On Gauss factorials and their connection to the cyclotomic \(\lambda \)-invariants of imaginary quadratic fields (Q2681251)

This paper exhibits an unexpected and very interesting link between the \(\lambda\)-invariants attached to the \(p\)-cyclotomic \(\mathbb Z_p\)-extension of certain imaginary fields \(K\) on the one side, and so-called 1-exceptional primes, arising from congruence properties of Gauss factorials, on the other. The prime \(p\) is always assumed to split in \(K\), which forces \(\lambda_p(K)\ge 1\). The main results of the paper give criteria for \(\lambda_p(K) \ge 2\). Earlier results in this direction were obtained, among others, by \textit{J. W. Sands} [Acta Arith. 65, No. 3, 243--248 (1993; Zbl 0789.11061)]. As an aside we note that another paper of \textit{J. W. Sands} [Proc. Am. Math. Soc. 112, No. 3, 671--684 (1991; Zbl 0735.11056)] shows that information about these \(\lambda\)-values may have an impact on the \(\lambda\)-invariants of the other (infinitely many) \(\mathbb Z_p\)-extensions of \(K\). Gauss factorials are defined via products much like factorials, however certain factors are left out. More precisely, \(N_p!\) is the product of all positive integers \(i\le N\) not divisible by \(p\). This agrees up to a sign with the value of the \(p\)-adic Gamma function \(\Gamma_p\) at \(N+1\); one wonders slightly why this is not mentioned. There are ``Wilson-like'' results about the congruence class of Gauss factorials modulo \(p\), and it is interesting to study congruences modulo \(p^2\). Cosgrave and Dilcher call a prime \(p\) congruent to \(\pm 1\) mod \(m\) 1-exceptional for \(m\) if the \((p-1)\)st power of \(((p^2-1)/m)_p!\) is 1 modulo \(p^2\). (Note that these expressions are hard to read in the paper, as the exponent \(p-1\) is placed directly above the index \(p\) and precedes the ! sign; it should come afterwards.) Two very important special cases of the main theorem then say: The prime \(p\) is 1-exceptional for \(m=3\) iff \(\lambda_p(\mathbb Q(\sqrt{-3})) \ge 2\); and \(p\) is 1-exceptional for \(m=4\) iff \(\lambda_p(\mathbb Q(i)) \ge 2\). The author uses this result in order to produce primes \(p\) having these latter properties that had not been previously known. The proof is a very nice combination of older results and new ideas. The main connection is provided by Theorem 3.3 which states for an imaginary quadratic field \(K\) in which the prime \(p\) splits: \(\lambda_p(K)>1\) if and only if a fairly complicated expression involving Gauss factorials is congruent to 1 mod \(p^2\). (The expression is manufactured to be 1 mod \(p\) all the time.) This theorem also needs the hypothesis that \(p\) does not divide \(h_K\). This allows to find an explicit generator for a suitable power of \(\overline{\mathfrak p}\), which is one of the two ideals above \(p\) in \(K\); the main ingredient is afforded by Jacobi sums. The Gross-Koblitz formula now permits to express the Jacobi sums via Gauss factorials. The missing link between the congruence condition and the statement \(\lambda_p(K)>1\) is provided by Gold's criterion (proved back in 1974!), which is a cute application of global class field theory, and was in slightly modified form already used in the 1993 paper of Sands mentioned above. As corollaries, the author also obtains criteria for \(\lambda_p(K)>1\) that bypass Gauss factorials, only invoking Bernoulli polynomials, or Euler numbers, or Glaisher numbers. These criteria do look simpler than the initial result given in Theorem 1.1. This is a well-written paper. The reviewer cannot resist mentioning an amusing typo in the bibliography: Math.~Comput.~has morphed into Math.~Compet.~(Mathematics of Competition?). Concluding remark: The author informs me that the power \(p^r\) in Theorems 1.1, 3.3, 3.7 should consistently be replaced by \(p=p^1\); in particular, read \(p^2\) for \(p^{2r}\). A corrigendum is planned.