4.1. Formalism of the Model
In this section, we describe the simulation of the indentation at an angle in an adhesive contact that is developed from experimental data provided in the previous sections of the article. Simulation setup is based on the method of dimensionality reduction (MDR [
31]), the schematics of which for normal adhesive contact are presented in
Figure 7.
Within MDR for an axially symmetrical indenter with three-dimensional profile
f(
r) an equivalent one-dimensional profile
g(
x) must be found according to Abel transform:
In our experiment, we use a spherical indenter, which, at small indentation depths
d, can be replaced by a paraboloid
f (
r) =
r2/(2
R) with sufficient accuracy, thus, according to the procedure (3) we obtain:
Then, the elastic half-space is replaced by an array of non-interacting springs (see
Figure 7b), each of them has both normal Δ
kz and tangential Δ
kx stiffness:
where Δ
x determines the discretization of the space (the numerical solution does not depend on Δ
x). In the considered case of contact of a rigid indenter with a soft elastic material, effective elastic
E* and shear
G* modules are defined by the expressions
where
E and
ν—elastic modulus and Poisson ratio of the material being indented. When profile
g(
x) is immersed into an array of non-interacting springs to a depth
d, compression of an individual spring with coordinate
x in the normal direction is
If there is an adhesion in the system, the outer springs are “pulled” to the indenter. As it can be seen from the
Figure 7b, these “adhesive” springs reduce the value of the normal force and increase the contact radius
a, which is calculated according to the Hess rule [
31]:
where Δ
lmax is a magnitude of the enlargement of border springs, and Δ
γ is an adhesion specific work. As it follows from (7) and (8) with account of Δ
lmax ≡
u(
a)
Normal contact force
FN is defined as a sum of the forces from individual springs in contact (both stretched and compressed):
The results (9), (10) are exactly the same as the relations from JKR theory [
11], concerning the normal contact force
FN, indentation depth
d and contact radius
a in the case of adhesive contact. A distinct feature of the experiment, schematically shown in
Figure 2, is the indentation at an angle
α to the elastomer surface where a tangential shift of the indenter is also present together with normal motion. Therefore, after the contact, the springs shown in the
Figure 7b undergo a tangential shift. The experiment shows that complex processes of contact boundary restructuring, propagation of elastic waves, etc., occur during the tangential shift. Moreover, the contact becomes asymmetrical (see
Videos S2–S4). As we have already noted above, various descriptions of these processes exist in the literature; however, there is still no complete understanding of what happens in adhesive contact during tangential shift. Here, we propose a simple model that neglects the effects of elastic wave propagation and the jump-like reconstruction of the contact boundary, but at the same time allows us to describe the evolution of such quantities as both components of the contact force and the contact area during the process of indentation at an angle precisely enough.
As it follows from numerous experiments [
37,
38], during tangential shear in the stationary sliding mode, shear stresses preserve their values at almost constant level
τ0 =
Fx/
A and do not depend on indentation depth
d. The model will be developed from this assumption. The following assumptions should be considered not as a rigorous MDR model but as an estimation of the sliding conditions. Let us consider the stationary sliding mode, with corresponding experimental data shown in
Figure 4. Estimation for the average tangential stress can be made using the parameters of the MDR model as
where tangential shift of all springs
ux(
xi) in contact is accumulated in sum. Here,
a is a radius of a contact, which within a discrete model is defined as
a = (
n/2)Δ
x, where
n is the total number of springs in contact, and Δ
x is the space discretization step defined above, which can also be interpreted as a width of the spring or the distance between them. In the case of stationary sliding, when all springs in contact have the same shift
ux(
x) ≡
, with respect to (5) and (6), we have
Expression (12) sets the maximum tension of the springs during tangential shear in the form
When exceeding this value, springs begin to slide. At this point, we would like to stress again that this is a condition of an “adhesive-like” sliding criterion, which is not valid for the considered system, but will provide a correct estimation of the critical tangential stress. It is worth noting that depends on the contact radius a. Hence, with an increase in the indentation depth, the maximum tangential tension of the springs also increases.
