Polymer molecular design

Molecule design and discovery in data-scarce(!) regimes

Identifying a polymer chemistry that meets the physical property requirements is a time- and resource-intensive process that involves many experiments and "Eureka!" moments. Even high-throughput screening or optimization based on virtual models (e.g., molecular dynamics) is inefficient because the molecular design space can be extremely large for polymer molecules. Additionally, there is a gap in interpreting and representing material structures between humans and computers. For example, in the case of polymers, while the aliphatic (linear carbon-carbon (C-C) chain) vs. aromatic (C-C rings) chemistry is interpretable to humans, it is not directly cognizant to a computer algorithm. How does one even translate symbolic representations of different molecules (e.g., 4,4’-(Propane-2,2-diyl)diphenol and 2,2’-[Oxybis(methylene)]bis(oxirane))) to a unified numerical representation that can be understood by an algorithm trying to optimize the polymer chemistry?  To this end, language- and graph-ML frameworks have remarkably emerged as the state-of-art solutions, borrowing tools from natural language processing and computer vision, respectively. However, key challenges persist. 

• They require large amounts of training data, often in tens of thousands of data points which are infeasible to obtain through expensive simulations. 

• The ML models often exhibit hallucination and predict chemically invalid, fictitious, or likely unsynthesizable molecules.

To address these challenges, we are developing a suite of data-efficient ML frameworks for polymer molecule design. Our approach centers on integrating physical and chemical knowledge with ML techniques, effectively overcoming issues related to data scarcity and model hallucination.

Deep learning reveals key predictors of thermal conductivity in covalent organic frameworks

The thermal conductivity of covalent organic frameworks (COFs), an emerging class of nanoporous polymeric materials, is crucial for many applications, yet the link between their structure and thermal properties is not well understood. From a dataset of over 2,400 COFs, we find that conventional features like density, pore size, void fraction, and surface area do not reliably predict thermal conductivity. To overcome this, we train an attention-based machine learning model that accurately predicts thermal conductivities, even for structures outside the training set. We then use the attention mechanism to understand why the model works. Surprisingly, dangling molecular branches emerge as key predictors of thermal conductivity, alongside conventional geometric descriptors like density and pore size. Our findings show that COFs with dangling functional groups exhibit lower thermal transfer capabilities than otherwise. Molecular dynamics simulations confirm this, revealing significant mismatches in the vibrational density of states due to the presence of dangling branches.

ML-guided inverse design and discovery of recyclable vitrimeric polymers

Vitrimer is a new, exciting class of sustainable polymers with the ability to heal due to their dynamic covalent adaptive network that can go through associative rearrangement reactions. However, a limited choice of constituent molecules restricts their property space, prohibiting full realization of their potential applications. To overcome this challenge, we couple molecular dynamics (MD) simulations and a graph variational autoencoder (VAE) machine learning model for inverse design of vitrimer chemistries with desired glass transition temperature (Tg) and synthesize a novel vitrimer polymer. We demonstrate high accuracy and efficiency of our framework in discovering novel vitrimers with desirable Tg beyond the training regime. The proposed framework offers an exciting tool for polymer chemists to design and synthesize novel, sustainable vitrimer polymers for a facet of applications. 

Publications

• P. Thakolkaran*, Y. Zheng*, Y. Guo, A. Vashisth*, S. Kumar*, Deep learning reveals key predictors of thermal conductivity in covalent organic frameworks, Arxiv:2409.06457.

• Y. Zheng, P. Thakolkaran, A.K. Biswal, J.A. Smith, Z. Lu, S. Zheng, B. H. Nguyen*, S. Kumar*, A. Vashisth*, AI-guided inverse design and discovery of recyclable vitrimeric polymers, Advanced Science, 2411385 (2024). [code] [data]