Nonlinearity-induced frequency mixing in AFM: novel contrast imaging with machine learning*

On complex thin films pertinent to lubrication and superhydrophilicity, dynamic AFM methods were used to sense both the elastic and viscous response to an AFM tip. A novel strategy in "multifrequency" AFM -- vibrating the AFM microcantilever (to which the tip is attached) at two tones near the fundamental flexural resonance -- generates dozens of intermodulation tones of response due to nonlinear tip-sample interaction (a well-known concept in electrical engineering usually in the context of AC signal distortion). This method¹ both expands contrast mechanisms -- with images of amplitude and phase at each mixing tone via 40 FPGA lock-in amplifiers -- and the ability to reconstruct the distance dependence of conservative and dissipative response at each image pixel via a 40-term discrete Fourier transform of tip motion. This reconstruction is an unprecedented capability in "force spectroscopy" (probing distance dependent interactions). Machine learning, to cluster force-spectroscopic fingerprints and thereby generate higher signal-to-noise images, is further discussed.

¹Intermodulation Products AB, Stockholm.