You are on page 1of 7

HW 4 Solutions

Math 115, Winter 2009, Prof. Yitzhak Katznelson

9.6 Let x1 = 1 and xn+1 = 3x2n for n ≥ 1.


a) If a = lim xn , then we have that a = lim xn+1 = lim 3x2n =
3 lim(xn · xn ) = 3(lim xn )2 = 3a2 , where the later steps use various
limit laws. Solving this equation yields two solutions; either a = 0 or
a = 1/3.
b) Yes, the limit exists because xn diverges to ∞. To prove this, we
can first show by induction that xn ≥ 3n−1 ; the basis for induction is
x1 = 1, which is true. And if xn ≥ 3n−1 , then xn+1 = 3x2n > 3xn ≥
3 · 3n−1 = 3n , so that is the induction step. Now pick an M > 0; we
want to find N such that n > N ⇒ xn > M. Let N = log3 (M) + 1.
Then if n > N, xn ≥ 3n−1 > 3N −1 = 3log3 (M ) = M. 
c) There is no contradiction, because when we assume that lim xn = a
and do algebra to figure out what a is, we’re implicitly assuming that
a is not ±∞. As you may remember from calculus, the rules of algebra
don’t always work with ±∞.

9.11 a) Suppose that lim sn = +∞ and inf{tn : n ∈ N} > −∞; we


want to show that lim(sn + tn ) = +∞. To do this, let t = inf{tn :
n ∈ N}; then t > −∞. Now pick M > 0. Since lim sn = ∞, there
exists N such that n > N implies that sn > max{M − t, 1}. Then
sn > M − t, so sn + t > M. But t = inf{tn }, and so tn ≥ t; thus
sn + tn ≥ sn + t > M whenever n > N. This is what we wanted. 
(Aside: The reason we use max{M − t, 1} instead of just M − t
is that technically the definition 9.8 only guarantees the existence of
such an N when M − t > 0, and M − t might not be positive. That
requirement is absolutely unnecessary, though, since the only thing we
really care about is large M, and any N that works for, say, 1 will also
work for any negative number. So we can replace “for M > 0” in 9.8
with “for any M”, and similarly with M < 0 in the second part of the
definition. I will not worry about whether M > 0 again.)
b) Show that if lim sn = +∞ and lim tn > −∞, then lim(sn + tn ) =
+∞. To do this, notice that tn is a convergent sequence, so it is
bounded (Theorem 9.1). Then there is an M such that |tn | ≤ M for
all n, so tn ≥ −M for all n, so inf{tn } ≥ −M > −∞. Then part a)
immediately gives us what we want. 
c) If lim sn = +∞ and tn is bounded, then just as in b), inf{tn } >
−∞. Thus by part a), we are done immediately, and lim(sn + tn ) =
+∞. 
1
2

9.12. Assume all sn 6= 0 and that the limit L = lim | sn+1 sn


| exists.
a) Show that if L < 1, then lim sn = 0.
First select a so that L < a < 1; notice that, say, the average of L
and 1 would work. Now, lim | sn+1 sn
| = L < a, so let ǫ = a − L; then
there exists an N where n > N ⇒ || sn+1
′ ′
sn
| − L| < ǫ. In particular, if
we let N = N + 1, then n ≥ N implies that | sn+1

