Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

The frictional force on sliding drops

The frictional force on sliding drops Joel Koplik Benjamin Levich Institute and Department of Physics City College of the City University of New York, New York, NY 10031 (Dated: September 18, 2018) The dynamic frictional force between solid surfaces in relative motion di ers from the static force needed to initiate motion, but this distinction is not usually thought to occur for liquid drops moving on a solid. Recent experiments [Gao, et al., Nature Phys. 114, 191 (2018)] have challenged this view, and claim to observe an analog of solid-on-solid friction for sliding drops. We use molecular dynamic simulations to investigate the forces that moving liquids exert on solids in several situations. In contrast to the indirect techniques required in laboratory experiments, the forces involved in friction are directly accessible in these calculations. We nd that, aside from possible inertial e ects due to the abrupt initiation of motion and aging e ects for uncon ned drops, the frictional forces are constant in time. I. INTRODUCTION pin de ection versus applied force converts the measured defection into the time-dependent force on the pin. The results indicate three regimes: a \static" regime where Liquids are often distinguished from solids by their re- the drop distorts but moves with the sliding stage, a sponse to shear: \A liquid cannot support a shear stress \threshold" regime where the drop begins to slip and and ows irreversibly and continuously when a stress is the force on the pin rises to a peak value, followed by a applied" (see, e.g., [1]). A partially-wetting drop resting constant-force regime where the drop in held in place by on an ideal, smooth and planar solid surface, held in place the pin while slipping over the sliding stage. This behav- by surface tension forces at the contact line, would then ior was observed to be robust in terms of drop and solid begin to slide at once under gravity once the surface tilts. materials and pulling speed and appears to be general, Realistically however [2], the drop would be held in place although one may wonder about complications due to the by surface heterogeneities and only begin to move when a motion of the liquid around the pin and the consequent critical tilt angle is reached. At that point, usually char- distortion of the drop surface. acterized in terms of the advancing and receding contact angles, sliding begins. If instead of the liquid a second These experimental results are quite surprising, and it solid were placed on the surface, once again a critical tilt would be desirable to have them con rmed independently angle would be needed to initiate motion but here the by another technique. While it is dicult to directly usual characterization is in terms of solid and dynamic access the force exerted on an sliding liquid drop in the friction. This is a distinction between the force needed to laboratory, it is straightforward to do so in a molecular initiate the motion, the point where the ratio of lateral dynamics (MD) simulation. In this calculation, the force to normal force equals the static friction coecient, and between each pair of atoms is computed and it is a simple the lesser force required to sustain the motion, where the matter of bookkeeping to identify and isolate the various ratio is the (smaller) dynamic friction coecient. Expla- forces exerted on the drop. In this paper we present nations of this phenomenon [3] involve mismatches in the the results of MD simulations of several con gurations in respective surface irregularities, changes in the degree of which liquids move on solids, to test for the presence of contact between the two solid surfaces, distortions of the any analog to static solid-on-solid friction. solid lattice and so on. These e ects would appear to What we actually measure is the force the liquid exerts be absent in the case of a liquid drop, which would ad- on a bounding solid. Ostensibly, as the title of the paper just itself to achieve complete contact with the solid, and suggests, we are interested in the frictional force exerted one would not expect a distinction between the forces on the liquid, but the force on the solid is both better (tilt angle) needed for the initiation and maintenance of de ned and more relevant. Instantaneously, the force on sliding motion. the liquid is just equal and opposite to the force on the A recent experiment by Gao et al. [4] concludes that in solid, by Newton's third law, but as discussed below force fact a sliding liquid drop does exhibit distinct static and uctuations in MD are so severe that it is necessary to dynamic friction regimes. In order to measure the force average over a nite time interval to obtain a robust value that a sliding drop exerts on a supporting solid experi- of a force. It is straightforward to follow a region of solid mentally an ingenious indirect technique was developed over time, since the structure is fairly rigid, and compute in which a capillary pin is embedded in a liquid drop the force on that region, but liquid atoms are subject to placed on a sliding stage and the pin's de ection mea- di usion and advection and any liquid region will change sured optically as the drop moves past. Calibration of its shape and location over time, while any xed region in a liquid will change its contents. The time-averaged force on the solid is thus well-de ned whereas that on a jkoplik@ccny.cuny.edu liquid may not be. Furthermore, in any experiment, such arXiv:1809.05577v1 [cond-mat.soft] 14 Sep 2018 2 as the one that motivated this paper, it is the force on a measuring probe that is detected. We rst consider the simpler case of Couette ow ini- tiated in a periodic channel by abruptly translating a bounding wall. This case has two advantages: an ana- lytic solution to the governing equations is available, and there is no moving contact line issue to complicate the problem. This simulation also allows us to identify some general features of force measurement and the initiation of motion, which reappear in drop motion. Next, we study isolated drops sliding on a solid surface due to an applied body force. Here we consider a partially-wetting drop which is equilibrated while resting on an atomistic solid surface and then subjected to a lateral body force, in imitation of a drop placed on a tilted plane in the presence of gravity. We consider both a uniform atom- ically smooth surface as well as a drop held in place by surface heterogeneity: an abrupt variation in wettability or a step change in surface height. The force measure- ment results here are quite consistent with the usual de- scription of this ow in terms of dynamic contact angles. Third, as a cleaner analog of the experiment of Gao et al., we study a cylindrical drop placed in the corner where two at surfaces intersect at 90 , and then translate on of the surfaces so as to drive the drop into the corner. In this way we avoid the complicating e ects of distorting the liquid vapor interface as the drop moves around the pin, present in the experiment [4]. Lastly we study a di- rect caricature of the experiment, using a spherical cap droplet on a sliding stage held in place by a xed pin. In all of these simulations we nd no evidence for en- hanced static solid-on-liquid forces beyond the e ects of ordinary uid viscosity and inertia. In some but not all cases we observe a signi cant peak in the force at early times, but we argue that this is the result of a step change in the force applied to the liquid, which is per- FIG. 1. Couette ow velocity pro le at 10 (top) and 200 fectly well accounted for by the viscous e ects built into (bottom). The MD results (20 realization average) are in the Navier-Stokes equations, along with some transient green (*), Newtonian Navier-Stokes in red (+) and shear- non-Newtonian e ects. thinning model in blue (). II. COUETTE FLOW understood that numerical results are given in units of 1=2 (length),  (energy), and  = m(=) (time), where The simulations are based on classical MD methods m is the common atomic mass. The parameter c is used [5, 6] and involve a drop composed of a generic Newto- to adjust the strength of the liquid-solid interaction and nian liquid made of tetramer molecules with Lennard- hence the wettability, as described below. V , which is LJ Jones interactions, adjacent to solid surfaces consisting cut o at r = 2:5, acts between each pair of atoms and of atoms tethered to lattice sites. This computational V acts between successive atoms in a four-monomer FENE framework is used throughout the paper. The interac- linear chain. The solid atoms are bound to lattice sites tions are using a harmonic potential with sti ness 100= and a local Nos e-Hoover thermostat xes the temperature at 12 6 r r T = 0:8=k . This particular liquid-vapor-solid system V (r) = 4  c LJ has been used extensively in our previous work [8, 9] and 1 r has the convenient features of short-range interactions, V (r) = k r ln 1 (1) FENE F easily-variable wettability and a sharp liquid/vapor in- 2 r terface. Furthermore, its properties have been previ- with parameters k = 30= and r = 1:5 (after ously measured at this temperature: bulk uid density F 0 Ref. [7]. In the following, if not stated explicitly it is 0:857 , viscosity 5:18m=( ) and