inspiration, but in reality, the
projects are simply given to students by professors at universities, who guide them through step by
step. The students work on subsubtasks like any diligent research assistant would, but the overall
direction, and even the choices of subtasks, are the professor’s.
Now, most professors won’t even work with undergraduates in research, because they lack the
background of graduate students. And high schoolers have even less background than undergrads.
2
knzhou.github.io
That means the professors that are willing to serve as research mentors are outnumbered by high
school students looking for them by a factor of a hundred to one, or perhaps even a thousand to
one. Given those odds, how can you find one?
In practice, there are two options: either cold email many professors and get extremely lucky, or
have an “in”. For example, you can be born to parents who themselves are professors, so they can
hire you in their own labs, or send you to their friends’ labs. Or you could just have very involved
parents, who will dig into the system and fight for you. You can go to an elite high school which
has a dedicated class that mass-produces winning science fair projects. If all else fails, you can shell
out for a few hours of Zoom calls with a past science fair winner, so you can ask them how they
did it. These days, they charge about $9,500 for the privilege.
Even once you have a mentor, your success in a science fair hinges on external factors. When you
do high school physics competitions, or research in graduate school, success is driven by internal
factors: you decide how many hours to work, which avenues look promising, and what to try next.
But a high schooler doing research almost never has the background to operate that way. Success
will only be possible if your mentor goes to the trouble of developing a very concrete, well-posed
subproblem for you to work on, while they handle all the creative direction and the details that
require advanced knowledge.1 The fact that this sounds like a pretty bad deal for the mentor (whose
student will soon go off to college, never to return) is precisely why you need an “in” to get one,
which in turn is highly correlated with your pedigree.
In my opinion, the best childhood activities are those directed as much as possible by the kids
themselves. The Online Physics Olympiad mentioned above is a great example: in a year where
most physics competitions were cancelled, high school students wrote, publicized, ran, and graded
a massive competition on their own, with zero budget. By contrast, a science fair is by definition
an exercise in appealing to adults, who craft the projects, pay for the equipment, run the fair, hear
the presentations, and judge the winners. And because there is so much adult involvement, and
the stakes in prestige are so high, the essential playfulness is lost. The mood is serious, or perhaps
even farcical. In a scene out of Bosch, feuding parents pull strings and throw piles of money with a
desperate energy, in the feverish pursuit of an Ivy League college.
Given all this, why do we place value on science fairs at all? Physics Olympiad problems are
merely exercises designed to be solvable by high schoolers who learn the subject well. But the end
goal of this learning to produce scientists who can produce completely new knowledge, years down
the line. A good science fair project, however, purportedly shows that the student can produce new
knowledge today. What could be a better signal of their potential?2
1
Some of my comments are specific to physics. For example, research projects in the social sciences, or in certain
areas of mathematics like combinatorics, require substantially less background. Here, a high schooler really could
do novel, interesting work reasonably independently. And if you can tinker, you can make good stuff with relatively
cheap electronics. On the other hand, for biology and chemistry, university mentorship and tight guidance is even
more necessary than in physics. You need the expensive equipment in university labs, and no professor is just going
to let you mess with it. If you’re there, they’ll be giving you very explicit instructions so you don’t waste their money.
2
The reason this feels so compelling is what Cal Newport calls the failed simulation effect. If some achievement
looks impressive, and you can’t see the steps one would take to get it, you naturally conclude that it must spring
from some exceptional genius. In reality, they usually come from parents’ pocketbooks or personal contacts. For
example, when you hear that, say, a 12 year old kid built a nuclear reactor in their garage, you should understand that
(1) a very inclusive definition of “nuclear reactor” is being used, with the products more closely resembling smoke
detectors and bananas than power plants, (2) there are established hobbyist communities for this kind of thing, and
step-by-step instructions, (3) the main reason few kids do it is that the materials cost tens of thousands of dollars and
require adult supervision to put together – aspects both carefully scrubbed out of the press release. (And who do you
think wrote the press release?) Of course I don’t mean to decry these kinds of activities; they’re probably a great way
for parents and their kids to bond and learn together. You just shouldn’t feel bad if you didn’t do the same as a kid.
3
knzhou.github.io
That’s the narrative, anyway, but I think an equally compelling argument could be made in the
opposite direction. What are the traits of good researchers? There are many, but they certainly
include a deep understanding of one subfield, a broad knowledge of all of physics, and the ability to
make insights into difficult problems they’ve never seen before. These underlying skills are precisely
what is trained by physics Olympiad problems, many of which illustrate the key insights behind
real breakthroughs in physics. If you really care about research, I think it’s better to spend your
high school years building the foundation that will someday make you a good researcher, rather
than impressing physics-illiterate adults with the superficial appearance of the final product.
I have nothing against science fair participants, and I’m friends with a bunch of national science
fair winners, many of whom also did Olympiads. The national winners are smart and dedicated.
But if you want to stand a chance to qualify for the national science fairs, and you don’t already
know how the game is played, you’re not going to find anything useful here, or in any book. You’re
not going to get anywhere with $20 or a handful of beans. You need to know the right people.
Programs
Okay, so how do you find the right people? I’ve gathered some advice from friends who have won
national science fairs. First, you can look at organized programs.
• If you are part of an underrepresented minority group, there are many well-organized and
well-funded summer programs which will take care of the process for you and let you meet a
group of similar students, such as SAMS, MITES, SMASH, YRP, SSRP, and many more.
• If you’re not part of such a group, you should avoid most of the well-promoted programs out
there. As one of many terrible examples3 , the Summer STEM Institute is a Zoom-based program
run by undergraduates with no experience formulating research questions (as mentioned above,
even national winners are simply told what to do by their mentors). Programs like these often
promise you’ll get a publishable paper after just a few hour-long meetings with a random college
student, which is completely absurd if you know anything about research. They also cost up
