When functions have no value(s):

Delta functions and distributions

Steven G. Johnson, MIT course 18.303 notes
Created October 2010, updated March 8, 2017.

Abstract x = 0. That is, one wouldRlike the function δ(x) = 0

for all x 6= 0, but with δ(x)dx = 1 for any in-
These notes give a brief introduction to the mo- tegration region that includes x = 0; this concept
tivations, concepts, and properties of distributions, is called a “Dirac delta function” or simply a “delta
which generalize the notion of functions f (x) to al- function.” δ(x) is usually the simplest right-hand-
low derivatives of discontinuities, “delta” functions, side for which to solve differential equations, yielding
and other nice things. This generalization is in- a Green’s function. It is also the simplest way to
creasingly important the more you work with linear consider physical effects that are concentrated within
PDEs, as we do in 18.303. For example, Green’s func- very small volumes or times, for which you don’t ac-
tions are extremely cumbersome if one does not al- tually want to worry about the microscopic details
low delta functions. Moreover, solving PDEs with in this volume—for example, think of the concepts of
functions that are not classically differentiable is of a “point charge,” a “point mass,” a force plucking a
great practical importance (e.g. a plucked string with string at “one point,” a “kick” that “suddenly” imparts
a triangle shape is not twice differentiable, making some momentum to an object, and so on. The prob-
the wave equation problematic with traditional func- lem is that there is no classical function δ(x) having
tions). Any serious work with PDEs will eventually these properties.
run into the concept of a “weak solution,” which is For example, one could imagine constructing this
essentially a version of the distribution concept. function as the limit:
(
1
∆x 0 ≤ x < ∆x
1 What’s wrong with functions? δ(x) = lim = lim δ∆x (x)?
∆x→0+ 0 otherwise ∆x→0+

The most familiar notion of a function f (x) is a map

For any ∆x > 0, the function δ∆x (x) at right has
from real numbers R to real numbers R (or maybe
integral = 1 and is zero except near x = 0. Un-
complex numbers C); that is, for every x you have
fortunately, the ∆x → 0 limit does not exist as an
a value y = f (x). Of course, you may know that
ordinary function: δ∆x (x) approaches ∞ for x = 0,
one can define functions from any set to any other
but of course ∞ is not a real number.
set, but at first glance it seems that R → R func-
Informally, one often sees “definitions” of δ(x) that
tions (and multi-dimensional generalizations thereof)
describe it as some mysterious object that is “not
are the best-suited concept for describing most phys-
quite” a function, which = 0 for x 6= 0 but is unde-
ical quantities—for example, velocity as a function
fined at x = 0, and which is “only really defined inside
of time, or pressure or density as a function of posi-
an integral” (where it “integrates” to 1).1 This may
tion in a fluid, and so on. Unfortunately, such func-
leave you with a queasy feeling that δ(x) is somehow
tions have some severe drawbacks that, eventually,
lead them to be replaced in some contexts by an- 1 Historically, this unsatisfactory description is precisely

other concept: distributions (also called generalized how δ(x) first appeared, and for many years there was a corre-
sponding cloud of guilt and uncertainty surrounding any usage
functions). What are these drawbacks? of it. Most famously, an informal δ(x) notion was popularized
by physicist Paul Dirac, who in his Principles of Quantum Me-
chanics (1930) wrote: “Thus δ(x) is not a quantity which can
1.1 No delta functions be generally used in mathematical analysis like an ordinary
function, but its use must be confined to certain simple types
For lots of applications, such as those involving PDEs of expression for which it is obvious that no inconsistency can
and Green’s functions, one would like to have a func- arise.” You know you are on shaky ground, in mathematics,
tion δ(x) whose integral is “concentrated” at the point when you are forced to appeal to the “obvious.”

