Nicholas Sullivan1, Peter Grutter2, Kirk H. Bevan1
1Department of Mining and Materials Engineering, McGill University,
2Department of Physics, McGill University,
Scanning probe microscopy (SPM) is a valuable technique by which one can investigate the physical characteristics of the surfaces of materials. However, its ability to gain broader statistical information is hampered by the time-consuming nature of running an experiment, the significant domain knowledge necessary to do so, and the decision-making of the individual researcher guiding the device. Recent studies have shown the value of machine learning-based automation in assisting these, from guiding tasks that require constant user vigilance to redefining the experimental workflow around more statistics-driven decision-making. A remaining limitation is reusability: the experiment is often designed in a device-specific manner, making it non-trivial to integrate created automation into other researchers' experiments. We present an automation framework for SPM (afspm) with the aim of decoupling device specifics from the experiment and any automation components developed. This design limits device-specifics in the software to a single component, the 'microscope translator', which translates between device-specific and afspm-generic terminology. By further defining responsibilities between the components of an experiment, we hope to help researchers develop reusable components that can be easily shared within the community (e.g., 'tip detector/corrector'). The goal is to make it easier to develop and share automation components among the SPM community. We test the framework on two different SPM devices (a softdB/gxsm home-built controller, and an Asylum MFP-3D system), validating the reusability of individual components on devices with different interfaces and programming language requirements.