The arithmetic of Fermat curves (Q1195995)

det \(p\) be an odd prime and \(F\) be the \(p\)-th Fermat curve given by \(x^ p+y^ p+z^ p=0\). Let \(\zeta\) be a \(p\)-th root of unity and \(C[p]\) the \(p\)-torsion in the ideal class group \(C\) of \(\mathbb{Q}(\zeta)\). The author proves that, if \(\text{rank}_{\mathbb{Z}/p\mathbb{Z}}C[p]<(p+5)/8\), then the number of \(\mathbb{Q}\)-rational points of \(F\) is at most \(2p-3\). ``Fermat's last theorem'' asserts that the number of \(\mathbb{Q}\)-rational points of \(F\) is at most three. Classically, there are many theorems varifying this statement under various restrictions on \((C[p]\), starting by Kummer's famous theorem that the assertion holds, if \(C[p]=0\) [\textit{E. Kummer}, ``Beweis des Fermatschen Satzes der Unmöglichkeit von \(x^ \lambda+y^ \lambda=z^ \lambda\) für eine unendliche Anzahl von Primzahlen \(\lambda\)'', Collected Papers, Vol. I, pp. 274-281 (1847)]. A result due to Eichler, which can be compared with the author's results, states that, if \(\text{rank}_{\mathbb{Z}/p\mathbb{Z}}C[p]<\sqrt p-2\), then there are no solutions to Fermat's equation, if \(p\nmid xyz\) [\textit{M. Eichler}, Acta Arith. 11, 129-131, 261 (1965; Zbl 0135.094)]. The author remarks that it seems likely that the hypotheses of Eichler's result as well as his result are satisfied for all primes \(p\). For \(p\leq 125.000\), the largest value of the rank is 5. Assuming Vandiver's conjecture, then the rank is equal to the index of irregularity, which is the number of Bernoulli numbers \(B_{_ 2k}\), with \(2\leq 2k\leq p-3\), which are divisible by \(p\). A probabilistic argument shows that this number is \(O(\log p/\log\log p)\) [ see \textit{L. C. Washington} ``Introduction to cyclotomic fields'' (1982; Zbl 0484.12001); example 6.6]. However the best known bound on the rank, as far as the author is concerned, is \(\text{rank}_{\mathbb{Z}/p\mathbb{Z}}C[p]<p/2\). Usually, arguments on ``Fermat's last theorem'' involved the method of infinite descent, which is done for the Fermat curve itself. In this paper, the author proves his result by using a descent argument on the Jacobian of \(F\) and recovering the information about the curve by applying Coleman's effective version of Chabauty's method [cf. \textit{R. F. Coleman}, Ann. Math., II. Ser. 121, 111-168 (1985; Zbl 0578.14038) and Duke Math. J. 52, 765-770 (1985; Zbl 0588.14015)]. In fact, the author works with quotients of Fermat curves. Let \(a\), \(b\) and \(c\) be integers such that \(\text{gcd}(abc,p)=1\) and \(a+b+c=0\). Let \(F_{a,b,c}\) be the complete nonsingular curve defined over \(\mathbb{Q}\) with affine equation \(y^ p=(-1)^ cx^ a(1-x)^ b\). The author proves that if \(((a^ ab^ bc^ c)^ p-a^ ab^ bc^ c)/p)\not\equiv 0\pmod p\) and \(\text{rank}_{\mathbb{Z}/p\mathbb{Z}}C[p]<(p+5)/8\), then the number of \(\mathbb{Q}\)-rational points of \(F_{a,b,c}\) is at most \(2p-3\). There is a map \(\varphi:F\to F_{a,b,c}\) defined by \((x,y)\mapsto(-x^ pz^{-p},x^ ay^ bz^ c)\), which is injective on \(F(\mathbb{Q})\). Hence the first result is obtained from the latter and the fact that for \(p\geq 5\) there exists a triple \((a,b,c)\) satisfying the above hypothesis (lemma 6).