sn
| < L + ǫ = a, i.e.
that |sn+1 | < a|sn |. So certainly |sN +1 | < a|sN |.
Now we see that |sN +2 | < a|sN +1 | < a2 |sN |. This pattern continues,
and we see that |sN +k | < ak |sN | for any k > 0. (Technically this
requires induction, but this is obvious enough that you could just state
it and say it’s proved by induction). Changing variables and letting
n = N + k, we see that for n > N, we have |sn | < an−N |sN |.
But now, look at lim an−N |sN |; |sN | is a number, so this limit is
|sN | lim an−N . Since a < 1, lim an = 0, so lim an−N = a−N lim an =
0. Thus the limit of an−N |sN | = 0. Since |sn | < an−N |sN | for all
sufficiently large n, the squeeze theorem tells us that lim sn = 0. 
b) Show that if L > 1, then lim |sn | = +∞.
Let tn = s1n . Then we notice that | tn+1
tn
sn
| = | sn+1 |. So: we know that
sn+1 sn+1
| sn | converges to L, that | sn | =
6 0 (because no sk is ever 0), and that
L 6= 0. Thus by Lemma 9.5, we conclude that | tn+1 tn
sn
| = | sn+1 | converges
to 1/L. And since L > 1, we know 1/L < 1, so we may apply part a)
to conclude that lim tn = 0. So of course lim |tn | = 0. And now: the
|sn | are all positive real numbers, and lim 1/|sn | = lim |tn | = 0, so by
Theorem 9.10, lim |sn | = +∞. 
n
9.15. Show that lim an! = 0 for all a ∈ R.
n
Pick a ∈ R, and let sn = an! . If a = 0, all the terms are 0 and we
are done immediately. If a 6= 0, then sn 6= 0 for all n. Now look at the
limit
an+1
sn+1 (n+1)! an+1 n! a
lim | | = lim | an | = lim | n | = lim | |
sn n!
a (n + 1)! n+1
1
= |a| lim = |a| · 0 = 0.
n+1
Thus L = lim | sn+1
sn
| exists and is equal to 0 < 1; so by problem 9.12
a), we conclude that lim sn = 0, as desired. 
n+1
9.18. a) To show that 1 + a + a2 + . . . + an = 1−a 1−a
for a 6= 1,
multiply both sides by 1 − a. We get the equation (1 − a) + (a − a2 ) +
(a2 − a3 ) + . . . + (an − an+1 ) = 1 − an+1 , which is true since all except
3

the first and the last terms on the left cancel. Of course this doesn’t
make sense for a = 1. 
b) Suppose |a| < 1; we want to find lim(1 + a + a2 + . . . + an ).
n+1
Well, by a), that is equal to lim 1−a 1−a
. By the limit laws, this is
1 n+1 n+1
1−a
(1 − lim a ). But |a| < 1, so lim a = a lim an = a · 0 = 0 (see
1
below). Thus the limit we want is 1−a . 
n
(Justification for saying that lim a = 0: if a = 0, it’s obvious. If
n
not, then no an is zero, so we can apply 9.12a): lim | a a+1 n | = lim |a| =
n
|a| < 1, so lim a = 0).
c) To calculate lim(1 + 31 + 19 + 27
1
+ . . . + 31n ), just apply part b) with
a = 31 ; the answer is 1−1 1 = 32 . 
3
d) What is lim(1 + a + a2 + . . . + an ) for a ≥ 1? Well, let sn =
1 + a + a2 + . . . + an . Then we’ll show lim sn = +∞; pick M > 0. Let N
be an integer bigger than M. If n > N, then sn = 1+a+a2 +. . .+an >
1 + a + a2 + . . . + aN . There are N + 1 terms in the sum, and since
a ≥ 1 each of them is at least 1, so the sum sn ≥ N + 1 > N > M.
This shows that lim sn = +∞. 

10.1. Which of the following sequences are nondecreasing? nonin-


creasing? bounded?
a) n1 is certainly nonincreasing, and is bounded: | n1 | ≤ 1.
n
b) (−1)
n2
is not monotone since it changes signs, but it is bounded:
(−1)n 1
| n2 | = | n2 | ≤ 1.
c) n5 is nondecreasing, but it is not bounded. (n5 ≥ n)
d) sin( nπ7
) is not monotone; it jumps around. But the range of sin
as a function is between −1 and 1, so | sin( nπ 7
)| ≤ 1 and it’s bounded.
e) (−2)n is not monotone, since it alternates sign. It’s not bounded
either, since |(−2)n | = |2n |, which goes to infinity.
f) 3nn is monotone nonincreasing. Proof: we want to show that 3n+1n+1 ≤
n n+1
3n
; i.e. that (multiplying by 3 ) n + 1 ≤ 3n, i.e. that 1 ≤ 2n. This
is true since n ≥ 1, so it’s monotone nonincreasing (which means it’s
bounded above by the first term). And now notice that 3nn is positive,
so it’s bounded below. This means it’s bounded.