1
not real or rigorous (and therefore anything based on u 6= 0. . . except for u(x) that are only nonzero for iso-
it may be suspect). For example, integration is an lated points. And with functions like S(x) there are
operation that is classically only defined for ordinary
R all sorts of apparently pointless questions about what
functions, so it may notR even be clear (yet) what “ ” value to assign exactly at the discontinuity. It may
means when we write “ δ(x)dx”. likewise seem odd to care about what value to assign
the slope of |x| at x = 0; surely any value in [−1, 1]
should do?
1.2 Not all functions are differentiable
Most physical laws can be written in the form of 1.4 Limits and derivatives
derivatives, but lots of functions are not differen-
cannot always be interchanged
tiable. Discontinuous functions arise all of the time
at the interface between two materials (e.g. think In numerical and analytical PDE methods, we are
of the density at the interface between glass and air). continually writing functions as limits. A Fourier se-
Of course, at a microscopic level you could argue that ries is a limit as the number of terms goes to infinity.
the quantum wavefunctions might be continuous, but A finite-difference method solves the problem in the
one hardly wishes to resort to atoms and quantum limit as ∆x → 0. We initially find the Green’s func-
mechanics every time a material has a boundary! tion as the limit of the response to a box-like function
The classic example of a discontinuous function is that is nonzero outside of a width ∆x, and then take
the Heaviside step function: the limit ∆x → 0. After all of these kinds of things,
( we eventually substitute our solution back into the
1 x≥0 PDE, and we assume that the limit of the solution is
S(x) = .
0 x<0 still a solution. In doing so, however, we usually end
up interchanging the limits and the differentiation.
The derivative S 0 (x) is zero everywhere except at For example, we usually differentiate a Fourier series
x = 0, where the derivative does not exist—the slope by:
at x = 0 is “infinity.”
R x Notice that S 0 (x) very much ∞ ∞ ∞
d X X d X
resembles δ(x), and ∞ δ(x )dx0 would certainly look
0
an sin(nπx) = an sin(nπx) = nπan cos(nπx),
dx n=1 dx
something like S(x) if it existed (since the “integral” n=1 n=1
of δ should be 1 for x > 0)—this is not a coincidence, but this is not always true if we interpret “=” in the
and it would be a shame not to exploit it! usual sense of being true for every x. The general
A function doesn’t need to be discontinuous to lack problem is that one cannot always interchange limits
a derivative. Consider the function |x|: its derivative and derivatives, i.e. dx d d
lim 6= lim dx . (Note that
is +1 for x > 0 and −1 for x < 0, but at x = 0 P∞ PN
the derivative doesn’t exist. We say that |x| is only n=1 is really a limit limN →∞ n=1 .)
A simple example of this is the function
“piecewise” differentiable. 
Note that S(x) is very useful for writing down all − x < −
 (
sorts of functions with discontinuities. |x| = xS(x) − 0 1 − < x < 
f (x) = x − ≤ x ≤  , f (x) = .
xS(−x), for example, and the δ∆x (x) “box” function  0 |x| > 
 x>

on the right-hand-side of the δ(x) “definition” above
can be written δ∆x (x) = [S(x) − S(x − ∆x)]/∆x. The limit as  → 0 of f (x) is simply zero. But
f0 (0) = 1 for all , so
1.3 Nagging worries about d   
d

discrepancies at isolated points 0= lim f 6= 1 = lim f .
dx →0 x=0
→0 dx x=0
When we try to do linear algebra with functions, we Notice, however, that this mismatch only occurs at
continually find ourselves worrying about excluding one isolated point x = 0. The same thing happens
odd caveats and counterexamples that have to do for Fourier series: differentiation term-by-term works
with finite discrepancies at isolated points. For ex- except at isolated points. Thus, this problem returns
ample, a Fourier series of a square-integrable function to the complaint in the previous section 1.3.
converges everywhere. . . except at isolated points of
discontinuity [like the point x = 0 for S(x), where
1.5 Too much information
a Fourier series would Rconverge to 0.5]. As an-
other example, hu, vi = ūv defines an inner prod- A function f (x) gives us a value at every point x, but
uct on functions and a norm kuk2 = hu, ui > 0 for does this really correspond to a measurable quantity

2
in the physical universe? How would you measure the [The generalization to functions φ(~x) for ~x ∈ Rd , with
velocity at one instant in time, or the density at one the distributions corresponding to d-dimensional in-
point in a fluid? Of course, your measurement device tegrals, is very straightforward, but we will stick to
could be very precise, and very small, and very fast, d = 1 for simplicity.] Second,4 we require that f {φ}
but in the end all you can ever measure are averages act like integration in that it must be linear :
of f (x) over a small region of space and/or time.
• f {αφ1 + βφ2 } = αf {φ1 } + βf {φ2 } for any num-
R But an average is the same thing as an integral bers α, β ∈ R and any φ1 , φ2 ∈ D.
f (x) over the averaging region. More generally, in-
stead of averaging f (x) uniformly in some region, we Thus, f is a linear map from D → R, and the set
could average with some weights φ(x) (e.g. our de- of all distributions for a given set of test functions
vice could be more sensitive to some points than oth- D is sometimes denoted D0 . There are two classes of
ers). Thus, the only physical question we can ever distributions: regular and singular distributions.
ask about a function is the value of an integral
Z ∞
f (x)φ(x)dx
2.1 Regular distributions
−∞ from ordinary functions f (x)
of f (x) against a test function φ(x). The most obvious way to define a distribution f {φ} is
But if all we can ever ask is such an integral, why simply an ordinary integral of an ordinary function:
are we worrying about isolated points? In fact, why given an ordinary function f (x), we can define the
do we even define f (x) to have values at points at distribution:
all? In a physical application of mathematics, per- Z ∞
haps we should only define things that we can mea-
f {φ} = f (x)φ(x)dx .
sure! This insight fixes every one of the problems −∞
above, and leads to the concept of distributions.
This is called a regular distribution.

