Signature ranks of units in cyclotomic extensions of abelian number fields (Q1741434)

For a totally real number field \(F\) of degree \(n\) over \(\mathbb{Q}\), one has the signature map from \(F^{\ast}\) to \(\mathbb{F}_2^n\) given by the signs of the element under the various embeddings. It is of particular interest to study the signatures of the cyclotomic units of the totally real field \(\mathbb{Q}(\zeta_m)^+\). The cyclotomic units in the case when \(m\) is a prime power are generated by \(-1\) and \(\frac{1 - \zeta_m^a}{(1- \zeta_m)\zeta_m^{(a-1)/2}}\) as \(a\) varies between \(2\) and \(\phi(m)/2 - 1\). In the general case, they are more complicated to describe explicitly and one then uses a construction of units due to Ramachandra. Ordering the elements of \(\mathrm{Gal}(\mathbb{Q}(\zeta_m)^+/\mathbb{Q})\), and ordering the \(\phi(m)/2\) generators of the group of cyclotomic units, the signatures of the group of cyclotomic units is conveniently described by a \((\phi(m)/2) \times (\phi(m)/2)\) matrix whose rows are the signatures of these generators in the \(\mathbb{F}_2\)-vector space \(\mathbb{F}_2^{\phi(m)/2}\). The rank of this matrix over \(\mathbb{F}_2\) is the rank of the group of signatures of the cyclotomic units of \(\mathbb{Q}(\zeta_m)^+\). Using this device, the authors first show that the rank of the signature group of the cyclomic units when \(m=p^n\) (\(p\) an odd prime) is at least \(\lfloor \log_2(p^n) \rfloor - 2\). An analogous result holds for \(p=2\) also but a classical result of Weber shows this rank is actually maximum possible; viz. \(2^{n-2}\). The authors point out that analogue of Weber's result about all signatures appearing does not hold for odd \(p\) in general -- for instance when \(p=29\). After studying the signatures of composites of field extensions, the authors are able to deduce the first main theorem of the paper which asserts: Suppose the positive integer \(m\) is odd or, it is divisible by \(4\). Then, the rank of the group of signatures of the group of cyclotomic units in \(\mathbb{Q}(\zeta_m)^+\) is at least \(\log_2(m) - 4 \omega(m) + 1\), where \(\omega(m)\) denotes the number of distinct prime factors of \(m\). In particular, the signature rank of the units in \(\mathbb{Q}(\zeta_m)^+\) tends to infinity with \(m\). As a consequence, they draw the interesting corollary: With the exception of \(m = 12\), no maximal real subfield of any cyclotomic field of \(m\)-th roots of unity has a fundamental system of units that are all totally positive. The heuristics concerning the ranks of cyclotomic unit groups suggest that ``most'' totally real fields have maximal unit signature ranks although these abelian extensions are not typical. In resonance with this heuristic, the authors prove that for infinitely many different families of cyclotomic fields, the units do indeed have nearly maximal signature rank. They show this by proving that the unit signature rank ``deficiency'' (that is, the difference \([F:\mathbb{Q}] -\) the rank of the group of ALL units) is bounded in finitely many composites of prime power cyclotomic towers over cyclotomic fields. In contrast to the above results, the authors prove: Suppose that there exist infinitely many cyclic cubic fields having a totally positive system of fundamental units (this is actually expected). Then, the difference between \(\phi(m)/2\) and the unit signature rank of \(\mathbb{Q}(\zeta_m)^+\) can be arbitrarily large.