I try to separate the following double sum

$$S=\sum_{i,j \in[-n,n], (i,j)\ne (0,0)} f_{ij}$$



$$S=\sum _{k=1}^n \sum _{j=1}^n (f(-k,-j)+f(-k,j)+f(k,-j)+f(k,j))$$

but I do not know why it does not work for several value of integers.

Could you help to get a formula for n dimensional sum


  • $\begingroup$ I try sum lattice sums and I need a general n dimension formula ,it is a way of begining ,Thanks @Jean Marie $\endgroup$ May 16, 2021 at 8:10
  • $\begingroup$ Your title is misleading: as I understand, you look for a closed-form formula... $\endgroup$
    – Jean Marie
    May 16, 2021 at 15:32
  • $\begingroup$ Yes pleases , in dimension two or three is will be ok $\endgroup$ May 16, 2021 at 16:29
  • $\begingroup$ What do you mean by "change of indices", by "separate" ? Explain with detailed sentences what you want to do. $\endgroup$
    – Jean Marie
    May 16, 2021 at 20:43
  • 1
    $\begingroup$ Your last expression is not the right one because you exclude for example indices (k=1,j=0) which should be included. $\endgroup$
    – Jean Marie
    May 17, 2021 at 8:09

1 Answer 1


Here is a numerical simulation giving the values of $S$ from $n=1$ to $n=100$.

A convergent behavior takes place with oscillations around a limit $\approx 1.164$.

enter image description here

but I am unable to find its exact value.

Nevertheless, this kind of sum has been studied: it is $b_2(1)$ with $b_2(s)$ given by formula (3), Problem n° 3 in this interesting [document] (https://www.davidhbailey.com//dhbpapers/tenproblems.pdf)

See as well here.

Both documents mention this formula:


which unfortunately is valid only for $s>1$.

Somewhat related: Infinite sum involving number of solutions to $k=i^2+j^2$

  • $\begingroup$ thanks @Jean Marie for the try maybe I express myself really bad what I want , I can calculation de numerical result but this is not the question, the double sum goes fron - Infinity t o infnity I need it is to get a representation of this sum for 0 to infnity for other reasons $\endgroup$ May 17, 2021 at 12:34
  • $\begingroup$ Could you say what you call "to get a representation" ? Is it a "closed-form formula" (see what this expression mean) ? And what do you mean about "n" dimensional sum in your question ? $\endgroup$
    – Jean Marie
    May 17, 2021 at 12:36
  • $\begingroup$ in Math languaje $$\sum _{k=-n}^n \sum _{j=-n}^n \to \sum _{k=1}^n \sum _{j=1}^n$$ $\endgroup$ May 17, 2021 at 18:22
  • $\begingroup$ You had asked a question previously a question on Madelung constant : you should link your questions in order to help people to help you. $\endgroup$
    – Jean Marie
    May 20, 2021 at 20:53

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.