$\newcommand{\absval}[1]{\left\lvert #1 \right\rvert}$ I am self-learning Real Analysis from Understanding Analysis by Stephen Abbot. I'd like someone to verify if my proofs/counterexamples about the below assertions are fine.

Consider each of the following propositions. Provide short proofs for those that are true and counterexamples for any that are not.

(a) If $\sum a_n$ converges absolutely, then $\sum a_n^2$ also converges absolutely.

(b) If $\sum a_n$ converges and $(b_n)$ converges, then $\sum a_n b_n$ converges.

(c) If $\sum a_n$ converges conditionally, then $\sum n^2 a_n$ diverges.

(d) If $\sum a_n$ with $a_n > 0$ is convergent, then is $\sum \sqrt{a_n}$ always convergent? Either prove it or give a counterexample.

(e) If $\sum a_n$ with $a_n > 0$ is convergent, then is $\sum \sqrt{a_n a_{n+1}}$ always convergent. Either prove it or give a counterexample.


(a) We know that, for real numbers $a,b \in \mathbf{R}$ \begin{align*} \absval{a}^2 + \absval{b}^2 \le (\absval{a} + \absval{b})^2 \end{align*}

So, we can write, \begin{align*} \absval{a_{m+1}}^2 + \ldots + \absval{a_{n}}^2 \le (\absval{a_{m+1}} + \ldots + \absval{a_{n}})^2 \end{align*}

As $\sum a_n$ is absolutely convergent, by the Cauchy criterion, given any $\epsilon > 0$, there exists $N \in \mathbf{N}$, such that \begin{align*} \absval{\absval{a_{m+1}} + \ldots + \absval{a_{n}}} < \sqrt{\epsilon} \end{align*}

for all $n > m \ge N$. Therefore,

\begin{align*} \absval{a_{m+1}}^2 + \ldots + \absval{a_{n}}^2 &\le (\absval{a_{m+1}} + \ldots + \absval{a_{n}})^2\\ &<(\sqrt{\epsilon})^2 = \epsilon \end{align*}

Thus, $\sum a_n^2$ is absolutely convergent.

(b) This proposition is false. As a counterexample, consider $a_n := \frac{(-1)^{n+1}}{\sqrt{n}}$ and $b_n := \frac{(-1)^{n+1}}{\sqrt{n}}$. $\sum a_n$ and $(b_n)$ are convergent, but $\sum a_n b_n$ is the harmonic series, which is well-known to diverge.

(c) I was not able to find any counter-examples. For example, if we define $a_n := \frac{(-1)^n}{n}$ then the series $\sum n^2 a_n$ diverges.

(d) No. As a counterexample, consider $a_n = \frac{1}{n^2}$. $\sum a_n$ is convergent, but $\sum \sqrt{a_n}$ is divergent.

(e) The geometric mean of two numbers is always less than or equal to the arithmetic mean. \begin{align*} \sqrt{a_n a_{n+1}} \le \frac{a_n + a_{n+1}}{2} \end{align*}

Both $(a_n)$ and $(a_{n+1})$ are convergent, and hence $\sum (a_n + a_{n+1})/2$ is also convergent by the algebraic limit theorem.

By the Comparison test, we must have $\sum \sqrt{a_n a_{n+1}}$ convergent.

  • $\begingroup$ what does converge conditionally means? $\endgroup$ – math Jan 1 at 17:41
  • $\begingroup$ $\sum a_n$ converges conditionally, if and only if, the series is convergent, but it is not absolutely convergent. $\endgroup$ – Quasar Jan 1 at 17:42
  • 1
    $\begingroup$ For c), if $\sum n^2a_n$ converges, then $n^2a_n\rightarrow0$; so eventually $|a_n|<1/n^2$. $\endgroup$ – David Mitra Jan 1 at 17:43
  • For a) I would have written :

Since $\sum |a_n|$ converges, there exists a rank $N$ such that $\forall n\geqslant N$, $|a_n|<1$. Then $\forall n\geqslant N, |a_n|^2 \leqslant |a_n|$. By theorem of comparison between positive terms series, we get that $\sum |a_n|^2$ converges.

  • For b), well done
  • For c), it's true

You can prove it by contraposition. If $\sum n^2 a_n$ converges then $n^2 a_n \rightarrow 0$ so there exists a rank $N$ such that $\forall n\geqslant N, n^2 |a_n| < 1$, so $\forall n\geqslant N, |a_n|<\frac{1}{n^2}$. By comparison, the serie converges absolutely.

  • For d), it's fine
  • For e), good too

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.