Maria Gabriela Jordão Oliveira, Nina Glaser

Determining eigenvalues is a computationally expensive task that is crucial for countless applications in natural sciences. Toward this end, we introduce the quantum block Krylov subspace projector (QBKSP) algorithm, a multireference quantum variant of the Lanczos algorithm designed to accurately compute low-lying eigenvalues, including degenerate states. We present three different compact quantum circuits to evaluate the required expectation values, each suited to different problem settings. To investigate the impact of the number and fidelity of the initial reference states, as well as time evolution duration, we perform error-free and limited-precision numerical simulations and quantum circuit simulations. The results demonstrate that using multiple initial reference states improves the convergence of the algorithm, especially in realistic precision-limited simulations and in cases where a single reference fails to simultaneously retrieve all desired eigenvalues. Furthermore, the QBKSP algorithm enables the computation of degenerate eigenstates and respective multiplicity by imposing appropriate convergence criteria.