liquid-vapor surface 3 tension 0:668= . For Couette ow a slab of liquid is placed between two solid plates, 15625 tetramer molecules in a cube of side L = 41:9 with 2916 solid atoms tethered in fcc lay- ers above and below, and equilibrated for 100 . The LJ interaction parameters are set to be c = 1 and ff c = 0:75 for uid- uid and uid-wall interactions, re- fw spectively. The bottom wall is then abruptly set into motion and translated steadily at velocity U = 0:1= . The resulting velocity pro le in the liquid at successive times is shown in the Fig. 1 at early (10 ) and late (200 ) times, and is seen to evolve into the expected linear form, with some velocity slip at the walls. The wall slip occurs because the uid-wall interaction is weak: the interaction parameters used correspond to a partially wetting liquid with a sessile drop contact angle around 90 . The MD results are compared to a solution of the Stokes equa- tion which incorporates the observed slip via a Navier boundary condition. This continuum ow eld is ob- tained equivalently from a semi-analytic solutions in the literature [10] or from a direct numerical integration. The early-time agreement is only approximate, but this is not FIG. 2. Average wall forces in Couette ow. Bottom wall: surprising given that we have imposed a shear stress MD results in red (+), Navier-Stokes in blue (x); top wall: instantaneously and examined the response after only MD results in green (*). Navier-Stokes in magenta (x). about 20 ps. In this situation one expects some transient elastic or more generally non-Newtonian behavior in the uid, and speci cally a delayed response to the imposed shear. In the literature one nds, for example, Couette (1) The MD force is noisy, because the intermolecular force is a rapidly varying function of interatomic spac- ow calculations using FENE-P dumbbell [11] which give a velocity pro le which lags behind the Navier-Stokes so- ing. The plotted curve is an average over 20 statistically lution at early times. Furthermore, in Fig. 1b we also independent realizations, obtained from di erent initial show a good t to the MD velocity pro le using a sim- atomic velocity values, and even then a further time av- ple shear-thinning model in which the viscosity varies as erage is needed in order to identify the trend. The uc- 1=2 tuations are much more severe at early times (t < 200 ) =   (1 + = ) , where and are the local and 0 0 0 global strain rates. We have also reproduced the MD because the data points are spaced by 5 and the result- ing averaging interval is short, while the data is more pro le using a simple Maxwell model (not shown). The point of this discussion of the velocity pro le is to argue stable at later times where the interval is 50 . that even for time-dependent ows, \experimental" MD (2) The continuum results are as expected. The lower results for velocity are consistent with the usual theoreti- wall is abruptly set into motion, giving a large stress at cal expectations for liquids, provided one takes account of the start (nominally in nite) which relaxes to a constant inertia and a possible transient non-Newtonian response value when the velocity pro le stabilizes at linear. The to abrupt changes in ow conditions. liquid adjacent to the upper wall is at rest until the vor- ticity disturbance from the lower wall reaches it, so this We now make the analogous comparison for the forces force is initially zero and increases to its steady value the liquid exerts on the solid. In an MD calculation as the uid accelerates. In the steady state regime the the force on any liquid atom is the simple sum of pair- wall forces are equal and opposite because the uid is wise forces exerted by all other atoms within interaction no longer accelerating and the forces must sum to zero. range, so each wall force is the sum of forces between all The force using the shear-thinning model is slightly dif- uid atoms and atoms in that wall, and can be isolated ferent at early times but still in the middle of the MD easily. The continuum force is just the shear stress times uctuations. the wall area, L @u=@y, where the sign re ects the (3) The MD force agrees with the continuum force, mod- direction of the normal to the interface, u and x are the ulo the uctuations. The uid begins to move immedi- streamwise velocity and coordinate, and y is the spanwise ately and the wall force does not exhibit stick-slip be- coordinate. The results for the lateral force on the two havior or any analog of solid-on-solid static friction. The walls in Couette ow are displayed in Fig. 2, along with sharp peak in the force on the moving wall at early times the corresponding continuum forces obtained from the results from inertia (included in the Newtonian contin- Newtonian calculation incorporating slip. (The two non- uum description) plus a bit of elasticity. Newtonian models mentioned above give similar results.) Several features are to be noted in this gure. The main issue in this paper is the possibility of a 4 FIG. 3. Sessile drop driven across an atomically-smooth homogeneous substrate at times 0, 500 and 2500 (top to bottom). shear stress enhancement at the onset of motion, and in this con guration the cause is the abrupt motion of the wall. Indeed, the stress peak at t = 0 is absent when the wall velocity is linearly ramped up from 0 to 0.1, in place of the step change illustrated in the gure. Like- wise, in a similar simulation of Poiseuille ow, even if a step pressure gradient is imposed the shear stress in- creases smoothly from zero. The distinction is completely accounted for by the Stokes equations: the analytic solu- tions for the velocity in start-up Couette and Poiseuille ow with a no-slip boundary condition, given for exam- ple in [12], directly indicate whether a stress peak is present. These solutions involve a Fourier series which converges for the shear stress at the bottom wall at t = 0 for Poiseuille ow but diverges in the Couette case. III. SESSILE DROPS FIG. 4. Center of mass (top) and lateral force (bottom) vs. The simplest uid mechanical analog of solid-solid fric- time for a sessile drop forced across the various substrates at tion involves a drop sliding on a solid surface. Although f = 0:001, averaged over 20 realizations. Smooth: red (+), these is an extensive literature on this problem [13, 14] chemical step: green (*), physical step: blue (). relatively little attention has been given to the forces in- volved, and we use MD simulations to disentangle them. We begin with a partially-wetting liquid on a uniform seen in Fig. 3, uctuate during the motion. A constant and atomically-smooth substrate, in the form of cylin- external force is formally equivalent to tilting the sub- drical cap drop shown in Fig. 3. The advantage of this strate in the presence of gravity, although realistically shape compared to a spherical cap is that the system is the values of f used here are much too large for this statistically homogeneous in the direction (z) normal to interpretation. The behavior is similar over a range of the plane of the gure, and the results can be averaged forcing values near this one, 5  10 to 0.005, but for over z. The wall- uid interaction strength is chosen as signi cantly lower values the force on the drop is unmea- c = 0:85 which gives an initial contact angle of 70 . ff surable because the signal is swamped by the uctuations After equilibration a constant force f = 0:001m= is and the drop motion itself becomes intermittent. As for applied to each atom and the drop translates, as shown in higher forcing values, at f = 0:01 the center of mass still the subsequent frames of the gure. The Reynolds num- moves with constant velocity although the drop becomes ber, Re  XH = where H is the equilibrium drop 0 0 increasingly elongated and at still higher values the drop height and X is the center of mass velocity (see below), tends to y o the substrate. is 0.0748 and the Capillary number Ca  X= is 0.149, Most realistic solid surfaces are heterogeneous and ir- with slightly di erent values in the other cases. As the regular, and we can address these complications by forc- drop moves its shape changes and exhibits distinct ad- ing the drop across a modi ed substrate involving either vancing and receding dynamic contact angles which, as a chemical or a physical heterogeneity. The drop is equi- 5 FIG. 5. Sessile drop driven across an substrate with a chem- ical step. Pinned at f = 0:0005 (top) and in motion for f = 0:001 at times 500 (center) and 2500 (bottom). librated as before but to the right of the drop there is either a chemical step where the wettability coecient c drops to 0.65 or a physical step where the height of fw the solid rises by one fcc unit cell, a distance of 1.17 here. When a lateral force is applied the drop tilts in the direction of the force, developing distinct advancing and receding angles,  , but moves only if the force is A;R large enough. We illustrate this behavior in Fig. 5 for the chemical step case, where the value of the threshold force needed to move the drop over the step can be obtained FIG. 6. Flow inside the drop on a homogeneous substrate. by a the following force balance [2]. Two-dimensional velocity eld in the moving center of mass In equilibrium on a homogeneous surface a drop is held frame (top) and average lateral velocity pro le in the center