Strong dynamical effects during stick-slip adhesive peeling

Abstract

We consider the classical problem of the stick-slip dynamics observed when peeling a roller adhesive tape at a constant velocity. From fast imaging recordings, we extract the dependence of the stick and slip phase durations on the imposed peeling velocity and peeled ribbon length. Predictions of Maugis and Barquins [in Adhesion 12, edited by K. W. Allen, Elsevier ASP, London, 1988, pp. 205–222] based on a quasistatic assumption succeed to describe quantitatively our measurements of the stick phase duration. Such a model however fails to predict the full stick-slip cycle duration, revealing strong dynamical effects during the slip phase.

---

1. Introduction

Everyday examples of adhesive peeling are found in applications such as labels, stamps, tape rollers, self-adhesive envelopes or post-it notes. During the peeling of those adhesives, a dynamic instability of the fracture process corresponding to a jerky advance of the peeling front called “stick-slip” may occur. This stick-slip instability has been an industrial concern since the 1950's because it leads to noise levels above the limits set by work regulations, to adhesive layer damage and/or to mechanical problems on assembly lines. This instability is still a limiting factor for industrial productivity due to the limitations of generic technical solutions applied to suppress it, such as anti-adhesive silicon coating.

From a fundamental point of view, the stick-slip instability of adhesive peeling is generally understood as the consequence of an anomalous decrease of the fracture energy Γ(ν_p) of the adhesive–substrate joint in a specific range of peeling front velocities ν_p.^1–8 Indeed, when the peeling process also involves a compliance between the point where the peeling velocity is imposed and the fracture front, this decreasing fracture energy naturally leads to oscillations of the fracture velocity ν_p around the mean velocity V imposed by the operator. Often, it is simply the peeled ribbon elasticity which provides compliance to the system. From a microscopic perspective, such an anomalous decrease of the fracture energy Γ(ν_p) (correctly defined for stable peeling only) could correspond (but not necessarily) to the transition from cohesive to adhesive failure^2,3 or between two different interfacial failure modes.^4,7 More fundamentally, this decrease of the fracture energy has been proposed to be the consequence of the viscous dissipation in the adhesive material.⁹ De Gennes¹⁰ further pointed out the probable fundamental role of the adhesive material confinement (which was evidenced experimentally in ref. 3) in such viscoelastic theory. Since then, it has however appeared that a model based on linear viscoelasticity solely cannot be satisfactory and that the role of creep, large deformations and temperature gradient in the adhesive material is important (ref. 11–14 and references therein).

Experimentally, the stick-slip instability was first characterized thanks to peeling force measurements which revealed strong fluctuations in certain ranges of peeling velocity.^1,3,5,6 Since then, it has also been studied through indirect measurements of the periodic marks left on the tape^5,6,15,16 or of the emitted acoustic noise.^17,18 Thanks to the progress in high speed imaging, it is now possible to directly access the peeling fracture dynamics in the stick-slip regime.^19–21

In the late 1980's, Barquins and co-workers^5,6 performed a series of peeling experiments of a commercial adhesive tape (3M Scotch® 602) at a constant pulling velocity V and for various lengths of peeled ribbon L. For the considered adhesive, the velocity range for which stick-slip was evidenced, thanks to peeling force fluctuation measurements, was shown to be 0.06 < V < 2.1 m s⁻¹. In a sub-range of unstable peeling velocity 0.06 < V < 0.65 m s⁻¹, the authors succeeded to access the stick-slip cycle duration thanks to the post-mortem detection of periodic marks left on the tape by stick-slip events. Moreover, they managed to model quantitatively the measured stick-slip period,^5,6 assuming the fracture dynamics to remain a quasistatic problem during the stick phase and backing on measurements of the stable branch of the fracture energy Γ(ν_p) at low peeling velocities below the instability onset.

