Ma 4111: Advanced Calculus
Solutions to Homework Assignment 3
Prof. Wickerhauser
Due Tuesday, October 9th, 2012
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.
1. Prove that every closed subset of R is the intersection of a countable collection of open sets.
Solution: Let S R be closed. Then R S is open, so by theorem 3.11 we can write R S =
S
k=1 Ak , S
where {Ak T: k = 1, 2, . . .} is a countable collection of disjoint open intervals of R. Thus
S = R k=1 Ak = k=1 (R Ak ), where each R Ak is the union of at most two disjoint closed
intervals. If Ak = (ak , bk ), then we can call the two intervals of its complement Lk = (, ak ] and
Rk = [bk , ), though one or both of these might be empty.
T def T
We now note that (, ak ] = j=1 Lkj where Lkj = (, ak +1/j) is open, and [bk , ) = j=1 Rkj
def T T
where Rkj = (bk 1/j, ) is also open. Thus Lkj Rkj is open and S = k=1 j=1 (Lkj Rkj ) is a
countable intersection of open sets since the indices are taken from the countable set Z+ Z+ . 2 2
For Problems 23, a set S Rn is called convex if for every pair of points x, y S and every real
number satisfying 0 < < 1 we have x + (1 )y S.
2. Prove that (a) an n-ball is convex; and (b) an n-dimensional open interval is convex.
Solution: For (b), we note that x (x, y) and y (x, y) implies ak xk bk and ak yk bk
for all k = 1, 2, . . . , n. But then since both and 1 are positive, we have ak + (1 )ak
xk + (1 )yk bk + (1 )bk for all k = 1, 2, . . . , n which means that x + (1 )y (x, y).
For (a), theorem 3.3 implies that if x B(x, r) and y B(x, r), then kx + (1 )y xk =
k[x x] + (1 )[y x]k kx xk + (1 )ky xk < r.
Remark. This proof also shows that the closed n-ball and the closed n-interval are convex. 2
3. Prove that the intersection of any collection of convex sets is convex.
T
Solution: Let X = KF K be an arbitrary intersection of convex sets K F and suppose x and
y belong to X. Then x and y belong to K for each K F and therefore x + (1 )y K for every
K F . Hence, x + (1 )y X. 2
4. Prove that the collection of isolated points of a set S Rn must be countable.
Solution: Around each isolated point x S is an open n-ball B(x) which contains no other points
of S. By theorem 3.27, there is an n-ball Ax with rational radius and rational center coordinates such
that x Ax B(x). The map x 7 Ax is a one-to-one correspondence between the isolated points of
S and a subset of the countable set of all open n-balls with rational center and radius. 2
1
5. (a) Give an example of a closed subset U R which is not bounded and an infinite open cover F of
U which has no finite subcover. (b) Give an example of a bounded subset B R which is not closed
and an infinite open cover F of B which has no finite subcover.
Solution: For (a) take U = R and consider the open cover {(n 1, n + 1) : n Z}. Any finite
subcover must be bounded both above and below, and thus could not cover R.
For (b) take B = (0, 1) and consider the open cover {( n1 , n2 ) : n Z+ }. Any finite subcover will give a
union that is bounded below by a positive number and will thus omit /2 (0, 1). 2
For Problems 6 and 7, let S be a subset of Rn and define a condensation point of S to be any point
x Rn such that every n-ball centered at x contains uncountably many points of S.
6. Prove that every uncountable subset S Rn contains a condensation point of S. (Hint: use the fact
that the countable union of countable sets is countable).
Solution: Suppose that no point of S is a condensation point of S. We will show that S must be
def
countable. But then for each x S there is some n-ball B(x) centered at x such that Ax = B(x) S
n
is
S countable. Then {B(x) : x S} is an open cover of S R and has a countable S subcover S
S k=1 B(x k ), defined by
S a countable subset {x k : k = 1, 2, . . .} S. But then S = S k=1 B(xk ) =
k=1 [S B(x k )] = k=1 Axk
is a countable union of countable sets Axk
, and is countable by theorem
2.27. 2
7. Assume that S is an uncountable subset of Rn . Let T be the collection of all condensation points of S.
Prove the following: (a) S T is countable; (b) S T is uncountable; (c) T is closed; (d) T contains
no isolated points.