in place by solid/liquid, solid/vapor and liquid/vapor of the drop in the substrate frame (bottom). surface tension forces at the contact line, with a lateral force balance given by Young's equation = SV SL cos  . For non-ideal surfaces there is a range of pos- Fig. 4 shows the drop center of mass X (t) moving with sible equilibrium contact angles. If a weak force f is nearly constant velocity from the start for both smooth applied to each atom, a drop tilts in the direction of the and heterogeneous surfaces. The total lateral force F (t) force but can remain at rest if the applied force is bal- the liquid exerts on the wall at f = 0:001 is given as a anced by the unbalanced surface tension forces per length function of time in Fig. 4. in all cases the wall force os- (cos  cos  ) at the receding contact line and simi- R 0 cillates about zero before the external force is applied at larly at the advancing line. Explicitly, the force balance time 100 and afterwards ramps monotonically up to to is fN = W (cos  cos  ), where N = 46800 is the R A a plateau. Only the transient behavior immediately after number of atoms in the drop and W length of the contact application of the external force shows a variation with line (the width of the drop in this situation, 51:3). For the structure of the substrate. The distinction between the chemical step, we observe that the drop appears sta- the size of the uctuations at early and late times is the tionary at force 0:0005, where it is pinned at the edge of same as in the previous Couette simulations: short and the step, but moves steadily at 0.0006 and higher values. long averaging intervals, respectively. The average force Using the drop density pro le at 0.0005 and de ning the in the plateau uctuates about the value of the net force liquid/vapor interface to be the contour where the den- applied to the liquid, Nf = 46:8, as expected because sity is half the bulk value, we estimate   67 and R the drop is not accelerating. It is possible to decompose 111 , giving f = 0:00055 from the force balance the wall force into capillary forces at the contact line equation, in agreement with the simulation. A similar plus frictional drag on the drop, but unfortunately the transition and corresponding force balance is found in dynamic contact angles could not be measured with any the case of a physical step. If there are no angles A;R accuracy, and we have not pursued this. in the equilibrium range satisfying the force balance the The motion of the liquid inside the drop is a combi- drop will move, and in the present simulations moving nation of rolling and slipping, as indicated in Fig. 6 for drops are expected to have have constant velocity, char- the smooth wall. The two-dimensional ow eld shown acteristic of linear friction in Stokes ow. is an average over 50 and evaluated in a reference frame We observe that, despite the shape uctuations seen moving with the drop center of mass; the resolution is in the gures, the motion is steady: for forcing f = 0:001 poor since the velocities involved are O(10 ) times the 6 random thermal velocity and the center of mass velocity is not exactly constant during the averaging interval, but rotation about the middle of the drop is evident. En- semble averaging tends to wash out the result for this two-dimensional eld, but is more e ective for the ve- locity pro le, which is instead evaluated in the reference frame of the solid wall and is also an average over the middle of the drop. The result is a roughly linear pro le as would correspond to rotation about the wall. (The up- permost points correspond to the liquid/vapor interfacial region where the density is falling o .) The presence of slip at the wall requires a small discussion. The region 0  y  1:71 is occupied by solid (at rest), and the lowest liquid point (at y = 2:565) is in the center of a nite-sized sampling bin in the liquid. The precise de - nition of the \solid wall" is always ambiguous at atomic resolution, but would certainly be somewhere in the re- gion 1:71  y  2:5, where u is non-zero. Aside from this ambiguity, as in Couette ow the presence of slip varies with the strength of the uid-wall interaction but the rolling motion is always present. The force exerted on the drop by the walls, can be ex- amined locally by computing the force as a function of the coordinate along the base of the drop. We use a sequence of slabs of wall, i  dx < x < (i + 1)  dx and 0 < y < L with dx = 1:71, and compute the force per area on each slab. (Note that this quantity is almost the x-y compo- nent of the uid's shear stress tensor at the wall. The quali cation is because, as in the slip discussion above, MD elds are always averages over a nite-sized sam- pling bin with the result assigned to its center, which is displaced from the wall by half the bin size. A further extrapolation would be required to determine the force at the wall.) In Fig. 7 we show the horizontal (x) and vertical (y) components of the force on the drop, along FIG. 7. Local elds for a drop on a homogeneous substrate, with the local density and lateral velocity, both in equi- at rest (top) and in motion (bottom) after 250 , in a single librium and while translating. The forces peak at the realization. Density: red (+), x-velocity: green (*), x-force: orange (x) and y-force: blue (). drop's two contact lines and are constant on average in the interior. In the x-direction, the signs of the two peaks correspond to the fact that the liquid/vapor surface ten- sion acts to contract the drop into to a circular cylinder in detail. The principal point we wish to make here is that when the drop moves the numerical values change but the attraction to the wall draws the edges of the drop but there is no qualitative di erence in local forces on outwards, corresponding to an inward force on the wall. The wall force in the interior of the drop is (statistically) the wall. The density is constant inside the drop, and its pro le simply shifts in x as the drop moves, while the constant, zero when the drop is in equilibrium and posi- tive during drop translation due to the liquid pulling to liquid velocity at the wall is constant, zero in equilibrium and approximately equal to the center of mass velocity the right. In the y-direction, the force peaks at the con- tact lines, again correspond to surface tension trying to during the motion. contract the drop by pulling upwards on the wall, while in the interior the force is a constant, corresponding to the Laplace pressure inside the drop. IV. CORNERED DROPS A similar but more detailed discussion of the variation of these local forces with contact angle and a comparison The experiments which motivated this study involved to Young's equation is given by Fernandez-Toledano, et spherical drops on a sliding stage held in place by a al., [15] for a di erent con guration involving a liquid measurement pin, and includes features of both of the bridge in equilibrium between two solid plane walls. In previous simulations. The liquid is forced into motion that case the drop is con ned by the walls but here is it abruptly as in Couette ow, with a localized impulsive is only attracted to one wall, so the force analysis di ers stress at a solid boundary, and then \slides" relative to 7 FIG. 8. Drop driven into a corner by motion of the bottom wall to the right, at times 0, 200 and 600 (top to bottom). The translation of the wall is shown by the red dot which indicates the location of an atom xed in the wall, the solid with (as we shall see) a combination of slip and rolling. Again we take advantage of MD's direct access FIG. 9. Streamlines (top) and local elds (bottom) for a to the force and use a simpler con guration with no vari- drop squeezed into a corner at u = 0:1: density: red (+), ation in one direction to improve the statistics. x-velocity: green (*), x-force: orange (x) and y-force: blue We place a drop in the corner formed by two distinct (). solid walls meeting at a right angle, as shown in Fig. 8, and choose the (partial) wetting coecient c = 0:75. ff The simulation box is a cube of side 68.4 containing wall is translated as a rigid (but slightly randomized) lat- 16,000 tetramer molecules, which initially ll an approx- tice. The reason is that at higher velocities the atom po- imate circular quadrant with a contact angle of 93 . The sitions lag signi cantly behind the tether positions unless vertical wall is xed in place while the horizontal wall is, a very high binding force is used, which would require a after equilibration, translated towards the xed wall at correspondingly very small time step to resolve. We have a constant velocity. Due to periodicity, the sliding wall veri ed that the uid motion and wall forces are not sen- can translate inde nitely. The gap where the two walls sitive to this modi cation. When the lower wall moves meet is the same size as the internal wall lattice spac- (right), the drop is squeezed into the corner, as shown in ing, and there is no interaction between the respective Fig. 8 for the u = 0:1 case, and evolves to a roughly wall atoms, and therefore no uid leakage or distortion time-independent con guration after several hundred  . of the corner or any solid frictional heat generated there. In the steady state, the advancing contact line at the hori- These simplifying properties would be dicult to