No label defined (Q4253162)

Let \(k\) be an algebraic number field and \({\mathbb A}_k\) the ring of adeles of \(k\). Denote by \(M_k\) the set of places of \(k\). For \(v\in M_k\) choose an absolute value \(|\cdot |_v\) such that \(|\cdot |_v\) extends the standard absolute value on \({\mathbb Q}\) if \(v\) is archimedean, and the \(p\)-adic absolute value if \(v\) lies above the prime \(p\). Then if \(l=[k:{\mathbb Q}]\) and \(l_v=[k_v:{\mathbb Q}_p]\) with \({\mathbb Q}_p={\mathbb R}\) if \(v\) is archimedean, define the normalized absolute value \(||\cdot ||_v=|\cdot |_v^{l_v/l}\). Finally, for \({\mathbf x}=(x_1 ,\ldots , x_m)\in k_v^m\) define the local height \(H_v({\mathbf x})=(\sum_{i=1}^m |x_i|_v^2)^{1/2l}\) if \(v\) is archimedean and \(H_v({\mathbf x})=(\max_{i=1}^m |x_i|_v)^{1/l}\) if \(v\) is non-archimedean. Given a matrix \(g\in GL(m,{\mathbb A}_k)\) define the twisted height of \({\mathbf x}=(x_1,\ldots ,x_m)\in k^m\) by \(H_g({\mathbf x})=\prod_{v\in M_k}H_v( g_v{\mathbf x})\), where \(g_v\) is the local component of \(g\) at \(v\). We write \(H\) for \(H_g\) if \(g\) is the identity. Define \(||\text{ det } g||_{\mathbb A}=\prod_{v\in M_k}||\text{ det } g_v||_v\). Denote by \(Gr (m,d,k)\) the collection of \(d\)-dimensional linear subspaces of \(k^m\). Given \({\mathcal A}\in Gr(m,d,k)\), define \(H_g({\mathcal A})=\prod_{v\in M_k} (g_v{\mathbf a}_1\wedge\cdots \wedge g_v{\mathbf a}_d)\) where \({\mathbf a}_1,\ldots ,{\mathbf a}_d\) is a basis of \({\mathcal A}\); this is independent of the choice of the basis. J. Thunder introduced the generalized Hermite constant \[ \gamma_{m,d}(k)=\max_{g\in GL(m,{\mathbb A}_k)}\min_{{\mathcal A}\in Gr(m,d,k)} H_g({\mathcal A})^2/||\text{ det} g||_{{\mathbb A}}^{d/m} \] and gave explicit upper and lower bounds for this quantity. Now let \(F=\sum_{i=1}^m\sum_{j=1}^m f_{ij}X_iX_j\) with \((f_{ij})\) a symmetric non-singular matrix with entries in \(k\) be a bilinear form. Let \(H(F)\) be the height of the coefficient vector \((f_{ij}:1\leq i,j\leq m)\). Let \(n\) be the largest integer for which \(k^m\) has an \(n\)-dimensional linear totally isotropic subspace \(U\), i.e. with \(F(U,U)=0\). Let \(O_F (k)\) be the orthogonal group of \(F\), i.e. the group of transformations \(g\in GL(m,k)\) with \(F(g{\mathbf x},g{\mathbf y})=F({\mathbf x},{\mathbf y})\) for \({\mathbf x},{\mathbf y}\in k^n\) and \(O_F({\mathbb A}_k)=O_F(k)\otimes_k {\mathbb A}_k\). The main result of the author is as follows: Let \(1\leq d\leq n\). Then for each \(g\in O_F({\mathbb A}_k)\) there is a \(d\)-dimensional totally isotropic subspace \({\mathcal X}_g\) of \(k^m\) such that \[ H_g({\mathcal X}_g)^2\leq \gamma_{m-d,1}(k)^{m-d}(2H(F))^{m-d}. \] This is an abstraction and slight improvement of a result by \textit{J. Vaaler} [Trans. Am. Math. Soc. 302, 281-296 (1987; Zbl 0617.10015)], which in turn was a generalization of \textit{H. P. Schlickewei}'s corresponding result over \({\mathbb Q}\) [Monatsh. Math. 100, 35-45 (1985; Zbl 0564.10012)].