Homework 3 Solutions

Math 171, Spring 2010

Please send corrections to [email protected]

17.4. Let {an } be a sequence with positive terms such that limn→∞ an = L > 0. Let x be a real number.
Prove that limn→∞ axn = Lx .

Solution. Let  > 0. By Theorem 17.4, note that L < (Lx + )1/x and L > (Lx − )1/x . Since
limn→∞ an = L, there exists some N such that n ≥ N implies an < (Lx + )1/x and an > (Lx − )1/x .
Hence by Theorem 17.4, for n ≥ N we have axn < Lx + and axn > Lx −. This shows limn→∞ axn = Lx .

19.3. Let 0 ≤ α < 1, and let f be a function from R → R which satisfies

|f (x) − f (y)| ≤ α|x − y|
for all x, y, ∈ R. Let a1 ∈ R, and let an+1 = f (an ) for n = 1, 2, .... Prove that {an } is a Cauchy
sequence.

Solution. First we prove by induction on n that |an+1 − an | ≤ αn−1 |a2 − a1 | for all n ∈ N. The base
case n = 1 is obvious. Assuming the formula is true when n = k, we show it is true for n = k + 1:
|ak+2 − ak+1 | = |f (ak+1 ) − f (ak )| ≤ α|ak+1 − ak | ≤ ααk−1 |a2 − a1 | = αk |a2 − a1 |
Hence, by induction, this formula is true for all n.
Note that if |a2 − a1 | = 0, then an = a1 for all n, and so the sequence is clearly Cauchy. Hence we
(1−α)
6 0. Now, given any  > 0, pick N such that αN −1 < |a
consider the case when |a2 − a1 | = 2 −a1 |
, which
n
we can do because 0 ≤ α < 1 =⇒ limn→∞ α = 0. Then, for any m, n ≥ N , with m ≥ n, we have
m−1
X
|am − an | ≤ |ai+1 − ai | by the triangle inequality
i=n
m−1
X
≤ αi−1 |a2 − a1 | by our formula above
i=n
m−1
X
= |a2 − a1 | αi−1
i=n
∞
X
≤ |a2 − a1 | αi−1
i=N
αN −1
= |a2 − a1 | by the formula for the sum of an infinite geometric series
1−α
<  by our choice of N.

Hence {an } is a Cauchy sequence.

20.6. Compute lim supn→∞ an and lim inf n→∞ an , where an is ...

Solution.
(a) n1
1
 1 1 1
Since we know limn→∞ n = 0, we know lim supn→∞ n = 0 = lim inf n→∞ n by Theorem 20.4.

(b) (1 + n1 )n
Since we know limn→∞ (1 + n1 )n = e, we know lim supn→∞ (1 + n1 )n = e = lim inf n→∞ (1 + n1 )n by
Theorem 20.4.

(c) (−1)n (1 − n1 )
Since −1 ≤ (1+ n1 )n ≤ 1 for all n, we have lim inf n→∞ (1+ n1 )n ≥ −1 and lim supn→∞ (1+ n1 )n ≤ 1.
Since the subsequence {a2n−1 } has limit −1 and the subsequence {a2n } has limit 1, we have
lim inf n→∞ (1 + n1 )n ≤ −1 and lim supn→∞ (1 + n1 )n ≥ 1. Hence lim inf n→∞ (1 + n1 )n = −1 and
lim supn→∞ (1 + n1 )n = 1.

20.7. Compute lim supn→∞ an and lim inf n→∞ an and La , where a1 , a2 , . . . is an enumeration of the ra-
tional numbers in the closed interval [0, 1].

Solution. We show that La = [0, 1]. It is easy to show La ⊂ [0, 1].

To show La ⊃ [0, 1], let s ∈ [0, 1]. First we consider the case s > 0. By Theorem, 17.1, there exists
an increasing rational sequence {rn } with limit s. As s > 0, for n sufficiently large we have rn ≥ 0,
so we may assume that rn ≥ 0 for all n, hence rn ∈ [0, 1] for all n. By induction on n, we define a
sequence {bn } which is a subsequence of both {an } and {rn }. For the base case, set b1 = r1 = ak for
some integer k. For the inductive step, suppose we have defined b1 , . . . , bn and bn = rl = ak . Since
a1 , a2 , . . . is an enumeration of the rational numbers, and since the set {rl+1 , rl+1 , . . . } is infinite but
{a1 , . . . ak } is finite, there exists some k 0 > k such that ak0 = rl0 for some l0 > l. Set bn+1 = rl0 = ak0 .
Note that {bn } is a subsequence of both {an } and {rn }. Since {bn } is a subsequence of {rn }, we
have limn→∞ bn = limn→∞ rn = s. Since {bn } is a subsequence of {an }, this shows s ∈ La . The
case s = 0 is analogous. Hence La ⊃ [0, 1], so La = [0, 1].
Since La = [0, 1], we have lim supn→∞ an = lub La = 1 and lim inf n→∞ an = glb La = 0.

