CHAPTER 3

Sequences and Series

3.1. Sequences and Their Limits

Definition (3.1.1). A sequence of real numbers (or a sequence in R) is a
function from N into R.

Notation.
(1) The values of X : N ! R are denoted as X(n) or xn, where X is the
sequence.
(2) (xn : n 2 N) or simply (xn) may denote a sequence — this is not the same
as {xn : n 2 N}.
(3) (x1, x2, . . . , xn, . . . ).
Example.
(1) (3n) = (3n : n 2 N) =
(3, 6, 9, . . . , 3n, . . . ).
(2) (1) = (1 : n 2 N) =
(1, 1, 1, . . . , 1, . . . ).
(3) ( 2)n = ( 2)n : n 2 N =
2, 4, 8, . . . , ( 2)n, . . . .
⇣1 1 ⌘ ⇣1 1 ⌘
n n
(4) + ( 1) = + ( 1) : n 2 N =
2 2 2 2
(0, 1, 0, 1, . . . , 0, 1, . . . ).

34
 3.1. SEQUENCES AND THEIR LIMITS 35
✓⇣ ⌘ 1 1 n◆ ✓⇣ ⌘ 1 1 ◆
n 2+2( 1) n 2+2( 1)n
(5) = :n2N =
2 2
⇣ n ⌘
1, 1, 1, 2, 1, 3, 1, 4, . . . , 1, , . . . .
2
Sequences may also be defined inductively or recursively.
Example.
(1) x1 = 5, xn+1 = 2xn 3 (n 1) gives
(5, 7, 11, 19, 35, . . . ).
(2) Fibonacci sequence: x1 = x2 = 1, xn+1 = xn 1 + xn (n 2) gives
(1, 1, 2, 3, 5, 8, 13, . . . ).
Definition (3.1.3). A sequence X = (xn) in R is said to converge to x 2 R,
or have limit x, if
8 ✏ > 0 9 K(✏) 2 N 3
8 n K(✏), |xn x| < ✏.
We write this as lim X = x, lim(xn) = x, lim xn = x, or
n!1
xn ! x as n ! 1.
A sequence that converges is called convergent, one that does not divergent.
⇣1⌘
Example. lim = 0.
n
Proof. Let ✏ > 0 be given. By the Archimedean property,
1
9 K(✏) 2 N 3 < ✏. Then,
K(✏)
1 1
for n K(✏),  , and so
n K(✏)
1 1 1
0 =  < ✏.
n n K(✏)
⇣1⌘
Thus lim = 0 by definition. ⇤
n
36 3. SEQUENCES AND SERIES

Calculator Visualisation
1 1
lim = lim = 0 if 8 ✏ > 0,
n!1 n x!1 x
with yM in = 0 ✏ and yM ax = 0 + ✏,
you can find K(✏) 2 N 3 if
xMin = K(✏) and xMax = 1E99,
the graph only enters the screen from the left and exits from the right.
Theorem (3.1.4 — Uniqueness of Limits). A sequence in R can have at
most one limit.
✏
Proof. [The technique.]
2
Suppose lim(xn) = x0 and lim(xn) = x00. By Theorem 2.1.9,
it suffices to show that |x0 x00| < ✏ 8 ✏ > 0, for then
|x0 x00| = 0 =) x0 = x00. Let ✏ > 0 be given.
✏
Since lim(xn) = x0, 9 K 0 2 N 3 8n K 0, |xn x0| < .
2
✏
Since lim(xn) = x00, 9 K 00 2 N 3 8n K 00, |xn x00| < .
2
Let K = max{K 0, K 00}. Then n K =) n K 0 and n K 00 =)
|x0 x00| = |x0 xn + xn x00| 
| {z }
smuggling
✏ ✏
|x0 xn| + |xn x00| < + = ✏. ⇤
2 2
 3.1. SEQUENCES AND THEIR LIMITS 37

Theorem (3.1.5). Let X = (xn) be a sequence in R, and let x 2 R. The

following are equivalent:
(a) X converges to x.
(b) 8 ✏ > 0, 9 K 2 N 3 8 n K, |xn x| < ✏.
(c) 8 ✏ > 0, 9 K 2 N 3 8 n K, x ✏ < xn < x + ✏.
(d) 8 ✏-nbhd. V✏(x) of x, 9 K 2 N 3 8 n K, xn 2 V✏(x).
Proof.
(a) () (b) by definition.
(b) () (c) () (d) since
|xn x| < ✏ () ✏ < xn x < ✏ () x ✏ < xn < x + ✏ ()
xn 2 V✏(x). ⇤

Technique
Given ✏ > 0. Produce or verify the existence of an integer K(✏) so that
n K(✏) =) |xn x| < ✏.
Sometimes |xn x| < ✏ can be converted, with reversible steps, to an inequality
of the form n > f (✏). Take K(✏) as the first integer greater than f (✏) (by the
Archimedean Property), K(✏) = [f (✏)] + 1, for example. Then
n K(✏) =) n > f (✏) =) |xn x| < ✏.
Example.
(1) lim(c) = c, c 2 R, i,e., xn = c 8n 2 N.
Proof. Given ✏ > 0. [To show 9 K(✏) 2 N 3 8 n K(✏), |c c| < ✏.]
|c c| = 0 < ✏ 8 n 2 N. Pick K(✏) = 1.
Then n K(✏) =) |c c| < ✏. ⇤
38 3. SEQUENCES AND SERIES
⇣ 1 ⌘ 1
(2) lim p = 0. xn = p here.
n n
Proof.
h 1 i
Given ✏ > 0. To show 9 K(✏) 2 N 3 n
K(✏) =) p 0 < ✏. Now
n
1 1 1 p 1
p 0 < ✏ () p < ✏ () < n () 2 < n.
n n ✏ ✏
h1i 1 1
Pick K(✏) = 2 + 1. Then n K(✏) =) n > 2 =) p 0 < ✏. ⇤
✏ ✏ n
⇣c⌘
(3) lim p = 0, c 2 R, p > 0.
n
Proof. Case c = 0 was Example 1, so assume c 6= 0. Given ✏ > 0.
c |c| |c| ⇣ |c| ⌘1/p
p
0 < ✏ () p < ✏ () < n () < n.
np n ✏ ✏
⇣ ⌘
|c| 1/p
Take K = + 1.
✏
⇣ |c| ⌘1/p c
Then n K =) n > =) p 0 < ✏. ⇤
✏ n
1 5 1, 000, 000!
Note. Thus xn = p
3
, xn = , and xn = all have limit 0.
n n5/4 n
 3.1. SEQUENCES AND THEIR LIMITS 39
⇣1⌘
(4) lim = 0.
2n
Proof. Given ✏ > 0.
1 1 1 n
0 < ✏ () < ✏ () < 2 ()
2n 2n ✏
1 ln ✏
ln < ln 2n () ln ✏ < n ln 2 () < n.
✏ ln 2
⇢ h
ln ✏ i
Take K = max 1, +1 .
ln 2
ln ✏ 1
Then n K =) n > =) n 0 < ✏. ⇤
ln 2 2
(5) Let xn = 1 + ( 1)n. X = (0, 2, 0, 2, . . . ).
lim(xn) does not exist.
Proof. [We use contradiction.]
Suppose lim(xn) = x. Then, 8 ✏ > 0, 9 K 2 N 3 8 n K, |xn x| < ✏.
In particular, for ✏ = 1, 9 K 2 N 3 8 n K, |xn x| < 1.
(
|0 x| < 1 for n odd
But ,
|2 x| < 1 for n even
so 2 = |2 x + x|  |2 x| + |x|  |2 x| + |x 0| < 1 + 1 = 2,
a contradiction.
Thus lim(xn) does not exist. ⇤
40 3. SEQUENCES AND SERIES
p
(6) Let xn = n. lim(xn) does not exist.
Proof. [We again use contradiction.]
p p
N 3 8 n K, | n x| < ✏
Suppose lim( n) = x.pThen, 8 ✏ > 0, 9 K 2 p
or, equivalently, ✏ < n x < ✏ or x ✏ < n < x + ✏.
p
Then, for ✏ = 1, 8 n K(1), n < x + 1 or n < (x + 1)2, contradicting the
Archimedean Property.
p
Thus lim( n) does not exist. ⇤

