Disentangling topographic contributions to near-field scanning microwave microscopy images

Highlights: • We present an empirical statistical model to estimate the topographic contribution to measured microwave images acquired by Near-field Scanning Microwave Microscopy (NSMM). Since topographic and material property variations can contribute to any NSSM measurement, our approach is critical for determination of the material property contribution to any NSSM measurement. • The prediction model parameters are determined from either reference sample or from the sample of interest by robust regression methods. • We apply our methods to thin Perovskite photovoltaic material samples and a GaN nanowire. • For the GaN, we show that our statistical prediction model is dramatically better than a prediction based on a simple physical model. -- Abstract: We develop empirical models to predict the contribution of topographic variations in a sample to near-field scanning probe microwave microscopy (NSMM) images. In particular, we focus on |S₁₁| images of a thin Perovskite photovoltaic material and a GaN nanowire. The difference between the measured NSMM image and this prediction is our estimate of the contribution of material property variations to the measured image. Prediction model parameters are determined from either a reference sample that is nearly free of material property variations or directly from the sample of interest. The parameters of the prediction model are determined by robust linear regression so as to minimize the effect of material property variations on results. For the case where the parameters are determined from the reference sample, the prediction is adjusted to account for instrument drift effects. Our statistical approach black is fully empirical black and thus complementary to current approaches based on physical models that are often overly simplistic.