David Gao1,2,3, Filippo Federici Canova1,4, Nikola Kolev3,5, Taylor Stock3,5, Neil Curson3,5
1Nanolayers Research Computing LTD
2Norwegian Institute of Science and Technology
3University College London
4Aalto University
5London Centre for Nanotechnology
Modern research techniques have massively increased our ability to generate large quantities of both experimental and simulated data. This data can be leveraged with data science, machine learning, and automation techniques to create significant impact in industry. Scanning probe lithography for example is a promising manufacturing technique for microelectronic and quantum devices that require atomic scale precision. However, typical laboratory techniques are not suitable for industrial scale manufacturing and require a specialist to perform repetitive, tedious, and prohibitively time-consuming tasks. The LabCore digital research platform and ALANN graphical user interface are being developed to bridge this gap by combining data management and machine learning tools with instrument control and automation capabilities. Our software automation plug-in connects directly to the scanner and allows the user to control all relevant imaging parameters, acquire scans, and automatically save them in a convenient format. At the heart of the software lies integrated advanced image filtering routines that remove slope, spikes and creep from the data, compensate for small tip changes and ultimately construct a binary map of the detected atomic step edges. Defects in atomic scale images are automatically detected and identified using machine-learning methods and a custom correlation algorithm realigns the SPM images as they are acquired by stitching their step maps together to eliminate drift and aid in navigation. ALANN constructs a map of the entire sample that can be used for navigation, instrument control to write user specified patterns, and autonomously alignment of consecutive lithography stages to ensure precise results.