betaFeedback

WorksPhysics and Astronomy

Static Friction Coefficient Depends on the External Pressure and Block Shape due to Precursor Slip

Abstract

Amontons' law states that the maximum static friction force on a solid object is proportional to the loading force and is independent of the apparent contact area. This law indicates that the static friction coefficient does not depend on the external pressure or object shape. Here, we numerically investigate the sliding motion of a 3D viscoelastic block on a rigid substrate using the finite element method (FEM). The macroscopic static friction coefficient decreases with an increase in the external pressure, length, or width of the object, which contradicts Amontons' law. Precursor slip occurs in the 2D interface between the block and substrate before bulk sliding. The decrease in the macroscopic static friction coefficient is scaled by the critical area of the precursor slip before bulk sliding. A theoretical analysis of the simplified models reveals that bulk sliding results from the instability of the quasi-static precursor slip caused by velocity-weakening local friction. We also show that the critical slip area determines the macroscopic static friction coefficient, which explains the results of the FEM simulation.

Introduction

A friction force prevents the relative sliding motion between two objects in contact. Friction plays a crucial role in various situations, such as the contact surface between the ground and the sole of a shoe, brakes and bearings in machines, and tectonic plates that cause earthquakes. Many studies on friction have been conducted, but the elucidation of the fundamental mechanism of friction is essential for science and technology[1, 2] [3, 4, 5, 6, 7].

Amontons’ law states that the maximum static friction force on a solid object is independent of the apparent contact area and proportional to the load[1, 2] [3, 4, 5, 6, 7]. This law has been taught in high school physics textbooks and is believed to hold true for diverse systems. When the friction force obeys Amontons’ law, the friction coefficient, which is the ratio of the friction force to the loading force, does not depend on the pressure, size, or object shape. On a rough frictional interface with numerous asperities, only a tiny fraction of the surfaces forms junctions, the so-called real contact points. Amontons’ law is explained by the proportionality of the total area of real contact points to the loading force[1, 2] [3, 4, 5, 6, 7, 8, 9].

The above explanation for the origin of Amontons’ law implicitly assumes uniformity of the stress field. Therefore, Amontons’ law is not expected to hold if a macroscopic deformation exists. In fact, recent numerical studies have reported the breakdown of Amontons’ law in macroscopic viscoelastic objects10,11, revealing that it is related to local quasi-static precursor slips before the onset of bulk sliding owing to non-uniform deformation10–28. The relationship between precursor slips and the breakdown of Amontons’ law has been confirmed previously in an experiment with an acrylic glass block12. However, previous studies have only investigated systems with a 1D frictional interface. Friction usually occurs in 2D interfaces of 3D objects. However, it is not clear whether the results in previous studies apply to more realistic 3D systems.

In this study, we numerically investigate the sliding motion of a 3D viscoelastic object on a rigid substrate using the finite element method (FEM). The macroscopic static friction coefficient decreases with an increase in the pressure or size of the object. The precursor slip propagates in a 2D frictional interface. Bulk sliding occurs when the area of the precursor slip reaches a critical value, which determines the macroscopic static friction coefficient. An analysis of the simplified models reveals that the instability of the precursor slip leads to bulk sliding.

Results

3D FEM simulation

We numerically investigate a viscoelastic block on a rigid substrate with width WW, length LL, and height HH along the xx-, yy-, and zz-axes, respectively, as shown in Figure 1 (see Methods for details). The area of the frictional interface is denoted by A0=LWA_0 = LW. The density, Young’s modulus, and Poisson’s ratio of the block are denoted by ρ\rho, EE, and ν\nu, respectively. The dissipation in the block is characterized by two viscosity coefficients: η1\eta_1 and η2\eta_2. We assume that Amontons’ law holds locally at the interface between the block and the rigid substrate (z=0z = 0), and the magnitude of the local frictional stress, σ(fric)(x,y)\sigma^{(\mathrm{fric})}(x,y) in the interface is locally determined as

Schematic of a 3D viscoelastic block on a fixed rigid substrate

Figure 1. Schematic of a 3D viscoelastic block on a fixed rigid substrate.

σ(fric)(x,y)=μ(v(x,y))p(x,y),(1)\sigma^{(\mathrm{fric})}(x,y) = \mu(v(x,y))p(x,y), \tag*{(1)}

where p(x,y)p(x,y) is the bottom pressure, and μ(v)\mu(v) is the friction coefficient, which depends on the magnitude of the local slip velocity v(x,y)v(x,y) when v(x,y)≠0v(x,y) \ne029. Here, μ(v)\mu(v) is characterized by the characteristic velocity of velocity-weakening friction vcv_{\mathrm{c}} and the local static and kinetic friction coefficients denoted by μS\mu_{\mathrm{S}} and μK\mu_{\mathrm{K}} (see Methods). The rigid rod quasi-statically pushes the center of the side surface along the yy direction. The macroscopic friction force FTF_{\mathrm{T}} is measured as the force on the rigid rod in the yy direction. The loading force applied to the top of the block is given by FN=PextA0F_{\mathrm{N}} = P_{\mathrm{ext}}A_0 with the external pressure to the top surface PextP_{\mathrm{ext}}.

The ratio FT/FNF_{\mathrm{T}}/F_{\mathrm{N}} is plotted against the displacement of the rigid rod UU for L/H=1L/H=1, W/H=2W/H=2, and Pext/E=0.006P_{\mathrm{ext}}/E=0.006 in Figure 2 (a). First, FT/FNF_{\mathrm{T}}/F_{\mathrm{N}} increases linearly with UU. After obtaining a maximum value lower than μS\mu_{\mathrm{S}}, FT/FNF_{\mathrm{T}}/F_{\mathrm{N}} rapidly decreases to a value close to μK\mu_{\mathrm{K}}. This rapid drop is associated with bulk sliding. The significant drop after the linear increase periodically repeats itself. This periodic behavior corresponds to the stick-slip motion of the object. The maximum value of FT/FNF_{\mathrm{T}}/F_{\mathrm{N}} represents the macroscopic static friction coefficient, μM\mu_{\mathrm{M}}. Figure 2 (b) and (Figure 2c) display the macroscopic static friction coefficient μM\mu_{\mathrm{M}} against pressure PextP_{\mathrm{ext}} for various L/HL/H and W/HW/H values, respectively. The magnitude of μM\mu_{\mathrm{M}} decreases with increasing PextP_{\mathrm{ext}}, which is qualitatively consistent with the results for a system with a 1D friction interface in ref. 10. The previous study reported the size dependence of μM\mu_{\mathrm{M}} while maintaining the aspect ratio L/H=2L/H=21, whereas Figure 2 (b) and (Figure 2c) demonstrate that the friction coefficient μM\mu_{\mathrm{M}} also decreases with increasing aspect ratios L/HL/H and W/HW/H. These results indicate that Amontons’ law breaks down in systems with 2D interfaces.

Ratio $F_{\mathrm{T}}/F_{\mathrm{N}}$ against displacement and macroscopic static friction coefficients against pressure

Figure 2. (a) Ratio FT/FNF_{\mathrm{T}}/F_{\mathrm{N}} against the displacement of the rigid rod UU for L/H=1L/H = 1, W/H=2W/H = 2, and Pext/E=0.006P_{\mathrm{ext}}/E = 0.006. The red horizontal line represents the macroscopic static friction coefficient μM\mu_{\mathrm{M}}. (b) Macroscopic static friction coefficient μM\mu_{\mathrm{M}} against pressure PextP_{\mathrm{ext}} for various L/HL/H values with W/H=1W/H = 1. The thin solid lines represent the analytical results with αA=0.2\alpha_{\mathrm{A}} = 0.2 given by equations (4) and (6). (c) Macroscopic static friction coefficient μM\mu_{\mathrm{M}} against PextP_{\mathrm{ext}} for various W/HW/H values with L/H=1L/H = 1. The thin solid lines represent the analytical results with αB=0.2\alpha_{\mathrm{B}} = 0.2 given by equations (6) and (11). The dotted and dashed lines represent μS\mu_{\mathrm{S}} and μK\mu_{\mathrm{K}}, respectively.

