Automating Scanning Tunneling Microscopy with AI/ML

                Scanning Tunneling Microscopy (STM) is a novel technique for obtaining atomic-resolution images of a sample surface and map the local density of states (LDOS). However, STM experiments tend to be time-consuming, difficult to reproduce, and require constant manual adjustment. To overcome these difficulties, we have developed machine-learning (ML)-assisted workflow that automates an experiment by controlling multiple instruments to correct for magnetic, piezoelectric, and thermal drift in real time. With this workflow, we were able to acquire the LDOS of superconductors as a function of temperature and magnetic field while compensating for drift in real time, which improves the reproducibility of data collected across multiple experiments. The STM automation significantly reduces the time and manpower for data acquisition, which will allow us to collect vast amount of data in reliable and systematic way.