10.6 a) Let (sn ) be a sequence such that for all n ∈ N, |sn+1 − sn | <
2 . We want to show that (sn ) is Cauchy (and hence convergent).
−n

To do this, pick ǫ > 0. Because lim 2−n = 0, there exists some N ∈ N


such that 2−N < ǫ. Now we claim that for m and k greater than N,
|sm − sk | < ǫ. We may assume “without loss of generality” that m ≤ k;
if not, switch the names of m and k.
4

By the generalized triangle inequality (you did this on homework),


we can write |sm −sn | ≤ |sm −sm+1 |+|sm+1 −sm+2 |+. . .+|sk−1 −sk |. The
hypothesis is that |sn+1 −sn | < 2−n , but that’s the same as |sn −sn+1 | <
2−n . So apply the hypothesis to each term in the sum, and we see that:

|sm − sk | ≤ 2−m + 2−m−1 + 2−m−2 + . . . + 2−k+1


1 1 1
= 2−m (1 +
+ + . . . + k−m−1 ).
2 4 2
By Problem 9.18, that sum is equal to

−m 1 − ( 21 )k−m 1
2 1 = 2−m+1 (1 − ( )k−m ) ≤ 2−m+1 ,
1− 2 2
and since m > N, m − 1 ≥ N, and this is less than or equal to 2−N ,
which is less than ǫ. So if m and k > N, then |sm − sk | < ǫ. Thus (sn )
is Cauchy. 
b) No, it is not true. To see this, let sn = 12 + 41 + 61 + . . . + 2n
1
. Then
1 1
|sn+1 − sn | = 2n+2 < n . However, sn does not converge: suppose for
contradiction that it does. Then the sequence 2sn converges also, and
the sequence 2sn = 1 + 21 + 13 + . . . + n1 . However, it is a standard result
that this sequence diverges to infinity. So we have a contradiction, and
sn does not converge, so it could not have been Cauchy. 
(To justify this “standard result”, note the following. The first term
is greater than 1/2. The second and third terms are each greater than
1/4, and there are two of them, so their sum is greater than 1/2. The
fourth-seventh terms are each greater than 1/8, and there are four of
them, so their sum is greater than 1/2. The eighth-fifteenth terms are
each greater than 1/16 and there are eight of them, so their sum is
greater than 1/2. In this way we see that the sum of the first 2N − 1
terms is greater than N · 1/2, so if n > 22M −1, then 2sn > 2M/2 = M,
so the sequence diverges to infinity).
(Also notice that the critical ingredient is the fact that the sum of
the terms 1/n does not converge. That’s exactly the criterion you need
for an argument of the form in a) to work).

10.7. Let S be a bounded nonempty subset of R and suppose sup S ∈ /


S. We want to show that there is a nondecreasing sequence (sn ) of
points in S such that lim sn = sup S.
We will construct this sequence by induction. First let s = sup S.
Since S is nonempty, there exists some element s1 of S. s1 < s, since
s∈/ S. Let d = s − s1 . Then |s1 − s| ≤ d.
5

Now suppose we have a nondecreasing sequence s1 , . . . , sn ∈ S with


d
the property that for each k between 1 and n, (s − sk ) ≤ 2k−1 . We
want to find an sn+1 ∈ S such that sn+1 ≥ sn and also (s − sn+1 ) ≤ 2dn ,
because then sn+1 would extend our sequence and ensure it has the
same properties. To do this, let t = 12 (sn + s). Then t < s, so t is not
an upper bound for S. So there is some element of S greater than t
(but not equal to s, since s ∈
/ S); call this sn+1 , and it’s clearly bigger
than sn . Now:
1 1 d d
s − sn+1 < s − t = (s − sn ) < n−1
= n.
2 22 2
So sn+1 extends our sequence. By construction, then, we can form
an infinite sequence (sn ) with the properties that:
i) sn ∈ S,
ii) sn is nondecreasing,
iii) 0 < s − sn < 2dn , where d = s − s1 .
But d is a fixed number, so lim 2dn = 0, so by the squeeze theorem
and iii), lim(s − sn ) = 0, so lim sn = s = sup S. Thus our sequence has
all the properties we want it to. 

