Multicalculus Test # 4

Dave L. Renfro
Louisiana School for Math, Science, and the Arts
DUE: Friday May 7, 1999

YOU MAY USE ANY NOTES, BOOKS, ETC. OF YOUR CHOOSING, BUT YOU ARE NOT ALLOWED
TO WORK TOGETHER OR OBTAIN HELP FROM ANY OTHER PERSON ON THIS. SEE ME IF YOU
HAVE QUESTIONS. THE STARRED (*) PARTS ARE OPTIONAL. (THIS INCLUDES PROBLEM 7). THEIR
SIGNIFICANCE TO YOU WILL BE EXPLAINED IN CLASS. FINALLY, YOU ONLY HAVE TO DO ONE OF
PROBLEMS 5 AND 6. (THE ONE THAT YOU DON’T DO WILL BE ANOTHER STARRED PROBLEM.)

1. [Functions
functions:
of Type xp sin (x q
)] Parts (c) and (d) involve the following
xn sin x 1
if x 6= 0
fn (x) =
0 if x=0

(a) Show that y = x2 sin x 3 has in…nite length for 0 < x 1.1 Include a labeled graph of
this function with your solution.

(b) Show that y = x3 sin x 2 has …nite length for 0 < x 1. Include a labeled graph of this
function with your solution.

(c) Show that f3 (x) is continuous at x = 0, f30 (0) exists, f30 (x) is continuous at x = 0, and
f300 (0) does not exist.2 Include a labeled graph of this function with your solution.

(d) Show that f4 (x) is continuous at x = 0, f40 (0) exists, f40 (x) is continuous at x = 0, f400 (0)
exists, and f400 (x) is not continuous at x = 0. Include a labeled graph of this function with
your solution.

(1*) Instead of (a)–(d) above, turn in all of Homework # 22.

2. [The Two–Path Test for Discontinuity] Recall the two–path test

for discontinuity (p. 979): If a function f (x; y) has di¤ erent limits along two di¤ erent
paths as (x; y) ! (a; b), then lim f (x; y) does not exist. I should also add the
(x; y)!(a; b)
(obvious) observation that if the function f (x; y) does not have a limit along some path
as (x; y) ! (a; b), then lim f (x; y) does not exist.
(x; y)!(a; b)

(a) Work exercises # 31–38 in Section 13.2 (p. 982).

(b) Show that f (x; y) = y x has the same limit along any 1’st quadrant polynomial approach
towards (0; 0) [i.e. along any path having the form y = axn , where x > 0 and n is a positive
a
integer], but the limit varies if approach paths of the form y = exp x are used, where
3
a > 0. As part of your solution, include carefully drawn graphs, for 0 < x < 0: 5, of
0: 5 1 2
y = x2 , y = x3 , y = x4 , y = exp x , y = exp x , and y = exp x together on the
same coordinate axis. You may sacri…ce some accuracy in order to give a more informative
and schematic presentation of the behavior of these graphs as x ! 0+ .
1 Parts (a) and (b) can be worked the same way as the two examples I’ve already shown you: (i) y = x sin x 1

has in…nite length for 0 < x 1, and (ii) y = x2

sin x 1has …nite length for 0 < x 1.
2 Recall that in class I’ve shown: (i) f (x) is continuous at x = 0 and f 0 (0) does not exist, and (ii) f2 (x) is
1 1
continuous at x = 0, f20 (0) exists, and f20 (x) is not continuous at x = 0.
3 Why do I need a > 0?
 1
exp( ax )
(2*) Verify analytically that if a > 0 and n is any positive integer, then lim+
x!0
xn = 0.
[You will …nd that a direct application of L’Hôpital’s rule doesn’t help. However, this
limit can be evaluated by using the changing variables method given (somewhere) in my
handwritten handout titled “Some Techniques for Evaluating Limits”. (But feel free to use
another (mathematically justi…ed) method if you …nd one.)]

3. [Partial Derivatives and the Gradient]

(a) Work problems # 33, 34, 48, and 49 in SECTION 13.3 (pp. 991–992).

(b) Work problems # 6–11 in SECTION 13.4 (p. 999).

(c) Work problems # 1, 4, 9, 12, 18, 19, and 21 in SECTION 13.5 (p.1007). [Note: The answer
to # 19 is wrong (at least, in some printings of your text.]

(d) Work problems # 1, 4, 9, and 12 in SECTION 13.6 (p. 1012).

