Troesch complexes and extensions of strict polynomial functors. (Q2885455)

This paper gives a new approach to many earlier results concerning Ext groups between strict polynomial functors. Fix a field \(k\) of positive characteristic \(p\). Recall from \textit{E. M. Friedlander} and \textit{A. Suslin} [Invent. Math. 127, No. 2, 209-270 (1997; Zbl 0945.14028)] that a strict polynomial functor is a functor \(F\) from finite dimensional vector spaces to finite dimensional vector spaces, enriched with data that make that the action of \(\mathrm{GL}_n(k)\) on \(F(k^n)\) is the restriction to \(\mathrm{GL}_n(k)\) of an action of the algebraic group \(\mathrm{GL}_n\). Thus Yoneda extensions in the category of strict polynomial functors give information on the representation theory of the algebraic group \(\mathrm{GL}_n\). This connection becomes quite precise in a limit \(n\to\infty\). Examples of strict polynomial functors are (co-)Schur functors and their Frobenius twists.NEWLINENEWLINE A Troesch complex is a complicated but explicit injective resolution of a Frobenius twist \(S^{n(r)}\) of a symmetric power functor \(S^n\). The \(S^n\) are injectives in the category of strict polynomial functors, and Troesch complexes tell about the effect of Frobenius twist on Ext groups. The author shows that many earlier results can be understood in terms of Troesch complexes.NEWLINENEWLINE He also uses Troesch complexes to derive a cohomological version of the First and Second Fundamental Theorem of Invariant Theory for \(\mathrm{GL}_n\), \(n\) large.NEWLINENEWLINE Let \(r\) be a positive integer and let \(F\), \(G\) be strict polynomial functors. He constructs a ``twisting spectral sequence'' NEWLINE\[NEWLINEE_2^{s,t}(F,G,r)=\text{Ext}^s(F,G(E_r\otimes I))\Rightarrow\text{Ext}^{s+t}(F^{(r)},G^{(r)}),NEWLINE\]NEWLINE where \(I\) denotes the identity functor and \(E_r=\text{Ext}^*(I^{(r)},I^{(r)})\) is a graded vector space with \(k\) in even degree \(2i\), \(0\leq i<p^r\) and zero in other degrees. The grading on \(E_r\) induces a \(\mathbb G_m\) action on \(E_r\), hence a \(\mathbb G_m\) action on \(G(E_r\otimes I)\), hence a grading on \(G(E_r\otimes I)\) and that grading is what \(t\) refers to. In all known cases the twisting spectral sequence collapses at the second page and the author conjectures that this always happens. In any case he shows the spectral sequence provides an effective tool for computations. Many classical results are now explained by collapsing for lacunary reasons.