Ma 4121: Introduction To Lebesgue Integration Solutions To Homework Assignment 6
Ma 4121: Introduction To Lebesgue Integration Solutions To Homework Assignment 6
Please return your solutions to the instructor by the end of class on the due date. You may collaborate
on these problems but you must write up your own solutions. Late homework will not be accepted.
2. Suppose that S ⊂ R2 is a measurable set with the property that λ(Sy ) = 0 for almost every y ∈ R,
where λ is 1-dimensional Lebesgue measure on R, and
def
Sy = {x ∈ R : (x, y) ∈ S}.
Prove that the 2-dimensional Lebesgue measure of S is zero. (Note: This is a partial converse to
Theorem 15.5 on p.412 of our text.)
1
def
Put f (x, y) = 1S (x, y). Then f ∈ M (R2 ). It suffices to show that
R
Solution: R2
f = 0.
2
Suppose first that S is bounded. Then f ∈ L(R ) and we may use Fubini’s theorem to evaluate
Z Z Z
f= G(y) = 0, since G(y) = f (x, y) = λ(Sy ) = 0, a.e.y ∈ R.
R2 y∈R x∈R
S∞
Now write S = n=1 Sn , where Sn = S ∩ [−n, n]2 . Then for each n, Sn is bounded and satisfies the
same hypotheses as S, so it has 2-dimensional Lebesgue measure 0. Thus the 2-dimensional Lebesgue
measure of S is at most the sum of countably many 0s, hence is zero. 2
3. Suppose that fi : R → R is defined and bounded on the compact interval [ai , bi ] ⊂ R. If fi ∈ L([ai , bi ])
for i = 1, . . . , n, prove that
! !
Z Z Z b1 bn
f1 (x1 ) · · · fn (xn ) d(x1 , . . . , xn ) = f1 (x1 ) dx1 ··· fn (xn ) dxn ,
Q a1 an
Solution: We prove this by induction on n. It is evidently true when n = 1, for then both sides are
the same.
Suppose the result holds for n −1; let g : Rn−1 → R be defined by g(x1 , . . . , xn−1 ) = f (x1 ) · · · f (xn−1 ).
def
This g is defined and bounded on Qn−1 = [a1 , b1 ] × · · · × [an−1 , bn−1 ]. For any fn ∈ L([an , bn ]), the
function gfn = g(x1 , . . . , xn−1 )fn (xn ) belongs to L(Q) by Lebesgue’s dominated convergence theorem
and we have
! "Z #
Z Z bn
g(x1 , . . . , xn−1 )fn (xn ) d(x1 , . . . , xn ) = g(x1 , . . . , xn−1 ) d(x1 , . . . , xn−1 ) fn (xn ) dxn ,
Q Qn−1 an
by Fubini’s theorem. Here g(x1 , . . . , xn−1 ) may be removed from the integral in xn since it has no
xn -dependence, and the remaining function fn is Lebesgue integrable by hypothesis. The result for n
now follows from the inductive hypothesis. 2
2 2
4. (a) Prove that R2 e−x −y = π by transforming the integral to polar coordinates
R
2 √
(b) Use part(a) to prove that R e−x = π.
R
2
(c) Use part (b) to prove that Rn e−kxk = π n/2 .
R
2
(d) Evaluate R e−tx for t > 0, and find t for which the value is 1.
R
Solution: (a) R2 = {(x, y) = (r cos θ, r sin θ) : 0 < r < ∞, 0 ≤ θ < 2π} ∪ {(0, 0)}, and this mapping
has Jacobian
cos θ sin θ
J=
−r sin θ r cos θ
with determinant |J| = |r| = r. Hence, by eq.30 on p.429 of our text,
Z Z
−x2 −y 2 2
e = e−r |J|,
R2 R2
2
after substituting r2 ← u and r dr ← 21 du.
2
−y 2 2 2
(b) Since e−x = e−x e−y , by using the results of exercise 3 above we get
Z Z Z Z 2
2
−y 2 2 2 2
e−x = e−x e−y = e−x ,
R2 x∈R y∈R x∈R
since the two integrals are the same except for the name of the variable. Combine with part (a) and
2 √
the observation that the integrals must be positive to conclude that R e−x = π.
R
2 √ √
e−tx
R
Thus x∈R
= π/ t, and choosing t = π gives
Z
2
e−πx = 1.
x∈R
2
5. Let Vn (a) denote the volume of the ball of radius a in Rn , that is, the n-dimensional Lebesgue measure
of the open set {x ∈ Rn : kxk < a}.
(a) Prove that Vn (a) = an Vn (1).
(b) Prove that, for n ≥ 3, we have the formula
Z 2π Z 1
2π
Vn (1) = Vn−2 (1) × (1 − r2 )n/2−1 r dr dθ = Vn−2 (1) .
0 0 n
π n/2
Vn (1) = ,
Γ( 12 n
+ 1)
3
(b) Write
Now suppose that n > 2 and that the equation holds for all k = 1, 2, . . . , n − 1. Then by (b),
4
Solution: Use the Tonelli-Hobson test, Th.15.8 on p.415 of our text. Since Q has 1-dimensional
measure zero, the set Q × R has 2-dimensional measure zero by exercise 2 above, so f agrees almost
2 2
everywhere with the function g(x, y) = e−x −y .
