SEQUENCES AND SERIES
Section 3.7 Introduction to Infinite Series
8. Investigate the convergence or the divergence of the following sequences:
(a) ð ffiffiffiffiffiffiffiffiffiffiffiffiffi n2þ2
p Þ; (b) ð ffiffiffi
pn
=ðn2þ1ÞÞ;
(c) ð ffiffiffiffiffiffiffiffiffiffiffiffiffi n2þ1 p = ffiffiffi
pn
Þ; (d) ðsin ffiffiffi
pn Þ:
9. LetðxnÞandðynÞbe sequences of positive numbers such that limðxn=ynÞ ¼ þ1, (a) Show that if limðynÞ ¼ þ1, then limðxnÞ ¼ þ1.
(b) Show that if (xn) is bounded, then limðynÞ ¼0.
10. Show that if limðan=nÞ ¼L, whereL>0, then limðanÞ ¼ þ1.
Just as a sequence may be indexed such that its first element is notx1, but isx0, orx5or x99, we will denote the series having these numbers as their first element by the symbols
X1
n¼0
xn or X1
n¼5
xn or X1
n¼99
xn:
It should be noted that when the first term in the series isxN, then the first partial sum is denoted by sN.
Warning The reader should guard against confusing the words ‘‘sequence’’ and ‘‘series.’’
In nonmathematical language, these words are interchangeable; however, in mathematics, these words are not synonyms. Indeed, a series is a sequenceS¼(sk) obtained from a given sequenceX¼(xn) according to the special procedure given in Definition 3.7.1.
3.7.2 Examples (a) Consider the sequenceX:¼ ðrnÞ1n¼0wherer2R, which generates the geometric series:
(3) X1
n¼0
rn¼1þrþr2þ þrnþ :
We will show that if jrj<1, then this series converges to 1=ð1rÞ. (See also Example 1.2.4(f).) Indeed, ifsn:¼1þrþr2þ þrnforn0, and if we multiplysn
byrand subtract the result fromsn, we obtain (after some simplification):
snð1rÞ ¼1rnþ1: Therefore, we have
sn 1
1r¼ rnþ1 1r; from which it follows that
sn 1 1r
jrjnþ1 j1rj:
Sincejrjnþ1!0 whenjrj<1, it follows that the geometric series converges to 1=ð1rÞ when jrj<1.
(b) Consider the series generated by
ð1Þn 1n¼0; that is, the series:
(4) X1
n¼0
ð1Þn ¼ ðþ1Þ þ ð1Þ þ ðþ1Þ þ ð1Þ þ :
It is easily seen (by Mathematical Induction) thatsn¼1 ifn0 is even andsn¼0 if nis odd; therefore, the sequence of partial sums isð1;0;1;0;. . .Þ. Since this sequence is not convergent, the series (4) is divergent.
(c) Consider the series
(5) X1
n¼1
1
nðnþ1Þ¼ 1 12þ 1
23þ 1
34þ :
3.7 INTRODUCTION TO INFINITE SERIES 95
By a stroke of insight, we note that 1
kðkþ1Þ¼1
k 1
kþ1:
Hence, on adding these terms fromk¼1 tok¼nand noting the telescoping that takes place, we obtain
sn ¼1
1 1
nþ1;
whence it follows that sn!1. Therefore the series (5) converges to 1. &
We now present a very useful and simplenecessarycondition for the convergence of a series. It is far from being sufficient, however.
3.7.3 The nth Term Test If the seriesP
xn converges, thenlimðxnÞ ¼0.
Proof. By Definition 3.7.1, the convergence ofP
xn requires that limðskÞexists. Since xn ¼snsn1, then limðxnÞ ¼limðsnÞ limðsn1Þ ¼0. Q.E.D.
Since the following Cauchy Criterion is precisely a reformulation of Theorem 3.5.5, we will omit its proof.
3.7.4 Cauchy Criterion for Series The seriesP
xn converges if and only if for every e>0there exists MðeÞ 2N such that if m>nMðeÞ,then
(6) jsmsnj ¼ jxnþ1þxnþ2þ þxmj<e:
The next result, although limited in scope, is of great importance and utility.