Homework
Pages 61-62 #5b,5d (Do not use Theorem 3.1.10 with these — work from the
definition)
Extra Problem: Prove ( 1)n diverges. (Hint: This is a translation of Exam-
ple 5 — watch your inequalities, though.)
Note. Sometimes it is awkward or impossible to solve |xn x| < ✏ for n.
In such cases, it may be possible to establish an inequality of the form
|xn x|  C|an|
where C > 0 and lim an = 0.
 3.1. SEQUENCES AND THEIR LIMITS 41

Theorem (3.1.10). Let (an) and (xn) be sequences in R, lim(an) = 0,

and x 2 R. If for some C > 0 and some m 2 N we have
|xn x|  C|an| 8 n m,
then lim(xn) = x.
Proof. Let ✏ > 0 be given. Since lim(an) = 0,
⇣✏⌘ ⇣✏⌘ ✏
9 Ka 2 N 3 8 n Ka , |an 0| < .
C C C
⇢ ⇣✏⌘
Let Kx(✏) = max m, Ka . Then
C
✏
8n Kx(✏), |xn  C|an|
x| |{z} <
|{z}⇣ ⌘ C · = ✏.
✏ C
n m n Ka
C
Thus lim(xn) = x. ⇤
Example.
⇣ 1 + ( 1)n ⌘
(7) lim = 0.
n
h ⇣ 1 1 1 ⌘i
Proof. X = 0, 1, 0, , 0, , . . . , 0, , . . . .
2 3 n
1 + ( 1)n 1 + ( 1)n 1+1 2
0 =  = !0
n n n n
by Example 3. The result follows from Theorem 3.1.10. ⇤
42 3. SEQUENCES AND SERIES
⇣ n+1 ⌘ 1
(8) lim = .
3n + 2 3
h ⇣2 3 4 5 6 7 n+1 ⌘i
Proof. X = , , , , , , . . . , ,... .
5 8 11 14 17 20 3n + 2
1
n+1 1 3n + 3 3n 2 1 1
= =  = 9 !0
3n + 2 3 3(3n + 2) 3(3n + 2) 3(3n) n
by Example 3. The result follows from Theorem 3.1.10. ⇤
⇣ n+1 ⌘
(9) lim 3 p = 0.
n + n
h ⇣ 3 4 5 6 n+1 ⌘i
Proof. X = 1, p , p , , p ,..., 3 p ,... .
8 + 2 27 + 3 66 125 + 5 n + n
n+1 n+1
3
p 0 = 3
p 
n + n n + n
n+1

n3
n + n 2n 2
= = !0
n3 n3 n2
by Example 3. The result follows from Theorem 3.1.10. ⇤
 3.1. SEQUENCES AND THEIR LIMITS 43
⇣ n3 + 3 ⌘ 1
(10) lim = .
2n3 n 2
h ⇣ 11 10 67 122 n3 + 3 ⌘i
Proof. X = 4, , , , ,..., 3 ,... .
14 17 124 245 2n n
n3 + 3 1 2n3 + 6 2n3 + n n+6
= = 
2n3 n 2 2(2n3 n) 2(2n3 n) |{z}
n 6
n+n 2n 1
= = !0
2(2n3 n3) 2n3 n2
by Example 3. The result follows from Theorem 3.1.10. ⇤

“Ultimate Behavior”

Definition (3.1.8). If X = (x1, x2, . . . , xn, . . . ) is a sequence in R and if

m 2 N, the m-tail of X is the sequence
Xm = (xm+n : n 2 N) = (xm+1, xm+2, . . . , xm+n, . . . ).
⇣ 1 1 1 ⌘
Example. The 4-tail of 1, , , . . . , , . . . is
2 3 n
⇣1 1 1 1 ⌘
X4 = , , , . . . , ,... .
5 6 7 4+n
Theorem (3.1.9). Let X = (xn : n 2 N) be a sequence and let m 2 N.
Then the m-tail Xm = (xm+n : n 2 N) converges () X converges. In
this case,
lim Xm = lim X.
Proof. Read in text — it is just a translation argument. ⇤
Homework
Pages 61-62 #6a, 6c, 10, 11
p p
Extra: Prove lim( n + 1 n) = 0.
Hint for #10: Look at Theorem 3.1.5(c) and pick the right ✏.
44 3. SEQUENCES AND SERIES

3.2. Limit Theorems

Definition (3.2.1). A sequence (xn) is bounded if
9 M > 0 3 |xn|  M 8 n 2 N.
Theorem (3.2.2). If (xn) converges, then (xn) is bounded.
Proof. Suppose lim(xn) = x and ✏ = 1.
Then 9 K(1) 2 N 3 8 n K(1), |xn x| < 1.
Then, for n K(1),
|xn| = |xn x + x|  |xn x| + |x| < 1 + |x|.
Let M = sup |x1|, |x2|, . . . , |xK(1) 1|, 1 + |x| .
Then |xn|  M 8 n 2 N. ⇤

Example. ( 1)n is bounded since |( 1)n|  1 8 n 2 N, but does not

converge. Thus, bounded 6=) convergent.
Example. (2n) diverges.
Proof. If (2n) converged, it would be bounded.
Thus 9 M > 0 3 |2n| = 2n  M 8 n 2 N.
Then n = log2 2n  log2 M 8 n 2 N,
contradicting the Archimedean Property. ⇤
 3.2. LIMIT THEOREMS 45

Theorem (3.2.3). Suppose lim(xn) = x and lim(yn) = y.