Now g > 0 on R2 , so |g| = g, and the iterated integral exists:
√ −y2
Z Z Z
|g(x, y)| = πe = π,
y∈R x∈R y∈R
using the results of exercise 4 above. Conclude that g ∈ L(R2 ), so f ∈ L(R2 ), with
R R R
f= g= |g| =
π. 2
7. Let f (x, y) = (x2 − y 2 )/(x2 + y 2 )2 for 0 ≤ x ≤ 1, 0 < y ≤ 1, and put f (0, 0) = 0. Prove that both
iterated integrals
Z 1 Z 1 Z 1 Z 1
f (x, y) dx dy, and f (x, y) dy dx
y=0 x=0 x=0 y=0
x2 − y 2
∂ y ∂ −x
2 2
= = .
∂y x + y (x2 + y 2 )2 ∂x x + y2
2
while 1
1 1 1 1
x2 − y 2
Z
−x −1
Z Z Z
dx dy = dx = dy < 0,
y=0 x=0 (x2 + y 2 )2 y=0 x + y2
2
y=0 y=0 1 + y2
where both integrals exist because the integrands are continuous and bounded. 2
8. Let f (x, y) = e−xy sin x sin y for x ≥ 0 and y ≥ 0, and let f (x, y) = 0 otherwise. Prove that both
iterated integrals
Z Z Z Z
f (x, y) dx dy, and f (x, y) dy dx
y∈R x∈R x∈R y∈R
/ L(R2 ). Explain why this does not contradict the Tonelli-Hobson test
exist and are equal, but that f ∈
(theorem 15.8, p.415).
Solution: If one of the iterated integrals exists, then so will the other, and they will be equal, for
the function satisfies f (x, y) = f (y, x) for all (x, y) ∈ R2 .
In Gradshteyn and Ryzhik, Table of Integrals, Series, and Products, 5th edition, equation 3.893(1) on
p.512 implies Z ∞
1
e−xy sin y dy = , if x > 0,
y=0 1 + x2
so that Z ∞ Z ∞ Z ∞
sin x
e−xy sin x sin y dx dy = dx < ∞,
y=0 x=0 x=0 1 + x2
5
by Lebesgue’s dominated convergence theorem and comparison with 1/(1 + x2 ) ∈ L(R).
Now suppose toward contradiction that |f | ∈ L(R2 ). Let
∞
def 1 [
B = {t ∈ R+ : | sin t| ≥ √ } = Bk ,
2 k=0
def
where Bk = [(k + 41 )π, (k + 34 )π]. Then for (x, y) ∈ B × B ⊂ R2 , we have | sin x sin y| ≥ 21 . We define
a function s : R2 → R as follows:
min{|f (s, t)| : (s, t) ∈ Bm × Bn }, if (x, y) ∈ Bm × Bn ,
s(x, y) =
0, otherwise.
Then s ∈ M (R2 ) since it is a countable sum of step functions, and 0 ≤ s ≤ |f | on R2 . Thus |f | ∈ L(R2 )
would imply that s ∈ L(R2 ). But in fact
1 −(m+ 3 )(n+ 3 )π2
s(x, y) = e 4 4 , if (x, y) ∈ Bm × Bn ,
2
and |Bm × Bn | = π 2 /4 for every n, m = 0, 1, 2, . . ., so
∞ ∞ ∞ ∞
π 2 X X 1 −( 3 +m)( 3 +n)π2 π 2 X X −mnπ2
ZZ
s= e 4 4 ≥ e
R2 4 m=0 n=0 2 8 m=1 n=1
Finally, we may use the integral test to show that the double series diverges, for
ZZ Z ∞ Z ∞ Z ∞
−xyπ 2 1 −xy 1 1
e = 2 e dy dx = 2 dx = +∞.
x>0,y>0 π x=0 y=0 π x=0 x
Solution: From Gradshteyn and Ryzhik, Table of Integrals, Series, and Products, 5th edition, we
have Z
dx 1
3
= ; (2.111(1), p.68)
(y + x) −2(y + x)2
and Z
x dx y 1
3
= − x + ; (2.114(2), p.69)
(y + x) 2 (y + x)2
Thus for y > 0, we have
1 1
x + y2 y
Z 1
(x − y) 2 −1
3
dx = − 2
+ 2
= .
0 (x + y) (x + y) x=0 (x + y) x=0 (1 + y)2
Similary, by exchanging x ↔ y and noting that f (x, y) = −f (y, x) for all x, y, we get
Z 1
(x − y) 1
3
dy = , for x > 0.
0 (x + y) (1 + x)2
6
Both of these functions have elementary antiderivatives, with which we may compute
1 Z 1 1 1 Z 1 1
−1
Z Z
1 1 1
f dx dy = =− , while f dy dx = =+ .
y=0 x=0 1 + y 0 2 x=0 y=0 1 + x 0 2
Since these are not equal, the original function f cannot belong to L(I) or else there would be a
contradiction with Fubini’s theorem. 2
10. Let I = [0, 1]×[1, +∞) and let f (x, y) = e−xy −2e−2xy if (x, y) ∈ I. Prove that f ∈/ L(I) by considering
the integrals Z ∞ Z 1 Z 1 Z ∞
f (x, y) dx dy, and f (x, y) dy dx
y=1 x=0 x=0 y=1
But
1 ∞ 1
e−x −x
Z Z Z Z Z
f = f (x, y) dy dx = [e − 1] dx,
x y x=0 y=1 0 x
while
∞ 1 ∞ 1
e−y −y e−1/z −1/z
Z Z Z Z Z Z
f = f (x, y) dx dy = [e − 1] dy = [e − 1] dz,
y x y=1 x=0 1 y 0 z
as shown by elementary methods, with the last equality following from the substitution y ← 1/z.
But for all z ∈ (0, 1), the continuous and bounded integrand functions satisfy
1 −1/z −1/z 1
e [e − 1] > e−z [e−z − 1],
z z
so the two iterated integrals exist but cannot be equal. 2