Tests for Nonabsolute Convergence

Let us recall first the following tests for absolute convergence:

1. Comparison Test 3. Root Test 5. Integral Test

2. Limit Comparison Test 4. Ratio Test 6. Raabe’s Test

and note that absolute convergence implies convergence of the series but not the other way around.
There are many series such as
∞ ∞
X (−1)n X (−1)n
√
3
and √ ,
n=1 n2 n=2
n2 − 1

which are convergent but not absolutely. So how are we going to prove or determine the convergence
of such type of series? We present here some tests for this case and show later that indeed, these
are convergent series.

1 Alternating Series Test

∞
X
Definition 1. A series of the form (−1)n+1 zn with zn ≥ 0 is called an alternating series.
n=1
.
Theorem 1 (Alternating Series Test). Let Z = {zn } be a decreasing sequence of strictly positive
∞
X
numbers with lim zn = 0. Then the alternating series (−1)n+1 zn is convergent.
n=1

Proof:
Let {sn } be the corresponding sequence of partial sums for the series (−1)n+1 zn . To prove
P
the theorem, we will consider the subsequence of {sn } with even indices and show that this is
convergent. Next, we will prove that the other half of {sn } is also convergent and converges to the
same limit.
Consider
s2n = z1 − z2 + z3 − z4 + · · · + z2n−1 − z2n
(1)
= (z1 − z2 ) + (z3 − z4 ) + · · · + (z2n−1 − z2n ).
Since zk − zk+1 ≥ 0, then each of the grouped terms in (1) is also nonnegative. This implies that
the subsequence {s2n } is increasing. Moreover, (1) can be rewritten as

s2n = z1 − (z2 − z3 ) − · · · − (z2n−2 − z2n−1 ) − z2n ,

and from this, it also follows that s2n ≤ z1 for all n ∈ N which makes this subsequence bounded.
Thus, from the Monotone Convergence Theorem, {s2n } converges in R.
Let s = lim s2n . We show next, that

{s2n+1 } → s. (2)

Indeed, s = lim s2n means that for every ε > 0, there exists N1 ∈ N such that for n ≥ N1 ,
ε
|s2n − s| < . (3)
2

1
Also, from the assumption that lim zn = 0, for every ε > 0, there exists N2 ∈ N such that for
n ≥ N2 ,
ε
|z2n+1 | < , (4)
2
by the uniqueness of the limit. Take K = sup{N1 , N2 } so that for every n ≥ K, in view of (3) and
(4), we get
|s2n+1 − s| = |s2n + z2n+1 − s|
ε ε
≤ |s2n − s| + |z2n+1 | < + = ε,
2 2
which proves (2). Hence, every partial sum of an odd number of terms is also within an ε−neighborhood
n is large enough. Thus, the convergence of {sn } follows since ε is arbitrary. Therefore, the
of s if P
series (−1)n+1 zn is convergent. 

2 The Dirichlet and Abel Tests

∞
X
Let us now study series of the type xn yn . We start by considering the arbitrary long sums of
n=1
this series.
. .
Lemma 2 (Abel’s Lemma or Partial Summation Formula). Let X = {xn } and Y = {yn } be
P .
sequences in R and let the partial sums of yn be denoted by {sn } with s0 = 0. If m > n ≥ 0,
then
Xm m−1
X
xk yk = (xm sm − xn+1 sn ) + (xk − xk+1 )sk . (5)
k=n+1 k=n+1

Proof:
Since yk = sk − sk−1 for k = 1, 2, . . ., the left side of (5) is the same as
m
X
xk (sk − sk−1 ).
k=n+1

Now, separating the terms with sn and sm and collecting the terms multiplying sn+1 , . . . , sm−1
gives the right side of (5). 

In addition to the Alternating Series Test, we have the following theorem.