3.7.5 Theorem Let(xn)be a sequence of nonnegative real numbers. Then the seriesP xn
converges if and only if the sequence S¼ ðskÞof partial sums is bounded. In this case, X1
n¼1
xn¼limðskÞ ¼supfsk:k2Ng:
Proof. Sincexn0, the sequenceSof partial sums is monotone increasing:
s1s2 sk :
By the Monotone Convergence Theorem 3.3.2, the sequenceS¼ ðskÞ converges if and only if it is bounded, in which case its limit equals supfskg. Q.E.D.
3.7.6 Examples (a) The geometric series (3) diverges if jrj 1.
This follows from the fact that the termsrndo not approach 0 whenjrj 1.
(b) The harmonic seriesX1
n¼1
1
n diverges.
Since the terms 1=n!0, we cannot use the nth Term Test 3.7.3 to establish this divergence. However, it was seen in Examples 3.3.3(b) and 3.5.6(c) that the sequence (sn) of partial sums is not bounded. Therefore, it follows from Theorem 3.7.5 that the harmonic series is divergent. This series is famous for the very slow growth of its partial sums (see the
discussion in Example 3.3.3(b)) and also for the variety of proofs of its divergence. Here is a proof by contradiction. If we assume the series converges to the numberS, then we have
S ¼ 1þ1 2
þ 1 3þ1
4
þ 1 5þ1
6
þ þ 1 2n1þ 1
2n
þ
> 1
2þ1 2
þ 1 4þ1
4
þ 1 6þ1
6
þ þ 1 2nþ 1
2n
þ
¼1þ1 2þ1
3þ þ1 nþ
¼S:
The contradiction S >S shows the assumption of convergence must be false and the harmonic series must diverge.
Note The harmonic series receives its musical name from the fact that the wavelengths of the overtones of a vibrating string are 1=2, 1=3, 1=4, . . . , of the string’s fundamental wavelength.
(c) The 2-seriesX1
n¼1
1
n2 is convergent.
Since the partial sums are monotone, it suffices (why?) to show that some subsequence of (sk) is bounded. If k1:¼211¼1, thensk1 ¼1. If k2:¼221¼3, then
sk2¼1 1þ 1
22þ 1 32
<1þ 2
22 ¼1þ1 2; and ifk3:¼231¼7, then we have
sk3¼sk2þ 1 42þ 1
52þ 1 62þ 1
72
<sk2þ 4
42<1þ1 2þ 1
22: By Mathematical Induction, we find that ifkj :¼2j1, then
0<skj <1þ12þ 12 2
þ þ 12 j1
:
Since the term on the right is a partial sum of a geometric series withr¼12, it is dominated by 1=112
¼2, and Theorem 3.7.5 implies that the 2-series converges.
(d) The p-seriesX1
n¼1
1
np converges whenp >1.
Since the argument is very similar to the special case considered in part (c), we will leave some of the details to the reader. As before, ifk1 :¼211¼1, then sk1¼1. If k2:¼221¼3, then since 2p<3p, we have
sk2 ¼ 1 1pþ 1
2pþ1 3p
<1þ 2
2p¼1þ 1 2p1: Further, ifk3:¼231, then (how?) it is seen that
sk3 <sk2þ 4
4p<1þ 1 2p1þ 1
4p1:
Finally, we let r:¼1=2p1; since p > 1, we have 0<r<1. Using Mathematical Induction, we show that ifkj:¼2j1, then
0<skj <1þrþr2þ þrj1< 1 1r:
Therefore, Theorem 3.7.5 implies that the p-series converges whenp>1.
3.7 INTRODUCTION TO INFINITE SERIES 97
(e) Thep-seriesX1
n¼1
1
np diverges when 0<p1.
We will use the elementary inequalitynpnwhenn2Nand 0<p1. It follows that 1
n 1
np for n2N:
Since the partial sums of the harmonic series are not bounded, this inequality shows that the partial sums of thep-series are not bounded when 0<p1. Hence thep-series diverges for these values of p.
(f) The alternating harmonic series, given by
(7) X1
n¼1
ð1Þnþ1
n ¼1
11 2þ1
3 þð1Þnþ1 n þ is convergent.