Figure 3 (a) shows the spatial distribution of the slip region with nonzero slip velocity in the frictional interface at z=0z=0 for U=U1,U2,U3U=U_1,U_2,U_3, and U4U_4 shown in Figure 2 (a). Here, we choose U1/L=50×10−3U_1/L=50\times10^{-3}, U2/L=56×10−3U_2/L=56\times10^{-3}, U3/L=61.38×10−3U_3/L=61.38\times10^{-3}, and U4/L=62.71×10−3U_4/L=62.71\times10^{-3}, which corresponds to the stationary stick-slip region. See Methods for the definition of the slip region. In Figure 3 (a), the local precursor slip starts from the region under the rigid rod for U=U1U=U_1. As UU increases (U2U_2 and U3U_3), the region expands gradually. After U=U3U=U_3, the entire area slips with v>vcv>v_{\mathrm{c}}, resulting in bulk sliding. Note that the slip occurs almost along the yy direction. Figure 3 (b) shows the area of precursor slip AA normalized by the area of frictional interface A0A_0 against displacement UU. First, the area of the precursor slip increases gradually with displacement UU. When the area AA reaches the critical area AcA_{\mathrm{c}} just before bulk sliding (dotted line), the propagation speed of the area suddenly increases. Owing to rapid propagation, AA reaches A0A_0 and then returns to 0. We demonstrate the normalized critical area Ac/A0A_{\mathrm{c}}/A_0 against pressure PextP_{\mathrm{ext}} in Figure 3 (c) and (Figure 3d) for various L/HL/H values with W/H=1W/H=1 and for various W/HW/H values with L/H=1L/H=1, respectively. The normalized critical area Ac/A0A_{\mathrm{c}}/A_0 decreases as PextP_{\mathrm{ext}}, L/HL/H, or W/HW/H increases. This decrease is similar to that of μM\mu_{\mathrm{M}} in Figure 2 (b) and (Figure 2c), respectively.

Spatial distribution of the slip region, normalized precursor slip area, and normalized critical area

Figure 3. (a) Spatial distribution of the slip region in the frictional interface at U=U1,U2,U3U=U_1,U_2,U_3, and U4U_4 for L/H=1L/H=1, W/H=2W/H=2, and Pext/E=0.006P_{\mathrm{ext}}/E=0.006. The yellow area represents the slip region. The rigid rod is pushing the block at (x/H,y/H)=(1,0)(x/H,y/H)=(1,0). (b) Normalized precursor slip area A/A0A/A_0 against displacement UU. The dotted line represents the normalized critical area Ac/A0A_{\mathrm{c}}/A_0. (c) Normalized critical area Ac/A0A_{\mathrm{c}}/A_0 against pressure PextP_{\mathrm{ext}} for various L/HL/H values with W/H=1W/H=1. The thin solid lines represent the analytical results with αA=0.2\alpha_{\mathrm{A}}=0.2 given by equation (4). (d) Normalized critical area Ac/A0A_{\mathrm{c}}/A_0 against PextP_{\mathrm{ext}} for various W/HW/H values with L/H=1L/H=1. The thin solid lines represent the analytical results with αB=0.2\alpha_{\mathrm{B}}=0.2 given by equation (11).

In Figure 4, we present the macroscopic friction coefficient μM\mu_{\mathrm{M}} against the normalized critical area Ac/A0A_{\mathrm{c}}/A_0 for various L/HL/H and W/HW/H values. The macroscopic friction coefficient μM\mu_{\mathrm{M}} for different L/HL/H and W/HW/H values approximately collapses onto a master curve, which indicates a linear increase in μM\mu_{\mathrm{M}} with Ac/A0A_{\mathrm{c}}/A_0. The minimum value close to Ac/A0=0A_{\mathrm{c}}/A_0=0 is almost equal to μK\mu_{\mathrm{K}}, whereas the maximum value at Ac/A0=1A_{\mathrm{c}}/A_0=1 is equal to μS\mu_{\mathrm{S}}.

Macroscopic static friction coefficient $\mu_\mathrm{M}$ against normalized critical area $A_\mathrm{c}/A_0$

Figure 4. Macroscopic static friction coefficient μM\mu_\mathrm{M} against the normalized critical area Ac/A0A_\mathrm{c}/A_0 for various L/HL/H and W/HW/H values. The solid line represents the analytical result given by equation (6). The dotted and dashed lines represent μS\mu_\mathrm{S} and μK\mu_\mathrm{K}, respectively.

Figure 5 shows the spatial distribution of the ratio σ(fric)/p\sigma^{(\mathrm{fric})}/p in the frictional interface for L/H=1L/H=1, W/H=2W/H=2, and Pext/E=0.006P_\mathrm{ext}/E=0.006 at U=U1,U2,U3U=U_1,U_2,U_3, and U4U_4. It should be noted that the direction of the frictional stress is almost opposite to the driving direction, that is, the yy direction. In the no-slip region, the local static friction can take any value for 0<σ(fric)/p<μS0<\sigma^{(\mathrm{fric})}/p<\mu_\mathrm{S}. Before the onset of precursor slip, that is, just after bulk sliding, σ(fric)/p\sigma^{(\mathrm{fric})}/p takes a value almost equal to μK\mu_\mathrm{K}, the local kinetic friction coefficient, in the entire interface, as explained below. At U=U1U=U_1, σ(fric)/p\sigma^{(\mathrm{fric})}/p reaches the local static friction coefficient, μS\mu_\mathrm{S}, near the rigid rod at (x/H,y/H)=(1,0)(x/H,y/H)=(1,0). As the displacement UU increases to U2U_2 and U3U_3, the area with σ(fric)/p≈μS\sigma^{(\mathrm{fric})}/p\approx\mu_\mathrm{S} gradually increases. The region of σ(fric)/p≈μS\sigma^{(\mathrm{fric})}/p\approx\mu_\mathrm{S} coincides with the local precursor slip region in Figure 3 (a). Except for the slip region, σ(fric)/p\sigma^{(\mathrm{fric})}/p remains approximately at μK\mu_\mathrm{K}. Immediately after U3U_3, bulk sliding with v>vcv>v_\mathrm{c} occurs, and the fast slip leads to σ(fric)/p=μK\sigma^{(\mathrm{fric})}/p=\mu_\mathrm{K} at U4U_4. Bulk sliding rapidly decelerates, and the slip velocity vv decreases to 0, when σ(fric)/p\sigma^{(\mathrm{fric})}/p increases to μS\mu_\mathrm{S} in the frictional interface. However, the internal deformation is not able to follow the rapid change, and the ratio of static frictional stress to bottom pressure finally returns to σ(fric)/p≈μK\sigma^{(\mathrm{fric})}/p\approx\mu_\mathrm{K} after bulk sliding. Consequently, σ(fric)/p\sigma^{(\mathrm{fric})}/p is almost equal to μK\mu_\mathrm{K} after bulk sliding. The macroscopic static friction coefficient μM\mu_\mathrm{M} is approximately expressed by the average of σ(fric)/p\sigma^{(\mathrm{fric})}/p over the entire frictional interface at U3U_3 immediately before bulk sliding. This result explains the dependence of μM\mu_\mathrm{M} on Ac/A0A_\mathrm{c}/A_0 shown in Figure 4, where μM\mu_\mathrm{M} approaches μS\mu_\mathrm{S} for Ac/A0=1A_\mathrm{c}/A_0=1.

Spatial distribution of the ratio of frictional stress to bottom pressure $\sigma^{(\mathrm{fric})}/p$

Figure 5. Spatial distribution of the ratio of frictional stress to bottom pressure σ(fric)/p\sigma^{(\mathrm{fric})}/p in the frictional interface for L/H=1L/H=1, W/H=2W/H=2, and Pext/E=0.006P_\mathrm{ext}/E=0.006 at U=U1,U2,U3U=U_1,U_2,U_3, and U4U_4. The rigid rod is pushing the block at (x/H,y/H)=(1,0)(x/H,y/H)=(1,0). The white area represents the region with p=0p=0 due to the lift of the bottom.

Analysis based on simplified models

To theoretically analyze the numerical results, we employ two simplified models, which explain the dependence of μM\mu_\mathrm{M} on L/HL/H and W/HW/H (see Supplementary Note online for details).

Model for large L/HL/H

To discuss the behavior of increasing L/HL/H while maintaining W/H=1W/H=1, we employ a 1D effective model, as shown in Figure 6 (a). The slip region propagates along the yy direction, as shown in the Supplementary Note and Supplementary Video S1. Therefore, in this model, the degrees of freedom in the zz and xx directions are neglected by assuming W/H≪L/HW/H \ll L/H, and the deformation is characterized only by the yy-dependent displacement in the yy direction, uy(y,t)u_y(y,t), at the interface z=0z=0. We also assume a uniform bottom pressure PextP_{\mathrm{ext}}. The equation of motion is given by

Schematics of simplified models for (a) $L/H \gg 1$ and (b) $W/H \gg 1$

Figure 6. Schematics of simplified models for (a) L/H≫1L/H \gg1 and (b) W/H≫1W/H \gg1.

