% vim: tw=50 % 31/01/2022 11AM \subsubsection*{Absolute Convergence} \begin{definition*} Take $a_n \in \CC$. If $\sum_{n = 1}^\infty |a_n|$ is convergent, then the series is called \emph{absolutely convergent}. \end{definition*} \begin{note*} Since $|a_N| \ge 0$, we can use the previous tests to check absolute convergence. This is particularly useful for $a_n \in \CC$. \end{note*} \begin{theorem} If $\sum a_n$ is absolutely convergent, then it is convergent. \end{theorem} \begin{proof} Suppose first $a_n \in \RR$. Let \[ v_N = \begin{cases} a_n & \text{if $a_n \ge 0$} \\ 0 & \text{if $a_n < 0$} \end{cases} \] \[ w_n = \begin{cases} 0 & \text{if $a_n \ge 0$} \\ -a_n & \text{if $a_n < 0$} \end{cases} \] \[ v_n = \frac{|a_n| + a_n}{2}, \qquad w_n = \frac{|a_n| - a_n}{2} \] Clearly, $v_n, w_n \ge 0$. Note $a_n = v_n - w_n$, and $|a_n| = v_n + w_n \ge v_n, w_n$. So if $\sum |a_n|$ converges, by comparison $\sum v_n$, $\sum w_n$ also converge, hence $\sum a_n$ converges. If $a_n \in \CC$, then $a_n = x_n + iy_n$. Now $|x_n|, |y_n| \le |a_n|$, so $\sum x_n$ and $\sum y_n$ are absolutely convergent, hence $\sum x_n$ and $\sum y_n$ converge. Since $a_n = x_n + iy_n$ we have that $\sum a_n$ converges as well. \end{proof} \subsubsection*{Examples} \begin{enumerate}[(1)] \item $\sum \frac{(-1)^{n + 1}}{n}$ converges but is \emph{not} absolutely convergent. \item $\sum_{n = 1}^\infty \frac{z^n}{2^n}$ for $z \in \CC$, then if $|z| < 2$ we have absolute convergence. If $|z| \ge 2$, $\left| \frac{z}{2} \right|^n \ge 1$, so $a_n$ does not tend to 0, hence the series diverges. \end{enumerate} \begin{definition*} If $\sum a_n$ converges, but $\sum |a_n|$ does not, it is said sometimes, that $\sum a_n$ is \emph{conditionally} convergent. \end{definition*} \noindent ``conditional'': because the sum to which the series converge is conditional on the order in which elements of the sequence are taken. If \emph{rearranged}, the sum is altered. \begin{example*} (Example Sheet 1, Q7) \begin{enumerate}[(i)] \item $1 - \frac{1}{2} + \frac{1}{3} - \frac{1}{4} + \cdots$. \item $1 + \frac{1}{3} - \frac{1}{2} + \frac{1}{5} + \frac{1}{7} - \frac{1}{4} + \frac{1}{9} + \cdots$. \end{enumerate} Let $s_n$ be the partial sum of (i) and $t_n$ the partial sum of (ii). Then $s_n \to s > 0$, and $t_n \to \frac{3s}{2}$. \end{example*} \noindent Rearrangement: \begin{definition*} Let $\sigma$ be a bijection of the positive integers, \[ a'_n = a_{\sigma(n)} \] is a rearrangement. \end{definition*} \begin{theorem} If $\sum a_n$ is absolutely convergent, every series consisting of the same terms in any order (i.e. a rearrangement) has the \emph{same sum}. \end{theorem} \begin{proof} We do the proof first for $a_n \in \RR$. Let $\sum a'_n$ be a rearrangement of $\sum a_n$. Let $s_n = \sum_{j = 1}^n a_j$ and $t_n = \sum_{j = 1}^n a'_j$, $s = \sum_{j = 1}^\infty a_j$. Suppose first that $a_n \ge 0$. Given $n$, we can find $q$ such that $sq$ satisfies \[ t_n \le sq \le s \] Now since $t_n$ is an increasing sequence bounded above, $t_n \to t$. Clearly $t \le s$. But by symmetry, $s \le t$, hence $t = s$. \\ If $a_n$ has any sign, consider $v_n$ and $w_n$ from theorem 1.13. Consider $\sum a'_n$, $\sum v'_n$ and $\sum w'_n$. Since $\sum |a_n|$ converges, both $\sum v_n$ and $\sum w_n$ converge. Use that $v_n, w_n \ge 0$ to deduce that $\sum v'_n = \sum v_n$ and $\sum w'_n = \sum w_n$. But $a_n = v_n - w_n$ hence $\sum a_n = \sum a'_n$. \\ For the case $a_n \in \CC$, write $a_n = x_n + iy_n$. Since $|x_n|, |y_n| \le |a_n|$, we have that $\sum x_n$ and $\sum y_n$ are absolutely convergent. By the previous case, $\sum x'_n = \sum x_n$ and $\sum y'_n = \sum y_n$ since $a'_n = x'_n + iy'_n \implies \sum a_n = \sum a'_n$. \end{proof}