Benchmarking DFT and Supervised Machine Learning: An Organic Semiconducting Polymer Investigation.

Area of Science:

• Materials Science
• Computational Chemistry
• Machine Learning

Background:

• Organic semiconductors are crucial for electronic devices.
• Accurate prediction of electronic properties like bandgap (BG) and ionization potential (IP) is essential for designing new materials.
• Computational methods are vital for material discovery but can be computationally expensive.

Purpose of the Study:

• To evaluate a simple linear regression machine learning model for predicting the BG and IP of organic polymers.
• To assess the accuracy of predictions using PBE and PW91 exchange-correlation functionals.
• To apply the validated model for predicting properties of novel polymer structures.

Main Methods:

• A training set of 22 organic semiconducting polymers was used.
• A supervised machine learning algorithm based on linear regression was employed.
• Density Functional Theory calculations with PBE and PW91 functionals were performed for property calculation.

Main Results:

• The linear regression model achieved high accuracy, with average percent errors below 3% for BG and 4% for IP.
• Calculations using PBE and PW91 functionals yielded reliable results when combined with the ML model.
• The method was successfully applied to predict BG and IP for new polymers and their derivatives.

Conclusions:

• Simple linear regression is a powerful and efficient tool for predicting key electronic properties of organic semiconductors.
• This approach facilitates rapid screening and design of new organic semiconducting materials.
• The study demonstrates the potential of machine learning in accelerating materials discovery for organic electronics.