Computational Supremacy of Quantum Eigensolver by Extension of Optimized Binary Configurations

Abstract

We developed a quantum eigensolver (QE) which is based on an extension of optimized binary configurations measured by quantum annealing (QA) on a D-Wave Quantum Annealer (D-Wave QA). This approach performs iterative QA measurements to optimize the eigenstates $\vert \psi \rangle$ without the derivation of a classical computer. The computational cost is $\eta M L$ for full eigenvalues $E$ and $\vert \psi \rangle$ of the Hamiltonian $\hat{H}$ of size $L \times L$, where $M$ and $\eta$ are the number of QA measurements required to reach the converged $\vert \psi \rangle$ and the total annealing time of many QA shots, respectively. Unlike the exact diagonalized (ED) algorithm with $L^3$ iterations on a classical computer, the computation cost is not significantly affected by $L$ and $M$ because $\eta$ represents a very short time within $10^{-2}$ seconds on the D-Wave QA. We selected the tight-binding $\hat{H}$ that contains the exact $E$ values of all energy states in two systems with metallic and insulating phases. We confirmed that the proposed QE algorithm provides exact solutions within the errors of $5 \times 10^{-3}$. The QE algorithm will not only show computational supremacy over the ED approach on a classical computer but will also be widely used for various applications such as material and drug design.