A Note on a Quantitative Form of the Solovay-Kitaev Theorem

TL;DR

This paper provides a quantitative analysis of the Solovay-Kitaev theorem, establishing bounds on the efficiency of universal gate sets in approximating arbitrary 1-qubit gates and relating it to geometric properties of certain lattice points.

Contribution

It introduces a new measure of efficiency for universal gate sets and derives bounds on their approximation capabilities using a quantitative framework.

Findings

Derived bounds on the measure K(T) for universal gate sets.

Connected approximation efficiency to geometric properties of lattice points.

Proposed a conjecture relating covering radius to the approximation process.

Abstract

The problem of finding good approximations of arbitrary 1-qubit gates is identical to that of finding a dense group generated by a universal subset of $SU(2)$ to approximate an arbitrary element of $SU(2)$. The Solovay-Kitaev Theorem is a well-known theorem that guarantees the existence of a finite sequence of 1-qubit quantum gates approximating an arbitrary unitary matrix in $SU(2)$ within specified accuracy $\varepsilon > 0$. In this note we study a quantitative description of this theorem in the following sense. We will work with a universal gate set $T$, a subset of $SU(2)$ such that the group generated by the elements of $T$ is dense in $SU(2)$. For $\varepsilon > 0$ small enough, we define $t_{\varepsilon}$ as the minimum reduced word length such that every point of $SU(2)$ lies within a ball of radius $\varepsilon$ centered at the points in the dense subgroup generated by $T$. For a measure of efficiency on T, which we denote $K(T)$, we prove the following theorem: Fix a $\delta$ in $[0, \frac{2}{3}]$. Choose $f: (0, \infty) \rightarrow (1, \infty)$ satisfying $\lim_{\varepsilon\to 0+}\dfrac{\log(f(t_{\varepsilon}))}{t_{\varepsilon}}$ exists with value $0$. Assume that the inequality $\varepsilon \leqslant f(t_{\varepsilon})\cdot 5^{\frac{-t_{\varepsilon}}{6-3\delta}}$ holds. Then $K(T) \leqslant 2-\delta$. Our conjecture implies the following: Let $\nu(5^{t_{\varepsilon}})$ denote the set of integer solutions of the quadratic form: $x_1^2+x_2^2+x_3^2+x_4^2=5^{t_{\varepsilon}}$. Let $M\equiv M_{S^3}(\mathcal{N})$ denote the covering radius of the points $\mathcal{N}=\nu(5^{t_{\varepsilon}})\cup\nu(5^{t_{\varepsilon}-1})$ on the sphere $S^{3}$ in $\mathbb{R}^{4}$. Then $M \sim f(\log N)N^{\frac{-1}{6-3\delta}}$. Here $N\equiv N(\varepsilon)=6\cdot5^{t_{\varepsilon}}-2$.