There is a beautiful formula to count the number of ideals $I$ in the ring of integers $\mathcal{O}_K$ of a number field $K$, given by

$$\sum_{n \leq X} a_n \sim C_K X,$$

where $a_n$ is the number of ideals in $\mathcal{O}_K$ of norm $n$ and $C_K$ is the residue at $s = 1$ of the Dedekind zeta function $\zeta_K$ of $K$. The quantity $C_K$ is used in the class number formula, see: https://en.wikipedia.org/wiki/Class_number_formula

Indeed, one can further refine the result by counting principal ideals. Let $b_n$ denote the number of principal ideals in $\mathcal{O}_K$ of norm $n$. Then

$$\displaystyle \sum_{n \leq X} b_n \sim \frac{C_K}{h_K} X,$$

where $h_K$ is the class number of $K$.

Now let $K = \mathbb{Q}(\sqrt{d})$, where $d$ is a positive fundamental discriminant. Let $\epsilon_d = u_0 + v_0 \sqrt{d}$, where $(u_0, v_0)$ is the smallest positive solution to the Pell equation $x^2 - dy^2 = \pm 4$. Then the regulator of $K$ is given by $\log \epsilon_d$, so the class number formula yields

$$\displaystyle C_K = \frac{2 h(d) \log \epsilon_d}{\sqrt{d}}.$$

Therefore, the number of principal ideals up to $X$ is given by

$$\displaystyle \sum_{n \leq X} b_n \sim \frac{2 \log \epsilon_d}{\sqrt{d}} X.$$

My question is, what about the error term? According to these notes due to Andrew Granville, one gets an error term of $O_d(X^{1/2})$. He did not specify the dependence on $d$. I believe it is not too difficult to obtain $O(\max\{u_0, v_0\} X^{1/2})$ from Granville's arguments, but this seems very large (at least, the dependence on $d$ of the error term is far worse than the dependence on $d$ of the main term). Is there a way to show that the dependence on $d$ is small, say $O(X^{1/2} \log \epsilon_d)$?


If $\zeta_K(s)$ has no zero for $\sigma \geq \rho$, you can take $O(x^{\rho+\epsilon})$.

This is proved for example in Lang's "Algebraic Number Theory" [XIII, §5, Theorem 6].

Even the simplest zero-free regions should give you a good enough estimate. Of course, under GRH you get $O(x^{1/2+\epsilon})$.

| cite | improve this answer | |
  • $\begingroup$ Is the constant in the big-$O$ absolute? Also, is there a version of Bombieri-Vinogradov that can give the above result 'on average' (instead of GRH) over a large range of $d$? $\endgroup$ – Stanley Yao Xiao Aug 7 '16 at 2:01
  • $\begingroup$ @StanleyYaoXiao I'm not sure. I doubt that you can go all the to $O(x^{1/2+\epsilon})$ with Bombieri-Vinogradov, but probably quite close. $\endgroup$ – Myshkin Aug 7 '16 at 5:01
  • $\begingroup$ Do you know who first proved this? Surely it was not Serge Lang. $\endgroup$ – KConrad Jul 8 '18 at 21:43

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.