Find $$\lim_{n\to\infty}\sum_{m=1}^{\infty}\frac{1}{m^2 + n^2}$$

I have the following theorems/definitions:

Theorem: (Dominated convergence theorem (DCT)). Let $(f_n)_{n\geq 1}$ be a sequence of functions such that $\lim\limits_{n\to\infty}f_n =f$ pointwise. If there exists an integrable function $g: S\to[0,\infty]$ such that $\left|f_n\right|\leq g$ for all $n\geq 1$, then $f$ is integrable and $$\lim\limits_{n\to\infty}\int_Sf_n d\mu = \int_Sf d\mu.$$

Def: Let $S = \mathbb{N}$ and $\mathcal{A} = \mathcal{P}(\mathbb{N})$. We write $\#A$ for the number of elements in a finite set $A$, and we set $\#A = \infty$ if $A$ is infinite. Let $\mu:\mathcal{A}\to[0,\infty]$ be given by $\mu(A) = \#A$. Then $\mu$ is a measure. Often $\mu$ is denoted by $\tau$ and is called the counting measure.

Exercise: Find $\lim\limits_{n\to\infty}\sum\limits_{m=1}^{\infty}\dfrac{1}{m^2 + n^2}$. Hint: use the counting measure and the DCT.

If I'm not mistaken then $\lim\limits_{n\to\infty}\dfrac{1}{m^2 + n^2} =0$ pointwise. That means that if I can think of an integrable function $g$ such that $\left|\dfrac{1}{m^2 + n^2}\right|\leq g$ for all $n\geq 1$, then $\lim\limits_{n\to\infty}\sum\limits_{m=1}^{\infty}\dfrac{1}{m^2 + n^2} = \sum\limits_{m=1}^{\infty}0\,d\mu$, by the DCT theorem. However I can't find such a function $g$ and I'm not sure how I can use the counting measure doing so..

Question: How do I find $\lim\limits_{n\to\infty}\sum\limits_{m=1}^{\infty}\dfrac{1}{m^2 + n^2}$ using the counting measure and the DCT?

Thanks in advance!

Edit: With some help from the commentsection I think I might have found the solution:

We have $\lim\limits_{n\to\infty}f_n = 0$ pointwise and $\left|f_n\right|\leq g$, $g$ integrable. Hence $\lim\limits_{n\to\infty}\int_S f_n d\mu = \int_S f d\mu$. If $\mu$ is the counting measure then $$\lim\limits_{n\to\infty}\sum\limits_{m=1}^\infty\dfrac{1}{m^2 + n^2} = \lim\limits_{n\to\infty}\int\limits_{m\in\mathbb{N}} f_n d\mu$$ Hence $\lim\limits_{n\to\infty}\sum\limits_{m=1}^\infty\dfrac{1}{m^2 + n^2}= \sum\limits_{m=1}^\infty 0 \,d\mu = 0$.

Is this correct?

  • 4
    $\begingroup$ Try: $g(m) = \frac{1}{m^2}$ $\endgroup$ – Mohammad W. Alomari Dec 22 '17 at 11:42
  • $\begingroup$ You can also use formula 6.3.13 people.math.sfu.ca/~cbm/aands/page_259.htm so it is $\sim\frac{\pi}{2n}\to 0$. $\endgroup$ – zwim Dec 22 '17 at 11:54
  • $\begingroup$ Just a comment : $\sum_{m \ge 1} \frac{1}{m^2+n^2} \le \sum_{m = 1}^N \frac{1}{m^2+n^2} +\sum_{m \ge N+1} \frac{1}{m^2} $. You can choose $N$ such that $\sum_{m \ge N+1} \frac{1}{m^2} < \epsilon$, then choose $n$ such that $\sum_{m = 1}^N \frac{1}{m^2+n^2} < \epsilon$. $\endgroup$ – reuns Dec 22 '17 at 11:55
  • $\begingroup$ @mwomath I thought of that, but I'm confused as to whether or not $g$ is integrable in that case, as $\frac{1}{m^2}$ is not a continuous function if $m\in\{1,2,3,\ldots\}$ right? $\endgroup$ – titusAdam Dec 22 '17 at 12:00
  • $\begingroup$ You need Lebesgue integrable function not continuous. Follow 'reuns' hint $\endgroup$ – Mohammad W. Alomari Dec 22 '17 at 12:05

You didn't (explicitely) check yet that the $g$, $f_n$ are measurable and that $g$ is integrable.

For the complete proof:

As mwomath stated, use $g(m)=\frac{1}{m^2}$. Both $f_n$ for all $n\geq 1$ and $g$ are measurable, because $\mathcal{A} = \mathcal{P}(\mathbb{N})$ (for any Borel-measurable subset $B$ we trivially have that $g^{-1}(B)\in \mathcal{P}(\mathbb{N})$ and $f_n^{-1}(B)\in \mathcal{P}(\mathbb{N})$).