(a) lim(xn + yn) = lim(xn) + lim(yn) = x + y.
✏
Proof. [The technique.]
2
✏
Since lim(xn) = x, 9 K1 2 N 3 8n K1, |xn
x| < .
2
✏
Since lim(yn) = y, 9 K2 2 N 3 8 n K2, |yn y| < .
2
Let K = max{K1, K2}. Then n K =) n K1 and n K2 =)
|(xn + yn) (x + y)| = |(xn x) + (yn y)| 
✏ ✏
|xn x| + |yn y| < + = ✏. ⇤
2 2
(b) lim(xn yn) = lim(xn) lim(yn) = x y.
Proof. Similar to the above. ⇤
46 3. SEQUENCES AND SERIES

(c) lim(xnyn) = lim(xn) lim(yn) = xy.

Proof. Let ✏ > 0 be given. Note
|xnyy xy| = (by smuggling)
|(xnyn xny) + (xny xy)| 
|(xnyn xny)| + |(xny xy)| 
|xn||yn y| + |y||xn x|.
[We are now able to gain control over all of the variable parts.]
Since lim(xn) = x:
(1) 9 M1 > 0 3 |xn|  M1 8 n 2 N by Theorem 3.2.2.
Let M = max M1, |y| .
✏
(2) 9 K1 2 N 3 8n K1, |xn x| < .
2M
✏
Since lim(yn) = y, 9 K2 2 N 3 8n K2, |yn y| < .
2M
Let K = max K1, K2 . Then, 8n K,
✏ ✏
|xnyy xy|  |xn||yn y| + |y||xn x|  M · +M · = ✏.
2M 2M
⇤

(d) lim(cxn) = c lim(xn) = cx for c 2 R.

Proof. This is a special case of (c). ⇤
 3.2. LIMIT THEOREMS 47
⇣1⌘ 1
(e) If lim(zn) = z, zn 6= 0 8 n 2 N, and z 6= 0, then lim = .
zn z
Proof. Let ✏ > 0 be given. Note
1 1 z zn 1
= = · |zn z|.
zn z znz |znz|
h 1 i
We need to find a bound for in the first factor.
|zn|
1
Let ↵ = |z| > 0. Since lim(zn) = z:
2
(1) 9 K1 2 N 3 8n K1, |zn z| < ↵. Then
↵< |zn z| 
|{z} |zn| |z| =)
Cor.2.2.4(a)
Th.2.2.2(c)
1 1 2
|z| = |z| ↵  |zn| =)  .
|2 {z } |zn| |z|
|zn| is bounded away from 0
1
(2) 9 K2 2 N 3 8n z| < ✏|z|2.
K2, |zn
2
Let K = max K1, K2 . Then, 8n K,
1 1 1 2 ⇣1 2
⌘
= · |zn z| < 2 ✏|z| = ✏.
zn z |znz| |z| 2
⇤
⇣x ⌘ lim(xn)
n
(f ) If lim(zn) = z, zn 6= 0 8 n 2 N, and z 6= 0, then lim = =
zn lim(zn)
x
.
z
Proof. This follows directly from parts (c) and (e) above. ⇤
48 3. SEQUENCES AND SERIES
⇣ 3n2 2⌘
Example. Find lim .
n2 + n
Proof.
⇣ 3n2 ✓ ◆
2⌘ 3 n22 lim 3 n22
lim = lim = =
n2 + n 1 + n1 lim 1 + n1
2
lim(3) lim n2 3 0
1 = = 3.
lim(1) + lim n
1 0
⇤
Theorem (3.2.4). If lim(xn) = x and xn 0 8 n 2 N, then x 0.
Proof. [Use contradiction by picking an appropriate ✏.]
Suppose x < 0 =) x > 0. Since lim(xn) = x,
for ✏ = x, 9 K 2 N 3 8 n K, |xn x| < x or
( x) < xn x< x or x + x < xn < x + x = 0.
Thus, for n = K, xK < 0, contradicting our hypotheses.
Thus x 0. ⇤
Theorem (3.2.5). If (xn) and (yn) are convergent sequences and if xn  yn
8 n 2 N, then lim(xn)  lim(yn).
Proof. Let zn = yn xn. Then zn 0 8 n 2 N,
so 0  lim(zn) = lim(yn) lim(xn) =) lim(xn)  lim(yn). ⇤
Theorem (3.2.6). If (xn) is convergent and a  xn  b 8 n 2 N, then
a  lim(xn)  b.
Proof. This follows from Theorem 3.2.5 by comparing (a) and (b) with
(xn). ⇤
 3.2. LIMIT THEOREMS 49

Theorem (3.2.7 — Squeeze Theorem). Suppose xn  yn  zn 8n 2 N

and lim(xn) = lim(zn). Then (yn) converges and
lim(xn) = lim(yn) = lim(zn).
Proof. Let w = lim(xn) = lim(zn). Given ✏ > 0.
9 K1 2 N 3 8 n K1, ✏ < xn w < ✏, and also
9 K2 2 N 3 8 n K2, ✏ < zn w < ✏.
Let K = max K1, K2 . Then for n K,
✏ |{z}
< xn w  yn w  zn < ✏ =) |yn
w |{z} w| < ✏.
n K1 n K2
Thus lim(yn) = w. ⇤
Note. The hypotheses of Theorem 3.2.4 thru Theorem 3.2.7 can be weak-
ened to apply to tails of the sequences rather than to the sequences themselves.
Example.
⇣ cos n ⌘
(1) Find lim .
n
1 cos n 1
Solution. 1  cos n  1 =)   .
n n n
⇣ 1⌘ ⇣1⌘
Since lim = lim = 0,
n n
⇣ cos n ⌘
lim = 0 by the Squeeze Theorem. ⇤
n
50 3. SEQUENCES AND SERIES

(2) Find lim n1/n .

