4.13.E: More Problems on Series of Functions
( \newcommand{\kernel}{\mathrm{null}\,}\)
Verify Note 3 and Example (\mathrm{c}) in detail.
Show that the so-called hyperharmonic series of order p,
\sum \frac{1}{n^{p}} \quad\left(p \in E^{1}\right) ,
converges iff p>1.
[Hint: If p \leq 1,
\sum_{n=1}^{\infty} \frac{1}{n^{p}} \geq \sum_{n=1}^{\infty} \frac{1}{n}=+\infty \quad(\text { Example }(\mathrm{b})) .
If p>1,
\begin{aligned} \sum_{n=1}^{\infty} \frac{1}{n^{p}} &=1+\left(\frac{1}{2^{p}}+\frac{1}{3^{p}}\right)+\left(\frac{1}{4^{p}}+\cdots+\frac{1}{7^{p}}\right)+\left(\frac{1}{8^{p}}+\cdots+\frac{1}{15^{p}}\right)+\cdots \\ & \leq 1+\left(\frac{1}{2^{p}}+\frac{1}{2^{p}}\right)+\left(\frac{1}{4^{p}}+\cdots+\frac{1}{4^{p}}\right)+\left(\frac{1}{8^{p}}+\cdots+\frac{1}{8^{p}}\right)+\cdots \\ &=\sum_{n=0}^{\infty} \frac{1}{\left(2^{p-1}\right)^{n}} . \end{aligned}
a convergent geometric series. Explain each step.]
\Rightarrow 3. Prove the refined comparison test:
(i) If two series of constants, \sum\left|a_{n}\right| and \sum\left|b_{n}\right|, are such that the sequence \left\{\left|a_{n}\right| /\left|b_{n}\right|\right\} is bounded in E^{1}, then
\sum_{n=1}^{\infty}\left|b_{n}\right|<+\infty \text { implies } \sum_{n=1}^{\infty}\left|a_{n}\right|<+\infty .
(ii) If
0<\lim _{n \rightarrow \infty} \frac{\left|a_{n}\right|}{\left|b_{n}\right|}<+\infty ,
then \sum\left|a_{n}\right| converges if and only if \sum\left|b_{n}\right| does.
What is
\lim _{n \rightarrow \infty} \frac{\left|a_{n}\right|}{\left|b_{n}\right|}=+\infty ?
[Hint: If (\forall n)|a_{n}| / |b_{n}| \leq K\), then |a_{n}| \leq K |b_{n}|.]
Test \sum a_{n} for absolute convergence in each of the following. Use Problem 3 or Theorem 2 or the indicated references.
(i) a_{n}=\frac{n+1}{\sqrt{n^{4}+1}}\left(\text { take } b_{n}=\frac{1}{n}\right);
(ii) a_{n}=\frac{\cos n}{\sqrt{n^{3}-1}}\left(\text { take } b_{n}=\frac{1}{\sqrt{n^{3}}} ; \text { use Problem } 2\right);
(iii) a_{n}=\frac{(-1)^{n}}{n^{p}}(\sqrt{n+1}-\sqrt{n}), p \in E^{1};
(iv) a_{n}=n^{5} e^{-n} (use Problem 18 of Chapter 3, §15);
(v) a_{n}=\frac{2^{n}+n}{3^{n}+1};
(vi) a_{n}=\frac{(-1)^{n}}{(\log n)^{q}} ; n \geq 2;
(vii) a_{n}=\frac{(\log n)^{q}}{n\left(n^{2}+1\right)}, q \in E^{1}.
[Hint for (vi) and (vii): From Problem 14 in §2, show that
\lim _{y \rightarrow+\infty} \frac{y}{(\log y)^{q}}=+\infty
and hence
\lim _{n \rightarrow \infty} \frac{(\log n)^{q}}{n}=0 .
Then select b_{n}.
Prove that \sum_{n=1}^{\infty} \frac{n^{n}}{n !}=+\infty.