Solution: For (a), note that if ST is uncountable then by Solution 6 it must contain a condensation
point of S T and thus of S. This contradicts the assumption that T contains all the condensation
points of S.
For (b), note that S is the disjoint union of S T and S T . Since S T is countable by part (a) but
the union S is uncountable, we conclude that S T must be uncountable.
For (c), let x be any accumulation point of T . Then any ball B(x; r) contains a point of T {x}, say
y. Since ky xk < r, we can find q > 0 such that q < r ky xk. We claim that B(y; q) B(x; r):
if x B(y; q), then kx xk kx yk + ky xk < q + ky xk < r ky xk + ky xk = r. But
B(y; q) contains uncountably many points of S since y T , so B(x; r) must also contain uncountably
many points of S. Since r > 0 was chosen arbitrarily we have shown that x T and we conclude that
T is closed.
For (d), note that if x T is an isolated point then there is a ball B(x) which contains no points of T
def
besides x. Consider the set Sx = S (B(x) {x}) S. This is uncountable, so by Solution 6 Sx
must contain a condensation point y of Sx . But y B(x) will also be a condensation point of S since
Sx S, so y T . But y 6= x since x
/ Sx , which contradicts our assumption that the only point of
T in B(x) is x. 2
For Problems 810, let kxk be the usual norm of x Rn , and define
n
def def
X
kxk1 = |xi |, kxk = max |xi |
i=1,2,...,n
i=1
for any x = (x1 , x2 , . . . , xn ) Rn .
2
8. Show that k k1 satisfies the norm axioms:
(i) kxk1 0, with equality if and only if x = 0.
(ii) kaxk1 = |a| kxk1 , for any scalar a and vector x.
(iii) kx + yk1 kxk1 + kyk1 for any vectors x, y Rn .
Solution: For (i), note that the sum of nonnegative numbers |x1 | + + |xn | is nonnegative and is
zero if and only if all of them are zero. For (ii), compute
n
X n
X n
X
kaxk1 = |axi | = |a| |xi | = |a| |xi | = |a| kxk1
i=1 i=1 i=1
For (iii), use the triangle inequality in R: |xi + yi | |xi | + |yi | for each i = 1, . . . , n, so
n
X n
X n
X n
X
|xi + yi | (|xi | + |yi |) = |xi | + |yi |.
i=1 i=1 i=1 i=1
2
9. Show that k k satisfies the three norm axioms of Problem 8.
Solution: For (i), note that the largest element of a set of nonnegative numbers |x1 |, . . . , |xn | is
nonnegative and is zero if and only if all of them are zero. For (ii), compute
kaxk = max |axi | = max |a| |xi | = |a| max |xi | = |a| kxk
i i i
For (iii), use the triangle inequality in R: |xi + yi | |xi | + |yi | for each i = 1, . . . , n, so
max |xi + yi | max (|xi | + |yi |) max |xi | + max |yi |.
i i i i
2
n
10. Find four positive constants A, B, C, D such that for every x in R we have
Akxk kxk1 Bkxk and Ckxk kxk Dkxk.
For full credit, find the largest A and C and the smallest B and D for which these inequalities hold,
and prove that no better numbers exist. (Hint: find examples in R1 or R2 which require the best
constants A, B, C, D.)
Solution: We claim that A = 1, B = n, C = 1n and D = 1 are the best constants:
1
kxk 1 kxk1 2 nkxk and kxk 3 kxk 4 kxk.
n
But first we show that they work:
(1) kxk kxk1 holds because the squares of the two nonnegative sides satisfy
2
|x1 |2 + . . . + |xn |2 |x1 | + . . . + |xn | .
All the additional cross terms on the right hand side are nonnegative.
(2) kxk1 nkxk holds by the CauchySchwarz inequality (theorem 1.23), using ai = xi and bi = 1
for all i = 1, . . . , n.
(3) 1 kxk kxk holds because |x1 |2 + . . . + |xn |2 n max{|xi |}2 .
n
(4) kxk kxk holds because max{|xi |2 } |x1 |2 + . . . + |xn |2 .
To show that no better constants exist, take x = (1, 0, . . . , 0) for (1) and (4), and take x = (1, 1, . . . , 1)
for (2) and (3). 2
