6

I am working from "Understanding Analysis" by Abbot and the following is an exercise that works through the proof of Abel's Test. I reproduce the question and the solution. I am confused at a section of the proof towards the end. Any clarifications would be great.

Abel's Test for convergence states that if the series $\sum_{n=1}^{\infty}x_n$ converges, and if $(y_n)$ is a sequence satisfying $y_1 \ge y_2 \ge \cdots \ge 0$, then the series $\sum_{n=1}^{\infty}x_ny_n$ converges.

(a) Assume that $\sum_{n=1}^{\infty}a_n$ has partial sums that are bounded by a constant $A>0$ and assume $b_1 \ge b_2 \ge \cdots \ge 0$. Use summation by parts to show that $|\sum_{j=1}^{n} a_jb_j | \le 2Ab_1$.

Let $A>0$ be an upper bound for the partial sums $s_n$ of $\sum_{n=1}^{\infty}a_n$, hence \begin{align} |\sum_{j=1}^{n} a_jb_j | &= |s_n b_{n+1} - s_m b_{m+1} + \sum_{j=m+1}^{n} s_j(b_j-b_{j+1})|\le \\ & \le Ab_{n+1} + Ab_{m+1} + A(b_{m+1}-b_{n+1})= \\ & =2Ab_{m+1} \le 2Ab_1 \end{align}

(b) Prove Abel's Test by setting $a_n = x_{m+n}$ and $b_n = y_{m+n}$.

To show that $\sum_{n=1}^{\infty}x_ny_n$ converges, we use the Cauchy Criterion. Let $\epsilon > 0$, we need to show that there exists an $N$ such that if $n > m \ge N$, it follows that $|\sum_{j=m+1}^{n}x_jy_j|<\epsilon$. Let $a_n = x_{m+n}$ and $b_n = y_{m+n}$, then from part (a), we have \begin{align} |\sum_{j=m+1}^{n}x_jy_j| = |\sum_{j=1}^{n-m}a_jb_j| \le 2Ab_1, \end{align} where $A$ is an upper bound on the partial sums of $\sum_{n=1}^{\infty}a_n=\sum_{j=m+1}^{\infty}x_j$. Since $\sum_{n=1}^{\infty}x_n$ converges, then by the Cauchy Criterion, we can pick $N$ such that $n > m \ge N$ implies $|\sum_{j=m+1}^{n}x_j| < \frac{\epsilon}{2y_1}$.

Up until this point, I clearly understand every step of the proof, but the following is when I get confused:

Looking again at what the constant $A$ represents, it follows that if $n > m \ge N$, then \begin{align} A \le |\sum_{j=m+1}^{n}x_j| < \frac{\epsilon}{2y_1}. \end{align}

How can the above be true? $A$ is the upper bound of the partial sum of the series $\sum_{j=m+1}^{\infty}x_j$, so by definition, we should have $|\sum_{j=m+1}^{n}x_j| \le A$ for all $n > m$. So, why is the above inequality $A \le |\sum_{j=m+1}^{n}x_j|$?

SwiftMo
  • 1,047
  • 1
  • 13
  • 23
  • Don't you also need the $y_n$ bounded above by the $x_n$? – lux May 30 '17 at 14:00
  • Can you clarify the info about the y sequence? – lux May 31 '17 at 00:57
  • The exercise does not provide any further information on the $y_n$, it simply says $(y_n)$ is a sequence satisfying $y_1 \ge y_2 \ge \cdots \ge 0$. – SwiftMo May 31 '17 at 01:59
  • You're quite right: my mistake – lux May 31 '17 at 03:58
  • My edit was to add a missing "$=$" in the proof of (a) and to add "$\le$" and "$=$" at both the beginnings and ends of lines, which is now a common style that makes reading easier. – DanielWainfleet May 31 '17 at 05:31
  • The equality $|\sum_{j=1}^{n} a_jb_j | = |s_n b_{n+1} - s_m b_{m+1} + \sum_{j=m+1}^{n} s_j(b_j-b_{j+1})|$ is wrong. – Philipp Oct 14 '21 at 15:36

1 Answers1

4

In the proof of Abel's Test, $A$ is a constant relative to $n$ but not constant relative to $m$ and should be written as $A_m,$ and should not be defined as just any upper bound for $\{|\sum_{j=1}^ka_j\}_k ,$ but as $\sup_k |\sum_{j=1}^ka_j|.$ Then $A_m=\sup_k|\sum_{j=1}^k|=\sup_k|\sum_{j=1}^kx_{m+j}|,$ which is as small as we want if we choose $m$ large enough. We obtain $$|\sum_{j=m+1}^nx_jy_j|=|\sum_{j=1}^{n-m}a_jb_j|\leq 2A_mb_1=2A_my_{m+1}$$ I suppose we should also write $a_{n,m}$and $b_{n,m}$ rather than $a_n$ and $b_n$.

  • Great answer, certainly defining $A_m$ as the least upper bound totally makes sense. – SwiftMo Jun 03 '17 at 03:07
  • As I tried to do the summation part that the OP mentioned \begin{eqnarray} |\sum_{j=1}^n a_j b_j|&=&\sum_{j=1}^n[(S_j-S_{j-1})b_j]\ &=&\sum_{j=1}^nS_jb_j-\sum_{j=1}^nS_{j-1}b_j\ &=&\begin{array}{c}\left(S_1b_1+S_{2}b_{2}+\cdots +\color{purple}{S_n b_n}\right)\-\ \left(\color{purple}{S_{0}b_1}+S_{1}b_{2}+\cdots+{S_{n-1} b_n}\right)\end{array}\ &=&[S_nb_n-S_{0}b_1]+S_1 (b_1-b_{2})+\cdots+S_{n-1}(b_{n-1}-b_n)\ &=&[S_nb_n-S_{0}b_1]+\sum_{j=1}^{n-1}[S_j(b_j-b_{j+1})] \end{eqnarray}It's not similar to OP. Why $?$ I know it's too late but I can't related it.Any help will be appreciated – emonHR Jul 28 '19 at 09:19