.
Theorem 3 (Dirichlet’s Test). Let X = {xn } be a decreasing sequence with lim xn = 0 and suppose
∞
X ∞
X
that the partial sums {sn } of yn are bounded. Thus, the series xn yn is convergent.
n=1 n=1

Proof:
Suppose |sn | ≤ B for all n ∈ N for some B > 0 and let m > n. Using Abel’s Lemma and the
fact that xk − xk+1 ≥ 0, it follows that
m
X m−1
X
xk yk = (xm sm − xn+1 sn ) + (xk − xk+1 )sk
k=n+1 k=n+1
m−1
X
≤ |xm sm | + |xn+1 sn | + |(xk − xk+1 )sk |
k=n+1
m−1
X
≤ (xm + xn+1 )B + (xk − xk+1 )B
k=n+1
= [(xm + xn+1 ) + (xn+1 − xm )] B
= 2xn+1 B.

2
 ∞
X
Since lim xk = 0, the convergence of xn yn follows now from the Cauchy Criterion for Series. 
n=1

Remark 2.1. One has the following:

∞
X ∞
X
1. The series is not concerned with the product of two series xn and yn since in general
n=1 n=1

∞
X ∞
X ∞
X
xn yn 6= xn yn .
n=1 n=1 n=1

∞
X
2. The converse of the of the theorem is not true. That is, the convergence of xn yn does not
n=1
∞
X
necessarily imply both xn decreasing to 0 and the sequence of partial sums {sn } of yn is
n=1
∞
X 1
bounded. One can take for example the series for p ∈ (0, 1]. Let
np+1
n=1

1 1
xn yn = · .
n np
∞
X
Both xn and yn decreases to 0 but neither of the sequence of the partial sums of xn or
n=1
∞
X
yn is bounded.
n=1

Another test we can apply to determine the convergence of a series is given by the next theorem.
.
Theorem 4 (Abel’s Test). Let X = {xn } be a convergent monotone sequence and suppose the
∞
X ∞
X
series yn is convergent. Thus, the series xn yn is also convergent.
n=1 n=1

Proof:
We will use the Dirichlet’s Test to prove this property.
Let {xn } be decreasing convergent sequence with limit x. Let

un = xn − x, n ∈ N,

and so un also decreases to 0. This equation gives us xn = x + un , and hence,

xn yn = xyn + un yn .
∞
X ∞
X
From the Dirichlet Test, un yn is convergent since yn is convergent. Furthermore, we also
n=1 n=1
∞
X ∞
X ∞
X
have xyn convergent because yn is. We then conclude that xn yn is also convergent.
n=1 n=1 n=1
On the other hand, for the case that {xn } is increasing with limit x, we let vn = x − xn for
n ∈ N so that {vn } decreases to 0. This provides xn = x − vn and that

xn yn = xyn − vn yn .

The conclusion then follows by using the same arguments as above. 

3
Remark 2.2. The information stated in Remark 2.1 are also true for Abel’s Test.

Before we deal with more series examples, in addition to the steps given in the previous section,
on has to take note of the following steps/reminders:

1. Test for possible divergence by using the simple test for divergence due to the Term Test.
That is, if the limit of the terms of the series is not 0, then the series diverges.

2. Test for possible absolute convergence by using any of the tests for absolute convergence. If
absolutely convergent, then the series is convergent, otherwise, proceed to the next step.

3. Use a test for nonabsolute convergence.

P∞ P∞
Example 1. Let us take series of the form n=1 an cos(nx) or n=1 an sin(nx).

(i) Since   
x 1 x
2 sin (cos(x) + · · · + cos(nx)) = sin n+ x − sin ,
2 2 2
it follows that if x 6= 2kπ (k ∈ N), then

n + 12 x − sin x




sin 2 1
| cos(x) + · · · + cos(nx)| = ≤
.
2 sin x2 x


sin 2

Hence, the Dirichlet Test implies that if {an } is decreasing with lim an = 0, then the series
∞
X
an cos(nx)
n=1

converges provided x 6= 2kπ.

(ii) Similarly, we have

x 
x  
1
2 sin (sin(x) + · · · + sin(nx)) = cos − cos n+ x ,
2 2 2

so that when x 6= 2kπ (k ∈ N), one has

1
| sin(x) + · · · + sin(nx)| ≤ x

.
sin 2

Hence, as in (i), if {an } is decreasing with lim an = 0, then the series

∞
X
an sin(nx)
n=1

converges by Dirichlet’s Test, provided x 6= 2kπ.

Note that the limit of the terms is 0 by the Squeeze Theorem for both series, and that one
X∞ X∞
cannot apply Abel’s Test since cos(nx) and sin(nx) are divergent.
n=1 n=1

Example 2. Investigate the convergence or divergence of the following series.

∞ ∞ ∞
X (−1)n X cos(n) X (−1)n
1. √ 2. 3. √
3
n2 n2 n2 − 1
n=1 n=1 n=2

4
 ∞ ∞ √ ∞
X
n 2n4 + 7 X
n+1 n X an
4. (−1) 6. (−1) 8. (−1)n , a ∈ [0, 1)
6n9 − 2n ln n n
n=1 n=2 n=1
∞ ∞
X 2n2 X 22n n2
5. (−1)n 7.
n=1
n2 + 1
n=2
en n! ln2 n

Solution:
1. We proceed in several steps.

(i) We first note that when n is even,

(−1)n 1
lim √ = lim √ = 0,
n→∞ 3 n2 n→∞ 3 n2

and when n is odd,

(−1)n −1
lim √3
= lim √
3
= 0,
n→∞ n2 n→∞ n2
so that
(−1)n
lim √ = 0.
n→∞ 3 n2

The Term Test then, is inconclusive.

(ii) Next, we check for possible absolute convergence. Taking the corresponding series of the
absolute values gives us
∞ ∞
X (−1)n X 1
√
3
= √
3
,
n 2 n 2
n=1 n=1
∞
2
X (−1)n
which is a divergent p−series since p = 3 is strictly less than 1. Hence, √
3
is not
n=1 n2
absolutely convergent.
1
(iii) Let us attempt to use the Alternating Series Test. Let zn = √
3
and consider {zn }.
n2
1
(iii.1) One has lim √ 3
= 0.
n→∞ n2 √
3
p
(iii.2) For all n ∈ N, we have n < n + 1 so that n2 < 3 (n + 1)2 . This hands us
1 1
p < √
3
,
3
(n + 1) 2 n2
which implies that zn+1 < zn for all n ∈ N. Thus, {zn } is decreasing.
From these results, hence, we can apply the Alternating Series Test and conclude that
the series is convergent.
∞
X (−1)n
The arguments from (i) - (iii), therefore, mean that the series √
3
is convergent, in
n=1 n2
particular, conditionally convergent.
cos(n)
2. First, the Squeeze Theorem gives us lim = 0 so that the Term Test is inconclusive.
n→∞ n2
That is, the series may or may not converge.
∞
X | cos(n)|
Consider now the series of positive terms given by . We have
n2
n=1

| cos(n)| 1
0< ≤ 2,
n2 n

5
 ∞
X 1
since | cos(n)| ≤ 1 for all n ∈ N. Since is a convergent p−series with p = 2 > 1, then
n2
n=1
∞
X | cos(n)|
the series is also convergent by the Comparison Test. This means that the series
n2
n=1
∞
X cos(n)
is absolutely convergent, and hence, convergent.
n2
n=1
1
One can also apply the result obtained in Example1 (i) above by taking an = 2 to conclude
n
∞
X cos(n)
the convergence of the series via Dirichlet’s Test.
n2
n=1

3. As in item 1, we will proceed step-by-step.

(i) Using the same argument as in Example 1 (i), one has
(−1)n
lim √ = 0,
n→∞ n2 − 1
and so the series may or may not converge.
∞
X 1
(ii) Let us now consider the associated series of positive values given by √ and
n=2
n2 − 1
∞ ∞
X 1 X 1 √
compare this to the series √ = . Note that n2 − 1 < n2 so that n2 − 1 < n
n=1 n2 n=1 n
for all n ∈ N. Thus, we have
1 1
< 2 .
n n −1
∞ ∞
X 1 X 1
The divergence of then follows from the divergence of by the Compar-
n2 − 1 n
n=1 n=1
ison Test.
Therefore, the series is not absolutely convergent.
1
(iii) Let zn = √ and we consider {zn }.
2
n −1
1
(iii.1) Checking the first condition of the Alternating Series Test gives us lim √ =
n→∞ 2
n −1
0.
(iii.2) FOr the second condition, we take the first derivative of zn with respect to n and
get
d −n
(zn ) = p < 0,
dn (n2 − 1)3
for all n ≥ 2.
Thus, the series is convergent by the Alternating Series Test.
From (i)-(iii), the series is convergent, in this case, conditionally.
4. First, we observe that (exercise)
4
 
n 2n + 7
lim (−1) = 0.
n→∞ 6n9 − 2n
∞
n4 1 X 2n4 + 7
Next we consider 9 = 5 and compare the series of absolute values given by
n n 6n9 − 2n
n=1
∞
X 1
to the series which is a convergent p−series with p = 5 > 1.
n5
n=1

6
 1 2n4 + 7
Let yn = and x n = . Thus,
n5 6n9 − 2n
7
xn 2+ n4 1
lim = lim 2 = < ∞.
n→∞ yn n→∞ 6 − 3
n5
∞ ∞
X X 2n4 + 7
This means that the series xn = is also convergent by the Limit Compar-
6n9 − 2n
n=1 n=1
∞
X 2n4 + 7
ison Test. Therefore, the series (−1)n is absolutely convergent, which makes it
6n9 − 2n
n=1
convergent.
∞
X 2n2 2n2
5. Note that the series (−1)n is alternating with z n = . Since
n2 + 1 n2 + 1
n=1

2n2
lim zn = lim = 2 6= 0,
n→∞ n→∞ n2 + 1

then
2n2
 
n
lim (−1) 2 6= 0,
n→∞ n +1
and so, the series diverges by failing the Term Test.

6. One can show that the terms of the series does not converge to 0 so that the series is divergent.
1 22n n2
7. Let xn = and y n = .
ln2 n en n!
(i) Taking the derivative of xn with respect to n yields
d −2
(xn ) = < 0,
dn n ln3 n
1
for n ≥ 2, so that {xn } is a decreasing sequence. Moreover, lim xn = lim = 0.
n→∞ n→∞ ln2 n
∞ ∞
X X 22n n2
(ii) Consider now the series yn = . One has
en n!
n=1 n=1

yn+1 22 (n + 1)
lim = lim = 0 < 1.
n→∞ yn n→∞ en2
∞
X
Thus, yn is convergent by the corollary to the Ratio Test.
n=1
∞
X 22n n2
Hence, the convergence of the series follows from either Dirichlet’s or Abel’s
n=2
en n! ln2 n
Test.

8. We can determine the convergence in two ways.

∞
(−1)n n
X
(i) Let yn = and xn = a . Thus, by the Alternating Series Test, yn is convergent.
n
n=1
On the other hand, since a ∈ [0, 1), then lim an = 0.Moreover, it can also be shown
n→∞
that an is decreasing. Hence, the convergence of the series follows by applying either the
Dirichlet’s or the Abel’s Test.

7
 1
(ii) The convergence of the series can also be shown by taking yn = (−1)n an and xn =
n
and then use either Dirichlet’s or Abel’s Test (exercise).
∞
X an
(iii) Taking the series of the absolute values gives us . The absolute convergence
n
n=1
follows then, by applying the corollary to the Root Test (exercise) and so the series is
convergent.