The reader should compare this series with the harmonic series in (b), which is divergent. Thus, the subtraction of some of the terms in (7) is essential if this series is to converge. Since we have
s2n¼ 1 11
2
þ 1 31
4
þ þ 1 2n1 1
2n
;
it is clear that the ‘‘even’’ subsequence ðs2nÞis increasing. Similarly, the ‘‘odd’’ subse- quence ðs2nþ1Þis decreasing since
s2nþ1¼1
1 1
21 3
1 41
5
1 2n 1
2nþ1
:
Since 0<s2n<s2nþ1=ð2nþ1Þ ¼s2nþ11, both of these subsequences are bounded below by 0 and above by 1. Therefore they are both convergent and to the same value. Thus the sequence (sn) of partial sums converges, proving that the alternating harmonic series (7) converges. (It is far from obvious that the limit of this series is equal to ln 2.) &
Comparison Tests
Our first test shows that if the terms of a nonnegative series are dominated by the corresponding terms of aconvergent series, then the first series is convergent.
3.7.7 Comparison Test Let X:¼(xn)and Y:¼(yn)be real sequences and suppose that for some K2N we have
(8) 0xnyn for nK:
(a) Then the convergence of P
yn implies the convergence ofP xn. (b) The divergence ofP
xn implies the divergence ofP yn. Proof. (a) Suppose thatP
yn converges and, givene>0, letMðeÞ 2Nbe such that if m>nMðeÞ, then
ynþ1þ þym<e:
Ifm>supfK; MðeÞg, then it follows that
0xnþ1þ þxmynþ1þ þym<e;
from which the convergence ofP
xn follows.
(b) This statement is the contrapositive of (a). Q.E.D.
Since it is sometimes difficult to establish the inequalities (8), the next result is frequently very useful.
3.7.8 Limit Comparison Test Suppose that X:¼ ðxnÞ and Y :¼ ðynÞ are strictly positive sequences and suppose that the following limit exists in R:
(9) r:¼lim xn
yn : (a) If r6¼0 thenP
xn is convergent if and only ifP
yn is convergent.
(b) If r¼0and ifP
yn is convergent, thenP
xn is convergent.
Proof. (a) It follows from (9) and Exercise 3.1.18 that there existsK2Nsuch that12r xn=yn2rfornK, whence
1 2r
yn xn ð2rÞyn for nK:
If we apply the Comparison Test 3.7.7 twice, we obtain the assertion in (a).
(b) Ifr¼0, then there existsK2N such that
0<xnyn for nK;
so that Theorem 3.7.7(a) applies. Q.E.D.
Remark The Comparison Tests 3.7.7 and 3.7.8 depend on having a stock of series that one knows to be convergent (or divergent). The reader will find that thep-series is often useful for this purpose.
3.7.9 Examples (a) The series X1
n¼1
1
n2þn converges.
It is clear that the inequality 0< 1
n2þn< 1
n2 for n2N is valid. Since the seriesP
1=n2 is convergent (by Example 3.7.6(c)), we can apply the Comparison Test 3.7.7 to obtain the convergence of the given series.
(b) The series X1
n¼1
1
n2nþ1 is convergent.
If the inequality
(10) 1
n2nþ1 1 n2
3.7 INTRODUCTION TO INFINITE SERIES 99
were true, we could argue as in (a). However, (10) isfalsefor alln2N. The reader can probably show that the inequality
0< 1
n2nþ1 2 n2
is valid for alln2N, and this inequality will work just as well. However, it might take some experimentation to think of such an inequality and then establish it.
Instead, if we takexn:¼1=ðn2nþ1Þandyn:¼1=n2;then we have xn
yn ¼ n2
n2nþ1¼ 1
1ð1=nÞ þð1=n2Þ!1:
Therefore, the convergence of the given series follows from the Limit Comparison Test 3.7.8(a).
(c) The series X1
n¼1
ffiffiffiffiffiffiffiffiffiffiffi1 nþ1
p is divergent.
This series closely resembles the series P 1= ffiffiffi
pn
, which is a p-series with p¼12; by Example 3.7.6(e), it is divergent. If we letxn:¼1= ffiffiffiffiffiffiffiffiffiffiffi
nþ1
p andyn:¼1= ffiffiffi pn
, then we have xn
yn ¼ ffiffiffin p ffiffiffiffiffiffiffiffiffiffiffi nþ1
p ¼ 1
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1þ1=n
p !1:
Therefore the Limit Comparison Test 3.7.8(a) applies.
