Troubleshooting unstable molecules in chemical space

Area of Science:

• Computational Chemistry
• Quantum Mechanics
• Materials Science

Background:

• Automated exploration of chemical compound space requires accurate minimum energy geometries.
• Maintaining correct bonding connectivities during geometry optimization is a significant challenge.
• Existing methods may struggle with structural stability and preserving intended molecular representations.

Purpose of the Study:

• To develop and validate an iterative high-throughput workflow for connectivity-preserving geometry optimizations.
• To address issues of geometric instability and unintended structural rearrangements in large chemical datasets.
• To identify and analyze unusual molecular structures, such as those with ultralong bonds.

Main Methods:

• An iterative workflow exploiting the proximity of quantum mechanical (QM) models for geometry optimization.
• Benchmarking on the QM9 dataset (133,885 small molecules with DFT-level properties).
• Utilizing Density Functional Theory (DFT) and post-DFT methods, coupled with topological electron density analysis.

Main Results:

• Successfully troubleshooted 2,988 out of 3,054 molecules with questionable geometric stability, preserving Lewis formula mapping.
• Identified 66 molecules as unstable using DFT and post-DFT methods; 52 contained -NNO- fragments, others had strained pyramidal sp2 carbon.
• Discovered ultralong C-C bonds (r > 1.70 Å) in the curated dataset, supported by topological analysis.

Conclusions:

• The proposed workflow effectively ensures connectivity preservation during geometry optimizations in large chemical datasets.
• The methodology successfully identifies and corrects structural instabilities, improving the reliability of quantum chemistry data.
• This strategy is vital for minimizing unintended structural changes and advancing big data generation in quantum chemistry.