High speed atomic force microscopy (AFM) force spectroscopy generates thousands of force-distance curves per second, each exhibiting complex tip-sample interactions. We present a GPU-accelerated tool for post-hoc analysis of 1D time series data of repeated experiments. Using dimensionality reduction (UMAP) and interactive clustering, the tool allows us to uncover and study relations between patterns. Physics-agnostic design makes it broadly applicable to serial measurements, like spectroscopy and indentation data.

Hydrogels produced from crosslinked Pluronic diacrylate (PluDA) exhibit a micellar structure. We probe the water-hydrogel interface using AFM in the PeakForce QNM mode (Bruker Nanowizard V) at 2000 force-distance curves per second. Our analysis identifies three distinct surface features with their respective tip-sample interactions. We show how the baseline of mostly repulsive interactions allows us to isolate event signatures: adhesion to micelle cores and furthermore the attachment, elongation and unbinding of single polymer strands. These signatures are inferred directly from force curves and visualized via overlays on AFM images. 

The approach enables materials scientists to uncover hidden patterns in their massively repeated experiments without requiring domain-specific preprocessing or assumptions, opening pathways for scalable analysis of generic serial measurement data.