ρu¨y(y,t)=∂σyy(y,t)∂y−μ(u˙y(y,t))PextαAH.(2)\rho\ddot{u}_y(y,t)=\frac{\partial\sigma_{yy}(y,t)}{\partial y}-\frac{\mu(\dot{u}_y(y,t))P_{\mathrm{ext}}}{\alpha_A H}. \tag*{(2)}

where u¨y\ddot{u}_y and u˙y\dot{u}_y are the second- and first-order time derivatives of the displacement, respectively. Here, αA\alpha_A represents the effect of the block thickness and is treated as a fitting parameter. The normal stress σyy\sigma_{yy} is given by

σyy(y,t)=E1∂uy∂y+ηt∂u˙y∂y(3)\sigma_{yy}(y,t)=E_1\frac{\partial u_y}{\partial y}+\eta_t\frac{\partial\dot{u}_y}{\partial y} \tag*{(3)}

with the elastic constant E1=E/{(1+ν)(1−ν)}E_1=E/\{(1+\nu)(1-\nu)\} and viscous constant ηt=η1(η1+2η2)/(η1+η2)\eta_t=\eta_1(\eta_1+2\eta_2)/(\eta_1+\eta_2) in the plane stress state by considering the block as a thin plate (see Methods).

The quasi-static solution uy(a)(y)u_y^{(a)}(y) of equation (3) with u¨y=u˙y=0\ddot{u}_y=\dot{u}_y=0 is obtained analytically, where the precursor slip area AA increases with UU (see Supplementary Note). A linear stability analysis reveals that the quasi-static solution becomes unstable, and bulk sliding occurs when AA reaches the critical area AcA_c owing to the competition between velocity-weakening friction and viscosity. The critical area AcA_c satisfies

π2ηt(AcA0)−2+2πLρE1(AcA0)−1=(μS−μK)PextL2vcαAH(4)\pi^2\eta_t\left(\frac{A_c}{A_0}\right)^{-2}+2\pi L\sqrt{\rho E_1}\left(\frac{A_c}{A_0}\right)^{-1}=\frac{(\mu_S-\mu_K)P_{\mathrm{ext}}L^2}{v_c\alpha_A H} \tag*{(4)}

(see Supplementary Note). For Ac/A0≪1A_c/A_0\ll1, equation (4) yields

AcA0≃π(μS−μKαA)−1/2(PextHηtvc)−1/2(LH)−1.(5)\frac{A_c}{A_0}\simeq\pi\left(\frac{\mu_S-\mu_K}{\alpha_A}\right)^{-1/2}\left(\frac{P_{\mathrm{ext}}H}{\eta_t v_c}\right)^{-1/2}\left(\frac{L}{H}\right)^{-1}. \tag*{(5)}

This equation indicates that the normalized critical area Ac/A0A_c/A_0 decreases as L/HL/H or PextP_{\mathrm{ext}} increases, which is consistent with the FEM results shown in Figure 3 (c). We plot Ac/A0A_c/A_0 obtained from equation (4) as thin solid lines in Figure 3 (c) by choosing the fitting parameter αA=0.2\alpha_A=0.2 to match the results of the FEM simulations. The analytical results semi-quantitatively reproduce the numerical results except for L/H=1L/H=1.

The quasi-static solution uy(a)(y)u_y^{(a)}(y) yields

μM=μK+(μS−μK)AcA0.(6)\mu_M=\mu_K+(\mu_S-\mu_K)\frac{A_c}{A_0}. \tag*{(6)}

This is consistent with the FEM simulations, as shown by the solid line in Figure 4. For Ac/A0≪1A_c/A_0\ll1, substituting equation (5) into this equation, we obtain

μM−μK≃π(μS−μK)12αA12(PextHηtvc)−12(LH)−1.(7)\mu_M-\mu_K\simeq\pi(\mu_S-\mu_K)^{\frac{1}{2}}\alpha_A^{\frac{1}{2}}\left(\frac{P_{\mathrm{ext}}H}{\eta_t v_c}\right)^{-\frac{1}{2}}\left(\frac{L}{H}\right)^{-1}. \tag*{(7)}

This equation indicates that the macroscopic static friction coefficient μM\mu_{\mathrm{M}} decreases as PextP_{\mathrm{ext}} or L/HL/H increases. We plot μM\mu_{\mathrm{M}} given by equations (4) and (6) as thin solid lines in Figure 2 (b), which semi-quantitatively reproduces the results of the FEM simulations except for L/H=1L/H=1.

In a previous study10, μM\mu_{\mathrm{M}} is obtained analytically as

μM−μK≃π2π2−4(μS−μK)23α13(PextHη1vc)−13(LH)−23.(8)\mu_{\mathrm{M}}-\mu_{\mathrm{K}} \simeq\frac{\pi^{2}}{\pi^{2}-4}(\mu_{\mathrm{S}}-\mu_{\mathrm{K}})^{\frac{2}{3}}\alpha^{\frac{1}{3}}\left(\frac{P_{\mathrm{ext}}H}{\eta_{1}v_{\mathrm{c}}}\right)^{-\frac{1}{3}}\left(\frac{L}{H}\right)^{-\frac{2}{3}} . \tag*{(8)}

for Ac/A0≪1A_{\mathrm{c}}/A_{0}\ll1 in a system with small L/HL/H. Here, α\alpha is the fitting parameter. The power-law exponents in equation (8) for the dependence on PextP_{\mathrm{ext}} and L/HL/H differ from those in equation (7). The present model assumes L/H≫1L/H\gg1, which results in a uniform bottom pressure, as shown in the Supplementary Note. For a small L/HL/H, the bottom pressure increases along the driving direction owing to the torque effect10,21, and the analytical results deviate from those of FEM due to the non-uniform pressure as shown in Figure 2 (b) and Figure 3 (c), which leads to different exponents from those in the present study.

Model for large W/HW/H

To discuss the behavior of increasing W/HW/H while maintaining L/H=1L/H=1, we employ a 1D effective model, as shown in Figure 6 (b). For W/H≫1W/H\gg1, the slip region propagates along the xx direction, as shown in Supplementary Note and Supplementary Video S2. Hence, in this model, we characterize the deformation only by the xx-dependent displacement in the yy direction, uy(x,t)u_y(x,t), at the interface z=0z=0 by assuming L/H≪W/HL/H\ll W/H. We also assume the rod size is sufficiently small and negligible. The equation of motion is given by

ρu¨y(x,t)=∂σxy(x,t)∂x−μ(u˙y(x,t))PextαBH.(9)\rho\ddot{u}_y(x,t)=\frac{\partial\sigma_{xy}(x,t)}{\partial x}-\frac{\mu(\dot{u}_y(x,t))P_{\mathrm{ext}}}{\alpha_{\mathrm{B}}H}. \tag*{(9)}

Here, αB\alpha_{\mathrm{B}} represents the effect of the block thickness and is treated as a fitting parameter. The shear stress σxy\sigma_{xy} is given by

σxy=E2∂uy∂x+η12∂u˙y∂x(10)\sigma_{xy}=E_{2}\frac{\partial u_y}{\partial x}+\frac{\eta_{1}}{2}\frac{\partial\dot{u}_y}{\partial x} \tag*{(10)}

with the elastic constant E2=E/{2(1+ν)}E_{2}=E/\{2(1+\nu)\} and the viscous constant η1/2\eta_{1}/2 (see Methods).

The quasi-static solution uy(a)(x)u_y^{(\mathrm{a})}(x) is also obtained analytically, where the precursor slip area AA increases with the value of UU (see Supplementary Note). The linear stability analysis reveals that the precursor slip becomes unstable, and bulk sliding occurs when AA reaches the critical area AcA_{\mathrm{c}} satisfying

2π2η1(AcA0)−2+4πWρE2(AcA0)−1=(μS−μK)PextW2vcαBH.(11)2\pi^{2}\eta_{1}\left(\frac{A_{\mathrm{c}}}{A_{0}}\right)^{-2}+4\pi W\sqrt{\rho E_{2}}\left(\frac{A_{\mathrm{c}}}{A_{0}}\right)^{-1}=\frac{(\mu_{\mathrm{S}}-\mu_{\mathrm{K}})P_{\mathrm{ext}}W^{2}}{v_{\mathrm{c}}\alpha_{\mathrm{B}}H}. \tag*{(11)}

For Ac/A0≪1A_{\mathrm{c}}/A_{0}\ll1, this equation yields

AcA0≃π(μS−μKαB)−12(PextH2η1vc)−12(WH)−1.(12)\frac{A_{\mathrm{c}}}{A_{0}}\simeq\pi\left(\frac{\mu_{\mathrm{S}}-\mu_{\mathrm{K}}}{\alpha_{\mathrm{B}}}\right)^{-\frac{1}{2}}\left(\frac{P_{\mathrm{ext}}H}{2\eta_{1}v_{\mathrm{c}}}\right)^{-\frac{1}{2}}\left(\frac{W}{H}\right)^{-1}. \tag*{(12)}

The power-law exponents for the pressure and aspect ratio are the same as those in equation (5). This equation indicates that Ac/A0A_{\mathrm{c}}/A_{0} decreases as PextP_{\mathrm{ext}} or W/HW/H increases. We plot Ac/A0A_{\mathrm{c}}/A_{0} given by equation (11) as thin solid lines in Figure 3 (d), which semi-quantitatively reproduces the results of the FEM analysis by choosing αB=0.2\alpha_{\mathrm{B}}=0.2 except for W/H=1W/H=1. For small W/HW/H, the size of the rod and the yy-dependence of the displacement become relevant, which leads to the deviation between the numerical and theoretical results.

