Chen Qu

Despite using fundamentally different machine learning architectures, PIP and PhysNet potentials yield remarkably consistent results for vibrational properties and tunneling dynamics, suggesting that both can reliably represent complex potential energy surfaces.