Vector Calculus

CHAPTER 2
Vector Functions
10. Curves in Space and Vector Functions

To describe the motion of a pointlike object in space, its position vector
must be specified at every moment of time. A vector is defined by three
components in a coordinate system. Therefore, the motion of the object
can be described by an ordered triple of real-valued functions of time. This
observation leads to the concept of vector-valued functions of a real variable.
Definition 10.1. (Vector Function).
Let D be a set of real numbers. A vector function r(t) of a real variable t is
a rule that assigns a unique vector to every value of t from D. The set D is
called the domain of the vector function.
Most commonly used rules to define a vector function are algebraic rules
that specify components of a vector function in a coordinate system as func-
tions of a real variable: r(t) = hx(t), y(t), z(t)i. For example,
√
r(t) = h 1 − t , ln(t) , t2 i or
√
x(t) = 1 − t , y(t) = ln(t) , z(t) = t2 .
Unless specified otherwise, the domain of the vector function is the set D
of all values of t at which the algebraic rule makes sense; that is, all three
components can be computed for any t from D. In the above example,
√
x(t) = 1 − t ⇒ −∞ < t ≤ 1 ,
y(t) = ln(t) ⇒ 0 < t < ∞ ,
z(t) = t2 ⇒ −∞ < t < ∞ .
The domain of the vector function is the intersection of the domains of its
components:
D = (−∞, 1] ∩ (0, ∞) ∩ (−∞, ∞) = (0, 1] .
The range of a vector function is the collection of all vectors r(t). It can
be visualized by viewing each r(t) as a position vector, that is, all vectors
from the range are depicted so that they have a common initial point (the
origin). The collection of all terminal points of position vectors defined by
a vector function r(t) is called a graph of the vector function.
Suppose that the components of a vector function r(t) are continuous
functions on an interval D = I = [a, b]. Consider all vectors r(t), as t ranges
over I, positioned so that their initial points are at a fixed point (e.g., the
origin of a coordinate system). Then the terminal points of the vectors r(t)
157
158 2. VECTOR FUNCTIONS
Figure 10.1. Left: The terminal point of a vector r(t)

whose components are continuous functions of t traces out
a curve in space. Right: Graphing a space curve. Draw a
curve in the xy plane defined by the parametric equations x =
x(t), y = y(t). It is a “shadow” of the original curve in the
xy plane and traced out by the vector R(t) = hx(t), y(t), 0i.
This planar curve defines a cylindrical surface in space. The
space curve in question lies in this surface. The space curve
is obtained by raising or lowering the points of the planar
curve along the surface by the amount z(t), that is, r(t) =
R(t) + ê3 z(t). In other words, the graph z = z(t) is wrapped
around the cylindrical surface.
form a curve in space as depicted in Figure 10.1 (left panel). The simplest
example is a straight line, which is described by a linear vector function
r(t) = r0 + tv. Thus, the graph of a vector function whose components are
continuous functions is a curve in space.
10.1. Graphing Space Curves. To visualize the shape of a curve C traced

out by a vector function, it is convenient to think about r(t) as a trajectory
of motion. The position of a particle in space may be determined by its
position in a plane and its height relative to that plane. For example, this
plane can be chosen to be the xy plane. Then
r(t) = hx(t), y(t), z(t)i
= hx(t), y(t), 0i + h0, 0, z(t)i
= R(t) + z(t)ê3 .
Consider the curve defined by the parametric equations x = x(t), y = y(t)
in the xy plane. It is a “shadow” made by the original curve in the xy
plane (if the light comes parallel to the z axis). One can mark a few points
along the shadow curve corresponding to particular values of t, say, Pn with
10. CURVES IN SPACE AND VECTOR FUNCTIONS 159
Figure 10.2. Graphing a helix. Left: The shadow curve

R(t) = hcos t, sin t, 0i is a circle of unit radius, traced out
counterclockwise. So the helix lies on the cylinder of unit
radius whose symmetry axis is the z axis. Middle: The
graph z = z(t) = t is a straight line that defines the height of
helix points relative to the circle traced out by R(t). Right:
The graph of the helix r(t) = R(t)+z(t)ê3 . As R(t) traverses
the circle, the height z(t) = t rises linearly. So the helix can
be viewed as a straight line wrapped around the cylinder.
coordinates (x(tn ), y(tn)), n = 1, 2, ..., N . Then the corresponding points

of the curve C are obtained from them by moving the points Pn along the
direction normal to the plane (i.e., along the z axis in this case), by the
amount z(tn ); that is, Pn goes up if z(tn ) > 0 or down if z(tn ) < 0. In other
words, as a particle moves along the curve x = x(t), y = y(t), it ascends or
descends according to the corresponding value of z(t).
The curve can also be visualized by using a piece of paper. Consider a
general cylinder with the horizontal trace being the curve x = x(t), y = y(t),
like a wall of the shape defined by this curve. Then make a graph of the
function z(t) on a piece of paper (wallpaper) and glue it to the wall so that
the t axis of the graph is glued to the curve x = x(t), y = y(t) while each
point t on the t axis coincides with the corresponding point (x(t), y(t)) of
the curve. After such a procedure, the graph of z(t) along the wall would
coincide with the curve C traced out by r(t). The procedure is illustrated
in Figure 10.1 (right panel).
Example 10.1. Graph the vector function r = hcos t, sin t, ti, where t
ranges over the real line.
Solution: It is convenient to represent r(t) as the sum of a vector in the
xy plane and a vector parallel to the z axis. In the xy plane, the curve
x = cos t , y = sin t ⇔ x2 + y 2 = 1
is the circle of unit radius traced out counterclockwise so that the point
(1, 0, 0) corresponds to t = 0. The circular motion is periodic with period 2π.
The height z(t) = t rises linearly as the point moves along the circle. Starting
from (1, 0, 0), the curve makes one turn on the surface of the cylinder of
unit radius climbing up by 2π. Think of a piece of paper with a straight
line depicted on it that is wrapped around the cylinder. Thus, the curve
traced by r(t) lies on the surface of a cylinder of unit radius and periodically
winds about it climbing by 2π per turn. Such a curve is called a helix. The
procedure is shown in Figure 10.2.
10.2. Limits and Continuity of Vector Functions.

Definition 10.2. (Limit of a Vector Function).
A vector r0 is called the limit of a vector function r(t) as t → t0 if
lim kr(t) − r0 k = 0 ;
t→t0
the limit is denoted as limt→t0 r(t) = r0 .

The left and right limits, limt→t− r(t) and limt→t+ r(t), are defined sim-
0 0
ilarly. This definition says that the length or norm of the vector r(t) − r0
approaches 0 as t tends to t0 . The norm of a vector vanishes if and only if
the vector is the zero vector. Therefore, the following theorem holds.
Theorem 10.1. (Limit of a Vector Function).
Let r(t) = hx(t), y(t), z(t)i and let r0 = hx0 , y0 , z0 i. Then the limit of a
vector function exists if and only if the limits of its components exist:
lim r(t) = r0 ⇐⇒ lim x(t) = x0 , lim y(t) = y0 , lim z(t) = z0 .
t→t0 t→t0 t→t0 t→t0
This theorem reduces the problem of finding the limit of a vector function
to the problem of finding limits of three ordinary functions.
Example 10.2. Let r(t) = hsin(t)/t , t ln t , (et − 1 − t)/t2 i. Find the
limit of r(t) as t → 0+ or show that it does not exist.
Solution: Recall from Calculus I that by the definition of the derivative
of f (t) = sin(t)
f (t) − f (0) sin t
f 0 (0) = lim = lim = cos 0 = 1
t→0 t t→0 t
because (sin t)0 = cos t. Since the limit exists, the left and right limits exist
and are equal to the limit. Thus, the limit of the first component is 1. The
limits of the other components may be investigated by l’Hospital’s rule. Let
us verify the hypotheses of l’Hospital’s rule. One has t ln(t) = ln(t)/t−1 so
that the second component is a ratio of two differentiable functions ln(t)
and t−1 that approach infinity as t → 0+ (the undetermined form ∞ ∞ ). The
third component is also a ratio of two differentiable functions that approach
zero as t → 0+ (the undetermined form 00 ). Then by l’Hospital’s rule for
either of these undetermined forms, if the limit of the ratio of the derivatives
exists, then the original limit exists and is equal to the limit of the ratio of
the derivatives. The existence of the limit of the ratio of the derivatives is
verified by the direct calculation:
ln t (ln t)0 t−1
lim t ln t = lim = lim = lim = − lim t = 0,
t→0+ t→0+ t−1 t→0+ (t−1 )0 t→0+ −t−2 t→0+
et − 1 − t et − 1 et 1
lim 2
= lim = lim = .
t→0+ t t→0+ 2t t→0+ 2 2
where l’Hospital’s rule has been used twice to calculate the last limit. Thus,
limt→0+ r(t) = h1, 0, 21 i.
Remark. Although l’Hospital’s rule is a powerful tool to resolve undeter-

mined forms, the use of basic properties of elementary functions and their
approximations by Taylor polynomials is a better (and often technically
simpler) method to analyze limits. Recall from Calculus II
t2 t3 t2
et = 1 + t + + + · · · = 1 + t + + O(t3 ) ,
2! 3! 2
t 3 t 5 t3
sin t = t − + + · · · = t − + O(t5 ) ,
3! 5! 6
2 4
t t t2
cos t = 1 − + + · · · = 1 − + O(t4 ) ,
2! 4! 2
p(p − 1) p(p − 1)(p − 2) 3
(1 + t)p = 1 + pt + t2 + t +···
2! 3!
p(p − 1) 2
= 1 + pt + t + O(t3 ) ,
2
1! 2! 3! t2
ln(1 + t) = t − t2 + t3 − t4 + · · · = t − + O(t3 ) ,
2! 3! 4! 2
where the symbol O(tn ) denotes terms of order at least tn , that is, O(tn ) =
M tn + possibly terms of higher order, for some constant M 6= 0. In other
words, the characteristic property of O(tn ) (the property that defines the
symbol O(tn )) is
O(tn )
lim n = M 6= 0 .
t→0 t
Taylor polynomial approximations of degree n are obtained from the corre-
sponding Taylor series by omitting terms O(tn+1 ) (which is justified when
the argument t is small) and used to analyze limits and local behavior of
functions near t = 0. For example, the limit of the third component can be
found as
et − 1 − t (1 + t + t2 /2 + O(t3 )) − 1 − t
=
t2 t2
2
t /2 + O(t ) 3
=
t2
1 1
= + O(t) →
2 2
as t → 0 because O(t) → 0. This method becomes superior to l’Hospital’s

rule when functions in questions cannot be expressed in elementary functions
(e.g. special functions) or such expressions are too complicated, while power
expansions are known for most functions encountered in applications. The
reader is encouraged to practice this method.
Definition 10.3. (Continuity of a Vector Function).
A vector function r(t), t ∈ [a, b], is said to be continuous at t = t0 ∈ [a, b] if
lim r(t) = r(t0 ) .
t→t0
A vector function r(t) is continuous in the interval [a, b] if it is continuous

at every point of [a, b].
By Theorem 10.1, a vector function is continuous if and only if all its
components are continuous functions.
Example 10.3. Let r(t) = hsin (2t)/t , t2 , eti for all t 6= 0 and r(0) =
h1, 0, 1i. Determine whether this vector function is continuous.
Solution: The components y(t) = t2 and z(t) = et are continuous for
all real t and y(0) = 0 and z(0) = 1. The component x(t) = sin(2t)/t is
continuous for all t 6= 0 because the ratio of two continuous functions is
continuous. Using the Taylor approximation
sin(2t) 2t + O(t3 )
x(t) = = = 2 + O(t2 ) ⇒ lim x(t) = 2 6= x(0) = 1;
t t t→0
that is, x(t) is not continuous at t = 0. Thus, r(t) is continuous everywhere,
but t = 0.
10.3. Space Curves and Continuous Vector Functions. A curve connecting

two points in space as a point set can be obtained as a continuous trans-
formation (or a deformation without breaking) of a straight line segment
in space. Conversely, every such space curve can be continuously deformed
to a straight line segment. So a curve connecting two points in space is a
continuous deformation of a straight line segment, and this deformation has
a continuous inverse.
A straight line segment can be viewed as an interval a ≤ t ≤ b (a set of
real numbers between a and b). Its continuous deformation can be described
by a continuous vector functions r(t) on [a, b]. So the range of a continuous
vector function defines a curve in space. Conversely, given a curve C as a
point set in space, one might ask the question: What is a vector function
that traces out a given curve in space? The answer to this question is not
unique. For example, a line L as a point set in space is uniquely defined by its
particular point and a vector v parallel to it. If r1 and r2 are position vectors
of two particular points of L, then both vector functions r1 (t) = r1 + tv and
r2 (t) = r2 − 2tv trace out the same line L because the vectors −2v and v
are parallel.
The following, more sophisticated example is also of interest. Suppose

one wants to find a vector function that traces out a semicircle of radius
R. Let the semicircle be positioned in the upper part of the xy plane:
x2 + y 2 = R2 and y ≥ 0. The following three vector functions trace out the
semicircle:
p
r1 (t) = ht, R2 − t2 , 0i , −R ≤ t ≤ R ,
r2 (t) = hR cos t, R sin t, 0i , 0 ≤ t ≤ π ,
r3 (t) = h−R cos t, R sin t, 0i , 0 ≤ t ≤ π .
This is easy to see by noting that the y components are non-negative in the
specified intervals and the norm of these vector functions is constant for any
value of t:
kri (t)k2 = x2i (t) + yi2 (t) = R2 , i = 1, 2, 3 ;
here ri = hxi , yi , zii. The latter means that the endpoints of the vectors
ri(t) always remain on the circle of radius R. It can therefore be concluded
that there are many vector functions whose ranges define the same curve in
space.
Another observation is that there are vector functions that trace out the
same curve in opposite directions as t increases from its smallest value a to
its largest value b. In the above example, the vector function r2 (t) traces
out the semicircle counterclockwise, while the functions r1 (t) and r3 (t) do so
clockwise. So a vector function defines the orientation of a curve. However,
this notion of the orientation of a curve should be regarded with caution
because a vector function may traverse its range (or a part of it) several
times. For example, the vector function
r(t) = hR cos t, R| sin t|, 0i
traces out the semicircle twice, back and forth, when t ranges from 0 to 2π.
The vector function
r(t) = (t2 , t2 , t2 )
is continuous on the interval [−1, 1] and traces out the straight line segment,
x = y = z, between the points (0, 0, 0) and (1, 1, 1) twice.
To emphasize the noted differences between space curves as point sets
and continuous vector functions, the notion of a parametric curve is intro-
duced.
Definition 10.4. (Parametric Curve)
A continuous vector function on an interval is called a parametric curve in
space.
If the range of a continuous vector function r(t) = hx(t), y(t), z(t)i, a ≤
t ≤ b coincides with a given curve C as a point set in space, then the vector
function is also called a parameterization of the curve C, the equations
C: x = x(t) , y = y(t) , z = z(t) , a ≤ t ≤ b,
are called parametric equations of C, and t is called a parameter. As noted, a

parameterization of a given space curve is not unique and there are different
parametric equations that describe the very same space curve. A parametric
curve r(t) is closed if
r(a) = r(b) ⇔ C is closed
(the initial and terminal points of a curve traversed by r(t) coincide). For
example, spatial curves that resemble the letter O or the infinity sign ∞ are
closed. A curve in space is said to be simple if, loosely speaking, it does
not intersect itself. To make this notion precise, it is rephrased in terms of
parametric curves. A non-closed parametric curve r(t) is called simple if
r(t1 ) 6= r(t2 ) for any t1 6= t2 ⇔ C is simple .
For example, curves resembling the letters M and N are simple, while a
curve that resembles the numeral 6 or the Greek letter α is not simple. A
closed parametric curve r(t) is simple if r(t1 ) 6= r(t2 ) for any t1 6= t2 , except
the case when t1 = a and t2 = b. A point set C is a simple curve if there
is a simple parametric curve whose range is C. A simple parametric curve
whose range is a space curve C is also called a simple parameterization of
the curve C. For example, a curve resembling the letter O is a simple closed
curve, whereas a closed curve resembling the infinity sign ∞ is not simple.
A simple parametric curve is always oriented. A vector function can also be
defined on an infinite interval, e.g., [a, ∞), or (−∞, b], or (−∞, ∞). In this
case, a parametric curve is said to be simple if it is simple on any subinterval
of the infinite interval.
Example 10.4. Find linear vector functions that orient the straight line
segment between r1 = h1, 2, 3i and r2 = h2, 0, 1i from r1 to r2 and from r2
to r1 .
Solution: The vector r2 − r1 = h1, −2, −2i is parallel to the line segment.
So the vector equation r(t) = r1 + t(r2 − r1 ) describes the line that contains
the segment in question. The vector r2 − r1 is directed from r1 to r2 .
Therefore when t increases from t = 0, the terminal point of r(t) goes along
the line from r1 toward r2 , reaching the latter at t = 1. Thus, the segment
is traversed from r1 to r2 by the vector function
r(t) = r1 + t(r2 − r1 ) = h1 + t, 2 − 2t, 3 − 2ti, 0 ≤ t ≤ 1.
Swapping the points r1 and r2 in the above argument, it is concluded that
the vector function
r(t) = r2 + t(r1 − r2 ) = h2 − t, 2t, 1 + 2ti, 0 ≤ t ≤ 1,
traverses the segment from r2 to r1 .
Example 10.5. Determine whether the parametric curve
r(t) = hcos t , sin(2t) , sin2 (2t)i , 0 ≤ t ≤ 2π ,
is closed or not and simple or not. Describe its shape and indicate the
orientation, if any.
Solution: Let us first investigate the vertical projection of the curve:
x = cos t , y = sin(2t) .
From the graphs of cos t and sin(2t), it is clear that this planar curve origi-
nates from the point (1, 0), goes into the first quadrant for 0 ≤ t ≤ π/2, then
crosses the origin at t = π/2. For π/2 ≤ t ≤ π, the curve lies in the third
quadrant and crosses the x axis at t = π. As t increases from π to 3π/2,
the curve is in the second quadrant and passes through the origin again at
t = 3π/2. Finally, for 3π/2 ≤ t ≤ 2π, the curve is in the fourth quadrant
and returns to the initial point at t = 2π. It resembles the infinity sign ∞.
Since r(0) = r(2π) = h1, 0, 0i, the curve is closed. The curve is not simple
because it intersects itself, r(π/2) = r(3π/2) = 0.
Now imagine a cylinder along the z axis with the cross section of the
shape of the infinity sign ∞. To visualize the curve, wrap the graph z =
z(t) = sin2 (2t) around this cylinder so that the zeros of the height z(t) at
t = 0, t = π/2, t = π, t = 3π/2, and t = 2π match respectively the points
(1, 0), (0, 0), (−1, 0), (0, 0), and (1, 0) on the cross section of the cylinder by
the xy plane. Clearly, the curve is oriented (the loop with x ≥ 0 is traversed
counterclockwise as viewed from the top of the z axis, while the loop with
x ≤ 0 is traversed clockwise).
10.4. Study Problems.

Problem 10.1. Find a vector function that traces out a helix of radius R
that climbs up along its axis by h per one turn. Is such a helix unique?
Solution: Let the helix axis be the z axis. By making the mechanical anal-
ogy with the motion of a particle along the helix in question, the motion in
the xy plane must be circular with radius R. Suitable parametric equations
of the circle are
x2 + y 2 = R2 ⇒ x(t) = R cos t , y(t) = R sin t .
With this parameterization of the circle, the motion has a period of 2π. On
the other hand, z(t) must rise linearly by h as t changes over the period.
Therefore, z(t) = ht/(2π). The vector function may be chosen in the form
r(t) = hR cos t, R sin t, ht/(2π)i .
Alternatively, one can take parametric equations of the circle in the form
x2 + y 2 = R2 ⇒ x(t) = R cos t , y(t) = −R sin t .
In the latter parameterization, the circle is traced out clockwise, whereas it is
traced out counterclockwise in the former parameterization. Consequently,
the vector function
r(t) = hR cos t, −R sin t, ht/(2π)i
Figure 10.3. Illustration to Study Problem 10.2. Left:

The curve lies on the cylinder of unit radius. It may be
viewed as the graph of z = sin(4t) on the interval 0 ≤ t ≤
2π wrapped around the cylinder. Top right: The circle
(shadow) traced out by R(t) = hcos t, sin t, 0i. It defines the
cylindrical surface on which the curve lies. Bottom right:
The graph z = z(t) = sin(4t), which defines the height of
points of the curve relative to the circle in the xy plane. The
dots on the t axis indicate the points of intersection of the
curve with the xy plane in the first quadrant.
also traces out a helix with required properties. The two helices are different
despite that they share the same initial and terminal points. One helix winds
about the z axis clockwise while the other counterclockwise.
Problem 10.2. Sketch and/or describe the curve traced out by the vector
function r(t) = hcos t, sin t, sin(4t)i if t ranges in the interval [0, 2π].
Solution: The vector function R(t) = hcos t, sin t, 0i traverses the circle
of unit radius in the xy plane, counterclockwise, starting from the point
(1, 0, 0). As t ranges over the specified interval, the circle is traversed only
once. The height z(t) = sin(4t) has a period of 2π/4 = π/2. Therefore,
the graph of sin(4t) makes four ups and four downs if 0 ≤ t ≤ 2π. The
curve r(t) = R(t) + ê3 z(t) looks like the graph of sin(4t) wrapped around
the cylinder of unit radius. It makes one up and one down in each quarter
of the cylinder. The procedure is shown in Figure 10.3.
Problem 10.3. Sketch and/or describe the curve traced out by the vector
function r(t) = ht cos t, t sin t, ti.
Figure 10.4. Illustration to Study Problem 10.3. Left:

The height of the graph relative to the xy plane (top). The
shadow curve R(t) = ht cos t, t sin t, 0i. For t ≥ 0, it looks
like an unwinding spiral (bottom). Right: For t > 0, the
curve is traversed by the point moving along the spiral while
rising linearly above the xy plane with the distance traveled
along the spiral. It can be viewed as a straight line wrapped
around the cone x2 + y 2 = z 2 .
Solution: The components of r(t) satisfy the equation

x2 (t) + y 2 (t) = t2 ⇒ x2 (t) + y 2 (t) = z 2 (t)
for all values of t. Therefore, the curve lies on the double cone x2 + y 2 = z 2 .
The shape of the planar parametric curve x = x(t), y = y(t) is easy to
understand in polar coordinates:
x = r cos ϕ , y = r sin ϕ ⇒ r = t, ϕ = t
The curve represents a rotational motion (the angle ϕ increases linearly with
time t) about the origin such that the distance from the origin r increases
linearly with the angle of rotation. This is a spiral in the xy plane.
To obtain the spatial curve in question, each point of the spiral must
be moved up or down in accord with the corresponding value of the height
z(t) = t. If t increases from t = 0, the curve in question is traced by a point
that rises linearly with the distance from the origin as it travels along the
spiral. If t decreases from t = 0, instead of rising, the point would descend
(z(t) = t < 0). So the curve winds about the axis of the double cone while
remaining on its surface. The procedure is shown in Figure 10.4.
Problem 10.4. Find the portion of the elliptic helix r(t) = h2 cos(πt),
t, sin(πt)i that lies inside the ellipsoid x2 + y 2 + 4z 2 = 13.
Solution: The helix here is called elliptic because it lies on the surface of an
elliptic cylinder. Indeed, in the xz plane, the parametric curve x = 2 cos(πt),
z = sin(πt) traverses the ellipse x2 /4 + z 2 = 1 because the latter equation
is satisfied for all real t:
x2 (t)
+ z 2 (t) = cos2 (πt) + sin2 (πt) = 1 .
4
Therefore the curve remains on the surface of the elliptic cylinder parallel
to the y axis. One turn around the ellipse occurs as t changes from 0 to 2
because the functions cos(πt) and sin(πt) have the period 2π/π = 2. The
helix rises by 2 along the y axis per turn because y(t) = t. Now, to solve
the problem, one has to find the values of t at which the helix intersects the
ellipsoid. The intersection happens when the components of r(t) satisfy the
equation of the ellipsoid, that is, when
x2 (t) + y 2 (t) + 4z 2 (t) = 1 ⇒ 4 + t2 = 13 ⇒ t = ±3 .
The position vectors of the points of intersection are r(±3) = h−2, ±3, 0i.
The portion of the helix that lies inside the ellipsoid corresponds to the
range −3 ≤ t ≤ 3.
Problem 10.5. Consider two curves C1 and C2 traced out by the vector
functions r1 (t) = ht2 , t, t2 + 2t − 8i and r2 (s) = h8 − 4s, 2s, s2 + s − 2i,
respectively. Do the curves intersect? If so, find the points of intersection.
Suppose two particles have the trajectories r1 (t) and r2 (t), where t is time.
Do the particles collide?
Solution: The curves intersect if there are values of the pair (t, s) such
that r1 (t) = r2 (s). This vector equation is equivalent the system of three
equations
  2
 x1 (t) = x2 (s)  t = 8 − 4s
r1 (t) = r2 (s) ⇔ y1 (t) = y2 (s) ⇒ t = 2s
 2
z1 (t) = z2 (s) t + 2t − 8 = s2 + s − 2