realize zontal wall (in the rest frame of this wall the contact line in a laboratory experiment, although the di erences in- advances to the left) increases to a velocity-dependent volve only the corner region which is not the focus of this value (100 to 145 ), while the static angle on the vertical study in any case. wall shows little change. The motion inside the drop is We consider a range of wall velocities 0.01 to 1.0= . rolling, as seen in Fig. 9, accompanied by some slip at At the lower velocities we use the same time step and wall the wall. The horizontal force the liquid exerts on the tethering strength as in the previous simulations, but for on the wall is negative, opposing the wall motion, with a the higher velocities the time step is lowered to 0.001 peak at the corner where a continuum no-slip ow eld and instead of tethering thermally active wall atoms, the would have a stress singularity, while the vertical force 8 FIG. 11. Time decay of uid density and velocity near the wall, and the x; y forces on the drop at u = 1:0; color code as in Fig. 7. squeezing the uid to the right and upwards and the uid resists this. The other notable feature of the forces is the systematic weak decay (towards zero) following the peak, surprising since one would have expected constant wall forces in an apparently steady ow. However, in this situation the drop is not con ned and is free to adjust its position. intuitively one might imagine the drop pushed upwards by repeated collisions with the sliding wall atoms and in fact precisely this behavior is observed. In Fig. 11 we plot the density and velocity in the uid adjacent to the wall at low and high velocities, normalized to their values at time when the wall begins to move. The initial peak FIG. 10. Time variation of the x (top) and y (bottom) force results from the abrupt start, following which there is on the sliding wall for various velocities. u = 0:1: red (+); a weak decay of the density, meaning uid moves away u = 0:25: green (*); u = 0:5: orange () and u = 1:0: blue from the wall, which produces parallel decays in the uid (x). velocity and wall forces. has the same interpretation as for sessile drops. At higher V. THE EXPERIMENT wall velocities, the liquid density pro le becomes slightly asymmetric and larger near the corner, the slip velocity increases, also asymmetrically but instead largest at the There remains the question of interpreting the exper- contact line, and the force pro les increase in magnitude iments of Gao et al. The principal distinction between but maintain the same shape. that experiment and the drop simulations here is the dis- The x and y components of the total force on the bot- tortion of the drop surface as it moves relative to the mea- tom wall are shown in Fig. 9, for di erent values of the suring pin. Rearrangement of the liquid drop's shape and wall velocity. In all cases there is an initial peak, resem- the contact lines on the pin and on the substrate would certainly a ect the force exerted We have carried out sev- bling that appearing in Couette ow, which we attribute to the inertial e ects of an impulsive start. The rea- eral simulations on this process, involving a spherical cap drop and a nite-sized obstacle pin, rather than a bar- son is that the duration of the stress peaks is 100-150 (note that wall motion begins at 100 ) which is compara- rier wall as in the previous section, but these simulations do not completely reproduce the experiments in terms of ble to the vorticity di usion time across the drop, R = with drop size R  35 and viscosity  = 0:518m=( , drop distortion and wall force. which characterizes ow development. The signs of the An example is given in Fig. 12 { a spherical cap is uid forces on the wall are negative because the walls are placed on a partially wetting substrate (c = 0:85) and ff 9 istic of solid/liquid friction. FIG. 12. Simulation of the experiment of Gao et al. [4]. Top: side and top view of the equilibrated system, Bottom: e ects of sliding the substrate to the left, at times 1000 and 5000 . equilibrated with a rectangular pin inserted in the drop from above. Most of the pin has the same interaction as the substrate, except that the upstream face is more strongly wetting (c = 1:2), in order to mimic the ex- ff periment where a metallic re ecting layer was added to improve the imaging. After equilibration the substrate is translated to the right at velocity u = 0:1 and the force FIG. 13. Force on the pin in the simulation shown in Fig. 12. on the pin recorded. Fig. 13 again shows a monotonic rise to a near-plateau, with no \static-friction" enhance- ment. At times beyond 5000 the drop becomes highly VI. CONCLUSIONS elongated and eventually detaches from the pin, but dur- ing the interval indicated the footprint of the drop on We have used MD simulations to investigate the pos- the substrate remains approximately circular. The re- sibility of an enhanced shear stress before a liquid begins sults are similar when the conditions of the simulation to move across a solid surface, analogous to the distinc- are varied (di erent wettabilities, di erent pin shapes, tion between static and dynamic friction when two solids di erent speeds, etc.), and provided the pin remains em- move relative to each other. In cases where the motion bedded inside the drop the force on the pin is roughly is driven by a constant body force applied to the inte- constant. Exceptions to the typical behavior are found rior of the liquid, even in a step-wise fashion, the stress when the pin is at the edge of the drop, when the liquid is found to increase monotonically from zero to a steady either rst wets or dewets the pin. In these cases the state value. An abrupt motion of a solid bounding sur- liquid/vapor interface does deform and a transient peak face, however, generates a large local strain which in turn or spike appears in the force. produces a peak in the shear stress. Such peaks are in - In contrast to the simulations, the experimental drops nite in the (mathematical) continuum limit but regular- do change shape and, in particular, their contact line ized to large but nite values in MD simulations and in length increases during the time interval when the force real life, but otherwise entirely in accord with the Navier- is enhanced. A further distinction in the simulations, a Stokes equations. Furthermore, even in situations where consequence of size limitations, is that the liquid climbing the liquid is pinned by inhomogeneities and requires a the wetting side of the pin has a density corresponding minimum threshold force for continuous motion, no force to liquid/vapor interfacial region and does not faithfully enhancement is found. The experiments [4] which mo- represent bulk liquid and may not exert the proper hy- tivated this work appear to incorporate changes in drop drodynamic drag. Nonetheless, it appears that the tran- shape and thereby involve interfacial dynamics as well as sient friction enhancement in the experiments is related wetting considerations, and do not provide evidence for to change of drop shape, and is not a general character- an analog of static solid/solid friction. 10 [1] J.C. Anderson, K.D. Leaver, R.D. Rawlings and J.M. channel, Phys. Fluids 18, 031104 (2006). Alexander, Materials Science, 4th ed., (Springer, New [9] J. Koplik and C. Maldarelli, Di usivity and hydrody- York, 1990). namic drag of nanoparticles at a vapor-liquid interface, [2] P.-G. de Gennes, F. Brochard-Wyart and D. Qu er e, Cap- Phys. Rev. Fluids 2, 024303 (2017). illarity and Wetting Phenomena: Drops, Bubbles, Pearls, [10] G. Kaoullas and G. C. Georgiou, Start-up and cessation Waves (Springer, New York, 2004). Newtonian Piseuille and Couette ows with dynamic wall [3] B. N. J. Persson, Sliding Friction, 2nd ed. (Springer, New slip, Meccanica 50, 1747 (2015). York, 2000). [11] Y. Mochimaru, Unsteady-state development of plane [4] N. Gao, F. Geyer, D. W. Pilat, S. Wooh, D. Vollmer, H.- Couette ow for viscoelsatic uids, J. Non-Newtonian J. But and R. Berger, How drops start sliding over solid Fluid Mech. 12, 135 (1983). surfaces, Nature Phys. 14, 191 (2018). [12] C. Pozrikidis, Introduction to Theoretical and Computa- [5] M. P. Allen and D. J. Tildesley, Computer Simulation of tional Fluid Dynamics (Oxford, New York, 1997). Liquids (Clarendon, Oxford, 1987). [13] D. Bonn, J. Eggers, J. Indekeu, J. Meunier and E. Rolley, [6] D. Frenkel and B. Smit, Understanding Molecular Simu- Wetting and spreading, Rev. Mod. Phys. 81. 739 (2009). lation, 2nd ed. (Academic, San Diego, 2002). [14] Special section containing papers on the dynamics of wet- [7] G. S. Grest and K. Kremer, Molecular dynamics simula- ting, J. Phys.: Condens. Matter, 21, 464117 (2009). tion for polymers in the presence of a heat bath, Phys. [15] J.-C. Fernandez-Toledano, T. D. Blake, P. Lambert and Rev. A 33, 3628 (1986). J. De Coninck, On the cohesion of uids and their adhe- [8] T. S. Lo, J. Koplik, M. Rauscher and S. Dietrich, Pearling sion to soilds, Adv. Colloid Interface Sci. 245, 102 (2017). instability of nanoscale uid ow con ned to a chemical http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Condensed Matter arXiv (Cornell University)

The frictional force on sliding drops

Koplik, Joel
Condensed Matter , Volume 2018 (1809) – Sep 14, 2018
Download PDF
10 pages

Loading next page...