In this article, we revisit these experiments by studying the stick-slip dynamics during the peeling of a roller adhesive tape at an imposed velocity. The principal improvement compared to Barquins's seminal work is that, thanks to a high speed camera coupled to image processing, we are able to access the dynamics of the peeling fracture front. We focus on the study of the duration of the stick-slip cycle and its decomposition into stick and slip events, for which data are inaccessible through other techniques. We present experimental data of the stick and slip durations for a wide range of imposed peeling velocities V and for different peeled ribbon lengths L. We show that the model proposed by Barquins and co-workers^5,6 describes the evolution of the duration of the stick phase, but fails to predict the duration of the whole stick-slip cycle due to unexpectedly long slip durations.

2. Experimental setup

In this section, we describe briefly the experimental setup which has already been presented in detail in a recent study.²¹ We peel an adhesive tape roller (3M Scotch® 600, made of a polyolefin blend backing coated with a layer of a synthetic acrylic adhesive, also studied in ref. 8, 19, and 21), mounted on a freely rotating pulley, by winding up the peeled ribbon extremity on a cylinder at a constant velocity V using a servo-controlled brushless motor (see Fig. 1). The experiments have been performed at a temperature of 23 ± 2° and a relative humidity of 45 ± 5%. The width of the tape is b = 19 mm, its thickness e = 38 μm and its Young's modulus E = 1.26 GPa.

Fig. 1 Schematic view of the experimental setup. The angles α and β are oriented clockwise and counterclockwise respectively. Roller diameter: 40 mm < 2R < 58 mm, roller and tape width: b = 19 mm.

Each experiment consists of increasing the winding velocity from 0 up to the target velocity V. Once the velocity V is reached, it is maintained constant for two seconds before decelerating the velocity back to zero. When stick-slip is present this 2-second stationary regime of peeling provides sufficient statistics to compute well converged stick-slip mean features. We have varied the imposed velocity V from 0.0015 to 2.5 m s⁻¹ for different values of the peeled tape length between L = 0.08 and 1.31 m. During the experiment, the peeled tape length L (Fig. 1) is submitted to variations, due to the stick-slip fluctuations and slow oscillations of the peeling point angular position, which however always remain negligible compared to its mean value (less than 0.3%).²¹

3. Peeling force measurement

Thanks to a force sensor (Interface® SML-5) on the holder maintaining the pulley, we are able to measure the mean value of the force F transmitted along the peeled tape during one experiment. When peeling is stable, we compute the strain energy release rate G from the mean value of the force F, following the traditional relationship for the peeling geometry^22,23for a peeling angle θ ≃ 90° (see Fig. 1). The quantity G corresponds to the amount of mechanical energy released by the growth of the fracture by a unit surface. The right-hand term of eqn (1) finally simply takes into account the work done by the operator but discards the changes in the elastic energy stored in material strains (term (F/b)²/2Ee in eqn (1))²³ which are negligible here. Indeed, the maximum encountered force in our experiments is typically of about 2 N, which gives F/b ≈ 100 J m⁻², to be compared to (F/b)²/2Ee ≈ 0.12 J m⁻².

In the context of elastic fracture mechanics, the condition for a fracture advance at a constant velocity ν_p is a balance between the release rate G and the fracture energy Γ(ν_p) required to peel a unit surface and accounting for the energy dissipation near the fracture front. When the fracture velocity ν_p approaches the Rayleigh wave velocity ν_R, Γ(ν_p) also takes into account the kinetic energy stored in material motions which leads to a divergence when ν_p → ν_R.²⁴ In our system, the strain energy release rate G, computed through eqn (1), therefore stands as a measure of the fracture energy Γ(ν_p) when the peeling is stable only, i.e. when ν_p is constant. We will nevertheless compute G for the experiments in the stick-slip regime for which the peeling fracture velocity ν_p(t) is strongly fluctuating in time. In such a case, G cannot be used as a measure of the fracture energy: it is simply the time average of the peeling force F in units of G.

In Fig. 2, we plot G as a function of the imposed peeling velocity V for three different peeled tape lengths L. When the peeling is stable, the peeling force is nearly constant in time, whereas it fluctuates strongly when the stick-slip instability is present. The standard deviation of these fluctuations is represented in Fig. 2 with error bars. Large error bars are indicative of the presence of stick-slip.