The macroscopic static friction coefficient μM\mu_{\mathrm{M}} is given by equation (6). For Ac/A0≪1A_{\mathrm{c}}/A_{0}\ll1, substituting equation (12) into equation (6), we obtain

μM−μK≃π(μS−μK)12αB12(PextH2η1vc)−12(WH)−1.(13)\mu_{\mathrm{M}}-\mu_{\mathrm{K}}\simeq\pi(\mu_{\mathrm{S}}-\mu_{\mathrm{K}})^{\frac{1}{2}}\alpha_{\mathrm{B}}^{\frac{1}{2}}\left(\frac{P_{\mathrm{ext}}H}{2\eta_{1}v_{\mathrm{c}}}\right)^{-\frac{1}{2}}\left(\frac{W}{H}\right)^{-1}. \tag*{(13)}

The macroscopic static friction coefficient μM\mu_{\mathrm{M}} decreases as PextP_{\mathrm{ext}} or W/HW/H increases. The thin solid lines shown in Figure 2 (c) are given by equations (6) and (11), and they semi-quantitatively reproduce the results of the FEM simulations except for W/H=1W/H=1.

Discussion

In this study, we numerically investigate the sliding motion of a 3D viscoelastic object using the FEM. The critical area of the precursor slip and macroscopic static friction coefficient decrease with an increase in the external pressure, length, or width of the object. The analysis based on the simplified models reveals that the stability condition determines the critical area of the precursor slip owing to the competition between the velocity-weakening friction and viscosity. The analysis explains the dependence of macroscopic static friction in the FEM simulations.

In a previous study10, the aspect ratio of the system is fixed at L/H=2L/H = 2 to investigate the size and load dependences of the precursor slip and the breakdown of Amontons’ law. For L/H=2L/H = 2, the nonuniformity of the bottom pressure is remarkable, which is considered to be the origin of the precursor slip and the breakdown of Amontons’ law. However, the present results with various aspect ratios show that the nonuniformity of shear stress also causes these behaviors without non-uniform pressure. Although the model considered in the previous study reproduces the results of systems with a smaller L/HL/H better, the simplified model in this study is more appropriate for systems with a large L/HL/H (see Supplementary Note).

The parameters for the FEM simulations employed here are those of a virtual material, and different from those of poly methyl methacrylate (PMMA) employed in experiments12, 15. We choose them to compare our results with the 2D simulations of previous studies10 and to reduce the computational load (see Methods). It also should be noted that the driving rod employed in experiments is hard but has finite stiffness, which is different from the rigid rod used in this study. The effect of the finite stiffness of the driving rod is considered to be small because it is taken into account as a deformation of the viscoelastic block around the driving point. In addition, we have ignored the aging effect7 in the local friction model because a previous experiment using PMMA15 indicates that the time scale of the aging is larger than that of the stick of the macroscopic stick-slip motion. The difference in the parameters, the driving method, and the local friction model may affect our results. However, FEM simulations employing similar parameters semi-quantitatively reproduce the external pressure dependence of the macroscopic static friction coefficient obtained in the experiment using PMMA12. The dependence of the macroscopic static friction coefficient on the aspect ratio for PMMA is also considered to be consistent with our present results. The dependence on material parameters, the driving methods, and the local friction model will be investigated in future work.

The dependence of the static and kinetic friction coefficients on the pressure or block shape has been studied in experiments using rubber blocks30–32. The results of these experiments are partially consistent with ours, but there is a difference in the dependence on the aspect ratio. In these experiments, the methods to change the aspect ratio and drive the block differ from those used in this study. For the rubber block, the local Amontons’ law used in this study may not be applicable because the real contact area can become comparable to the apparent contact area, which contradicts the assumption of the Amontons’ law. We need further investigations to determine the origin of the difference.

Recent numerical simulations of spring-block models have shown that the friction coefficient changes with the geometric pattern of the frictional interface33–36. However, our results indicate that an object shape can also control the macroscopic static friction coefficient. This might lead to new insights into methods for controlling friction in various objects, including shoe soles and tires.

Precursor slip has been investigated experimentally for the sliding motion of PMMA blocks based on fracture mechanics26, 37–41. Such a precursor slip is related to pre-earthquakes that occur a few days or months before a major earthquake42–44, which are studied using frictional spring-block models45. However, these studies have focused on 1D frictional interfaces or discrete models, which differ from 2D friction interfaces in more realistic systems. Our results for a 3D system with a 2D interface will provide new insights into the precursor slip observed in realistic situations.

Methods

Setting of system

The equation of motion for a viscoelastic body is given by

ρu¨=∇⋅σ(14)\rho\ddot{u} = \nabla\cdot\boldsymbol{\sigma} \tag*{(14)}

with displacement u\boldsymbol{u}, stress σ\boldsymbol{\sigma}, and second-order time derivative u¨\ddot{\boldsymbol{u}} of displacement. The stress σ\boldsymbol{\sigma} is given by the sum of the elastic stress σ(E)\boldsymbol{\sigma}^{(\mathrm{E})} obeying Hooke’s law and the viscous stress σ(V)\boldsymbol{\sigma}^{(\mathrm{V})}, which is proportional to the strain rate. We assume that the viscoelastic body is isotropic. The elastic stress tensor σij(E)\sigma_{ij}^{(\mathrm{E})} is given by

σij(E)=E1+νϵij+νE(1+ν)(1−2ν)ϵkkδij(15)\sigma_{ij}^{(\mathrm{E})} = \frac{E}{1+\nu}\epsilon_{ij} + \frac{\nu E}{(1+\nu)(1-2\nu)}\epsilon_{kk}\delta_{ij} \tag*{(15)}

with the Kronecker delta δij\delta_{ij} and the strain tensor ϵij\epsilon_{ij}. The viscous stress tensor σij(V)\sigma_{ij}^{(\mathrm{V})} is given by

σij(V)=η1ϵ˙ij+η2ϵ˙kkδij(16)\sigma_{ij}^{(\mathrm{V})} = \eta_{1}\dot{\epsilon}_{ij} + \eta_{2}\dot{\epsilon}_{kk}\delta_{ij} \tag*{(16)}

with the strain rate tensor ε˙ij\dot{\varepsilon}_{ij}46. The boundary conditions for the top surface at z=Hz = H are σzz=−Pext\sigma_{zz} = -P_{\mathrm{ext}} and σzx=σzy=0\sigma_{zx} = \sigma_{zy} = 0. At the free surface for x=0,Wx = 0,W or y=0,Ly = 0,L, we assume σ⋅n=0\boldsymbol{\sigma} \cdot\mathbf{n} = 0 with the normal vector n\mathbf{n} of the surface. The boundary conditions at the contact surface with a rigid rod (y=0y = 0) are given by σyx=σyz=0\sigma_{yx} = \sigma_{yz} = 0 and u˙y=Vrod\dot{u}_y = V_{\mathrm{rod}}, where u˙y\dot{u}_y is the velocity in the yy direction and VrodV_{\mathrm{rod}} is the velocity of rigid rod. At the bottom of the block (z=0z = 0) in contact with a rigid substrate, the bottom pressure p=−σzzp = -\sigma_{zz} is determined such that the displacement uzu_z in the zz direction is 0. However, the bottom pressure is limited to p≥0p \ge0. The region of the bottom surface with uz>0u_z > 0 and p=0p = 0 becomes a free surface with σ⋅n=0\boldsymbol{\sigma} \cdot\mathbf{n} = 0. The boundary condition in the tangential direction at the bottom with p>0p > 0 is given by

t=−vvμ(v)p(17)\mathbf{t} = -\frac{\mathbf{v}}{v}\mu(v)p \tag*{(17)}

with the tangential stress vector t(x,y)=(σzx,σzy)\mathbf{t}(x,y) = (\sigma_{zx},\sigma_{zy}), local slip velocity vector v(x,y)=(u˙x,u˙y)\mathbf{v}(x,y) = (\dot{u}_x,\dot{u}_y), velocity u˙x\dot{u}_x in the xx direction, and velocity u˙y\dot{u}_y in the yy direction. The direction of the frictional stress is opposite to that of the local slip velocity. Frictional stress is defined as σ(fric)(x,y)=∣t∣\sigma^{(\mathrm{fric})}(x,y) = |\mathbf{t}|. The slip velocity is defined as v(x,y)=∣v(x,y)∣v(x,y) = |\mathbf{v}(x,y)|.