Solution. [This one is tricky.]
For n > 1, n1/n > 1, so xn = n1/n = 1 + tn, where tn = n1/n 1 > 0. Then,
from the Binomial theorem,
n(n 1)
n = (1 + tn)n = 1 + ntn + t2n + positive terms,
2
so p
n(n 1) 2 2 2
tn < n =) t2n < =) tn < p .
2 n 1 n 1
Thus
p p
2 2
1 < xn = 1 + tn < 1 + p =) 1 < xn < 1 + p .
n 1 n 1
⇣ p ⌘
2
Since lim(1) = lim 1 + p = 1,
n 1
lim(xn) = lim n1/n = 1 by the Squeeze Theorem. ⇤
⇣p
n 2
(3) Find lim n .
Solution.
p
n 2 p p p p
lim n = lim n n · n n = lim n n · lim n n = 1 · 1 = 1.
⇤
Theorem (3.2.9). Suppose lim(xn) = x. Then lim(|xn|) = |x|.
Proof. We know |xn| |x|  |xn x|. Thus, given ✏ > 0,
if 9 K 2 N 3 8 n K, |xn x| < ✏, we also get |xn| |x| < ✏. ⇤
 3.2. LIMIT THEOREMS 51

Theoremp (3.2.10). Suppose lim(xn) = x and xn 0 8 n 2 N. Then

p
lim( xn) = x.
Proof. [Using the conjugate].
First, x 0 by Theorem 3.2.4. Let ✏ > 0 be given.
Case x = 0 9 K 2 N 3 8 n K, |xn 0| < ✏2 ()
p
0  xn < ✏2 () 0  xn < ✏ () |xn 0| < ✏.
p p
Case x > 0 Then x > 0. 9 K 2 N 3 8 n K, |xn x| < x✏ =)
p p p p
p p ( xn x) · ( xn + x)
| xn x| = p p =
xn + x
p
|xn x| |xn x| x✏
p p  p < p = ✏.
xn + x x x
⇤
Homework
Pages 69-70 #1d, 5b, 6bd (both find and prove)
⇣x ⌘ ⇣
n
Extra: If lim = x 6= 0, then (xn) is not bounded. Hint: Prove (xn)
n⇣ ⌘ ⌘
xn
bounded =) lim =0 .
n
52 3. SEQUENCES AND SERIES

3.3. Monotone Sequences

Definition (3.3.1). Let X = (xn) be a sequence.
X is increasing if x1  x2  · · ·  xn  xn+1  · · · .
X is decreasing if x1 x2 ··· xn xn+1 ···.
X is monotone if it is either increasing or decreasing.
Example.
(1) (1, 1, 2, 3, 5, 8, . . . ) is increasing.
(2) For 0 < b < 1, (b, b2, b3, . . . ) is decreasing.
(3) (2, 0, 2, 0, . . . ) is not monotone.
(4) (4, 2, 1, 3, 3, 5, 5, 7, 7, . . . ) is ultimately increasing.

Theorem (3.3.2 — Monotone Convergence Theorem (MCT)). A mono-

tone sequence converges () it is bounded. Further:
(a) if X = (xn) is a bounded, increasing sequence, then lim(xn) = sup{xn}.
(b) if X = (xn) is a bounded, decreasing sequence, then lim(xn) = inf{xn}.

Proof. (=)) Follows from Theorem 3.2.2.

((=) (a) Since (xn) is bounded,
9 M > 0 3 |xn|  M 8 n 2 N =) xn  M 8 n 2 N.
Thus, by Completeness, x? = sup{xn} exists. Let ✏ > 0 be given.
By Property S, 9 K 2 N 3 x? ✏ < xK (xK = s✏).
Since (xn) is increasing, xK  xn 8 n K. Thus, 8 n K,
? ? ?
x ✏
| {z } < xK  xn
|{z}  x < x
| + }✏ =) |xn
{z x?| < ✏.
Thus lim(xn) = x?.
(b) Similar to (a), except uses Property I. ⇤
 3.3. MONOTONE SEQUENCES 53

Example. Determine
p whether lim(xn) exists and, if so, its value where
x1 = 1 and xn+1 = 1 + xn for n 1.
Solution.
q r q
p p p p
x2 = 1 + 1 = 2, x3 = 1 + 2, x4 = 1 + 1 + 2 , . . .
(a) [Show monotone increasing.]
p
x1 < x2 since 1 < 2. Assume xn  xn+1.
p p
Then xn+1 = 1 + xn  1 + xn+1 = xn+2,
so by induction xn  xn+1 8 n 2 N.
Thus (xn) is increasing.
(b) [Show (xn) is bounded above by 2 using induction.]
x1 = 1 < 2. Suppose xn  2. Then
p p p p
xn+1 = 1 + xn  1 + 2 = 3 < 4 = 2.
Thus, by induction, xn  2 8 n 2 N,
and so 2 is an upper bound of (xn).
(c) Thus lim(xn) = x for some x 2 R by the MCT.
Since (xn+1) is a tail of (xn), lim(xn+1) = x also. Then
p p
x = lim(xn+1) = lim( 1 + xn) = lim(1 + xn) =
p p
lim(1) + lim(xn) = 1 + x =)
p
1± 5
x2 = 1 + x =) x2 x 1 = 0 =) x = .
2
p p
1 5 1+ 5
Since < 0, we conclude x = lim(xn) = . ⇤
2 2
54 3. SEQUENCES AND SERIES

Note.
(1) An increasing sequence is bounded below by its first term. Thus if x? =
sup{xn : n 2 N},
M = max |x1|, |x?|
is a bound for the sequence.
(2) A decreasing sequence is bounded above by its first term.

Homework
Page 77 # 1, 2
Hint for # 2: (a) Show xn xn+1 0 8 n 2 N. Thus (xn) is decreasing.
(b) Show (xn) is bounded below. Then (xn) is bounded by M = max |x1|, |l.b.| .
(c) Find and solve an equation to get x = lim(xn).
Example.
(2) Establish convergence or divergence of (xn) where
1 1 1
xn = 1 + + + · · · + .
1! 2! n!
1
Solution. xn+1 = xn + > xn, so (xn) is increasing.
(n + 1)!
1 1
Noting that < n , we have
(n + 1)! 2
⇣ 1 1 1 1 ⌘
xn < 1 + 1 + + 2 + 3 + · · · + n 1 =
2 2 2 2
1 ( 12 )n ⇣ 1 ⌘n 1
1+ 1 =1+2 < 3,
1 2 2
so (xn) is bounded above and so (xn) converges by the MCT.
Although we now know the limit exists, we do not have a technique for finding
the exact limit. ⇤
 3.3. MONOTONE SEQUENCES 55

(3)

Let An = the sum of the semicircular areas.