20.9. Let {an } be a bounded sequence such that every convergent subsequence of {an } has a limit L.
Prove that limn→∞ an = L.

Solution.
Method 1: Note that La = {L}. Hence lim supn→∞ an = lub(La ) = L = glb(La ) = lim inf n→∞ an .
So by Theorem 20.4, limn→∞ an = L.

Method 2: Suppose for a contradiction that {an } does not have limit L. Hence there exists an
 > 0 such that for any integer N , there exists some n > N with |an − L| > . This allows us to
define n1 < n2 < n3 . . . such that |ani − L| >  for all i. Since {ani } is a bounded sequence, by
Bolzano-Weierstrass it has a convergent subsequence, which clearly does not converge to L. This is
a contradiction, and so it must be that limn→∞ an = L.

20.13. Let {an } and {bn } be sequences such that {an } is convergent and {bn } is bounded. Prove that

lim sup(an + bn ) = lim sup an + lim sup bn

n→∞ n→∞ n→∞

and
lim inf (an + bn ) = lim inf an + lim inf bn
n→∞ n→∞ n→∞

Solution. Since {an } is convergent, it is bounded. So Theorem 20.6 gives us

lim sup(an + bn ) ≤ lim sup an + lim sup bn

n→∞ n→∞ n→∞
2
 Let limn→∞ an = L. Note if l ∈ Lb , then l = limk→∞ bnk for some nk . Hence
lim (ank + bnk ) = lim ank + lim bnk = L + l
k→∞ k→∞ k→∞

so L + l ∈ La+b . This shows that

lim sup(an + bn ) = lub La+b ≥ L + lub Lb = lim sup an + lim sup bn
n→∞ n→∞ n→∞

We’ve shown
lim sup(an + bn ) = lim sup an + lim sup bn
n→∞ n→∞ n→∞
The proof that
lim inf (an + bn ) = lim inf an + lim inf bn
n→∞ n→∞ n→∞
is analogous.

20.20. Let {an } be a sequence of positive numbers such that limn→∞ an = L. Prove that limn→∞ (a1 a2 · · · an )1/n =
L.

Solution. Let  > 0. Then there exists some N such that n ≥ N implies an ≤ L + . Note we
have
(a1 · · · aN +m )1/(N +m) = (a1 · · · aN )1/(N +m) (aN +1 · · · aN +m )1/(N +m)
≤ (a1 · · · aN )1/(N +m) (L + )m/(N +m)
= (a1 · · · aN (L + )−N )1/(N +m) (L + )
Hence we have
lim sup(a1 a2 · · · an )1/n = lim sup(a1 · · · aN +m )1/(N +m)
n→∞ m→∞

≤ lim sup(a1 · · · aN (L + )−N )1/(N +m) (L + ) by Theorem 20.5

m→∞
=L+ by Theorem 16.4
Hence lim supn→∞ (a1 a2 · · · an )1/n ≤ L +  for all  > 0, so lim supn→∞ (a1 a2 · · · an )1/n ≤ L. Anal-
ogously, one can show that lim inf n→∞ (a1 a2 · · · an )1/n ≥ L. Hence by Theorem 20.2 and 20.4, we
have limn→∞ (a1 a2 · · · an )1/n = L.

21.2. Let An = lub{an , an+1 , . . . } and Bn = glb{an , an+1 , . . . } for n = 1, 2, . . . . Compute An , Bn ,

limn→∞ An , and limn→∞ Bn , where an =

Solution.
(a) (−1)n
Clearly An = 1 and Bn = −1 so limn→∞ An = 1 and limn→∞ Bn = −1.

(b) n1
1
Clearly An = n and Bn = 0 so limn→∞ An = 0 and limn→∞ Bn = 0.

(c) (1 + n1 )n
By Theorem 16.6 the sequence {(1 + n1 )n } is increasing and convergent with limit e. So An = e
and Bn = (1 + n1 )n so limn→∞ An = e and limn→∞ Bn = e.
(−1)n
(d) n ( (
1 −1
n n even n+1 n even
We have An = 1
and Bn = −1
. So limn→∞ An = 0 and limn→∞ Bn = 0.
n+1 n odd n n odd
(e) (−1)n (1 − n1 )
3
 We have An = 1 and Bn = −1, so limn→∞ An = 1 and limn→∞ Bn = −1.
P∞ 1
22.4. Prove that the series n=1 n(n+1) converges and find its sum.

