Elliptic curves and lower bounds for class numbers (Q2187113)

Let \({\mathbb Q}(\sqrt{-D})\) be an imaginary quadratic field where \(-D\) is a fundamental discriminant, and let \(h(-D)\) be its class number. In the present paper the authors obtain lower bounds for \(h(-D)\) for certain values of \(D\) by using elliptic curves to construct ideal class pairings, that is, maps \[E({\mathbb Q})\times E_{-D}({\mathbb Q})\rightarrow{\text {CL}}(-D),\] where \(E/{\mathbb Q}\) is given by \[E: y^2=x^3+a_4x+a_6,\] where \(a_4,a_6\in {\mathbb Z}\), \(j(E)\) is its \(j\)-invariant, \(\Delta(E)\) is its discriminant, \(E({\mathbb Q})\) has rank \(r=r(E)\geq 1\), \(E_{-D}/{\mathbb Q}\) is its quadratic twist \[E_{-D}: -D\cdot \left (\frac y 2\right )^2=x^3+a_4x+a_6,\] and \(Q_D=(u,v)\in E_{-D}({\mathbb Q})\) is an integer point, where \(v\neq 0\), with \(v\) even if \(-D\) is odd. In more detail, let \(P=(\frac A {C^2}, \frac B {C^3})\in E({\mathbb Q})\), with \(A,B,C\in {\mathbb Z}\), and \(Q=(\frac u {w^2}, \frac v {w^3})\in E_{-D}({\mathbb Q})\) with \(u,v,w\in {\mathbb Z}\), not necessarily in lowest terms. Suppose \(v\neq 0\), with \(v\) even if \(-D\) is odd. Let \(\alpha:=|Aw^2-uC^2|\) and \(G:=\gcd(\alpha, C^6v^2)\). For \(\ell \in {\mathbb Z}\) let \[F_{P,Q}(X,Y)=\frac \alpha G \cdot X^2+\frac{2w^3B+\ell \cdot \frac \alpha G}{C^3v}\cdot XY+\frac{(2w^3B+\ell \cdot \frac \alpha G)^2+C^6v^2D}{4C^6v^2\cdot \frac \alpha G}\cdot Y^2.\] The authors prove Theorem 2.1. Assuming the notation and hypotheses above, \(F_{P,Q}(X,Y)\) is well defined (e.g. there is such an \(\ell\)) in \({\text {CL}}(-D)\). Moreover, if \((P_1,Q_1)\) and \((P_2,Q_2)\) are two such pairs for which \(F_{P_1,Q_1}(X,Y)\) and \(F_{P_2,Q_2}(X,Y)\) are \({\text {SL}}_2({\mathbb Z})\)-equivalent, then \(\frac{\alpha_1}{G_1}=\frac{\alpha_2}{G_2}\) or \(\frac{\alpha_1 \alpha_2}{G_1G_2}>D/4\). Now let \(\Omega_r:=\pi^{\frac r 2}/\Gamma(\frac r 2 +1)\), and let \(R_{\mathbb Q}(E)\) and \(E_\text{tor}({\mathbb Q})\) be the regulator and torsion subgroup, respectively. Let \(h_W(\cdot)\) and \(\widehat h(\cdot)\) be the logarithmic height and canonical height, respectively. For a basis \(\{P_1,\ldots, P_r\}\) of \(E({\mathbb Q})/E_\text{tor}({\mathbb Q})\) define its diameter by \[d(E)=\max_{\delta_i\in\{\pm1,0\}}2\widehat h \left (\sum_{i=1}^r\delta_iP_i\right ).\] Also, define \[c(E):=\frac{|E_\text{tor}({\mathbb Q})|}{\sqrt{R_{\mathbb Q}(E)}}\cdot \Omega_r,\] \[\delta(E):=\frac1 8h_W(j(E))+\frac {1}{12}h_W(\Delta(E))+\frac5 3,\] and \[T_E(D,Q_D):=\frac1 4\log\left(\frac {D}{(1+|u|)^2}\right )-\delta(E).\] Using Theorem 2.1 the authors prove Theorem 1.1. Assuming the hypotheses above, if \((1+|u|)^2\exp(4\delta(E)+d(E))<D\leq \frac{(1+|u|)^2u^2}{v^4}\), then \[h(-D)\geq \frac{c(E)}{2}\cdot \left (T_E(D,Q_D)^{\frac r 2}-r\sqrt{d(E)}\cdot T_E(D,Q_D)^{\frac{r-1}{2}}\right ).\] Now let \(t\in {\mathbb Z}\). With the aim of improving, for a certain family of discriminants, the effective lower bound obtained by \textit{J. Oesterlé} [Astérisque 121--122, 309--323 (1985; Zbl 0551.12003)], the authors define \(-D_E(t):=-4(t^3+a_4t-a_6)\) and \(Q_{D_E(t)}:=(-t,1)\), and using Theorem 2.1 prove Theorem 1.2. If \(\varepsilon >0\), then there is an effectively computable constant \(N(E,\varepsilon)<0\) such that for negative fundamental discriminants of the form \(-D_E(t)\), where \(t\in {\mathbb Z}\) and \(-D_E(t) < N(E,\varepsilon)\), we have \[h(-D_E(t))\geq \frac1 2 \left (\frac{c(E)}{\sqrt{12^r}}-\varepsilon \right )\cdot \log(D_E(t))^{\frac r 2}.\] As a final application of Theorem 2.1 the authors conclude their paper with the consideration of an infinite family of curves \[E_{a,b}: y^2=x^3-a^2x+b^2\] of discriminant \(\Delta_{a,b}:=-16(27b^4-4a^6)\). Let \(D_{a,b}(t):=4(t^3-a^2t-b^2)\), and for \(a\) and \(b\) positive integers let \[c_{a,b}^{(2)}:=\frac{\Omega_2}{12\cdot \widehat h(P_{\max}^{(2)})} \hskip .25 in \text{and} \hskip .25 in c_{a,b^3}^{(3)}:=\frac{\Omega_3}{24\sqrt{3}\cdot \widehat h(P_{\max}^{(3)})^{\frac 3 2}},\] where \(P_{\max}^{(2)}\in \{(0,b),(-a,b)\}\subset E_{a,b}(\mathbb Q)\) and \(P_{\max}^{(3)}\in \{(0,b^3),(-a,b^3), (-b^2,ab)\}\subset E_{a,b^3}(\mathbb Q)\) are chosen to have the largest canonical height. The authors prove Theorem 1.3. If \(a\) and \(b\) are positive integers, then the following are true: \begin{itemize} \item[(1)] If \(a\gg_b 1\) (resp. \(b\gg_a 1\)), then \(r(E_{a,b}(\mathbb Q))\geq 2\). Moreover, if \(\varepsilon >0\), then for sufficiently large fundamental discriminants \(-D_{a,b}(t)<0\) in absolute value we have \[h(-D_{a,b}(t))\geq (c_{a,b}^{(2)}-\varepsilon)\cdot \log(D_{a,b}(t)).\] \item[(2)] If \(a\gg_b 1\) (resp. \(b\gg_a 1\)), then \(r(E_{a,b^3}(\mathbb Q))\geq 3\). Moreover, if \(\varepsilon >0\), then for sufficiently large fundamental discriminants \(-D_{a,b^3}(t)<0\) in absolute value we have \[h(-D_{a,b^3}(t))\geq (c_{a,b^3}^{(3)}-\varepsilon)\cdot \log(D_{a,b^3}(t))^{\frac 3 2}.\] \end{itemize} The authors point out that Theorem 1.3 is effective and often improves the bound obtained in the paper cited above.