Improved Simulation of Quantum Circuits by Fewer Gaussian Eliminations

Abstract: We show that the cost of strong simulation of quantum circuits using $t$ $T$ gate magic states exhibits non-trivial reductions on its upper bound for $t=1$, $t=2$, $t=3$, and $t=6$ with odd-prime-qudits. This agrees with previous numerical bounds found for qubits. We define simulation cost by the number of terms that require Gaussian elimination of a $t \times t$ matrix and so capture the cost of simulation methods that proceed by computing stabilizer inner products or evaluating quadratic Gauss sums. Prior numerical searchs for qubits were unable to converge beyond $t=7$. We effectively increase the space searched for these non-trivial reductions by $>10^{10^4}$ and extend the bounds to $t=14$ for qutrits. This is accomplished by using the Wigner-Weyl-Moyal formalism to algebraically find bounds instead of relying on numerics. We find a new reduction in the upper bound from the $12$-qutrit magic state of ${3^{\sim 0.469t}}$, which improves on the bound obtained from the $6$-qutrit magic state of ${3^{\sim 0.482t}}$.