Latent unified smooth Hamiltonians for excited state chemistry

The paper introduces a neural network architecture that learns a latent, implicit basis representation of the electronic-state Hamiltonian, enabling a unified treatment of ground and excited states, conical intersections, and non‑adiabatic couplings. The model is trained on realistic photochemical systems—thymine and azobenzene—and accurately reproduces energies, oscillator strengths, and critical geometries such as conical intersections and excited‑state minima. It also captures Berry phase accumulation around conical intersections and can be extended to learn other operators like transition dipole moments.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Machine Learning.

• llms