[Hint: Show that n^{n} / n ! does not tend to 0.]
Prove that \lim _{n \rightarrow \infty} \frac{x^{n}}{n !}=0.
[Hint: Use Example (d) and Theorem 4.]
Use Theorems 3,5,6, and 7 to show that \sum\left|f_{n}\right| converges uniformly on B, provided f_{n}(x) and B are as indicated below, with 0<a<+\infty and b \in E^{1}. For parts (\text { ix })-(\text { xii }), find M_{n}=\max _{x \in B}\left|f_{n}(x)\right| and use Theorem 3 . (Calculus rules for maxima are assumed known.)
(i) \frac{x^{2 n}}{(2 n) !} ;[-a, b].
(ii) (-1)^{n+1} \frac{x^{2 n-1}}{(2 n-1) !} ;[-a, b].
(iii) \frac{x^{n}}{n^{n}} ;[-a, a].
(iv) n^{3} x^{n} ;[-a, a](a<1).
\left.\text { (v) } \frac{\sin n x}{n^{2}} ; B=E^{1} \text { (use Problem } 2\right).
(vi) e^{-n x} \sin n x ;[a,+\infty).
(vii) \frac{\cos n x}{\sqrt{n^{3}+1}} ; B=E^{1}.
(viii) a_{n} \cos n x, with \sum_{n=1}^{\infty}\left|a_{n}\right|<+\infty ; B=E^{1}.
(ix) x^{n} e^{-n x} ;[0,+\infty).
(x) x^{n} e^{n x} ;\left(-\infty, \frac{1}{2}\right].
(xi) (x \cdot \log x)^{n}, f_{n}(0)=0 ;\left[-\frac{3}{2}, \frac{3}{2}\right].
(xii) \left(\frac{\log x}{x}\right)^{n} ;[1,+\infty).
(xiii) \frac{q(q-1) \cdots(q-n+1) x^{n}}{n !}, q \in E^{1} ;\left[-\frac{1}{2}, \frac{1}{2}\right].
\Rightarrow 8. (Summation by parts.) Let f_{n}, h_{n}, and g_{n} be real or complex functions (or let f_{n} and h_{n} be scalar valued and g_{n} be vector valued). Let f_{n}= h_{n}-h_{n-1}(n \geq 2) . Verify that (\forall m>n>1)
\begin{aligned} \sum_{k=n+1}^{m} f_{k} g_{k} &=\sum_{k=n+1}^{m}\left(h_{k}-h_{k-1}\right) g_{k} \\ &=h_{m} g_{m}-h_{n} g_{n+1}-\sum_{k=n+1}^{m-1} h_{k}\left(g_{k+1}-g_{k}\right) . \end{aligned}
[Hint: Rearrange the sum.]
\Rightarrow 9. (Abel's test.) Let the f_{n}, g_{n}, and h_{n} be as in Problem 8, with h_{n}=\sum_{i=1}^{n} f_{i} . Suppose that
(i) the range space of the g_{n} is complete;
(ii) \left|g_{n}\right| \rightarrow 0 (uniformly) on a set B ; and
(iii) the partial sums h_{n}=\sum_{i=1}^{n} f_{i} are uniformly bounded on B ; i.e.,
\left(\exists K \in E^{1}\right)(\forall n) \quad\left|h_{n}\right|<K \text { on } B .
Then prove that \sum f_{k} g_{k} converges uniformly on B if \sum\left|g_{n+1}-g_{n}\right| does.
\text { (This always holds if the } g_{n} \text { are real and } g_{n} \geq g_{n+1} \text { on } B .)
[Hint: Let \varepsilon>0 . Show that
(\exists k)(\forall m>n>k) \quad \sum_{i=n+1}^{m}\left|g_{i+1}-g_{i}\right|<\varepsilon \text { and }\left|g_{n}\right|<\varepsilon \text { on } B .