Substituting the second equation t = 2s into the first equation, one finds that
(2s)2 = 8 − 4s whose solutions are s = −2 and s = 1. One has yet to verify
that the third equation holds for the pairs (t, s) = (−4, −2) and (t, s) = (2, 1)
(otherwise, the z components do not match). A simple calculation shows
that indeed both pairs satisfy the equation. So the position vectors of the
points of intersection are
r1 (−4) = r2 (−2) = h16, −4, 0i and r1 (2) = r2 (1) = h4, 2, 0i .
Although the curves along which the particles travel intersect, this does
not mean that the particles would necessarily collide because they may not
arrive at a point of intersection at the same moment of time, just like two
cars traveling along intersecting streets may or may not collide at the street
intersection. The collision condition is more restrictive:
r1 (t) = r2 (t) ;
the time of collision t must satisfy three conditions (the particles happen to
be at the same time and the same point). For the problem at hand, these
conditions cannot be fulfilled for any t because, among all the solutions of
r1 (t) = r2 (s), there is no solution for which t = s. Thus, the particles do
not collide.
Problem 10.6. Find a vector function that traces out the curve of in-
tersection of the paraboloid z = x2 + y 2 and the plane 2x + 2y + z = 2
counterclockwise as viewed from the top of the z axis.
Solution: One has to find the components x(t), y(t), and z(t) such that
they satisfy the equations of the paraboloid and plane simultaneously for
all values of t. This ensures that the endpoint of the vector r(t) remains
on both surfaces, that is, traces out their curve of intersection (see Fig-
ure 10.5). Consider first the motion in the xy plane described by the vector
function hx(t), y(t), 0i. Solving the plane equation for z, z = 2 − 2x − 2y,
and substituting the solution into the paraboloid equation, one finds
2 − 2x − 2y = x2 + y 2 ⇔ 4 = (x + 1)2 + (y + 1)2
by completing the squares. This equation describes a circle of radius 2
centered at (−1, −1). Its parametric equations may be chosen as
x = x(t) = −1 + 2 cos t , y = y(t) = −1 + 2 sin t .
Figure 10.5. Illustration to Study Problem 10.6. The

curve is an intersection of the paraboloid and the plane P.
It is traversed by the point simultaneously moving counter-
clockwise about the circle (shadow) in the xy plane (indicated
by P0 ) and rising so that it remains on the paraboloid.
By construction, this circle is the vertical projection, hx(t), y(t), z(t)i →

hx(t), y(t), 0i, of the curve of intersection onto the xy plane (the plane P0
in Figure 10.5). As t increases from 0 to 2π, the circle is traced out coun-
terclockwise as required (the clockwise orientation can be obtained, e.g.,
by reversing the sign of sin t). The height along the curve of intersection
relative to the xy plane is z(t) = 2 − 2x(t) − 2y(t). Thus,
r(t) = h−1 + 2 cos t, −1 + 2 sin t, 6 − 4 cos t − 4 sin ti ,
where t ∈ [0, 2π].
Problem 10.7. Let v(t) → v0 and u(t) → u0 as t → t0 . Prove the limit
law for vector functions: limt→t0 (v(t) · u(t)) = v0 · u0 using only Definition
10.2. Then prove this law using Theorem 10.1 and basic limit laws for
ordinary functions.
Solution: The idea is similar to the proof of the basic limit laws for or-
dinary functions given in Calculus I. One has to find an upper bound for
|v · u − v0 · u0 | in terms of kv − v0 k and ku − u0 k. By Definition 10.2, the
latter quantities converge to zero as t → t0 . The conclusion should follow
from the squeeze principle. Consider the identities:
v · u − v0 · u0 = (v − v0 ) · u + v0 · u − v0 · u0
= (v − v0 ) · u + v0 · (u − u0 )
= (v − v0 ) · (u − u0 ) + (v − v0 ) · u0 + v0 · (u − u0 )
It follows from the inequality 0 ≤ |a + b| ≤ |a| + |b| and the Cauchy-Schwarz
inequality (Theorem 3.2) |a · b| ≤ kakkbk that
0 ≤ |v · u − v0 · u0 |
≤ |(v − v0 ) · (u − u0 )| + |(v − v0 ) · u0 | + |v0 · (u − u0 )|
≤ kv − v0 kku − u0 k + kv − v0 kku0 k + kv0 kku − u0 k
By Definition 10.2, kv − v0 k → 0 and ku − u0 k → 0 as t → t0 . So the right
side of the above inequality converges to zero. By the squeeze principle, it
is then concluded that |v · u − v0 · u0 | → 0 as t → t0 , which proves the
assertion. A proof based on Theorem 10.1 is simpler. If vi (t) and ui (t),
i = 1, 2, 3, are components of v(t) and u(t), respectively, then by Theorem
10.1 vi (t) → v0i and ui (t) → u0i as t → t0 . Hence,

lim v(t) · u(t) = lim v1 (t)u1 (t) + v2 (t)u2 (t) + v3 (t)u3 (t)
t→t0 t→t0
= lim v1 (t)u1 (t) + lim v2 (t)u2 (t) + lim v3 (t)u3 (t)
t→t0 t→t0 t→t0
= v01 u01 + v02 u02 + v03 u03
= v0 · u0
where the basic limit laws for ordinary functions have been used.
10.5. Exercises.
1–5. Find the domain of each of the following vector functions:
1. r(t) = h t, 2 t
√ t , 2e it ;
2. r(t) = h √t, t , e i ;
3. r(t) = h 9 − t2 , ln t, cos t i ;
2
4. r(t) = h ln(9 √ − t ), ln √ |t|, (1 + t)/(2 + t) i ;
5. r(t) = h t − 1, ln t, 1 − t i .
6–15. Find each of the following limits or show that it does not exist:
√
6. limt→1 h √t, 2 − t − t2 , 1/(t2 − 2) i ;
7. limt→1 h t, 2 − t − t2 , 1/(t2 − 1) i ;
8. limt→0 h et , sin t, t/(1 − t) i ;
9. limt→∞ h e−t , 1/t2 , 4 i ; √ √
10. limt→∞ h e−t , (1 − t√ 2 )/t2 , 3 t/( t + t) i ;
11. limt→−∞ h 2, t2 , 1/ 3 t √ i;
12. limt→0+ h (e2t − 1)/t, ( 1 + t − 1)/t, t ln t i ;
13. limt→0 h sin2 (2t)/t2 , t2 + 2, √ (cos t − 1)/t2 i ;
14. limt→0 h (e2t − 1)/t, t cot t, 1 + t i ;
15. limt→∞ h e2t / cosh2 t, t2012 e−t , e−2t sinh2 t i .
16–22. Sketch each of the following curves and identify the direction in
which the curve is traced out as the parameter t increases:
16. r(t) = ht, cos(3t), sin(3t)i ;
17. r(t) = h2 sin(5t), 4, 3 cos(5t)i ;
18. r(t) = h2t sin t, 3t cos t, ti ;
19. r(t) = hsin t, cos t, ln ti ;
20. r(t) = ht, 1 − t, (t − 1)2 i ;
21. r(t) = ht2 , t, sin2 (πt)i √;
22. r(t) = hsin t, sin t, 2 cos ti .
23. Two objects are said to collide if they are at the same position at the
same time. Two trajectories are said to intersect if they have common points.
Let t be the physical time. Let two objects travel along the space curves
r1 (t) = ht, t2 , t3 i and r2 (t) = h1 + 2t, 1 + 6t, 1 + 14ti. Do the objects collide?
Do their trajectories intersect? If so, find the collision and intersection
points.
24. Find a simple parameterization of the curve of intersection of the surfaces
x2 + y 2 /4 + z 2 /9 = 1, y ≥ 0, and z = x2 . Sketch the curve.
25–32. Find two vector functions that traverse a given curve C in the
opposite directions if C is the curve of intersection of two surfaces:
25. y = x2 and z = 1 ;
26. x = sin y and z = x ;
27. x2 + y 2 = 9 and z = xy ;
28. x2 + y 2 = z 2 and x + y + z = 1 ;
29. z = x2 + y 2 and y = x2 ;
30. x2 /4 + y 2 /9 = 1 and x + y + z = 1 ;
31. x2 /2 + y 2 /2 + z 2 /9 = 1 and x − y = 0 ;
32. x2 + y 2 − 2x = 0 and z = x2 + y 2 .
33. Specify the parts of the curve r(t) = hsin t, cos t, 4 sin2 ti that lie above
the plane z = 1 by restricting the range of the parameter t.
34. Find the values of the parameters a and b at which the curve r(t) =
h1 + at2 , b − t, t3 i passes through the point (1, 2, 8).
35–39. Find the values of a, b, and c, if any, for which each of the following
vector functions is continuous: r(0) = ha, b, ci and, for t 6= 0,
35. r(t) = ht, cos2 t, √ 1 + t + t2 i ;
36. r(t) = ht, cos2 t, 1 + t2 i ;
37. r(t) = ht, cos2 t, ln |t|i ;
38. r(t) = hsin(2t)/t, sinh(3t)/t, t ln |t|i ;
39. r(t) = ht cot(2t), t1/3 ln |t|, t2 + 2i .
40. Suppose that the limits limt→a v(t) and limt→a u(t) exist. Prove the
basic laws of limits for the following vector functions:
lim(v(t) + u(t)) = lim v(t) + lim u(t),
t→a t→a t→a
lim(sv(t)) = s lim v(t),
t→a t→a
lim(v(t) · u(t)) = lim v(t) · lim u(t),
t→a t→a t→a
lim(v(t) × u(t)) = lim v(t) × lim u(t).
t→a t→a t→a
41. Prove the last limit law in Exercise 40 directly from Definition 10.2, i.e.,
without using Theorem 10.1. Hint: see Study Problem 10.7.
42–47. Let
√
v(t) = h(e2t − 1)/t, ( 1 + t − 1)/t , t ln |t|i ,
u(t) = hsin2 (2t)/t2 , t2 + 2, (cos t − 1)/t2 i ,
w(t) = ht2/3 , 2/(1 − t), 1 + t − t2 + t3 i .
Use the basic laws of limits established in Exercise 40 to find:
42. limt→0 (2v(t) − u(t) + w(t)) ;
43. limt→0 (v(t) · u(t)) ;
44. limt→0 (v(t) × u(t)) ;
45. limt→0 [w(t) · (v(t) × u(t))] ;
46. limt→0 [w(t) × (v(t) × u(t))] ;
47. limt→0 [w(t) × (v(t) × u(t)) + v(t) × (u(t) × w(t)) +
u(t) × (w(t) × v(t))] .
48. Suppose that the vector function v(t) × u(t) is continuous where u(t)
and v(t) are non-vanishing. Does this imply that both vector functions v(t)
and u(t) are continuous? Support your arguments by examples.
49. Suppose that the vector functions v(t) × u(t) and v(t) 6= 0 are contin-
uous. Does this imply that the vector function u(t) is continuous? Support
your arguments by examples.
11. DIFFERENTIATION OF VECTOR FUNCTIONS 173
11. Differentiation of Vector Functions

Definition 11.1. (Derivative of a Vector Function).
Suppose a vector function r(t) is defined on an interval [a, b] and t0 ∈ [a, b].
If the limit
r(t0 + h) − r(t0 ) dr
lim = r0 (t0 ) = (t0 )
h→0 h dt
exists, then it is called the derivative of a vector function r(t) at t = t0 ,
and r(t) is said to be differentiable at t0 . For t0 = a or t0 = b, the limit
is understood as the right (h > 0) or left (h < 0) limit, respectively. If the
derivative exists for all points in [a, b], then the vector function r(t) is said
to be differentiable on [a, b].
It follows from Theorem 10.1 that a vector function is differentiable if

and only if all its components are differentiable:
D x(t + h) − x(t) y(t + h) − y(t) z(t + h) − z(t) E
r0 (t) = lim , ,
h→0 h h h
0 0 0
(11.1) = hx (t) , y (t) , z (t)i .
For example,
r(t) = hsin(2t), t2 − t, e−3t i ⇒ r0 (t) = h2 cos(2t), 2t − 1, −3e−3t i.
Definition 11.2. (Continuously Differentiable Vector Function).

If the derivative r0 (t) is a continuous vector function on an interval [a, b],
then the vector function r(t) is said to be continuously differentiable on [a, b].
Higher-order derivatives are defined similarly: the second derivative is

the derivative of r0 (t):
r00 (t) = (r0(t))0 = hx00 (t) , y 00 (t) , z 00 (t)i ,
the third derivative is the derivative of r00(t):
r000(t) = (r00 (t))0 = hx000 (t) , y 000 (t) , z 000 (t)i ,
and so on
r(n)(t) = (r(n−1)(t))0 = hx(n)(t) , y (n) (t) , z (n) (t)i ,
provided all components are differentiable sufficiently many times.
11.1. Differentiation Rules. The following rules of differentiation of vector

functions can be deduced from the rule (11.1).
Theorem 11.1. (Differentiation Rules).

Suppose u(t) and v(t) are differentiable vector functions and f (t) is a real-
valued differentiable function. Then
dh i
v(t) + u(t) = v 0 (t) + u0 (t),
dt
dh i
f (t)v(t) = f 0 (t)v(t) + f (t)v 0 (t),
dt
dh i
v(t) · u(t) = v 0 (t) · u(t) + v(t) · u0 (t),
dt
dh i
v(t) × u(t) = v 0 (t) × u(t) + v(t) × u0 (t).
dt
dh i
v(f (t)) = f 0 (t)v 0(f (t)).
dt
The proof is based on a straightforward use of the rule given in Eq.
(11.1) and basic rules of differentiation for ordinary functions and left as an
exercise to the reader.
Example 11.1. Find the first and second derivatives of the vector func-
tion r(t) = (a + t2 b) × (c − td) where a, b, c, and d are constant vectors.
Solution: By the product rule,
r0 (t) = (a + t2 b)0 × (c − td) + (a + t2 b) × (c − td)0
= 2tb × (c − td) − (a + t2 b) × d
r00 (t) = (2tb)0 × (c − td) + 2tb × (c − td)0 − (a + t2 b)0 × d
= 2b × (c − td) − 2tb × d − 2tb × d
= 2b × c − 6tb × d
Alternatively, the cross product can be calculated first and then differenti-
ated:
r(t) = a × c − ta × d + t2 b × c − t3 b × d
r0 (t) = −a × d + 2tb × c − 3t2 b × d
r00 (t) = 2b × c − 6tb × d

11.2. Differential of a Vector Function. If r(t) is differentiable, then

(11.2) r(t + h) − r(t) = hr0 (t) + hu(h) ,
where u(h) approaches the zero vector, u(h) → 0, as h → 0. Indeed, by the
definition of the derivative,
r(t + h) − r(t)
lim u(h) = lim − r0 (t) = 0 .
h→0 h→0 h
Therefore, the components of the difference r(t + h) − r(t) − hr0 (t) converge
to 0 faster than h. Suppose that r0 (t0 ) does not vanish. Consider a linear
vector function L(t) with the property L(t0 ) = r(t0 ). Its general form is
L(t) = r(t0 ) + (t − t0 )v ,
where v is a constant vector. For t close to t0 , L(t) is a linear approximation
to r(t) in the sense that the approximation error kr(t) − L(t)k becomes
smaller with decreasing |t − t0 |. It follows from Eq.(11.2) that
r(t) − L(t) = (r0 (t0 ) − v)h + hu(h) , h = t − t0 .
By the triangle inequality
kak − kbk ≤ ka + bk ≤ kak + kbk
the error of the linear approximation is bounded as
kr(t) − L(t)k
kr0 (t0 ) − vk − ku(h)k ≤ ≤ kr0 (t0 ) − vk + ku(h)k .
|h|
Since ku(h)k tends to zero as h → 0, the lower and upper bounds (the left
and right sides of the inequality) approach the same value kr0(t0 ) − vk. This
allows us to conclude that the approximation error decreases linearly with
decreasing h:
kr(t) − L(t)k = kr0 (t0 ) − vk |h| , h → 0,
0 0
provided v 6= r (t0 ). Only when v = r (t0 ), the approximation error de-
creases faster than h:
kr(t) − L(t)k = ku(h)k |h| , h → 0.
Thus, the linear vector function
(11.3) L(t) = r(t0 ) + r0 (t0 )(t − t0 )
is the best linear approximation of r(t) near t = t0 . Provided the derivative
does not vanish, r0 (t0 ) 6= 0, the linear vector function L(t) defines a line
passing through the point r(t0 ). This line is called the tangent line to the
curve traced out by r(t) at the point r(t0 ).
The analogy can be made with the tangent line to the graph y = f (x)
at a point (x0 , y0 ) where y0 = f (x0 ). The equation of the tangent line is
y = y0 + f 0 (x0 )(x − x0 )
(recall Calculus I). The graph is a curve in the xy plane whose parametric
equations are
C: y = f (x) ⇔ x = t , y = f (t) ⇔ r(t) = ht, f (t)i .
Put r(t0 ) = hx0 , y0 i. Then the tangent line is traversed by the linear vector
function (11.3) where
r0 (t0 ) = h1, f 0 (t0 )i .
Figure 11.1. Left: A secant line through two points of

the curve, P0 and Ph . As h gets smaller, the direction of
the vector P0~Ph = r(t0 + h) − r(t0 ) becomes closer to the
tangent to the curve at P0 . Right: The derivative r0 (t)
defines a tangent vector to the curve at the point with the
position vector r(t). It also specifies the direction in which
r(t) traverses the curve with increasing t. T̂(t) is the unit
tangent vector.
The parametric equations of this line define the tangent line to the graph:
L: x = x0 +(t−t0 ) , y = y0 +f 0 (t0 )(t−t0 ) ⇒ y = y0 +f 0 (x0 )(x−x0 )
because x − x0 = t − t0 .
Definition 11.3. (Differential of a Vector Function).
Let r(t) be a differentiable vector function. Then the vector
dr(t) = r0 (t) dt
is called the differential of r(t).
In particular, the derivative is the ratio of the differentials, r0 (t) = dr/dt.
Recall that the differential dt is an independent variable that describes in-
finitesimal variations of t such that higher powers of dt can be neglected. In
this sense the definition of the differential is the linearization of Eq. (11.2) in
dt = h (when terms decreasing to zero faster than h are neglected). At any
particular t = t0 , the differential dr(t0 ) = r0 (t0 )dt 6= 0 defines the tangent
line
L(t) = r(t0 ) + dr(t0 ) = r(t0 ) + r0 (t0 )dt, t = t0 + dt .
Thus, the differential dr(t) at a point of the curve r(t) is the increment of
the position vector along the line tangent to the curve at that point.
11.3. Geometrical Significance of the Derivative. Among all the lines

through a particular point r(t0 ) of the curve traversed by a vector function
r(t), the line parallel to the derivative r0 (t0 ) has been shown to approximate
best the curve near r(t0 ). Let P0 and Ph have position vectors r(t0 ) and
r(t0 + h). Then
−−−→
P0 Ph = r(t0 + h) − r(t0 )
−−−→
is a secant vector. As h → 0, the direction of the vector P0 Ph becomes
tangential to the curve as depicted in Figure 11.1. On the other hand, it
follows from (11.2) that, for small enough h = dt,
−−−→
P0 Ph = dr(t0 ) = r0 (t0 )h , h → 0 .
This qualitative geometrical consideration justifies that the line through
P0 and parallel to the derivative r0 (t0 ) 6= 0 was called tangent. So, the
geometrical significance of the derivative of a vector function is that the
vector r0 (t) 6= 0 is tangent to the curve traversed by the vector function r(t).
The direction of the tangent vector also defines the orientation of the curve,
i.e., the direction in which the curve is traced out by r(t).
Example 11.2. Find the line tangent to the curve r(t) =
h2t, t2 − 1, t3 + 2ti at the point P0 (2, 0, 3).
Solution: By the geometrical property of the derivative, a vector parallel
to the line is v = r0 (t0 ), where t0 is the value of the parameter t at which
r(t0 ) = h2, 0, 3i is the position vector of P0 . Therefore, t0 = 1. Then
v = r0 (1) = h2, 2t, 3t2 + 2i = h2, 2, 5i .

t=1
Parametric equations of the line through P0 = (2, 0, 3) and parallel to v are

x = 2 + 2t , y = 2t , z = 3 + 5t .

11.4. Smooth curves. Let r(t) be a parameterization of a curve C. Can a

tangent line be defined at a point of C where the derivative r0 (t) vanishes?
It turns out that the answer depends on intrinsic geometrical properties of
the curve C and not so much on a particular parameterization of C that
happens to have the zero derivative at a particular point.
Suppose first that the derivative r0 (t) exists and does not vanish. Then,
at any point of the curve traced out by r(t), a unit tangent vector can be
defined by
r0 (t)
T̂(t) = 0 .
kr (t)k
In Section 10.3, spatial curves were identified with continuous vector func-
tions. Intuitively, a smooth curve as a point set in space should have a unit
tangent vector that is continuous along the curve. Recall also that, for any
curve as a point set in space, there are many vector functions whose range
coincides with the curve.
Definition 11.4. (Smooth Curve).

