Sustainable Materials Acceleration Platforms—A Pathway to Democratizing AI in Materials Science

Materials scientists are called upon to solve grand societal challenges, such as climate change, by developing sustainable materials and technologies. To do so, they are rapidly adopting AI methods that may cause a severe crisis in sustainability in our research institutions. To support this shift, we have seen the emergence of materials acceleration platforms (MAPs), which automate data generation tasks and facilitate digitization of materials laboratory data by emulating human research workflows. However, while automation is proven to reduce labor, simply repeating existing costly, wasteful and environmentally harmful tasks to generate bigger datasets will exacerbate budgets and environmental sustainability problems in the short to medium terms, causing a challenge to mass adoption. A recent report by the Royal Society of Chemistry (RSC) has highlighted the need to design and adopt Sustainable Laboratory Practices that reduce the environmental impact of research by implementation of Life Cycle Assessment (LCA). Use of such design tools can also help reduce the overall cost of data generation.<br/><br/>Here, we discuss a sustainable MAP, known as RoboMapper, designed to generate data with an order of magnitude less cost, environmental impact and time compared to existing MAPs. The sustainable MAP is designed in conjunction with an evironmental economist using comparative LCA analysis of data generated using traditional workflows, traditional automation and alternative approaches, to identify the main bottleneck for sustainability. To our surprise, we find that materials characterization is the primary bottleneck and source of environmental impact with material waste and single use plastics following in distant second. We demonstrate how RoboMapper workflow can be implemented collaboratively in a multi-institutional setting through examples in perovskite and polymer research.