 
/lp/arxiv-cornell-university/the-frictional-force-on-sliding-drops-MQQ67bzm0u

References (15)

ISSN
2469-990X
eISSN
ARCH-3331
DOI
10.1103/PhysRevFluids.4.014001
Publisher site
See Article on Publisher Site

Abstract

Joel Koplik Benjamin Levich Institute and Department of Physics City College of the City University of New York, New York, NY 10031 (Dated: September 18, 2018) The dynamic frictional force between solid surfaces in relative motion di ers from the static force needed to initiate motion, but this distinction is not usually thought to occur for liquid drops moving on a solid. Recent experiments [Gao, et al., Nature Phys. 114, 191 (2018)] have challenged this view, and claim to observe an analog of solid-on-solid friction for sliding drops. We use molecular dynamic simulations to investigate the forces that moving liquids exert on solids in several situations. In contrast to the indirect techniques required in laboratory experiments, the forces involved in friction are directly accessible in these calculations. We nd that, aside from possible inertial e ects due to the abrupt initiation of motion and aging e ects for uncon ned drops, the frictional forces are constant in time. I. INTRODUCTION pin de ection versus applied force converts the measured defection into the time-dependent force on the pin. The results indicate three regimes: a \static" regime where Liquids are often distinguished from solids by their re- the drop distorts but moves with the sliding stage, a sponse to shear: \A liquid cannot support a shear stress \threshold" regime where the drop begins to slip and and ows irreversibly and continuously when a stress is the force on the pin rises to a peak value, followed by a applied" (see, e.g., [1]). A partially-wetting drop resting constant-force regime where the drop in held in place by on an ideal, smooth and planar solid surface, held in place the pin while slipping over the sliding stage. This behav- by surface tension forces at the contact line, would then ior was observed to be robust in terms of drop and solid begin to slide at once under gravity once the surface tilts. materials and pulling speed and appears to be general, Realistically however [2], the drop would be held in place although one may wonder about complications due to the by surface heterogeneities and only begin to move when a motion of the liquid around the pin and the consequent critical tilt angle is reached. At that point, usually char- distortion of the drop surface. acterized in terms of the advancing and receding contact angles, sliding begins. If instead of the liquid a second These experimental results are quite surprising, and it solid were placed on the surface, once again a critical tilt would be desirable to have them con rmed independently angle would be needed to initiate motion but here the by another technique. While it is dicult to directly usual characterization is in terms of solid and dynamic access the force exerted on an sliding liquid drop in the friction. This is a distinction between the force needed to laboratory, it is straightforward to do so in a molecular initiate the motion, the point where the ratio of lateral dynamics (MD) simulation. In this calculation, the force to normal force equals the static friction coecient, and between each pair of atoms is computed and it is a simple the lesser force required to sustain the motion, where the matter of bookkeeping to identify and isolate the various ratio is the (smaller) dynamic friction coecient. Expla- forces exerted on the drop. In this paper we present nations of this phenomenon [3] involve mismatches in the the results of MD simulations of several con gurations in respective surface irregularities, changes in the degree of which liquids move on solids, to test for the presence of contact between the two solid surfaces, distortions of the any analog to static solid-on-solid friction. solid lattice and so on. These e ects would appear to What we actually measure is the force the liquid exerts be absent in the case of a liquid drop, which would ad- on a bounding solid. Ostensibly, as the title of the paper just itself to achieve complete contact with the solid, and suggests, we are interested in the frictional force exerted one would not expect a distinction between the forces on the liquid, but the force on the solid is both better (tilt angle) needed for the initiation and maintenance of de ned and more relevant. Instantaneously, the force on sliding motion. the liquid is just equal and opposite to the force on the A recent experiment by Gao et al. [4] concludes that in solid, by Newton's third law, but as discussed below force fact a sliding liquid drop does exhibit distinct static and uctuations in MD are so severe that it is necessary to dynamic friction regimes. In order to measure the force average over a nite time interval to obtain a robust value that a sliding drop exerts on a supporting solid experi- of a force. It is straightforward to follow a region of solid mentally an ingenious indirect technique was developed over time, since the structure is fairly rigid, and compute in which a capillary pin is embedded in a liquid drop the force on that region, but liquid atoms are subject to placed on a sliding stage and the pin's de ection mea- di usion and advection and any liquid region will change sured optically as the drop moves past. Calibration of its shape and location over time, while any xed region in a liquid will change its contents. The time-averaged force on the solid is thus well-de ned whereas that on a jkoplik@ccny.cuny.edu liquid may not be. Furthermore, in any experiment, such arXiv:1809.05577v1 [cond-mat.soft] 14 Sep 2018 2 as the one that motivated this paper, it is the force on a measuring probe that is detected. We rst consider the simpler case of Couette ow ini- tiated in a periodic channel by abruptly translating a bounding wall. This case has two advantages: an ana- lytic solution to the governing equations is available, and there is no moving contact line issue to complicate the problem. This simulation also allows us to identify some general features of force measurement and the initiation of motion, which reappear in drop motion. Next, we study isolated drops sliding on a solid surface due to an applied body force. Here we consider a partially-wetting drop which is equilibrated while resting on an atomistic solid surface and then subjected to a lateral body force, in imitation of a drop placed on a tilted plane in the presence of gravity. We consider both a uniform atom- ically smooth surface as well as a drop held in place by surface heterogeneity: an abrupt variation in wettability or a step change in surface height. The force measure- ment results here are quite consistent with the usual de- scription of this ow in terms of dynamic contact angles. Third, as a cleaner analog of the experiment of Gao et al., we study a cylindrical drop placed in the corner where two at surfaces intersect at 90 , and then translate on of the surfaces so as to drive the drop into the corner. In this way we avoid the complicating e ects of distorting the liquid vapor interface as the drop moves around the pin, present in the experiment [4]. Lastly we study a di- rect caricature of the experiment, using a spherical cap droplet on a sliding stage held in place by a xed pin. In all of these simulations we nd no evidence for en- hanced static solid-on-liquid forces beyond the e ects of ordinary uid viscosity and inertia. In some but not all cases we observe a signi cant peak in the force at early times, but we argue that this is the result of a step change in the force applied to the liquid, which is per- FIG. 1. Couette ow velocity pro le at 10 (top) and 200 fectly well accounted for by the viscous e ects built into (bottom). The MD results (20 realization average) are in the Navier-Stokes equations, along with some transient green (*), Newtonian Navier-Stokes in red (+) and shear- non-Newtonian e ects. thinning model in blue (). II. COUETTE FLOW understood that numerical results are given in units of 1=2 (length),  (energy), and  = m(=) (time), where The simulations are based on classical MD methods m is the common atomic mass. The parameter c is used [5, 6] and involve a drop composed of a generic Newto- to adjust the strength of the liquid-solid interaction and nian liquid made of tetramer molecules with Lennard- hence the wettability, as described below. V , which is LJ Jones interactions, adjacent to solid surfaces consisting cut o at r = 2:5, acts between each pair of atoms and of atoms tethered to lattice sites. This computational V acts between successive atoms in a four-monomer FENE framework is used throughout the paper. The interac- linear chain. The solid atoms are bound to lattice sites tions are using a harmonic potential with sti ness 100= and a local Nos e-Hoover thermostat xes the temperature at 12 6 r r T = 0:8=k . This particular liquid-vapor-solid system V (r) = 4  c LJ has been used extensively in our previous work [8, 9] and 1 r has the convenient features of short-range interactions, V (r) = k r ln 1 (1) FENE F easily-variable wettability and a sharp liquid/vapor in- 2 r terface. Furthermore, its properties have been previ- with parameters k = 30= and r = 1:5 (after ously measured at this temperature: bulk uid density F 