Why ice is so slippery

Abstract

The origin of ice's slipperiness has long puzzled scientists. To resolve this question, we simulate ice- glass (amorphous silica) friction at the nanoscale from first principles and upscale to the macroscale using a frictional heating model. We find that nanoscale simulations alone cannot capture the correct velocity dependence of ice friction, resulting in an overestimated coefficient of friction. By properly accounting for frictional heating, we find a strong increase in contact temperature toward the melting point, even under modest motion of 1 millimeter with velocities above 0.1 m/s, yielding excellent agreement with experimental friction data across a wide range of velocities. While the initial formation of a lubricating film on ice may occur without heating, the ultimate slipperiness of ice hinges on frictional heating, as proposed by Bowden and Hughes in 1939, but without incorporating melting.

Figures (7)

• Figure 1: Nanoscale vs. macroscale ice friction.a Setup for nanoscale MD simulations of amorphous silica sliding on ice. b Friction stress as a function of contact temperature for different sliding velocities from nanoscale MD simulations, with fitted equation \ref{['eq:high_friction']} (dashed lines). c Ice friction coefficient as a function of velocity for different temperatures with equation \ref{['eq:high_friction']} directly, without frictional heating, comparing with experiments. d Illustration of macroscale friction, where nanoscale represents one asperity contact. e Contact temperature as a function of velocity for different background temperatures using the frictional heating model (equation \ref{['eq:surface_temperature']}). f Macroscale ice friction coefficient using the frictional heating model, comparing with experiments. Contact temperatures for nanoscale MD simulations are reported in ℃ relative to the DC-R$^2$SCAN model melting point of 289K. Experimental friction coefficient data are from ice--glass friction experiments canale2019nanorheologymiyashita2023slidingakkokParametersAffectingKinetic1987 and curling stone experiments penner2001physicsnyberg2013asymmetrical, which is composed of granite with $\sim$70% content of silica.
• Figure 2: Nanoscale vs. macroscale premelting film properties.a premelting film thickness and b viscosity estimated directly from nanoscale MD simulations, and to macroscale using the frictional heating model. Nanoscale is reported for contact temperature, while macroscale is reported for background temperature.
• Figure 3: Parity plots for ice--glass friction modeling vs. data. a Nanoscale MD data vs. fitted model predictions for friction stress equation \ref{['eq:high_friction']}, corresponding to Figure \ref{['fig:upscaling']}b. b Experimental vs. modeled macroscale friction coefficient, corresponding to Figure \ref{['fig:upscaling']}f. c MD data vs. fitted model predictions for premelting film thickness, corresponding to Figure \ref{['fig:film-properties']}a.
• Figure S1: Setup for nanoscale ice--glass simulations.
• Figure S2: Validation of DC-R$^2$SCAN MLIP for physical properties.a Radial distribution function at 300K and 0.1MPa for experiment, and simulations using $NVT$ ensemble at experimental density, $NPT$ ensemble at 0.1MPa, 300K. b Density of liquid water and ice from experiments and simulations. c Potential energy for liquid--ice coexistence simulations at different temperatures. d Diffusion coefficient of liquid water for 0.1MPa for model and experiments holz2000temperatureeasteal1989diaphragmmills1973self. e Viscosity of liquid water at 0.1MPa for model and experiments Lemmon2021dehaoui2015viscosity. f Structure factor of amorphous silica from experiments and GAP model erhardMachinelearnedInteratomicPotential2022, and simulations at different annealing rates.