Rigidity and semi-invariants in Drinfeld modules (Q1907841)

Let \(k\) be an algebraically closed field with \(X\) and \(Y\) two abelian varieties over \(k\). Let \(f: X\to Y\) be any morphism of \(X\) and \(Y\) as \(k\)-varieties. Then it is a well known and quite wonderful fact that one may write \(f(x)= h(x) +y\) where \(h: X\to Y\) is a morphism of abelian varieties and \(y\in Y\). In particular let \(f:X\to Y\) be a morphism which has the property that it takes the set of \(n\)-torsion points on \(X\) to the set of \(n\)-torsion points on \(Y\) for all positive integer \(n\). Then one sees immediately that \(f\) must be a morphism of abelian varieties. Let \(X\) be a curve over the finite field \(\mathbb{F}_r\) where \(r= p^n\). Let \(\infty\in X\) be a fixed closed point and let \({\mathbf A}\) be the affine ring of \(X_\infty\). Let \(L\) be an ``\({\mathbf A}\)-field'', that is \(L\) is a field over \(\mathbb{F}_r\) with an \(\mathbb{F}_r\)-algebra map \(\iota: {\mathbf A}\to L\). In this setup, one has the theory of Drinfeld modules. As is well known, Drinfeld modules are analogous to elliptic curves and abelian varieties. As such, it is reasonable to ask whether there are also rigidity theorems for them. The first and obvious result is a negative one: obviously not all polynomials (i.e., maps from \(\mathbb{G}_a\) to itself) are morphisms of Drinfeld modules. So any sort of result in this area must take into account the Drinfeld module actions themselves and not just underlying variety \((\mathbb{G}_a)\). In the paper being reviewed, the author establishes the following theorems. Theorem 1. Let \(L\) be as above. Let \(\varphi\) and \(\psi\) be two Drinfeld modules over \(L\) (not necessarily of the same rank). Let \(g(x)\) be a polynomial with coefficients in \(L\) with the property that it maps \(\varphi [a]\) (i.e., the \(a\)-division points of \(\varphi\), etc.) into \(\psi [a]\) (as sets) for all \(0\neq a\in {\mathbf A}\). Then \(g\in \Hom_L (\varphi, \psi)\). Theorem 2. Let \(\varphi\) be a Drinfeld module over \(L\) where we now assume that \(L\) has generic characteristic (i.e., \(\iota: {\mathbf A}\to L\) is injective). Suppose that \(g(x)\in L[x]\) and at least one of the following hold: 1. There is an infinite \({\mathbf A}\)-submodule of the torsion module associated to \(\varphi\) such that \(g\) maps \(S\) to itself (as a set). 2. There are infinitely many \(a\in {\mathbf A}\) for which \(g\) maps \(\varphi [a]\) to itself (as sets). Then \(g(x)= j(x)+ c\) where \(j\in \text{End}_L (\varphi)\) and \(c\in L\) is a torsion point. A homomorphism version of Theorem 2 is also established predicated on a very reasonable conjecture of ``Serre-type'' on the image of the absolute Galois group on Tate modules. Finally, the reader can clearly see the analogy between Theorem 1 and the corresponding abelian variety result presented above.