A self-driving laboratory for optimizing thin-film materials

To satisfy the evolving needs of the many industries that employ thin-film materials, new materials and deposition methods are continually being optimized. These optimizations are often slow and empirical because thin-film materials and deposition methods are complex. In this thesis I develop a new tool for accelerating the optimization of thin-film materials: a self-driving laboratory (SDL). This SDL consists of a suite of automated film synthesis and characterization stations linked together by robots and controlled by an experiment-planning algorithm. This SDL optimizes thin-film materials by iteratively planning and executing film synthesis experiments in a fully autonomous loop. I first demonstrate the SDL by using it to autonomously maximize the hole mobility of films of spiro-OMeTAD, a p-type organic semiconductor used in perovskite solar cells. The SDL automatically deposited films containing varying amounts of a dopant, annealed these films for varying durations, and then characterized the films. The SDL identified the doping ratio and annealing time that maximize the hole mobility by performing 35 cycles of iterative experimentation under the control of a Bayesian optimization algorithm. This study showed that an SDL can autonomously optimize film composition and processing parameters without ongoing human intervention. I then upgrade the SDL to optimize combustion-synthesized palladium films for multiple objectives simultaneously. Under the control of a multiobjective Bayesian optimization algorithm, the SDL autonomously identified a range of combustion synthesis conditions yielding optimal trade-offs between the conflicting objectives of conductivity and annealing temperature for drop-cast films. Simulated optimization runs indicated that the self-driving laboratory achieved this result in 10 times fewer experiments than would have been required using a grid search. The combustion synthesis conditions identified by the self-driving laboratory enable the spray coating of uniform palladium films with moderate conductivity (1.1E5 S/m) at 191 °C. Spray coating at 226 °C yields films with conductivities (2.0E6 S/m) comparable to those of sputtered films (2.0 to 5.8E6 S/m). This work shows how a self-driving laboratory can efficiently identify film synthesis conditions yielding optimal trade-offs between conflicting objectives.Science, Faculty ofChemistry, Department ofGraduat