A point set C in space is called a smooth curve if C is a curve for which
there exists a simple, continuously differentiable parameterization r(t) whose
derivative does not vanish and, if the curve is closed r(a) = r(b), then in
addition r0 (a) = sr0 (b) for some s > 0.
Clearly, a smooth curve always has a tangent line at any point. It is its
characteristic property. The tangent line is parallel to the unit tangent vector
at that point. A smooth parametric curve r(t) is oriented by the direction
of the unit tangent vector T̂(t). If a smooth curve is closed r(a) = r(b), then
the unit tangent vectors T̂(a) and T̂(b) must coincide, which is guaranteed
by the condition r0 (a) = sr0 (b) for some s > 0. Indeed,
1 s s
T̂(a) = 0 r0 (a) = 0
r0 (b) = T̂(b) = T̂(b) if s > 0 .
kr (a)k ksr (b)k |s|
If s < 0, then T̂(a) = −T̂(b) and the unit tangent vector is discontinuous.
How does one determine whether a given parametric curve is smooth or
not? Note that if r0 (t) is continuous and never vanishes, then T̂(t) is contin-
uous. In particular, with the definition above, a smooth curve does indeed
have a continuous unit tangent vector. Therefore, if a curve does not have a
continuous unit tangent vector, it cannot be smooth. This enables us to con-
clude that some curves are not smooth, based on properties deduced from
a single parameterization. This is important because one cannot possibly
test all parameterizations to see whether one of them meets the conditions
in Definition 11.4. The following example illustrates this concept.
Example 11.3. Investigate whether the planar curves traversed by the
vector functions r1 (t) = ht3 , t2 i and r2 (t) = ht3 , t5 i are smooth.
Solution: The first component of each vector function x(t) = t3 is one-to-
one for all t (x(t1 ) = x(t2 ) implies t1 = t2 ). Therefore both the curves are
simple. The vector functions are continuously differentiable everywhere,
r01 (t) = h3t2 , 2ti , r02 (t) = h3t2 , 5t2 i ,
but both the derivatives vanish at the origin, r01 (0) = r02 (0) = 0. The unit
tangent vectors T̂1 (0) and T̂2 (0) are not defined at t = 0 for the given
parameterizations. Let us investigate the left and right limits of the unit
tangent vectors at t = 0. One has
kr01 (t)k = (9t4 + 4t2 )1/2 = 2|t|(1 + 9t2 /4)1/2 .
Therefore
1 1 D 3t2 t E
0
T̂1 (t) = r (t) = , .
kr01 (t)k 1 1 + 9t2 /4 2|t| |t|
p
If t > 0, then t/|t| = 1 and, if t < 0, then t/|t| = −1. Hence,

lim T̂1 (t) = h0, 1i , lim T̂1 (t) = h0, −1i .
t→0+ t→0−
This shows that the unit tangent vector is discontinuous at the point r(0) =
0 and the curve is not smooth. To visualize the curve near this point, let us
solve the equation x = t3 for t, t = x1/3 , and substitute the latter into y = t2
to obtain y = x2/3 . So the curve traversed by r1 (t) is the graph y = x2/3 ,
which has a cusp at x = 0. (The graph lies in the positive half-plane y ≥ 0
and approaches the y axis tangentially, forming a horn-like shape at the
origin.) It should be stressed that the presence of a cusp was established
from a singe (given) parameterization of the curve.
For the second curve, a similar analysis shows that
kr02 (t)k = (9t4 + 25t8 )1/2 = 3t2 (1 + 25t4 /9)1/2 ,

1 1 D 5t2 E
T̂2 (t) = 0 r02 (t) = p 1, ,
kr2(t)k 1 + 25t4 /9 3
lim T̂2 (t) = lim T̂2 (t) = h1, 0i .
t→0+ t→0−
The latter property allows us to define the unit tangent at the point where
the derivative vanishes by T̂2 (0) = h1, 0i so that T̂2 (t) is continuous every-
where (in particular, limt→0 T̂2 (t) = T̂2 (0)) and, hence, the curve is smooth.
The fact that the derivative of a vector function that traverses a smooth
curve vanishes at some point of the curve does not contradict Definition
11.4 because there exists another vector function R(s) = hs, s5/3i that has
the same range since R(t3 ) = r2 (t) and whose derivative is continuous and
never vanishes, R0 (s) = h1, 5s2/3/3i =6 0 for all s. It is easy to see that the
5/3
second curve is the graph y = x which has no cusp at x = 0 (it is tangent
to the x axis and has an inflection point at x = 0).
Problem 11.1. Determine whether the cycloid x = a(t − sin t),

y = a(1 − cos t) is smooth, where a > 0 is a parameter. If it is not smooth at
particular points, investigate its behavior near those points by approximating
the cycloid by the graph of a power function.
Solution: The existence of a continuous unit tangent vector has to be

verified. Let r(t) = hx(t), y(t)i. Since x0 (t) = a(1 − cos t) ≥ 0 for all t, and
x0 (t) = 0 only when t is a multiple of 2π, x(t) is monotonically increasing.
In particular, x(t) is one-to-one, so C is simple. Since y 0 (t) = a sin t, the
derivatives x0 (t) and y 0 (t) vanish simultaneously if and only if t = 2πn for
some integer n. Thus, r0 (t) 6= 0 unless t = 2πn, so C is smooth except
possibly at the points r(2πn) = h2πna, 0i; that is, the portion of C between
two consecutive such points is smooth, but it is not yet known whether C
is smooth at those points. Since
q
kr0 (t)k = a 2(1 − cos t) = a 4 sin2 (t/2) = 2a| sin(t/2)| ,
p
the components of the unit tangent vector for t 6= 2πn are

x0 (t) y 0 (t) sin t
T1 (t) = 0 = | sin(t/2)| , T2 (t) = 0 = .
kr (t)k kr (t)k 2| sin(t/2)|
Owing to the periodicity of the sine and cosine functions, it is sufficient to
investigate the point corresponding to t = 0. If there exists a continuous
unit tangent vector, then the limit limt→0 T̂(t) should exist and be the unit
tangent vector at the point corresponding to t = 0. By Theorem 10.1, the
existence of the limits of the components T1 (t) and T2 (t) as t → 0 has to be
verified. Evidently, T1 (t) → 0 as t → 0, but the limit limt→0 T2 (t) does not
exist. Indeed, by the trigonometric identity sin t = 2 sin(t/2) cos(t/2) the
left and right limits are different:
2 sin(t/2) cos(t/2)
lim T2 (t) = lim = lim cos(t/2) = 1 ,
t→0+ t→0+ 2 sin(t/2) t→0+
2 sin(t/2) cos(t/2)
lim T2 (t) = lim = − lim cos(t/2) = −1 .
t→0 − t→0 − −2 sin(t/2) t→0−
Therefore the left and right limits of the unit tangent vector do not coincide
at t = 0:
lim T̂(t) = h0, 1i = 6 h0, −1i = lim T̂(t) .
t→0+ t→0−
This means that T̂(t) cannot be continuously extended across the point
(0, 0), so C is not smooth there (as well as at (2πn, 0)).
A local behavior of the cycloid near (0, 0) may be investigated by us-
ing the Taylor polynomial approximations near t = 0 of the trigonometric
functions involved:
sin t = t − t3 /6 + O(t5 ) , cos t = 1 − t2 /2 + O(t4 ) .
So neglecting terms t4 and higher for small t, the cycloid is approximated
by the curve
a a
x = a(t − sin(t)) ≈ t3 , y = a(1 − cos(t)) ≈ t2 .
6 2
Expressing t via x from the first equation and substituting it into the other
equation,
6x 1/3 a 6x 2/3 9a 1/3
t= ⇒ y= = cx2/3 , c = ,
a 2 a 2
it is concluded that near the point (0, 0) the cycloid behaves as the graph
y = cx2/3 , which has a cusp at (0, 0).
Problem 11.2. Prove that, for any smooth curve on a sphere, a tangent
vector at any point P is orthogonal to the vector from the sphere center to
the point P .
Solution: Let r0 be the position vector of the center of a sphere of radius
R. The position vector r of any point of the sphere satisfies the equation
kr − r0 k = R ⇒ (r − r0 ) · (r − r0 ) = R2
because kak2 = a · a for any vector a. Let r(t) be a vector function that
traces out a curve on the sphere. In particular, the parameterization of the
curve can be chosen so that r0 (t) 6= 0 (the curve is smooth). Then, for all
values of t,
(r(t) − r0 ) · (r(t) − r0 ) = R2 .
Differentiating both sides of the latter relation (put u(t) = v(t) = r(t) + r0
in the third equation in Theorem 11.1) and using the basic properties of the
dot product, one infers
r0 (t) · (r(t) − r0 ) + (r(t) − r0 ) · r0 (t) = 0 ,
2 r0 (t) · (r(t) − r0 ) = 0 ,
r0 (t) · (r(t) − r0 ) = 0 ⇐⇒ r0 (t) ⊥ r(t) − r0 .
If r(t) is the position vector of P and O is the center of the sphere, then
−−→
OP = r(t) − r0 , and hence the tangent vector r0 (t) at P is orthogonal to
−−→
OP for any t or at any point P of the curve.
Problem 11.3. Let u(t) be differentiable and ku(t)k = k where k is a
constant. Show that u(t) and u0 (t) are orthogonal. Use the result to find
the unit tangent vector as a function of t to the exponential helix r(t) =
het cos t, et sin t, et i. Is the curve smooth?
Solution: Since ku(t)k2 = u(t) · u(t), by differentiating the equation
u(t) · u(t) = k2 , one infers that the derivative u0 (t) is orthogonal to u(t)
(see the third equation in Theorem 11.1):
0
u(t) · u(t) = 0 ⇒ 2u0 (t) · u(t) = 0 ⇔ u0 (t) ⊥ u(t) .
The vector function r(t) can be written in the form

r(t) = et u(t) + et ê3 , u(t) = hcos t, sin t, 0i , ku(t)k = 1 .
By the rules of differentiation (see the second equation in Theorem 11.1):

r0 (t) = et u(t) + et u0 (t) + et ê3 = et u(t) + u0 (t) + ê3 .
The vectors u(t), u0 (t) = h− sin t, cos t, 0i,√and ê3 are mutually orthogonal
unit vectors so that ku(t) + u0 (t) + ê3 k = 3 (the length of the diagonal of
a cube with edges of unit length). Therefore
√ 1 1
kr0 (t)k = 3 et ⇒ T̂(t) = 0 r0 (t) = √ u(t) + u0 (t) + ê3
kr (t)k 3
1
= √ hcos t − sin t, sin t + cos t, 1i .
3
The reader is advised to compare the technicalities involved if T̂(t) is com-
puted by differentiating r(t) component-wise. The function T̂(t) is contin-
uous because its components are continuous. So the curve is smooth.
11.6. Exercises.
1–6. Find the derivatives and differentials of each of the following vector
functions:
1. r(t) = h1, 1 + t, 1 + t3 i ;
2. r(t) = hcos t, sin2 (t), t2 i ;
3. r(t) = hln(t),
√ e2t√, te−t i ;
4. r(t) = h t − 2, t2 − 4, ti ;
3
5. r(t) = a + bt2 − cet ;

6. r(t) = ta × (b − cet ) .
7. Sketch the curve traversed by the vector function r(t) = h2, t − 1, t2 + 1i.
Indicate the direction in which the curve is traversed by r(t) with increasing
t. Sketch the position vectors r(0), r(1), r(2) and the vectors r0 (0), r0 (1),
r0 (2). Repeat the procedure for the vector function R(t) = r(−t) = h2, −t−
1, t2 + 1i for t = −2, −1, 0.
8–12. Determine if the curve traced out by each of the following vector
functions is smooth for a specified interval of the parameter. If the curve is
not smooth at a particular point, graph it near that point.
8. r(t) = ht, t2 , t3 i, 0 ≤ t ≤ 1 ;
9. r(t) = ht2 , t3 , 2i, −1 ≤ t ≤ 1 ;
10. r(t) = ht1/3 , t, t3 i, −1 ≤ t ≤ 1 ;
11. r(t) = ht5 , t3 , t4 i, −1 ≤ t ≤ 1 ;
12. r(t) = hsin3 t, 1, t2 i, −π/2 ≤ t ≤ π/2 .
13. Determine whether a cardioid described by the polar graph r = 1 −
a cos θ, 0 ≤ θ ≤ 2π, is smooth, where |a| ≤ 1. Sketch the cardioid for a = 0,
a = ±1/2, and a = ±1. Hint: to find parametric equations of the cardioid,
use the relations between the polar and rectangular coordinates and θ as a
parameter.
14–15. Find the parametric equations of the tangent line to each of the
following curves at a specified point:
14. r(t) = ht2 − t,√ t3 /3, 2ti, P0 = (6, 9, 6);
15. r(t) = hln t, 2 t, t2 i, P0 = (0, 2, 1) .
16–17. Find the unit tangent vector to the curve traversed by the specified
vector function at the given point P0 :
16. r(t) = h2t + 1, 2 tan−1 t, e−t i, P0 (1, 0, 1) ; √
17. r(t) = hcos(ωt), cos(3ωt), sin(ωt)i, P0 (1/2, −1, 3/2) , where ω
is a positive constant.
18. Find r0 (t) · r00 (t) and r0 (t) × r00 (t) if r(t) = ht, t2 − 1, t3 + 2i.
19. Is there a point on the curve r(t) = ht2 − t, t3 /3, 2ti at which the
tangent line is parallel to the vector v = h−5/2, 2, 1i? If so, find the point.
20. Let r(t) = het , 2 cos t, sin(2t)i. Use the best linear approximation
L(t) near t = 0 to estimate r(0.2). Use a calculator to assess the accuracy
kr(0.2)−L(0.2)k of the estimate. Repeat the procedure for r(0.7) and r(1.2).
Compare the errors in all three cases.
21. Find the point of intersection of the plane y + z = 3 and the curve
r(t) = hln t, t2 , 2ti. Find the angle between the normal of the plane and
the tangent line to the curve at the point of intersection.
22. Does the curve r(t) = h2t2 , 2t, 2−t2 i intersect the plane x+y +z = −3?
If not, find a point on the curve that is closest to the plane. What is the
distance between the curve and the plane. Hint: Express the distance be-
tween a point on the curve and the plane as a function of t, then solve the
extreme value problem.
23. Find the point of intersection of two curves r1 (t) = ht, 1 − t, 3 + t2 i and
r2 (s) = h3 − s, s − 2, s2 i. If the angle at which two curves intersect is defined
as the angle between their tangent lines at the point of intersection, find the
angle at which the above two curves intersect.
24. State the condition under which the tangent lines to the curve r(t)
at two distinct points r(t1 ) and r(t2 ) are intersecting, or skew, or parallel.
Let r(t) = h2 sin(πt), cos(πt), sin(πt)i, t1 = 0, and t2 = 1/2. Determine
whether the tangent lines at these points are intersecting and, if so, find the
point of intersection.
25. Suppose a smooth curve r(t) does not intersect a plane through a point
P0 and orthogonal to a vector n. Assume that, among the points on the
curve, there is one that is closest to the plane. What is the angle between n
and a tangent vector to the curve at the point that is the closest to the plane?
26. Suppose r(t) is twice differentiable. Show that (r(t) × r0 (t))0 =
r(t) × r00(t).
27. Suppose that r(t) is differentiable three times. Show that
0 00 0 0 000
[r(t) · (r (t) × r (t))] = r(t) · (r (t) × r (t)).
28. Let r(t) be a differentiable vector functions. Show that
(kr(t)k)0 = r(t) · r0 (t)/kr(t)k at all t for which r(t) 6= 0.
29. A space warship can fire a laser cannon forward along the tangent
line to its trajectory. If the trajectory is traversed by the vector function
r(t) = ht, t, t2 +4i in the direction of increasing t and the target is the sphere
x2 + y 2 + z 2 = 1, find the part of the trajectory in which the laser cannon
can hit the target. Hint: If a line L is tangent to the trajectory at t = t0 ,
then the target is hit when the distance between L and the origin is less or
equal 1. State this geometrical condition as an algebraic condition on t0 .
To solve this algebraic condition, show that the trajectory is a parabola in
the plane y = x. So, find points on the parabola at which its tangent is at
a distance less or equal 1 to the origin.
30–32. A plane normal to a curve at a point P0 is the plane through P0
whose normal is tangent to the curve at P0 . For each of the following curves
find a suitable parameterization, the tangent line, and the normal plane at
a specified point:
30. y = x, z = x2 , P0 = (1, 1, 1) ;
31. x2 + z 2 = 10, y 2 + z 2 = 10, P0 = (1, 1, 3) ;
32. x2 + y 2 + z 2 = 6, x + y + z = 0, P0 = (1, −2, 1) ;
33. Show that tangent lines to a circular helix have a constant angle with
the axis of the helix.
34. Consider a line through the origin. Any such line sweeps a circular cone
when rotated about the z axis and, for this reason, is called a generating
line of a cone. Prove that the curve r(t) = (et cos t, et sin t, et ) intersects all
generating lines of the cone x2 + y 2 = z 2 at the same angle. Hint: Show
that parametric equations of a line in the cone are x = s cos θ, y = s sin θ,
z = s. Define the points of intersection of the line and the curve and find
the angle at which they intersect.
12. INTEGRATION OF VECTOR FUNCTIONS 185
12. Integration of Vector Functions

Definition 12.1. (Definite Integral of a Vector Function).
Let r(t) be defined on the interval [a, b]. The vector whose components are the
definite integrals of the corresponding components of r(t) = hx(t), y(t), z(t)i
is called the definite integral of r(t) over the interval [a, b] and denoted as
Z b DZ b Z b Z b E
(12.1) r(t) dt = x(t) dt , y(t) dt , z(t) dt .
a a a a
If the integral (12.1) exists, then r(t) is said to be integrable on [a, b].
By this definition, a vector function is integrable if and only if all its
components are integrable functions. Recall that a continuous real-valued
function is integrable. Therefore, the following theorem holds.
Theorem 12.1. If a vector function is continuous on the interval [a, b],
then it is integrable on [a, b].
Example 12.1. Find the integral of r(t) = ht/π, sin t, cos ti over the
interval [0, π].
Solution: The components of r(t) are continuous on [0, π]. Therefore, by
the fundamental theorem of calculus,
Z π DZ π Z π Z π E
r(t)dt = (t/π)dt, sin t dt, cos t dt
0 0 0 0
D t2 π π πE
= , − cos t , sin t = hπ/2, 2, 0i .
2π 0 0 0

Definition 12.2. (Indefinite Integral of a Vector Function).
A vector function R(t)Ris called an antiderivative of r(t) if R0 (t) = r(t).
The indefinite integral r(t)dt of a vector function r(t) is the collection of
all antiderivatives R(t) of r(t).
The indefinite integral of r(t) can also be viewed as the most general
vector function whose derivative is r(t). Let R(t) = hX(t), Y (t), Z(t)i be an
antiderivative of r(t) = hx(t), y(t), z(t)i on an interval. Then the functions
X(t), Y (t), and Z(t) are antiderivatives of x(t), y(t), and z(t), respectively.
Recall from Calculus I that two functions that have derivatives equal on an
interval can differ at most by a constant function on this interval. Therefore
Z Z Z
x(t) dt = X(t) + +c1 , y(t) dt = Y (t) + c2 , z(t) dt = Z(t) + c3 ,
where c1 , c2 , and c3 are constants. The latter relations can be combined

into a single vector relation:
Z
r(t) dt = R(t) + c ,
where c is an arbitrary constant vector.

Recall further from Calculus I that for a function x(t) continuous on

[a, b], its particular antiderivative is given by
Z t
X(t) = x(u)du, a ≤ t ≤ b ,
a
and X(t) is differentiable on (a, b) and satisfies the condition X(a) = 0.

Therefore a particular antiderivative of a continuous vector function r(t) is
Z t
R(t) = r(u)du, a ≤ t ≤ b .
a
The vector function R(t) is differentiable on (a, b) and satisfies the condi-
tion R(a) = 0. The indefinite integral on [a, b] is obtained by adding a
constant vector, R(t) → R(t) + c. This observation allows us to extend the
fundamental theorem of calculus to vector functions.
Theorem 12.2. (Fundamental Theorem of Calculus for Vector Functions).
If r(t) is continuous on [a, b], then
Z b
r(t) dt = R(b) − R(a)
a
where R(t) is any antiderivative of r(t), that is, a vector function such that
R0 (t) = r(t).
Example 12.2. Find r(t) if r0 (t) = h2t, 1, 6t2i and r(1) = h2, 1, 0i.
Solution: Taking the antiderivative of r0 (t), one finds
Z
r(t) = h2t, 1, 6t2 i dt = ht2 , t, 3t3 i + c.
The constant vector c is determined by the condition r(1) = h2, 1, 0i, which
gives
h1, 1, 3i + c = h2, 1, 0i ⇒ c = h2, 1, 0i − h1, 1, 3i = h1, 0, −3i ,
and, hence, r(t) = ht2 + 1, t, 3t3 − 3i.
In general, the solution of the equation r0 (t) = v(t) satisfying the con-
dition r(t0 ) = r0 can be written in the form
Z t
r0 (t) = v(t) and r(t0 ) = r0 ⇒ r(t) = r0 + v(u) du
t0
if v(t) is a continuous vector function on an interval (t0 lies in the in-

terval of continuity). As noted above, if the integrand is a continuous
function on an interval, then the derivative of the integral with respect
to its upper limit
R t is the value of the integrand at that limit. Therefore,
r0 (t) = (d/dt) t0 v(u) du = v(t), and hence r(t) is an antiderivative of v(t).
When t = t0 , the integral vanishes and r(t0 ) = r0 as required.
12.1. Applications to Mechanics. Let r(t) be the position vector of a par-

ticle as a function of time t. The first derivative
r0 (t) = v(t)
is called the velocity of the particle. The magnitude of the velocity vector
v(t) = kv(t)k is called the speed. The speed of a car is a number shown on
the speedometer. The velocity defines the direction in which the particle
travels and the instantaneous rate at which it moves in that direction. The
second derivative
r00(t) = v 0 (t) = a(t)
is called the acceleration. If m is the mass of a particle and F is the force act-
ing on the particle, then according to Newton’s second law, the acceleration
and force are related as
F = ma .
If the time is measured in seconds, the length in meters, and the mass in
kilograms, then the force is given in newtons, 1 N = 1 kg · m/s2 .
If the force is known as a vector function of time, then Newton’s second
law determines a particle’s trajectory. The problem of finding the trajectory
amounts to reconstructing the vector function r(t) if its second derivative
1
r00 (t) = F(t)
m
is known; that is, r(t) is given by a second antiderivative of (1/m)F(t). In-
deed, the velocity v(t) is an antiderivative of (1/m)F(t), and the position
vector r(t) is an antiderivative of the velocity v(t). As shown in the previ-
ous section, an antiderivative is not unique, unless its value at a particular
point is specified. So the trajectory of motion is uniquely determined by
Newton’s equation, provided the position and velocity vectors are specified at
a particular moment of time, for example,
r(t0 ) = r0 , v(t0 ) = v0 .
The latter conditions are called initial conditions. Given the initial condi-
tions, the trajectory of motion is uniquely defined by the relations:
1 t
Z Z t
(12.2) v(t) = v0 + F(u) du , r(t) = r0 + v(u) du
m t0 t0
if the force is a continuous vector function of time. If the force is piecewise
continuous, then the initial value problem has to be solved for each interval
of continuity. The values of r(t) and v(t) at the end point of a preceding
interval of continuity serve as the initial values for the next adjacent interval
of continuity.
Remark. If the force is a function of a particle’s position, F = F(r(t)), then
the Newton’s equation becomes a system of ordinary differential equations
that is a set of some relations between components of the vector function
r(t), its derivatives, and time. Solving the initial value problem for a system
of differential equations requires a special technique to integrate such equa-

tions. Sometimes such problems can be solved by simpler means. Examples
are given in Study Problems 12.3, 15.4, and 15.6.
Example 12.3. (Motion Under a Constant Force).
Prove that the trajectory of motion under a constant force is a parabola if
the initial velocity is not parallel to the force.
Solution: Let F be a constant force. Without loss of generality, the initial
conditions can be set at t = 0, r(0) = r0 , and v(0) = v0 . Then
1 t t
Z
v(t) = v0 + Fdu = v0 + F ,
m 0 m
Z t
t2
r(t) = r0 + v(u)du = r0 + tv0 + F.
0 2m
If the vectors v0 and F are parallel, then they are proportional, v0 = cF.
In this particular case, the trajectory r(t) = r0 + (ct + t2 /(2m))F = r0 + sF
lies in the straight line through r0 and parallel to F. The parameter s =
ct + t2 /(2m) defines the position of the particle on the line as a function of
time. Otherwise, the vector r(t) − r0 is a linear combination of two non-
parallel vectors v0 and F and hence must be orthogonal to n = F×v0 by the
geometrical property of the cross product. Therefore, the particle remains
in the plane through r0 that is parallel to F and v0 or orthogonal to n:
(r(t) − r0 ) · n = 0 , n = F × v0
(see Figure 12.1, left panel). The shape of a space curve does not depend
on the choice of the coordinate system. Let us choose the coordinate system
such that the origin is at the initial position r0 and the plane in which the
trajectory lies coincides with the zy plane so that F is parallel to the z axis.
In this coordinate system,
r0 = h0, 0, 0i , F = h0, 0, −F i , v0 = h0, v0y , v0z i .
The parametric equations of the trajectory of motion assume the form
x = 0, y = v0y t , z = v0z t − t2 F/(2m) .
The substitution of t = y/v0y from the second equation into the third equa-
tion yields
2
F v0y v0z
z=− y2 + y,
2m v0y
which defines a parabola in the zy plane. Thus, the trajectory of motion un-
der a constant force is a parabola through the point r0 that lies in the plane
containing the force and initial velocity vectors F and v0 . The parabola is
concave in the direction of the force. In Figure 12.1, the force vector points
downward and the trajectory is concave down.
Figure 12.1. Left: Motion under a constant force F. The

trajectory is a parabola that lies in the plane through the
initial point of the motion r0 and orthogonal to the vector
n = F × v0 , where the initial velocity v0 is assumed to be
non-parallel to the force F. Right: Motion of a projectile
thrown at an angle θ and an initial height h. The trajectory
is a parabola. The point of impact defines the range L(θ).
12.2. Motion Under a Constant Gravitational Force. The magnitude of

the gravitational force that acts on an object of mass m near the surface
of the Earth is mg, where g ≈ 9.8 m/s2 is a universal constant called the
acceleration of a free fall. According to Example 12.3, any projectile fired
from some point follows a parabolic trajectory. This fact allows one to
predict the exact positions of the projectile and, in particular, the point at
which it impacts the ground. In practice, the initial speed v0 of the projectile
and angle of elevation θ at which the projectile is fired are known (see
Figure 12.1, right panel). Some practical questions are: At what elevation
angle is the maximal range reached? At what elevation angle does the range
attain a specified value (e.g., to hit a target)?
To answer these and related questions, choose the coordinate system
such that the z axis is directed upward from the ground and the parabolic
trajectory lies in the zy plane. The projectile is fired from the point (0, 0, h),
where h is the initial elevation of the projectile above the ground (firing from
a hill). In the notation of Example 12.3, F = h0, 0, −mgi (the z component
of F is negative because the gravitational force is directed toward the ground,
while the z axis points upward), v0y = v0 cos θ, and v0z = v0 sin θ. The
trajectory is
1
y = tv0 cos θ , z = h + tv0 sin θ − gt2 , t ≥ 0 .
2
It is interesting to note that the trajectory is independent of the mass of
the projectile. Light and heavy projectiles would follow the same parabolic
trajectory, provided they are fired from the same position, at the same speed,
and at the same angle of elevation. The height of the projectile relative to
the ground is given by z(t). The horizontal displacement is y(t). Let tL > 0
be the moment of time when the projectile lands; that is, when t = tL , the
height vanishes, z(tL ) = 0. A positive solution of this equation is
q
v0 sin θ + v02 sin2 θ + 2gh
tL = .
g
The distance L traveled by the projectile in the horizontal direction until it
lands is the range:
L = y(tL ) = tL v0 cos θ .
Consider the case when the projectile is fired from the ground, h = 0.
Then the travel time of the projectile and the range are:
2v0 sin θ v02 sin(2θ)
h=0 ⇒ tL = , L= .
g g
The range attains its maximal value v02 /g when the projectile is fired at an
angle of elevation θ = π/4. The angle of elevation at which the projectile
hits a target at a given range L = L0 is
1 L g
0
θ = sin−1 .
2 v02
Note that this relation makes sense only if L0 g/v02 ≤ 1. It is impossible to
hit the target at a range that exceeds the maximal range, L0 > v02 /g.
If h 6= 0, the angle at which L = L(θ) attains its maximal values can be
found by solving the equation L0 (θ) = 0, which defines critical points of the
function L(θ). The angle of elevation at which the projectile hits a target at
a given range is found by solving the equation L(θ) = L0 . The technicalities
are left to the reader.
Remark. In reality, the trajectory of a projectile deviates from a parabola
because there is an additional force acting on a projectile moving in the
atmosphere, the friction force. The friction force depends on the velocity of
the projectile. A wind creates an additional force, a drag force (a projec-
tile is dragged in the direction of the wind). So a more accurate analysis
of the projectile motion in the atmosphere requires methods of differential
equations.