Let us consider the simulation procedure using an example according to scenario (A) that is shown in
Figure 2. At a given angle
α the indenter moves in the normal and tangential directions with velocities
vz =
vsin
α and
vx =
vcos
α, which specify its normal and tangential displacements as functions of time
t:
During the indenter motion, profile g(x) is immersed at depth d. At the same time, at every iteration, according to Hess rule (8) the springs that are in contact are determined, which defines the contact radius a = (n/2)Δx. According to (7), the strain of the individual spring in normal direction can be found. Shifts of the springs in tangential direction ux(xi) are the same as the shift of the indenter after the moments of the contact of these springs with an indenter, while ux(xi) < (a).
For those springs where tension
ux(
xi) exceeds critical value
(
a), the tension magnitude is set to be equal
ux(
xi) =
(
a). Normal
FN and tangential
Fx forces are calculated as a sum of all forces in contact:
After the indenter reaches the maximum indentation depth
dmax, it is pulled out of the elastomer in the vertical direction with the velocity components
vz = −
v,
vx = 0. In this case, all the above described actions are performed. Another feature of the simulation is that when new springs come into contact (in the indentation phase), the value of the specific work of adhesion Δ
γ0 is assigned to them, while during the detachment (pull-off phase), a different value Δ
γ1 is used. When the condition Δ
γ1 > Δ
γ0 is true, the developed model reproduces experimentally observed secondary adhesive hysteresis [
34,
39].
Since in the pull-off phase the value Δ
γ1 is applied only to those springs that have already come into contact, when changing the direction of movement of the indenter, the size of a contact area
A is preserved for some time, which is clearly observable in experiments on normal indentation (see
Figure 3).
The model reported in this subsection, describes the contact between a rigid indenter and an elastic half-space, where the tangential and normal contacts are considered independently of each other. In fact, MDR makes it possible to take into account the thickness of the elastomer being indented [
47], as well as the relation between normal and tangential contact, as it was carried out, for example, in [
48]. A simpler model was deliberately chosen by us in order to show the main patterns of indentation at an angle, which are described in the next subsection of the work. Description of more subtle effects, such as the effect of tangential shear on contact strength in the normal direction, propagation of elastic waves, friction at the contact boundary, loss of contact area symmetry, etc., requires the construction of more complex models and is not the aim of this work. Some of the abovementioned effects were described in the framework of a dynamic model, proposed by us in [
38].
4.2. Results of the Simulation
Figure 8 shows the dependencies of the main parameters, obtained from the simulation of indentation according to scenario (A), the schematics of which are shown in
Figure 2. The dependencies, shown in
Figure 8, relate to the conditions of a real-life experiment with the experimental data presented in
Figure 5.
Figure 8 does not show the case with pure normal indentation at
α = 90°. Additionally, the panel (f) that shows relation
Lvertical/
Lhorisontal in experimental
Figure 5 is also absent in the
Figure 8 due to the fact that in MDR simulations the contact area is considered as circular and ratio
Lvertical/
Lhorisontal is always equal to one.
As it follows from the
Figure 8, the simulation results reproduce performed experiment qualitatively correct (see
Figure 5). The most important differences between the experiment and simulation, as well as possible explanations for such differences and suggestions for further improvement of the model, are described below.