The frictional stress σ(fric)\sigma^{(\mathrm{fric})} is given by equation (1). In the case v(x,y)=0v(x,y) = 0, the frictional stress is balanced with the local shear stress, where the maximum magnitude of the former is given by μSp(x,y)\mu_{\mathrm{S}}p(x,y). The local friction coefficient μ(v)\mu(v) linearly decreases from μS\mu_{\mathrm{S}} to μK\mu_{\mathrm{K}} for 0<v≤vc0 < v \le v_{\mathrm{c}} and μK\mu_{\mathrm{K}} for v>vcv > v_{\mathrm{c}}. Amontons’ law is expected to hold locally if the local region considered in the frictional interface contains a sufficiently large number of real contact points and has negligibly small spatial variations in internal stress8, 9, 47.

To treat static friction in the numerical simulation, we introduce a small velocity scale vev_{\mathrm{e}}. The local friction coefficient μ(v)\mu(v) is given by

μ(v)={μSv/ve,0≤v≤veμS−(μS−μK)v/vc,ve<v<vcμK,v≥vc(18)\mu(v) = \begin{cases} \mu_{\mathrm{S}}v/v_{\mathrm{e}}, & 0 \le v \le v_{\mathrm{e}} \\ \mu_{\mathrm{S}}-(\mu_{\mathrm{S}}-\mu_{\mathrm{K}})v/v_{\mathrm{c}}, & v_{\mathrm{e}} < v < v_{\mathrm{c}} \\ \mu_{\mathrm{K}}, & v \ge v_{\mathrm{c}} \end{cases} \tag*{(18)}

We consider the limit ve→+0v_{\mathrm{e}} \to+0. The region with 0≤v≤ve0 \le v \le v_{\mathrm{e}} corresponds to static friction. The slip area AA is defined as the region with v>vev > v_{\mathrm{e}}.

Details of 3D FEM simulation

The viscoelastic block is divided into cubes with length Δx\Delta x consisting of six tetrahedra. The displacements and velocities within each element are approximated using a linear interpolation. We choose the characteristic velocity ve/Vrod=2.5×10−2v_{\mathrm{e}}/V_{\mathrm{rod}} = 2.5 \times10^{-2} such that ve/Vrod≪1v_{\mathrm{e}}/V_{\mathrm{rod}} \ll1 is satisfied. In the FEM simulations, we select Δx/H=1/40\Delta x/H = 1/40, Δt/(Hρ/E)≈10−6\Delta t/(H\sqrt{\rho/E}) \approx10^{-6}, and Vrodρ/E=2×10−5V_{\mathrm{rod}}\sqrt{\rho/E} = 2 \times10^{-5}. We have confirmed that the numerical results do not change, even if we use smaller values.

First, we apply an external uniform pressure PextP_{\mathrm{ext}} to the top surface and relax the system to an equilibrium state. After relaxation, the center of the side surface (x,y,z)=(W/2,0,H/2)(x,y,z) = (W/2,0,H/2) is pushed along the yy direction by a rigid rod from time t=0t = 0 with a sufficiently slow speed VrodV_{\mathrm{rod}}. The displacement of the rigid rod is denoted by U(t)=VrodtU(t) = V_{\mathrm{rod}}t. The length of one side of a rigid square rod is 0.1H0.1H, and the height of its center from the bottom is 0.5H0.5H.

Details of analysis based on simplified models

Model for large L/HL/H: The second term on the right-hand side of equation (2) represents local friction. Here, we assume a constant bottom pressure given by PextP_{\mathrm{ext}}, which is verified in the FEM simulations for L/H≫1L/H \gg1 as shown in the Supplementary Note and Supplementary Video S1. The local friction coefficient μ\mu is expressed as a function of v=∣u˙y∣v = |\dot{u}_y|. Note that 0≤μ≤μS0 \le\mu\le\mu_{\mathrm{S}} when v=0v = 0. The boundary conditions are ∂uy(y=L,t)/∂y=0\partial u_y(y = L,t)/\partial y = 0 and uy(y=0,t)=U(t)u_y(y = 0,t) = U(t). In our analysis, we set the origin of UU immediately after the bulk sliding and assume that the ratio of the frictional stress to PextP_{\mathrm{ext}} is equal to μK\mu_{\mathrm{K}} at U=0U = 0.

Model for large W/HW/H: The second term on the right-hand side of equation (9) represents the friction. The bottom pressure is almost independent of xx in the FEM simulations, as shown in the Supplementary Note and Supplementary Video S2. Therefore, we assume a constant bottom pressure given by PextP_{\mathrm{ext}}. The boundary conditions are ∂uy(∣x∣=W/2,t)/∂x=0\partial u_y(|x| = W/2,t)/\partial x = 0 and uy(x=0,t)=U(t)u_y(x = 0,t) = U(t).

Parameters

The parameters for the viscoelastic object are chosen as ν=0.34\nu= 0.34, η1/(HρE)=2\eta_1/(H\sqrt{\rho E}) = 2, and η2/η1=1\eta_2/\eta_1 = 1, whereas we set the parameters for the friction as μS=0.38\mu_{\mathrm{S}} = 0.38, μK=0.1\mu_{\mathrm{K}} = 0.1, and vcρ/E=3.4×10−4v_{\mathrm{c}}\sqrt{\rho/E} = 3.4 \times10^{-4}, following previous FEM simulations10. These values are different from those adopted for the experiment using PMMA10, 12. The parameters for the PMMA blocks12 are estimated as L/H=5L/H = 5, W/H=0.25W/H = 0.25, Pext/E≈3×10−4P_{\mathrm{ext}}/E \approx3 \times10^{-4}, ν=0.4\nu= 0.4, μS=1.2\mu_{\mathrm{S}} = 1.2, and μK=0.2\mu_{\mathrm{K}} = 0.2, and much smaller vcρ/Ev_{\mathrm{c}}\sqrt{\rho/E} and η1/(HρE)\eta_1/(H\sqrt{\rho E}) are used in the previous study10.

Data availability

The datasets used and/or analyzed during the current study available from the corresponding author on reasonable request.

Acknowledgements

This study was supported by JSPS KAKENHI Grant Numbers JP19K03670, JP20K03792, JP21H01006, and JP22J20527. The numerical simulations were partially conducted on the supercomputer systems in ISSP, University of Tokyo, Japan, and in YITP, Kyoto University, Japan. We would like to thank Editage (www.editage.com) for English language editing.

Author contributions statement

W.I. conducted the FEM simulations and analysis based on simplified models. All authors analyzed the results and reviewed the manuscript.

Additional information

Supplementary information accompanies this paper at https://www.nature.com/srep

Competing interests: The authors declare no competing interests.

Supplementary Information

Supplementary Video (Titles and legends)

Supplementary Video S1. Spatial distributions of different quantities in the frictional interface for Pext/E=0.012P_{\mathrm{ext}}/E = 0.012, L/H=4L/H = 4, and W/H=1W/H = 1.

Left panel shows the ratio FT/FNF_{\mathrm{T}}/F_{\mathrm{N}} against the displacement of the rigid rod UU. (a) Spatial distribution of the slip region, which is represented by the yellow area. (b) Spatial distribution of the ratio σ(fric)/p\sigma^{(\mathrm{fric})}/p. The white area represents the region with p=0p = 0 due to the lift at the bottom. (c) Spatial distribution of pp.

Supplementary Video S2. Spatial distributions of different quantities in the frictional interface for Pext/E=0.012P_{\mathrm{ext}}/E = 0.012, L/H=1L/H = 1, and W/H=4W/H = 4.

Left panel shows the ratio FT/FNF_{\mathrm{T}}/F_{\mathrm{N}} against UU. (a) Spatial distribution of the slip region, which is represented by the yellow area. (b) Spatial distribution of the ratio σ(fric)/p\sigma^{(\mathrm{fric})}/p. The white area represents the region with p=0p = 0 due to the lift at the bottom. (c) Spatial distribution of pp.

Supplementary Note

In this supplementary note, we provide a detailed description of the analysis presented in the main text. § 13 describes the FEM results for the spatial distribution of quantities at the frictional interface. Details of the analytical calculation based on the simplified models are presented in § 14. In § 15, we compare our results with those of a previous study[1].

