A nonconductive sample only becomes measurable through its metallic coating — which then conducts heat in parallel. The additive quantity is axial thermal conductance. Each point is a measurement at one film electrical conductance (thicker film → more parallel channel), and the composite conductance rises with the film’s share; the extrapolated intercept at zero film is the bare substrate’s conductance. The plotted differential extrapolation never imports a Lorenz number — the film term lives in the fitted slope. The slider drives a separate route for contrast: a single-point Wiedemann–Franz correction, whose result shifts with the assumed Lorenz number while the plot above stays untouched.
Model: composite AXIAL CONDUCTANCE with parallel channels, Kcomposite = Ks + Kf, the film term tied to its axial electrical conductance through Kf ∝ LfGeT (all normalized), Gaussian noise per point, least squares extrapolated to zero film. Two routes are deliberately contrasted: the plotted differential extrapolation uses no assumed Lorenz number (the film term is a fitted slope), while the slider illustrates the bias an assumed Lorenz number would inject into a separate single-point subtraction, evaluated first-order at the mid-range film share. Simplifications stated: film heat-capacity share, area/thickness errors, resistance error and their covariances are omitted. Real corrections work in the full composite model with the film’s measured resistance and, where accuracy demands, its measured effective Lorenz number. Schematic teaching tool, not an instrument.