On lifting maps between Weyl modules: Can bad shapes be resolved by better shapes? (Q1328485)

Ever since the appearance of [\textit{K. Akin} and the first author, Adv. Math. 72, 171-210 (1988; Zbl 0681.20028)], in which the existence of finite projective resolutions of Weyl modules was proven, there has been a good deal of interest in describing such resolutions explicitly. An explicit resolution of two-rowed skew shapes was given elsewhere, but as yet no characteristic-free resolution of the general \(n\)-rowed skew shape (or even partition) has been made explicit. At one point in these investigations, it was observed that three-rowed ``bad'' shapes have a ``resolution'' in terms of ``better'' three-rowed skew shapes and that these resolutions very closely resembled in form the projective resolutions of two-rowed shapes. This observation initiated a program to resolve bad shapes in terms of better shapes. (Definitions of these terms are given in Section 2.) A positive spurt to this campaign was provided by the successful resolution of the worst bad shapes in terms of better ones. The way to proceed to the next step seemed clear. Unfortunately, despite persistent efforts to implement the next step, there always seemed to be some trouble. In characteristic zero, the use of the equivalence of the functors, \(\text{Hom}_G(L_\lambda, L_\mu \otimes X) \approx \text{Hom}_G (L_{\lambda/\mu}, X)\), helped to circumvent the problems, but it was soon clear that this equivalence is not true in every characteristic. (Here, too, is an interesting problem: we know that the functor \(\text{Hom}_G (L_\lambda, L_\mu \otimes X)\) is representable. What is it represented by?) But basically, as we show in Section 3, the obstacle lies in trying to lift a map between bad shapes to a map between associated good shapes. Since this lifting could be effected in characteristic zero, we had to find an arithmetic way to study the liftings of maps to see what was really occurring. Using the method of computing weight modules described in [loc. cit.] in connection with calculation of intertwining numbers, we produced an example to show that in general the maps we were trying to lift could not be lifted. In the final section of this paper, we give this example and show that our lifting problem is not solvable in characteristic two.