Closed‐Loop Multi‐Objective Optimization for Cu–Sb–S Photo‐Electrocatalytic Materials’ Discovery

Abstract

Copper antimony sulphides are regarded as promising catalysts for photoelectrochemical water splitting because of their earth abundance and broad light absorption. The unique photoactivity of copper antimony sulphides is dependent on their various crystalline structures and atomic compositions. Here, we built a closed‐loop workflow that explores Cu‐Sb‐S compositional space to optimise its photoelectrocatalytic hydrogen evolution from water, by integrating a high‐throughput robotic platform, characterization techniques, and machine learning (ML) optimization workflow. The multiple‐objective optimization model discovered optimum experimental conditions after only nine cycles of integrated experiments‐machine learning loop. Photocurrent testing at 0 V versus reversible hydrogen electrode (RHE) confirmed the expected correlation between the materials properties and photocurrent. An optimum photocurrent of ‐186 µA cm−2 was observed on Cu‐Sb‐S in the ratio of 9:45:46 in the form of single‐layer coating on F‐doped SnO2 (FTO) glass with a corresponding bandgap of 1.85 eV and 63.2% Cu1+/Cu species content. The targeted intelligent search revealed a non‐obvious CuSbS composition that exhibited 2.3 times greater activity than baseline results from random sampling.This article is protected by copyright. All rights reserved

A multiobjective closed-loop approach towards autonomous discovery of electrocatalysts for nitrogen reduction

“…Recent examples using this approach with Bayesian optimization are experimental electrochemical reactor protocols 33 and electrocatalyst stoichiometry for fixed composition. 34 A drawback of using a summation is the need for choosing appropriate weights for each term in the sum. Previous domain knowledge can guide this selection, but could be a barrier towards application in new domains.…”

Accelerating the Design of Multishell Upconverting Nanoparticles through Bayesian Optimization

“…To more efficiently design UCNPs for targeted applications, we looked toward machine learning (ML) approaches, which have emerged as powerful tools for accelerating the design of other complex materials − and nanostructures. − Although ML has been used to analyze spectroscopic data , and images , from UCNP experiments, it has not yet been applied to the discovery or recommendation of new UCNP structures. One promising ML approach, Bayesian optimization (BO), , has been used for the experimental design of nanoparticles, − photocatalysts, phase-change materials, and alloys, and for the acceleration of microscopy . Unlike gradient-based inverse design methods , and stochastic methods such as genetic algorithms, which require large training data sets, BO is a sample efficient algorithm that searches for optimal outcomes starting with a small amount of initial data.…”