Now we have $\int\limits_{m\in\mathbb{N}} g\ d\mu=\sum\limits_{m=1}^\infty\dfrac{1}{m^2}=\frac{\pi^2}{6}<\infty$, so $g$ is integrable and obviously $|f_n|=\frac{1}{m^2+n^2}\leq \frac{1}{m^2}=g$ for all $n\geq 1$.

Indeed $\lim\limits_{n\to\infty}f_n = 0$ pointwise, so by the DCT you get $$\lim\limits_{n\to\infty}\sum\limits_{m=1}^\infty\dfrac{1}{m^2 + n^2} = \lim\limits_{n\to\infty}\int\limits_{m\in\mathbb{N}} f_n d\mu=\int\limits_{m\in\mathbb{N}} \lim\limits_{n\to\infty}f_n d\mu=\int\limits_{m\in\mathbb{N}} 0 d\mu=0$$


Dominating Function $$ \frac1{m^2+n^2}\le\frac1{m^2} $$ and $$ \begin{align} \sum_{m=1}^\infty\frac1{m^2} &\le1+\sum_{m=2}^\infty\frac1{m(m-1)}\\ &=2 \end{align} $$

Using Riemann Sums $$ \begin{align} \lim_{n\to\infty}\sum_{m=1}^\infty\frac1{m^2+n^2} &=\lim_{n\to\infty}\frac1n\lim_{n\to\infty}\frac1n\sum_{m=1}^\infty\frac1{\frac{m^2}{n^2}+1}\\ &=\lim_{n\to\infty}\frac1n\int_0^\infty\frac1{1+x^2}\,\mathrm{d}x\\[6pt] &=\lim_{n\to\infty}\frac\pi{2n}\\[9pt] &=0 \end{align} $$

Exact Sum and Asymptotic Approximation $$ \begin{align} \sum_{m=1}^\infty\frac1{m^2+n^2} &=-\frac1{2in}\sum_{m=1}^\infty\left(\frac1{m+in}+\frac1{-m+in}\right)\tag1\\ &=-\frac1{2n^2}-\frac1{2in}\sum_{m\in\mathbb{Z}}\frac1{m+in}\tag2\\ &=-\frac1{2n^2}-\frac1{2in}\pi\cot(\pi in)\tag3\\ &=-\frac1{2n^2}+\frac1{2n}\pi\coth(\pi n)\tag4\\ &=\frac\pi{2n}-\frac1{2n^2}+O\!\left(\frac1{ne^{2n}}\right)\tag5 \end{align} $$ Explanation:
$(1)$: partial fractions
$(2)$: rewrite the sum as a principal value sum
$(3)$: equation $(7)$ from this answer
$(4)$: $\cot(iz)=-i\coth(z)$ [Exact Sum]
$(5)$: $\coth(x)=1+O\!\left(e^{-2x}\right)$ as $x\to\infty$ [Asymptotic Approximation]

  • $\begingroup$ @user296113: it is the product of two limits. One is the Riemann Sum approximation of $\int_0^\infty\frac{\mathrm{d}x}{1+x^2}=\frac\pi2$ and the other is $\lim\limits_{n\to\infty}\frac1n=0$. How is that not rigorous? $\endgroup$ – robjohn Dec 23 '17 at 4:53
  • $\begingroup$ Sorry, I see now it's ok and infact all this work deserve at least +1. $\endgroup$ – user296113 Dec 23 '17 at 8:37

Because $(m+n)^2 \le 2(m^2 + n^2),$

$$\sum_{m=1}^{\infty}\frac{1}{2(m^2+n^2)} \le \sum_{m=1}^{\infty}\frac{1}{(m+n)^2} = \sum_{m=n+1}^{\infty}\frac{1}{m^2}.$$

Since $\sum_m 1/m^2<\infty,$ the last sum $\to 0$ as $n\to \infty,$ proving the limit in question is $0.$


You can use the following identity $$ \frac{1}{m^2+n^2}=\int_{0}^{+\infty}\frac{\sin(m x)}{m}e^{-nx}\,dx $$ (which is a simple consequence of integration by parts) and recall that $$ \sum_{m\geq 1}\frac{\sin(m x)}{m} $$ is the Fourier series of a sawtooth wave, in particular a bounded function $w(x)$ (each partial sum is uniformly bounded between $-\text{Si}(\pi)$ and $\text{Si}(\pi)$). By the DCT it follows that $$ \lim_{n\to+\infty}\sum_{m\geq 1}\frac{1}{m^2+n^2}=\lim_{n\to +\infty}\underbrace{\int_{0}^{+\infty}w(x) e^{-nx}\,dx}_{O\left(\frac{1}{n}\right)}=0.$$


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.