Then use Problem 8 to show that
\left|\sum_{i=n+1}^{m} f_{i} g_{i}\right|<3 K \varepsilon .
Apply Theorem 3^{\prime} of §12.]
\Rightarrow \mathbf{9}^{\prime}. Prove that if \sum a_{n} is a convergent series of constants a_{n} \in E^{1} and if \left\{b_{n}\right\} is a bounded monotone sequence in E^{1}, then \sum a_{n} b_{n} converges.
[Hint: Let b_{n} \rightarrow b. Write
a_{n} b_{n}=a_{n}\left(b_{n}-b\right)+a_{n} b
\left.\text { and use Problem } 9 \text { with } f_{n}=a_{n} \text { and } g_{n}=b_{n}-b .\right]
\Rightarrow 10. Prove the Leibniz test for alternating series: If \left\{b_{n}\right\} \downarrow and b_{n} \rightarrow 0 in E^{1}, then \sum(-1)^{n} b_{n} converges, and the sum \sum_{n=1}^{\infty}(-1)^{n} b_{n} differs from s_{n}=\sum_{k=1}^{n}(-1)^{k} b_{k} by b_{n+1} at most.
\Rightarrow 11. (Dirichlet test.) Let the f_{n}, g_{n}, and h_{n} be as in Problem 8 with \sum_{n=0}^{\infty} f_{n} uniformly convergent on B to a function f, and with
h_{n}=-\sum_{i=n+1}^{\infty} f_{i} \text { on } B .
Suppose that
(i) the range space of the g_{n} is complete; and
(ii) there is K \in E^{1} such that
\left|g_{0}\right|+\sum_{n=0}^{\infty}\left|g_{n+1}-g_{n}\right|<K \text { on } B .
Show that \sum f_{n} g_{n} converges uniformly on B.
[Proof outline: We have
\left|g_{n}\right|=\left|g_{0}+\sum_{i=0}^{n-1}\left(g_{i+1}-g_{i}\right)\right| \leq\left|g_{0}\right|+\sum_{i=0}^{n-1}\left|g_{i+1}-g_{i}\right|<K \quad \text { by (ii). }
Also,
\left|h_{n}\right|=\left|\sum_{i=0}^{n} f_{i}-f\right| \rightarrow 0 \text { (uniformly) on } B
by assumption. Hence
(\forall \varepsilon>0)(\exists k)(\forall n>k) \quad\left|h_{n}\right|<\varepsilon \text { on } B .
Using Problem 8, obtain
(\forall m>n>k)\left|\sum_{i=n+i}^{m} f_{i} g_{i}\right|<2 K \varepsilon .]
Prove that if 0<p \leq 1, then \sum \frac{(-1)^{n}}{n^{p}} converges conditionally.
[Hint: Use Problems 11 and 2 .]
\Rightarrow 13. Continuing Problem 14 in §12, prove that if \sum\left|f_{n}\right| and \sum\left|g_{n}\right| converge on B (pointwise or uniformly), then so do the series
\sum\left|a f_{n}+b g_{n}\right|, \sum\left|f_{n} \pm g_{n}\right|, \text { and } \sum\left|a f_{n}\right| .
\left[\text { Hint } :\left|a f_{n}+b g_{n}\right| \leq|a|\left|f_{n}\right|+|b|\left|g_{n}\right| . \text { Use Theorem } 2 .\right]
For the rest of the section, we define
x^{+}=\max (x, 0) \text { and } x^{-}=\max (-x, 0) .
\Rightarrow 14. Given \left\{a_{n}\right\} \subset E^{*} show the following:
(i) \sum a_{n}^{+}+\sum a_{n}^{-}=\sum\left|a_{n}\right|.
(ii) If \sum a_{n}^{+}<+\infty or \sum a_{n}^{-}<+\infty, then \sum a_{n}=\sum a_{n}^{+}-\sum a_{n}^{-}.
(iii) If \sum a_{n} converges conditionally, then \sum a_{n}^{+}=+\infty=\sum a_{n}^{-}.