(e) Work problem # 34 in SECTION 13.5 (p. 1009). In addition, show that
! !
! f g rf rg f
r =
g g2

! !
(3*-a) Let !
r = x i + y j be the radial vector from the origin. Then the magnitude of !
r,
1
given by r = x2 + y 2 2 , is a real–valued function of two real variables. Verify (by explicit
partial di¤erentiation with respect to x and y) that
! n !
r (r ) = nrn 2
r
!
r
holds for any real number n. Note that if r^ = r is the unit vector associated with r, then
! n
r (r ) = nrn 1^ r.
! !
(3*-b) Let r2 = r r = @x @2 @2
2 + @y 2 be the Laplace operator in TWO dimensions. Verify (by
explicit partial di¤erentiation with respect to x and y) that

r2 (rn ) = n2 rn 2

1
for any real number n, where r = x2 + y 2 2
.
! !
(3*-c) Let r2 = r r = @x @2 @2 @2
2 + @y 2 + @z 2 be the Laplace operator in THREE dimensions.
Verify (by explicit partial di¤erentiation with respect to x and y) that

r2 (rn ) = n (n + 1) rn 2

1
for any real number n, where r = x2 + y 2 + z 2 2 . Parts (c) and (d) show that expressions
for r2 for radial functions (functions depending only on r) can depend on the number of
2
independent variables involved. Indeed, if f = f (r), then r2 f = ddrf2 + 1r drdf
in TWO
2
dimensions, while r2 f = ddrf2 + 2r dr
df
in THREE dimensions.4 You should be able to verify
these more general observations, but I’m not speci…cally asking that you do so for this test.
[This more general result for TWO dimensions is immediate from (e), by the way.]
4 This distinction caused your teacher at least 90 minutes of hair–pulling anguish in trying to …gure out why the

result of the computation in part (c) was di¤erent in some of the books I was looking at!
(3*-d) Let z = F (x; y), where x = r cos and y = r sin . Then from the chain rule (speci…cally,
the version given on the middle of p. 997) we have

@z @z
= Fx cos + Fy sin and = Fx (r sin ) + Fy (r cos ) :
@r @
[Hey, if r = 1 (e.g. on the unit circle), don’t these equations remind you of the rotation transformation
equations? Yes, there is a connection. As you progress through multicalculus in college (or, in the case of
one of you, LSMSA’s advanced calculus), you might want to check out a book from the library on vector
and tensor analysis and see if you can …gure out what’s REALLY going on.] Now let z = G (r; ) be
the function that gives z in polar coordinates. Show that
1 1
Grr + Gr + 2 G = Fxx + Fyy :
r r
In other words, you’re going to show that the Laplace operator in polar coordinates is given
by
@2 1 @ 1 @2
r2 = 2
+ + 2 2:
@r r @r r @
HINT: Start with the formulas for @z @z
@r and @ and take further derivatives with respect to
r and with respect to so as to obtain expressions for Grr and G . (You’ll have to use the
product rule along with the chain rule.5 ) Then substitute the expressions you have for Grr ,
Gr , and G into the left–hand side of the relation you are to prove and simplify.

(3*-e) Let n be a positive integer. Using the formula for r2 in polar coordinates (see (e) above),
show that each of the following two functions is harmonic:

(i) rn cos (n ) and (ii) rn sin (n ) :

Note these are the real and imaginary parts (in polar coordinates) of the complex–valued
n
function of one complex variable given by f (z) = z n , since z n = rei = rn ei(n ) =
n
r [cos (n ) + i sin (n )]. Therefore, it follows that these functions are harmonic from what
I’ve told you in some recent handouts. Nonetheless, for this problem I want you to verify
this by explicitly evaluating r2 for these functions.

