Snippet portion: From Iwaniec and Kowalski's Analytic Number Theory:

If the class number $h=h(D)$ is small, then there are only few prime ideals $\bf{p}$ of degree one with small norm. Indeed, if $p=\bf{p \bar{p}}$ with $(\bf{p},\bf{\bar{p}})=1$, then $\bf{p}^h$ is a principal ideal generated by $\frac{1}{2}(m+n\sqrt{D})$ with $n \ne 0$, when $p^h = \frac{1}{4}(m^2-n^2 D) \ge \frac{|D|}{4}$.

Therefore the least prime $p_1 = p_1(D)$ with $\chi_D(p_1)=1$ satisfies $p_1 \ge {(\frac{|D|}{4})}^{1/h}$.

Hence $\chi_D(n)$ agrees with $\mu(n)$ on all squarefree numbers $n \le {(\frac{|D|}{4})}^{1/h}$ with $(n,{(\frac{|D|}{4})}^{1/h})=1$. This property is not likely to hold in long segments (because $\chi_D$ is periodic while $\mu$ is not), therefore this suggests that h is rather large.

Question portion: Although the above argument would not work in a Real quadratic field ($D > 0$ so the last inequality in the first paragraph does not hold), it seems that if we replace the class number h with h times the regulator this should work.

Any ideas on how to actually show this?


2 Answers 2


The following is not a full answer, but perhaps gives you an idea of how to approach the result.

Let us consider the claim $$ p^{hR} \ge \Big(\frac{D}4\Big) $$ for the smallest noninert prime. I first show that the inequality holds whenever $p \ge 11$. In fact, we have $h \ge 1$ and $R \ge \frac12 \log D + O(1)$: the latter claim is clear since the fundamental unit $(t+u\sqrt{D})/2$ satisfies $u \ge 1$ and $t \approx u \sqrt{D}$. Since $\log p > 2$, the claimed inequality follows.

Thus we only have to look at small primes. If $R$ is large enough, the claim holds. If $R$ is as small as possible, the following happens. If $u = 1$, the discriminant must have a very special form. One possibility is $D = 4n^2 +1$. Here the fundamental unit is $\varepsilon = 2n+\sqrt{D}$ (unless $n = 1$), so $R \approx \frac12 \log D$. On the other hand, "Davenport's Lemma" shows that the minimal nontrivial norm is $n = N((2n+1+\sqrt{D})/2)$; see

  • Ankeny, Chowla & Hasse, On the class-number of the maximal real subfield of a cyclotomic field J. Reine Angew. Math. 217, 217-220 (1965);

    a different idea for a proof of this and similar results can be found in

  • Lemmermeyer & Pethö, Simplest cubic fields, Manuscr. Math. 88, No.1, 53-58 (1995).

Observe that we only have to check the claimed inequality for $n \le 7$.

Since the result holds for $R$ large and very small, I would guess that its proof is within reach. Good luck!


This is a bit of a follow up to Franz's answer, but is also a bit different.

Dirichlet's class number formula tells us that $$hR = \frac{\mu(K)\sqrt{|D|}}{2^r(2\pi)^s}L(1,\chi)$$

A particular case of a famous result of Siegel says that for some positive constant $c$ $$L(1, \chi) > c|D|^{-1/4}$$

Combine to get that for some constant $c'$ $$hR > c'\sqrt[4]{|D|}$$

Notice that if $|D| > 2^{48}$ then $lg(|D|/4) < \sqrt[8]{|D|}$, so if $|D| > max\{2^{48},c'^{-8}\}$ then $$2^{hR} > 2^{c' \sqrt[4]{|D|}} > 2^{c'c'^{-1}\sqrt[8]{D}} > |D|/4$$ as required.

  • $\begingroup$ The $c$ of Siegel is ineffective. Instead via $L$-functions you get $2^{hR}= 2^{2L(1\,chi)\sqrt d}$, so you need $(2\log 2)L(1,\chi)\sqrt D\ge\log (D/4)$ if I am correct. But this is not known effectively, as the best bounds are from the Gross-Zagier theorem and Goldfeld's work, and would be $\log(D)^{1-\epsilon}$ on the right. $\endgroup$
    – Junkie
    May 7, 2010 at 4:16
  • $\begingroup$ True. But at least we know one exists. $\endgroup$ May 7, 2010 at 6:02

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.