Energy landscape structure of small graph isomorphism under variational optimization

We investigate a quadratic unconstrained binary optimization formulation of the graph isomorphism problem using the quantum approximate optimization algorithm and the variational quantum eigensolver. For small graph instances, we observe that isomorphic pairs exhibit consistent clustering in variational energies, indicating that the Hamiltonian successfully encodes structural features. However, we demonstrate that low variational energy alone is an unreliable certifier of isomorphism due to the high probability of converging to infeasible states that violate bijection constraints. To address this, we analyze optimization trajectories rather than final energies, consistently outperforming naive energy thresholding, although absolute performance remains limited. Our results characterize the current limits of variational algorithms for graph isomorphism, positioning energy landscape analysis as a diagnostic tool rather than a scalable decision procedure in the noisy intermediate scale quantum regime.