2 Distributions
Not all ordinary functions R f (x) define regular dis-
tributions: we must have f φ finite for all φ ∈ D.
Rb
The old kind of function is a map from R → R: given This reduces to requiring that a |f (x)|dx < ∞ for
an x, we get a value f (x). Following section 1.5, all intervals [a, b] (f is “locally integrable”).
this is too much information; we can only ask for an
“average” value given some weight function φ(x). So, 2.2 Singular distributions
we make a new definition of “function” that provides & the delta function
this information, and only this information:
Although the integral of an ordinary function is one
• f is a rule that given any test function φ(x) re- way to define a distribution, it is not the only way.
turns a number f {φ}.2 For example, the following distribution is a perfectly
This new definition of a “function” is called a dis- good one:
tribution or a generalized function. We are no δ{φ} = φ(0) .
longer allowed to ask the value at a point x. This
This rule is linear and continuous, there are no
will fix all of the problems with the old functions from
weird infinities, nor is there anything spooky or non-
above. However, we should be more precise about our
rigorous. Given a test function φ(x), the δ(x) distri-
definition. First, we have to specify what φ(x) can
bution is simply the rule that gives φ(0) from each φ.
be:
δ(x), however, does not correspond to any ordinary
• φ(x) is an ordinary function R → R (not a dis- function—it is not a regular distribution—so we call
tribution) in some set D. We require φ(x) to be it a singular distribution.
infinitely differentiable. We also require φ(x) to R
be nonzero only in some finite region (the “sup- • Notation: when we write “ δ(x)φ(x)dx,” we
port” of φ): φ is a smooth “bump” function.3 don’t mean an ordinary integral, we really mean
δ{φ} = φ(0).
2 Many authors use the notation hf, φi instead of f {φ}, but

we are already using h·, ·i for inner products and I don’t want 4 There is also a third requirement: f {φ} must be contin-

to confuse matters. uous, in that if you change φ(x) continuously the value of
3 Alternatively, one sometimes loosens this requirement to f {φ} must change continuously. In practice, you will never
merely say that φ(x) must → 0 quickly as x → ±∞, and in violate this condition unless you are trying to construct per-
particular that xn φ(x) → 0 as x → ±∞ for all integers n ≥ 0. verse counter-examples.

3
Furthermore, when we say “f = δ(x − x0 ),” we just S(x):
mean the distribution f {φ} = φ(x0 ). (This kind Z ∞ Z ∞
of “abuse” of notation is too convenient, and too S{φ} = S(x)φ(x)dx = φ(x)dx.
widespread, to discard in the name of purity.) −∞ 0

If we look at the “finite-∆x delta” approximation It immediately follows that the distributional deriva-
δ∆x (x) from section 1.1, that defines the regular dis- tive of the step function is
tribution: Z ∞
0 0 ∞
Z ∆x S {φ} = S{−φ } = − φ0 (x)dx = −φ(x)|0
1 0
δ∆x {φ} = φ(x)dx,
∆x 0 = φ(0) −  φ(∞)
 = φ(0).

