On singularity properties of word maps and applications to probabilistic Waring type problems (1912.12556v2)
Abstract: We study singularity properties of word maps on semisimple algebraic groups and Lie algebras, generalizing the work of Aizenbud-Avni in the case of the commutator map. Given a word $w$ in a free Lie algebra $\mathcal{L}{r}$, it induces a word map $\varphi{w}:\mathfrak{g}{r}\rightarrow\mathfrak{g}$ for every semisimple Lie algebra $\mathfrak{g}$. Given two words $w_{1}\in\mathcal{L}{r{1}}$ and $w_{2}\in\mathcal{L}{r{2}}$, we define and study the convolution of the corresponding word maps $\varphi_{w_{1}}*\varphi_{w_{2}}:=\varphi_{w_{1}}+\varphi_{w_{2}}:\mathfrak{g}{r_{1}+r_{2}}\rightarrow\mathfrak{g}$. We show that for any word $w\in\mathcal{L}{r}$ of degree $d$, and any simple Lie algebra $\mathfrak{g}$ with $\varphi{w}(\mathfrak{g}{r})\neq0$, one obtains a flat morphism with reduced fibers of rational singularities (abbreviated an (FRS) morphism) after taking $O(d{4})$ self-convolutions of $\varphi_{w}$. We deduce that a group word map of length $\ell$ becomes (FRS) at $(e,\ldots,e)\in G{r}$ after $O(\ell{4})$ self-convolutions, for any semisimple algebraic group $G$. We furthermore bound the dimensions of the jet schemes of the fibers of Lie algebra word maps, and the fibers of group word maps in the case where $G=\mathrm{SL}{n}$. For the commutator word $\nu=[X,Y]$, we show that $\varphi{\nu}{*4}$ is (FRS) for any semisimple Lie algebra, obtaining applications in representation growth of compact $p$-adic and arithmetic groups. The singularity properties we consider, such as the (FRS) property, are intimately connected to the point count of fibers over finite rings of the form $\mathbb{Z}/p{k}\mathbb{Z}$. This allows us to relate them to properties of some natural families of random walks on finite and compact $p$-adic groups. We explore these connections, and provide applications to $p$-adic probabilistic Waring type problems.
Paper Prompts
Sign up for free to create and run prompts on this paper using GPT-5.
Collections
Sign up for free to add this paper to one or more collections.