Sarnak's conjecture in quantum computing, cyclotomic unitary group coranks, and Shimura curves

Abstract

Sarnak's conjecture in quantum computing concerns when the groups $\operatorname{PU}_2$ and $\operatorname{PSU}_2$ over cyclotomic rings $\mathbb{Z}[\zeta_n, 1/2]$ with $\zeta_n=e^{2\pi i/n}$, $4|n$, are generated by the Clifford-cyclotomic gate set. We previously settled this using Euler-Poincar\'{e} characteristics. A generalization of Sarnak's conjecture is to ask when these groups are generated by torsion elements. An obstruction to this is provided by the corank: a group $G$ has $\operatorname{corank} G >0$ only if $G$ is not generated by torsion elements. In this paper we study the corank of these cyclotomic unitary groups in the families $n=2^s$ and $n=3\cdot 2^s$, $n\geq 8$, by letting them act on Bruhat-Tits trees. The quotients by this action are finite graphs whose first Betti number is the corank of the group. Our main result is that for $n=2^s$ and $n=3\cdot 2^s$ the corank groups doubly exponentially in $s$ as $s\rightarrow \infty$; it is $0$ precisely when $n=8,12, 16,24$ and indeed the cyclotomic unitary groups are generated by torsion elements (in fact by the Clifford-cyclotomic gates) for these $n$. We give explicit lower bounds for the corank in two different ways. The first is to bound the isotropy subgroups in the action on the tree by explicit cyclotomy. The second is to relate our graphs to Shimura curves over $F_n=\mathbb{Q}(\zeta_n)^+$ via interchanging local invariants and applying a result of Selberg and Zograf. We show that the cyclotomy arguments give the stronger bounds. In a final section we execute a program of Sarnak to show that our results for the $n=2^s$ and $n=3\cdot 2^s$ families are sufficient to give a second proof of Sarnak's conjecture.