On the proof of Stark's theorem (Q2533230): Difference between revisions

The author presents a proof that the only imaginary quadratic fields of class number one are those classically known. The proof is along the lines of \textit{K. Heegner}'s proof [Math. Z. 56, 227--253 (1952; Zbl 0049.16202)] (whose gaps were recently filled in by \textit{M. Deuring} [Invent. Math. 5, 169--179 (1968; Zbl 0155.38001)]) rather than along the lines of the first proof by \textit{H. M. Stark} [Mich. Math. J. 14, 1--27 (1967; Zbl 0148.27802)]. The fields in question are reduced to \(\mathbb Q(- p)^{1/2}\), \(p\) prime \(\equiv \pm 2\pmod 5\), by elementary consideration. From the Dedekind eta function, \(\eta(z)\) one forms \(\varphi(z) = -5^{-1/2}\eta_0\eta_1\eta_{-1}/\eta_\infty\eta_2\eta_{-2}\) and $\psi(z) = -5^{-1/2}\eta_0\eta_2eta_{-2}/\eta_\infty\eta_1\eta_{-1}\) where \(\eta_k=\eta((z + k)/5)\), \(\eta_\infty=\eta(5z)\). The 10 functions \(\varphi(z + k)\), \(\varphi(-1/(z +k))\) \((k\) modulo 5) are seen to be conjugates under the modular group and satisfy the equation \[ 0 = (\varphi - \varepsilon^3) ((\varphi - \varepsilon) (\varphi^2+\varepsilon+\varepsilon^2))^3+ 5^{-5/2} j(z)\varphi^5 \tag{9}\] for \(j(z)\) the modular function and \(\varepsilon = (1 + 5^{1/2})/2\), (likewise an equation (10) holds for \(\psi\) with \(\varepsilon^{-1}\) replacing \(\varepsilon\). On the other hand by an application of the Dirichlet-Kronecker limit formula \(\varepsilon^{m_p} = \varphi(\omega)\) for \(p\equiv 3\pmod 5\) and \(\varepsilon^{m_p} = \psi(\omega)\) for \(p\equiv 2\pmod 5\) where \(2m_p\) is the class number of \(\mathbb Q(-5p)^{1/2}\) and \(\omega = (5p + (-p)^{1/2})/2\). Now \(j(\omega) = q^3\) for some integer \(q\) (and class number one) so (9) and (10) become the diophantine equation \[ 5^{-1/2} (E - E^{-1}) (5 (\varepsilon^{\pm 1} E - (\varepsilon^{\mp 1}E^{-1}) (\varepsilon^{\pm 4}E^2 + (\varepsilon^{\mp 4}E^{-2}+1))^3 = q^3 \tag{12} \] where \(m_p=2l \pm 3\) and \(E = \varepsilon^l= \varepsilon_1^2\). Thus (12) is a diophantine equation in \(x\) and \(y\) where \(\varepsilon_1= (x + y5^{1/2})/2\) and it has only the classically known solutions corresponding to \(m_7=1\), \(m_{43}=1\), \(m_{43}=7\), \(m_{67}=9\), \(m_{163}=15\). In conclusion the author considers the connection between the diophantine equation (12) and quadratic class number as he asks whether this is "harmony" in the spirit of Leibniz and Hilbert or "thrift" in the spirit of Dedekind.