which is just the average of φ(x) in [0, ∆x]. Now, But this is exactly the same as δ{φ}, so we immedi-
0
however, viewed as a distribution, the limit ∆x → 0 ately conclude: S = δ .
5
is perfectly well defined: lim∆x→0 δ∆x {φ} = φ(0) = Since any function with jump discontinuities can be
δ{φ}, i.e. δ∆x → δ. written in terms of S(x), we find that the derivative
Of course, the δ distribution is not the only singular of any jump discontinuity gives a delta function mul-
distribution, but it is the most famous one (and the tiplied by the magnitude of the jump. For example,
one from which many other singular distributions are consider:
(
built). We will see more examples later. x2 x < 3
f (x) = = x2 + (x3 − x2 )S(x − 3).
x3 x ≥ 3
2.3 Derivatives of distributions
The distributional derivative works just like the ordi-
& differentiating discontinuities nary derivative, except that S 0 = δ, so6
How do we define the derivative f 0 of a distribution? f 0 (x) = 2x + (3x2 − 2x)S(x − 3) + (33 − 32 )δ(x − 3)
Well, at the very least we want it to be the same as (
the ordinary derivative when f is a regular distribu- 2x x < 3
= 18δ(x − 3) + ,
tion f (x) that happens to be differentiable
R 0 in the or- 3x2 x ≥ 3
0
dinary sense. In that case, f {φ} = f (x)φ(x)dx =
− f (x)φ0 (x)dx = f {−φ0 }, where we have integrated
R where of course by f 0 (x) I mean the distribution
by parts and used the fact that φ(x) is zero outside a f 0 {φ}. It is common to be “casual” with notation in
finite region to eliminate the boundary terms. This this way for distributions, treating them like ordinary
is such a nice result that we will use it to define the functions, but you have to remember that you can’t
derivative of any distribution: evaluate them at any point x, you can only evaluate
them for test functions φ(x).
• The distributional derivative f 0 of f {φ}
is given by the distribution f 0 {φ} = f {−φ0 } , 2.4 Isolated points
where φ0 (x) is the ordinary derivative of φ(x). With ordinary functions, we had to make lots of
caveats about isolated points. No more with distri-
(This is sometimes also called a weak derivative.)
butions. The key point is that two different ordinary
Since the test functions φ(x) were required to be in-
functions can define the same distribution. Consider,
finitely differentiable, we have a remarkable conse-
for example, the function
quence: every distribution is infinitely differentiable (
(in the distributional sense). 1 x=0
For example, since δ{φ} = φ(0), it immediately f (x) = .
0 otherwise
follows that the derivative of a delta function is the
distribution δ 0 {φ} = δ{−φ0 } = −φ0 (0). This is not a delta function: it is finite at x = 0, and
The most important consequence of this definition is a perfectly acceptable function. It also defines a
is that even discontinuous functions are differ- regular distribution:
entiable as distributions, and their derivatives give Z ∞
delta functions for each discontinuity. Consider the f {φ} = f (x)φ(x)dx = 0.
−∞
regular distribution S defined by the step function
6 I’m being a bit glib here. How do we know that the product
5 We have used the fact that φ(x) is required to be continu- rule works the same? Below, we will rigorously define what it
ous, from which one can show that nothing weird can happen means to multiply a distribution by a smooth function like
with the average of φ(x) as ∆x → 0. x3 − x2 , and the ordinary product rule will follow.

