Quantum Power Method by a Superposition of Time-Evolved States

We propose a quantum-classical hybrid algorithm of the power method, here dubbed as quantum power method, to evaluate hat{cal H}ⁿ |ψrangle with quantum computers, where n is a nonnegative integer, hat{cal H} is a time-independent Hamiltonian of interest, and |ψrangle is a quantum state. We show that the number of gates required for approximating hat{cal H}ⁿ scales linearly in the power and the number of qubits, making it a promising application for near term quantum computers. Using numerical simulation, we show that the power method can control systematic errors in approximating the Hamiltonian power {hat{cal H}ⁿ} for n as large as 100. As an application, we combine our method with a multireference Krylov-subspace-diagonalization scheme to show how one can improve the estimation of ground-state energies and the ground-state fidelities found using a variational-quantum-eigensolver scheme. Finally, we outline other applications of the quantum power method, including several moment-based methods. We numerically demonstrate the connected-moment expansion for the imaginary-time evolution and compare the results with the multireference Krylov-subspace diagonalization.