Stacked Ensemble Machine Learning for Range-Separation Parameters.

Area of Science:

• Computational Chemistry
• Materials Science
• Machine Learning Applications

Background:

• Density functional theory (DFT) is crucial for high-throughput materials and drug discovery.
• Self-interaction error significantly limits DFT's accuracy for organic semiconductors.
• Optimally tuned range-separated hybrid (OT-RSH) functionals address this error but are computationally expensive.

Purpose of the Study:

• To develop an accelerated computational method for determining the range-separation parameter (ω) in OT-RSH functionals.
• To create a machine learning model that predicts ω based on molecular structure and electronic properties.
• To provide a cost-effective alternative to traditional OT-RSH calculations for organic semiconductors.

Main Methods:

• Development of a stacked ensemble machine learning model (ML-ωPBE).
• Training the model using a diverse dataset of 1970 organic molecules.
• Validation of the model's accuracy and efficiency on an independent set of 1956 molecules.

Main Results:

• ML-ωPBE achieves a mean absolute error of 0.00504 a₀⁻¹ for optimal ω values.
• The machine learning approach reduces computational cost by 2.66 orders of magnitude compared to nonempirical OT-ωPBE.
• The model demonstrates comparable predictive accuracy for optical properties of organic semiconductors.

Conclusions:

• The proposed machine learning model offers a highly efficient and accurate method for calculating OT-RSH parameters.
• ML-ωPBE significantly accelerates the discovery and design of organic semiconducting materials.
• This work overcomes a major bottleneck in applying advanced DFT methods to organic electronics.