Data-Driven Acceleration of the Experimental Synthesis of Quantum Materials

Quantum materials are made of solid crystals that possess various quantum properties, such as entanglement and coherence. These properties could lead to important technological advances in electronics and energy efficiency. Despite their scientific and industrial relevance, the creation of these materials remains vastly empirical, with a majority of processes being governed by individual practical methods acquired through experience. This project proposes developing a system based on data-driven discovery to help guide the experimental process in order to identify regions of parameter space that are most likely to be viable and to accelerate the optimization of desired material properties. Currently, there is not enough experimental data to support this data-driven method. To address this, we are using data from available scientific literature to begin. This data is often disorganized, limiting its usage in data processing.

In order to access and extract the literature data, we will first manually label relevant parts of scientific papers which contain the process to create materials. These texts will be used as a training set to train a neural network, which will then automatically identify the paragraphs that are relevant to the composition of the material. These selected paragraphs will then be analyzed using natural language processing, a technique that allows to identify the specific steps of the composition process such as temperature, amount of chemical and other variables. Finally, the chemical will receive its unique “recipe” which will be uploaded to a database for facilitated access.

Faculty Advisor: Linda Petzold