Problem 12.1. The acceleration of a particle is a = h2, 6t, 0i. Find the
position vector of the particle and its velocity in two units of time t if the
particle was initially at the point (−1, −4, 1) and had the velocity h0, 2, 1i.
Solution: The velocity vector is
Z
v(t) = a(t) dt = h2t, 3t2 , 0i + c .
The constant vector c is fixed by the initial condition v(0) = h0, 2, 1i, which
yields c = h0, 2, 1i. Thus, v(t) = h2t, 3t2 + 2, 1i and v(2) = h4, 14, 1i. The
position vector is
Z
r(t) = v(t) dt = ht2 , t3 + 2t, ti + c .
Here the constant vector c is determined by the initial condition r(0) =

h−1, −4, 1i, which yields c = h−1, −4, 1i. Thus, r(t) = ht2 −1, t3 +2t−4, t+1i
and r(2) = h3, 8, 3i. Aternatively, the problem can also be solved using
Eqs. (12.2):
Z t Z t
v(t) = h0, 2, 1i + a(u)du = h0, 2, 1i + h2, 6u, 0idu = h2t, 3t2 + 2, 1i ,
0 0
Z t
r(t) = h−1, −4, 1i + v(u)du = ht2 − 1, t3 + 2t − 4, t + 1i .
0

Problem 12.2. Show that if the velocity and position vectors of a particle
remains orthogonal during the motion, then the trajectory lies on a sphere.
Solution: If v(t) = r0 (t) and r(t) are orthogonal, then
r0 (t) · r(t) = 0
for all t. On the other hand
(r(t) · r(t))0 = r0 (t) · r(t) + r(t) · r0 (t) = 2r0 (t) · r(t) = 0 .
Therefore r(t) · r(t) = R2 = const or kr(t)k = R for all t; that is, the particle
remains at a fixed distance R from the origin all the time.
Problem 12.3. A charged particle moving in a magnetic field B is subject
to the Lorentz force F = (e/c)v × B, where e is the electric charge of the
particle and c is the speed of light in vacuum. Assume that the magnetic
field is a constant vector parallel to the z axis and the initial velocity is
v(0) = hv⊥ , 0, vki (here v⊥ and vk are the components of the initial velocity
in the direction perpendicular and parallel to the magnetic field, respectively).
Show by verifying Newton’s second law mr00(t) = F(t) that the trajectory is
a helix:
eB v⊥
r(t) = hR sin(ωt), R cos(ωt), vk ti, ω = , R= ,
mc ω
where B = kBk is the magnitude of the magnetic field and m is the particle
mass.
Solution: Newton’s second law reads
e
mv 0 = v × B.
c
Put B = h0, 0, Bi. Then

v = r0 = hωR cos(ωt), −ωR sin(ωt), vk i ,
v × B = h−ωRB sin(ωt), −ωRB cos(ωt), 0i,
ω
v 0 = h−ω 2 R sin(ωt), −ω 2 R cos(ωt), 0i = v ×B.
B
The substitution of these relations into Newton’s second law yields
e mω e eB
mv 0 = v × B ⇒ v×B = v×B ⇒ ω = .
c B c mc
Since v(0) = hωR, 0, vki = hv⊥ , 0, vki, it follows that R = v⊥ /ω.
Remark. The rate at which the helix rises along the magnetic field is
determined by the magnitude (speed) of the initial velocity component vk
parallel to the magnetic field, whereas the radius of the helix is determined
by the magnitude of the initial velocity component v⊥ perpendicular to the
magnetic field. A particle makes one full turn about the magnetic field in
time T = 2π/ω = 2πmc/(eB), that is, the larger the magnetic field, the
faster the particle rotates about it. If vk = 0, then a charged particle is
trapped by a magnetic field (it remains on a circular orbit). This effect is
used in a device called a cyclotron to trap charged particles in a magnetic
field. The magnetic field is made so that it has a constant magnitude but its
direction changes slowly in space, remaining tangential to a circle. Even if
vk 6= 0, the particle is winding about this circle and thus remains trapped is
a spatial region that looks like a doughnut (think of a particle moving along
a helix whose axis is a circle of radius that is larger than the radius of the
helix). An accurate mathematical description of the motion of a charged
particle in a cyclotron would require the methods of differential equations
because the magnetic field is a function of position in space. This “trapping”
effect of a magnetic field is responsible for a beautiful natural phenomenon
known as polar lights (see Study Problem 15.4).
12.4. Exercises.
1–7. Find the indefinite and definite integrals over specified intervals for
each of the following functions:
1. r(t) = h1, 2t, 3t2 i, 0 ≤ t ≤ 2 ;
2. r(t) = hsin t,√t3 , cos ti,
√ −π ≤ t ≤ π ;
2
3. r(t) = ht , t 1 − t, ti, 0 ≤ t ≤ 1 ;
4. r(t) = ht ln t, t2 , e2t i, 0 ≤ t ≤ 1 ;
5. r(t) = h2 sin t cos t, 3 sin t cos2 t, 3 sin2 t cos ti, 0 ≤ t ≤ π/2 ;
6. r(t) = a + cos(t)b, 0 ≤ t ≤ π, a and b are constant vectors ;
7. r(t) = a × (u0 (t) + b), 0 ≤ t ≤ 1, if u0 (t) is continuous and
u(0) = a and u(1) = a − b .
8–11. Find r(t) if the derivatives r0 (t) and r(t0 ) are given:
8. r0 (t) = h1, 2t, 3t2 i, r(0) = h1, 2, 3i;
√
9. r0 (t) = ht − 1, t2 , ti, r(1) = h1, 0, 1i ;
10. r0 (t) = hsin(2t) , 2 cos t, sin2 ti, r(π) = h1, 2, 3i;
11. r0 (t) = h2t, et , 4t3 i , r(0) = h1, 3, 0i .
12–14. Find the solution r(t) of each of the following initial value problems:
12. r00 (t) = h0, 2, 6ti, r(0) = h1, 2, 3i, r0 (0) = h1, 0, −1i ;
13. r00 (t) = ht1/3 , t1/2 , 6ti, r(1) = h1, 0, −1i, r0 (0) = h1, 2, 0i;
14. r00 (t) = h− sin t, cos t, 1/ti, r(π) = h1, −1, 0i, r0 (π) = h−1, 0, 2i .
15. Solve the equation r00(t) = a where a is a constant vector if r(0) = b
and r(t0 ) = c for some t = t0 6= 0.
16. Find the most general vector function whose nth derivative vanishes,
r(n)(t) = 0, in an interval.
17. Show that a continuously differentiable vector function r(t) satisfying
the equation r00 (t) × r0 (t) = 0, where r0 (t) is never zero, traverses a straight
line (or a part of it).
18. If a particle was initially at point (1, 2, 1) and had velocity v = h0, 1, −1i,
find the position vector of the particle after it has been moving with accel-
eration a(t) = h1, 0, ti for 2 units of time.
19. A particle of unit mass moves under a constant force F. If a particle
was initially at the point r0 and passed through the point r1 after 2 units of
time, find the initial velocity of the particle. What was the velocity of the
particle when it passed through r1 ?
20. A particle of mass of 1 kg was initially at rest. Then during 2 seconds a
constant force of magnitude of 3 N was applied to the particle in the direc-
tion of h1, 2, 2i. How far is the particle from its initial position in 4 seconds?
21. The velocity of a particle is v(t) = h2t, 5, 2t − 16i. Find its position r(t)
when the speed of the particle is minimal if r(0) = 0.
22. A projectile is fired from the ground at an initial speed of 400 m/s and at
an angle of elevation of 30◦ . Find the range of the projectile, the maximum
height reached, and the speed at impact.
23. A ball of mass m is thrown southward into the air at an initial speed
of v0 at an angle of θ to the ground. An east wind applies a steady force of
magnitude F to the ball in a westerly direction. Find the trajectory of the
ball. Where does the ball land and at what speed? Find the deviation of
the impact point from the impact point A when no wind is present. Is there
any way to correct the direction and the initial speed in which the ball is
thrown so that the ball still hits A? Is it possible to achieve the goal only
by adjusting the direction, while keeping the initial speed fixed?
24. A rocket burns its on-board fuel while moving through space. Let
v(t) and m(t) be the velocity and mass of the rocket at time t. It can
be shown that the force exerted by the rocket jet engines is m0 (t)vg , where
vg is the velocity of the exhaust gases relative to the rocket. Show that
v(t) = v(0) − ln(m(0)/m(t))vg. The rocket is to accelerate in a straight line
from rest to twice the speed of its own exhaust gases. What fraction of its
initial mass would the rocket have to burn as fuel?
25. The acceleration of a projectile is a(t) = h0, 2, 6ti. The projectile is shot
from (0, 0, 0) with an initial velocity v(0) = h1, −2, −10i. It is supposed to
destroy a target located at (2, 0, −12). The target can be destroyed if the
projectile’s speed is at least 3.1 at impact. Will the target be destroyed?
13. ARC LENGTH OF A CURVE 195
13. Arc Length of a Curve

Let C be a simple curve in space. Then there is a simple parameteriza-
tion r(t), a ≤ t ≤ b, of C. The vector function r(t) traverses the curve C and
establishes a one-to-one correspondence between points of C and points of
the interval [a, b] (except possibly for the case when C is closed, r(a) = r(b)),
i.e., the end points of the interval correspond to the same point of the curve
if the curve is closed). Consider a partition of the interval [a, b],
a = t0 < t1 < t2 < · · · < tN −1 < tN = b .
This partition induces a partition of the curve which is a collection of points

of C, Pk , k = 0, 1, ..., N , whose position vectors are r(tk ). In particular, P0
and PN are the end points of the curve (see Fig. 13.1 (left panel)). If the
curve is closed, then P0 = PN . Let DN = maxk (tk − tk−1 ) be the maximal
length among all the partition intervals. A partition is said to be refined
if DN 0 < DN for N 0 > N . Clearly it is possible to refine a partition so
that DN → 0 as N → ∞. Note that for each partition there is at least one
partition interval whose length is DN . Therefore a refinement is obtained by
adding a partition point in each partition interval whose length is DN (e.g.,
by adding the midpoint of each interval of length DN to the partition). So
by taking N large enough DN can be made smaller than any preassigned
positive number. So it will always be assumed that under a refinement of a
partition, DN → 0 as N → ∞.
Definition 13.1. (Arc Length of a Curve).

Let C be a simple curve in space. Let a collection of points Pk be a partition
of C, k = 0, 1, ..., N, and |Pk−1 Pk | be the distance between two neighboring
partition points (P0 = PN if the curve is closed). The arc length of a curve
C is the limit
N
X
L = lim |Pk−1 Pk |,
N →∞
k=1
where the partition is refined as N → ∞, provided it exists and is indepen-

dent of the choice of partition. If L < ∞, the curve is called measurable or
rectifiable.
The geometrical meaning of this definition is rather simple. Here the sum
of |Pk−1 Pk | is the length of a polygonal path with vertices at P0 , P1 ,..., PN
in this order. As the partition becomes finer and finer, this polygonal path
approaches the curve more and more closely (see Figure 13.1, left panel).
In certain cases, the arc length is given by the Riemann integral.
Figure 13.1. Left: The arc length of a curve is defined

as the limit of the sequence of lengths of polygonal paths
through partition points of the curve. Right: Natural pa-
rameterization of a curve. Given a point A of the curve, the
arc length s is counted from it to any point P of the curve.
The position vector of P is a vector R(s). If the curve is
traced out by another vector function r(t), then there is a
relation s = s(t) such that r(t) = R(s(t)).
Theorem 13.1. (Arc Length of a Curve).

Suppose that a curve C has a simple, continuously differentiable parameter-
ization r(t), a ≤ t ≤ b. Then
Z b
L= kr0 (t)k dt .
a
A complete proof of the theorem goes beyond the scope of the course
and is given in advanced calculus courses (it requires the concept of uniform
continuity). Nevertheless the result may be understood from the following
consideration. By the hypothesis, the curve C is simple. Hence, given a
partition tk of [a, b] such that t0 = a < t1 < · · · < tN −1 < tN = b, there is a
unique polygonal path with vertices Pk on C whose length is
N
X N
X
LN = |Pk−1 Pk | = krk − rk−1 k .
k=1 k=1
where rk = r(tk ). Put ∆tk = tk − tk−1 > 0, k = 1, 2, ..., N. Under a

refinement of the partition DN = maxk ∆tk → 0 as N → ∞ and therefore
∆tk → 0 for all k as N → ∞. By the second hypothesis, r(t) traversing C is
differentiable. Put r0k−1 = r0 (tk−1 ). The differentiability of r(t) implies that
rk − rk−1 = r0k−1 ∆tk + uk ∆tk
(see (11.2)). By the triangle inequality (3.4),

kr0k−1 k∆tk − kuk k∆tk ≤ krk − rk−1 k ≤ kr0k−1 k∆tk + kuk k∆tk .
The lower and upper bounds for the length of the polygonal path are ob-
tained by taking the sum over k in this inequality:
N
X N
X N
X
kr0k−1 k∆tk − EN ≤ LN ≤ kr0k−1 k∆tk + EN , EN = kuk k∆tk .
k=1 k=1 k=1
By the third hypothesis, the derivative r0 (t)
is continuous and so is its norm
kr0 (t)k. A continuous function on [a, b] is integrable. Therefore its Riemann
sum converges:
N
X Z b
kr0k−1 k∆tk → kr0 (t)k dt = L as N → ∞ .
k=1 a
under a refinement of a partition for any choice of sample points (in this
case sample points are the left endpoints of partition intervals kr0k−1 k =
kr0 (tk−1 )k). The conclusion of the theorem follows from the squeeze princi-
ple: LN → L as N → ∞ if
N
X
lim EN = lim kuk k∆tk = 0 .
N →∞ N →∞
k=1
Put MN = maxk kuk k (the largest kuk k for a given partition size N ). Then
N
X N
X
EN = kuk k∆tk ≤ MN ∆tk = MN (b − a) .
k=1 k=1
So, it is sufficient to show that MN → 0 as N → ∞. In other words, all
kuk k ≤ MN converge to zero uniformly under a refinement of the partition.
By Eq. (11.2), kuk k converges to 0 as tk → tk−1 for a fixed k. So, intuitively
MN should converge to 0 because tk → tk−1 (or ∆tk → 0) for all k under
a refinement of the partition. This conclusion can indeed be rigorously
established: If r0 (t) is continuous on a closed interval [a, b], then 0 ≤ kuk k ≤
MN → 0 as N → ∞.
A few remarks on the use of Theorem 13.1 are in order. The length of
a curve that has a simple, continuously differentiable parameterization r(t)
on an infinite interval is defined as an improper integral. For example, if
a ≤ t < ∞, then
Z b
L = lim kr0 (t)k dt .
b→∞ a
The curve has a finite length if the integral converges (L < ∞).
The length is additive. So, if a curve C can be partitioned into finitely
many simple pieces and each piece admits a simple, continuously differ-
entiable parameterization, then the length of each piece can be found by
Theorem 13.1 and the length of C is the sum of lengths of each piece. This
observation allows us to use Theorem 13.1 to compute the length of curves
that are not simple.
Furthermore let r(t) be continuously differentiable on [a, b] but does not
necessarily define a one-to-one correspondence with its range C. Then the
Rb
integral a kr0 (t)k dt exists but may not coincide with the length of the curve
C as a point set in space because r(t) may traverse C or parts of C several
times. Nevertheless, the value of the integral may be useful in applications.
Suppose r(t) is a trajectory of a particle. Then its velocity is v(t) = r0 (t)
and its speed is v(t) = kv(t)k. The distance traveled by the particle in the
time interval [a, b] is given by
Z b Z b
D= v(t) dt = kr0 (t)k dt.
a a
If a particle travels along the same space curve (or some of its parts) several
times, then the distance traveled does not coincide with the arc length L of
the curve, D ≥ L.
Example 13.1. Find the arc length of the curve r(t) = ht2 , 2t, ln ti,
1 ≤ t ≤ 2.
Solution: The derivative r0 (t) = h2t, 2, 1/ti is continuous on [1, 2]. Its
norm is r r
0 2
1 1 2 1
kr (t)k = 4t + 4 + 2 = 2t + = 2t + .
t t t
Therefore, by Theorem 13.1,
Z 2 Z 2
0 1 2 2
L= kr (t)kdt = 2t + dt = t2 + ln t = 3 + ln 2 .
1 1 t 1 1

Example 13.2. Find the arc length of one turn of a helix of radius R
that rises by h per each turn.
Solution: Let the helix axis be the z axis (see Study Problem 10.1). The
helix is traced out by the vector function
r(t) = hR cos t, R sin t, th/(2π)i .
One turn is in a one-to-one correspondence with the interval t ∈ [0, 2π]
(because z(t) = th/(2π) is one-to-one). So r(t) is a simple, continuously
differentiable parameterization of one turn of the helix. Therefore,
kr0 (t)k = kh−R sin t, R cos t, h/(2π)ik = R2 + (h/(2π))2 .
p
So the norm of the derivative turns out be constant. The arc length is
Z 2π Z 2π
kr0 (t)k dt = R2 + (h/(2π))2
p p
L= dt = (2πR)2 + h2 .
0 0
This result is rather easy to obtain without calculus. The helix lies on a cylin-
der of radius R. If the cylinder is cut parallel to its axis and unfolded into a
strip, then one turn of the helix becomes the hypotenuse of the right-angled
triangle with catheti 2πR and h. The result follows from the Pythagorean
theorem. This consideration also shows that the length does not depend on
whether the helix winds about its axis clockwise or counterclockwise.
13.1. Reparameterization of a Curve. In Section 10.3, it was shown that

a space curve can be traversed by different vector functions. These vector
functions are different parameterizations of the same curve. For example, a
semicircle of radius R is traversed by the vector functions
r(t) = hR cos t, R sin t, 0i , t ∈ [0, π] ,
p
R(u) = hu, R2 − u2 , 0i , u ∈ [−R, R] .
They are related to one another by the composition rule:
R(u) = r(t(u)) , t(u) = cos−1 (u/R) or
r(t) = R(u(t)) , u(t) = R cos t .
This example illustrates the concept of a reparameterization of a curve. A
reparameterization of a curve is a change of the parameter that labels points
of the curve. It merely reflects a simple fact that there are many different
vector functions which traverse the same space curve.
Definition 13.2. (Reparameterization of a Curve).
Let r(t) traverse a curve C if t ∈ [a, b]. Let g(u) be a continuous one-to-
one function on an interval [a0 , b0 ] whose range is the interval [a, b], i.e., for
every a ≤ t ≤ b there is just one a0 ≤ u ≤ b0 such that t = g(u) and vice
versa. The vector function R(u) = r(g(u)) is called a reparameterization of
C.
The geometrical properties of the curve (e.g., its shape or length) do
not depend on a parameterization of the curve because the vector functions
r(t) and R(u) have the same range. A reparameterization of a curve is
a technical tool to find parametric equations of the curve convenient for
particular applications.
13.2. A Natural Parameterization of a Smooth Curve. Suppose one is trav-

eling along a highway from town A to town B and comes upon an accident.
How can the location of the accident be reported to the police? If one has
a GPS navigator, one can report coordinates on the surface of the Earth.
This implies that the police should use a specific (GPS) coordinate system
to locate the accident. Is it possible to avoid any reference to a coordinate
system? A simpler way to define the position of the accident is to report
the distance traveled from A along the highway to the point where the ac-
cident happened (by using, e.g., mile markers). No coordinate system is
needed to uniquely label all points of the highway by specifying the distance
from a particular point A to the point of interest along the highway. This
observation can be extended to all smooth curves (see Figure 13.1, right
panel).
Definition 13.3. (Natural or Arc Length Parameterization).
Let C be a smooth curve of length L. Let r(t), t ∈ [a, b], be a simple,
continuously differentiable parameterization of C such that r0 (t) 6= 0. Then
the arc length s = s(t) of the portion of the curve between r(a) and r(t) is a
function of the parameter t:
Z t
s = s(t) = kr0 (u)kdu , 0 ≤ s ≤ L .
a
The vector function R(s) = r(t(s)) is called a natural or arc length param-
eterization of C, where t(s) is the inverse function of s(t).
For a smooth curve, the function r(t) is continuously differentiable and,
hence, kr0 (t)k is continuous on [a, b]. Therefore, the derivative s0 (t) exists
and is obtained by differentiating the integral with respect to its upper limit:
d t 0
Z
s0 (t) = kr (u)kdu = kr0(t)k > 0 ,
dt a
which is possible because the integrand kr0 (u)k is continuous on [a, b]. The
derivative s0 (t) is strictly positive because r0 (t) 6= 0 for a smooth curve. The
existence of the inverse function of s(t) is guaranteed by the inverse function
theorem proved in Calculus I:
Theorem 13.2. (Inverse Function Theorem).
Let s(t), a ≤ t ≤ b, have a continuous derivative such that s0 (t) > 0 for
a < t < b. Then there exists an inverse differentiable function t = t(s),
c < s < d, and t0 (s) = 1/s0 (t), where t = t(s) on the right side.
Thus, the condition s0 (t) = kr0 (t)k > 0 guarantees the existence of a
one-to-one correspondence between the variables s and t and the existence
of the differentiable inverse function t = t(s). Let r(t) = hx(t), y(t), z(t)i be
parametric equations of a smooth curve C. Then the parametric equations
of C in the natural parameterization have the form
R(s) = hx(t(s)), y(t(s)), z(t(s))i.
Example 13.3. Reparameterize the helix from Example 13.2 r(t) =
hR cos t, R sin t, th/(2π)i with respect to arc length measured from the point
(R, 0, 0) in the direction of increasing t.
Solution: The point (R, 0, 0) corresponds to t = 0. Then
Z t
L t Lt 2πs
Z
s(t) = kr0 (u)k du = du = ⇒ t(s) =
0 2π 0 2π L
p
where L = (2πR)2 + h2 is the arc length of one turn of the helix (see
Example 13.2). Therefore
R(s) = r(t(s)) = hR cos(2πs/L), R sin(2πs/L), hs/Li
In particular, R(0) = hR, 0, 0i and R(L) = hR, 0, hi are the position vector
of the end points of one turn of the helix as required.
Example 13.4. Find the coordinates of a point P that is 5π/3
units of length away from the point (4, 0, 0) along the helix r(t) =
h4 cos(πt), 4 sin(πt), 3πti. Is there only one such point of the helix?
Solution: If R(s) is the natural parameterization of the helix where s is
counted from the point (4, 0, 0), then the position vector of the point in
question is given by R(5π/3). Thus, the first task is to find R(s). One has
r0 (u) = h−4π sin(πu), 4π cos(πu), 3πi ⇒ kr0 (u)k = 5π.
The initial point of the helix corresponds to t = 0. So the arc length counted
from (4, 0, 0) as a function of t is
Z t Z t
0 s
s(t) = kr (u)k du = 5π du = 5πt ⇒ t(s) = .
0 0 5π
The natural parameterization reads
R(s) = r(t(s)) = h4 cos(s/5), 4 sin(s/5), 3s/5i.
√
The position vector of P is R(5π/3) = h2, 2 3, πi. There are two points of
the helix at the specified distance from (4, 0, 0) because the arclength can be
counted in two directions from a given point. Note that s(t) defined above is
the arc length parameter counted from (4, 0, 0) in the direction of increasing
t (upward along the helix, t > 0). Accordingly, s(t) can also be counted
in the direction of decreasing t (downward along the helix, t < 0). In the
latter case, s(t) = −5πt √ > 0. Hence, the position vector of the other point
is R(−5π/3) = h2, −2 3, −πi.
It follows from Theorem 13.2 that the derivative of a vector function
that traverses a smooth curve C with respect to the natural parameter, the
arclength, is a unit tangent vector to the curve. Indeed, by the chain rule
applied to the components of the vector function:
dr(t) D dx(t) dy(t) dx(t) E
= , , = hx0 (t)t0 (s), y 0 (t)t0 (s), z 0 (t)t0 (s)i
ds ds ds ds
1 1
= t0 (s)hx0 (t), y 0 (t), z 0 (t)i = 0 r0 (t) = 0 r0 (t) = T̂(t)
s (t) kr (t)k
Thus, for a natural parameterization r(s) of a smooth curve C, the derivative
r0 (s) is a unit tangent vector to C, kr0 (s)k = 1. In Example 13.3, a natural
parameterization of a helix of radius R with one turn of a length L was
obtained. The derivative of the corresponding vector function reads:
d
R0 (s) = hR cos(2πs/L), R sin(2πs/L), hs/Li
dt
= h−(2πR/L) sin(2πs/L), (2πR/L) cos(2πs/L), h/Li
kR0 (s)k = (2πR/L)2 + (h/L)2 = 1 ,
p
where the fundamental

p trigonometric identity cos2 θ + sin2 θ = 1 and the
relation L = (2πR)2 + h2 were used.
By definition, the arc length is independent of a parameterization of a
space curve. For smooth curves, this can also be established through the
change of variables in the integral that determines the arc length. Indeed,
let r(t), t ∈ [a, b], be a simple, continuously differentiable parameterization
of a curve C of length L. Consider the change of the integration variable
t = t(s), s ∈ [0, L]. Then by the inverse function theorem s0 (t) = kr0 (t)k
and ds = s0 (t) dt = kr0 (t)k dt. Thus,
Z b Z L
0
kr (t)k dt = ds = L
a 0
for any parameterization of the curve C satisfying the hypotheses of Theo-

rem 13.1.
13.3. Study problems.
Problem 13.1.pFind the arclength of the part of the curve of intersection

p
of the cone z = x2 + y 2 and the cylindrical surface y = x tan x2 + y 2
from the origin to the point (0, π/2, π/2). Hint: Use polar coordinates to
obtain a parameterization of the curve.
Solution: Let x = r cos θ and y = r sin θ. In the polar coordinates the

equation of the cylindrical surface is simplified to tan θ = tan r or r = θ.
The latter is a polar graph that describes a spiral unwinding from the origin
counterclockwise. The spiral is the cross section of the cylindrical surface
in the xy plane. Using θ as a parameter so that z = r = θ, the curve of
intersection is traversed by the vector function
r(θ) = hθ cos θ, θ sin θ, θi = θhcos θ, sin θ, 1i = θû(θ) + θê3 ,
where û(θ) = hcos θ, sin θ, 0i and 0 ≤ θ ≤ π/2 in order to select the part of
the curve in question (note that r(0) = 0 and r(π/2) = h0, π/2, π/2i). This
parameterization is simple (because z(θ) = θ is one-to-one) and continuously
differentiable. By the product rule
r0 (θ) = û(θ) + θû0 (θ) + ê3 .