(iv) If \sum\left|a_{n}\right|<+\infty, then for any \left\{b_{n}\right\} \subset E^{1},
\sum\left|a_{n} \pm b_{n}\right|<+\infty \text { iff } \sum\left|b_{n}\right|<\infty ;
\text { moreover, } \sum a_{n} \pm \sum b_{n}=\sum\left(a_{n} \pm b_{n}\right) \text { if } \sum b_{n} \text { exists. }
\left.\text { [Hint: Verify that }\left|a_{n}\right|=a_{n}^{+}+a_{n}^{-} \text {and } a_{n}=a_{n}^{+}-a_{n}^{-} . \text {Use the rules of } §4 .\right]
\Rightarrow 15. (Abel's theorem.) Show that if a power series
\sum_{n=0}^{\infty} a_{n}(x-p)^{n} \quad\left(a_{n} \in E, x, p \in E^{1}\right)
converges for some x=x_{0} \neq p, it converges uniformly on \left[p, x_{0}\right] (or \left.\left[x_{0}, p\right] \text { if } x_{0}<p\right).
[Proof outline: First let p=0 and x_{0}=1 . Use Problem 11 with
f_{n}=a_{n} \text { and } g_{n}(x)=x^{n}=(x-p)^{n} .
As f_{n}=a_{n} 1^{n}=a_{n}\left(x_{0}-p\right)^{n}, the series \sum f_{n} converges by assumption. The convergence is uniform since the f_{n} are constant. Verify that if x=1, then
\sum_{k=1}^{\infty}\left|g_{k+1}-g_{k}\right|=0 ,
and if 0 \leq x<1, then
\sum_{k=0}^{\infty}\left|g_{k+1}-g_{k}\right|=\sum_{k=0}^{\infty} x^{k}|x-1|=(1-x) \sum_{k=0}^{\infty} x^{k}=1 \quad \text { (a geometric series). }
Also, \left|g_{0}(x)\right|=x^{0}=1 . Thus by Problem 11(\text { with } K=2), \sum f_{n} g_{n} converges uniformly on [0,1], proving the theorem for p=0 and x_{0}=1 . The general case reduces to this case by the substitution x-p=\left(x_{0}-p\right) y . Verify!]
Prove that if
0<\underline{\lim} a_{n} \leq \overline{\lim } a_{n}<+\infty ,
then the convergence radius of \sum a_{n}(x-p)^{n} is 1.
Show that a conditionally convergent series \sum a_{n}\left(a_{n} \in E^{1}\right) can be rearranged so as to diverge, or to converge to any prescribed sum s. [Proof for s \in E^{1} : Using Problem 14(iii), take the first partial sum
a_{1}^{+}+\cdots+a_{m}^{+}>s .
Then adjoin terms
-a_{1}^{-},-a_{2}^{-}, \ldots,-a_{n}^{-}
until the partial sum becomes less than s . Then add terms a_{k}^{+} until it exceeds s. Then adjoin terms -a_{k}^{-} until it becomes less than s, and so on.
As a_{k}^{+} \rightarrow 0 and a_{k}^{-} \rightarrow 0 (why?), the rearranged series tends to s . (Why?)
Give a similar proof for s=\pm \infty. Also, make the series oscillate, with no sum.]
Prove that if a power series \sum a_{n}(x-p)^{n} converges at some x=x_{0} \neq p, it converges absolutely (pointwise) on G_{p}(\delta) if \delta \leq\left|x_{0}-p\right|.
[Hint: By Theorem 6, \delta \leq\left|x_{0}-p\right| \leq r\left(r=\text { convergence radius). Fix any } x \in G_{p}(\delta) .\right. Show that the line \overrightarrow{p x}, when extended, contains a point x_{1} such that |x-p|< \left|x_{1}-p\right|<\delta \leq r . By Theorem 6, the series converges absolutely at x_{1}, hence at x as well, by Theorem 7.]