GCP (Gauss' Circle Problem) asks for a closed form for the number of square-lattice points inside a circle, centered at the origin, of radius $r$.

Let's denote by $N(r)$ the number of these points. Then, $N(r)$ is the number of integer solutions (pairs of integers $x$ and $y$) to the inequality

$$x^2+y^2 \le r^2$$

But, what would happen if, instead of setting the center of the circle at the origin, we moved the circle $1/2$ units in the X-axis? The number of lattice points $N^*(r)$ would be the number of integer solutions to

$$(x+1/2)^2+y^2 \le r^2$$

It is easy to show that $N^*(r)$ would also be the number of solutions to

$$x^2+(y+1/2)^2 \le r^2$$

For last, let's denote by $N^{**}(r)$ the number of lattice points of a circle centered at $(1/2, 1/2)$; that is, the number of integer solutions to

$$(x+1/2)^2+(y+1/2)^2 \le r^2$$

Then, my main question is:

Is it possible to get either $N^*(r)$ or $N^{**}(r)$ if only $N(r)$ and $r$ are given?

I found this post about a generalization of this problem, but I would prefer having an exact answer to this question, rather than an approximate result involving big-O notation. It does not mind if any floor function is needed.

Thank you.

  • 1
    $\begingroup$ This seems to be a duplicate of your question math.stackexchange.com/questions/2243695/… on MSE, which has been receiving answers and comments $\endgroup$ – Yemon Choi Apr 24 '17 at 17:45
  • 1
    $\begingroup$ @Yemom Due to an error in my question, I asked there if there was any direct relationship between the counting functions without knowing $r$ and here about the case $r$ was given. This is a much more difficult problem, so I decided to post it here $\endgroup$ – user3141592 Apr 24 '17 at 17:48
  • 1
    $\begingroup$ I thought about editing the question to add that new fact, but, as I already had an answer, I decided to create this new post for not to revive the old one. $\endgroup$ – user3141592 Apr 24 '17 at 17:49
  • $\begingroup$ Since it's clearly possible to compute any of $N(r)$, $N^*(r)$, or $N^{**}(r)$ from just $r$ itself if you allow floor functions and square-root functions (just sum the number of $y$-values for a given $x$, over all possible values of $x$), maybe you could clarify exactly what sort of formula you're looking for. $\endgroup$ – Greg Martin Feb 21 '18 at 16:44
  • $\begingroup$ @GregMartin Well, it would be great to find a closed form for the relationship between these two functions and $N(r)$, but maybe those floor functions and square root can be a good point to start from. $\endgroup$ – user3141592 Feb 21 '18 at 17:18

I am skeptical, in view of:

Bleher, Pavel M.; Dyson, Freeman J., The variance of the error function in the shifted circle problem is a wild function of the shift, Commun. Math. Phys. 160, No.3, 493-505 (1994). ZBL0808.11058.

  • $\begingroup$ Thank you very much! That result is interesting (and somewhat disappointing, since it is less probable for me to find the exact answer of the question). On the other hand, I hope that there is some kind of closed form (or maybe involving the floor function) relationship between the original problen and the shifted one for the specific values $\alpha=(0,1/2)$ or $\alpha=(1/2,1/2)$. $\endgroup$ – user3141592 Apr 24 '17 at 20:31

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.