Let Ln = the sum of the semicircumferences.
It appears lim An = 0 and lim Ln = 1. Well,
n!1 n!1
⇡ ⇣ 1 ⌘2 ⇡ 1
An = n · · = · ! 0 as n ! 1,
2 2n 8 n
but ⇣1⌘ ⇡
Ln = n · ⇡ · = 8 n 2 N,
2n 2
⇡
so lim Ln = .
n!1 2
56 3. SEQUENCES AND SERIES

Problem (Page 77 # 10). Establish convergence or divergence of (yn) where

1 1 1
yn = + + ··· + 8 n 2 N.
n + 1
| {z } n + 2 2n
|{z}
largest smallest
term term
Solution. It might seem obvious that lim(yn) = 0, but incorrect.
Note that
1 1 1 1
yn + ··· + =n· =
|2n {z 2n} 2n 2
n terms
and
1 1 1 n
yn  + ··· + =n· = < 1,
n
| + 1 {z n + 1
} n + 1 n + 1
n terms
so (yn) is bounded by 1.
Now
1 1 1 1 1
yn+1 = + + ··· + + + ,
n+2 n+3 2n 2n + 1 2n + 2
so
1 1 1 1 1 2
yn+1 yn = + = + =
2n + 1 2n + 2 n + 1 2n + 1 2n + 2 2n + 2
1 1 1
= > 0,
2n + 1 2n + 2 (2n + 1)(2n + 2)
so (yn) is increasing. Thus (yn) converges by the MCT and
1
 lim(yn)  1.
2
Can we find lim(yn)?
 3.3. MONOTONE SEQUENCES 57

Note n n
X 1 X 1 1
yn = = ·
k n
.
n+k 1+ n
k=1 k=1
[What is this latter sum?]

yn is a right-hand Riemann sum for

1
f (x) = for 0  x  1.
1+x
Thus Z 1
1 1
lim(yn) = dx = ln|1 + x| = ln 2 ln 1 = ln 2.
0 1+x 0
⇤
58 3. SEQUENCES AND SERIES

3.4. Subsequences and the Bolzano-Weierstrass Theorem

Definition (3.4.1). Let X = (xn) be a sequence and let
n1 < n2 < · · · < nk < · · ·
be a strictly increasing sequence of natural numbers. Then the sequence
X 0 = (xnk ) = (xn1 , xn2 , . . . , xnk , . . . )
is a subsequence of X.
⇣ 1 ⌘ ⇣1 1 1 1 ⌘
Example. Let X = = , , ,..., ,... .
2n 2 4 6 2n
Some subsequences:
⇣ 1 ⌘ ⇣1 1 1 1 ⌘
(1) (xnk ) = = , , ,..., ,... .
4k 4 8 12 4k
⇣ 1 ⌘ ⇣1 1 1 1 ⌘
(2) (xnk ) = = , , ,..., ,... .
4k 2 2 6 10 4k 2
⇣ 1 ⌘ ⇣1 1 1 1 ⌘
(3) (xnk ) = = , , , . . . , 2, . . . .
4k 2 4 16 36 4k
⇣ 1 ⌘ ⇣1 1 1 1 ⌘ ⇣1 1 1 1 ⌘
(4) (xnk ) = = , , ,..., ,... = , , ,..., ,... .
(2k)! 2! 4! 6! (2k)! 2 24 720 (2k)!
(5) X itself
(6) Any tail of X
In general, to form a subsequence of X, just pick out any infinite selection of
terms of X going from left to right.
 3.4. SUBSEQUENCES AND THE BOLZANO-WEIERSTRASS THEOREM 59

Theorem (3.4.2). If X = (xn) converges to x, so does any subsequence

(xnk ).
Proof. Let ✏ > 0 be given. Since (xn) converges to x,
9K2N3 8n K, |xn x| < ✏. Since
n1 < n2 < · · · < nk < · · ·
is an increasing sequence in N, nk k 8 k 2 N.
Let K 0 = nK . Then, 8 nk K 0 = nK , nk K =) |xnk x| < ✏.
Thus (xnk ) converges to x. ⇤
1
Example. For c > 1, find lim(c n ) if it exists.
Solution.
1
(a) xn = c n > 1 8 n 2 N, so (xn) is bounded below.
1 1 1 1
(b) xn xn+1 = c n c n+1 = c n+1 c n(n+1) 1 > 0 8 n 2 N,
so (xn) is decreasing.
(c) Thus lim(xn) = x exists.
[Using a subsequence to find x.]
1 1 1 1
2 2
Now x2n = c 2n = c n = xn , so
1 1
2
x = lim(x2n) = lim xn = x 2 =)
x2 = x =) x2 x = 0 =) x(x 1) = 0 =) x = 0 or x = 1.
Since xn > 1 8 n 2 N, lim(xn) = 1. ⇤
60 3. SEQUENCES AND SERIES

Theorem (3.4.7 — Monotone Subsequence Theorem). If X = (xn) is a

sequence in R, then there is a subsequence of X that is monotone.
Proof. We will xm a peak if n m =) xn  xm (i.e, if no term to the
right of xm is greater than xm).
Case 1 : X has infinitely many peaks.
Order the peaks by increasing subscripts. Then
xm1 xm2 ··· xmk ··· ,
so
(xm1 , xm2 , . . . , xmk , . . . )
is a decreasing subsequence.
Case 2 : X has finitely many (maybe 0) peaks.
Let xm1 , xm2 , . . . , xmr denote these peaks.
Let s1 = mr + 1 (the first index past the last peak) or s1 = 1 if there are no
peaks.
Since xs1 is not a peak, 9 s2 > s1 3 xs1 < xs2 .
Since xs2 is not a peak, 9 s3 > s2 3 xs2 < xs3 .
Continuing, we get an increasing subsequence. ⇤

Theorem (3.4.8 — Bolzaono-Weierstrass Theorem). A bounded sequence

of real numbers has a convergent subsequence.
Proof. If X = (xn) is bounded, by the Monotone Subsequence Theorem it
has a monotone subsequence X 0 which is also bounded. Then X 0 is convergent
by the MCT. ⇤
 3.4. SUBSEQUENCES AND THE BOLZANO-WEIERSTRASS THEOREM 61

Theorem (3.4.4). Let X = (xn) be a sequence. The following are equiv-

alent:
(a) (xn) does not converge to x 2 R.
(b) 9 ✏0 > 0 3 8 k 2 N, 9 nk 2 N 3 nk k and |xnk x| ✏0.
(c) 9 ✏0 > 0 and a subsequence X 0 = (xnk ) of X 3 |xnk x| ✏0 8 k 2 N.
Proof.
[(a) =) (b)] This is the negative of the definition of convergence.
[(b) =) (c)] Take the ✏0 from (b).
Let n1 2 N 3 |xn1 x| ✏0.
Let n2 2 N 3 n2 > n1 and |xn2 x| ✏0.
Let n3 2 N 3 n3 > n2 and |xn3 x| ✏0.
Continuing, we generate the subsequence.
[(c) =) (a)] Suppose X = (xn) has a subsequence X 0 = (xnk ) satisfying (c).
If xn ! x, so would (xnk ) ! x. Then 9 K 2 N 3 8k K, |xnk x| < ✏0.
But this contradicts (c). ⇤
62 3. SEQUENCES AND SERIES
⇣ p
n⇡ ⌘ 2
Example. cos does not converge to .
4 2
⇣ p p p p
n⇡ ⌘ ⇣ 2 2 2 2 ⌘
Proof. cos = , 0, , 1, , 0, , 1, . . . .
4 2 2 2 2
p
2
Let ✏0 = . 8 k 2 N, let nk = 8k + 3.
4
⇣ ⌘ ⇣ p ⌘
(8k + 3)⇡ 2
Then (xnk ) = cos = .
4 2
p p p p
2 2 2 p 2
Then 8 k 2 N, xnk = = 2 = ✏0.
2 2 2 4
⇣ ⌘ p
n⇡ 2
Thus cos does not converge to . ⇤
4 2
Theorem (3.4.5 — Divergence Criterion). If a sequence X = (xn) has
either of the following properties, then X is divergent.
(a) X has two convergent subsequences X 0 = (xnk ) and X 00 = (xrk ) whose
limits are not equal.
(b) X is unbounded.