Spatial distributions of quantities in the frictional interface

This section presents the spatial distributions of the quantities in the frictional interface for the 3D FEM simulations, which supports the assumption used to derive the simplified models in the main text. The spatial distributions for L/H=4L/H = 4 and W/H=1W/H = 1 are shown in Figure S1, corresponding to Supplementary Video S1. Figure S1 and (Figure S1) demonstrate the spatial distributions of the slip region and the ratio σ(fric)/p\sigma^{(\mathrm{fric})}/p, respectively. Both distributions exhibit a slight dependency on xx. Therefore, in the simplified model with L/H≫1L/H \gg1, we neglect the xx dependence. The spatial distribution of pressure pp at the bottom is shown in Figure S1. The pressure pp is approximately equal to the pressure PextP_{\mathrm{ext}} at the top surface of the entire bottom plane. Based on this numerical result, we assume that the bottom pressure is uniform in the simplified model for L/H≫1L/H \gg1.

Spatial distributions of different quantities in the frictional interface

Figure S1. Spatial distributions of different quantities in the frictional interface for Pext/E=0.012P_{\mathrm{ext}}/E = 0.012, L/H=4L/H = 4, W/H=1W/H = 1, and U/L=2.6×10−2U/L = 2.6 \times10^{-2}. The rigid rod pushes the block at (x/H,y/H)=(0.5,0)(x/H,y/H) = (0.5,0). (a) Spatial distribution of the slip region, which is represented by the yellow area. (b) Spatial distribution of the ratio σ(fric)/p\sigma^{(\mathrm{fric})}/p. (c) Spatial distribution of pp.

The spatial distributions in the frictional interface for W/H=4W/H = 4 with L/H=1L/H = 1 are shown in Figure S2, corresponding to Supplementary Video S2. Figure S2 and (Figure S2) demonstrate the spatial distributions of the slip region and the ratio σ(fric)/p\sigma^{(\mathrm{fric})}/p, respectively. The regions for precursor slip and σ(fric)/p≈μS\sigma^{(\mathrm{fric})}/p \approx\mu_{\mathrm{S}} propagate from the center at x/H=2x/H = 2, which is different from those for L/H≫1L/H \gg1 shown in Figure S1. The characteristic behavior is the extension of the slip region along the xx direction, as shown in Supplementary Video S2. Therefore, in the model for W/H≫1W/H \gg1, we neglect the dependence on yy. The spatial distribution of pressure pp at the bottom is shown in Figure S2. The variation in pp is not strong. Therefore, we assume a constant pressure in the simplified model for W/H≫1W/H \gg1.

Spatial distributions of different quantities in the frictional interface

Figure S2. Spatial distributions of different quantities in the frictional interface for Pext/E=0.012P_{\mathrm{ext}}/E = 0.012, L/H=1L/H = 1, W/H=4W/H = 4, and U/L=1.08×10−1U/L = 1.08 \times10^{-1}. The rigid rod pushes the block at (x/H,y/H)=(2,0)(x/H,y/H) = (2,0). (a) Spatial distribution of the slip region, which is represented by the yellow area. (b) Spatial distribution of the ratio σ(fric)/p\sigma^{(\mathrm{fric})}/p. (c) Spatial distribution of pp.

Analysis based on the simplified models

This section presents details of the analytical calculations based on the simplified models. We derive the results for L/H≫1L/H \gg1 and W/H≫1W/H \gg1 shown in § 14.1 and § 14.3.

Model for large L/HL/H

_Quasi-static solution_

The quasi-static solution uy0(y)u_{y0}(y) of equations (2) and (3) in the main text with the local friction coefficient μ=μK\mu= \mu_{\mathrm{K}} and acceleration u¨y(y,t)=0\ddot{u}_y(y,t) = 0 at U=0U = 0 is given by

uy0(y)=μKPextE1αAH(y22−Ly).(19)u_{y0}(y) = \frac{\mu_{\mathrm{K}}P_{\mathrm{ext}}}{E_1\alpha_A H}\left(\frac{y^2}{2}-Ly\right). \tag*{(19)}

The quasi-static solution uy(a)(y)u_y^{(a)}(y) for U>0U > 0 is given by

uy(a)(y)={uy1(y),0≤y≤luy0(y),l<y≤L(20)u_y^{(a)}(y) = \begin{cases} u_{y1}(y), & 0 \le y \le l \\ u_{y0}(y), & l < y \le L \end{cases} \tag*{(20)}

where ll is the length of the precursor slip. The slip area AA is given by

A=lW.(21)A = lW. \tag*{(21)}

Here, we set the local friction coefficient μ\mu to be equal to μS\mu_{\mathrm{S}} for 0≤y≤l0 \le y \le l, and the displacement and stress are continuous at y=ly = l:

uy1(y=l)=uy0(y=l)(22)u_{y1}(y=l) = u_{y0}(y=l) \tag*{(22)}
duy1dy∣y=l=duy0dy∣y=l.(23)\left.\frac{d u_{y1}}{d y}\right|_{y=l} = \left.\frac{d u_{y0}}{d y}\right|_{y=l}. \tag*{(23)}

which gives

uy1(y)=uy0(y)+(μS−μK)Pext2E1αAH(y2−2ly+l2)(24)u_{y1}(y) = u_{y0}(y) + \frac{(\mu_{\mathrm{S}}-\mu_{\mathrm{K}})P_{\mathrm{ext}}}{2E_1\alpha_A H}\left(y^2-2ly+l^2\right) \tag*{(24)}

and

l={2E1αAHU(μS−μK)Pext}12.(25)l = \left\{\frac{2E_1\alpha_A H U}{(\mu_{\mathrm{S}}-\mu_{\mathrm{K}})P_{\mathrm{ext}}}\right\}^{\frac{1}{2}}. \tag*{(25)}

From equations (S3) and (S7), AA is obtained as

AW={2E1αAHU(μS−μK)Pext}12.(26)\frac{A}{W} = \left\{\frac{2E_1\alpha_A H U}{(\mu_{\mathrm{S}}-\mu_{\mathrm{K}})P_{\mathrm{ext}}}\right\}^{\frac{1}{2}}. \tag*{(26)}

_Stability analysis_

We introduce the fluctuation δuy(y,t)\delta u_y(y,t) as

δuy(y,t)=uy(y,t)−uy(a)(y).(27)\delta u_y(y,t) = u_y(y,t)-u_y^{(a)}(y). \tag*{(27)}

For y>ly > l, we assume δuy(y,t)=0\delta u_y(y,t) = 0. Substituting equation (S9) into the equation of motion (2), we obtain

ρδu¨y=E1∂2δuy∂y2+ηt∂2δu˙y∂y2+(μS−μK)δu˙yPextvcαAH.(28)\rho\ddot{\delta u}_y = E_1\frac{\partial^2 \delta u_y}{\partial y^2}+\eta_t\frac{\partial^2 \dot{\delta u}_y}{\partial y^2}+\frac{(\mu_{\mathrm{S}}-\mu_{\mathrm{K}})\dot{\delta u}_y P_{\mathrm{ext}}}{v_c\alpha_A H}. \tag*{(28)}

Because δuy=0\delta u_y = 0 at y=0y = 0 and ll, δuy(y,t)\delta u_y(y,t) can be expressed as

δuy(y,t)=∑m=1∞uymeλmtsin⁡kmξ(29)\delta u_y(y,t) = \sum_{m=1}^{\infty}u_{ym}e^{\lambda_m t}\sin k_m \xi \tag*{(29)}

with a positive integer mm and eigenvalue λm\lambda_m of the time evolution operator, where kmk_m and ξ\xi are defined as

km=mπ(30)k_m = m\pi \tag*{(30)}
ξ=yl,(31)\xi= \frac{y}{l}, \tag*{(31)}

respectively. Substituting equation (S11) into equation (S10), multiplying by 2sin⁡knξ2\sin k_n\xi with a positive integer nn, and integrating for 0<y<l0<y<l, we obtain

{ρλm2+ηtkm2l2λm+E1km2l2−(μS−μK)PextvcαAHλm}uneλmt=0.(32)\left\{\rho\lambda_m^2+\eta_t\frac{k_m^2}{l^2}\lambda_m+E_1\frac{k_m^2}{l^2}-\frac{(\mu_S-\mu_K)P_{\mathrm{ext}}}{v_c\alpha_AH}\lambda_m\right\}u_n e^{\lambda_m t}=0. \tag*{(32)}

From equations (S3) and (S14), we obtain

ρL2λm2+km2ηt(AA0)−2λm+km2E1(AA0)−2−(μS−μK)PextL2vcαAHλm=0.(33)\rho L^2\lambda_m^2+k_m^2\eta_t\left(\frac{A}{A_0}\right)^{-2}\lambda_m+k_m^2E_1\left(\frac{A}{A_0}\right)^{-2}-\frac{(\mu_S-\mu_K)P_{\mathrm{ext}}L^2}{v_c\alpha_AH}\lambda_m=0. \tag*{(33)}

The second term on the left-hand side of equation (S15) corresponds to the viscosity, while the fourth term on the same side represents the velocity-weakening friction. For A/A0≪1A/A_0\ll1, we can neglect the fourth term and λm\lambda_m satisfies Re⁡λm<0\operatorname{Re}\lambda_m<0. For a larger A/A0A/A_0, the second and third terms can be neglected, and one of the solutions satisfies Im⁡λm=0\operatorname{Im}\lambda_m=0 and Re⁡λm>0\operatorname{Re}\lambda_m>0.