4. [Multi– Variable Taylor Expansion]: If f is a real–valued function of

the two real variables x and y that is su¢ ciently well behaved in a neighborhood of (x; y) =
(a; b),6 then

f (x; y) = f (a; b)
+ [fx (a;
h b) (x a) + fy (a; b) (y a)] i
1 2 2
+ 2! fxx (a; b) (x a) + 2fxy (a; b) (x a) (y b) + fyy (a; b) (y b)
+

This corresponds to the one variable expansion

1 2
f (x) = f (a) + f 0 (a) (x a) + f 00 (a) (x a) +
2!
5 For @ @ @
instance, @r
(Fx cos ) = [using the product rule] @r
(Fx ) cos + Fx @r
(cos ) = [using the chain rule]

@Fx @x @Fx @y
+ cos + Fx 0 = Fxx cos2 + Fxy cos sin ;
@x @r @y @r
@
where @r (cos ) = 0 is used.
6 The precise hypothesis is that f is analytic at (a; b), the same hypothesis that ensures the validity of the one

variable Taylor expansion. We didn’t discuss this matter in any detail but, for what it’s worth, this hypothesis is
a bit stronger than being in…nitely di¤erentiable at every point in some neighborhood of (a; b).
 By replacing (x a) with x [hence, x is replaced with a + x], we get the equivalent
expansion
1 2
f (a + x) = f (a) + f 0 (a) x + f 00 (a) ( x) +
2!
The two variable version of this other one variable expansion is

f (a + x; b + y) = f (a; b)
+ [fx (a;
h b) x + fy (a; b) y] i
1 2 2
+ 2! fxx (a; b) ( x) + 2fxy (a; b) ( x) ( y) + fyy (a; b) ( y)
+

Here is a neat way to describe this two variable expansion. Let

! x ! a ! x
X = ; X0 = ; and X = :
y b y

Then
h ! ! i 1 ! ! 2
f = f + r X (f ) ! + r X (f )
!
X0 +
!
X
!
X0 X0 2! !
X0

1 ! ! 3
+ r X (f ) + ,
3! !
X0

! ! !
where the subscripts refer to evaluations and r X represents the dot product of r
! ! !
with X . The various powers of r X are to be expanded “algebraically”, meaning
@ @2
(for example) that the square of @x represents @x 2 , and so on in the obvious manner. [I

say “obvious”, since you already have the expansion written out in terminology you should
understand. Just make the obvious identi…cations needed for the expansions to be equal!]
!
Note how this last expansion looks just like the one variable expansion, with r replacing
ordinary di¤erentiation, dot product replacing ordinary multiplication, etc. The neat thing
about writing the expansion this way is that if you decide to make everything three variables
(or more), the form of the expansion remains the same!

(a) Let f be a real–valued function of three variables x, y, and z, and put

0 1 0 1 0 1
x a x
! @ ! ! @
X = y A ; X0 = @ b A ; and X = y A:
z c z

Express the following in “expanded component form” (i.e. like the terms in the version of
the two variable Taylor series at the top of this page):
h ! ! i
r X (f ) ! =
X0
?
! ! 2
r X (f )
!
= ?
X0

! ! 3
r X (f )
!
= ?
X0

For the last one, you will have to cube a trinomial. Here is a neat way to obtain the pattern.
 3
I found the following interesting technique for evaluating (A + B + C) on page 74 of Webster
Wells, Advanced Course in Algebra, D. C. Heath & Co., 1904. Clearly, all the expanded
terms will be of the third degree. These terms fall into three classes: terms that are the cube of a
variable, terms that involve the square of a variable, and a term involving ABC. Clearly, the term
involving A3 has coe¢ cient 1. Therefore by symmetry, this is also true for the terms involving
B 3 and C 3 . Moreover, it is evident that the term involving A2 B has coe¢ cient 3. (There are
three ways to choose exactly one B when three choices with replacement are made from the set
fA; B; Cg.) Therefore, by symmetry, this is also true for the terms AB 2 , AC 2 , A2 C, etc. Hence,
3
(A + B + C) = A3 + B 3 + C 3 + 3 A2 B + A2 C + AB 2 + B 2 C + AC 2 + BC 2 + k ABC

for some constant k. We can …nd k by substituting in A = B = C = 1 and solving for k. This
gives k = 6, and so we have
3
(A + B + C) = A3 + B 3 + C 3 + 3 A2 B + A2 C + AB 2 + B 2 C + AC 2 + BC 2 + 6ABC:

(b) Use the 2’nd order Taylor polynomial for functions of 3 variables to …nd the quadratic
approximation to
0 1 0 1
a 0
f (x; y; z) = ex sin y cos z about @ b A = @ 0 A :
c 0

(c) Use sin y cos z = 21 sin (y z) + 12 sin (y + z) [see the appropriate trig. formula on page 5
of your text’s trig. appendix (or see equation (2) on p. 632)] and substitution into the one
variable Taylor series for the SINE function to obtain an expansion for sin y cos z (up to
and including third degree terms is enough). Then multiply this expansion by the quadratic
expansion for ex , namely 1 + x + 12 x2 , keeping only terms of degree 2 or less. Show that the
answer you get is the same as what you got in part (b).