The vectors û(θ), û0 (θ) = h− sin θ, cos θ, 0i, and ê3 are mutually orthogonal
unit vectors. Therefore kr0 (θ)k2 = 1 + θ2 + 1 = 2 + θ2 . The arclength is
r ! r π
Z πp 2 2 2
2 θ θ θ θ
L= 2 + θ2 dθ = ln √ + 1 + +√ 1+
0 2 2 2 2
0
r ! r
π π2 π π2
= ln √ + 1+ + √ 1+ .
2 2 8 2 2 8
The integration can be carried out using the substitution
√ √ p √
θ = 2 sinh t , dθ = 2 cosh tdt , 2 + θ2 = 2 cosh t .
Then
Z p Z Z
2 2
2 + θ dθ = 2 cosh (t) dt = (1 + cosh(2t)) dt
1
= t+ sinh(2t) = t + sinh(t) cosh(t)
2
√ p
= t + (θ/ 2) 1 + θ2 /2
The variable t is expressed via θ by solving the quadratic equation in et :
√ √
2 θ = 2 sinh t = et − e−t ⇒ et = θ/ 2 + 1 + θ2 /2
p
and taking the natural logarithm of both sides of the latter relation.
13.4. Exercises.
1–6. Find the arc length of each of the following parametric curves:
1. r(t) = h3 cos t, 2t, 3 sin ti, −2 ≤ t ≤ 2 ;
2. r(t) = h2t, t3 /3, t2 i, 0 ≤ t ≤ 1 ;
3. r(t) = h3t2 ,√4t3/2 , 3ti, 0 ≤ t ≤ 2 ;
4. r(t) = het , 2 t, e−t i, −1 ≤ t ≤ 1 ;
5. r(t) = hcosh t, sinh t, ti, 0 ≤ t ≤ 1 ;
6. r(t) = hcos t − t sin t, sin t + t cos t, t2 i, 0 ≤ t ≤ 2π ; Hint: Find
the decomposition r(t) = v(t) + tw(t) + t2 ê3 where v, w, and ê3
are mutually orthogonal, and v 0 (t) = w(t), w0 (t) = −v(t). Use the
Pythagorean theorem to calculate kr0 (t)k .
7. Find the arc length of the curve r(t) = he−t cos t, e−t sin t, e−t i, 0 ≤ t < ∞.
Hint: Put r(t) = e−t u(t), differentiate, show that u(t) is orthogonal to u0 (t),
and use the Pythagorean theorem to calculate kr0 (t)k .
8. Find the arc length of the portion of the helix r(t) = hcos t, sin t, ti that
lies inside the sphere x2 + y 2 + z 2 = 2.
9. Find the arc length of the portion of the curve r(t) = h2t, 3t2 , 3t3 i that
lies between the planes z = 3 and z = 24.
10. Find the arc length of the portion of the curve r(t) = hln t, t2 , 2ti that lies
between the points of intersection of the curve with the plane y − 2z + 3 = 0.
11. Let C be the curve of intersection of the surfaces z 2 = 2y and 3x = yz.
Find the length of C from the origin to the point (36, 18, 6).
12–15. For each of the following curves defined by given equations with a
parameter a > 0, find suitable parametric equations and evaluate the arc
length between a given point A and and a generic point B = (x0 , y0 , z0 ):
12. y = a sin−1 (x/a), z = (a/4) ln[(a − x)/(a + x)], A = (0, 0, 0);
13. (x − y)2 = a(x + y), x2 − y 2 = 9z 2 /8, A = (0, 0, 0) ; Hint: Use
new variables u = x+y and v = x−y to find parametric equations ;
14. x2 + y 2 = az, y = x tan(z/a), A = (0, 0, 0); Hint: Use polar
coordinates in the p xy−plane to find parametric equations ;
2 2 2 2
15. x + y + z = a , x2 + y 2 cosh(tan−1 (y/x)) = a, A = (a, 0, 0) ;
Hint: Represent the second equation as a polar graph .
16–20. Reparameterize each of the following curves with respect to the arc
length measured from the point where t = 0 in the direction of increasing t:
16. r(t) = ht, 1 − 2t, 5 + 3ti ;
17. r(t) = t22t ê + ( t22+1 − 1)ê3 ;
+1 1
18. r(t) = hcosh t, sinh t, ti ;
19. x = a(t − sin t), y = a(1 − cos t), z = 0, a > 0 , and 0 ≤ t ≤ 2π ;
20. r(t) = et hcos t, sin t, 1i .
21. A particle travels from the point (R, 0, 0) into the positive quadrant
along a helix of radius R that rises h units of length per turn. If the z axis is
the symmetry axis of the helix, find the position vector of the particle after
it travels the distance 4πR.
22. A particle travels along a curve traversed by the vector function√r(u) =
hu, cosh u, sinh ui from the point (0, 1, 0) with a constant speed 2 m/s
so that its x coordinate increases. Find the position of the particle in one
second.
23. Let C be a smooth closed curve whose arclength is L. Let r(t), a ≤ t ≤ b,
be a simple, continuously differentiable parameterization of C. Prove that
there is a number a ≤ t∗ ≤ b such that kr0 (t∗ )k = L/(b − a). Hint: Recall
the integral mean value theorem.
24. A particle travels in space a distance D in time T . Show that there is a
moment of time 0 ≤ t ≤ T at which the speed of the particle coincides with
the average speed D/T .
14. CURVATURE OF A SPACE CURVE 205
14. Curvature of a Space Curve

Consider two curves passing through a point P . Both curves bend at P .
Which one bends more than the other and how much more? The answer
to this question requires a numerical characterization of bending, that is, a
number computed at P for each curve with the property that it becomes
larger as the curve bends more. Naturally, this number should not depend
on a parameterization of a curve. Suppose that a curve is smooth so that a
unit tangent vector can be attached to every point of the curve. A straight
line does not bend (does not “curve”) so it has the same unit tangent vector
at all its points. If a curve bends, then its unit tangent vector must change
along the curve.
The position on the curve can be specified in a coordinate- and parame-
terization-independent way by the arclength s counted from a particular
point of the curve. If T̂(s) is the unit tangent vector as a function of s, then
its derivative T̂0 (s) vanishes for a straight line (see Figure 14.1), while this
is not the case for a general smooth curve. However, T̂0 (s) is not a number
and cannot serve as the thought-after numerical measure of bending. Let
∆θ be the angle between T̂(s) and T̂(s0 ) for some s0 . Since T̂(s) and T̂(s0 )
have the same (unit) length, a non-vanishing derivative T̂0 (s0 ) implies that
the rate of change ∆θ/∆s, where ∆s = s − s0 , cannot vanish too. Yet, by
geometrical reasonings, the rate ∆θ/∆s is expected to increase as the curve
bends more at a point corresponding to s = s0 . So, the rate ∆θ/∆s in the
limit s → s0 might be a good numerical measure of bending of a curve at a
point.
By definition,
T̂(s) − T̂(s0 )
T̂0 (s0 ) = lim .
s→s0 s − s0
Figure 14.1. Left: A straight line does not bend. The

unit tangent vector has zero rate of change relative to the arc
length parameter s. Right: Curvature of a smooth curve.
The more a smooth curve bends, the larger the rate of change
of the unit tangent vector relative to the arc length parame-
ter becomes. So the magnitude of the derivative (curvature)
kT̂0 (s)k = κ(s) can be taken as a geometrical measure of
bending.
By continuity of the unit tangent vector

∆θ → 0 as ∆s → 0 .
Since T̂ is a unit vector, kT̂k = 1 and T̂ · T̂ = 1 and hence
(T̂(s) − T̂(s0 )) · (T̂(s) − T̂(s0 )) = 2 − 2 cos ∆θ = 4 sin2 (∆θ/2)
√
Recall that kak = a · a for any vector a. Therefore the magnitude
kT̂(s) − T̂(s0 )k = 2 sin(∆θ/2)
because ∆θ ≥ 0. In the limit ∆s → 0, the sinus function can be approxi-
mated by its argument: sin x = x + O(x3 ) = x(1 + O(x2 )). Therefore
2 sin(∆θ/2) ∆θ ∆θ
kT̂0 (s0 )k = lim = lim 1 + O(∆θ2 ) = lim ,
∆s→0 ∆s ∆s→0 ∆s ∆s→0 ∆s
and the number kT̂0 (s0 )k can be used as a numerical measure of bending
(or curvature) of a curve at a point.
Definition 14.1. (Curvature of a Smooth Curve).
Let C be a smooth curve and let its unit tangent vector T̂(s) be a differen-
tiable function of the arc length counted from a particular point of C. The
number
d
κ(s) = T̂(s)
ds
is called the curvature of C at the point corresponding to the value s of the
arc length.
Let r(s) be the natural parameterization of a smooth curve (the parame-
ter s is the arc length measured from a particular point on the curve). Then,
as shown in the previous section, r0 (s) = T̂(s), and therefore
r0 (s) = T(s) ⇒ κ(s) = kT0 (s)k = kr00 (s)k .
Example 14.1. Find the curvature of a helix of radius R that rises the
distance h per turn.
Solution: In Example 13.3, the natural parameterization of the helix is
obtained
r(s) = hR cos(2πs/L), R sin(2πs/L), hs/Li.
p
where L = (2πR)2 + h2 is the arc length of one turn. Differentiating this
vector function twice with respect to the arc length parameter s,
r00 (s) = h−(2π/L)2R cos(2πs/L), −(2π/L)2 R sin(2πs/L), 0i
= −(2π/L)2R hcos(2πs/L), sin(2πs/L), 0i ,
R
κ(s) = kr00(s)k = (2π/L)2R = 2 ,
R + (h/2π)2
where the relation khcos u, sin u, 0ik = 1 has been used. So the helix has a
constant curvature.
In practice, finding the natural parameterization of a smooth curve might

be a tedious technical task. Therefore, a question of interest is to develop
a method to calculate the curvature in any parameterization. Let r(t) be a
vector function in [a, b] that traces out a smooth curve C. The unit tangent
vector as a function of the parameter t has the form
r0 (t)
T̂(t) = .
kr0 (t)k
So, to calculate the curvature as a function of t, the relation between the
derivatives d/ds and d/dt has to be found. If s = s(t) is the arc length as
a function of t (see Definition 13.3), then by the inverse function theorem
(Theorem 13.2), there exists an inverse differentiable function t = t(s) that
expresses the parameter t as a function of the arc length s and
dt(s) 1 1
= = 0 .
ds (ds(t)/dt) kr (t)k
By the chain rule:
d dt d 1 d
T̂ = T̂ = 0 T̂
ds ds dt kr (t)k dt
and therefore
kT̂0 (t)k
(14.1) κ(t) = .
kr0 (t)k
Note that the existence of the curvature requires that r(t) be twice differen-
tiable because T̂(t) is proportional to r0 (t). The ratio in Eq. (14.1) makes
sense if r0 (t) 6= 0. Differentiation of the unit vector T̂ can sometimes be a
rather technical task, too. The following theorem provides a more conve-
nient way to calculate the curvature.
Theorem 14.1. (Curvature of a Curve).
Let a smooth curve be traced out by a twice-differentiable vector function
r(t) such that r0 (t) 6= 0. Then the curvature is
kr0 (t) × r00 (t)k
(14.2) κ(t) = .
kr0 (t)k3
Proof. Put v(t) = kr0 (t)k. With this notation,
r0 (t) = v(t)T̂(t) .
Differentiating both sides of this relation using the product rule, one infers
v 0 (t) 0
(14.3) r00 (t) = v 0 (t)T̂(t) + v(t)T̂0 (t) = r (t) + v(t)T̂0 (t) .
v(t)
Since the cross product of two parallel vectors vanishes, it follows from (14.3)
that

(14.4) r0 (t) × r00(t) = v(t) r0 (t) × T̂0 (t) .
Now recall that ka × bk = kakkbk sin θ where θ is the angle between vectors
a and b. Therefore,
(14.5) kr0 (t) × r00 (t)k = v(t)kr0(t) × T̂0 (t)k = kr0(t)k2 kT̂0 (t)k sin θ,
where θ is the angle between T̂0 (t) and the tangent vector r0 (t). Since T̂(t)
is a unit vector, one has kT̂(t)k2 = T̂(t) · T̂(t) = 1. By taking the derivative
of both sides of the latter relation, it is concluded that the vectors T̂0 (t) and
r0 (t) are orthogonal:
π
T̂0 (t) · T̂(t) = 0 ⇔ T̂0 (t) ⊥ T̂(t) ⇔ T̂0 (t) ⊥ r̂0 (t) ⇔ θ =
2
because r0 (t) is parallel to T̂(t). Hence, sin θ = 1. Then the claim (14.2)
follows from (14.1) and (14.5):
kT0 (t)k kr0 (t) × r00 (t)k kr0 (t) × r00 (t)k
κ(t) = = = .
kr0 (t)k kr0 (t)k3 sin θ kr0 (t)k3

Example 14.2. Find the curvature of the curve r(t) = hln t, t2 , 2ti at
the point P0 (0, 1, 2).
Solution: The point P0 corresponds to t = 1 because r(1) = h0, 1, 2i
coincides with the position vector of P0 . Hence, one has to calculate κ(1):
r0 (1) = ht−1 , 2t, 2i = h1, 2, 2i ⇒ kr0 (1)k = 3 ,

t=1
r00 (1) = h−t−2 , 2, 0i = h−1, 2, 0i

t=1
r0 (1) × r00 (1) = h−4, −2, 4i = 2h−2, −1, 2i ,
kr0 (1) × r00(1)k 2 kh2, 1, −2ik 6 2
κ = κ(1) = 0 3
= 3
= = .
P0 kr (1)k 3 27 9

Equation (14.2) can be simplified in two particularly interesting cases.
If a curve is planar (i.e., it lies in a plane), then, by choosing the coordinate
system so that the xy plane coincides with the plane in which the curve lies,
one has r(t) = hx(t), y(t), 0i. Since r0 and r00 are in the xy plane, their cross
product is parallel to the z axis:
r = hx(t), y(t), 0i ⇒ r0 × r00 = h0, 0, x0y 00 − x00 y 0 i .
The substitution of this relation into Eq. (14.2) leads to the following result.
Corollary 14.1. (Curvature of a Planar Curve).
The curvature of a planar smooth curve r(t) = hx(t), y(t), 0i, where the
derivatives x0 (t) and y 0 (t) do not vanish simultaneously, is
|x0 y 00 − x00 y 0 |
κ= .
[(x0)2 + (y 0 )2 ]3/2
A further simplification occurs when the planar curve is a graph y =

f (x). The graph is traced out by the vector function r(t) = ht, f (t), 0i.
Then, in Corollary 14.1, x0 (t) = 1, x00 (t) = 0, and y 00 (t) = f 00 (t) = f 00 (x),
which leads to the following result.
Corollary 14.2. (Curvature of a Graph).
The curvature of the graph y = f (x) is
|f 00 (x)|
κ(x) = .
[1 + (f 0 (x))2 ]3/2
14.1. Geometrical Significance of the Curvature. Let us calculate the cur-

vature of a circle of radius R. One has
x(t) = R cos t x0 (t) = −R sin t x00 (t) = −R cos t
⇒ 0 ⇒
y(t) = R sin t y (t) = R cos t y 00 (t) = −R sin t
By Corollary 14.1
(x0 )2 + (y 0 )2 = R2 R2 1
⇒ κ= =
x0 y 00 − x00 y 0 = R2 R 3 R
Therefore, the curvature is constant along the circle and equals a reciprocal
of its radius. The fact that the curvature is independent of its position on
the circle can be anticipated from the rotational symmetry of the circle (it
bends uniformly). Naturally, if two circles of different radii pass through the
same point, then the circle of smaller radius bends more as it lies inside the
bigger circle. Note also that the curvature has the dimension of the inverse
length. This motivates the following definition.
Definition 14.2. (Radius of Curvature).
The reciprocal of the curvature of a curve is called the radius of curvature
ρ = 1/κ.
The radius of curvature is a function of a point on the curve. Let a
planar curve have a curvature κ at a point P . Consider a circle of radius
ρ = 1/κ through the same point P that also has the same tangent line. Then
the center of such a circle lies on the line through P that is perpendicular to
the tangent line. There are only two circles with the described properties,
only only one of them has the same concavity at P as the curve relative
to the tangent line. In other words, the curve and the circle “bend” in the
same way (or direction) at P , have the same tangent line and the curvature.
This circle is called the osculating circle of the curve at a point P . One can
say the curve looks like a circle of radius 1/κ (in units of length) near P .
Example 14.3. Find the osculating circle for the graph y = cos(2x) at
the point (0, 1).
Solution: Since y 0 (0) = −2 sin 0 = 0, the tangent line to the graph is
horizontal y = 1 and the y axis is the line normal to the tangent line at
(0, 1). Therefore the center of the osculating circle lies on the y axis down
Figure 14.2. Left: Radius of curvature. A smooth curve

near a point P can be approximated by a portion of a circle
of radius ρ = 1/κ. The curve bends in the same way as a
circle of radius that is the reciprocal of the curvature. A
large curvature at a point corresponds to a small radius of
curvature. Middle: Osculating plane and osculating circle.
The osculating plane at a point P contains the tangent vector
T̂ and its derivative T̂0 at P and hence is orthogonal to n =
T̂ × T̂0 . The osculating circle lies in the osculating plane, it
has radius ρ = 1/κ, and its center is at a distance ρ from
P in the direction of T̂0 . One says that the curve “bends”
in the osculating plane. Right: For a curve traced out by a
vector function r(t), the derivatives r0 and r00 at any point P0
lie in the osculating plane through P0 . So the normal to the
osculating plane can also be computed as n = r(t0 )0 × r00 (t0 ),
where r(t0 ) is the position vector of P0 .
from y = 1 because the graph of cos(2x) is concave downward near x = 0.

The curvature of the graph at x = 0 is found by Corollary 14.2: y 0 (x) =
−2 sin(2x), y 00 (x) = −4 cos(2x) and
|y 00 (0)| | − 4|
κ(0) = 0 2 3/2
= = 4.
(1 + (y (0)) ) 1
So the radius of curvature is ρ0 = ρ(0) = 1/κ(0) = 1/4. The center (xc , yc )
of the osculating circle lies on the y axis ρ0 units of length down from y = 1,
that is, at (xc , yc ) = (0, 3/4). The equation of the osculating circle is
(x − xc )2 + (y − yc )2 = ρ20 or x2 + (y − 43 )2 = 1
16 .