Homework
Pages 84-85 # 4b, 9 (Hint: Use Theorem 3.4.4), 11 (Hint: What is the only
possible limit?), 14 (extra credit)
 3.4. SUBSEQUENCES AND THE BOLZANO-WEIERSTRASS THEOREM 63

Theorem (3.4.9). Let X = (xn) be a bounded sequence such that every

convergent subsequence converges to x. Then lim(xn) = x.
Proof. Let M be a bound for X. Suppose xn 6! x. By Theorem 3.4.4,
9 ✏0 > 0 and a subsequence X 0 = (xnk ) 3 |xnk x| ✏0 8 k 2 N.
Now M is also a bound for X 0 = (xnk ),
so it has a convergent subsequence X 00 = (xnkr ) with lim(xnkr ) = x.
Then 9 K 2 N 3 8 r K, |xnkr x| < ✏0, a contradiction. ⇤
Example. We cannot drop the bounded hypothesis:
⇣ 1 1 1 ⌘
1, , 3, , 5, , . . . .
2 4 6
64 3. SEQUENCES AND SERIES

3.5. The Cauchy Criterion

Example. Suppose lim(xm xn) = 0? Does lim(xn) necessarily exist?
p
NO! xn = n is a counterexample.
Definition (3.5.1). A sequence X = (xn) is a Cauchy sequence if
8 ✏ > 0 9 H(✏) 2 N 3 8 n, m H(✏) with n, m 2 N, |xn xm| < ✏
.
Lemma (3.5.3). If X = (xn) converges, then X is Cauchy.
✏
Proof. [Another argument.]
2
✏
Suppose lim(xn) = x. Given ✏ > 0, 9 K 2 N 3 8 n K, |xn x| < .
2
Let H = K. Then, for m, n H = K,
✏ ✏
|xn xm| = |(xn x) + (x xm)|  |xn x| + |xm x| < + =✏
2 2
Thus (xn) is Cauchy. ⇤
Note. (xn) is not Cauchy if
9 ✏0 > 0 3 8H 2 N, 9 n, m H 3 |xn xm| ✏0
.
p
Example. (xn) = n is not Cauchy.
Proof. Let ✏0 = 1 and H 2 N be given.
p p p
Let m = H, so xm = m = H.
p
Since n is unbounded,
p p
9n2N3 | n H| 1
where n H.
p p
Thus | xn xm| 1. ⇤
 3.5. THE CAUCHY CRITERION 65

Lemma (3.5.4). Cauchy sequences are bounded.

Proof. Let X = (xn) be Cauchy and ✏ = 1
9H2N3 8n H, |xn xH | < 1.
Then
|xn| |xH |  xn xH < 1 =)
1 < |xn| |xH | < 1 =) |xn| < |xH | + 1.
Let M = max |x1|, |x2|, . . . , |xH 1|, |xH | + 1 .
Then |xn|  M 8n 2 N. ⇤
Theorem (3.5.5 — Cauchy Convergence Criterion). A sequence is con-
vergent () it is Cauchy.
✏
Proof. [Yet another argument.]
2
(=)) Lemma 3.5.3
((=) Let X = (xn) be Cauchy. Then X is bounded,
so by B-W, X has a convergent subsequence, say X 0 = (xnk ) ! x?.
[To show lim(xn) = x?.] Let ✏ > 0 be given.
✏
Since (xn) is Cauchy, 9 H 2 N 3 8 n, m H, |xn xm| < .
2
✏
Since lim(xnk ) = x?, 9 K 2 N 3 K H and |xK x?| < .
2
✏
But |xn xK | < also. Then, for n H,
2
✏ ✏
|xn x?| = |(xn xK ) + (xK x?)|  |(xn xK )| + |(xK x?)| < + = ✏.
2 2
Thus lim(xn) = x?. ⇤
66 3. SEQUENCES AND SERIES
⇣1⌘
Example. (xn) = is Cauchy.
n
Proof.
Given ✏ > 0. WLOG (without loss of generality), suppose n m.
Then
1 1 1 1 1 1
= < (= m > .
< ✏ |{z}
n m m n m ✏
if
h1i
Take H = + 1. Then
✏
1 1 1
n m H =) n m> =) < ✏.
✏ n m
⇤
 3.5. THE CAUCHY CRITERION 67
⇣ 1 1 1⌘
Problem (Page 91 # 2b). Show 1 + + + · · · + is Cauchy.
2! 3! n!
Proof.
h
In this proof we use the facts that 2n 1  n! (Example 1.2.4(e)) and that
2 n 1 rn+1 i
1 + r + r + ··· + r = . Given ✏ > 0. WLOG, suppose n m.
1 r
⇣ 1 1 1⌘ ⇣ 1 1 1⌘
|xn xm| = 1 + + + · · · + 1 + + + ··· + =
2! 3! n! 2! 3! m!
1 1 1 1 1 1
+ + ··· +  m + m+1 + · · · + n 1 
(m + 1)! (m + 2)! n! 2 2 2
1⇣ 1 1 ⌘ 1 1 1 n m
2 2n m 1
1 + + ··· n m 1 = m · = 
2m 2 2 2 1 12 2n 1
2n m 1 1 m 1 ln ✏ ln ✏
= < ✏ (= < 2 (= < m 1 (= 1 < m.
2n 1 2m 1 ✏ ln 2 ln 2
n h ln ✏ i o
Choose H = max 1, 1 + 1 . Then
ln 2
ln ✏
n m H =) m > 1 =) |xn xm| < ✏.
ln 2
⇤
Definition (3.5.7). A sequence X = (xn) is contractive if
9 0 < C < 1 3 |xn+2 xn+1|  C|xn+1 xn| 8 n 2 N.
C is the constant of the contractive sequence.
68 3. SEQUENCES AND SERIES

Theorem (3.5.8). Every contractive sequence is convergent.

