R.J. Sheridan, I. Saito, L.C. Brinson
Duke University,
United States
Keywords: nanoindentation, interphase, metrology, inference, interpretable ML
Summary:
Polymer nanocomposites have been shown to exhibit a variety of enhanced properties such as modulus and toughness. The origin of these enhancements has been assigned in part to a region of polymer material near the nanofiller interfaces with altered properties, known as interphase. Our group has previously measured the apparent elastic modulus of the interphase region with Atomic Force Microscopy Nanomechanical Mapping (AFM-NM) by constructing thin films of styrene-butadiene rubber on carbon substrates and cross-sectioning with argon ion miller. Through comparisons against FE indentation simulations at soft/rigid interfaces, a method was developed to quantitatively infer the size of interphase layers in the carbon/rubber system. This talk explores the generality of the interphase inference technique, by validating the underlying FE indentation simulation experimentally using defined-radius probes and an idealized soft/rigid material interface, including an artificial interphase with controlled stiffness and size. Our materials are an array of thermoset polymers with tunable glass transition temperatures and rubbery moduli, which we designed to obtain nearly ideal adhesive indentations. We analyze our results across three probe radii and five interphase thicknesses to find a scope of validity for interphase inference, defined in terms of a scale-free (nondimensional) description of indentation parameters. The key outcome of the analysis is an estimated lower bound on the length of an interphase near a rigid substrate that can be measured under any given AFM-NM imaging conditions. We then apply a physics-constrained, interpretable machine learning algorithm to the raw force-distance data to support intuition about the origin of the confounding physical effects that induce these metrology limits, and make some proposals for minimizing the limiting physical effects.