The fluctuation δuy(y,t)\delta u_y(y,t) becomes unstable when Re⁡λm>0\operatorname{Re}\lambda_m>0; however, the oscillatory instability with Im⁡λm≠0\operatorname{Im}\lambda_m\ne0 is suppressed by the static friction force because the oscillatory motion is accompanied by backward motion. Bulk sliding occurs when the eigenvalue satisfies

Im⁡λm=0 and Re⁡λm>0(34)\operatorname{Im}\lambda_m=0\ \text{and}\ \operatorname{Re}\lambda_m>0 \tag*{(34)}

with m=1m=1. We obtain equation (4) in the main text for the critical area AcA_c from this condition using equation (S15), which gives

AcA0=πηtL(−ρE1+ρE1+(μS−μK)PextηtvcαAH).(35)\frac{A_c}{A_0}=\frac{\pi\eta_t}{L\left(-\sqrt{\rho E_1}+\sqrt{\rho E_1+\frac{(\mu_S-\mu_K)P_{\mathrm{ext}}\eta_t}{v_c\alpha_AH}}\right)}. \tag*{(35)}

Macroscopic friction coefficient

The loading force FNF_N is given by

FN=PextA0.(36)F_N=P_{\mathrm{ext}}A_0. \tag*{(36)}

The driving force FTF_T is balanced with the total friction force, except for the duration of bulk sliding. It should be noted that the precursor motion is quasi-static. FTF_T is given by the integral of the local friction force as

FT=W∫0LμPext dy.(37)F_T=W\int_0^L\mu P_{\mathrm{ext}}\,dy. \tag*{(37)}

Using equations (2), (3), (S1), (S3), (S2), (S5), and (S6), we obtain

FT=αAHW∫0Ldσyy(y)dy dy=E1αAHW{∫0ld2uy1dy2 dy+∫lLd2uy0dy2 dy}=−E1αAHWduy1dy∣y=0=PextA0{μK+(μS−μK)AA0}.(38)\begin{aligned} F_T&=\alpha_AHW\int_0^L\frac{d\sigma_{yy}(y)}{dy}\,dy\\ &=E_1\alpha_AHW\left\{\int_0^l\frac{d^2u_{y1}}{dy^2}\,dy+\int_l^L\frac{d^2u_{y0}}{dy^2}\,dy\right\}\\ &=-E_1\alpha_AHW\left.\frac{du_{y1}}{dy}\right|_{y=0}\\ &=P_{\mathrm{ext}}A_0\left\{\mu_K+(\mu_S-\mu_K)\frac{A}{A_0}\right\}. \tag*{(38)} \end{aligned}

Because the macroscopic static friction coefficient μM\mu_M is the ratio of the driving force to the loading force, FT/FNF_T/F_N, at A=AcA=A_c, equation (6) in the main text is obtained from equations (S18) and (S20).

Model for large W/HW/H

Quasi-static solution

The quasi-static solution uy0(x)u_{y0}(x) of equations (9) and (10) in the main text at U=0U=0 is obtained as

uy0(x)=μKPext2E2αBH(x2−Wx)(39)u_{y0}(x)=\frac{\mu_K P_{\mathrm{ext}}}{2E_2\alpha_B H}\left(x^2-Wx\right) \tag*{(39)}

with the boundary conditions, local friction coefficient μ=μK\mu=\mu_K, and acceleration u¨y(x,t)=0\ddot{u}_y(x,t)=0. For U>0U>0, the quasi-static solution uy(a)(x)u_y^{(a)}(x) is given by

uy(a)(x)={uy1(x),0≤x≤w2,uy0(x),w2<x≤W2.(40)u_y^{(a)}(x)= \begin{cases} u_{y1}(x), & 0\le x\le\frac{w}{2},\\ u_{y0}(x), & \frac{w}{2}<x\le\frac{W}{2}. \end{cases} \tag*{(40)}

with the width ww of the precursor slip related to AA as

A=wL.(41)A=wL. \tag*{(41)}

Setting μ\mu equal to μS\mu_S for 0≤x≤w/20\le x\le w/2, the displacement and stress are continuous at x=w/2x=w/2:

uy1(x=w2)=uy0(x=w2)(42)u_{y1}\left(x=\frac{w}{2}\right)=u_{y0}\left(x=\frac{w}{2}\right) \tag*{(42)}
duy1dx∣x=w2=duy0dx∣x=w2,(43)\left.\frac{du_{y1}}{dx}\right|_{x=\frac{w}{2}}=\left.\frac{du_{y0}}{dx}\right|_{x=\frac{w}{2}}, \tag*{(43)}

where we obtain uy1(x)u_{y1}(x) as

uy1(x)=uy0(x)+(μS−μK)Pext2E2αBH(x2−wx+w24)(44)u_{y1}(x)=u_{y0}(x)+\frac{(\mu_S-\mu_K)P_{\mathrm{ext}}}{2E_2\alpha_B H}\left(x^2-wx+\frac{w^2}{4}\right) \tag*{(44)}

with

w=2{2E2αBHU(μS−μK)Pext}12.(45)w=2\left\{\frac{2E_2\alpha_BHU}{(\mu_S-\mu_K)P_{\mathrm{ext}}}\right\}^{\frac{1}{2}}. \tag*{(45)}

From equation (41), AA can be expressed as

AL=2{2E2αBHU(μS−μK)Pext}12.(46)\frac{A}{L}=2\left\{\frac{2E_2\alpha_BHU}{(\mu_S-\mu_K)P_{\mathrm{ext}}}\right\}^{\frac{1}{2}}. \tag*{(46)}

Stability analysis

We introduce the fluctuation δuy(x,t)\delta u_y(x,t) as

δuy(x,t)=uy(x,t)−uy(a)(x).(47)\delta u_y(x,t)=u_y(x,t)-u_y^{(a)}(x). \tag*{(47)}

Substituting equation (47) into equation (9), we obtain

ρδu¨y=E2∂2δuy∂x2+η12∂2δu˙y∂x2+(μS−μK)Pextδu˙yvcαBH.(48)\rho\ddot{\delta u}_y=E_2\frac{\partial^2\delta u_y}{\partial x^2}+\frac{\eta_1}{2}\frac{\partial^2\dot{\delta u}_y}{\partial x^2}+\frac{(\mu_S-\mu_K)P_{\mathrm{ext}}\dot{\delta u}_y}{v_c\alpha_BH}. \tag*{(48)}

Assuming δuy(x,t)=0\delta u_y(x,t)=0 for x>w/2x>w/2 and x=0x=0, δuy(x,t)\delta u_y(x,t) is expressed as

δuy(x,t)=∑m=1∞uymeλmtsin⁡kmξ,(49)\delta u_y(x,t)=\sum_{m=1}^{\infty}u_{ym}e^{\lambda_m t}\sin k_m\xi, \tag*{(49)}

where kmk_m and ξ\xi are given by

km=mπ(50)k_m=m\pi \tag*{(50)}
ξ=2xw.(51)\xi=\frac{2x}{w}. \tag*{(51)}

Here, mm is a positive integer and λm\lambda_m is the eigenvalue. Substituting equation (S31) into equation (S30), multiplying by 2sin⁡knξ2\sin k_n\xi, and integrating for 0<x<w/20 < x < w/2, we obtain

{ρλm2+2η1km2w2λm+4E2km2w2−(μS−μK)PextvcαBHλm}uneλmt=0.(52)\left\{\rho\lambda_m^2+2\eta_1\frac{k_m^2}{w^2}\lambda_m+4E_2\frac{k_m^2}{w^2}-\frac{(\mu_S-\mu_K)P_{\mathrm{ext}}}{v_c\alpha_BH}\lambda_m\right\}u_ne^{\lambda_m t}=0. \tag*{(52)}

From equations (S23) and (S34), we obtain

ρW2λm2+2km2η1(AA0)−2λm+4km2E2(AA0)−2−(μS−μK)PextW2vcαBHλm=0.(53)\rho W^2\lambda_m^2+2k_m^2\eta_1\left(\frac{A}{A_0}\right)^{-2}\lambda_m+4k_m^2E_2\left(\frac{A}{A_0}\right)^{-2}-\frac{(\mu_S-\mu_K)P_{\mathrm{ext}}W^2}{v_c\alpha_BH}\lambda_m=0. \tag*{(53)}