to $10,000, which means you need to pay over $1,000 per hour of actual mentorship! Make no
mistake: these programs are just a trap for desperate and unwise parents, run by deeply cynical
kids looking to make a quick buck. They embody everything wrong with the science fair and
college admissions process.
• Anyway, if you’re aiming for pure math research, one excellent program is MIT PRIMES,
which matches high school students with qualified mentors for free. Admissions are relatively
transparent, requiring the solution of a problem set; Olympiad qualifications help. Also see this
advice from the Euler Circle.
• For general science research, the best program is RSI, which is also free. RSI is an extremely
selective program, and one of its explicit goals is to allow its students to go to any selective
college they want, which is achieved by very strong recommendation letters. However, because
of its prestige, if you don’t come from a very small state, you usually need to already have
research experience to stand a chance of admission.
• There are also great physics summer programs that are more about learning than research.
3
It’s not worth giving a list of programs to watch out for, because a new one starts up every month. However, it’s
easy to spot them using common sense. Just divide the cost with the number of hours you actually meet one-on-one
with a qualified mentor (college students don’t count, some graduate students do), and see if the result is sensible.
4
knzhou.github.io
– ISSYP is an nearly free program covering topics in theoretical physics, held at the beautiful
Perimeter Institute. QCSYS is a free program run by Perimeter’s partner university,
focusing on quantum cryptography.
– SPARC is a free program which covers “applied qualitative thinking”, which means an array
of fun topics such as game theory, economics, and cognitive science. Though it doesn’t have
much to do with physics, it attracts many people in the Olympiad community and is a lot
of fun, since there’s plenty of free time for discussions. A major part of the application is a
set of interesting, open-ended questions.
– SSP is a relatively expensive but fun program for high school juniors. While it bills itself
as a research program, the real point is community building – hanging out with friends all
night in a big observatory, taking data. Admissions have a strong “personality” component.
Zoom-based camps are much less fun than in-person camps, because all the informal socialization
is lost, leaving only dry lectures. I wouldn’t recommend paying money for a Zoom-based camp.
I also wouldn’t recommend paying for a summer camp at a prestigious university which centers
around taking a standard college course; they tend to be extremely overpriced, and you can get
the same knowledge elsewhere.
Starting Research
While organized programs are fun, outside of pure math, most science fair winners don’t get their
projects from them. The main method is reaching out to professors directly.
• The hardest stumbling block, at least for me, was realizing this is a sensible thing to try. As
mentioned above, it’s highly unlikely that you could do anything more efficiently than a professor
or even a graduate student, since you are missing years of knowledge. But professors really do
choose to mentor high school students sometimes, for a variety of reasons: your time is free,
you could become more useful if you stick around, and they might like you for some reason.
• If you have the privilege of personally knowing lots of professors already, you’re very lucky; stop
reading, and just ask them the next time you bump into them at the country club. Otherwise,
to get started, browse the faculty listings at nearby universities. Read their biographies and
websites and see what sparks your interest. Exclude any professors that are no longer taking
students, such as emeriti. Consider friendly faculty who care about mentoring; a professor
happy to work with undergraduates, and who publicly encourages them to apply, is much more
likely to be willing to work with high school students.
• Don’t be overly picky at this stage. For example, don’t fall into the trap of insisting on a string
theory project because you once saw a Michio Kaku video – you won’t get anywhere that way.
Also don’t laser focus on the highest ranked university within a thousand miles. Cutting-edge
research is done at hundreds of universities in the United States. The most important factor is
that the university be close enough for you to regularly show up.
• Pick several professors to send a cold email to. Don’t be demanding; be aware that you’re
asking for a big favor. Also be aware that professors typically skim through hundreds of emails
a day. Your first email should briefly convey the following information:
– Your interest in physics. Don’t try to exaggerate here, because it will be obvious if you
do. For example, if you’ve learned some quantum mechanics by reading a textbook, that’s
5
knzhou.github.io
a great thing to mention. But don’t say you’ve been dying to work on superconducting
spintronics, or whatever it is the professor works on, since infancy. A decent number of
professors have heard about Olympiads, so mentioning your awards can help.
– Your ability to code. In order for a high school student to be helpful at all, the professor
needs to be able to find a specific, reasonably sized task for them to do, requiring less
technical background. In the vast majority of cases, this means you’ll end up coding, which
is an essential part of all fields of physics. If you have experience coding, even if it wasn’t
in physics, you should mention it.
– The amount of time you can commit. Summer tends to be a good time to start, because
you’ll be free almost all the time. Don’t put a hard time limit (e.g. “I need a result within 3
months for this science fair”) because research doesn’t work that way. Most decent projects
take a year or more.
No matter what your qualifications are, the vast majority of requests will be declined; that’s
normal. Remember, you only need one!
• Here are some points of email etiquette:
– In the United States, titles generally don’t matter as much as in Europe or Asia, but it’s
still best to address professors as Prof. Lastname and postdocs as Dr. Lastname in a first
email. If they reply to you and sign off with their first name, feel free to use it.
– Sometimes, high school students have elaborate email signatures detailing their class year,
address, school, favorite color, and more. Only administrators use these in real life. Just
sign off with “Thanks”, followed by your name.
– Don’t send anything that demands substantial energy for the professor to parse, such as a
multi-page resume with every minor honor you’ve ever gotten.
– If you don’t get a response in a week, your email could have just fallen down in their inbox,
so you can send a single reminder. Don’t send multiple reminders and don’t be annoying.
• Remember that research takes a long time, and don’t obsess about jumping to a final result
for the science fair. You’ll probably need weeks of reading and listening just to figure out the
basics of what your problem is about. Sometimes, months of work will yield nothing at all!
If you stick around in your research group, you’ll find that the real benefit of mentorship has nothing
to do with winning the science fair. You’ll gain a much deeper appreciation of how science works
by being part of a real scientific community. If you work hard and show that you can be a useful
part of this community, your mentor will be able to say a lot more about your abilities than a high
school teacher ever could.

