On Shalev's Quantitative Version of Thompson's Conjecture

The classical Thompson's conjecture asserts that in any finite nonabelian simple group $G$, there exists a conjugacy class $S$ whose square is equal to the entire group. Shalev (Annals Math., 2009) conjectured that a much stronger quantitative version holds: there exists $ϵ>0$ such that for any conjugacy class $S$ with $|S|\ge |G{|}^{1-ϵ}$, we have $G\setminus \{1\}\subseteq S^2$. While the conjecture turns out to fail in general, we prove a close variant of the conjecture over ${\mathrm{PSL}}_n(2)$: There exist $ϵ>0$ and $N\in \mathbb{N}$ such that for all $n\ge N$, every conjugacy class $S\subseteq {\mathrm{PSL}}_n(2)$ with $|S|\ge |{\mathrm{PSL}}_n(2){|}^{1-ϵ}$ satisfies $\{x\in {\mathrm{PSL}}_n(2):\mathrm{supp}(x)\ne 1\}\subseteq S{S}^{-1}.$ On the other hand, we show that for each $ϵ>0$, there exist conjugacy classes $S\subset {\mathrm{PSL}}_n(8)$ of size $>|{\mathrm{PSL}}_n(8)){|}^{1-ϵ}$ for which $S{S}^{-1}$ misses even elements with support size linear in $n$. In addition, for every odd $n\ge 5$, we construct an inverse-closed normal set in ${\mathrm{PSL}}_n(5)$ with fixed positive density whose square misses a nonidentity element. This answers on the negative a question of Shalev (ECM, 2021) and disproves a conjecture of Pyber (Kourovka notebook, 2022).