5. [Di¤erentiating Under the Integral Sign]:

R
Sometimes it is easier to
1 10
solve a harder problem! An example is trying to evaluate 0 x ln x 1 dx. We will …nd (see
R1
below) that the more general integral 0 x ln x 1 dx is “easier”to …nd. The following is from
pp. 86–87 of Richard Feynman, Surely You’re Joking, Mr. Feynman, W. W. Norton
& Company, 1985:

So every physics class, I paid no attention to what was going on with Pascal’s Law, or whatever they
were doing. I was up in the back with this book: “Advanced Calculus”, by Woods. Bader [Feynman’s
High School Physics teacher, who loaned Feynman his copy of Wood’s book] knew I had studied
“Calculus for the Practical Man” a little bit, so he gave me the real works— it was for a junior or
senior course in college. It had Fourier series, Bessel functions, determinants, elliptic functions— all
kinds of wonderful stu¤ that I didn’t know anything about.
That book also showed how to di¤erentiate parameters under the integral sign— it’s a certain opera-
tion. It turns out that [it’s] not taught very much in the universities; they don’t emphasize it. But I
caught on how to use that method, and I used that one [**] tool again and again. So because I was
self–taught using that book, I had peculiar methods of doing integrals.
The result was, when guys at MIT or Princeton had trouble doing a certain integral, it was because
they couldn’t do it with the standard methods they had learned in school. If it was contour integration,
they would have found it; if it was a simple series expansion, they would have found it. Then I come
along and try di¤erentiating under the integral sign, and often it worked. So I got a great reputation
for doing integrals, only because my box of tools was di¤erent from everybody else’s, and they had
tried all their tools on it before giving the problem to me.
Here
R 1 x is 1an example of the method Feynman is referring to. Let’s say we want to evaluate
0 ln x dx for some …xed 0. The problem is that none of the usual integration techniques
will allow you to …nd the antiderivative of the integrand. In fact, I believe the antiderivative of
this particular integrand cannot be expressed
R1 in terms of the so called elementary functions. So
what are we to do? Letting F ( ) = 0 x ln x 1 dx, we have7
Z 1 Z 1 Z 1
0 d x 1 d x 1 x ln x
F ( ) = dx = dx = dx
d 0 ln x 0 d ln x 0 ln x
Z 1 1
1 +1 1
= x dx = x = :
0 +1 0 +1

Integrating with respect to gives F ( ) = ln ( + 1) + C. We …nd C by using a value of for

R1 0 R1
which we know what F ( ) is. For instance, F (0) = 0 xln x 1 dx = 0 (0) dx = 0. By
plugging = 0 into F ( ) = ln ( + 1) + C, we get C = 0. Therefore,
Z 1
x 1
dx = ln ( + 1) :
0 ln x

In particular, the integral mentioned earlier has the value ln (11). [You might want to numerically
check this on your calculator.]

Use the method described above, or another method of your choice, to …nd an exact analytic
expression in terms of the elementary functions for
Z 2" #
ln 1 + 12 cos x
dx :
0 cos x
Z 2 h i
HINT: Let F ( ) = ln(1 + cos cos x)
cos x dx . After we …nd an expression for F ( ), we
0
simply plug in = cos 1 21 . First, …nd an integral expression for F 0 ( ) by di¤erentiating under
the integral sign. After a couple of steps, you’ll …nd that you can factor ( sin ) out from the
integral sign (a constant, as far as integration with respect to x is concerned). To evaluate the
integral you have left, observe that it corresponds to integral # 74 at the back of your text (the
table of integrals) when a = 1, b = 1, and c = cos . (Note the condition
p b2 > c2 is satis…ed for
all 0 < < 2 .) Since we’re only interested in …rst quadrant angles, 1 cos2 = jsin j winds
up being just sin . Also, you might want to make use of the trig. identity tan2 (A) = 11 + cos(2A)
cos(2A)

