Universal Phase Identification of Block Copolymers from Physics-informed Machine Learning

Elizabeth Murphy,Xinyi Fang,Phillip Kohl,Youli Li,Craig Hawker,Christopher Bates,Mengyang Gu

DOI: https://doi.org/10.26434/chemrxiv-2024-9zjt0

2024-05-31

Abstract:Block copolymers play a vital role in materials science due to their widely studied self-assembly behavior. Traditionally, exploring the phase space of block copolymer self-assembly and associated structure–property relationships involves iterative synthesis, characterization, and theory, which is labor-intensive both experimentally and computationally. Here, we introduce a versatile, high-throughput workflow towards materials discovery that integrates controlled polymerization and automated chromatographic separation with a novel physics-informed machine learning (ML) algorithm for the rapid analysis of small-angle X-ray scattering (SAXS) data. Leveraging the expansive and high-quality experimental datasets generated by automated chromatography, this machine learning method effectively reduces data dimensionality by extracting chemical-independent features from SAXS data. This new approach allows for the rapid and accurate prediction of morphologies without repetitive manual analysis, achieving out-of-sample predictive accuracy of around 95% for both novel and existing materials in the training dataset. By focusing on a subset of samples with large predictive uncertainty, only a small fraction of the samples needs to be inspected to further improve accuracy and achieve near-perfect predictions. In summary, the synergistic combination of controlled synthesis, automated chromatography, and data-driven analysis creates a powerful workflow that markedly expedites the discovery of structure–property relationships in advanced soft materials.

Chemistry

What problem does this paper attempt to address?

This paper aims to address the challenges in the phase - state identification of block copolymers during self - assembly. Traditional methods involve a large amount of synthesis, characterization, and theoretical research, which are not only time - consuming and labor - intensive but also very intensive in both experiments and calculations. Specifically, the paper attempts to solve the following main problems: 1. **Improve the efficiency of phase - state identification**: Traditional phase - state identification methods require manual analysis of each small - angle X - ray scattering (SAXS) data, which is very time - consuming and error - prone. The paper proposes a physics - informed machine - learning algorithm that can quickly and automatically extract key features from SAXS data, thereby achieving efficient and accurate phase - state prediction. 2. **Reduce data dimensions**: Traditional machine - learning methods usually deal with complete intensity curves, which contain a large amount of noise and redundant information. By extracting chemically - independent features, the paper significantly reduces the data dimensions and improves the accuracy and robustness of the model. 3. **Improve prediction accuracy**: The method in the paper achieves a prediction accuracy of about 95% on both new and existing materials in the training set. By examining samples with high prediction uncertainty, the prediction accuracy is further improved and eventually approaches 100%. 4. **Adapt to multiple materials**: The method proposed in the paper is applicable not only to specific types of block copolymers but also to new materials with different chemical compositions, demonstrating its wide applicability and versatility. 5. **Automate laboratory work - flow**: By combining automated chromatographic separation techniques and advanced statistical models, the paper establishes a high - throughput work - flow, which greatly accelerates the discovery process of new materials and provides in - depth understanding of the structure - performance relationship. In conclusion, through the introduction of an innovative physics - informed machine - learning algorithm, this paper solves several key problems in the phase - state identification of block copolymers and provides a powerful tool for research in the field of materials science.

Universal Phase Identification of Block Copolymers from Physics-informed Machine Learning

Accelerated discovery and mapping of block copolymer phase diagrams

Autonomous Construction of Phase Diagrams of Block Copolymers by Theory-Assisted Active Machine Learning

Surface-induced Phase Transitions in Dense Nanoparticle Arrays of Lamella-Forming Diblock Copolymers

Computational Phase Discovery in Block Polymers

Accelerated Prediction of Phase Behavior for Block Copolymer Libraries Using a Molecularly Informed Field Theory

Interpretable Machine Learning Models for Phase Prediction in Polymerization-Induced Self-Assembly

Prediction of Diblock Copolymer Morphology via Machine Learning

Discrete Block Copolymers with Diverse Architectures: Resolving Complex Spherical Phases with One Monomer Resolution

Simple Calibration of Block Copolymer Melt Models

Development of a deep-learning model for phase-separation structure of diblock copolymer based on self-consistent field analysis

Machine Learning and Polymer Self-Consistent Field Theory in Two Spatial Dimensions

Adaptively driven X-ray diffraction guided by machine learning for autonomous phase identification

Morphology and Phase Diagram of ABC Linear Triblock Copolymers: Parallel Real-Space Self-Consistent-field-theory Simulation.

Neural Network-Driven Exploration of Solvophilic Block Size Effects in Polymerization-Induced Self-Assembly: From 2D To 3D Comprehensive Pseudo-Phase Diagram

The Block Copolymer Phase Behavior Database

Discovering Ordered Phases of Block Copolymers: New Results from a Generic Fourier-space Approach.

Phase behavior of triblock copolymer and homopolymer blends: Effect of copolymer topology

Enhancing Copolymer Property Prediction through the Weighted-Chained-SMILES Machine Learning Framework

Stabilizing hexagonally close-packed phase in single-component block copolymers through rational symmetry breaking

Accelerated Design of Block Copolymers: An Unbiased Exploration Strategy via Fusion of Molecular Dynamics Simulations and Machine Learning