0 Ref. [7]. In the following, if not stated explicitly it is 0:857 , viscosity 5:18m=( ) and liquid-vapor surface 3 tension 0:668= . For Couette ow a slab of liquid is placed between two solid plates, 15625 tetramer molecules in a cube of side L = 41:9 with 2916 solid atoms tethered in fcc lay- ers above and below, and equilibrated for 100 . The LJ interaction parameters are set to be c = 1 and ff c = 0:75 for uid- uid and uid-wall interactions, re- fw spectively. The bottom wall is then abruptly set into motion and translated steadily at velocity U = 0:1= . The resulting velocity pro le in the liquid at successive times is shown in the Fig. 1 at early (10 ) and late (200 ) times, and is seen to evolve into the expected linear form, with some velocity slip at the walls. The wall slip occurs because the uid-wall interaction is weak: the interaction parameters used correspond to a partially wetting liquid with a sessile drop contact angle around 90 . The MD results are compared to a solution of the Stokes equa- tion which incorporates the observed slip via a Navier boundary condition. This continuum ow eld is ob- tained equivalently from a semi-analytic solutions in the literature [10] or from a direct numerical integration. The early-time agreement is only approximate, but this is not FIG. 2. Average wall forces in Couette ow. Bottom wall: surprising given that we have imposed a shear stress MD results in red (+), Navier-Stokes in blue (x); top wall: instantaneously and examined the response after only MD results in green (*). Navier-Stokes in magenta (x). about 20 ps. In this situation one expects some transient elastic or more generally non-Newtonian behavior in the uid, and speci cally a delayed response to the imposed shear. In the literature one nds, for example, Couette (1) The MD force is noisy, because the intermolecular force is a rapidly varying function of interatomic spac- ow calculations using FENE-P dumbbell [11] which give a velocity pro le which lags behind the Navier-Stokes so- ing. The plotted curve is an average over 20 statistically lution at early times. Furthermore, in Fig. 1b we also independent realizations, obtained from di erent initial show a good t to the MD velocity pro le using a sim- atomic velocity values, and even then a further time av- ple shear-thinning model in which the viscosity varies as erage is needed in order to identify the trend. The uc- 1=2 tuations are much more severe at early times (t < 200 ) =   (1 + = ) , where and are the local and 0 0 0 global strain rates. We have also reproduced the MD because the data points are spaced by 5 and the result- ing averaging interval is short, while the data is more pro le using a simple Maxwell model (not shown). The point of this discussion of the velocity pro le is to argue stable at later times where the interval is 50 . that even for time-dependent ows, \experimental" MD (2) The continuum results are as expected. The lower results for velocity are consistent with the usual theoreti- wall is abruptly set into motion, giving a large stress at cal expectations for liquids, provided one takes account of the start (nominally in nite) which relaxes to a constant inertia and a possible transient non-Newtonian response value when the velocity pro le stabilizes at linear. The to abrupt changes in ow conditions. liquid adjacent to the upper wall is at rest until the vor- ticity disturbance from the lower wall reaches it, so this We now make the analogous comparison for the forces force is initially zero and increases to its steady value the liquid exerts on the solid. In an MD calculation as the uid accelerates. In the steady state regime the the force on any liquid atom is the simple sum of pair- wall forces are equal and opposite because the uid is wise forces exerted by all other atoms within interaction no longer accelerating and the forces must sum to zero. range, so each wall force is the sum of forces between all The force using the shear-thinning model is slightly dif- uid atoms and atoms in that wall, and can be isolated ferent at early times but still in the middle of the MD easily. The continuum force is just the shear stress times uctuations. the wall area, L @u=@y, where the sign re ects the (3) The MD force agrees with the continuum force, mod- direction of the normal to the interface, u and x are the ulo the uctuations. The uid begins to move immedi- streamwise velocity and coordinate, and y is the spanwise ately and the wall force does not exhibit stick-slip be- coordinate. The results for the lateral force on the two havior or any analog of solid-on-solid static friction. The walls in Couette ow are displayed in Fig. 2, along with sharp peak in the force on the moving wall at early times the corresponding continuum forces obtained from the results from inertia (included in the Newtonian contin- Newtonian calculation incorporating slip. (The two non- uum description) plus a bit of elasticity. Newtonian models mentioned above give similar results.) Several features are to be noted in this gure. The main issue in this paper is the possibility of a 4 FIG. 3. Sessile drop driven across an atomically-smooth homogeneous substrate at times 0, 500 and 2500 (top to bottom). shear stress enhancement at the onset of motion, and in this con guration the cause is the abrupt motion of the wall. Indeed, the stress peak at t = 0 is absent when the wall velocity is linearly ramped up from 0 to 0.1, in place of the step change illustrated in the gure. Like- wise, in a similar simulation of Poiseuille ow, even if a step pressure gradient is imposed the shear stress in- creases smoothly from zero. The distinction is completely accounted for by the Stokes equations: the analytic solu- tions for the velocity in start-up Couette and Poiseuille ow with a no-slip boundary condition, given for exam- ple in [12], directly indicate whether a stress peak is present. These solutions involve a Fourier series which converges for the shear stress at the bottom wall at t = 0 for Poiseuille ow but diverges in the Couette case. III. SESSILE DROPS FIG. 4. Center of mass (top) and lateral force (bottom) vs. The simplest uid mechanical analog of solid-solid fric- time for a sessile drop forced across the various substrates at tion involves a drop sliding on a solid surface. Although f = 0:001, averaged over 20 realizations. Smooth: red (+), these is an extensive literature on this problem [13, 14] chemical step: green (*), physical step: blue (). relatively little attention has been given to the forces in- volved, and we use MD simulations to disentangle them. We begin with a partially-wetting liquid on a uniform seen in Fig. 3, uctuate during the motion. A constant and atomically-smooth substrate, in the form of cylin- external force is formally equivalent to tilting the sub- drical cap drop shown in Fig. 3. The advantage of this strate in the presence of gravity, although realistically shape compared to a spherical cap is that the system is the values of f used here are much too large for this statistically homogeneous in the direction (z) normal to interpretation. The behavior is similar over a range of the plane of the gure, and the results can be averaged forcing values near this one, 5  10 to 0.005, but for over z. The wall- uid interaction strength is chosen as signi cantly lower values the force on the drop is unmea- c = 0:85 which gives an initial contact angle of 70 . ff surable because the signal is swamped by the uctuations After equilibration a constant force f = 0:001m= is and the drop motion itself becomes intermittent. As for applied to each atom and the drop translates, as shown in higher forcing values, at f = 0:01 the center of mass still the subsequent frames of the gure. The Reynolds num- moves with constant velocity although the drop becomes ber, Re  XH = where H is the equilibrium drop 0 0 increasingly elongated and at still higher values the drop height and X is the center of mass velocity (see below), tends to y o the substrate. is 0.0748 and the Capillary number Ca  X= is 0.149, Most realistic solid surfaces are heterogeneous and ir- with slightly di erent values in the other cases. As the regular, and we can address these complications by forc- drop moves its shape changes and exhibits distinct ad- ing the drop across a modi ed substrate involving either vancing and receding dynamic contact angles which, as a chemical or a physical heterogeneity. The drop is equi- 5 FIG. 5. Sessile drop driven across an substrate with a chem- ical step. Pinned at f = 0:0005 (top) and in motion for f = 0:001 at times 500 (center) and 2500 (bottom). librated as before but to the right of the drop there is either a chemical step where the wettability coecient c drops to 0.65 or a physical step where the height of fw the solid rises by one fcc unit cell, a distance of 1.17 here. When a lateral force is applied the drop tilts in the direction of the force, developing distinct advancing and receding angles,  , but moves only if the force is A;R large enough. We illustrate this behavior in Fig. 5 for the chemical step case, where the value of the threshold force needed to move the drop over the step can be obtained FIG. 6. Flow inside the drop on a homogeneous substrate. by a the following force balance [2]. Two-dimensional velocity eld in the moving center of mass In