Fig. 2 Mean value of the peeling force F, in units of the strain energy release rate G = F/b, as a function of V for 3 different peeled tape lengths L. Stars report the data of Barquins and Ciccotti⁸ for the same adhesive. The solid line is a power law fit G = 137V^0.146 of the data in the low velocity stable branch. Error bars represent the standard deviation of the force fluctuations during one experiment.

Between V = 0.0015 m s⁻¹ and V = 0.10 ± 0.03 m s⁻¹, we observe that G = F/b increases slowly with V and that its temporal fluctuations are nearly zero, revealing that the peeling is stable. This increasing branch G(V) is therefore a measure of the adhesive fracture energy Γ(ν_p = V) = G(V) for V < 0.10 ± 0.03 m s⁻¹. Our results are compatible with the data reported by Barquins and Ciccotti⁸ for the same adhesive tape (see Fig. 2). However, they explored a much larger range of velocities in this stable branch of peeling, down to V = 10⁻⁵ m s⁻¹. By using both series of measurements, it is reasonable to model the stable peeling branch with a power law, G(V) = aVⁿ, with n = 0.146 and a = 137. For 0.10 ± 0.03 m s⁻¹ < V ≤ 2.5 m s⁻¹, we observe that the measured value of G(V) decreases with V. This tendency, which was already observed in previous experiments,²⁵ is accompanied with the appearance of temporal fluctuations which are the trace of the stick-slip instability. From these data, we can estimate the onset of the instability to be V_a = 0.10 ± 0.03 m s⁻¹. The measured decreasing branch of G(V) for V > V_a appears as a direct consequence of the anomalous decrease of the fracture energy at the origin of the instability. It is important to note that the measured mean value of G = F/b is nearly independent of the length of peeled ribbon L. This result is natural in the stable peeling regime but was a priori unknown in the stick-slip regime.

Barquins and Ciccotti⁸ succeeded to measure a second stable peeling branch for V ≥ 19 m s⁻¹. This increasing branch constitutes a measure of the peeling fracture energy Γ(ν_p = V) = G(V) in a fast and stable peeling regime. In ref. 8, this branch is inferred to exist for velocities even lower than V = 19 m s⁻¹, although it was not possible to measure it. Backing on the data of ref. 6 for a very close adhesive, one can however guess that the local minimum value of G(V), corresponding to a velocity in the range 2.5 m s⁻¹ < V < 19 m s⁻¹, would be bound by G_0,1 = 18 < G < G_0,2 = 33 J m⁻².

4. Peeling point dynamics

The local dynamics of the peeling point is imaged using a high speed camera (Photron FASTCAM SA4) at a rate of 20 000 fps. The recording of each movie is triggered once the peeling has reached a constant average velocity V ensuring that only the stationary regime of the stick-slip is studied. Through direct image analysis,²¹ the movies allow access to the curvilinear position of the peeling point [small script l]_α(t) = Rα in the laboratory frame (with α being the angular position of the peeling point and R being the roller diameter, α > 0 in Fig. 1). Image correlations on the adhesive tape roller contrast pattern further allow direct access to its angular velocity dβ/dt in the laboratory frame (where β is the unwrapped angular position of the roller, β > 0 in Fig. 1, [small script l]_β(t) = Rβ). We finally compute numerically the curvilinear position [small script l]_p(t) = [small script l]_β(t) + [small script l]_α(t) and velocity ν_p(t) = d[small script l]_p/dt of the peeling point in the roller reference frame.

The curvilinear position of the peeling point [small script l]_α(t) in the laboratory frame is actually estimated from the position of the peeled ribbon at a small distance of 0.30 ± 0.05 mm from the peeling fracture front on the roller surface. We therefore do not detect strictly the peeling fracture front position but a very close quantity only. This procedure can consequently introduce some bias in our final estimation of the fracture front velocity ν_p(t). This bias is notably caused by the changes in the radius of curvature of the tape at the junction with the substrate which are due to the force oscillations in the peeled tape characteristics of the stick-slip instability. Such an effect actually biases the measurement toward larger velocities during the stick phase and lower velocities during the slip phase. Another effect that leads to uncertainties in velocity measurement is the emission of a transverse wave in the peeled tape when the fracture velocity abruptly changes at the beginning and at the end of slip phases.