Let us now view a planar curve in space. Then there are infinitely many
circles of radius 1/κ through the point P on the curve that share the same
tangent line with the curve at P . By what principle can one select the
osculating circle out of all such circles? Clearly, the same question arises for
a general smooth curve in space. There are infinitely many planes containing
the tangent line of a curve at a particular point. All such planes are obtained
from another by rotations about the tangent line, and in any such plane one
can take a circle of radius 1/κ through P . An example of the osculating
circle to a planar curve already answered the question about the selection
principle: the osculating circle should provide the best approximation to
the curve near P . So, what is left is to quantify the notion of the “best
approximation”.
Let r(s) be a natural parameterization of a curve such that the position
vector of P is r(0) (i.e., the arc length parameter is measured from P ). Then
r0 (s) = T̂(s) , r00 (s) = T̂0 (s) ⊥ T̂(s) , kr00 (s)k = κ(s) .
The natural parameter s is equal to the arclength when counted in the
direction of T̂(0) from P and is equal to the negative of the arclength when
counted in the opposite direction from P along the curve. Let R(s) be a
natural parameterization of the circle of radius ρ that and passes through P
(with the same agreement about negative and positive s as for the curve).
Let us demand that the circle and the curve have the same tangent line and
the same curvature at P . This implies that
r(0) = R(0) , r0 (0) = R0 (0) = T̂(0) , kr00(0)k = kR00 (0)k = κ(0) .
Note that the unit tangent vectors to the curve and circle may have opposite
directions in general. However this only refers to an agreement about the
direction in which the curves are traced out by r(s) and R(s) and has nothing
to do with the shape of the curves. The equality of the unite tangent vectors
means that s increases in the same direction along the curve and circle (or
they are oriented in the same way).
Consider the quantity kr(s) − R(s)k. It has a simple meaning. It defines
the distance between two points that are at a distance s from P along the
curve and along the circle. So, for a circle that gives the “best approxima-
tion”, this distance should be smallest for any given (small) value of s. Let
us adopt this principle to figure out for which circle kr(s)−R(s)k is minimal.
To accomplish this task, recall from Calculus 1 that the Taylor polynomial
of the second degree approximates best a twice differentiable function x(s)
near a point, say s = 0 among all quadratic polynomials sharing the same
value x(0) and the same slope x0 (0) of the tangent line:
1
x(s) = x(0) + x0 (0)s + x00 (0)s2 + s2 u(s)
2
The error of the approximation s2 u(s) tends to zero faster than s2 , that is,
u(s) → 0 as s → 0. In other words, put
x(s) − p (s)
p(s) = x(0) + x0 (0)s + ax2 , u(s) = .
s2
Then the error of the approximation s2 u(s) tends to zero faster than s2 ,
that is,
x(s) − p(s)
lim u(s) = lim =0
s→0 s→0 s2
if and only if a = 21 x00 (0). In this sense, the Taylor polynomial provides the
best approximations near s = 0 to a twice-differentiable function.
Using the Taylor approximation for each component of the vector func-
tions r(s) and R(s), one finds
r(s) = r(0) + r0 (0)s + 12 r00 (0)s2 + s2 u(s) ,
R(s) = R(0) + R0 (0)s + 12 R00 (0)s2 + s2 v(s) ,
where u(s) → 0 and v(s) → 0 as s → 0. Since the curve and the circle have a
common point and the same unit tangent vector at s = 0, the approximation
error reads
1
kr(s) − R(s)k = kr00 (0) − R00 (0) + w(s)ks2 ,
2
where w(s) = 2(u(s) − v(s)) → 0 as s → 0. Therefore
kr(s) − R(s)k 1
lim 2
= lim kr00 (0) − R00 (0) + w(s)k
s→0 s 2 s→0
1 00
= kr (0) − R00 (0)k = 0
2
if and only if
r00 (0) = R00 (0) .
In other words, the error of the approximation decreases to zero in the
fastest way (faster than s2 ) if and only if the circle and and the curve have
the same derivative of the unit tangent vector T̂0 (0) with respect to the
natural parameter at the point P . This is a stronger condition than just the
equality of the magnitudes κ(0) = kr00 (0)k = kR00 (0)k.
The vectors T̂(0) and T̂0 (0) are orthogonal. Hence there exists a unique
plane through P that contains these vectors. The circle that provides the
best approximation to the curve near P lies in this plane as the plane and
the circle have the same T̂(0) and T̂0 (0). The radius of a circle is always
perpendicular to the tangent line through the end point of the radius. The
line through P and parallel to T̂0 (0) is perpendicular to the tangent line
and, hence, passes through the center of the circle. By the geometrical
interpretation of the derivative of a vector function the vector T0 (s) points
in the direction in which the curve bends as seen in Figs. 14.1 and 14.2
(the left panel); the curve is concave in the direction of T̂0 (s). This implies
that the best approximating circle must lie in the plane through P that
contains the vectors T̂(0) and T̂0 (0), and the center of the circle must be
ρ(0) = 1/κ(0) units of length from the point P in the direction of T̂0 (0) (so
that the curve and the circle have the same concavity at P ).
Definition 14.3. (Osculating Plane and Circle).
The plane through a point P of a curve that is parallel to the unit tangent
vector T̂ and its derivative T̂0 6= 0 at P is called the osculating plane at P .
The circle of radius ρ = 1/κ, where κ is the curvature at P , through P that
lies in the osculating plane and whose center is in the direction of T̂0 from
P is called the osculating circle at P .
Unit normal vector. Suppose that the curvature does not vanish along a
curve, κ 6= 0. Therefore the derivative T̂0 of the unit tangent vector does
not vanish and one can define the unit vector
1 1
N̂(s) = T̂0 (s) = T̂0 (s) , κ(s) 6= 0 ,
0
kT̂ (s)k κ(s)
that is parallel to T̂0 at each point of the curve. This vector is called the
normal of the curve. By construction, the unit normal and tangent vectors
are perpendicular:
T̂(s) ⊥ N̂(s) ⇔ T̂(s) · N̂(s) = 0
at any point on the curve.
Let r0 be the position vector of a particular point P0 of the curve and
T̂0 and N̂0 be the unit tangent and normal vectors at P0 . Then any vector
in the osculating plane through P0 is a linear combination of T̂0 and N̂0 . In
other words, T̂0 and N̂0 form an orthonormal basis in the osculating plane.
By Definition 14.3, the position vector of the center of of the osculating
circle at P0 is
1
rc = r0 + ρ0 N̂0 , ρ = .
κ0
Since any vector in the osculating plane is a linear combination of T̂0 and
N̂0 , then parametric equations of the osculating circle may be taken in the
form
R(t) = rc + a(t)T̂0 + b(t)N̂0
where the functions a(t) and b(t) should satisfy the condition that the dis-
tance between a point on the circle to its center is ρ0 for all t:
kR(t) − rc k = ρ0 .
Owing to the orthogonality of T̂0 and N̂0 , that is, T̂0 · N̂0 = 0, this condition
implies that
ρ20 = (r(t) − rc ) · (r(t) − rc )
= (a(t)T̂0 + b(t)N̂0 ) · (a(t)T̂0 + b(t)N̂0 )
= a2 (t)T̂0 · T̂0 + b2 (t)N̂0 · N̂0
= a2 (t) + b2 (t)
Thus, the components of the vector R(t) − rc in the orthonormal basis T̂0
and N̂0 satisfy an equation of the circle of radius ρ0 and may chosen in the
form a(t) = sin t and b(t) = ±ρ0 cos t (the choice of the sign determines the
direction in which the circle is traversed). If one takes a(t) = −ρ0 cos t, then
R(0) = r0 (the point P0 corresponds to t = 0 in the parametric equations
of the osculating circle through P0 ). This proves the following theorem.
Theorem 14.2. (Parametric equations of the osculating circle)

Suppose that a smooth curve C has a nonzero curvature κ0 at a point P0 . Let
T̂0 and N̂0 be the unit tangent and normal vectors at P0 . Then parametric
equations of the osculating circle through P0 are
1
R(t) = r0 + ρ0 (1 − cos t)N̂0 + ρ0 sin t T̂0 , ρ0 = , 0 ≤ t ≤ 2π .
κ0
An example of constructing parametric equations of the osculating circle
is given in Study Problem 14.4. When the curve in question is planar, the
osculating circle may be found by simpler means.
Theorem 14.3. (Equation of the Osculating Plane).
Let a curve C be traced out by a twice-differentiable vector function r(t).
Let P0 be a point of C such that its position vector is r(t0 ) = hx0 , y0 , z0 i
at which the vector n = r0 (t0 ) × r00 (t0 ) does not vanish. An equation of the
osculating plane through P0 is
n1 (x − x0 ) + n2 (y − y0 ) + n3 (z − z0 ) = 0 , n = hn1 , n2 , n3 i.
Proof. It follows from (14.3) that the second derivative r00(t0 ) lies in the
osculating plane because it is a linear combination of T̂(t0 ) and T̂0 (t0 ).
Hence, the osculating plane contains the first and second derivatives r0 (t0 )
and r00 (t0 ). Therefore, their cross product n = r0 (t0 )×r00 (t0 ) is perpendicular
to the osculating plane, and the conclusion of the theorem follows.
Example 14.4. For the curve r(t) = ht, t2 , t3 i, find the osculating plane
through the point (1, 1, 1).
Solution: The point in question corresponds to t = 1. Then
r0 (1) = h1, 2t, 3t2i = h1, 2, 3i
t=1
r00 (1) = h0, 2, 6ti = h0, 2, 6i
t=1
Therefore, the normal of the osculating plane is
n = r0 (1) × r00(1) = h1, 2, 3i × h0, 2, 6i = h6, −6, 2i .
The osculating plane is 6(x − 1) − 6(y − 1) + 2(z − 1) = 0 or 3x − 3y + z = 1.

Problem 14.1. Show that any smooth curve whose curvature vanishes is
a portion of a straight line.
Solution: Let r(s) be a natural parameterization of a smooth curve. Then
the derivative is a unit tangent vector to the curve, T̂(s) = r̂0 (s). By the
definition of the curvature, κ(s) = kT̂0 (s)k = kr00(s)k. If κ(s) = 0, then
r00 (s) = 0 for all s. Therefore the unit tangent vector r0 (s) = T̂ is a constant
vector. The integration of this relation yields a vector equation of a straight

line through some point r0 and parallel to T̂:
κ(s) = 0 ⇔ r00 (s) = 0 ⇔ r0 (s) = T̂ ⇔ r(s) = r0 + T̂s .

Problem 14.2. (Curvature of a Polar Graph)
Find the curvature of a polar graph r = f (θ) where f (θ) and f 0 (θ) do not
vanish simultaneously, and f is twice differentiable. In particular, calculate
the curvature of the spiral r = θ.
Solution: Using the relations x = r cos(θ), y = r sin θ between the rect-
angular and polar coordinates, parametric equations of the polar graph can
be written in the form
r(θ) = hf (θ) cos θ, f (θ) sin θ , 0i = f (θ)û(θ)
One has û0 (θ) = h− sin θ, cos θ, 0i, which is a unit vector orthogonal to
û(θ), and û00 (θ) = −û(θ). By the product rule for differentiation and by the
properties of the cross product:
r0 (θ) = f 0 (θ)û(θ) + f (θ)û0 (θ) ,

r00 (θ) = f 00 (θ) − f (θ) û(θ) + 2f 0 (θ)û0 (θ) ,

r0 (θ) × r00 (θ) = 2(f 0 (θ))2 û(θ) × û0 (θ) + f (θ) f 00 (θ) − f (θ) û0 (θ) × û(θ)

= 2(f 0 (θ))2 − f (θ)f 00 (θ) + (f (θ))2 û(θ) × û0 (θ) .
Since û(θ), û0 (θ), and û(θ) × û0 (θ) are mutually orthogonal unit vectors,
1/2
kr0 (θ)k = (f 0 (θ))2 + (f (θ))2 ,
kr0 (θ) × r00 (θ)k = 2(f 0 (θ))2 − f (θ)f 00 (θ) + (f (θ))2 ,

kr0 (θ) × r00 (θ)k
κ(θ) =
kr0 (θ)k3
2(f 0 (θ))2 − f (θ)f 00 (θ) + (f (θ))2
= 3/2 .
0 2
(f (θ)) + (f (θ)) 2
For the spiral f (θ) = θ, f 0 (θ) = 1, and f 00 (θ) = 0. Therefore

2 + θ2
κ(θ) =
(1 + θ2 )3/2
is the curvature of the spiral.
Problem 14.3. Find the maximal curvature of the graph of the exponen-
tial, y = ex , and the point(s) at which it occurs.
Solution: The curvature of the graph y = ex is calculated by Corollary

14.2:
|y 00 | ex
κ(x) = = .
[1 + (y 0 )2 ]3/2 (1 + e2x )3/2
Critical points are determined by κ0 (x) = 0 or
ex (1 + e2x )1/2 (1 − 2e2x)

κ0 (x) = =0 ⇒ 2e2x = 1
(1 + e2x )3
ln 2
⇒ x=− .
2
From the shape of the graph of the exponential, it is clear that the found crit-
ical point corresponds to the (absolute) maximum of κ(x) (maximal bend-
ing) and
1 √
(1/ 2) 2
κmax = κ − ln(2) = 3/2
= √ ,
2 [1 + (1/2)] 3 3
where the relation e2x = 1/2 at the critical point has been used.
Problem 14.4. Let a curve C be the intersection of two quadric surfaces

4z = x2 and 6y = xz. Find an equation of the osculating plane and para-
metric equations of the osculating circle through the point (2, 1/3, 1).
Solution: One has to find a parameterization of the curve of intersection.

If x = x(t), y = y(t), and z = z(t) are parametric equations of the curve of
intersection, then the functions x(t), y(t), and z(t) must satisfy 4z = x2 and
6y = xz for all values of the parameter t. Put x = 2t. Then it follows from
4z = x2 and 6y = xz that z = t2 and y = t3 /3. Thus, the curve is traversed
by the vector function
r(t) = h2t, t3 /3, t2 i
which is continuously differentiable. The point of interest corresponds to

t = 1. The curve is smooth because x0 (t) = 2 > 0 so that x(t) is one-to-
one by the inverse function theorem (the parameterization is simple) and
r0 (t) 6= 0. Then using Theorem 14.3 the osculating plane P is found:
r0 (t) = h2, t2, 2ti , r00 (t) = h0, 2t, 2i ,

n = r0 (1) × r00 (1) = h2, 1, 2i × h0, 2, 2i = 2h−1, −2, 2i
P : −1(x − 1) − 2(y − 1) + 2(z − 1) = 0 or x + 2y − 2z = 1 .
Parametric equations of the osculating circle are found by using Theorem

14.2. One has
p
kr0 (t)k = 4 + t4 + 4t2 = (2 + t2 )2 = 2 + t2 ,
p
1 1
T̂(t) = 2
h2, t2, 2ti ⇒ T̂0 = T̂(1) = h2, 1, 2i ,
2+t 3
2t 1
T̂0 (t) = − h2, t2 , 2ti + h0, 2t, 2i ,
(2 + t2 )2 2 + t2
2 2 2
T̂0 (1) = h−2, 2, 1i ⇒ kT̂0 (1)k = kh−2, 2, 1ik = ,
9 9 3
1 1
N̂0 = T̂0 (1) = h−2, 2, 1i ,
kT̂0 (1)k 3
kT̂0 (1)k 2 1 9
κ(1) = 0
= ⇒ ρ0 = = ,
kr (1)k 9 κ(1) 2
where Eq. (14.1) has been used to calculate the curvature. Let r0 = h1, 1, 1i.
Then the vector function
9 9
R(t) = r0 + (1 − cos t)N̂0 + sin t T̂0 , 0 ≤ t ≤ 2π ,
2 2
traverses the osculating circle through r0 . Put R(t) = hX(t) , Y (t) , Z(t)i.
Then parametric equations of the osculating circle are
x = X(t) = −1 + 3 cos t + 3 sin t ,
y = Y (t) = 4 − 3 cos t + 32 sin t ,
5
z = Z(t) = 2 − 23 cos t + 3 sin t .
It is not difficult to verify that the circle lies in the osculating plane by sub-
stituting these equations into the equation of the osculating plane obtained
above.
Problem 14.5. Consider a helix r(t) = hR cos(ωt), R sin(ωt), hti, where
R, ω, and h are numerical parameters. The arclength of one turn of the helix
is a function of the parameter ω, L = L(ω), and the curvature at any fixed
point of the helix is also a function of ω, κ = κ(ω). Use only geometrical
arguments (no calculus) to find the limits of L(ω) and κ(ω) as ω → ∞.
Solution: The vector function r(t) traces out one turn of the helix when t
ranges over the period of cos(ωt) or sin(ωt) (i.e., over the interval of length
2π/ω). Thus, the helix rises by 2πh/ω = H(ω) along the z axis per each
turn. When ω → ∞, the height H(ω) tends to 0 so that each turn of the
helix becomes closer and closer to a circle of radius R. Therefore,
L(ω) → 2πR (the circumference)
1
κ(ω) → (the curvature of the circle)
R
as ω → ∞.
A calculus approach requires a lot more work to establish this result:

Z 2π/ω
2π p
L(ω) = kr0 (t)k dt = (Rω)2 + h2
0 ω
p
= 2π R2 + (h/ω)2 → 2πR,
kr0 (t) × r00(t)k Rω 2 [(Rω)2 + h2 ]1/2

κ(ω) = =
kr0 (t)k3 [(Rω)2 + h2 ]3/2
R 1
= 2 2
→
R + (h/ω) R
as ω → ∞.
14.3. Exercises.
1–10. Find the curvature of each of the following parameterized curves as a
function of the parameter t, and find the radius of curvature at the indicated
point P :
1. r(t) = ht, 1 − t, t2 + 1i, P = (1, 0, 2);
2. r(t) = ht2 , t, 1i, P = (4, 2, 1);
3. y = sin(x/2), P = (π, 1) ;
4. r(t) = h4t3/2 , −t2 , ti, P = (4, −1, 1) ;
5. x = 1 + t2 , y = 2 + t3 , P = (2, 1) ;
6. x = et cos t, √y = 0, z = et sin t, P = (1, 0, 0);
7. r(t) = hln t, t, t2 i, P = (0, 1, 1);
8. r(t) = hcosh√t, sinh t, 2 + ti, P = (1, 0, 2);
9. r(t) = het , 2 t, e−t i, P = (1, 0, 1) ;
10. r(t) = hsin t − t cos t, t, cos t + t sin ti, P = (0, 0, 1) .
11. Find the curvature of r(t) = ht, t2 /2, t3 /3i at the point of its intersection
with the surface z = 2xy + 2/3.
12. Find the maximal and minimal curvatures of the graph y = cos(ax) and
the points at which they occur. Sketch the graph for a = 1 and mark the
points of the maximal and minimal curvature, local maxima and minima of
cos x, and the inflection points.
13. Find the maximal and minimal curvatures of the graph y = 1/x.
14. Use a geometrical interpretation of the curvature to guess the point on
the graphs y = ax2 and y = ax4 where the maximal curvature occurs. Then
verify your guess by calculations.
15–17. Let f (x) be twice continuously differentiable function and κ(x) be
the curvature of the graph y = f (x).
15. Does κ attain a local maximum value at every local minimum and
maximum of f ? If not, state an additional condition on f under
which the answer to this question is affirmative.
16. Prove that κ = 0 at inflection points of the graph.
17. Show by an example that the converse of the statement in Exercise

16 is not true, i.e., that the curvature vanishes at x = x0 does not
imply that the point (x0 , f (x0 )) is an inflection point.
18. Let f be twice differentiable at x0 . Let T2 (x) be its Taylor polynomial
of the second degree about x = x0 . Compare the curvatures of the graphs
y = f (x) and y = T2 (x) at x = x0 .
19–21. Find the curvature of each of the following polar graphs:
19. r = 1 + cos θ ;
20. r = eθ ;
21. r = | sin(3θ)| .
22–26. Find the equation of the osculating circle for each of the following
planar curves at a specified point P :
22. y = x2 , P = (0, 0) ;
23. y = x − x2 /4, P = (2, 1) ;
24. y = 1/x, P = (1, 1) ;
25. x = a(t − sin t), y = a(1 − cos t) (a cycloid) , P = (a(π/2 − 1), a) ;
26. x = cos t, y = sin(2t), P = (1, 0) and P = (−1, 0) .
27. Find the maximal and minimal curvature of the ellipse x2 /a2 +y 2 /b2 = 1,
a > b, and the points where they occur. Give the equations of the osculating
circles at these points.
28. Let r(t) = ht3 , t2 , 0i. This curve is not smooth and has a cusp at t = 0.
Find the curvature for t 6= 0 and investigate its limit as t → 0.
29. Show that the cardioid r = 1 + cos θ is not smooth at the origin. In-
vestigate the curvature of the cardioid as the origin is approached along the
cardioid.
30–31. Find an equation of the osculating plane for each of the following
curves at a specified point:
30. r(t) = h4t3/2 ,√−t2 , ti, P = (4, −1, 1) ;
31. r(t) = hln t, t, t2 i, P = (0, 1, 1) .
32. Find an equation for the osculating and normal planes for the curve
r(t) = h ln(t) , 2t , t2 i at the point P0 of its intersection with the plane y−z =
1. A plane is normal to a curve at a point if the tangent to the curve at
that point is normal to the plane.
33. Is there a point on the curve r(t) = ht, t2 , t3 i where the osculating
plane is parallel to the plane 3x − 3y + z = 1?
34. Prove that the trajectory of a particle has a constant curvature if the
particle moves so that the magnitudes of its velocity and acceleration vectors
are constant.
35. Consider a graph y = f (x) such that f 00 (x0 ) 6= 0. At a point (x0 , y0 ) on
the curve, where y0 = f (x0 ), find the equation of the osculating circle in the
form (x − a)2 + (y − b)2 = R2 . Hint: Show first that the vector h1, f 0(x0 )i
is tangent to the graph and a vector orthogonal to it is h−f 0 (x0 ), 1i. Then
consider two cases f 00 (x0 ) > 0 and f 00 (x0 ) < 0.
36. Let a smooth curve r = r(t) be planar and lie in the xy plane. At a
point (x0 , y0 ) on the curve, find the equation of the osculating circle in the
form (x − a)2 + (y − b)2 = R2 . Hint: Use the result of Study Problem 14.4 to
express the constants a, b, and R via x0 , y0 , and the curvature at (x0 , y0 ).
37. Find parametric equations of the osculating circle to the curve r(t) =
h4t3/2 , −t2 , ti at the point P = (4, −1, 1) by using the method of Study
Problem 14.4.
15. APPLICATIONS TO MECHANICS AND GEOMETRY 221
15. Applications to mechanics and geometry

15.1. Tangential and Normal Accelerations. Let r(t) be the trajectory of
a particle (t is time). Then v(t) = r0 (t) and a(t) = v 0 (t) are the velocity
and acceleration of the particle. The magnitude of the velocity vector is the
speed, v(t) = kv(t)k. If T̂(t) is the unit tangent vector to the trajectory,
then T̂0 (t) is orthogonal to it. The unit vector N̂(t) = T̂0 (t)/kT̂0 (t)k is called
a unit normal to the trajectory. In particular, the osculating plane at any
point of the trajectory contains T̂(t) and N̂(t). The differentiation of the
relation v(t) = v(t)T̂(t) (see (14.3)) shows that that acceleration always lies
in the osculating plane:
a = v 0 T̂ + v T̂0 = v 0 T̂ + vkT0 kN̂ .
Furthermore, substituting the relations κ = kT̂0 k/v and ρ = 1/κ into the
latter equation, one finds (see Figure 15.1, left panel) that
a = aT T̂ + aN N̂,
v·a
aT = v 0 = T̂ · a = ,
v
v2 kv × ak
aN = κv 2 = = .
ρ v
Definition 15.1. (Tangential and Normal Accelerations).

Scalar projections aT and aN of the acceleration vector onto the unit tan-
gent and normal vectors at any point of the trajectory of motion are called
tangential and normal accelerations, respectively.
The tangential acceleration aT determines the rate of change of a par-

ticle’s speed, while the normal acceleration appears only when the particle
makes a “turn”. In particular, a circular motion with a constant speed,
v = v0 , has no tangential acceleration, aT = 0, and the normal acceleration
is constant:
v02
v = v0 = const ⇒ aT = v 0 = 0 , aN = κv02 = .
R
because the curvature of a circle is the reciprocal of its radius, κ = 1/R.
Example 15.1. Let r(t) = ht, t2 /2, t3/6i be the position vector of a point
particle as a function of time t. Find, the velocity, speed, acceleration, tan-
gential and normal accelerations of the particle at the point P = (2, 2, 4/3).
Figure 15.1. Left: Decomposition of the acceleration a

of a particle into normal and tangential components. The
tangential component aT is the scalar projection of a onto
the unit tangent vector T̂. The normal component is the
scalar projection of a onto the unit normal vector N̂. The
vectors r and v are the position and velocity vectors of the
particle. Right: The tangent, normal, and binormal vectors
associated with a smooth curve. These vectors are mutually
orthogonal and have unit length. The binormal is defined by
B̂ = T̂ × N̂. The shape of the curve is uniquely determined
by the orientation of the triple of vectors T̂, N̂, and B̂ as
functions of the arc length parameter up to general rigid
rotations and translations of the curve as the whole.
Solution: The particle reaches the point P at t = 2 because r(2) =

h2, 2, 4/3i is the position vector of P . Therefore
v = v(2) = r0 (2) = h1, t, t2/2i = h1, 2, 2i ,
P t=2
v = v(2) = kv(2)k = kh1, 2, 2ik = 3 ,

P
a = a(2) = r00 (2) = h0, 1, ti = h0, 1, 2i ,

P t=2
a(2) · v(2) 6
aT = aT (2) = = = 2,
P v(2) 3
kv(2) × a(2)k kh2, −2, 1ik 3
aN = aN (2) = = = = 1.
P v(2) 3 3

