Machine Learning Parameters of Optimally Tuned Range-Separated Hybrid Functionals for Transition Metal Complexes.

Area of Science:

• Computational Chemistry
• Materials Science
• Machine Learning

Background:

• Optimally tuned range-separated hybrid (OT-RSH) functionals offer high accuracy for electronic structure calculations.
• Determining optimal parameters for these functionals can be computationally expensive, especially for large systems like transition metal complexes (TMCs).
• Developing efficient methods to predict these parameters is crucial for accelerating materials discovery.

Purpose of the Study:

• To develop a machine learning (ML) approach for predicting optimal range separation parameters in TMCs.
• To reduce the computational cost of electronic structure calculations using OT-RSH functionals.
• To enable accurate and efficient high-throughput screening of TMCs for optoelectronic applications.

Main Methods:

• A dataset of 4380 TMCs was compiled from the tmQM database.
• Each TMC was represented by a 62,087-dimensional multiple-fingerprint feature (MFF) vector.
• Support vector machine (SVM) regression models were trained to predict the optimal range separation parameter.

Main Results:

• The developed machine-learned functional, ML-LC-PBE0*, predicts ionization potentials comparable to fully tuned calculations.
• This ML approach significantly reduces computational cost compared to traditional methods.
• ML-LC-PBE0* demonstrates superior performance over conventional global and fixed-parameter RSH functionals when compared to DLPNO-CCSD(T) reference values.

Conclusions:

• The ML approach provides an efficient and accurate framework for predicting optimal parameters in RSH functionals for TMCs.
• This method facilitates high-throughput screening and electronic structure prediction.
• The findings pave the way for discovering novel TMC-based optoelectronic materials.