Fig. 3(a) and (c) represent the fracture position [small script l]_p(t) and velocity ν_p(t) as a function of time for a typical experiment performed at V = 0.55 m s⁻¹ and L = 0.47 m. In these figures, we observe alternate phases of slow (stick phase) and fast (slip phase) peeling which are the signatures of the stick-slip motion. These large velocity fluctuations are quite regular in terms of duration and to a lesser extent in terms of amplitude at least at the considered peeling velocity. Our general data analysis further consists of the decomposition of the signal of instantaneous peeling velocity ν_p(t) into stick-slip cycles by setting the beginning of each cycle at times t_n (n denoting the n^th cycle) when ν_p(t_n) = V and dν_p(t_n)/dt < 0. From these data, we extract the duration T_ss of each stick-slip cycle for which we define a rescaled time t′ = (t − t_n)/T_ss. We further compute the phase averaged evolution of the peeling fracture velocity ν_p(t′) from t′ = 0 to 1 considering all the stick-slip cycles in one experiment. With this procedure, we finally extract for each peeling velocity V and peeled tape length L the typical fracture velocity evolution during a stick-slip cycle getting rid of intrinsic fluctuations of the stick-slip period. In Fig. 3(b) and (d), we show the phase averaged position and velocity profiles, corresponding to Fig. 3(a) and (c) respectively, as a function of t = t′〈T_ss〉 (〈〉 denotes the ensemble averaged value over all the cycles in one experiment).

Fig. 3 (a) Peeling point position [small script l]_p(t) in the roller reference frame for an experiment performed at V = 0.55 m s⁻¹ and L = 0.47 m. The dashed line shows [small script l]_p = Vt, with V the average peeling velocity. (b) Corresponding phase averaged peeling point position as a function of t′〈T_ss〉 (see the main text). (c and d) Corresponding instantaneous (c) and phase averaged (d) peeling point velocity ν_p. The dashed horizontal lines show the average peeling velocity V and the continuous horizontal lines show 3V_a. In (b) and (d), the vertical lines show the transitions between the stick (ν_p < 3V_a) and the slip (ν_p > 3V_a) phases.

From these phase averaged velocity profiles, we define, for each experimental condition V and L, stick events as continuous periods during which ν_p(t) < 3V_a and slip events as continuous periods during which ν_p(t) > 3V_a. According to the model of Barquins et al.^5,6 a natural threshold in order to separate the stick and slip phases is the onset of the instability V_a (as defined in Fig. 2). However, as discussed previously, due to the procedure used for the detection of the peeling point, our measurement of the fracture velocity can be affected by biases caused by the variation of the tape curvature at the peeling point and by the propagation of transverse waves in the tape. The effect of the latter can be observed in Fig. 3(d) in the early stage of the stick phase. In order to avoid taking into account the velocity biases in the decomposition of the stick-slip cycle, we chose for the threshold separating the stick and slip phases a value little larger than the “theoretical” threshold V_a, that is to say 3V_a.

Finally, as we have shown recently in ref. 21, when the peeling velocity V is increased, low frequency pendular oscillations of the peeling angle θ develop. Due to the dependence of the stick-slip instability onset on the mean peeling angle, these oscillations lead to intermittence in the stick-slip dynamics for peeling velocities V > 1.5 m s⁻¹. We therefore exclude the experiments with V > 1.5 m s⁻¹in the sequel. For the studied experiments, we have a mean peeling angle 〈θ〉 = 90 ± 3° with slow temporal variations in the range Δθ = ±15° during one experiment.