4
Think about it: no matter what φ(x) is, the integral switch to interpreting them as distributions when we
must give zero because it is only nonzero (by a fi- run into difficulty (e.g. derivatives of discontinuities,
nite amount) at a single point, with zero area (“zero etcetera). Since the rules for distribution operations
measure” for the pure-math folks). Thus, in the dis- are all defined to be consistent with those for func-
tribution sense, we can say perfectly rigorously that tions in the case of regular distributions, this doesn’t
usually cause any trouble.
f =0 However, it is good to define some of the important
even though f (x) 6= 0 in the ordinary-function sense! operations on distributions precisely. All you have to
In general, any two ordinary functions that only do is to explain what the operation doesR to test func-
differ (by finite amounts—not delta functions!) at tions, usually defined by analogy with f (x)φ(x)dx
isolated points (a “set of measure zero”) define the for regular distributions. Here are a few of the most
same regular distribution. We no longer have to basic operations:
make caveats about isolated points—finite values at • differentiation: f 0 {φ} = f {−φ}
isolated points make no difference to a distribution.
For example, there are no more caveats about the • addition: (f1 + f2 ){φ} = f1 {φ} + f2 {φ}
Fourier series or Fourier transforms: they converge,
period, for distributions.7 • multiplication by smooth functions (including
Also, there is no more quibbling about the value of constants) s(x): [s(x) · f ]{φ} = f {s(x)φ(x)}
things like S(x) right at the point of discontinuities.
• product rule for multiplication by smooth func-
It doesn’t matter, for distributions. Nor is there quib-
tions s(x): [s·f ]0 {φ} = [s·f ]{−φ0 } = f {−sφ0 } =
bling about the derivative of things like |x| right at
f {s0 φ − (sφ)0 } = f {s0 φ} + f 0 {sφ} = [s0 · f ]{φ} +
the point of the slope discontinuity. It is an easy mat-
[s · f 0 ]{φ} = [s0 · f + s · f 0 ]{φ}.
ter to show that the distributional derivative of |x| is
simply 2S(x) − 1, i.e. it is the regular distribution • translation: [f (x − y)]{φ(x)} = f {φ(x + y)}
corresponding to the function that is +1 for x > 0
and −1 for x < 0 (with the value at x = 0 being 1
• scaling [f (αx)]{φ(x)} = α f {φ(x/α)}
rigorously irrelevant).
If you are not sure where these rules come from, just
2.5 Interchanging R plugging them into a regular distribution f {φ} =
try
f (x)φ(x)dx, and you’ll see that they work out in
limits and derivatives the ordinary way.
With a distribution, limits and (distributional)
derivatives can always be interchanged. This is
tremendously useful when talking about PDEs and 4 Problems with distributions
convergence of approximations. In the distribution
Unfortunately, distributions are not a free lunch; they
sense, the Fourier series can always be differentiated
come with their own headaches. There are two ma-
term-by-term, for example.
jor difficulties, one of which is surmountable and the
This is easy to prove. Suppose that the distribu-
other is not:
tions fn → f as n → ∞. That is, for any φ(x),
fn {φ} → f {φ}. Since this is true for any φ(x), • Boundary conditions: since distributions do not
it must be true for −φ0 (x), and hence fn0 {φ} = have values at individual points, it is not so easy
fn {−φ0 } → f {−φ0 } = f 0 {φ} as n → ∞. Q.E.D. to impose boundary conditions on the solutions
if they are viewed as distributions—what does it
mean to set u(0) = 0? There are ways around
3 Rules for distributions this, but they are a bit cumbersome, especially
For the most part, in 18.303, we will cheat a bit. in more than one dimension.
We will treat things as ordinary functions when- • Multiplication: it is not generally meaningful
ever we can, using the ordinary operations, and only to multiply distributions. The simplest exam-
7 Technically, we have to choose distributions with the right ple is the delta function: what would δ(x)2 be?
set of test functions. The right test functions for the general δ∆x (x)2 is okay, but its limit as ∆x → 0 does
Fourier transform on the real line are those for which xn φ(x) →
0 as x → ±∞ for any n > 0, i.e. φ(x) vanishes faster than any
not exist even as a distribution (the amplitude
polynomial. The resulting distributions are called tempered goes as 1/∆x2 while the integral goes as ∆x, so
distributions, and are the domain of the Fourier transform. it diverges).

5
For linear PDEs, lack of multiplication is not such is true if
a big problem, but it does mean that we need to ÂG(x, x0 ) = δ(x − x0 )
be careful about Hilbert spaces: if we think of the
and the integrals are re-interpreted as evaluating a
solutions u(x) as distributions, we have a problem
distribution.
because hu, ui may not be defined—the set of distri-
What does this equation mean? For any x 6= x0
butions does not form a Hilbert space. (Technically,
[or for any φ(x) with φ(x0 ) = 0] we must have
we can make something called a rigged Hilbert space ∂2
that includes distributions, but I don’t want to go ÂG(x, x0 ) = 0 = − ∂x 0
2 G(x, x ), and this must mean

there.) that G(x, x ) is a straight line for x < x0 and x > x0 .