(d) The series X1
n¼1
1
n!is convergent.
It would be possible to establish this convergence by showing (by Induction) that n2<n!forn4, whence it follows that
0< 1 n!< 1
n2 for n4:
Alternatively, if we let x:¼1=n!andyn:¼1=n2, then (whenn4) we have 0xn
yn ¼n2
n!¼ n
12 ðn1Þ< 1 n2!0:
Therefore the Limit Comparison Test 3.7.8(b) applies. (Note that this test was a bit troublesome to apply since we do not presently know the convergence of any series for which the limit ofxn=ynis really easy to determine.) &
Exercises for Section 3.7
1. LetP
anbe a given series and letP
bnbe the series in which the terms are the same and in the same order as inP
anexcept that the terms for whichan¼0 have been omitted. Show that Panconverges toAif and only ifP
bnconverges toA.
2. Show that the convergence of a series is not affected by changing afinitenumber of its terms. (Of course, the value of the sum may be changed.)
3. By using partial fractions, show that (a) X1
n¼0
1 nþ1
ð Þðnþ2Þ¼1; (b)X1
n¼0
1 aþn
ð Þðaþnþ1Þ¼1
a>0;ifa>0: (c) X1
n¼1
1
n nð þ1Þðnþ2Þ¼1 4: 4. IfP
xnandP
ynare convergent, show thatP xnþyn
ð Þis convergent.
5. Can you give an example of a convergent seriesP
xnand a divergent seriesP
ynsuch that PðxnþynÞis convergent? Explain.
6. (a) Calculate the value of P1
n¼2
2=7
ð Þn. (Note the series starts atn¼2.) (b) Calculate the value of P1
n¼1ð1=3Þ2n. (Note the series starts atn¼1.) 7. Find a formula for the series P1
n¼1r2nwhenj jr <1.
8. Let r1;r2;. . .;rn;. . . be an enumeration of the rational numbers in the interval [0,1].
(See Section 1.3.) For a givene>0, put an interval of lengthenabout thenth rational number rnforn¼1;2;3;. . .;and find the total sum of the lengths of all the intervals. Evaluate this number fore¼0:1 ande¼0:01.
9. (a) Show that the series P1
n¼1cosnis divergent.
(b) Show that the series P1
n¼1
cosn
ð Þ=n2is convergent.
10. Use an argument similar to that in Example 3.7.6(f) to show that the series P1
n¼1 1 ð Þpffiffinn is convergent.
11. IfP
anwithan>0 is convergent, then isP
a2nalways convergent? Either prove it or give a counterexample.
12. IfP
anwithan>0 is convergent, then isP ffiffiffiffiffipan
always convergent? Either prove it or give a counterexample.
13. IfP
anwithan>0 is convergent, then isP ffiffiffiffiffiffiffiffiffiffiffiffiffipananþ1
always convergent? Either prove it or give a counterexample.
14. If P
anwithan>0 is convergent, and ifbn:¼ða1þ þanÞ=nforn2N;then show that Pbnis always divergent.
15. Let P1
n¼1
a nð Þbe such thatða nð ÞÞis a decreasing sequence of strictly positive numbers. Ifs nð Þ denotes thenth partial sum, show (by grouping the terms insð Þ2n in two different ways) that
1
2ðað Þ þ1 2að Þ þ þ2 2nað Þ2n Þ sð Þ 2n að Þ þ1 2að Þ þ þ2 2n1a2n1
það Þ2n . Use these inequalities to show thatP1
n¼1
a nð Þconverges if and only ifP1
n¼1
2nað Þ2n converges. This result is often called theCauchy Condensation Test; it is very powerful.
16. Use the Cauchy Condensation Test to discuss thep-series P1
n¼1
1=np
ð Þforp>0.
17. Use the Cauchy Condensation Test to establish the divergence of the series:
(a) X 1
nlnn, (b) X 1
nðlnnÞðln lnnÞ,
(c) X 1
nðlnnÞðln lnnÞðln ln lnnÞ.
18. Show that ifc>1, then the following series are convergent:
(a) X 1
nðlnnÞc, (b) X 1
nðlnnÞðln lnnÞc.
3.7 INTRODUCTION TO INFINITE SERIES 101