5. Stick-slip cycle duration

From the signal of peeling point position [small script l]_p(t) (see Fig. 3(a)), we define the stick-slip amplitude A_ss as the distance travelled by the fracture during a stick-slip cycle. In Fig. 4, we report this amplitude A_ss for each stick-slip event as a function of the corresponding stick-slip period T_ss, for all events in 6 different experiments. These data gather close to the curve A_ss = VT_ss. The large spread of the data along the curve A_ss = VT_ss reflects the statistics of the stick-slip cycle amplitude and duration which could be due for instance to adhesive heterogeneities. In contrast, the dispersion of the data around the curve A_ss = VT_ss is much smaller. It actually estimates the discrepancy between the imposed velocity V and the averaged fracture velocity for each stick-slip cycle. The observed small discrepancy actually traces back both measurement errors in the instantaneous fracture velocity and intrinsic fluctuations of the dynamics.

Fig. 4 Stick-slip amplitude A_ss as a function of stick-slip period T_ss for each stick-slip cycle in 6 different experiments with L = 0.47 and 1.31 m and V = 0.30, 0.55 and 1.00 m s⁻¹. The lines represent the curves A_ss = VT_ss.

In Fig. 4, one can already see that the statistically averaged values of A_ss and T_ss increase with L for a given peeling velocity V. In the following, we will focus on the study of the statistical average 〈T_ss〉 of the duration of the stick-slip oscillation and its decomposition into stick and slip phases with in mind the aim of testing the description of Barquins, Maugis and co-workers.^5,6 There is no need to study the averaged stick-slip amplitude 〈A_ss〉 since it is unequivocally related to 〈T_ss〉 through 〈A_ss〉 = V 〈T_ss〉.

In Fig. 5(a), we plot the mean stick-slip duration T_ss as a function of V for three different lengths L of the peeled ribbon. The data correspond to the average 〈T_ss〉 and the error bars correspond to the standard deviation of the statistics of T_ss over all the stick-slip events in each experiment. In the following, since we will consider the averaged values only, we will skip the brackets 〈〉. At first sight, it appears that, within the error bars, the stick-slip duration T_ss is stable over the major part of the explored range of peeling velocity V. One can however note that, independently of L, T_ss tends to decrease with V for V ≤ V_c = 0.6 ± 0.1 m s⁻¹. Such behavior is compatible with the observations of Barquins et al.⁵ but appears here over a rather limited velocity range. The characteristic velocity V_c = 0.6 ± 0.1 m s⁻¹ above which T_ss is nearly constant seems not to depend strongly on the length of the peeled ribbon L.

Fig. 5 Average stick-slip cycle duration T_ss as a function of the average peeling velocity V, for different lengths of the peeled ribbon L. (b) Average stick-slip, stick and slip durations as functions of the average peeling velocity for L = 0.47 m. Each data point corresponds to the average and each error bar to the standard deviation of the statistics over one experiment.

In Fig. 5(b), we show the mean durations of stick and slip events, T_stick and T_slip respectively, as a function of the imposed peeling velocity V for the experiments performed with the peeled length L = 0.47 m. Interestingly, we observe that the stick and slip phases evolve differently with V: the stick duration decreases with V, while the slip duration increases over the whole explored range of V. In consequence, the ratio T_stick/T_slip, presented in Fig. 6, decreases with V from T_stick/T_slip ∼ 4 ± 1 down to T_stick/T_slip ∼ 0.3 ± 0.2. Such behavior of T_stick/T_slip appears to be very little dependent on L according to Fig. 6. For V ≥ 0.90 ± 0.05 m s⁻¹, T_stick/T_slip becomes smaller than 1, meaning that the slip phase is longer than the stick one. Our data therefore show that it is not possible to neglect the slip duration compared to the stick duration in general.

Fig. 6 T _stick/T_slipvs. V for 3 different L. Each data point corresponds to the average and each error bar to the standard deviation of the statistics over one experiment.

6. Model

In this section, we compare our experimental data with the model proposed by Barquins, Maugis and co-workers in ref. 5 and 6. This model is based on measurements of the stable branch of the fracture energy Γ(ν_p) for low peeling velocities below the instability onset V_a, and on the following assumptions:

• During the stick phase, the equilibrium between the instantaneous energy release rate G = F/b and the fracture energy Γ(ν_p) (of the low velocity stable branch) is still valid dynamically, i.e. G(t) = Γ(ν_p(t)).

