First degree cohomology of Specht modules and extensions of symmetric powers (Q1721957)

Consider a finite symmetric group $\Sigma_d$ over an algebraically closed field $K$ of characteristic $p > 0$. For a partition $\lambda$ of $d$, let $\mathrm{Sp}(\lambda)$ denote the associated Specht module. The goal of the paper is understanding the first cohomology group $H^1(\Sigma_d,\mathrm{Sp}(\lambda))$. The strategy is to exploit known relationships between the representation theory of symmetric groups and that of general linear groups to ultimately reframe the problem to one involving extensions for a Borel subgroup. In particular, suppose that $\lambda$ is a partition of $d$ with $n$-parts. For $N \geq n$, consider the general linear group $\mathrm{GL}_N(K)$ and the Borel subgroup $B$ (implicitly depending on $N$) of lower triangular matrices. The technical focus of the paper is on computing $\mathrm{Ext}^1_{B}(S^dE,K_{\lambda})$, where $S^dE$ is the $d$th symmetric power of the natural module $E$ for $\mathrm{GL}_N(K)$ and $K_{\lambda}$ denotes the one-dimensional $B$-module associated to $\lambda$. Precise descriptions of all $\lambda$ for which this $\mathrm{Ext}$-group is non-zero are given. A key idea here is making use of an equivalence of categories (that holds in a more general setting) between $B$-modules and $(\mathrm{hy}(U),T)$-modules, where $U$ denotes the unipotent radical of $B$, $\mathrm{hy}(U)$ denotes the hyperalgebra of $U$, and $T$ denotes the torus of diagonal matrices in $\mathrm{GL}_N(K)$. This allows the authors to make explicit use of basis elements in the hyperalgebra. The nature of the results differ depending on whether $\lambda$ is a James partition or not. A James partition is one of at least length two satisfying a certain vanishing condition on binomial coefficients. For example, a two-part partition $\lambda = (a,b)$ is James if and only if $\binom{a+i}{i} = 0 \pmod p$ for $1 \leq i \leq b$. For a James partition $\lambda$ of length $n$, the dimension of $\mathrm{Ext}^1_{B}(S^dE,K_{\lambda})$ is bounded by $n-1$ and is described precisely in terms of the notion of ``$p$-segments'' of $\lambda$. In the non-James case, the dimension is at most 1 and the answer differs for $p = 2$ vs. $p > 2$. For $p \geq 3$, the above $\mathrm{Ext}$-computations over $B$ give the desired computation of $H^1(\Sigma_d,\mathrm{Sp}(\lambda))$. For $p = 2$, one does not have complete information, as we only know that $\dim H^1(\Sigma_d,\mathrm{Sp}(\lambda)) \geq \dim\mathrm{Ext}^1_{B}(S^dE,K_{\lambda})$. The authors use their results to address several questions posed by \textit{D. Hemmer} [J. Algebra 322, No. 5, 1498--1515 (2009; Zbl 1187.20062)] on the vanishing/non-vanishing of $H^1(\Sigma_d,\mathrm{Sp}(\lambda))$. The authors also observe that their methods can be used to recover James' result that $H^0(\Sigma_d,\mathrm{Sp}(\lambda))$ is non-zero (and equals $K$) if and only if $\lambda$ is a James partition. Further, in the process of developing their techniques, the authors recover a number of other results, including Verma's conjecture that there is an equivalence of categories between modules for an algebraic group and its hyperalgebra; a description of extensions between standard induced modules for the special linear group $\mathrm{SL}_2(K)$, first given by \textit{K. Erdmann} [Math. Z. 218, No. 3, 447--459 (1995; Zbl 0824.20039)]; and a result of \textit{C. Weber} [J. Algebra 392, 23--41 (2013; Zbl 1292.20061)] on the vanishing of $H^1(\Sigma_d,\text{Sp}(\lambda))$ for certain $\lambda$.