Kuroda's formula and arithmetic statistics (Q2069636)

A beautiful formula due to \textit{S. Kuroda} [Nagoya Math. J. 1, 1--10 (1950; Zbl 0037.16101)] from 1950 relates the class number of a multiquadratic field \(K\) with those of its quadratic subfields \(k_1, \cdots, k_n\). The unit index \(Q(K) := [O_K^{\ast}: O_{k_1}^{\ast} \cdots O_{k_n}^{\ast}]\). In the biquadratic case which corresponds to \(n=3\), the formula shows \(h(K) = \frac{Q(K)}{4} h(k_1)h(k_2)h(k_3)\) where \(h(F)\) denotes the class number of a field \(F\). Also, it is known that the unit index \(Q(K)\) takes one of the three values \(1,2,4\). In the paper under review, the authors fix a positive, square-free integer \(d>1\) and consider biquadratic fields \(K = \mathbb{Q}(\sqrt{p},\sqrt{d})\) for various primes \(p\) not dividing \(d\). They are interested in determining the three densities of primes \(p\) (if they exist) for which the unit index is one of the possibilities \(1,2,4\). Moreover, they are looking for a governing field \(M_d\) which governs the density; viz., a Galois extension \(M_d\) along with a class function \(\phi_d\) from \(\mathrm{Gal}(M_d/\mathbb{Q})\) to \(\{1,2,4\}\) such that for a prime \(p\) unramified in \(M_d\), one has \(\phi_d(\mathrm{Frob}_p) = i\) if and only if the corresponding unit index for \(K= \mathbb{Q}(\sqrt{d}, \sqrt{p})\) is \(i\) (\(i=1,2,4\)). If there is a governing field, the density can be computed directly from the Chebotarev density theorem. Assume that \(d\) has no prime factor congruent to \(3\) mod \(4\) (this is equivalent to the \(2\)-ranks of the class group and the narrow class group being equal for the field \(\mathbb{Q}(\sqrt{d})\)). Also, supposing that \(p \equiv 1\) mod \(4\), and that the \(4\)-ranks of the class group and the narrow class group of \(\mathbb{Q}(\sqrt{d})\) and of \(\mathbb{Q}(\sqrt{dp})\) are zero, the unit index can be proved to be either \(1\) or \(2\). After proving a result about when the unit index is \(2\) under the above assumptions, the authors make a general conjecture. If \(d\) has \(t\) distinct prime divisors, and \(p\) does not divide \(d\), then the number \(m_{d,p}\) of prime divisors of \(m\) which split completely in \(\mathbb{Q}(\sqrt{p})\) is among \(\{0,1, \cdots, t-1\}\). Under the \(2\)-rank and \(4\)-rank assumptions on the class group and the narrow class group of \(\mathbb{Q}(\sqrt{d})\), the conjecture asserts that for any given \(m\), the density of the set \(P_{d,m}\) of primes \(p \equiv 1\) mod \(4\) not dividing \(d\), for which \(m_{d,p}=m\) and the unit index is \(2\) equals \(\frac{1}{2^{t-1}}\). The main result proves the conjecture for \(m=t-1\) and \(m=t-2\). In addition, for \(m \leq t-3\), a partial result is proved showing that the density is at the most \(1/2^m\). The authors use Rédei matrices.