• The peeled ribbon remains fully stretched during the peeling, which means

where u is the elongation of the tape with Young's modulus E and thickness e.

• The slip duration is negligible compared to the stick duration.

Backing on these hypotheses, it is possible to derive a prediction for the stick-slip duration T_ss. Introducing the inverted function ν_p = Γ⁻¹(G) and noting that du/dt = V − ν_p (see the next paragraph and ref. 21), eqn (2) leads to the dynamical relationship

which can be integrated over the stick phase to getG_a is the maximum value of Γ(ν_p) at the end of the “slow” stable branch Γ_slow(ν_p). G₀ is the minimum value of Γ(ν_p) at the beginning of the “fast” stable branch Γ_fast(ν_p) (see Fig. 2) and is assumed to be also the value of G at which the stick phase starts on the slow branch after a slip phase.

In this model, the ribbon is assumed to remain taut during the whole stick-slip cycle. In order to challenge the validity of this hypothesis, let us estimate the evolution of the elongation u(t) of the tape as a function of time. If we note P(t) the peeling point position and M the point where the peeled tape is wound, we can define the quantity u(t) as the difference between the distance and the length of the peeled tape in the unstrained state. If u(t) is positive, this quantity indeed measures the elongation of the tape as in eqn (2), whereas it measures the excess of slack tape if it is negative. Following ref. 21, one can show that

Since in our experiments the peeling angle θ is close to 90° and the roller rotation velocity Rdβ/dt sticks to the imposed peeling velocity V to a precision always better than ±1.5%,²¹ we finally have . The elongation/slack u(t) increases with during the stick phase and decreases with the same amplitude during the slip phase. This compensation is ensured by the fact that the averaged velocity over the stick-slip cycle matches the imposed velocity V, i.e., and is valid whether or not the tape remains always taut during the stick-slip cycle.

To test the relevance of the hypothesis of a tape always in tension, one can actually compare the increase/decrease Δu of the quantity u(t) during the stick/slip phase with the one predicted by the quasistatic model of Barquins and co-workers

for an always taut tape. Throughout our data, the relative discrepancy (Δu_theo − Δu)/Δu is typically less than 15% which confirms the relevance of the assumption of a tape in tension during the whole stick-slip cycle.

An equivalent but more instructive way to test the model of Barquins and co-workers is to integrate numerically eqn (4) and compare it with experimental measurements of the stick duration. To do so, we use the fit of the data of the energy release rate G(V) in Fig. 2, i.e. G(V) = Γ_slow(V) = aVⁿ, with n = 0.146 and a = 137. The value of G₀ is affected by a significant uncertainty in our data. We will therefore use two different guesses corresponding to the limit values introduced in page 3 (see G_0,1 and G_0,2 in Fig. 2). These values of G₀ correspond to two limit values of the fracture velocity at the beginning of the stick phase: V_0,1 = 10⁻⁶ m s⁻¹ measured in another adhesive but with a close behavior,⁶ and V_0,2 = 6.3 × 10⁻⁵ m s⁻¹ which is an upper limit for V₀ according to the data of Fig. 2.

In the inset of Fig. 7(b), we report the measured data for T_ss/L as a function of V for three different lengths L as well as the predictions of eqn (4) with V_0,1 (solid line) and V_0,2 (dashed line). The model appears compatible with the experimental data only for a marginal range of very low peeling velocities. Once V > 0.5 m s⁻¹, the measured values of T_ss/L indeed deviate more and more from the theoretical prediction. A first natural explanation for this discrepancy is that the assumption of a negligible slip duration T_slip (barely verified for low velocities for which 0.25 < T_slip/T_stick < 0.5) becomes more and more false as V is increased (see Fig. 6).