To gain an intuitive understanding of the tangential and normal accel-
erations, consider a car moving along a road. The speed of the car can be
changed by pressing the gas or brake pedals. When one of these pedals is
suddenly pressed, one can feel a force along the direction of motion of the car
(the tangential direction). The car speedometer also shows that the speed
changes, indicating that this force is due to the acceleration along the road
(i.e., the tangential acceleration aT = v 0 6= 0). When the car moves along a
straight road with a constant speed, its acceleration is 0. When the road
takes a turn, the steering wheel must be turned in order to keep the car on
the road, while the car maintains a constant speed. In this case, one can
feel a force normal to the road. It is larger for sharper turns (a larger cur-
vature or a smaller radius of curvature) and also grows when the same turn
is passed with a greater speed. This force is due to the normal acceleration,
aN = v 2 /ρ, and is called a centrifugal force. By Newton’s law, its magnitude
is
mv 2
F = maN = mκv 2 = ,
ρ
where m is the mass of a moving object (e.g., a car).
When making a turn, the car does not slide off the road as long as the
friction force between the tires and the road compensates for the centrifugal
force. The maximal friction force depends on the road and tire conditions
(e.g., a wet road and worn tires reduce substantially the maximal friction
force). The centrifugal force is determined by the speed (the curvature of the
road is fixed by the road shape). So, for a high enough speed, the centrifugal
force can no longer be compensated for by the friction force, and the car
would skid off the road. For this reason, suggested speed limit signs are
often placed at highway exits. If one drives a car on a highway exit with
a speed twice as high as the suggested speed, the risk of skidding off the
road is quadrupled, not doubled, because the normal acceleration aN = v 2 /ρ
quadruples when the speed v is doubled.
Example 15.2. A road takes a turn that has a parabolic shape,
x2
y= , −R < x < R ,
2R
where (x, y) are coordinates of points of the road and R is a constant (all
measured in meters). A safety assessment requires that the normal acceler-
ation on the road should not exceed a threshold value am meters per second
squared to avoid skidding off the road. If a car moves with a constant speed
v meters per second along the road, find the portion of the road where the
car might skid off the road.
Solution: The normal acceleration of the car as a function of position (not
time!) is
aN (x) = κ(x)v 2 .
By Corollary 14.2, the curvature of the graph y = x2 /(2R) is
|y 00 | 1
κ(x) = = .
[1 + (y 0 )2 ]3/2 R[1 + (x/R)2]3/2
The maximal curvature and hence the maximal normal acceleration are
attained at x = 0. So, if the speed v is such that
v2
amax
p
N = aN (0) = < am ⇒ v < v max = Ram ,
R
no accident can happen. Thus, a suggested speed limit sign v < vmax can
be placed for this turn.
If a car makes the turn exceeding the speed limit vmax , then the danger-
ous part of the road, where an accident can happen, is determined by the
inequality
aN (x) ≥ am ,
v2 √ v 4/3
≥ am ⇒ |x| ≤ R ν − 1 , ν = .
R[1 + (x/R)2]3/2 vmax
The constant ν always exceeds 1 if v > vmax . Since the turn corresponds
to the range |x| < R, the dangerous
√ part of the turn is determined by the
smallest number of R and R ν − 1. Thus, the car can skid off the road
when moving on the part of the road corresponding to the interval
√
−Rµ ≤ x ≤ Rµ , µ = min{1, ν − 1} .
Note that an accident can happen anywhere on the turning part of the road
if the speed at which the car enters into this part of the road exceeds the
speed limit just by the factor 23/4 ≈ 1.68 (it corresponds to ν = 2).
15.2. Torsion and Frenet-Serret Formulas. The shape of a space curve as

a point set is independent of a parameterization of the curve. A natural
question arises: What parameters of the curve determine its shape? Suppose
the curve is smooth enough so that the unit tangent vector T̂(s) and its
derivative T̂0 (s) can be defined as functions of the arc length s counted from
an endpoint of the curve. Let N̂(s) be the unit normal vector of the curve.
If the curvature κ(s) does not vanish, then N̂(s) is uniquely defined by the
relation T̂0 (s) = κ(s)N̂(s) because T̂0 (s) and T̂(s) are orthogonal and, by
definition, kT̂0 (s)k = κ(s).
Definition 15.2. (Binormal Vector).
Let T̂ and N̂ be the unit tangent and normal vectors at a point of a curve.
The unit vector B̂ = T̂ × N̂ is called the binormal (unit) vector.
So, with every point of a smooth curve with a non-vanishing curvature,
one can associate a unique triple of mutually orthogonal unit vectors so that
one of them is tangent to the curve while the other two span the plane normal
to the tangent vector (normal to the curve). If the curvature vanishes, one
can still define two unit mutually orthogonal vectors in the plane normal to
the curve. However, any rotation in this plane would produce another pair
of unit vectors with the same properties so that the triple of unit vectors is
not unique.
By a suitable rotation, the triple of vectors T̂, N̂, and B̂ can be oriented
parallel to the axes of any given coordinate system, that is, parallel to ê1 ,
ê2 , and ê3 , respectively. Indeed, T̂ and N̂ can always be made parallel to
ê1 and ê2 . Then, owing to the relation ê1 × ê2 = ê3 , the binormal must
be parallel to ê3 . In other words, T̂, N̂, and B̂ define a local right-handed
coordinate system at each point of the curve. The orientation of the unit
tangent, normal, and binormal vectors relative to some coordinate system
depends on the point of the curve. The triple of these vectors can only rotate
as the point slides along the curve (the vectors are mutually orthogonal and
unit at any point). Therefore, the rates with respect to the arc length at
which these vectors change must be characteristic for the shape of the curve
(see Figure 15.1, right panel).
By the definition of the curvature, T̂0 (s) = κ(s)N̂(s). Suppose κ(s) 6= 0
so that N̂(s) is uniquely defined. Next, consider the rate:
B̂0 = (T̂ × N̂)0 = T̂0 × N̂ + T̂ × N̂0 = T̂ × N̂0
because T̂0 (s) is parallel to N̂(s). It follows from this equation that B̂0 is
perpendicular to T̂, and, since B̂ is a unit vector, its derivative must also
be perpendicular to B̂. Thus, B̂0 must be parallel to N̂. This conclusion
establishes the existence of another scalar quantity that characterizes the
curve shape.
Definition 15.3. (Torsion of a Curve).
Let N̂(s) and B̂(s) be unit normal and binormal vectors of the curve as
functions of the arc length s. Then
dB̂(s)
= −τ (s)N̂(s)
ds
and the number τ (s) is called the torsion of the curve.
By definition, the torsion is measured in units of a reciprocal length, just
like the curvature, because the unit vectors T̂, N̂, and B̂ are dimensionless.
Example 15.3. Use the natural parameterization of a helix given in
Example 13.3 to find the unit tangent, normal, and binormal vectors, T̂,
N̂, and B̂, as functions of the arclength parameter. Express the torsion and
curvature of the helix in terms its radius R and the height per turn h.
Solution: To simplify notations, put ω = 2π/L and a = h/(2π) so that
ω = (R2 + a2 )−1/2 and the natural parameterization from Example 13.3
reads
r(s) = hR cos(ωs), R sin(ωs), aωsi .
Then using the definitions of T̂, N̂, B̂, the curvature, and the torsion,
T̂(s) = r0 (s) = h−Rω sin(ωs), Rω cos(ωs), aωi ,
T̂0 (s) = Rω 2 h− cos(ωs), − sin(ωs), 0i ,
R h
κ(s) = kT̂0 (s)k = Rω 2 = 2 , a=
R + a2 2π
1
N̂(s) = T̂0 (s) = h− cos(ωs), − sin(ωs), 0i ,
κ(s)
B̂(s) = T̂(s) × N̂(s) = haω sin(ωs), −aω cos(ωs), Rωi

B̂0 (s) = haω 2 cos(ωs), −aω 2 sin(ωs), 0i = −aω 2 N̂(s) ,
a h
τ (s) = aω 2 = 2 2
, a= .
R +a 2π
So, a helix has a constant torsion and a constant curvature.
At any point of a curve, the binormal B̂ is perpendicular to the osculating
plane. So, if the curve is planar, then B̂ does not change along the curve,
B̂0 (s) = 0, because the osculating plane at any point coincides with the plane
in which the curve lies. A planar curve has no torsion. Thus, the torsion is
a local numerical characteristic that determines how fast the curve deviates
from the osculating plane while bending in it with some curvature radius.
It follows from the relation N̂ = B̂ × T̂ (compare ê2 = ê3 × ê1 ) that
N̂0 = (B̂ × T̂)0 = B̂0 × T̂ + B̂ × T̂0 = −τ N̂ × T̂ + κB̂ × N̂ = τ B̂ − κT̂ ,
where the definitions of the torsion and curvature have been used. The
obtained rates of change of the unit vectors are known as the Frenet-Serret
formulas or equations:
(15.1) T̂0 (s) = κ(s)N̂(s) ,
(15.2) N̂0 (s) = −κ(s)T̂(s) + τ (s)B̂(s) ,
(15.3) B̂0 (s) = −τ (s)N̂(s) .
The Frenet-Serret equations form a system of differential equations for the
components of T̂(s), N̂(s), and B̂(s). If the curvature and torsion are con-
tinuous functions on an interval 0 ≤ s ≤ L, then the system can be proved
to have a unique solution on this interval for every given set of the vectors
T̂(0), N̂(0), and B̂(0) at an initial point of the curve. A geometrical mean-
ing of the Frenet-Serret equations is that they determine how the triple of
unit mutually orthogonal vectors T̂, N̂, and B̂ at a point P on the curve
is rotated as the point P moves along the curve and that this rotation is
uniquely determined by the shape of the curve.
In particular, for the helix discussed in Example 15.3, the unit vectors
T̂, N̂, and B̂ rotates about the z axis, while remaining mutually orthog-
onal, at a constant rate relative to the distance s traveled along the helix
that is determined by the curvature and torsion of the helix. Indeed, by
Study Problem 1.2 variations of the angle θ = sω in T̂(s), N̂(s), and B̂(s)
correspond to rotations about the z axis. The angle of rotation changes
at a constant rate dθ/ds = ω = (κ2 + τ 2 )−1/2 determined by the constant
curvature and torsion of the helix. On the other hand, the geometrical pa-
rameters of the helix (its radius R and the height h per turn) are uniquely
determined by the curvature and torsion as Example 15.3 shows. So the
shape of a helix is determined by its curvature and torsion. The observation
that the curvature and the torsion determine the shape of a smooth curve is
of general nature.
Theorem 15.1. (Shape of a Smooth Curve in Space).

Given the curvature and torsion as continuous functions of the arclength
parameter of a smooth curve, the curve is uniquely determined by them up
to rigid rotations and translations of the curve as a whole, provided the
curvature is nowhere zero.
A proof of Theorem 15.1 requires a proof of the uniqueness of a solution
to the Frenet-Serret equations, which goes beyond the scope of this course. It
should also be noted that the Frenet-Serret equations have a unique solution
even without the condition that the curvature does not vanish. However,
wherever κ(s) = 0, the objects N̂(s) and τ (s) no longer have the meaning
that they were defined to have earlier in this section. Solutions of the Frenet-
Serret equations with κ = 0 and various τ 6= 0 may describe the same curve
in space. For example, if κ(s) = τ (s) = 0, then T̂(s) = T̂(0), N̂(s) = N̂(0),
and B̂(s) = B̂(0). If r(s) is a natural parameterization of a curve, then
r0 (s) = T̂(s) = T̂(0). The integration of this equation yields r(s) = r0 +
T̂(0)s, where r0 is a constant vector, which is a straight line (or a part
of it if the range of s is restricted). On the other hand, if κ(s) = 0 and
τ (s) = τ0 6= 0, then it follows from Eq. (15.1) that unit tangent vector is
still a constant vector T̂(s) = T̂(0) and, hence, the curve is a line. As has
been noted earlier, the choice of N̂ and B̂ is not unique wherever κ(s) = 0
(they can be chosen up to any rotation in the normal plane). This freedom
shows up in the solution of Eqs. (15.2) and (15.3) (which is easy to verify
by differentiation)
N̂(s) = cos(τ0 s)N̂(0) + sin(τ0 s)B̂(0) ,
B̂(s) = cos(τ0 s)B̂(0) − sin(τ0 s)N̂(0) .
This solution describes a rotation through the angle θ = τ0 s of the initial
unit vectors N̂(0) and B̂(0) in the plane orthogonal to the line as the point
travels the distance s along the line. This rotation is not associated with
the shape of the curve (a line parallel to T̂(0) in the case considered) and
has no geometrical significance whatsoever. This illustrates the necessity of
the hypothesis of a non-vanishing curvature in Theorem 15.1.
Example 15.4. Use the Frenet-Serret equations to prove that a curve
with a constant curvature κ(s) = κ0 6= 0 and zero torsion τ (s) = 0 is a circle
(or its portion) of radius R = 1/κ0 .
Solution: A vector function r(s) that satisfies the Frenet-Serret equations
is sought in the basis of the initial tangent, normal, and binormal vectors:
ê1 = T̂(0), ê2 = N̂(0), and ê3 = B̂(0). Since the torsion is 0, the binormal
does not change along the curve, B̂(s) = ê3 . The curve is planar and lies in
a plane orthogonal to ê3 . Any unit vector T̂ orthogonal to ê3 can always be
written as T̂ = cos ϕê1 + sin ϕê2 where ϕ = ϕ(s) such that ϕ(0) = 0. Owing
to the relations ê1 × ê2 = −ê2 × ê1 = ê3 , a unit vector N̂ orthogonal to T̂
such that T̂ × N̂ = B̂ = ê3 must have the form N̂ = − sin ϕê1 + cos ϕê2 .
Equation (15.1) gives
T̂0 = −ϕ0 sin ϕê1 + ϕ0 cos ϕ ê2 = ϕ0 N̂ = κ0 N̂ ⇒ ϕ0 (s) = κ0
and therefore ϕ(s) = κ0 s because ϕ(0) = 0. For a natural parameterization
of the curve, r0 (s) = T̂(s). Hence,
r0 (s) = cos(κ0 s)ê1 + sin(κ0 s)ê2 ,
r(s) = r0 + κ−1 −1
0 sin(κ0 s)ê1 − κ0 cos(κ0 s)ê2
where r0 is a constant vector. Put R = 1/κ0 . By the Pythagorean theorem,
the distance between any point of the curve and a fixed point r0 is constant:
1
kr(s) − r0 k2 = 2 sin2 (κ0 s) + cos2 (κ0 s) = R2 .
κ0
Since the curve is planar, it is a circle (or its portion) of radius R.
A calculation of the torsion based on Definition 15.3 requires a natural
parameterization of a smooth. The following theorem provides the method
to compute torsion of a curve by using any suitable parameterization.
Theorem 15.2. (Torsion of a Curve).
Let r(t) be three times differentiable vector function that traverses a smooth
curve whose curvature does not vanish. Then the torsion of the curve is
(r0 (t) × r00 (t)) · r000(t)
τ (t) = .
kr0 (t) × r00(t)k2
Proof. Put kr0(t)k = v(t) (if s = s(t) is the arc length as a function of t,
then s0 = v). By (14.3) and the definition of the curvature,
(15.4) r00 = v 0 T̂ + κv 2 N̂ ,
and by (14.4) and the definition of the binormal,
(15.5) r0 × r00 = v T̂ × r00 = κv 3 B̂.
Differentiation of both sides of (15.4) gives
r000 = v 00 T̂ + v 0 T̂0 + (κ0 v 2 + 2κvv 0 )N̂ + κv 2 N̂0 .
The derivatives T̂0 (t) and N̂0 (t) are found by making use of the differentia-
tion rule d/ds = (1/s0 (t))(d/dt) = (1/v)(d/dt) in the Frenet-Serret equations
(15.1) and (15.2):
T̂0 = κv N̂ , N̂0 = −κv T̂ + τ v B̂ .
Therefore,
(15.6) r000 = (v 00 − κ2 v 3 )T̂ + (3κvv 0 + κ0 v 2 )N̂ + κτ v 3 B̂.
By (15.5), (15.6), and the mutual orthogonality of the tangent, normal, and
binormal vectors
(r0 × r00) · r000 = κv 3 B̂ · r000 = κ2 v 6 τ B̂ · B̂ = κ2 v 6 τ .
Therefore,
(r0 × r00) · r000 (r0 × r00 ) · r000 kr0 × r00k
τ= = , κ = ,
κ2 v 6 kr0 × r00 k2 v3
and the conclusion of the theorem follows from Theorem 14.1.
Relation (15.5) shows that B̂ is the unit vector in the direction of r0 ×r00 .
This observation offers a more convenient way for calculating the unit bi-
normal vector than its definition because it uses any parametrization of a
smooth curve. The unit tangent, normal, and binormal vectors at a partic-
ular point r(t0 ) of the curve r(t) are
r0 (t0 )
(15.7) T̂(t0 ) = ,
kr0 (t0 )k
r0 (t0 ) × r00(t0 )
(15.8) B̂(t0 ) = 0 ,
kr (t0 ) × r00(t0 )k
(15.9) N̂(t0 ) = B̂(t0 ) × T̂(t0 ) .
Example 15.5. Find the unit tangent, normal, and binormal vectors
and the torsion of the curve r(t) = hln t, t, t2 /2i at the point (0, 1, 1/2).
Solution: The point in question corresponds to t = 1. Therefore
√
r0 (1) = ht−1 , 1, ti = h1, 1, 1i ⇒ kr0 (1)k = 3
t=1
00 −2
r (1) = h−t , 0, 1i = h−1, 0, 1i
t=1
r000(1) = h2t−3 , 0, 0i = h2, 0, 0i

t=1
√
r0 (1) × r00(1) = h1, −2, 1i ⇒ kr0 (1) × r00(1)k = 6
T̂(1) = √1 h1, 1, 1i
3
1
B̂(1) = √ h1, −2, 1i
6
N̂(1) = √ 1√ h1, −2, 1i × h1, 1, 1i = 1
√ h−3, 0, 3i
6 3 3 2
= √1 h−1, 0, 1i
2
(r0 (1) × r00 (1)) · r000(1) 2 1
τ (1) = = =
kr0 (1) × r00 (1)k2 6 3

Remark. A smooth curve C has a unit tangent vector at a point P . So a

small part of the curve (a part of a small arclength s) containing P can be
approximated by a piece of the tangent line of the same arclength s. If the
curve C has a nonzero curvature at P , then a better approximation is given
by a part of the osculating circle of the arclength s (see Study Problem 14.4).
If the curve C has a nonzero torsion at P , an even more accurate approx-
imation is provided by a curve through P that has the same unit tangent
vector at P , and a constant curvature and torsion equal to the curvature

and torsion of the curve C at P . By Theorem 15.1 such a curve is unique.
As shown in Example 15.3 (see also Study Problem 15.3), it is a helix whose
radius and length of each turn are uniquely determined by the curvature and
torsion. These three successively more accurate approximations do not refer
to any particular coordinate system or any particular parameterization of C
as the approximation curves are fully determined as the point sets in space
by the geometrical invariants of the curve C at P : the unit tangent vector,
curvature, and torsion.

Problem 15.1. Find the position vector r(t) of a particle as a function of
time t if the particle moves clockwise along a circular path of radius R in
the xy plane through r(0) = hR, 0, 0i with a constant speed v0 .
Solution: For a circle of radius R in the xy plane through the point
(R, 0, 0),
r(t) = hR cos ϕ, R sin ϕ, 0i ,
where ϕ = ϕ(t) such that ϕ(0) = 0. Then the velocity is
v(t) = r0 (t) = ϕ0 h−R sin ϕ, R cos ϕ, 0i .
Hence, the condition kv(t)k = v0 yields R|ϕ0(t)| = v0 or ϕ(t) = ±(v0 /R)t
and
r(t) = hR cos(ωt), ±R sin(ωt), 0i ,
where ω = v0 /R is the angular velocity. The second component must be
taken with the minus sign because the particle revolves clockwise with in-
creasing t (the second component should become negative immediately after
t = 0).
Problem 15.2. Let the particle position vector as a function of time t be
r(t) = hln(t), t2 , 2ti, t > 0. Find the speed, tangential and normal acceler-
ations, the unit tangent, normal, and binormal vectors, and the torsion of
the trajectory at the point P0 (0, 1, 2).
Solution: By Example 14.2, the velocity and acceleration vectors at P0 are
v = h1, 2, 2i and a = h−1, 2, 0i. So the speed is v = kvk = 3. The tangential
acceleration is aT = v · a/v = 1. As v × a = 2h−2, −1, 2i, the normal
acceleration is aN = kv × ak/v = 6/3 = 2. The unit tangent, binormal, and
normal vectors are obtained by Eqs. (15.7)–(15.9), where t0 = 1:
1 1
T̂ = v = h1, 2, 2i ,
v 3
1 1
B̂ = v × a = h−2, −1, 2i ,
kv × ak 3
1
N̂ = T̂ × B̂ = h−2, 2, −1i .
3
The torsion at P0 is calculated by Theorem 15.2 with t = 1. One has

r000(1) = h2/t2 , 0, 0i|t=1 = h2, 0, 0i = b. Therefore,
8 2
τ (1) = (v × a) · b/kv × ak2 = − =− .
36 9

Problem 15.3. (Curves with Constant Curvature and Torsion).
Prove that all curves with a constant curvature κ(s) = κ0 6= 0 and a constant
torsion τ (s) = τ0 6= 0 are helices by integrating the Frenet-Serret equations.
Solution: It follows from (15.1) and (15.3) that the vector
w = τ T̂ + κB̂
does not change along the curve, w0 (s) = 0. Indeed, because κ0 (s) = τ 0 (s) =
0, one has
w0 = τ T̂0 + κB̂0 = (τ κ − τ κ)N̂ = 0 .
By the Pythagorean theorem, kwk = (κ20 + τ02 )1/2 . Consider two new unit
vectors orthogonal to N̂:
1
ŵ = w = sin α T̂ + cos α B̂, û = cos α T̂ − sin α B̂ ,
kwk
where cos α = κ0 /ω, sin α = τ0 /ω, and ω = (κ20 + τ02 )1/2 . By construction,
the unit vectors û, ŵ, and N̂ are mutually orthogonal unit vectors, which is
easy to verify by calculating the corresponding dot products, û·û = ŵ· ŵ = 1
and û · ŵ = 0. Also,
ŵ × û = − cos2 α T̂ × B̂ + sin2 α B̂ × T̂ = (cos2 α + sin2 α)N̂ = N̂ .
By differentiating the vector û and using the Frenet-Serret equations,
û0 = cos α T̂0 − sin α B̂0 = (κ0 cos α + τ0 sin α)N̂ = ω N̂ .
Since ŵ(s) = ŵ(0) is a constant unit vector, it is convenient to seek a
solution in an orthonormal basis such that ê3 = ŵ(0) and ê1 × ê2 = ê3 . In
this basis û = cos ϕ ê1 + sin ϕ ê2 , where ϕ = ϕ(s), is a unit vector in the
plane orthogonal to ê3 . The orientation of the basis vectors in the plane
orthogonal to ê3 is defined up to a general rotation about ê3 . This freedom
is used to set ê1 = û(0), which implies that the function ϕ(s) satisfies the
condition ϕ(0) = 0. Then the unit normal vector in this basis is
N̂ = ŵ × û = cos ϕ ê3 × ê1 + sin ϕ ê3 × ê2 = cos ϕ ê2 − sin ϕ ê1
and
û0 = −ϕ0 sin ϕ ê1 + ϕ0 cos ϕ ê2 = ϕ0 N̂ .
The comparison of this representation of û0 with that obtained before yields
ϕ0 (s) = ω or ϕ(s) = ωs owing to the condition ϕ(0) = 0. Expressing the
vector T̂ via û and ŵ,
T̂ = cos α û + sin α ŵ ,
one infers (compare Example 15.4)

