Separating Motion Artifacts from Biomarker Signals in Flexible MoS2 FET Cortisol Sensors: A Coupled Multiphysics Model and Finite-Element Analysis of Strain-Binding Cross-Talk and Mitigation Strategies

Abstract Flexible molybdenum disulfide (MoS2) field-effect transistor (FET) biosensors can monitor cortisol in sweat, but their atomically thin channel is strongly piezoresistive, so body movement produces current changes that resemble hormone signals. We present a six-stage multiphysics model that links wrist and elbow flexion and walking to a strain-equivalent false cortisol concentration. Channel strain is computed by plane-strain finite-element analysis in the open-source solver Elmer, verified against laminate beam theory, and drives compact models of piezoresistance, Schottky-barrier lowering, strain-binding cross-talk, aptamer kinetics and Debye screening. For a 25 µm polyimide patch the model predicts wrist flexion produces a current artifact of 44% of the full-scale cortisol response and walking 5 to 11 ng/mL of false cortisol, with about 70% of the artifact arising at the Schottky contacts. Low-pass filtering leaves the non-zero mean of unipolar flexion (22% of full scale). Twin-sensor subtraction is limited by gauge mismatch and by a cross-talk floor proportional to aptamer occupancy, 1.0 to 2.8% of full scale across the physiological range at the wrist. With a 2.5 µm substrate and matched twin channels the residual falls to 0.9 to 1.3 ng/mL, not consistently below a 1 ng/mL target at the wrist for the assumed affinity. A rigid island under the channel retains 18 to 109% of the baseline strain under end-moment loading and about 300% when the skin imposes the curvature, so it fails to isolate the channel in bending. The predicted magnitudes scale with the assumed threshold shift; design guidelines by body location are given.