Efficient exploration of compositional space for high-performance copolymers via Bayesian optimization

Area of Science:

• Polymer Science
• Materials Chemistry
• Computational Materials Science

Background:

• Traditional copolymer design relies on inefficient trial-and-error methods.
• Achieving multiple, conflicting properties like moisture resistance, thermal stability, and high modulus in polycyanurates is challenging.
• Intrinsic trade-offs limit simultaneous property improvements.

Purpose of the Study:

• To develop an efficient Bayesian optimization (BO)-guided method for designing co-cured polycyanurates.
• To expedite the discovery of copolymers with improved comprehensive properties.
• To overcome the limitations of traditional trial-and-error approaches in materials design.

Main Methods:

• Developed a Bayesian optimization (BO)-guided strategy.
• Utilized molecular simulations for knowledge integration and benchmarking.
• Conducted experimental validation of designed copolymers.

Main Results:

• Successfully identified co-cured polycyanurates with low water uptake, high glass transition temperature, and high Young's modulus.
• Achieved significant property improvements in a few experimental iterations.
• Demonstrated the efficacy of the BO-guided method in accelerating materials discovery.

Conclusions:

• The developed BO-guided method offers an efficient pathway for designing high-performance copolymers.
• This approach overcomes the inherent trade-offs in property optimization.
• Provides a framework for the efficient design of other advanced polymer materials.