Efficient Quantum Information-Inspired Ansatz for Variational Quantum Eigensolver Algorithm: Applications to Atomic Systems

Abstract

We present a quantum information-inspired ansatz for the variational quantum eigensolver (VQE) and demonstrate its efficacy in calculating ground-state energies of atomic systems. Instead of adopting a heuristic approach, we start with an approximate multi-qubit target state and utilize two quantum information-theoretic quantities, i.e., von Neumann entropy and quantum mutual information, to construct our ansatz. The quantum information encoded in the target state helps us to design unique blocks and identify qubit pairs that share maximum quantum correlations among them in the multi-qubit system, thereby enabling us to deterministically place two-qubit entanglers in the suitably constructed parametrized quantum circuit. We find that our approach has the advantage of reduced circuit depth compared to the unitary coupled-cluster (UCC) ansatz (the gold standard for VQE), and yet yields accurate results. To test the performance of our ansatz, we apply it to compute ground-state energies of atomic systems. We find that for up to 12 qubits (or 12 spin orbitals) noiseless calculation, the proposed ansatz yields energies with 99.99% accuracy relative to the complete active space configuration interaction values, while utilizing only two blocks, which contain at most 99% fewer 2-qubit gates than the UCC ansatz.