Gaussian Process Regression Aided Spiral Scanning on Polaritonic Media

Integration time and signal-to-noise are inextricably linked when performing scanning probe measurements such as in scanning near-field optical microscopy (SNOM). Since these measurements define a large lower bound on the measurement time, we used a combination of Gaussian process regression with sparse spiral scanning in order to bypass this constraint. Our study demonstrates that this approach, when used to image graphene/α-RuCl₃ charge-transfer polaritons and hBN phonon polaritons, results in key features such as damping and dispersion that are in good agreement with those extracted from traditional raster scans with the same integration time per pixel and dimensions. Most significantly, the gaussian process aided sparse spiral scan has roughly 9 times less data than raster scans and offers a commensurate 9 times decrease in measurement time to raster scans.