Fig. 7 (a) T_slip/L, (b) T_stick/L and T_ss/L (insert) vs. V for 3 different L. Each data point corresponds to the average and each error bar to the standard deviation of the statistics over one experiment. In (a), the curve close to the x-axis represents the theoretical prediction for a quasistationary slip phase. In (b), the lines show the predictions of eqn (4) with V_a = 0.10 m s⁻¹ and V_0,1 = 10⁻⁶ m s⁻¹ (solid line) or V_0,2 = 6.3 × 10⁻⁵ m s⁻¹ (dashed line).

In Fig. 7(b) we therefore directly plot T_stick/L as a function of V, along with the prediction (4). One can note that the theoretical predictions using the two limit guesses for V₀ are not very different. A first interesting result is that the stick duration appears, to the first order, proportional to the peeled tape length L as evidenced by the reasonable collapse of the data T_stick/L on a master curve, which is compatible with the analytical prediction of the model (4). But more importantly, we observe that for the range of velocities explored, the model for T_stick, which does not use any adjustable parameter, reproduces very well the experimental data.

Obviously, one can consider an equivalent quasistationary approximation during the slip phase in order to predict the slip duration using Γ_fast⁻¹(G) instead of Γ_slow⁻¹(G) in eqn (4). Here, Γ_fast⁻¹(G) corresponds to the inverse of the energy fracture G = Γ_fast(ν_p) in the fast and “stable” peeling regime of Fig. 2. The integration using the model of the fast branch Γ_fast(V) = 6.5 × 10⁻⁵V^4.5 (see Fig. 2) however leads to values of T_slip always 2 orders of magnitude smaller than the experimental values as evidenced in Fig. 7(a). It is however worth noting that the collapse of the data T_slip/L for the different L values shows that T_slip increases nearly linearly with L.

7. Discussion

In this paper, we report experiments on a roller adhesive tape peeled at a constant velocity focusing on the regime of stick-slip instability. From fast imaging recordings, we extract the dependence of the stick and slip phase durations on the imposed peeling velocity V and peeled ribbon length L.

The stick phase duration T_stick of the stick-slip oscillations is shown to be nearly proportional to the peeled tape length L and to decrease with the peeling velocity V. These data moreover appear in quantitative agreement with the predictions of a model proposed by Barquins, Maugis and co-workers in ref. 5 and 6 which do not introduce any adjustable parameter. This successful comparison confirms the relevance of the two main assumptions made in the model: (i) the tape remains in tension during the whole stick-slip cycle; (ii) the principle of an equilibrium between the instantaneous energy release rate G(t) = F(t)/b and the fracture energy Γ(ν_p(t)), as measured in the steady peeling regime, is valid dynamically during the stick phase.

Describing the peeling dynamics as a function of time t by the knowledge of the fracture velocity ν_p(t) and of the force F(t) = bG(t) in the peeled tape, the considered model further assumes that the system jumps instantaneously, at the end of the stick phase, from the “slow” stable branch to the “fast” stable branch of the steady fracture energy G = Γ(ν_p) and then instantaneously backward from the “fast” branch to the “slow” branch at the end of the slip phase. In such a framework, reproducing the assumptions (i) and (ii) for the slip phase leads to a prediction for the slip duration. We have shown that this prediction is at least hundred times smaller than the slip phase duration T_slip measured in our experiments. We actually report that in contrast to what is finally proposed in ref. 5 and 6, the slip duration T_slip cannot be neglected compared to the stick one T_stick, since it is at best 4 times smaller, and becomes even larger than T_stick for V ≥ 0.90 ± 0.05 m s⁻¹.

These last experimental results account for the existence of strong dynamical effects during the slip phase which can therefore not be described by a quasistatic hypothesis. These dynamical effects could be due to the inertia of the ribbon close to the fracture front. Some models also predict a strong influence of the roller inertia.^18,26 Notably, thanks to numerical computation, De and Ananthakrishna²⁶ have shown that for certain values of the roller inertia, the slip phase could consist of several jumps from the “fast stable” branch to the “slow stable” branch in the (ν_p, G = Γ(ν_p)) diagram. Such a process would certainly produce a longer slip time than expected in the framework of Barquins's model. It would be most interesting to confront our experimental observations to the predictions of this model, based on a detailed set of dynamical equations and ad hoc assumptions made on the velocity dependence of Γ. However, such a comparison is not straightforward in our current setup since we do not have the temporal and spatial resolutions to detect such eventual fast oscillations. Besides, in order to obtain a quantitative comparison, measurement of the instantaneous peeling force F(t) is required but it remains a challenge.

