η = shear stress / shear rate ≈ γ / (dv_x/dy) — from the velocity profile
Temperature (model units — liquid above ≈1300):
velocity profile vx(y): the flow the shear delivers. Linear (on the dashed line) ⇒ thermostat-limited (measuring friction Γ, not η); curved/parabolic ⇒ real viscous transport (η from the curvature).
Caveats. Model-time (no calibrated Pa·s); under-coordinated weak-H-bond water;
confined droplet + body-force shear (no periodic boundaries). This reads trends, not an absolute
viscosity. The honest test: does the dissipation rate fall as T rises (correct sign), and is it
higher at higher density (HDL more viscous)?