0

To satisfy, the boundary conditions, this straight line

must pass through zero at 0 and L, and hence G(x, x0 )
must look like αx for x < x0 and β(x − L) for x > x0
5 The weak form of a PDE for some constants α and β.
G(x, x0 ) had better be continuous at x = x0 , or
Suppose we have a linear PDE Âu = f . We want otherwise we would get a delta function from the
to allow f to be a delta function etcetera, but we first derivative—hence α = β(x0 − L)/x0 . The first
still want to talk about boundary conditions, Hilbert derivative ∂x ∂
G(x, x0 ) then gives α for x < x0 and β
spaces, and so on for u. There is a relatively simple 0
for x > x . What about the next derivative? Since
compromise for linear PDEs, called the weak form ∂ 0
∂x G(x, x ) is discontinuous, it doesn’t have an ordi-
of the PDE or the weak solution. This concept nary second derivative at x = x0 , but as a distribu-
can roughly be described as requiring Âu = f only ∂2 0
tion it is no problem: ∂x 2 G(x, x ) is zero everywhere
in the “weak” sense of distributions (i.e., integrated (the derivative of a constant) plus a delta function
against test functions, taking distributional deriva- δ(x − x0 ) multiplied by β − α, the size of the jump.
tives in the case of a u with discontinuities), but re- Thus, − ∂x ∂2 0 0
2 G(x, x ) = ÂG = (α − β)δ(x − x ), and
quiring u to be in a more restrictive class, e.g. regu- from above we must have α − β = 1. Combined with
lar distributions corresponding to functions satisfying the equation for α from continuity of G, we obtain
the boundary conditions in the ordinary sense but β = −x0 /L and α = 1 − x0 /L, exactly the same as
having some continuity and differentiability so that our result from class (which we got by a more labo-
hu, ui and hu, Âui are finite (i.e., living in an appropri- rious method).
ate Sobolev space). Even this definition gets rather
technical very quickly, especially if you want to allow
delta functions for f (in which case u can blow up at Further reading
the point of the delta for dimensions > 1). Nailing
down precisely what spaces of functions and opera- • I. M. Gel’fand and G. E. Shilov, Generalized
tors one is dealing with is where a lot of technical Functions, Volume I: Properties and Operations
difficulty and obscurity arises in functional analysis. (New York: Academic Press, 1964). [Out of
However, for practical engineering and science appli- print, but still my favorite book on the subject.]
cations, we can get away with being a little careless in
precisely how we define the function spaces because • Robert S. Strichartz, A Guide to Distribu-
the weird counterexamples are usually obviously un- tion Theory and Fourier Transforms (Singapore:
physical. The key insight of distributions is that what World Scientific, 1994).
we care about is weak equality, not pointwise equality, • Jesper Lutzen, The Prehistory of the Theory of
and correspondingly we only need weak derivatives Distributions (New York: Springer, 1982). [A
(integration by parts, or technically “bilinear forms”). fascinating book describing the painful historical
process that led up to distribution theory.]

6 Green’s functions
Now that we have distributions, Green’s functions are
much easier to work with. Consider, for example, the
∂2
Green’s function G(x, x0 ) of  = − ∂x 2 on [0, L] with

Dirichlet (zero) boundaryR conditions. If Âu = f is

to be solved by u(x) = G(x, x0 )f (x0 )dx0 , then we
must have Âu = [ÂG(x, x0 )]f (x0 )dx0 = f (x), which
R

6
Distributions in a nutshell
• Delta functions are okay. You can employ their
informal description without guilt because there
is a rigorous definition to fall back on in case of
doubt.
R∞
– An “integral” −∞ δ(x − x0 )φ(x)dx just
0
Rmeans φ(x0 ). Integrals over0
finite domains
Ω
δ(x − x )φ(x) give φ(x ) if x0 is in the
0
interior of Ω and 0 if x is outside Ω, but
are undefined (or at least, more care is re-
quired) if x0 is on the boundary dΩ.
• When in doubt about how to compute f 0 (x),
integrate Rby parts againstR a test function to
see what f 0 (x)φ(x) = − f (x)φ0 (x) does (the
“weak” or “distributional” derivative).
– A derivative of a discontinuity at x0 gives
δ(x−x0 ) multiplied by the size of the discon-
tinuity [the difference f (x0+ )−f (x0− )], plus
the ordinary derivative everywhere else.
– This also applies
√ to differentiating func-
tions like 1/ x that have finite integrals
but whose classical derivatives have diver-
gent integrals—applying the weak deriva-
tive instead produces a well-defined distri-
bution. [For example, this procedure fa-
mously yields ∇2 1r = −4πδ(~x) in 3d. It also
can lead to “Hadamard regularizations,” the
famous Cauchy principal value, and other
regularizations of apparently divergent in-
tegrals that are often presented in a myste-
rious fashion: a unique regularization arises
because the integrand came from a distribu-
tional derivative or limit.]
• All that matters in the distribution (weak) sense
is the integral of a function times a smooth, lo-
calized test function φ(x).
R Anything that doesn’t
change such integrals f (x)φ(x), like finite val-
ues of f (x) at isolated points, doesn’t matter.
(That is, whenever we use “=” for functions we
are almost always talking about weak equality.)
• Interchanging limits and derivatives is okay in
the distribution sense. Differentiating Fourier
series (and other expansions in infinite bases)
term-by-term is okay.
• In practice, we only ever need to solve PDEs in
the distribution sense (a “weak” solution): inte-
grating the left- and right-hand sides against any
test functions must give the same number, with
all derivatives taken in the weak sense.