at some point. (I’ll leave it to you as to what you want to use for “A”.) O–K, now you should
be ready to plug in the limits. If you’ve been really careful, you’ve noticed that we’re actually
dealing with an IMPROPER integral! (Because the denominator of the integrand equals zero
when x = 2 .) So we actually have to take a limit when plugging in the limits of integration. But
you should have an antiderivative that’s continuous at x = 2 , so the limit of the antiderivative as
x ! 2 (Why is the limit from the left of 2 ?) is just what you get when you plug x = 2 into the
antiderivative.8 After plugging in the limits and making what should be obvious simpli…cations
(one of which is that tan 4 = 1), you’ll have a super–simple expression for F 0 ( ). Integrate this
expression with respect to to get F ( ) (Don’t forget the + C!) and then …nd C by observing
that F 2 = 0. Make sure you show some steps to convince me that you know why F 2 = 0.
For instance, F 2 is NOT an integral whose limits are 2 to 2 ! Finally, the integral you want
is F cos 1 12 . Since this is an improper integral, you might …nd that your calculator takes a
long time to give a reasonable numerical approximation to it.
7 The rigorous justi…cation for why these limiting operations can be interchanged in this particular case involves

careful analysis beyond the scope of our course. Let me just say that you can’t always do this!
8 The example I worked earlier was also an improper integral, and the same explaination I give here justi…es

what I did there.

6. [Fourier Series]: The Fourier expansion of f (x) on the interval ( ; ) is
1
X
f (x) A0 + [ An cos (nx) + Bn sin (nx) ] ;
n=1

where
Z
1
A0 = f (x) dx
2
Z
1
An = f (x) cos (nx) dx
Z
1
Bn = f (x) sin (nx) dx :

The sense in which the sign can be replaced by an = sign depends on results that need
not concern us now. It will be su¢ cient for our purposes that if f (x) has a continuous
derivative on ( ; ), then this series converges “in a nice way”to f (x) for each x 2 ( ; ).
The formulas I gave for An and Bn are surprisingly easy to obtain, IF we are allowed to
ASSUME the existence of such an expansion and IF we are allowed to ASSUME that
the series converges “in a nice way” (e.g. uniform convergence, for instance).9 To …nd a
formula for the Bn ’s, multiply both sides of the Fourier series expansion for f (x) by sin (mx),
integrate both sides from to , and then evaluate some de…nite integrals:
1
X
f (x) sin (mx) = A0 sin (mx) + [ An cos (nx) sin (mx) + Bn sin (nx) sin (mx) ]
n=1

Z Z
f (x) sin (mx) dx = A0 sin (mx) dx
1 Z
X Z
+ An cos (nx) sin (mx) dx + Bn sin (nx) sin (mx) dx
n=1

At this point we observe that

Z Z
0 if m 6= n
An cos (nx) sin (mx) = 0 and Bn sin (nx) sin (mx) =
Bm if m=n

(This follows easily by using identities such as sin A cos B = 21 [sin (A + B) + sin (A B)].),
and so we get everything on the right–hand side equal to zero except for the integral involving
Bn when n takes on the value m:
Z Z
f (x) sin (mx) dx = Bm sin (mx) sin (mx) dx = Bm :

This gives the formula for Bn I gave when you make a “m ! n dummy variable substitution”.
To …nd An , do the same thing except start by multiplying both sides by cos (mx). Fourier
expansions are unique:
f (x) = A0 + A1 cos x + B1 sin x + A2 cos 2x + B2 sin 2x +
f ( x) = A0 + A1 cos ( x) + B1 sin ( x) + A2 cos ( 2x) + B2 sin ( 2x) +
= A0 + A1 cos x B1 sin x + A2 cos 2x B2 sin 2x +

