Autopylot: Pragmatic Benchmarking of Excited-State Electronic Structure.

Area of Science:

• Computational chemistry
• Photochemistry
• Quantum chemistry

Background:

• Accurate excited-state modeling is vital for understanding photochemical reaction pathways and mechanistic reliability.
• Current methods for selecting electronic structure models rely heavily on manual screening and chemical intuition, often neglecting crucial regions of the potential energy surface.
• This manual approach is time-consuming and may not identify the most suitable method for a given problem.

Purpose of the Study:

• To develop an automated workflow for benchmarking excited-state electronic structure methods.
• To enable efficient and reliable selection of computational methods for photochemical studies.
• To guide users toward optimal cost-accuracy trade-offs in excited-state modeling.

Main Methods:

• Developed Autopylot, a Python package for automated excited-state method benchmarking.
• Implemented a workflow comparing single-structure absorption spectra against a reference across multiple geometries.
• Incorporated computational time as a key metric for evaluating method efficiency.

Main Results:

• Autopylot successfully automated the benchmarking of excited-state electronic structure methods.
• The package accurately described both the Franck-Condon region and excited-state minima.
• Benchmarking on 28 small organic molecules identified accurate methods within minutes, closely reproducing reference spectra.

Conclusions:

• Autopylot significantly advances the automated selection of excited-state electronic structure methods.
• The package offers a flexible and efficient solution for computational chemists.
• This work paves the way for high-throughput, automated method selection in photochemical research.