(1) As
Figure 5a shows, the adhesive strength of the contact in the normal direction decreases with decreasing of the angle value
α. Here, by adhesive strength we mean the absolute value of the normal force |
FN,min| during pull-off phase at
FN < 0 N, i.e., the force caused by adhesion, because the same external force must be applied to completely destroy the contact. The maximum adhesive strength is observed in a purely normal contact (see dependence
FN(
d) in
Figure 3a for pull-off phase). At the presence of tangential shift, the strength of the contact in the normal direction decreases. In the simulation, we considered tangential and normal contact independently, therefore, for all angles
α magnitudes of |
FN,min| are the same (see
Figure 8a). Moreover, independent consideration of normal and tangential contact leads to the situation where dependencies
A(
t) in
Figure 8c at the beginning of the pull-off phase exhibit intervals with the constant size of the contact area
A. Tangential force
Fx also does not change within these time intervals of constant contact area
A, due to the absence of the tangential movement during the indenter pull-off in the pure normal direction. This constrain can be avoided by introducing the coupling between normal and tangential contact as it was carried out in [
48] for instance. However, application of the criterium, proposed in [
48], in our simulation did not lead to the expected results and therefore additional studies on this matter are needed.
(2) Maximum values of the normal
FN and tangential
Fx forces, as well as the size of a contact area
A and average contact pressure <
p> obtained in experiments exceed those from simulations. The reason for this is that in the experiment, indentation was performed in an elastomer layer with a limited thickness
h = 5 mm, while simulation was carried out for a half-space, for which
h → ∞. In the case of an elastic layer, the stiffness of the contact is significantly higher (especially for elastomers) [
32], which leads to increased values compared to the half-space case. To take into account the limited thickness of the layer, it is possible to use the generalized MDR proposed in a recent work [
47], which, however, contains a description of the modeling procedure only for normal contact.
(3) In the pull-off phase of the detachment of the indenter from the elastomer in the normal direction the tangential stresses in all cases increase to the maximum stationary value
τ0 = 42 kPa (see
Figure 8e) in the simulation, while in the experiment stresses <
τ> in the pull-off phase are characterized by a rapid growth (see
Figure 5e). In simulations, the limit for
τ0 values caused by the use of the springs sliding criteria (12), (13)—during the pull-off phase contact radius
a decreases, therefore, according to (13), there comes a moment when all the springs in contact begin to slide due to a decrease in the value of their critical tension
, and thus providing constant tangential stresses
τ0 over all contact area. The rapid growth of stresses
τ0, observed in experiments during pull-off, may be related to contact strengthening in time, and also to the viscoelasticity of the elastomer. Strengthening of the contact leads to the fact that the value
τ0 increases as well as adhesive strength in normal direction |
FN,min|. At the same time, viscoelasticity leads to a decrease in the velocity of rubber slip over the indenter, and, as a result, to the growth of
τ0 during the indenter motion. In the used estimation based on MDR, it is possible to take into account both contact strengthening (by increasing the value of the specific work of adhesion Δ
γ with time) and viscoelasticity (by using Kelvin–Voigt elements instead of springs shown in
Figure 7b). However, before such a modification of the model, it is necessary to find out the true causes of the described behavior first, which requires additional experiments that are beyond the scope of the proposed work.
(4) The experiment shows the moments of abrupt increase in the contact area
A when new regions of rubber are attached to the indenter, and the rearrangement of the contact boundary develops in different ways on the front and back sides of the contact. It is clearly observable in the experimental dependencies
A(
t), <
p>(
t) and
τ(
t) presented in the
Figure 5 for an angle
α = 10° (see also
Supplementary Video S3), at conditions close to tangential shear, as well as during pure tangential contact (see
Figure 4 and
Supplementary Video S2). In the simulation, the contact area grows smoothly, since its increase occurs only due to indentation in the normal direction (adhesive JKR contact). This simplification of the model is related to the paragraph (2) above, when the normal and tangential contact are considered independently.
Note that mentioned simplifications of the model lead to visible differences between experiment and simulation only when the angle at which the indentation is performed deviates significantly from the value α = 90°, that is corresponding to normal contact. In the case of indentation at angles close to 90° (for example, 80°, 70° and 60°), the experiment and theory give almost qualitatively identical results, since at such angles, in the indentation phase newly attached rubber regions move together with the indenter without slipping.
