Fully Automated Optimization of Robot-Based MOF Thin Film Growth via Machine Learning Approaches

Metal-organic frameworks (MOFs), have emerged as ideal class of materials for the identification of structure-property relationships and for the targeted design of multifunctional materials for diverse applications. While the powder form is most common, for the integration of MOFs into devices, typically thin films of surface anchored MOFs (SURMOFs), are required. Although the quality of SURMOFs emerging from layer-by-layer approaches is impressive, previous works revealed that the optimum growth conditions are very different between different types of MOFs and different substrates. Furthermore, the choice of appropriate synthesis conditions (e.g., solvents, modulators, concentrations, immersion times) is crucial for the growth process and needs to be adjusted for different substrates. Machine learning (ML) approaches show great promise for multi-parameter optimization problems such as the above discussed growth conditions for SURMOF on a particular substrate. Here, this work presents an ML-based approach allowing to quickly identify optimized growth conditions for HKUST-I SURMOFs with high crystallinity and uniform orientation. This process can subsequently be used to optimize growth on other types of substrates. In addition, an analysis of the results allows to gain further insights into the factors governing the growth of MOF thin films.

Automated summary

Metal-organic frameworks (MOFs) are ideal materials for studying structure-property relationships and designing multifunctional materials, with surface anchored MOFs (SURMOFs) being commonly used in devices. Previous methods for optimizing SURMOF growth conditions have limitations, but this work presents a machine learning-based approach that quickly identifies optimized growth conditions for high-quality SURMOFs. It also provides insights into the factors influencing MOF thin film growth.