Vibrational energies for HFCO using a neural network sum of exponentials potential energy surface

Problem

Accurately computing the vibrational energies of formyl fluoride (HFCO) requires an accurate six-dimensional potential energy surface fitted to expensive ab initio data.

Why it matters

Accurate potential energy surfaces enable reliable prediction and interpretation of experimental infrared spectra of molecules.

What the AI did

A neural network sum-of-exponentials approach fit a six-dimensional PES to CCSD(T)-F12 ab initio data; vibrational energies for HFCO and DFCO were then computed and validated with the MCTDH approach.

Results

PES fit RMSE = 10 cm−1 up to 10,000 cm−1 above zero-point energy; fundamental and all vibrational states up to 5000 cm−1 more accurate than previous MP2-based PES

Impact

The new PES outperformed the prior MP2-based surface and covers most experimentally measured IR data for HFCO/DFCO.

Record

Importance rating

More in Physics & Astronomy