Inverse design of soft materials via a deep learning-based evolutionary strategy.

Gabriele M Coli,Marjolein Dijkstra,Laura Filion,Emanuele Boattini

doi:10.1126/sciadv.abj6731

Abstract

Colloidal self-assembly—the spontaneous organization of colloids into ordered structures—has been considered key to produce next-generation materials. However, the present-day staggering variety of colloidal building blocks and the limitless number of thermodynamic conditions make a systematic exploration intractable. The true challenge in this field is to turn this logic around and to develop a robust, versatile algorithm to inverse design colloids that self-assemble into a target structure. Here, we introduce a generic inverse design method to efficiently reverse-engineer crystals, quasicrystals, and liquid crystals by targeting their diffraction patterns. Our algorithm relies on the synergetic use of an evolutionary strategy for parameter optimization, and a convolutional neural network as an order parameter, and provides a way forward for the inverse design of experimentally feasible colloidal interactions, specifically optimized to stabilize the desired structure.

Highlights

Self-assembly of colloidal particles is ubiquitous in nature and is considered to be of paramount importance for the design of novel functional materials
Tremendous efforts have been devoted to the “forward design” problem: Which structures with what properties are formed for a given colloidal building block under what circumstances? A major drawback of this approach is that the number of possible building blocks and thermodynamic conditions is intractably large, making a systematic exploration of these design spaces extremely demanding
Our inverse design method (IDM) combines the covariance matrix adaptation evolutionary strategy (CMA-ES) for parameters optimization and a convolutional neural network (CNN) for the fitness evaluation, which are both described in detail in Methods

Summary

Introduction

Self-assembly of colloidal particles is ubiquitous in nature and is considered to be of paramount importance for the design of novel functional materials. Viruses, lipid bilayers, tissues, atomic and molecular crystals, liquid crystals, and nanoparticle superlattices are all self-assembled from smaller components in a highly intricate way. The structure of such an assembly is determined by the interactions of the building blocks and by the thermodynamic conditions, e.g., pressure, temperature, or composition. Understanding the relation between building blocks and self- assembled arrangements is essential for materials design, as the physical properties of materials are intimately related to the structure. A major drawback of this approach is that the number of possible building blocks and thermodynamic conditions is intractably large, making a systematic exploration of these design spaces extremely demanding Tremendous efforts have been devoted to the “forward design” problem: Which structures with what properties are formed for a given colloidal building block under what circumstances? A major drawback of this approach is that the number of possible building blocks and thermodynamic conditions is intractably large, making a systematic exploration of these design spaces extremely demanding

Methods

Results

Conclusion