A random walk in higher arithmetic (Q1085202)

The authors apply the method of random walk on lattices to establish transcendence results. The new part of the paper concerns the proof of a special case of the Grothendieck conjecture (G.C.) on linear differential equations, here viewed as a transcendence statement for periods of solutions. Namely, they treat the case when the monodromy group is a priori known to be a subgroup of GL(m, \({\bar {\mathbb{Q}}})\), \(m= order\) of the differential equation. The idea of proof is interesting. It involves showing that a linear differential equation defined over \({\bar {\mathbb{Q}}}(x)\) can be reduced, by a rational change of the independent variable \(x'=f(x)\), to a differential equation (of higher order) over \({\bar {\mathbb{Q}}}(x')\) with singularities only at \(x'=0,1,\infty\). This permits uniformization by modular functions and relates the monodromy group to SL(2, \({\mathbb{Z}}).\) After that the proof is only sketched and promized to appear elsewhere. Since the authors do mention here their previous results on the G.C. and Lamé equation [(*) Lect. Notes Math. 1135, 52-100 (1984; Zbl 0565.14010)] we believe it useful to point out here an error in (*), that had previously escaped to us. The error is in Theorem 7.2 of (*), where it is asserted that the Lamé operator \[ L_{n,B}=D^ 2+(f'/f)D- (n(n+1)x+B)/f \] with integer \(n>0\), never satisfies the assumptions of the G.C.. On the contrary \(L_{1,0}\) with \(f(x)=4x^ 3-1\), has two solutions u,v such that: \[ uv=x,\quad u/v=((y-i)/(y+i))^{1/3},\quad y^ 2=f(x),\quad i^ 2=-1, \] and therefore has a full set of algebraic solutions. We will come back to this question elsewhere.