κ0 κ0 τ0
r0 (s) = T̂(s) = cos(ωs) ê1 + sin(ωs) ê2 + ê3
ω ω ω
where r(s) is a natural parameterization of the curve. The integration of
this equation gives
κ0 τ0
r(s) = r0 + R sin(ωs) ê1 − R cos(ωs) ê2 + hs ê3 , R = 2 , h = ,
ω ω
This is a helix of radius R whose axis goes through the point r0 parallel to
ê3 ; the helix climbs along its axis by 2πh/ω per each turn.
Remark. The fact that a curve with constant nonzero curvature and torsion
is a helix can also be established by Theorem 15.1. By Example 15.3, a
helix has constant nonzero curvature and torsion. By Theorem 15.1, the
curve with such curvature and torsion is unique modulo rigid rotations and
translations. Thus, the curve is a helix.
Problem 15.4. (Motion in a Constant Magnetic Field, Revisited).
The force acting on a charged particle moving in the magnetic field B is
given by F = (e/c)v × B, where e is the electric charge of the particle, c
is the speed of light, and v is its velocity. Show that the trajectory of the
particle in a constant magnetic field is a helix whose axis is parallel to the
magnetic field by integrating Newton’s equations of motion.
Solution: In contrast to Study Problem 12.3, here the shape of the tra-
jectory is to be obtained directly from Newton’s second law with arbitrary
initial conditions. Choose the coordinate system so that the magnetic field
is parallel to the z axis, B = Bê3 , where B is the magnitude of the magnetic
field. Newton’s law of motion, ma = F, where m is the mass of the particle,
determines the acceleration, a = µv × B = µBv × ê3 , where µ = e/(mc).
First, note that v30 = a3 = ê3 ·a = 0. Hence, v3 = vk = const (the component
of the velocity parallel to the magnetic field remains constant).
Second, by the geometrical property of the cross product the accelera-
tion and velocity remain orthogonal during the motion, and therefore the
tangential acceleration vanishes, aT = v ·a = 0. Hence, the speed of the par-
ticle is a constant of motion, v = v0 (because v 0 = aT = 0). Let us make the
orthogonal decomposition of v relative to the magnetic field: v = v⊥ + vk ê3 ,
where v⊥ is in the xy plane. Since kvk = v0 , the magnitude of v⊥ is also
constant, kv⊥k = v⊥ = (v02 − vk2 )1/2 . The velocity vector can therefore be
written in the form v = hv⊥ cos ϕ, v⊥ sin ϕ, vk i, where the function ϕ = ϕ(t)
is to be determined by the equations of motion:
a = µB v × ê3 = µB hv⊥ sin ϕ, −v⊥ cos ϕ, 0i ,
a = v 0 = ϕ0 h−v⊥ sin ϕ, v⊥ cos ϕ, 0i .
It follows from the comparison of these expressions that ϕ0 (t) = −µB or
ϕ(t) = −µBt + ϕ0 = ωt + ϕ0 where ω = eB/(mc) is the so-called cyclotron
frequency and the integration constant ϕ0 is determined by the initial veloc-

ity: v(0) = hv⊥ cos ϕ0 , v⊥ sin ϕ0 , vk i, i.e., tan ϕ0 = v2 (0)/v1 (0). Integration
of the equation
r0 (t) = v(t) = hv⊥ cos(ωt − ϕ0 ), −v⊥ sin(ωt − ϕ0 ), vk i
yields the trajectory of motion:
r(t) = r0 + hR sin(ωt − ϕ0 ), R cos(ωt − ϕ0 ), vkti,
where R = v⊥ /ω. This equation describes a helix of radius R whose axis
goes through r0 parallel to the z axis. So a charged particle moves along
a helix that winds about force lines of the magnetic field. The particle
revolves in the plane perpendicular to the magnetic field with frequency
ω = eB/(mc). In each turn, the particle moves along the magnetic field a
distance h = 2πvk /ω. In particular, if the initial velocity is orthogonal to
the magnetic field (i.e., vk = 0), then the trajectory is a circle of radius R.
The polar lights. The Sun produces a stream of charged particles (the
solar wind). The magnetic field of the Earth plays the role of a shield from
the solar wind as it traps the particles forcing them to travel along its force
lines that are arcs connecting the magnetic poles of the Earth (which ap-
proximately coincide with the south and north poles). As a result, the solar
wind particles can penetrate the lower atmosphere only near the magnetic
poles of the Earth causing a spectacular phenomenon, the polar lights, by
colliding with molecules of the oxygen and nitrogen in the atmosphere.
Problem 15.5. Suppose that the force acting on a particle of mass m
is proportional to the position vector of the particle (such forces are called
central). Prove that the angular momentum of the particle, L = mr × v, is
a constant of motion (i.e., dL/dt = 0).
Solution: Since a central force F is parallel to the position vector r, their
cross product vanishes, r × F = 0. By Newton’s second law, ma = F and
hence mr × a = 0. Therefore,
dL
= m(r × v)0 = m(r0 × v + r × v 0 ) = mr × a = 0,
dt
where r0 = v, v 0 = a, and v × v = 0 have been used.
Problem 15.6. (Kepler’s Laws of Planetary Motion). Newton’s law of grav-
ity states that two masses m and M at a distance r are attracted by a force
of magnitude GmM/r 2 , where G is the universal constant (called Newton’s
constant). Prove Kepler’s laws of planetary motion:
1. A planet revolves around the Sun in an elliptical orbit with the Sun at
one focus.
2. The line joining the Sun to a planet sweeps out equal areas in equal times.
3. The square of the period of revolution of a planet is proportional to the
cube of the length of the major axis of its orbit.
Solution: Let the Sun be at the origin of a coordinate system and let
r be the position vector of a planet. The mass of the Sun is much larger
than the mass of a planet and therefore a displacement of the Sun due to
the gravitational pull from a planet can be neglected (e.g., the Sun is about
332946 times heavier than the Earth). Let r̂ = r/r be the unit vector parallel
to r. Then the gravitational force is
GM m GM m
F=− r̂ = − r,
r2 r3
where M is the mass of the Sun and m is the mass of a planet. The minus
sign is necessary because an attractive force must be opposite to the position
vector. By Newton’s second law, the trajectory of a planet satisfies the
equation ma = F and hence
GM
a=− r.
r3
The gravitional force is a central force, and, by Study Problem 15.5, the
vector r × v = l is a constant of motion. One has v = r0 = (rr̂)0 = r 0 r̂ + rr̂0 .
Using this identity, the constant of motion can also be written as
l = r × v = rr̂ × v = r(r 0 r̂ × r̂ + rr̂ × r̂0 ) = r 2 (r̂ × r̂0 ).
Using the rule for the double cross product (see Study Problem 4.4), one
infers that
GM
a × l = − 2 r̂ × l = −GMr̂ × (r̂ × r̂0 ) = GMr̂0 ,
r
where r̂ · r̂ = 1 has been used. On the other hand,
(v × l)0 = v 0 × l + v × l0 = a × l
because l0 = 0. It follows from these two equations that
(15.10) (v × l)0 = GMr̂0 =⇒ v × l = GMr̂ + c,
where c is a constant vector. The motion is characterized by two constant
vectors l and c. It occurs in the plane through the origin that is orthogonal
to the constant vector l because l = r × v must be orthogonal to r. It
also follows from (15.10) and l · r̂ = 0 that the constant vectors l and c
are orthogonal because l · c = 0. It is therefore convenient to choose the
coordinate system so that l is parallel to the z axis and c to the x axis as
shown in Figure 15.3 (left panel).
The vector r lies in the xy plane. Let θ be the polar angle of r (i.e.,
r · c = rc cos θ, where c = kck is the length of c). Then
r · (v × l) = r · (GMr̂ + c) = GM r + rc cos θ.
On the other hand, using a cyclic permutation in the triple product,
r · (v × l) = l · (r × v) = l · l = l 2 ,
Figure 15.2. Left: The setup of the coordinate system for

the derivation of Kepler’s first law. Right: An illustration
to the derivation of Kepler’s second law.
where l = klk is the length of l. The comparison of the last two equations
yields the equation for the trajectory:
ed
l 2 = r(GM + c cos θ) =⇒ r = ,
1 + e cos θ
where d = l 2 /c and e = c/(GM ). This is the polar equation of a conic section
with focus at the origin and eccentricity e (see Calculus II). Thus, all possible
trajectories of any massive body in a solar system are conic sections! This
is a quite remarkable result. Parabolas and hyperbolas do not correspond
to a periodic motion. So a planet must follow an elliptic trajectory with the
Sun at one focus. All objects coming to the solar system from outer space
(i.e., those that are not confined by the gravitational pull of the Sun) should
follow either parabolic or hyperbolic trajectories.
To prove Kepler’s second law, put r̂ = hcos θ, sin θ, 0i and hence r̂0 =
h−θ sin θ, θ0 cos θ, 0i. Therefore,
0
l = r 2 (r̂ × r̂0 ) = h0, 0, r 2θ0 i =⇒ l = r 2 θ0 .

The area of a sector with angle dθ swept by r is dA = 12 r 2 dθ (see Calculus
II; the area bounded by a polar graph r = r(θ)). Hence,
dA 1 dθ l
= r2 = .
dt 2 dt 2
For any moments of time t1 and t2 , the area of the sector between r(t1 ) and
r(t2 ) is
Z t2 Z t2
dA l l
A12 = dt = dt = (t2 − t1 ).
t1 dt t1 2 2
Thus, the position vector r sweeps out equal areas in equal times (see Fig-
ure 15.3, right panel).
Kepler’s third law follows from the last equation. Indeed, the entire area
of the ellipse A is swept when t2 − t1 = T is the period of the motion. If the
major and minor axes of the ellipse are 2a and 2b, respectively, a > b, then
A = πab = lT /2 and T = 2πab/l. Now recall that ed = b2 /a for an elliptic
conic section (see Calculus II) or b2 = eda = l 2 a/(GM ). Hence,
4π 2 a2 b2 4π 2 3
T2 = = a .
l2 GM
Note that the proportionality constant 4π 2 /(GM ) is independent of the mass
of a planet; therefore, Kepler’s laws are universal for all massive objects
trapped by the Sun (planets, asteroids, and comets).
15.4. Exercises.
1–7. For each of the following trajectories of a particle, find the velocity,
speed, and the normal and tangential acceleration as functions of time, and
their values at a specified point P :
1. r(t) = ht, 1 − t, t2 + 1i , P = (1, 0, 2) ;
2. r(t) = ht2 , t, 1i , P = (4, 2, 1) ;
3. r(t) = h4t3/2 ,√−t2 , ti , P = (4, −1, 1) ;
4. r(t) = hln t, t, t2 i , P = (0, 1, 1) ;
5. r(t) = hcosh√t, sinh t, 2 + ti , P = (1, 0, 2);
6. r(t) = het , 2 t, e−t i , P = (1, 0, 1);
7. r(t) = hsin t − t cos t, t2 , cos t + t sin ti , P = (0, 0, 1) .
8. Find the normal and tangential accelerations of a particle with the posi-
tion vector r(t) = ht2 + 1, t, t2 − 1i when the particle is at the least distance
from the origin.
9. Find the tangential and normal accelerations of a particle with the po-
sition vector r(t) = hR sin(ωt + ϕ0 ), −R cos(ωt + ϕ0 ), v0ti, where R, ω, ϕ0 ,
and v0 are constants (see Study Problem 15.4).
10. The shape of a winding road can be approximated by the graph y =
L cos(x/L), where the coordinates are in miles and L = 0.1 mile. The condi-
tion of the road is such that if the normal acceleration of a car on it exceeds
0.13g, where g is the acceleration of the free fall, the car may skid off the
road. Recommend a speed limit for this portion of the road.
11. A particle moves along the curve y = x2 + x3 in the direction of increas-
ing x. If the acceleration of the particle at the point (1, 2) is a = h−3, −1i,
find its normal and tangential accelerations.
12. Suppose that a particle moves so that its tangential acceleration aT is
constant, while the normal acceleration aN remains 0. What is the trajec-
tory of the particle?
13. Suppose that a particle moves in a plane so that its tangential accelera-
tion aT remains 0, while the normal acceleration aN is constant. What is the
trajectory of the particle? Hint: Investigate the curvature of the trajectory.
14. A race car moves with a constant speed v0 along an elliptic track
x2 /a2 + y 2 /b2 = 1, a > b. Find the maximal and minimal values of the
magnitude of its acceleration and the points where they occur.
15. Does there exist a curve with zero curvature and a non-zero torsion?
Explain the answer.
16–20. For each of the following curves, find the unit tangent, normal, and
binormal vectors and the torsion at a specified point P :
16. r(t) = ht, 1 − t, t2 + 1i , P = (1, 0, 2) ;
17. r(t) = ht3 , t2 , 1i , P = (8, 4, 1);
18. r(t) = h4t3/2 , √−t2 , ti , P = (4, −1, 1) ;
19. r(t) = hln t, 2 t, t2 i , P = (0, 2, 1);
20. r(t) = hcosh t, sinh t, 2 + ti , P = (1, 0, 2).
21. Let r(t) = hcos t + t sin t, sin t − t cos t, t2 i. Find the speed, the tangential
and normal accelerations, the curvature and torsion, and the unit tangent
vector, normal, and binormal as functions of time t.
Hint: To simplify calculations, find the decomposition r(t) = v(t) − tw(t) +
t2 ê3 where v, w, and ê3 are mutually orthogonal unit vectors such that
v 0 (t) = w(t), w0 (t) = −v(t). Use the properties of the cross products of
mutually orthogonal unit vectors.
22. Let C be the curve of intersection of an ellipsoid x2 /a2 +y 2 /b2 +z 2 /c2 = 1
with the plane 2x − 2y + z = 0. Find the torsion and the binormal B̂ along
C.
Selected Answers and Hints to Exercises

Section 10. 1. −∞ < t < ∞. 2. t ≥ 0. 3. 0 < t ≤ 3. 4. The
union of three open intervals (−3, −2), (−2, 0), and (0, 3). 5. t = 1. 6.
h1, 0, −1i. 7. The limit does not exist. 8. h1, 0, 0i. 9. h0, 0, 4i. 10.
h0, −1, 0i. 11. The limit does not exist. 12. h2, 12 , 0i. 13. h4, 2, − 12 i.
14. h2, 1, 1i. 15. h4, 0, 14 i. 16. A helix about the x axis, radius is 1; the
helix rises by 2π/3 units per turn. 17. An ellipse that is the intersection
of the elliptic cylinder x2 /4 + z 2 /9 = 1 and the coordinate plane y = 4.
19. The curve is obtained by wrapping the graph of ln t on the cylinder
x2 + y 2 = 1. The line parallel to the z axis through the point (0, 1, 0) (it
lies on the cylinder) is a vertical asymptote of the curve as t → 0+ . 21.
The curve is obtained by wrapping the graph of sin2 (πt) to the parabolic
cylinder x = y 2 . 22. The curve is the circle of intersection of the sphere
x2 + y 2 + z 2 = 2 and the plane y = x. 23. No collision. Two points
of intersection: r1 (1) = r2 (0) = h1, 1, 1i and r1 (2) = r2 (1/2) = h2, 4, 8i.
24. x = t, z = t2 , y = 2(1 − t2 − t4 /9)1/2. 25. r(t) = ht, t2 , 1i. 26. r(t) =
hsin t, t, sin ti. 27. r(t) = h3 cos t, 3 sin t, 29 sin(2t)i. 29. r(t) = ht, t2 , t2 +t4 i.
30. r(t) = h2 cos t, 3 sin t, 1 − 2 cos t − 3 sin ti. 31. r(t) = hcos t, cos t, 3 sin ti.
32. r(t) = h1 + cos t, sin t, 2 + 2 cos ti. 33. π6 ≤ t ≤ 5π 7π
6 and 6 ≤ t ≤ 6 .
11π
34. a = 0 and b = 4. 35. r(0) = h0, 1, 1i. 36. r(0) = h0, 1, 1i. 37. No such
r(0) exists. 38. r(0) = h2, 3, 0i. 39. r(0) = h 12 , 0, 2i. 42. h0, 1, 32 i. 43. 9.
44. h− 14 , 1, 2i. 45. 4. 46. h3, − 14 , 12 i. 47. 0 (by the Jacobi identity).
Section 11. 1. r0 (t) = h0, 1, 3t2i. 2. r0 (t) = h− sin t, sin(2t), 2ti. 3. r0 (t) =
h 1t , 2e2t, e−t(1 − t)i. 4. r0 (t) = h 13 (t − 2)−2/3 , t(t2 − 4)−1/2 , 1i. 5. r0 (t) =
2tb − et c. 6. r0 (t) = a × b − et (t + 1)a × c. 8. smooth everywhere. 9.
smooth everywhere except the point (0, 0, 2). 12. smooth everywhere except
the point (0, 1, 0). 14. x = 6+5t, y = 9+9t, z = 6+2t. 15. x = t, y = 2+t,
z = 1 + 2t. 16. h 23 , 23 , − 13 i. 18.r0 · r00 = 4t + 18t3 , r0 × r00 = h6t2 , −6t, 2i.
19. Yes, at the point r(−2) = h6, − 83 , −4i. 21. The point of intersection is
4
r(1) = h0, 1, 2i, the angle is cos−1 ( 3√ ). 22. No intersection; the distance
2√
between the curve and the plane is 4/ 3. 23. The point of intersection is
r1 (1) = r2 (2) = h1, 0, 4i; the angle is cos−1 ( √13 ). 25. π/2. 30. x = 1 + t,
y = 1 + t, z = 1 + 2t (tangent line); x + y + 2z = 4 (normal plane). 31.
x = 1 + 3t, y = 1 + 3t, z = 3 − t (tangent line); 3x + 3y − z = 3 (normal
plane). 32. x = 1 + t, y = −2, z = 1 − t (tangent line); x − z = 0 (normal
plane).
Section 12. 1. h2, 4, 8i (definite integral). 2. h0, 0, 0i (definite integral).

3. h 13 , 15
4 2
, 3 i (definite integral). 4. h− 41 , 13 , 12 (e2 − 1)i (definite integral). 5.
h1, 1, 1i (definite integral). 6. πa (definite integral). 7. 0 (definite integral).
8. h1+t, 2+t2 , 3+t3 i. 9. h 12 t2 −t+ 32 , 13 t3 − 13 , 23 t3/2 + 13 i. 10. h 32 − 12 cos(2t), 2+
2 sin(t), 3 − π2 + 2t − 14 sin(2t)i. 12. ht + 1, t2 + 2, t3 − t + 3i. 13. ht + 28 9 7/3
(t −
SELECTED ANSWERS AND HINTS TO EXERCISES 239
2
1), 2t + 15 (2t5/2 − 17), t3 − 2i. 14. h1 + sin t, −2 − cos t, t(1 + ln(t/π)) −
πi. 16. cn tn−1 + cn−1 tn−2 + · · · + c2 t + c1 where ck , k = 1, 2, ..., n, are
arbitrary constant vectors. Alternatively, r(t) = hPn−1 (t), Qn−1 (t), Rn−1 (t)i
where Pn−1 , Qn−1 , and Rn−1 are polynomials of degree at most n − 1. 18.
r(2) = h3, 4, 31 i. 20. 18 meters. 22. The maximal height is approximately
2 kilometers, the range is approximately 14 kilometers, and the speed at
impact is 400 m/s. 23. Let the x axis be from east to west, the y axis
from north to south, the z axis is vertical, and the origin at the initial
point. Then the point of impact is x = 2v02 F sin2 θ/(mg 2 ), y = v02 sin(2θ)/g,
z = 0. Change the initial velocity by adding a nonzero x component: v0 =
hu, v0 cos θ, v0 sin θi where u = −F v0 sin θ/(mg), then the impact point is at
x = z = 0, y = v02 sin(2θ)/g as if F = 0.
√ −1 −1
√ √
Section
√ 13. 1. 4 13 2. 7/3. 3. 18. 4. 2(e − e ). 5. (e − e )/ 2. 7. 3.
8. √2 2. 9. √ 23. 10. 8 + ln 3. 11. 42. 14. Parametric equations are r(t) =
√
ha t cos t, a t sin t, ati; the arc length is a t0 (1 + 2t0 /3) where t0 = z0 /a.
16. R(s) = h √s14 , 1 − √2s14 , 5 + √3s14 i. 17. R(s) = sin(s)ê1 + cos(s)ê3 (the
√ √
R(s) = h 1 + s2 /2, s/ 2, sinh−1 (s/ 2i; recall that
p
curve is a circle). 18.
√ √
sinh−1 (y) = ln(y+ y 2 + 1). 22. The position vector is hln(1+ 2), 2, 1i.
p
Section 14. 1. κ(t) = (1 + 2t2 )−3/2, κ(1) = 3−3/2. 2. κ(t) = 2(1 + 4t2 )3/2,
κ(2) = 2/173/2. 3. κ(x) = 14 | sin(x/2)|(1 + 41 cos2 (x/2))−3/2, κ(π) = 14 .
√ √
9. κ(t) = 2(et + e−t )−2 , κ(0) = 2−3/2 . 11. 2/3. 12. x = πn/a
where n is any integer. 17. y = x4 , x0 = 0. 18. The curvatures are
3
equal. 19. κ(θ) = 2√ 2
(1 + cos θ)−1/2 (the curve is not smooth at the origin,
√
θ = π, where the curvature becomes infinite). 20. κ(θ) = e−θ / 2. 22.
x2 + (y − R)2 = R2 where R = 1/κ(0) = 1/2. 24. (x − 2)2 + (y − 2)2 = 2.
27. κmax = a/b2 occurs at (±a, 0), κmin = b/a2 occurs at (0, ±b); the
osculating circle at (a, 0) is (x − a + b2 /a)2 + y 2 = b4 /a2 ; the osculating
circle at (0, b) is x2 + (y − b + a2 /b)2 = a4 /b2 . 28. κ(t) → ∞ as t → 0. 30.
2x + 3y − 6z = −1. 31. 6x − 16y + z = −15. 33. Yes, at (1, 1, 1).
√
Section
√ 15. 1. v(t) = h1,
√ −1, 2ti, a(t) = h0, 0, 2i, v(t) = 2 + 4t2 , aT (t) =
2 2
4t/ 2 + 4t , aN (t) = 2/ 1 + 2t . The point √ P corresponds to t√= 1. 2.
v(t) = h2t, 1, 2 2
√ 0i, a(t) = h2, 0, 0i, v(t) = 1 + 4t , aT (t) = 4t/ 1 + 4t ,
2
aN√(t) = 2/ 1 + 4t . The p point P corresponds √ to t = 2. 3. v(t) =
h6 2
p t, −2t, 1i, a(t) = h3/ (t), −2, 0i, v(t) = 4t + 36t + 1, kv × ak =
36t + 4 + 9/t, aT (t) = (18 + 4t)/v(t), √ aN (t) 2= kv × ak/v(t). The√ point2P
1 1
corresponds to t = 1. 4. v(t) = t h1, t/2, 2t i, a(t) = t2 h−1, − t/4, 2t i,
q
v(t) = 1t 1 + 4t4 + t/2, kv × ak = t13 94 t5 + 16t4 + t/16, v · a = t13 (4t2 −
p
1 − t/8), aT (t) = v · a/v(t), aN (t) √= kv × ak/v(t). The point P corresponds

to t = 1. 8. aT = 0 and aN = 2 2. 10. The maximal speed is 51 km/h
or 32 mph. A suitable speed limit is 30 mph or 50 km/h. 11. A tangent

vector is h1, 5i, and a normal vector h−5, 1i. Hence the magnitudes √ of the
tangential√and normal accelerations are, respectively, aT = −8/ 26 and
aN = 14/ 26. 12. A straight line. 13. A circle. 14. Maximal acceler-
ation v02 a/b2 occurs at (±a, 0). 17. T̂ = √110 h3, 1, 0i, N̂ = √110 h1, −3, 0i,
B̂ = h0, 0, −1i, τ = 0. 18. T̂ = √1 h6, −2, 1i, N̂ = √1 h−9, −38, −22i,
41 7 41
1 3
B̂ = 7 h2, 3, −6i, τ = − 49 . 19. 1
T̂ = √16 h1, 1, 2i, N̂ = √606 h−17, −11, 14i,
B̂ = √ 1 h6, −8, 1i, τ = 12
B̂ = 13 h2, −2, 1i, τ = 0.
101 101 . 22.

Vector Calculus

Uploaded by

Copyright:

Available Formats

Vector Calculus

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

Vector Calculus

Uploaded by

Copyright:

Available Formats

CHAPTER 2

10. Curves in Space and Vector Functions

Figure 10.1. Left: The terminal point of a vector r(t)

10.1. Graphing Space Curves. To visualize the shape of a curve C traced

Figure 10.2. Graphing a helix. Left: The shadow curve

coordinates (x(tn ), y(tn)), n = 1, 2, ..., N . Then the corresponding points

10.2. Limits and Continuity of Vector Functions.

the limit is denoted as limt→t0 r(t) = r0 .

Remark. Although l’Hospital’s rule is a powerful tool to resolve undeter-

as t → 0 because O(t) → 0. This method becomes superior to l’Hospital’s

A vector function r(t) is continuous in the interval [a, b] if it is continuous

10.3. Space Curves and Continuous Vector Functions. A curve connecting

The following, more sophisticated example is also of interest. Suppose

are called parametric equations of C, and t is called a parameter. As noted, a

10.4. Study Problems.

Figure 10.3. Illustration to Study Problem 10.2. Left:

Figure 10.4. Illustration to Study Problem 10.3. Left:

Solution: The components of r(t) satisfy the equation

Figure 10.5. Illustration to Study Problem 10.6. The

By construction, this circle is the vertical projection, hx(t), y(t), z(t)i →

11. Differentiation of Vector Functions

It follows from Theorem 10.1 that a vector function is differentiable if

r(t) = hsin(2t), t2 − t, e−3t i ⇒ r0 (t) = h2 cos(2t), 2t − 1, −3e−3t i.

Definition 11.2. (Continuously Differentiable Vector Function).

Higher-order derivatives are defined similarly: the second derivative is

r00 (t) = (r0(t))0 = hx00 (t) , y 00 (t) , z 00 (t)i ,

the third derivative is the derivative of r00(t):

r000(t) = (r00 (t))0 = hx000 (t) , y 000 (t) , z 000 (t)i ,

r(n)(t) = (r(n−1)(t))0 = hx(n)(t) , y (n) (t) , z (n) (t)i ,

provided all components are differentiable sufficiently many times.

11.1. Differentiation Rules. The following rules of differentiation of vector

Theorem 11.1. (Differentiation Rules).

11.2. Differential of a Vector Function. If r(t) is differentiable, then

Figure 11.1. Left: A secant line through two points of

11.3. Geometrical Significance of the Derivative. Among all the lines

v = r0 (1) = h2, 2t, 3t2 + 2i = h2, 2, 5i .

Parametric equations of the line through P0 = (2, 0, 3) and parallel to v are

11.4. Smooth curves. Let r(t) be a parameterization of a curve C. Can a

Definition 11.4. (Smooth Curve).

If t > 0, then t/|t| = 1 and, if t < 0, then t/|t| = −1. Hence,

kr02 (t)k = (9t4 + 25t8 )1/2 = 3t2 (1 + 25t4 /9)1/2 ,

11.5. Study Problems.

Problem 11.1. Determine whether the cycloid x = a(t − sin t),

Solution: The existence of a continuous unit tangent vector has to be

the components of the unit tangent vector for t 6= 2πn are

The vector function r(t) can be written in the form

5. r(t) = a + bt2 − cet ;

12. Integration of Vector Functions

where c1 , c2 , and c3 are constants. The latter relations can be combined

where c is an arbitrary constant vector.

Recall further from Calculus I that for a function x(t) continuous on

and X(t) is differentiable on (a, b) and satisfies the condition X(a) = 0.

if v(t) is a continuous vector function on an interval (t0 lies in the in-

12.1. Applications to Mechanics. Let r(t) be the position vector of a par-

of differential equations requires a special technique to integrate such equa-

Figure 12.1. Left: Motion under a constant force F. The

12.2. Motion Under a Constant Gravitational Force. The magnitude of

12.3. Study Problems.

Here the constant vector c is determined by the initial condition r(0) =

Put B = h0, 0, Bi. Then