equilibrium on a homogeneous surface a drop is held frame (top) and average lateral velocity pro le in the center in place by solid/liquid, solid/vapor and liquid/vapor of the drop in the substrate frame (bottom). surface tension forces at the contact line, with a lateral force balance given by Young's equation = SV SL cos  . For non-ideal surfaces there is a range of pos- Fig. 4 shows the drop center of mass X (t) moving with sible equilibrium contact angles. If a weak force f is nearly constant velocity from the start for both smooth applied to each atom, a drop tilts in the direction of the and heterogeneous surfaces. The total lateral force F (t) force but can remain at rest if the applied force is bal- the liquid exerts on the wall at f = 0:001 is given as a anced by the unbalanced surface tension forces per length function of time in Fig. 4. in all cases the wall force os- (cos  cos  ) at the receding contact line and simi- R 0 cillates about zero before the external force is applied at larly at the advancing line. Explicitly, the force balance time 100 and afterwards ramps monotonically up to to is fN = W (cos  cos  ), where N = 46800 is the R A a plateau. Only the transient behavior immediately after number of atoms in the drop and W length of the contact application of the external force shows a variation with line (the width of the drop in this situation, 51:3). For the structure of the substrate. The distinction between the chemical step, we observe that the drop appears sta- the size of the uctuations at early and late times is the tionary at force 0:0005, where it is pinned at the edge of same as in the previous Couette simulations: short and the step, but moves steadily at 0.0006 and higher values. long averaging intervals, respectively. The average force Using the drop density pro le at 0.0005 and de ning the in the plateau uctuates about the value of the net force liquid/vapor interface to be the contour where the den- applied to the liquid, Nf = 46:8, as expected because sity is half the bulk value, we estimate   67 and R the drop is not accelerating. It is possible to decompose 111 , giving f = 0:00055 from the force balance the wall force into capillary forces at the contact line equation, in agreement with the simulation. A similar plus frictional drag on the drop, but unfortunately the transition and corresponding force balance is found in dynamic contact angles could not be measured with any the case of a physical step. If there are no angles A;R accuracy, and we have not pursued this. in the equilibrium range satisfying the force balance the The motion of the liquid inside the drop is a combi- drop will move, and in the present simulations moving nation of rolling and slipping, as indicated in Fig. 6 for drops are expected to have have constant velocity, char- the smooth wall. The two-dimensional ow eld shown acteristic of linear friction in Stokes ow. is an average over 50 and evaluated in a reference frame We observe that, despite the shape uctuations seen moving with the drop center of mass; the resolution is in the gures, the motion is steady: for forcing f = 0:001 poor since the velocities involved are O(10 ) times the 6 random thermal velocity and the center of mass velocity is not exactly constant during the averaging interval, but rotation about the middle of the drop is evident. En- semble averaging tends to wash out the result for this two-dimensional eld, but is more e ective for the ve- locity pro le, which is instead evaluated in the reference frame of the solid wall and is also an average over the middle of the drop. The result is a roughly linear pro le as would correspond to rotation about the wall. (The up- permost points correspond to the liquid/vapor interfacial region where the density is falling o .) The presence of slip at the wall requires a small discussion. The region 0  y  1:71 is occupied by solid (at rest), and the lowest liquid point (at y = 2:565) is in the center of a nite-sized sampling bin in the liquid. The precise de - nition of the \solid wall" is always ambiguous at atomic resolution, but would certainly be somewhere in the re- gion 1:71  y  2:5, where u is non-zero. Aside from this ambiguity, as in Couette ow the presence of slip varies with the strength of the uid-wall interaction but the rolling motion is always present. The force exerted on the drop by the walls, can be ex- amined locally by computing the force as a function of the coordinate along the base of the drop. We use a sequence of slabs of wall, i  dx < x < (i + 1)  dx and 0 < y < L with dx = 1:71, and compute the force per area on each slab. (Note that this quantity is almost the x-y compo- nent of the uid's shear stress tensor at the wall. The quali cation is because, as in the slip discussion above, MD elds are always averages over a nite-sized sam- pling bin with the result assigned to its center, which is displaced from the wall by half the bin size. A further extrapolation would be required to determine the force at the wall.) In Fig. 7 we show the horizontal (x) and vertical (y) components of the force on the drop, along FIG. 7. Local elds for a drop on a homogeneous substrate, with the local density and lateral velocity, both in equi- at rest (top) and in motion (bottom) after 250 , in a single librium and while translating. The forces peak at the realization. Density: red (+), x-velocity: green (*), x-force: orange (x) and y-force: blue (). drop's two contact lines and are constant on average in the interior. In the x-direction, the signs of the two peaks correspond to the fact that the liquid/vapor surface ten- sion acts to contract the drop into to a circular cylinder in detail. The principal point we wish to make here is that when the drop moves the numerical values change but the attraction to the wall draws the edges of the drop but there is no qualitative di erence in local forces on outwards, corresponding to an inward force on the wall. The wall force in the interior of the drop is (statistically) the wall. The density is constant inside the drop, and its pro le simply shifts in x as the drop moves, while the constant, zero when the drop is in equilibrium and posi- tive during drop translation due to the liquid pulling to liquid velocity at the wall is constant, zero in equilibrium and approximately equal to the center of mass velocity the right. In the y-direction, the force peaks at the con- tact lines, again correspond to surface tension trying to during the motion. contract the drop by pulling upwards on the wall, while in the interior the force is a constant, corresponding to the Laplace pressure inside the drop. IV. CORNERED DROPS A similar but more detailed discussion of the variation of these local forces with contact angle and a comparison The experiments which motivated this study involved to Young's equation is given by Fernandez-Toledano, et spherical drops on a sliding stage held in place by a al., [15] for a di erent con guration involving a liquid measurement pin, and includes features of both of the bridge in equilibrium between two solid plane walls. In previous simulations. The liquid is forced into motion that case the drop is con ned by the walls but here is it abruptly as in Couette ow, with a localized impulsive is only attracted to one wall, so the force analysis di ers stress at a solid boundary, and then \slides" relative to 7 FIG. 8. Drop driven into a corner by motion of the bottom wall to the right, at times 0, 200 and 600 (top to bottom). The translation of the wall is shown by the red dot which indicates the location of an atom xed in the wall, the solid with (as we shall see) a combination of slip and rolling. Again we take advantage of MD's direct access FIG. 9. Streamlines (top) and local elds (bottom) for a to the force and use a simpler con guration with no vari- drop squeezed into a corner at u = 0:1: density: red (+), ation in one direction to improve the statistics. x-velocity: green (*), x-force: orange (x) and y-force: blue We place a drop in the corner formed by two distinct (). solid walls meeting at a right angle, as shown in Fig. 8, and choose the (partial) wetting coecient c = 0:75. ff The simulation box is a cube of side 68.4 containing wall is translated as a rigid (but slightly randomized) lat- 16,000 tetramer molecules, which initially ll an approx- tice. The reason is that at higher velocities the atom po- imate circular quadrant with a contact angle of 93 . The sitions lag signi cantly behind the tether positions unless vertical wall is xed in place while the horizontal wall is, a very high binding force is used, which would require a after equilibration, translated towards the xed wall at correspondingly very small time step to resolve. We have a constant velocity. Due to periodicity, the sliding wall veri ed that the uid motion and wall forces are not sen- can translate inde nitely. The gap where the two walls sitive to this modi cation. When the lower wall moves meet is the same size as the internal wall lattice spac- (right), the drop is squeezed into the corner, as shown in ing, and there is no interaction between the respective Fig. 8 for the u = 0:1 case, and evolves to a roughly wall atoms, and therefore no uid leakage or distortion time-independent con guration after several hundred  . of the corner or any solid frictional heat generated