8

I want to prove that $\sum_{1\leq m^2+n^2\leq R^2}{\frac{1}{m^2+n^2}}=2\pi\log R+O(1)$ as $R\rightarrow\infty$.

For this, I'm trying to approximate the sum by using the integral $\int_{1\leq r\leq R}{\frac{1}{x^2+y^2}}dxdy=\int_{0}^{2\pi}\int_{1}^{R}{\frac{1}{r}}drd\theta=2\pi\log R$

How can I show that $\sum_{1\leq m^2+n^2\leq R^2}{\frac{1}{m^2+n^2}}=\int_{1\leq r\leq R}{\frac{1}{x^2+y^2}}dxdy$ as $R\rightarrow\infty$ rigorously?

Any help will be appreciated.

alemonk
  • 683

2 Answers2

3

$$ \begin{align} \sum_{m,n=1}^R\frac1{m^2+n^2} &=2\sum_{m=1}^R\sum_{n=1}^m\frac1{m^2+n^2}-\sum_{n=1}^R\frac1{2n^2}\\ &=2\sum_{m=1}^R\frac1m\sum_{n=1}^m\frac{\frac1m}{1+\left(\frac nm\right)^2}-\sum_{n=1}^R\frac1{2n^2}\\ \end{align} $$ Comparing the Riemann sum $\sum\limits_{n=1}^m\frac{\frac1m}{1+\left(\frac nm\right)^2}$ and the integral $\int_0^1\frac{\mathrm{d}x}{1+x^2}$, we see that $$ \sum_{n=1}^m\frac{\frac1m}{1+\left(\frac nm\right)^2}=\frac\pi4+O\left(\frac1m\right) $$ Since $\sum\limits_{m=1}^n\frac1m=\log(n)+O(1)$, and $\sum\limits_{n=1}^\infty\frac1{2n^2}=\frac{\pi^2}{12}$, we get that $$ \sum_{m,n=1}^R\frac1{m^2+n^2}=\frac\pi2\log(R)+O(1)\tag{1} $$ Furthermore, since $\log(R/\sqrt2)=\log(R)+O(1)$, $$ \sum_{m,n=1}^{R/\sqrt2}\frac1{m^2+n^2}=\frac\pi2\log(R)+O(1)\tag{2} $$ The sum over the terms where $m=0$ or $n=0$ is handled by $$ \sum_{m=1}^\infty\frac1{m^2}=\frac{\pi^2}6=O(1)\tag{3} $$ Since $$ \overbrace{\left\{1\le m,n\le R/\sqrt2\right\}}^{(2)}\subset\left\{1\le m^2+n^2\le R^2\text{ and }m,n\ge1\right\}\subset\overbrace{\left\{1\le m,n\le R\right\}}^{(1)} $$ and accounting for the four quadrants and the four borders where $m=0$ or $n=0$, we get $$ \sum_{1\le m^2+n^2\le R^2}\frac1{m^2+n^2}=2\pi\log(R)+O(1) $$

robjohn
  • 353,833
  • I think you meant to have the upper index be $R$, not $R^2$. Also, there might be some strange things going on with the notation with respect to one of the indices being $0$ or negative? – heropup Nov 22 '15 at 22:52
  • @heropup: Oh, I misread the subscript as $1\le m,n\le R^2$. I have fixed this. – robjohn Nov 22 '15 at 22:59
  • @heropup: Oh, I see that there are now 4 quadrants... I have fixed this, too. Now I think I have corrected all the problems from my misreading. – robjohn Nov 22 '15 at 23:18
1

Just an idea:

Let's use $$x-1<[x]\leq x$$ $$\sum \frac{1}{m^2+n^2}=\int \frac{1}{[x]^2+[y]^2}$$ Then $$\int \frac{1}{x^2+y^2}\leq\int \frac{1}{[x]^2+[y]^2}<\int \frac{1}{(x-1)^2+(y-1)^2}$$ And hopefully we can show $$\lim_{R\to\infty}\int \frac{1}{x^2+y^2}=\lim_{R\to\infty}\int \frac{1}{(x-1)^2+(y-1)^2}$$ ?

Kay K.
  • 10,141