Acknowledgements

This work was supported by the French ANR through Grant “STICKSLIP” no. 12-BS09-014. We thank Costantino Creton and Matteo Ciccotti for fruitful discussions and Matteo Ciccotti for sharing data with us.

References

1. J. L. Gardon, J. Appl. Polym. Sci., 1963, 7, 625–641.
2. A. N. Gent and R. P. Petrich, Proc. R. Soc. London, Ser. A, 1969, 310, 433–448.
3. D. W. Aubrey, G. N. Welding and T. Wong, J. Appl. Polym. Sci., 1969, 13, 2193–2207.
4. D. W. Aubrey and M. Sherriff, J. Polym. Sci., 1980, 18, 2597–2606.
5. M. Barquins, B. Khandani and D. Maugis, C. R. Acad. Sci., Ser. II: Mec., Phys., Chim., Sci. Terre Univers, 1986, 303, 1517–1519.
6. D. Maugis and M. Barquins, in Adhesion 12, ed. K. W. Allen, Elsevier ASP, London, 1988, pp. 205–222.
7. C. Derail, A. Allal, G. Marin and P. Tordjeman, J. Adhes., 1997, 61, 123–157 , 1998, 68, 203–228.
8. M. Barquins and M. Ciccotti, Int. J. Adhes. Adhes., 1997, 17, 65–68.
9. D. Maugis, J. Mater. Sci., 1985, 20, 3041–3073.
10. P.-G. de Gennes, Langmuir, 1996, 12, 4497–4500.
11. A. N. Gent, Langmuir, 1996, 12, 4492–4496.
12. G. Carbone and B. N. J. Persson, Phys. Rev. Lett., 2005, 95, 114301.
13. E. Barthel and C. Fretigny, J. Phys. D: Appl. Phys., 2009, 42, 19.
14. H. Tabuteau, S. Mora, M. Ciccotti, C.-Y. Hui and C. Ligoure, Soft Matter, 2011, 7, 9474–9483.
15. J. L. Racich and J. A. Koutsky, J. Appl. Polym. Sci., 1975, 19, 1479–1482.
16. G. Ryschenkow and H. Arribart, J. Adhes., 1996, 58, 143–161.
17. M. Gandur, M. Kleinke and F. Galembeck, J. Adhes. Sci. Technol., 1997, 11, 11–28.
18. M. Ciccotti, B. Giorgini, D. Vallet and M. Barquins, Int. J. Adhes. Adhes., 2004, 24, 143–151.
19. P.-P. Cortet, M. Ciccotti and L. Vanel, J. Stat. Mech.: Theory Exp., 2007, P03005.
20. S. T. Thoroddsen, H. D. Nguyen, K. Takehara and T. G. Etoth, Phys. Rev. E: Stat., Nonlinear, Soft Matter Phys., 2010, 82, 046107.
21. P.-P. Cortet, M.-J. Dalbe, C. Guerra, C. Cohen, M. Ciccotti, S. Santucci and L. Vanel, Phys. Rev. E: Stat., Nonlinear, Soft Matter Phys., 2013, 87, 022601.
22. R. S. Rivlin, Paint Technol., 1944, 9, 215–216.
23. K. Kendall, J. Phys. D: Appl. Phys., 1975, 8, 1449–1452.
24. L. B. Freund, Dynamic fracture mechanics, Cambridge University Press, London, 1998.
25. Y. Yamazaki and A. Toda, Physica D, 2006, 214, 120–131.
26. R. De and G. Ananthakrishna, Eur. Phys. J. B, 2008, 61, 475–483.

---