there. In the steady state, the advancing contact line at the hori- These simplifying properties would be dicult to realize zontal wall (in the rest frame of this wall the contact line in a laboratory experiment, although the di erences in- advances to the left) increases to a velocity-dependent volve only the corner region which is not the focus of this value (100 to 145 ), while the static angle on the vertical study in any case. wall shows little change. The motion inside the drop is We consider a range of wall velocities 0.01 to 1.0= . rolling, as seen in Fig. 9, accompanied by some slip at At the lower velocities we use the same time step and wall the wall. The horizontal force the liquid exerts on the tethering strength as in the previous simulations, but for on the wall is negative, opposing the wall motion, with a the higher velocities the time step is lowered to 0.001 peak at the corner where a continuum no-slip ow eld and instead of tethering thermally active wall atoms, the would have a stress singularity, while the vertical force 8 FIG. 11. Time decay of uid density and velocity near the wall, and the x; y forces on the drop at u = 1:0; color code as in Fig. 7. squeezing the uid to the right and upwards and the uid resists this. The other notable feature of the forces is the systematic weak decay (towards zero) following the peak, surprising since one would have expected constant wall forces in an apparently steady ow. However, in this situation the drop is not con ned and is free to adjust its position. intuitively one might imagine the drop pushed upwards by repeated collisions with the sliding wall atoms and in fact precisely this behavior is observed. In Fig. 11 we plot the density and velocity in the uid adjacent to the wall at low and high velocities, normalized to their values at time when the wall begins to move. The initial peak FIG. 10. Time variation of the x (top) and y (bottom) force results from the abrupt start, following which there is on the sliding wall for various velocities. u = 0:1: red (+); a weak decay of the density, meaning uid moves away u = 0:25: green (*); u = 0:5: orange () and u = 1:0: blue from the wall, which produces parallel decays in the uid (x). velocity and wall forces. has the same interpretation as for sessile drops. At higher V. THE EXPERIMENT wall velocities, the liquid density pro le becomes slightly asymmetric and larger near the corner, the slip velocity increases, also asymmetrically but instead largest at the There remains the question of interpreting the exper- contact line, and the force pro les increase in magnitude iments of Gao et al. The principal distinction between but maintain the same shape. that experiment and the drop simulations here is the dis- The x and y components of the total force on the bot- tortion of the drop surface as it moves relative to the mea- tom wall are shown in Fig. 9, for di erent values of the suring pin. Rearrangement of the liquid drop's shape and wall velocity. In all cases there is an initial peak, resem- the contact lines on the pin and on the substrate would certainly a ect the force exerted We have carried out sev- bling that appearing in Couette ow, which we attribute to the inertial e ects of an impulsive start. The rea- eral simulations on this process, involving a spherical cap drop and a nite-sized obstacle pin, rather than a bar- son is that the duration of the stress peaks is 100-150 (note that wall motion begins at 100 ) which is compara- rier wall as in the previous section, but these simulations do not completely reproduce the experiments in terms of ble to the vorticity di usion time across the drop, R = with drop size R  35 and viscosity  = 0:518m=( , drop distortion and wall force. which characterizes ow development. The signs of the An example is given in Fig. 12 { a spherical cap is uid forces on the wall are negative because the walls are placed on a partially wetting substrate (c = 0:85) and ff 9 istic of solid/liquid friction. FIG. 12. Simulation of the experiment of Gao et al. [4]. Top: side and top view of the equilibrated system, Bottom: e ects of sliding the substrate to the left, at times 1000 and 5000 . equilibrated with a rectangular pin inserted in the drop from above. Most of the pin has the same interaction as the substrate, except that the upstream face is more strongly wetting (c = 1:2), in order to mimic the ex- ff periment where a metallic re ecting layer was added to improve the imaging. After equilibration the substrate is translated to the right at velocity u = 0:1 and the force FIG. 13. Force on the pin in the simulation shown in Fig. 12. on the pin recorded. Fig. 13 again shows a monotonic rise to a near-plateau, with no \static-friction" enhance- ment. At times beyond 5000 the drop becomes highly VI. CONCLUSIONS elongated and eventually detaches from the pin, but dur- ing the interval indicated the footprint of the drop on We have used MD simulations to investigate the pos- the substrate remains approximately circular. The re- sibility of an enhanced shear stress before a liquid begins sults are similar when the conditions of the simulation to move across a solid surface, analogous to the distinc- are varied (di erent wettabilities, di erent pin shapes, tion between static and dynamic friction when two solids di erent speeds, etc.), and provided the pin remains em- move relative to each other. In cases where the motion bedded inside the drop the force on the pin is roughly is driven by a constant body force applied to the inte- constant. Exceptions to the typical behavior are found rior of the liquid, even in a step-wise fashion, the stress when the pin is at the edge of the drop, when the liquid is found to increase monotonically from zero to a steady either rst wets or dewets the pin. In these cases the state value. An abrupt motion of a solid bounding sur- liquid/vapor interface does deform and a transient peak face, however, generates a large local strain which in turn or spike appears in the force. produces a peak in the shear stress. Such peaks are in - In contrast to the simulations, the experimental drops nite in the (mathematical) continuum limit but regular- do change shape and, in particular, their contact line ized to large but nite values in MD simulations and in length increases during the time interval when the force real life, but otherwise entirely in accord with the Navier- is enhanced. A further distinction in the simulations, a Stokes equations. Furthermore, even in situations where consequence of size limitations, is that the liquid climbing the liquid is pinned by inhomogeneities and requires a the wetting side of the pin has a density corresponding minimum threshold force for continuous motion, no force to liquid/vapor interfacial region and does not faithfully enhancement is found. The experiments [4] which mo- represent bulk liquid and may not exert the proper hy- tivated this work appear to incorporate changes in drop drodynamic drag. Nonetheless, it appears that the tran- shape and thereby involve interfacial dynamics as well as sient friction enhancement in the experiments is related wetting considerations, and do not provide evidence for to change of drop shape, and is not a general character- an analog of static solid/solid friction. 10 [1] J.C. Anderson, K.D. Leaver, R.D. Rawlings and J.M. channel, Phys. Fluids 18, 031104 (2006). Alexander, Materials Science, 4th ed., (Springer, New [9] J. Koplik and C. Maldarelli, Di usivity and hydrody- York, 1990). namic drag of nanoparticles at a vapor-liquid interface, [2] P.-G. de Gennes, F. Brochard-Wyart and D. Qu er e, Cap- Phys. Rev. Fluids 2, 024303 (2017). illarity and Wetting Phenomena: Drops, Bubbles, Pearls, [10] G. Kaoullas and G. C. Georgiou, Start-up and cessation Waves (Springer, New York, 2004). Newtonian Piseuille and Couette ows with dynamic wall [3] B. N. J. Persson, Sliding Friction, 2nd ed. (Springer, New slip, Meccanica 50, 1747 (2015). York, 2000). [11] Y. Mochimaru, Unsteady-state development of plane [4] N. Gao, F. Geyer, D. W. Pilat, S. Wooh, D. Vollmer, H.- Couette ow for viscoelsatic uids, J. Non-Newtonian J. But and R. Berger, How drops start sliding over solid Fluid Mech. 12, 135 (1983). surfaces, Nature Phys. 14, 191 (2018). [12] C. Pozrikidis, Introduction to Theoretical and Computa- [5] M. P. Allen and D. J. Tildesley, Computer Simulation of tional Fluid Dynamics (Oxford, New York, 1997). Liquids (Clarendon, Oxford, 1987). [13] D. Bonn, J. Eggers, J. Indekeu, J. Meunier and E. Rolley, [6] D. Frenkel and B. Smit, Understanding Molecular Simu- Wetting and spreading, Rev. Mod. Phys. 81. 739 (2009). lation, 2nd ed. (Academic, San Diego, 2002). [14] Special section containing papers on the dynamics of wet- [7] G. S. Grest and K. Kremer, Molecular dynamics simula- ting, J. Phys.: Condens. Matter, 21, 464117 (2009). tion for polymers in the presence of a heat bath, Phys. [15] J.-C. Fernandez-Toledano, T. D. Blake, P. Lambert and Rev. A 33, 3628 (1986). J. De Coninck, On the cohesion of uids and their adhe- [8] T. S. Lo, J. Koplik, M. Rauscher and S. Dietrich, Pearling sion to soilds, Adv. Colloid Interface Sci. 245, 102 (2017). instability of nanoscale uid ow con ned to a chemical

Journal

Condensed Matter – arXiv (Cornell University)

Published: Sep 14, 2018

There are no references for this article.