Proof. [We prove the sequence to be Cauchy, and thus convergent.]
Let X = (xn) be a contractive sequence. 8 n 2 N,
|xn+2 xn+1|  C|xn+1 xn|  C 2|xn xn 1|  · · ·  C n|x2 x1|.
Then, WLOG for m > n,
|xm xn|  |xm xm 1|  |xm x |  · · ·  |xn+1
1 xn| 
| {zm 2 }
smuggling + triangle inequality
Cm 2
+ C m 3 + · · · + C n 1 |x2 x1| =
Cn 1
Cm n 1
+ C m n 2 + · · · + 1 |x2 x1| =
⇣ Cm n ⌘
n 1 1
C |x2 x1| 
1 C
⇣ 1 ⌘
n 1
C |x2 x1| ! 0 as n ! 1
1 C
since lim(C n) = 0. Thus (xn) is Cauchy, and so convergent. ⇤
 3.5. THE CAUCHY CRITERION 69

1
Example. x1 = 1, x2 = 2, xn = (xn 2 + xn 1) for n 3.
2
⇣ 3 7 13 27 ⌘
(xn) = 1, 2, , , , , . . . .
2 4 8 16
(a) (xn) is contractive. Thus (xn) converges.
Proof.
1 1 1 1
|xn+2 xn+1| = (xn + xn+1) xn+1 = xn xn+1 = |xn+1 xn|.
2 2 2 2
⇤
(b) Note that
1 1
|xn+1 xn| = |x
1 2
x1| = .
2n 2n 1

(c) [To find lim(x2n+1) = lim(xn).]

1 1 1
x2n+1 x2n 1 = (x2n 1 + x2n) x2n 1 = x2n x2n 1 =
2 2 2
1 1 1 1
|x2n x2n 1| = · 2n 2 = 2n 1 .
2 2 2 2
Thus
1 1 1
x2n+1 = x2n 1 + 2n 1 = x2n 3 + 2n 3 + 2n 1 = · · · =
2 2 2
1 1 1 1 1 h 1 1 1 i
1 + + 3 + 5 + · · · + 2n 1 = 1 + 1 + 2 + 4 + · · · + 2n 2 =
2 2 2 2 2 2 2 2
1 h 1 1 1 i 1 h 1 ⇣ 1 ⌘ 2 ⇣ 1 ⌘n 1i
1 + 1 + 2 + 4 + · · · + 2n 2 = 1 + 1 + + + ··· + =
2 2 2 2 2 4 4 4
1 n 1
1 1 4 1 4 4 1
n
1+ · = 1 + · =
2 1 14 2 4 1
1
4 4n 1 2 5
1+ !1+ = as n ! 1.
6 3 3
5
Thus lim(xn) = lim(x2n+1) = .
3
70 3. SEQUENCES AND SERIES

3.6. Properly Divergent Sequences

Definition. Let (xn) be a sequence.
(a) We say (xn) tends to +1 and write lim(xn) = +1 if
8 ↵ 2 R 9 K(↵) 2 N 3 8 n K(↵), xn > ↵.
(b) We say (xn) tends to 1 and write lim(xn) = 1 if
8 2 R 9 K( ) 2 N 3 8 n K( ), xn < .
We say (xn) is properly divergent in either case.

Example. For C > 1, lim(C n) = +1

Proof. Let ↵ 2 R be given. [How to express C > 1.]
C = 1 + b where b > 0. By the Archimedean Property,
↵
9 K(↵) 2 N 3 K(↵) > . Then 8 n K(↵),
b
C n = (1 + b)n |{z} 1 + nb > 1 + ↵ > ↵.
Bernoulli
n
Thus lim(C ) = +1. ⇤
Homework
Page 91 # 2a, 3b, 7, 9
 3.7. INTRODUCTION TO INFINITE SERIES 145

3.7. Introduction to Infinite Series

Definition (3.7.1). If X = (xn) is a sequence in R, then the infinite series

(or just series) generated by X is the sequence S = (sk ) defined by
s1 = x1
s2 = s1 + x2 (= x1 + x2)
...
sk = sk 1 + xk (= x1 + x2 + · · · + xk )
...

The xn are the terms of the series and the sk are the partial sums of the series.
If lim S exists, we say the series is convergent and call this limit the sum or
value of the series. If this limit does not exist, we say this series S is divergent.
Notation. 1
X X X
(xn) or xn or xn
n=1
We can also use 1 1
X X
xn or xn
n=0 n=5
If the first term of the series is xN , then the first partial sum is sN .
146 3. SEQUENCES AND SERIES

Example.
1
X
(1) rn = 1 + r + r2 + · · · + rn + · · · (geometric series)
n=0

sn = 1 + r + r2 + · · · + rn
rsn = r + r2 + · · · + rn + rn+1
sn(1 r) = sn rsn = 1 rn+1
For r 6= 1,
rn+1
1
sn =
1 r
Then 1
X 1 rn+1 1
n
r = lim = if |r| < 1,
n=0
n!1 1 r 1 r
1
X
and rn diverges if |r| 1.
n=0

1
X 1 1 1 1
(2) = + + + ···.
n=1
n(n + 1) 1 · 2 2 · 3 3 · 4
1 1 1
Since = ,
k(k + 1) k k + 1
⇣ 1⌘ ⇣1 1⌘ ⇣1 1⌘ ⇣1 1 ⌘
sn = 1 + + + ··· + =)
2 2 3 3 4 n n+1
1
sn = 1 =)
n+1
X1
1 ⇣ 1 ⌘
= lim 1 = 1.
n=1
n(n + 1) n!1 n + 1
 3.7. INTRODUCTION TO INFINITE SERIES 147
1
X
(3) ( 1)n = 1 1+1 1 + ···.
n=0
1
X
S = (sn) = (1, 0, 1, 0, . . . ) diverges =) ( 1)n diverges.
n=0
P
Theorem (3.7.3 —nth Term Test). If xn converges, lim(xn) = 0.
P
Proof. xn converges =) s = lim(sn) exists =)
s = lim(sn 1) =)
lim(xn) = lim(sn sn 1) = lim(sn) lim(sn 1) = s s = 0.
⇤
Example.
(4) Geometric series with |r| 1 diverges since (rn) diverges.
X1
1 1 1 1
(5) For p = 1 + p + p + ··· + p + ···,
n=1
n 2 3 n
⇣ 1 ⌘
lim(xn) = lim p = 0.
n
But
1 1 1
sn = 1 + p + p + · · · + p
2 3 n
1 1 1 1 1 p
p + p + p + ··· + p = n · p = n
n n n n n
| {z }
n terms
X1
p 1
Thus lim(sn) lim( n) ! 1, so p diverges.
x=1
n
Note. This implies the converse of Theorem 3.7.3 is not true.
148 3. SEQUENCES AND SERIES
1
X
(6) Consider cos n.
n=1
Assume lim(cos n) = 0 =) lim(cos2 n) = 0 =)
lim(sin2 n) = lim(1 cos2 n) = lim(1) lim(cos2 n) = 1 0 = 1..
Then lim(sin2 2n) = 0 as a subsequence.
Now sin 2n = 2 sin n cos n =) sin2 2n = 4 sin2 n cos2 n =)
lim(sin2 2n) = 4 lim(sin2 n) lim(cos2 n) = 4 · 1 · 0 = 0 6= 1,
a contradiction.
1
X
Thus lim(cos n) 6= 0 =) cos n diverges.
n=1
Theorem (7.3.4 — Cauchy Criterion for Series).
P
xn converges () 8 ✏ > 0 9 M 2 N 3 if m > n M , then
|sm sn| = |xn+1 + xn+2 + · · · + xm| < ✏.