10.11. Let t1 = 1 and tn+1 = [1 − 4n1 2 ] · tn for n ≥ 1.


a) We want to show that lim tn exists. Because monotone bounded
sequences converge, if we can show that tn is monotone nonincreasing
and bounded below, we will be done.
Notice that 0 < 4n1 2 < 1, so 0 < 1 − 4n1 2 < 1 also. Now t1 is positive,
and if tn is positive then so is tn+1 (which is tn multiplied by a positive
number), and so by induction all tn are positive. And since all tn are
positive and tn+1 = c · tn for a positive number c < 1, tn+1 < tn . So tn
are bounded below by zero and the sequence is monotone nonincreasing.
Thus it converges, and so lim tn exists. 
b) You aren’t supposed to know the answer to this question. If you
write out a few terms, you might see that it seems to converge around
some value, but you’d probably never guess that it’s 2/π.

1
10.12. Let t1 = 1 and tn+1 = [1 − (n+1)2 ] · tn for n ≥ 1.

a) To show that lim tn exists, the reasoning is precisely identical


to that of Problem 10.11 a) - show it’s monotone nonincreasing and
bounded below.
b) Again, you aren’t supposed to know the answer to this question
instantly. You might guess it’s 1/2, but you couldn’t prove that.
6

c) To show that tn = n+1


2n
, use induction. The basis for induction
1+1
is that t1 = 2 = 1, which is true by hypothesis. Now assume that
tn = n+1
2n
; then

1 n+1 (n + 1)2 − 1) n+2 (n + 1) + 1


tn+1 = [1 − 2
] = = = ,
(n + 1) 2n 2n(n + 1) 2(n + 1) 2(n + 1)
so tn+1 = (n+1)+1
2(n+1)
, and this is the inductive step. So by induction,
n+1
tn = 2n for all n. 
d) Now what’s lim tn = lim n+12n
? Well, it’s lim( 12 + 2n
1
) = 21 +lim 2n
1
=
1
2
. That’s it. 

11.5. Let (qn ) be an enumeration of all the rationals in the interval


(0,1].
a) We claim that the set of subsequential limits of (qn ) is the closed
interval [0,1]. This would be easy if we didn’t have to worry about the
ordering of the (qn ).
First we claim that given any number a ∈ [0, 1], there is a sequence of
distinct rational numbers (rn ) in (0,1] that converge to it. The proof is
as follows: for any n, consider the interval (a−1/n, a+1/n). As long as
a ∈ [0, 1] and n > 2, either (a, a + 1/n) or (a − 1/n, a) will be contained
in (0, 1]. So pick r1 and r2 arbitrary, and then for each successive n
let rn be a rational number in that open interval (whichever works for
that a). Since there are infinitely many rational numbers in each open
interval, we can always pick rn to be different from any of the finitely
many previous terms. This will do it, because we’ll have |rn −a| < 1/n,
and so lim rn exists and equals a.
Now (qn ) is an enumeration of the rationals in (0, 1], and so each
rn = qin for some in ∈ N. The problem is that (rn ) might be in a
different order than the (qn ), and thus not a subsequence.
Consider the sequence in - remember that in is “the place in the
list of qn where rn occurs.” By Theorem 11.3, in has a monotonic
subsequence. But in is a sequence of distinct natural numbers, and
so that subsequence cannot be nonincreasing! If it were, you would
eventually get to 1, and then have nothing left to be the next term. So
that monotonic subsequence must be nondecreasing - in fact increasing,
since the terms are distinct. Call the increasing subsequence ikn . But
then let sn = rkn . The ikn are a subsequence of in , and so kn is
increasing, so we have that sn is a subsequence of rn . But also, sn is
equal to the kn th term of qn , so sn is a subsequence of qn .
7

Since rn converges to a, so does its subsequence sn . And sn is a


subsequence of qn . Therefore a is a subsequential limit. 
(Yes, this is rather horrible. I’m curious - can any of you figure out
a faster way to do it? If so, let me know).
b) To find lim sup qn and lim inf qn , use Theorem 11.7, part ii). The
lim sup qn is equal to sup Q, where Q is the set of subsequential limits
of (qn ). But we know Q = [0, 1], so lim sup qn = 1. Similarly, by the
other part of Theorem 11.7 ii), lim inf qn = inf Q = 0. 

You might also like