Agents improve selecting key orbitals for molecular excited state calculations
El Agente Quntur: Agentic Active-Space Selection for CASSCF
Multiagent Systems
Summary
Picking the right pieces to study in complex molecules is tricky and usually needs expert judgement and a lot of trial and error. The authors used a system of smart computer agents driven by advanced language models to automatically choose and adjust these key pieces, called active spaces, for electronic calculations. Their method was tested on various molecular excited states and showed more accurate results than previous automatic methods. This approach helps reduce human effort in setting up these complex calculations while improving accuracy. However, the authors note there is still room to make the process faster and more consistent.
What this means in practice
- •For computational chemists: Automatically select and refine active spaces for better accuracy in excited-state molecular simulations.
- •For pharmaceutical modeling teams: Improve computational predictions by reducing setup time and errors in electronic structure calculations for drug-like molecules.
Authors
Juan B. Pérez-Sánchez, Viviana Roman-Ventura, Olalla Nieto-Faza, Aiwei Yin, Alán Aspuru-Guzik, Varinia Bernales
Abstract
Multireference calculations often require substantial human expertise and time to choose and refine an active space for the electronic states or chemical process of interest. This involves selecting which orbitals to include, verifying that the calculation describes the intended excited state, and deciding whether the results have converged. Here, we ask whether scientific agentic systems (harnesses) can make these decisions reliably. We use the El Agente Quntur multi-agent harness, with GPT-5.6 Sol and GPT-5.6 Terra as the underlying LLM components, to explore 32 diverse singlet excitations at fixed molecular geometries. The agents planned and executed calculations, choosing and refining active spaces under prompts that specified equal-weight two-singlet state-averaged CASSCF followed by strongly contracted NEVPT2 with def2-TZVPD. We evaluate excitation energies against published CASPT2 reference values at matching geometries and compare performance with six deterministic Active Space Finder protocols. Within the El Agente Quntur system, agents detect missing active orbitals, compare candidate excited states, and revise state assignments as calculations progress. Agentic execution reduced mean absolute error by 48\% relative to the state-of-the-art Active Space Finder protocol. We note that computational traces vary substantially across runs in both active space selection strategies and cost, even for the same molecule. Our results underscore the advantage of agentic active space selection over fixed workflows, yet leave room to improve efficiency through runtime-aware decision-making, reduced inter-agent communication overhead and redundancy, more systematic checks of state identity, and clearer stopping criteria.