Designing building blocks of covalent organic frameworks through on-the-fly batch-based Bayesian optimization.

Area of Science:

• Materials Science
• Computational Chemistry
• Organic Chemistry

Background:

• Covalent organic frameworks (COFs) are crystalline porous polymers with tunable properties.
• Designing COFs with specific functionalities, like high hole conductivity, is challenging due to the vast chemical space.
• Traditional methods for exploring COF building blocks are computationally expensive and time-consuming.

Purpose of the Study:

• To develop a computational method for designing covalent organic frameworks (COFs) with high hole conductivity.
• To efficiently navigate the large design space of COF building blocks.
• To identify novel and promising COF candidates for enhanced electronic applications.

Main Methods:

• Utilized a Bayesian optimization (BO) algorithm to sample the COF building block space.
• Employed a molecular generation algorithm to ensure the creation of valid, rigid, three-fold symmetric molecules.
• Trained surrogate models to predict COF conductivity properties (level alignment and reorganization free energy) at low computational cost.

Main Results:

• Identified several promising COF candidates with high predicted hole conductivity after 20 training steps.
• Discovered both novel molecular motifs and variations of known structures.
• The BO approach significantly reduced the computational cost of material screening.

Conclusions:

• Bayesian optimization is an effective strategy for accelerating the discovery of functional COFs.
• The identified candidates warrant further investigation for their potential in electronic devices.
• Computational screening can significantly streamline the materials design process.