Maximizing information extraction from AFM nanoindentation — Best practices to algorithms

R. Sheridan, C. Brinson
Duke University,
United States

Keywords: nanoindentation, best practices, information, inference, parametric modeling


As the appetite for reliable, reusable soft materials data grows, AFM nanoindentation is poised to be a major source of this information, considering the high spatial resolution and sampling rates available on modern instrumentation. Reusable data must be collected to an even higher standard of experimental planning and bias elimination than is commonly achieved even in peer-reviewed publications. In this presentation, we share highlights from our recent review containing best practices and recommendations for collecting data, including (a) identifying and reducing, eliminating, or controlling for physical phenomena that interfere with material property measurements and (b) selecting imaging parameters that limit or even eliminate the propagation of spring constant calibration error into modulus measurements. Property measurements, including modulus, require fitting a model to infer parameters – a fraught process that – even after applying best practices – results in a fit that is inevitably imperfect, implying a hidden, incalculable bias in the inferred parameters. This presentation will discuss a model and fitting algorithm that provides an improvement to the status quo, mainly in its robustness and minimization of manual parameter twiddling, then subsequently paint a picture for a machine-learning enabled strategy to transfer and re-use property measurements even in the face of these imperfect models.