Drinfeld-Stuhler modules (Q2319810)

This article introduces Drinfeld-Stuhler modules, develops their basic theory and clarifies their relationship to previously defined categories. Let \(F\) be the function field of a smooth and geometrically irreducible projective curve \(C\) over a finite field \(\mathbb{F}_q\). We fix a closed point \(\infty\in C\) and let \(\mathbb{C}_\infty:=\overline{F}_\infty\). The open affine \(C-\{\infty\}\) is the spectrum of a Dedeking ring \(A\). Let \(L\) be a field that is an \(A\)-algebra by virtue of a morphism \(\gamma:A\to L\). As a reminder and motivator, a Drinfeld \(A\)-module over \(L\) is an injective ring homomorphism \[ \phi:A\to \mathrm{End}(\mathbb{G}_{a,L})\cong L[\tau],\, a\to\phi_a \] (where \(L[\tau]\) is the ring of skew polynomials over \(L\)) such that \(\partial \phi_a\), the constant term of \(\phi_a\), equals \(\gamma(a)\) for all \(a\in A\). There is a notion of rank so that \(\ker\phi_a\cong(A/a)^{\mathrm{rank}\phi}\) for all \(a\) coprime to \(\ker\gamma\). The category of Drinfeld modules of rank \(r\) over \(\mathbb{C}_ \infty\) is equivalent to the category of lattices \(\Lambda\subset\mathbb{C}_ \infty\) of rank \(r\). One can thus draw an analogy with characteristic \(0\), where one thinks of Drinfeld modules of rank \(0\) as \(\mathbb{G}_a\), of rank \(1\) as \(\mathbb{G}_m\) or CM elliptic curves, and of rank \(2\) as elliptic curves (while for rank \(>2\) there is no characteristic \(0\) analogue). These are, in a sense, ``one-dimensional'' objects. A function field analogue of higher-dimensional abelian varieties was introduced by Stuhler in the form of \(\mathcal{D}\)-elliptic sheaves [\textit{G. Laumon} et al., Invent. Math. 113, No. 2, 217--338 (1993; Zbl 0809.11032)]. These are generalisations of Drinfeld's elliptic sheaves, which correspond to Drinfeld modules under the shtuka dictionary. The question then is what sort of higher-dimensional modules should correspond to \(\mathcal{D}\)-elliptic sheaves. While this concept has appeared implicitly in the previous literature, the article fills this gap with the definition of Drinfeld-Stuhler modules. The advantage of these modules is that they are more elementary and easier to work with for readers who are not acquainted with the language of elliptic sheaves. Let \(D\) be a central simple algebra over \(F\) of dimension \(d^2\) which splits at \(\infty\). Let \(\mathcal{O}_D\) be a maximal \(A\)-order of \(D\). A \textit{Drinfeld--Stuhler \(\mathcal{O}_D\)-module} over \(L\) is an injective ring homomorphism \[ \phi:\mathcal{O}_D\to\mathrm{End}(\mathbb{G}_{a,L}^d)\cong M_d(L[\tau]),\, a\mapsto \phi_a \] such that (i) for all \(a\in \mathcal{O}_D\cap D^\times\), \(\phi_a\) is surjective and \(\#\ker\phi_a=\#(\mathcal{O}_D/a)\); (ii) the composition \(A\to\mathcal{O}_D\xrightarrow{\phi} M_d(L[\tau])\xrightarrow{\partial} M_d(L)\) maps \(a\in A\) to \(\gamma(a)\mathrm{I}_d\). With a precise notion of rank, condition (i) can be restated as demanding that \(\phi\) has rank \(1\), although one could also extend the definition to Drinfeld-Stuhler modules of higher rank. One important difference to Drinfeld \(A\)-modules exploited in the proof of the local Langlands conjectures is that the moduli of Drinfeld-Stuhler modules are already projective when \(D\) is non-split. After establishing basic properties in analogy to Drinfeld modules, the author proves a Morita equivalence between Drinfeld-Stuhler \(M_d(A)\)-modules and Drinfeld \(A\)-modules of rank \(d\). Hereafter, it is shown that the category of Drinfeld-Stuhler \(\mathcal{O}_D\)-modules is equivalent to the categories of \(\mathcal{D}\)-elliptic sheaves (modulo an action), Anderson \(\mathcal{O}_D\)-motives and, over \(\mathbb{C}_\infty\), rank \(1\) \(\mathcal{O}_D\)-lattices \(\Lambda\subset\mathbb{C}_\infty\). Using these equivalences, the theories of complex multiplication and supersingularity are developed with results similar to those for Drinfeld modules. The last section considers fields of moduli of a Drinfeld-Stuhler module \(\phi\) over \(L^{\mathrm{sep}}\), i.e.\ fields \(L\) such that the action of \(\mathrm{Gal}(L^{\mathrm{sep}}/L)\) fixes the isomorphism class of \(\phi\). In analogy with results by Shimura and Jordan for abelian varieties, it gives two sufficient conditions for a field of moduli to be a field of definition, and shows that, as for elliptic curve, fields of moduli of Drinfeld \(A\)-modules are always fields of definition. Whether the same could be true for all Drinfeld-Stuhler modules is raised as an open question.