If f (x) = f ( x), then from the equality of the COSINE coe¢ cients we get A0 = A0 ,
A1 = A1 , A2 = A2 , etc. [i.e. nothing new] and from the equality of the SINE coe¢ cients
9 This latter hypothesis allows us to integrate term–by–term, among other things.
 we get B1 = B1 , B2 = B2 , etc. [i.e. all the Bn ’s have to be zero]. In the same
way, uniqueness of Fourier expansions implies that the Fourier expansion of an odd function
contains only sine functions. By the way, note that this equality of coe¢ cients business is
the “lead player” in this proof, and not any speci…c fact about sines and cosines other than
sines are odd functions and cosines are even functions. Therefore, the same proof works
for Taylor expansions about x = 0. Hence, the Taylor expansion about x = 0 of an even
function should include only even powers of x [look at the expansions of cos x, sin x2 , etc.]
and the Taylor expansion about x = 0 of an odd function should include only odd powers
of x [look at the expansions of sin x, tan x, etc.]. Let’s consider “even” and “odd” as types
of symmetry that a function can have. Our conclusion is that if a function satis…es one of
these types of symmetry, then we know something nontrivial about its Fourier expansion
and about its Taylor expansion (about x = 0). This idea, that symmetry principles can tell
you something nontrivial about an expansion, is one of those important things that keep
recurring in mathematics, physics, chemistry, engineering, etc. For example, the s, p, d, etc.
orbitals that you study in chemistry are the functions that an arbitrary (bound) state of the
hydrogen atom can be expanded in terms of.10 (Chemists also use them to expand the wave
function of a multi–electron atom, but in this case the expansion is only an approximation
of the true wave function, even if you sum the entire resulting in…nite series of functions in
such an expansion.) Often, if a speci…c hydrogen wave function is known to have some type
of symmetry (e.g. relative to a point, a line, or a plane), then you can use this knowledge
to obtain something nontrivial about the expansion of its wave function in a series of the
hydrogen atom eigenfunctions.11 This can save an enormous amount of computation in some
situations.
(a) Find the Fourier expansion for f (x) = x2 on the interval ( ; ). Compute enough of the
coe¢ cients An ’s and Bn ’s so that it is clear what the general pattern is.

(b) Let Fn (x) be the Fourier expansion of f (x) = x2 up to and including the cos (nx) and
sin (nx) terms. Give carefully drawn and labeled graphs of y = x2 , y = F2 (x), and y = F6 (x)
on the same coordinate system, for 10 x 10. Use three di¤erent colored pens for these
graphs, or use highlight markers to highlight your graphs with three di¤erent colors. (You
can come by my o¢ ce to use my highlight markers for this if you want to.) Therefore, you
will draw ONE carefully labeled coordinate system, and on this coordinate system I
want to see three graphs in three di¤erent colors. Of course, use a calculator to get these
graphs.

(c) Plug x = 0 into the Fourier expansion you found in (a) and use what you get to show that
2
1 1 1 n+1 1
= 1 + + + ( 1) +
12 4 9 16 n2
1 0 By the way, Schrödinger’s equation involves the Laplace operator r2 . Schrödinger’s equation for the hydrogen

atom is (typically) solved by …rst recasting everything into spherical coordinates, which means that you need to
know what r2 is in spherical coordinates. [This is odd ... I can’t …nd any discussion of r2 in other coordinate
systems in your text! This is a serious de…ciency in my opinion. (In fact, the index doesn’t even have a listing for
“Laplace” or “Laplacian”.)]
1 1 For those who took linear algebra last semester: The linear operator that arises from the Schrödinger

equation for the hydrogen atom (see below) and the eigenfunctions for this linear operator are just the eigenvectors
for this linear operator in the sense that you studied in linear algebra. The corresponding eigenvalues are the
energies for those particular “eigenstates” of the hydrogen function. Here is the Schrödinger equation for the wave
function = (x; y; z) of a hydrogen atom, where h is Planck’ ps constant, e is the electron charge, m is the mass
of the electron, E is the total energy of the electron, and r = x2 + y 2 + z 2 :
8 2m e2
r2 + E+ = 0:
h r
2
h e
If you de…ne L by L = 8 2m
r2 r
, then it is trivial to verify that L ( 1 + 2 ) = L ( 1 ) + L ( 2 ) and
L (c ) = cL ( ) (i.e. L is a linear operator). Note that Schrödinger’s equation is now simply L ( ) = E , which is
an eigenvalue problem.
7*. ample
[Integrals Containing Gaussian Functions]: First, read over Ex-
4 on pp. 1070–1071. There is a very well known “trick” for evaluating the de…nite
R1 x2
R1 x2
integral 0
e dx. Let I = 0
e dx. Then
Z 1 Z 1 Z 1 Z 1
2
2 x2 x2 y2
I = e x dx e dx = e dx e dy :
0 0 0 0