Solution. We use partial fractions to rewrite the terms. We try

1 A B
= + =⇒ 1 = (n + 1)A + nB = A + n(A + B) =⇒ A = 1 and B = −1.
n(n + 1) n n+1
1 1 1
Indeed, we have n(n+1) = n − n+1 . So the n-th partial sum is

sn = a1 + a2 + ... + an
1 1 1 1 1 1 
= − + − + ... + −
1 2 2 3 n n+1
1
=1−
n+1
This is an example of a telescoping series. Since
∞ ∞ 1

lim sn = lim 1 − = 1,
n=1 n=1 n+1
P∞ 1
P∞ 1
we have that n=1 n(n+1) converges, and n=1 n(n+1) = 1.
P∞ P∞ P∞
23.5. Give an example of divergent series n=1 an and n=1 bn such that n=1 (an + bn ) converges.
P∞ P∞
Solution. Let an = 1 and bn = −1 for all n. Then n=1 an and n=1 bn are geometric series
with r =
P1, and hence diverge by Theorem 22.4. However, an + bn = 0 for all n so the n-th partial
∞ P∞
sum of n=1 (an + bn ) is zero for all n, giving n=1 (an + bn ) = 0 converges.
P∞
24.9. Prove that if {an } is a decreasing
P∞ sequence of positive numbers and n=1 an converges, then
limn→∞ nan = 0. Deduce that n=1 n1s diverges if 0 ≤ s ≤ 1.
Pn
Solution. Let  > 0. Let sn = k=1 ak . Since {sn } converges, {sn } is Cauchy so there ex-
ists some N such that n, m ≥ N implies |sn − sm | < . In particular, for n ≥ N we have
(n − N )an ≤ aN +1 + · · · + an = |sn − sN | < . Hence limn→∞ (n − N )an = 0. By Theorem
22.3, we have limn→∞ N an = N limn→∞ an = 0 too. So by Theorem 12.2, we have
lim nan = lim (n − N )an + lim N an = 0
n→∞ n→∞ n→∞

so we’re done.
Now, note that for 0 ≤ s ≤ 1, we have limn→∞ n n1s = limn→∞ n1−s 6= 0. By the above, this
P∞
means that n=1 n1s cannot converge, that is, it diverges.

25.2. Let {an } satisfy the hypotheses

P∞ of the alternating series test. Let {sn } denote the sequence of par-
tial sums of the series n=1 (−1)n+1 an . Prove that the sequence {s2n−1 } is decreasing and bounded
below by 0.

Solution. Note s2(n+1)−1 = s2n−1 − a2n + a2n+1 ≤ s2n−1 since a2n+1 ≤ a2n . Hence the sequence
{s2n−1 } is decreasing.
Note s2n−1 = (a1 − a2 ) + (a3 − a4 ) + · · · + (a2n−3 − a2n−2 ) + a2n−1 , where each term in parenthesis
is bounded below by 0, and a2n−1 is also bounded below by 0. Hence {s2n−1 } is bounded below by 0.

Give an example of a sequence {an } of positive numbers such that limn→∞ an = 0, but the series
25.4. P
∞ n+1
n=1 (−1) an diverges.

4
 (
2
n odd
Solution. Let an = n+1 . Then limn→∞ an = 0 but the (2n − 1)-th partial sum of
0 n even
P∞ P∞ P∞
n=1P(−1)n+1 an is equal to the n-th partial sum of n=1 n1 . Since n=1 n1 diverges by Corollary
∞
24.3, n=1 (−1)n+1 an also diverges.
P∞ P∞
26.4. Prove that if n=1 an converges absolutely, then n=1 a2n converges.

Solution. By Theorem 22.3, we have limn→∞ |an | = 0. So by Theorem 13.2, {|an |} is bounded.
P∞ {an } is P
Hence bounded. We apply part (i) of Theorem 26.4 (replacing {bn } with
∞ 2
P∞{an })
2
to see that
a
n=1 n na = a
n=1 n converges absolutely. Since all terms are non-negative, n=1 n converges.
a
an+1 1/n
26.8. Let {an } be a sequence of positive numbers. Prove that if limn→∞ an = L then limn→∞ an = L.
Deduce limn→∞ (n!)n1/n = e.

Solution. Define the sequence {bn } by b1 = a1 and bn = an−1 an

for n ≥ 2. Since limn→∞ aan+1
n
= L,
we have limn→∞ bn = L. Note that an = b1 b2 · · · bn . Applying exercise 20.20 to the sequence {bn },
we get
lim an1/n = lim (b1 b2 · · · bn )1/n = L
n→∞ n→∞

nn
Now, let an = n! . Note that
(n+1)n+1 (n+1)n+1
an+1 (n+1)! n+1 n+1 n 1
lim = lim nn = lim = lim ( ) = lim (1 + )n = e
n→∞ an n→∞
n!
n→∞ nn n→∞ n n→∞ n
By the conclusion above, we have
n
lim = lim a1/n
n =e
n→∞ (n!)1/n n→∞