LadderGen: a large-scale generative library of ladder polymers for membrane separations

Area of Science:

• Polymer Science
• Materials Science
• Computational Chemistry

Background:

• Ladder polymers offer unique properties like rigidity and microporosity for gas separation membranes.
• Limited synthetic accessibility of ladder polymers restricts their practical application.
• Developing new ladder polymers is crucial for advancing membrane technology.

Purpose of the Study:

• To create a large-scale hypothetical library of ladder polymers using machine learning.
• To identify promising ladder polymer candidates for gas separation membranes.
• To accelerate the discovery of next-generation materials for sustainable applications.

Main Methods:

• Utilized template polymerization reactions and generative machine learning (ML) models.
• Constructed a library of 0.8 million hypothetical ladder polymer structures.
• Employed ML models to predict key membrane properties (Tg, FFV, O2/N2 selectivity/permeability).
• Validated predictions using high-fidelity molecular dynamics simulations.

Main Results:

• Generated an extensive library of novel ladder polymer structures.
• Identified potential candidates approaching or exceeding the 2008 Robeson upper bound.
• ML predictions were validated by molecular dynamics simulations for top candidates.
• Demonstrated the capability of LadderGen for rapid screening and discovery.

Conclusions:

• LadderGen significantly expands the chemical space of ladder polymers.
• The framework enables accelerated *de novo* discovery of high-performance gas separation membranes.
• This approach facilitates the development of materials for sustainable energy and environmental solutions.