RThe
1
second equality
x2
R 1 follows
y2
from
R 1 thez2fact that
R 1x is 2a dummy variable. This means that
0
e dx = 0
e dy = 0
e dz = 0
e d = . (Indeed, each of these is
X4 X4 X 4
p
equal to 12 .) This is analogous to the fact that n2 = k2 = z2 = Next,
n=1 k=1 z=1
we have Z 1 Z 1 Z 1 Z 1
x2 y2 y2 x2
e dx e dy = e dy e dx:
0 0 0 0
R1 2 R1 y2
This follows from the fact that 0 e y dy is a constant and so we can put 0 e dy inside
the integration with respect to x. Next up is
Z 1 Z 1 Z 1Z 1
2 2 2 2
e y dy e x dx = e x e y dx dy;
0 0 0 0

which follows from Fubini’s theorem (an iterated integral equals a double integral).12 Finally,
we have Z 1Z 1 Z 2 Z 1
2 2 2
e (x +y ) dx dy = e r r dr d ;
0 0 =0 r=0
where I converted to polar coordinates. Note that the region of integration is the …rst
quadrant, which can be described in polar coordinates as the set of points (r; ) such that
0 2 and 0 r < 1. In making the change from Cartesian coordinates to polar
coordinates, dx dy changes to r dr d . [See equation (7) at the bottom of page 1069. A
general way of changing coordinates is discussed in Section 14.7, which is VERY important.]
R1 2
You should now be able to carry out this integration to get I 2 = 4 . Therefore, 0 e x dx =
1 p
2 . This is (essentially) integral #140 in your text’s table of integrals.13 For part (a),
note that 1 < x; y < 1 limits correspond to 0 < 2 and 0 r < 1.

(a*) Use the technique described above to show that

Z 1 r
I0 exp (a + bi) x2 dx =
1 a + bi

when a and b are real numbers with a > 0. [We need a > 0 so that this improper integral
actually converges.] You may evaluate Rthe various complex–valued function integrations
formally. That is, you may assume that erx dx = 1r erx + C for any complex constant r.
1 2 Yes,
I know Fubini’s theorem is only stated for integration over a bounded rectangle. O–K, so there are
R 2 R 2 R 2 R 2
extra limits via 01 e x dx = lim 0a e x dx and 01 e y dy = lim 0b e y dy. There are theorems in (college)
a!1 b!1
advanced calculus that tell you when you’re allowed to take limits under the integral sign:
Z 1 Z 1 Z a Z b Z a Z b
2 2 2 2
y2 x2
e y dy e x dx = lim lim e y dy e x dx = lim lim e dy e dx
0 0 a!1 0 b!1 0 a!1 b!1 0 0
Z aZ b Z 1Z 1
y2 x2 x2 y2
= lim lim e e dy dx = e e dx dy:
a!1 b!1 0 0 0 0
R1
The …rst equality is the de…nition of 0 , the second equality comes from interchanging one of the limits with
an integration [this can be done under certain conditions which need not concern us (an appropriate condition is
satis…ed in our present case)], the third equality is Fubini’
R sR theorem for the rectangle [0; a] [0; b], and the fourth
equality is the de…nition the improper double integral 01 01 (which I couldn’t …nd mentioned in your text prior
to exercise 12 on page 1071; in fact, I can’t …nd improper multiple integrals mentioned anywhere in your text).
1 3 The technique I just went through is outlined in problem # 28 on page 1101 of your text.
 R1
Although 1
exp (a + bi) x2 dx does not converge if a = 0 in the sense that our textbook’s
R1 2
de…nition requires (Section 8.6), it is useful in quantum mechanics to “de…ne” 1 e ibx dx as
R1 2 p p p p
lim+ 1 e (a + bi)x dx, which results in the value bi = b i = b e
4 i.
a!0

(b*) By combining the technique from (a) above with a variation of the technique explained in
R 1 n (a + bi)x2
problem 3, it is possible to evaluate 1
x e dx for any positive integer n when
a > 0. Actually, if n is odd, the integral is zero
R 1 2 (a + bi)x2 since the integrand
R 1 will be an odd function.
4 (a + bi)x2
I want you to evaluate I2 1
x e dx and I4 1
x e dx.

R1 R1
HINT: Put I0 = 1
e (a + bi)x2
dx . Then d
da (I0 ) = d
1 da
e ax2 ibx2
dx =
R1 2 2
d
1
x2 e ax ibx dx, which implies (I0 ) = I2 . Hence, all you have to do is di¤erenti-
da
ate I0 with respect to a, multiply by 1, and you’ll get I2 . I’ll leave it to you to …gure out how to
get I4 . [Make sure you give me an appropriate explanation.. Don’t just di¤erentiate something
and say this is the answer! Give some sort of justi…cation like I did with I2 .]