Taking into account the spring slip criterion (13), and assuming that indenter trajectory is known (14), the conditions when slippage of the springs begins can be easily determined. Now, we consider indentation according to scenario (A) in the phase of indentation, when the specific work of adhesion Δ
γ0 is small, therefore, with sufficient accuracy, the contact can be considered as non-adhesive in the normal direction. At this, the contact radius is defined as [
49]
According to (13) and (14), sliding of springs in contact in the phase of immersion of the indenter at an angle
α occurs when the general displacement of the indenter
x =
vtcos
α becomes equal to critical displacement
(
a) (13). If the contact radius is defined according to (16), it can be determined that slippage will start if the indentation depth
d exceeds a critical value:
which depends on the indentation angle
α. According to (17), at indentation angles
α < 90° slipping will occur anyways (with the growth of
α slipping appears at larger indentation depth
d). Dependence
dcrit(
α) is shown in
Figure 9a in solid line.
Figure 9a represents a diagram with two regions, the “sliding-ind” (here ”ind” denotes indentation) region, characterized by sliding of the springs, and the “no-sliding-ind” region, where during the entire indentation phase, new springs move in the tangential direction together with the indenter without slipping after attachment. The solid curve in the figure shows the dependence
dcrit(
α), defined by the expression (17). According to this dependence, at maximal indentation depth
dmax = 0.3 mm sliding in the indentation phase takes place if
α <
αcHertz ≈ 23.59°. However, main disadvantage of estimation (17) is the definition of the contact radius
a (16) obtained for non-adhesive contact. At small magnitudes of the specific work of adhesion Δ
γ, such an approach is precise enough. However, if Δ
γ is not necessarily small, for example in the pull-off phase, adhesion plays a crucial role. When the simple definition (16) is not adopted, dependence
dcrit(
α) is defined by the solution of the system of equations:
where first equation follows from (13) and (14) at
x =
(
a), while second equation is a relation (9), needed for contact radius
a determination. Obtained from (18) dependence
dcrit(
α), calculated at Δ
γ = 0.01 J/m
2 for the indentation phase (see parameters in the caption of
Figure 8), is shown in
Figure 9a in dashed line and located slightly above the “non-adhesive” dependence (17). Thus, taking into account adhesion widens the diagram region related to the absence of sliding (“no-sliding-ind”), at the same time at the maximum indentation depth
dmax = 0.3 mm slips occur at smaller, comparing to non-adhesive case, angle
αcJKR ≈ 22.87°.
Dependencies shown in
Figure 8, automatically relate to the above analysis of the slippage criterion, since they are obtained from the simulations. However, experimental data presented in
Figure 5 also show the absence of the slipping mode at large angles
α, that confirms Equations (17) and (18). Note that the above-mentioned complex dynamic processes take place within the slip mode, associated with a jump-like increase in the contact area, the propagation of elastic waves, etc. Qualitatively, these processes were described in our recent paper [
38], where the influence of the indentation depth
dmax on the tangential contact in the presence of adhesion was studied. In [
38], an experiment was also conducted and a numerical model was proposed.
Figure 10 shows the results of the simulation according to scenario (B), shown in
Figure 2. Thus, results presented in
Figure 10, should be compared to the experimental data shown in
Figure 6. As it follows from the comparison, the simulation results and experimental data differ similarly to the scenario (A), therefore we will not discuss them in detail again. Let us discuss, however, the sliding criterion, similar to the above described expressions (17) and (18), that were obtained for scenario (A). In scenario (B) after the indentation to the maximal depth
dmax = 0.3 mm, the indenter is shifted along the trajectory that is defined by equations:
where
is time, measured from the moment when the indenter stopped to immerse, i.e., from the beginning of the tangential shift. Expressions (13), (16) and (19) together lead to the quadratic equation for critical value of the indentation depth
dcrit, at which springs start to slip:
with two positive roots.