The second term on the left-hand side of equation (S35) represents the viscosity, while the fourth term on the same side originates from the velocity-weakening friction. Similar to the case for L/H≫1L/H \gg1, bulk sliding occurs when the eigenvalue satisfies equation (S16) with increasing A/A0A/A_0 owing to the instability caused by the velocity-weakening friction. From this condition, we obtain equation (11) in the main text for the critical area AcA_c, which gives

AcA0=πη1W(−ρE2+ρE2+(μS−μK)Pextη12vcαBH).(54)\frac{A_c}{A_0}=\frac{\pi\eta_1}{W\left(-\sqrt{\rho E_2}+\sqrt{\rho E_2+\frac{(\mu_S-\mu_K)P_{\mathrm{ext}}\eta_1}{2v_c\alpha_BH}}\right)}. \tag*{(54)}

Macroscopic friction coefficient

The loading force FNF_N is given by equation (S18). The driving force FTF_T is balanced with the integral of the local friction force as

FT=L∫−w/2w/2μPextdx=2L∫0w/2μPextdx.(55)F_T=L\int_{-w/2}^{w/2}\mu P_{\mathrm{ext}}dx=2L\int_0^{w/2}\mu P_{\mathrm{ext}}dx. \tag*{(55)}

Using equations (9), (10), (S21), (S23), (S22), (S25), and (S26), we obtain

FT=2αBHL∫0w/2dσxy(x)dxdx=2E2αBHL{∫0w/2d2uy1dx2dx+∫w/2wd2uy0dx2dx}=−2E2αBHLduy1dx∣x=0=PextA0{μK+(μS−μK)AA0}.(56)\begin{aligned} F_T&=2\alpha_BHL\int_0^{w/2}\frac{d\sigma_{xy}(x)}{dx}dx\\ &=2E_2\alpha_BHL\left\{\int_0^{w/2}\frac{d^2u_{y1}}{dx^2}dx+\int_{w/2}^{w}\frac{d^2u_{y0}}{dx^2}dx\right\}\\ &=-2E_2\alpha_BHL\left.\frac{du_{y1}}{dx}\right|_{x=0}\\ &=P_{\mathrm{ext}}A_0\left\{\mu_K+(\mu_S-\mu_K)\frac{A}{A_0}\right\}. \tag*{(56)} \end{aligned}

Because the macroscopic static friction coefficient μM\mu_M is the ratio of the driving force to the loading force, FT/FNF_T/F_N, at A=AcA=A_c, we obtain equation (6) in the main text from equations (S18) and (S38).

Comparison with the analytical result from a previous study

In this section, we compare our results with those of ref. [1]. A previous study discussed the dependence of the precursor slip and friction coefficient on system size with a constant aspect ratio L/H=2L/H=2. In the analysis, the authors considered the nonuniformity of the bottom pressure as the origin of the precursor slip and demonstrated the breakdown of Amontons’ law. The bottom pressure was shown to increase along the driving direction owing to the torque effect[1, 2] and was approximated by p(y)=2Pexty/Lp(y)=2P_{\mathrm{ext}}y/L. The equation of motion is given by

ρu¨y(y,t)=∂σyy(y,t)∂y+σzy(y,t)−μ(u˙y)p(y)αH.(57)\rho\ddot{u}_y(y,t)=\frac{\partial\sigma_{yy}(y,t)}{\partial y}+\frac{\sigma_{zy}(y,t)-\mu(\dot{u}_y)p(y)}{\alpha H}. \tag*{(57)}

where the stresses σyy\sigma_{yy} and σzy\sigma_{zy} are expressed as

σyy=E1∂uy∂y+η1∂u˙y∂y(58)\sigma_{yy}=E_1\frac{\partial u_y}{\partial y}+\eta_1\frac{\partial\dot{u}_y}{\partial y} \tag*{(58)}
σzy=E2U−uyH/2+η12Vrod−u˙yH/2(59)\sigma_{zy}=E_2\frac{U-u_y}{H/2}+\frac{\eta_1}{2}\frac{V_{\mathrm{rod}}-\dot{u}_y}{H/2} \tag*{(59)}

with the fitting parameter α\alpha. The boundary conditions are as

∂uy∂y∣y=0=∂uy∂y∣y=L=0.(60)\left.\frac{\partial u_y}{\partial y}\right|_{y=0}=\left.\frac{\partial u_y}{\partial y}\right|_{y=L}=0. \tag*{(60)}

Using the same procedure as that in § 14, the equation for the critical area AcA_{\mathrm{c}} is obtained as

π2ηt4(AcA0)−2+η1α(LH)2+2Lρ{π2E14(AcA0)−2+2E2α(LH)2}=π2−4π2(μS−μK)PextL2αvcHAcA0.(61)\frac{\pi^2\eta_{\mathrm{t}}}{4}\left(\frac{A_{\mathrm{c}}}{A_0}\right)^{-2}+\frac{\eta_1}{\alpha}\left(\frac{L}{H}\right)^2+2L\sqrt{\rho\left\{\frac{\pi^2E_1}{4}\left(\frac{A_{\mathrm{c}}}{A_0}\right)^{-2}+\frac{2E_2}{\alpha}\left(\frac{L}{H}\right)^2\right\}} =\frac{\pi^2-4}{\pi^2}\frac{(\mu_{\mathrm{S}}-\mu_{\mathrm{K}})P_{\mathrm{ext}}L^2}{\alpha v_{\mathrm{c}}H}\frac{A_{\mathrm{c}}}{A_0}. \tag*{(61)}

For Ac/A0≪1A_{\mathrm{c}}/A_0\ll1, Ac/A0A_{\mathrm{c}}/A_0 is expressed as

AcA0≃π2π2−4(μS−μKα)−13(PextHηtvc)−13(LH)−23,(62)\frac{A_{\mathrm{c}}}{A_0}\simeq\frac{\pi^2}{\pi^2-4}\left(\frac{\mu_{\mathrm{S}}-\mu_{\mathrm{K}}}{\alpha}\right)^{-\frac{1}{3}}\left(\frac{P_{\mathrm{ext}}H}{\eta_{\mathrm{t}}v_{\mathrm{c}}}\right)^{-\frac{1}{3}}\left(\frac{L}{H}\right)^{-\frac{2}{3}}, \tag*{(62)}

where the exponents for PextP_{\mathrm{ext}} and L/HL/H differ from those in equation (5) in the main text. The difference in the exponents results from the different assumptions for the bottom pressure adopted in both studies, i.e., uniform pressure in the present study and non-uniform pressure in the previous study. Note that the macroscopic friction coefficient μM\mu_{\mathrm{M}} in this model is also given by equation (6) in the main text.

In Figure S3, we show μM\mu_{\mathrm{M}} against PextP_{\mathrm{ext}} for various lengths, L/HL/H with W/H=1W/H=1, obtained from the FEM simulations in the present study. The analytical results in ref. 1 given by equations (61) and (6), and in the present study given by equations (35) and (6) are also shown. The fitting parameter α\alpha is determined such that the analytical result matches the FEM results for L/H=2L/H=2 and Pext/E>0.01P_{\mathrm{ext}}/E>0.01. Figure S3 shows that the analytical results in the previous study better reproduce the results of the FEM analysis for L/H≤2L/H\le2. This is because the nonuniformity of the bottom pressure is significant for L/H≤2L/H\le2. However, the analytical results deviate from the FEM analysis for L/H≥4L/H\ge4, where the bottom pressure is almost uniform, as shown in Figure S1 and Supplementary Video S1 (c). Instead, the analytical results of the present study agree with the FEM analysis for L/H≥4L/H\ge4, as shown in Figure 3 and Figure S3.

Dependence of $\mu_M$ on $P_{\mathrm{ext}}$ for various $L/H$ values with $W/H=1$

Figure S3. Dependence of μM\mu_M on PextP_{\mathrm{ext}} for various L/HL/H values with W/H=1W/H=1. The thin solid lines represent the analytical results in ref. 1 given by equations (61) and (6) with the fitting parameter α=0.05\alpha=0.05. The thin dash-dotted lines represent the analytical results in the present study given by equations (35) and (6) with the fitting parameter αA=0.2\alpha_{\mathrm{A}}=0.2. The dotted and dashed lines represent μS\mu_{\mathrm{S}} and μK\mu_{\mathrm{K}}, respectively.

References

  1. [1]Otsuki, M. & Matsukawa, H. Systematic breakdown of Amontons’ law of friction for an elastic object locally obeying Amontons’ law. Sci. Reports 3, 1586, DOI: https://doi.org/10.1038/srep01586 (2013).
  2. [2]Scheibert, J. & Dysthe, D. K. Role of friction-induced torque in stick-slip motion. Europhys. Lett. 92, 54001, DOI: https://doi.org/10.1209/0295-5075/92/54001 (2010).

Paper details

vlnc:W_dpubkz5w
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 10

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents