Section 11.6 Absolute Convergence and the Ratio and Root Tests
EX.21
Let an= n2+ 1 2n2+ 1
n
. lim
n→∞
p|an n| = lim
n→∞
n2+ 1 2n2+ 1 = 1
2 < 1, so the given series is absolutely convergent by the Root Test.
EX.27
Let an = (−1)n−11 · 3 · 5 · · · (2n − 1)
(2n − 1)! . Apply the Ratio Test,
n→∞lim
an+1 an
= lim
n→∞
1 · 3 · · · (2n + 1) (2n + 1)!
(2n − 1)!
1 · 3 · · · (2n − 1) = lim
n→∞
1
2n = 0 < 1 Therefore the given series is absolutely convergent by the Ratio Test.
EX.28
First observe that the given series can be expressed as
∞
P
n=1
2 · 6 · · · (4n − 2)
5 · 8 · · · (3n + 2). Let an= 2 · 6 · · · (4n − 2) 5 · 8 · · · (3n + 2). Apply the Ratio Test,
n→∞lim
an+1 an
= lim
n→∞
2 · 6 · · · (4n + 2) 5 · 8 · · · (3n + 5)
5 · 8 · · · (3n + 2)
2 · 6 · · · (4n − 2) = lim
n→∞
4n + 2 3n + 5 = 4
3 > 1 Therefore the given series is divergent by the Ratio Test.
EX.29
Since 2 · 4 · 6 · · · 2n
n! = 2n(1 · 2 · 3 · · · n)
n! = 2n, the given series is a divergent geometric series with
|r| = 2 > 1.
EX.31
Since a1 = 2 > 0 and an > 0 ⇒ an+1 = 5n + 1
4n + 3an > 0 for all n > 1, we have an > 0 for all n ≥ 1 by mathematical induction and hence we can apply the Ratio Test to this series. lim
n→∞
an+1 an
=
n→∞lim
5n + 1 4n + 3 = 5
4 > 1, so the given series is divergent by the Ratio Test.
EX.38
(a) First we observe that for all m ≥ 1,
an+m= rn+m−1an+m−1 = rn+m−1rn+m−2an+m−2 = · · · = rn+m−1· · · rn+1an+1 Since {rn} is decreasing and rn+1 < 1,
Rn = an+1+ an+2+ an+3+ an+4+ · · ·
= an+1+ rn+1an+1+ rn+2rn+1an+1+ rn+3rn+2rn+1an+1+ · · ·
= an+1(1 + rn+1+ rn+2rn+1+ rn+3rn+2rn+1+ · · · )
≤ an+1(1 + rn+1+ r2n+1+ r3n+1+ · · · )
= an+1
1 − rn+1
(b) Since {rn} is increasing, rn ≤ L for all n ≥ 1. To prove this, suppose that there exist an integer N such that rN > L. Then
rn> rN > L for all n ≥ N + 1 ⇒ lim
n→∞rn 6= L a contradiction. Thus, because L < 1
Rn = an+1(1 + rn+1+ rn+2rn+1+ rn+3rn+2rn+1+ · · · )
≤ an+1(1 + L + L2+ L3+ · · · )
= an+1 1 − L
EX.39
(a) s5 = 1
1 · 21 + 1
2 · 22 + 1
3 · 23 + 1
4 · 24 + 1
5 · 25 = 661
960 ≈ 0.6885 rn = n2n
(n + 1)2n+1 = n
2(n + 1) increases as n → ∞ and lim
n→∞rn = 1
2. By the result of Problem 38 (b), the error in using s5 is
R5 ≤ a6
1 − 1/2 = 1/(6 · 26)
1/2 = 1
192 ≈ 5.208 × 10−3 (b) 1/(n + 1)2n+1
1 − 1/2 < 0.00005 ⇒ 20000 ≤ (n + 1)2n From the following table, we know that the above inequality holds whenever n ≥ 11.
n 8 9 10 11 12
(n + 1)2n 2304 5120 11264 24576 53244 Choose n = 11. The sum of the series is approximated by s11=
11
P
n=1
1
n2n ≈ 0.693109 2
EX.43
(a) P an is absolutely convergent ⇒ P an is convergent. Since P a+n/P a−n is the sum/difference of two convergent sequences times a constant (1/2), it is convergent as well.
(b) Notice that |an| = 2a+n− an = an− 2a−n. IfP a+n orP a−n converges, thenP |an|, as the difference of two convergent sequences, converges as well, which contradicts to the fact that P an is not absolutely convergent. Therefore both P a+n and P a−n are divergent.
EX.44
Following the hint, for any real number r we will construct a rearranged sequence {ˆan} of {an} such that it sums up to r. Here we will assume that r ≥ 0. For r < 0, just interchange the role of a+n and a−n in this proof.
First, let ˆsn denote the partial sum of the sequence {ˆan} and define ˆs0 = 0.
Next, take the following steps recursively to construct {ˆan}:
(1) Let σ0 = l+0 = l−0 = 0.
(2) For m ≥ 1
Case I. If m is odd, m = 2k − 1, find l+k such that
ˆ sσm−1+
l+k−1
X
n=l+k−1+1
a+n ≤ r < ˆsσm−1+
l+k
X
n=l+k−1+1
a+n (*)
Let σm = σm−1 + (lk+− lk−1+ ).
For i = 1, 2, · · · , l+k − l+k−1, assign ˆaσm−1+i = al+
k−1+i. Case II. If m is even, m = 2k, find lk− such that
ˆ sσm−1+
l−k−1
X
n=l−k−1+1
a−n ≥ r > ˆsσm−1+
l−k
X
n=l−k−1+1
a−n (**)
Let σm = σm−1 + (lk−− lk−1− ).
For i = 1, 2, · · · , l−k − l−k−1, assign ˆaσm−1+i = al−
k−1+i.
where the zeros terms in {a+n} and {a−n} are discarded and the subscripts are relabeled.
Note that the existence of lk+ and lk− in (*) and (**) is guaranteed by the fact that both P a+n and P a−n are divergent which is proved in Problem 43 (b). The other thing to notice is that the constructed sequence {ˆan} is indeed a rearrangement of {ˆan}, that is, this provides an 1-to-1 and onto map from {an} to {ˆan} (check it yourself!).
Finally, let us check whether
∞
P
n=1
ˆ
an = r. Given σm−1 < n ≤ σm, by the construction process we have the error
|ˆsn− r| ≤ max(|ˆsσm−1 − r|, |ˆsσm − r|) ≤ max(|ˆaσm−1|, |ˆaσm|) Since P an is conditionally convergent, lim
n→∞an = 0. So lim
n→∞ˆan = 0 and hence lim
n→∞|ˆsn − r| = 0.
Therefore
∞
P
n=1
ˆ
an = r and we complete the proof.
The following figure should help you understand the construction process:
4
Figure 1: *
EX.45
(a) Suppose that P n2an converges. Then lim
n→∞n2an = 0 ⇒ lim
n→∞|n2an| = lim
n→∞n2|an| = 0. By applying the Comparison Test for the seriesP |an| with the convergent p-seriesP 1
n2 (p = 2 > 1), we have lim
n→∞
|an|
1/n2 = lim
n→∞n2|an| = 0. This implies that P |an| is convergent and hence P an is absolutely convergent, a contradiction. Thus, P n2an is divergent.
(b) Let an = (−1)n−11
n. By the Alternating Series Test, P an is conditionally convergent. But P nan=P(−1)n−1 is divergent because the limit of its partial sum does not exist.