The smaller root
dcrit ≤
dmax defines the slip condition when pulling-off the indenter according to the scenario (B). In
Figure 9b, this root relates to the lowest solid line
dcrit(
α), that divides the diagram region
d <
dmax = 0.3 mm into two parts: “sliding-det” with sliding (here “det” denotes detachment) and “no-sliding-det” without sliding. The diagram region at
d <
dmax relates to the performed experiment (see
Figure 6) and simulation (see
Figure 10). As it follows from
Figure 9b, during the indenter pull-off from the initial indentation depth
dmax at an angle
α, if
α < 90° springs begin to slip in a certain moment of time.
Moreover, the smaller the angle, the earlier the slip begins, since the depth of indentation
d in the experiment decreases, starting from the value of
dmax (see the first equation in (19)). Note that with a purely tangential movement (
α = 0°), slippage on the diagram occurs already at
d =
dmax, since the indentation depth
d in the experiment is equal to the maximum value
dmax and does not change (case with the absence of normal movement of the indenter, is shown in
Figure 4). When the indenter is pulled-off in the normal direction (
α = 90°), there is no slippage (case with the absence of tangential movement, is shown in
Figure 3). The above conclusions are confirmed by the experiment, in which a region
τ0 = const is observed within whole range of angles
α (see
Figure 6e). The length of this region, however, significantly decreases with the growth of an angle
α. Similar behavior is also observed in simulations (see
Figure 10e).
Dependence
dcrit(
α), which is located above the value
d =
dmax = 0.3 mm in the
Figure 9b, relates to the situation when the indenter is not withdrawing from the elastomer after reaching the critical indentation depth, but continues to immerse into elastomer, moving at an angle
α instead of indentation in the normal direction. This curve also divides the area of the diagram at
d >
dmax = 0.3 mm into two parts with (”sliding-ind”) and without (“no-sliding-ind”) sliding. Plotted together by solid lines in the
Figure 9b, both dependencies
dcrit(
α) define a complete diagram of indentation modes, for the cases of indentation and pull-off of the indenter at an angle
α from the initial value
d =
dmax = 0.3 mm.
The refined dependence
dcrit(
α) with taking into account the adhesion in the normal direction is given by the solution of the system of equations (compare with (18))
where in first equation “–” sign relates to the scenario (B), when the trajectory of the indenter is defined by Equation (19). The choice of the “+” sign in this equation describes a situation in which the indenter after reaching the maximum indentation depth
dmax during the indentation in the normal direction continues immersing into the elastomer at an angle
α. Dependencies, defined by Equation (21), are shown in the
Figure 9b by dashed lines. The upper dashed curve was obtained at a smaller value of the specific work of adhesion Δ
γ = Δ
γ0 = 0.01 J/m
2, because it relates to the indentation phase. Therefore, the difference between the “adhesion-free” curve (solid line) and the curve with adhesion in the normal direction (dashed line) is not significant here. However, for the pull-off phase (curves below
dmax value), the difference is more significant, since the pull-off phase is characterized by a significantly larger value Δ
γ = Δ
γ1 = 0.15 J/m
2. Moreover, the dashed curve, obtained by taking into account the adhesion in the normal direction, in a certain range of angles
α located below the
d = 0 mm axis, since due to adhesion contact it also exists at
d < 0 mm.
Besides the experiments described above performed with an indenter with a radius of
R = 30 mm, we conducted a similar series of experiments with an indenter of a larger radius
R = 100 mm. In order to not overload the article, we do not describe these experiments here, but they are available as video files in the
Supplementary Materials (Videos S5–S8), which show the evolution of the contact area and main parameters. The difference between the corresponding experiments with indenters of different radii is only the value of the radius
R, all other conditions of the experiment were the same. In addition to the video files, the
Supplementary Materials contain a file named “
Figures S1–S4” with the dependences of the main parameters of the system obtained for the indenter with the radius
R = 100 mm. Presented dependencies are similar to the data described above for the case
R = 30 mm.