Kevin Zhou All Handouts

Uploaded by

Copyright:

Available Formats

Kevin Zhou All Handouts

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Kevin Zhou All Handouts

Uploaded by

Copyright:

Available Formats

knzhou.github.

Advice For Introductory Physics

What should I know before I start learning physics?

What should I know before I start entering physics competitions?

How do I start learning physics?

How can I tell if I understand algebra-based introductory physics?

What are some good introductory books at each level?

– Hewitt, Conceptual Physics.

What makes a competition prep program effective?

Is prep program X, book Y, or course Z enough for USAPhO?

Do I really have to learn X if I want to win competitions?

And what about those weird things I learned in middle school?

memorization of arbitrary rules, occasionally punctuated by cutesy crafts5 . Thankfully, none of

How should I prepare for the F = ma exam?

Sir, what are your tips to crack JEE?

What’s the best way to spend my time learning?

How should I self-study from Halliday, Resnick, and Krane?

What are some important traps to avoid?

Do I have enough talent to succeed?

01W – 22.5 minutes, 01m – 45 minutes, 01h – 100 minutes.

2 Textbooks and Resources

• There are a number of compilations of Russian physics problems.

• 2 units of problem solving.

– M4: oscillations. Damped/driven oscillators, normal modes, small oscillations, adiabaticity.

– E1: electrostatics. Coulomb’s law, Gauss’s law, potentials, conductors.

– R1: kinematics. Lorentz transformations, Doppler effect, acceleration, classic paradoxes.

– W1: wave equation, standing waves, music, interferometry. (M4 required)

• 3 units of modern physics.

Frequently Asked Questions

Who has used the handouts?

What’s the catch?

Why are you giving everything away for free?

Do you offer tutoring for the problem sets?

What is the schedule for releasing the handouts?

How do I know if I’m ready to start using the handouts?

How long should it take to go through them?

How did you choose the problems?

Can I ask you about the problems?

What if I find an error in the solutions?

(b) What are the readings on the two spring scales?

(a) Show that g can be measured as

[3] Problem 14. A cue ball is a uniform sphere of radius R.

(a) Which orbit has the most angular momentum?

(b) Which orbit has the highest total energy?

(c) On which orbit is the largest speed acquired?

In all cases, justify your answer carefully.

[2] Problem 17. A comet passes by Sun as shown, in a parabolic path.

2 Problem Solving Skills

(a) Using the relation E = hf , find the dimensions of h.

[2] Problem 25. A half-infinite line has linear charge density λ.

(c) Sketch the electric field lines everywhere.

(a) Consider the vector field

(b) Find the potential difference between the plates.

(c) Find the force between the plates.

[3] Problem 28. A battery is connected to an RC circuit as shown.

(a) Solve for the charge on the capacitor as a function of time.

(a) What is the induced current in the ring?

(a) Find the magnetic field everywhere.

(a) What is the net momentum p acquired by the loop?

(b) How is momentum conserved in this setup?

01W – 22.5 minutes, 01m – 45 minutes, 01h – 100 minutes.

[m] = kg, [~] = J · s = kg m2 s−1 , [e2 /4π0 ] = J · m = kg m3 s−2 .