Conical intersections play a vital role in photochemical processes. The standard quantum chemistry approach to study conical intersections between ground and excited states are the state-average multi-configurational methods, which at least require solving an active space problem whose computational cost on classical computers scales exponentially in the worst case. Quantum computing offers an alternative tool to solve this problem, however, its applicability to study conical intersections, in particular, on real quantum hardware remains to be explored. In this work, we realize a hybrid quantum-classical state-average complete active space self-consistent field method based on the variational quantum eigensolver (VQE-SA-CASSCF) for the first time on a programmable superconducting quantum processor, and applied it to study conical intersections of two prototypical systems - ethylene (C2H4) and triatomic hydrogen (H3). We show that a combination of different strategies can lead to a qualitatively correct reproduction of conical intersections using VQE-SA-CASSCF, including improving the stability of quantum hardware, reducing the depth of variational circuits, grouping Pauli terms to minimize measurements, and applying appropriate error mitigation. These results allow us to identify the challenges to be overcome in the future and pave the way for using quantum computers to study conical intersections of more complex systems.
Comment: This paper includes 18 pages, 11 figures and 91 references. This paper has not been submitted to any other journals