PTheorem (3.7.5). Let (xn) be a sequence of nonnegative numbers. Then

xn converges () S = (sk ) is bounded. In this case,
Xn
xn = lim(sk ) = sup{sk : k 2 N}.
i=1

Proof. Since xn 0 8 n 2 N, (sk ) is monotone increasing.

By the MCT, S = (sk ) converges () it is bounded, in which case
lim(sk ) = sup{sk }.
⇤
 3.7. INTRODUCTION TO INFINITE SERIES 149

Example.
1
X 1
(7) The harmonic series diverges. The proof is similar to that of Example
n=1
n
5.
X1
1
(8) The p-series p
converges for p > 1.
n=1
n
Proof. Since (sk ) is monotone, we need only to show (sk ) is bounded. But
it suffices to show that some subsequence is bounded.
Let k1 = 21 1 = 1, so sk1 = 1.
Let k2 = 22 1 = 3. Then, since 2p < 3p,
1 ⇣1 1⌘ 2 1
sk2 = p + p + p < 1 + p = 1 + p 1 .
1 2 3 2 2
Let k3 = 23 1 = 7 =)
⇣1 1 1 1
sk3 = sk2 + p + p + p + p
4 5 6 7
4 1 1
< sk2 + p < 1 + p 1 + p 1 .
4 2 4
Continuing inductively, if kj = 2j 1,
1 1 1
0 < skj = 1 + p 1 + p 1 + · · · + j 1 p 1
2 4 (2 )
1 1 1
= 1 + p 1 + p 1 2 + ··· + p 1 j 1
2 (2 ) (2 )
1
< .
1 2p1 1
X1
1
Thus (skj ) is bounded and so p
converges for p > 1. ⇤
n=1
n
150 3. SEQUENCES AND SERIES

X1
1
(9) The p-series p
diverges for p  1.
n=1
n
1 ⇣1⌘
p
For p  0, =n with p 0 =) lim p 6= 0.
np n
1 1
For 0 < p  1, np < n =)  p .
n n
Since the partial sumes of the harmonic series are not bounded, neither are the
X1
1
partial sums here. Thus p
diverges for p  1.
n=1
n

Theorem (9.3.2 — Alternating Series Test). Let Z = (zn) be a decreasing

sequence of strictly positive numbers with lim(zn) = 0. Then the alternating
1
X
series ( 1)n+1zn converges.
n=1

Proof. s2n = (z1 z2) + (z3 z4) + · · · + (z2n 1 z2n),

and since zk zk+1 > 0, (s2n) is increasing.
Since s2n = z1 (z2 z3) (z4 z5) ··· (z2n 2 z2n 1) z2n,
s2n < z1 8 n 2 N =) (s2n) converges by the MCT.
Suppose lim(s2n) = s. Then, given ✏ > 0, 9 K 2 N 3 for n K,
✏ ✏
|s2n s| < and |z2n+1| < .
2 2
Then for n K,
✏ ✏
|s2n+1 s| = |s2n + z2n+1 s|  |s2n s| + |z2n+1| <
+ = ✏.
2 2
Thus, for n large enough, each partial sum is within ✏ of s, so lim(sn) = s and
X1
( 1)n+1zn = s. ⇤
n=1
 3.7. INTRODUCTION TO INFINITE SERIES 151

Example.
X1
1
(10) The alternating harmonic series ( 1)n+1 converges.
n=1
n
1
X
Note. If ( 1)n+1zn = s, |s sn|  zn+1.
n=1

X1 X10
1 1 1627 1
For ( 1)n+1 , s ( 1)n+1 = s < .
n=1
n n=1
n 2520 11

Theorem (3.7.7 — Comparison Test). Let (xn) and (yn) be sequences

where 0  xn  yn for n K 2 N. Then
P P
(a) yn converges =) xn converges.
P P
(b) xn diverges =) yn diverges.
Proof.
P
(a) Suppose yn converges and ✏ > 0 is given.
9M 2N3 m>n M =) yn+1 + · · · + ym < ✏.
Let M 0 = max{K, M }. Then m > M 0 =)
0  xn+1 + · · · + xm  yn+1 + · · · + ym < ✏,
P
so xn converges.
(b) Contrapositive of (a). ⇤
152 3. SEQUENCES AND SERIES

Example.
X1
1
(11) 2+n
converges since
n=1
n
X1
1 1 1
< 8 n 2 N and converges (p = 2).
n2 + n n2 n=1
n2

1
X X 1 1
1
(12) . Seems to be like .
n=1
n2 n+1 n=1
n2

1 1
But > 8n 2 N.
n2 n + 1 n2
1 2
We can show 0 < 2  2 8n 2 N , but this is not obvious.
n n+1 n

Theorem (3.7.8 — Limit Comparison Test).⇣Suppose ⌘ (xn) and (yn) are

xn
strictly positive sequences and suppose r = lim exists.
yn
P P
(a) If r 6= 0, xn converges () yn converges.
P P
(b) If r = 0 and yn converges, then xn converges.
Proof.
⇣x ⌘
n
(a) Since r = lim 6= 0, 9 K 2 N 3
yn
1 xn
r  2r for n K =)
2 yn
⇣1 ⌘
r yn  xn  (2r)yn for n K.
2
The result follows by applying comparison twice.
(b) If r = 0, 9 K 2 N 3 0 < xn  yn for n K.
The result follows by comparison. ⇤
 3.7. INTRODUCTION TO INFINITE SERIES 153

Example.
1
X 1
(12 continued) .
n=1
n2 n+1
Since
✓ 1 ◆ ⇣ ⌘ ✓ ◆
n2 n+1 n2 1
r = lim 1 = lim = lim 1 = 1,
n2
n2 n+1 1 n + n12
1
X X 1 1
1
converges by limit comparison since converges.
n=1
n2 n+1 n=1
n2

1
X 1
(13) p .
n=1
3n + 1
X1 X1
1 1
We compare with p = 1/2
, which diverges as a p-series with p  1.
n=1
n n=1
n
✓p 1 ◆ ⇣ 1/2 ⌘ ✓h i1/2◆
3n+1 n n
r = lim = lim = lim =
p1 (3n + 1)1/2 3n + 1
n
 ⇣ n ⌘ 1/2 ⇣ 1 ⌘ 1
lim = =p =6 0.
3n + 1 3 3
1
X 1
Thus p diverges by limit comparison.
n=1
3n + 1