Unsupervised search of low-lying conformers with spectroscopic accuracy: A two-step algorithm rooted into the island...

Area of Science:

• Computational Chemistry
• Spectroscopy
• Artificial Intelligence

Background:

• Spectroscopic studies traditionally focused on small, semi-rigid molecules.
• Recent shifts target flexible molecules with complex energy landscapes, posing challenges for traditional analysis.
• Artificial intelligence (AI) offers powerful unsupervised methods for exploring molecular flexibility.

Purpose of the Study:

• To present an AI-driven, two-stage quantum mechanical (QM/QM') exploration and refinement strategy for flexible molecules.
• To evaluate the performance of semi-empirical methods in the exploration phase.
• To compare stochastic and meta-heuristic algorithms for conformational space exploration.

Main Methods:

• Integration of unsupervised AI algorithms with a QM/QM' strategy.
• Utilizing a user-friendly graphical interface for the computational workflow.
• Testing various semi-empirical methods and comparing exploration algorithms (stochastic vs. meta-heuristic).
• Application of a polarizable continuum model for condensed-phase simulations.

Main Results:

• Demonstrated efficient exploration of conformational space for medium-sized chromophores, including amino acids.
• Showcased the applicability and efficiency of the method for condensed-phase systems using polarizable continuum models.
• Achieved remarkable agreement between computed and experimental vibrational circular dichroism spectra for rhodium complexes with flexible ligands.

Conclusions:

• The presented unsupervised AI-driven QM/QM' approach effectively handles flexible molecules in both gas and condensed phases.
• The methodology provides a computationally efficient and accurate tool for analyzing complex molecular systems.
• This work advances the study of flexible molecules through the synergistic application of AI and quantum chemistry.