You are on page 1of 17

Eur. J. Mech.

A/Solids 18 (1999) 499–515


 Elsevier, Paris

Material compressibility effects for the squeeze of very thin films

François Auslender, Mohammed Trifa, François Sidoroff 1


Laboratoire de Tribologie et Dynamique des Systèmes, UMR 5513, Ecole Centrale de Lyon, 69131 Ecully, France
(Received 26 May 1997; revised and accepted 23 February 1999)

Abstract – The purpose of the present work is the study and clarification of the essential role played by the material compressibility of very thin solid
layers such as for instance appear in surface force experiments. For the sake of simplicity, attention will be focused on an elastic film squeezed between
rigid surfaces. The starting point is the classical lubrication theory reformulated within the framework of linear elasticity (Incompressible Reynolds
Model). This model is then extended to include compressibility effects. This extension is based on a phenomenological analysis of the stress and strain
fields obtained in some simple model cases and which are shown to correspond to an oedometric squeeze inside the contact with boundary corrections.
This results in the oedometric Reynolds model which is then applied to two special cases: a uniform film of constant thickness under plane strain
condition and the interfacial film occurring in the sphere/plane contact. These two problems confirm the fact that compressibility cannot be neglected
for very thin films and show that the governing adimensional parameters combine compressibility and thinness.  Elsevier, Paris
thin films / squeeze / nanorheology / compressibility

1. Introduction

The mechanics of very thin layers is an important field in many scientific or engineering problems. The
starting point of the work which will be presented here is essentially related to the nanorheological analysis
of confined liquid layers as it appears in surface force apparatus (Israelachvili, 1992; Tonck et al., 1988).
Experimental evidence as well as physical understanding clearly show a solid like structuration of the liquid
near the walls requiring, for a correct mechanical interpretation of the experimental results, refinement of
the usual models (Chan and Horn, 1985; Monfort and Hadziioannou, 1988) in order to take into account the
heterogeneous structure of the interfacial film as well as a significant understanding of the influence of the
involved material parameters. Such an analysis has been presented in (Sidoroff and Auslender, 1996; Auslender,
1996) which was essentially based on an extension of Reynolds equation to an heterogeneous viscoelastic
film under harmonic solicitation. When compared with experiments, this analysis provided a qualitatively
correct picture but required for the confined solid like layer an elastic stiffness much smaller (several orders of
magnitude in fact) than the values which could reasonably be expected.
The explanation advanced for this discrepancy was the fact that the proposed analysis assumed incompress-
ibility. It remains however (i) to develop a proper extension of the mechanical model to the compressible case
and (ii) to understand how material compressibility can play such an important role for this kind of problem
whereas it can reasonably be neglected for so many other problems.
These are in fact the essential issues to be dealt with in the present paper. For the sake of clarity, we will focus
on a very simple problem: the squeeze of an homogeneous or heterogeneous thin linear elastic film between
two rigid bodies. Extension to the practical situation investigated in (Tonck et al., 1988; Monfort et al., 1991;

1 E-mail: francois.sidoroff@ec-lyon.fr.
500 F. Auslender et al.

Georges et al., 1993) (viscoelastic harmonic response) is however straightforward through the use of complex
harmonic variables and responses. More generally, it is believed that the physical insight to be gained can be
used for many other problems, in particular connected with interfaces in composite materials and structures.

2. General background

2.1. Formulation of the squeeze problem

Let us consider a thin elastic film confined between two rigid bodies respectively defined in the natural
reference configuration by z = x3 = 0 and z = h(xα ) (Greek indices α, β take the values 1, 2 while Latin
indices i, j take the values 1, 2, 3 with the usual summation convention). Denoting by u the displacement
field, the boundary conditions for the squeeze problem are
uα = u3 = 0 for z = 0,
(1)
uα = 0, u3 = δD for z = h(xα )
to be completed with free boundary conditions at infinity or on the outer lateral surface. Here δD denotes the
prescribed vertical displacement.
In particular, we shall be interested in the global squeeze stiffness k
ZZ
F = kδD, F= σ33 dS. (2)
z=0

Special attention will be focused on the plane/plane case (figure 1b)

h(xα ) = h = cte (3)

and on the sphere/plane case (figure 1c)


r2
h(r) = D + , (4)
2R
where D is the sphere plane separation and R the sphere radius.

2.2. Elastic formulation of Reynolds equation

Lubrication theory and Reynolds equation (Cameron, 1967) are usually developed in fluid mechanics from
Stokes equations. From a formal point of view these equations are identical with Navier’s equations in
linear elasticity with the correspondence velocity → displacement and strain rate → strain. The Reynolds
approximation can be developed just as well in elasticity and it appears as the appropriate model for the analysis
of very thin elastic or viscoelastic layer.
The investigated elastic problem for an incompressible layer is formulated as follows
divσ = 0, (5)
εii = 0, (6)
σ = −p1 + 2Gε (7)
from the equilibrium equations (5) and the incompressible Hooke’s law (6), (7).
Material compressibility effects for the squeeze of very thin films 501

Figure 1. The squeeze problem. 1.a General case; 1.b The plane/plane case; 1.c The sphere/plane case.

Integrating ∂uz /∂z as given by (6) with respect to z and using the boundary conditions (1), the following
“flow equation” is obtained
Z h(xα )
∂qα
−divq = − = δD with qα = uα dz. (8)
∂xα 0

Classical lubrication theory (Cameron, 1967) is essentially based on the following assumptions:
(a) the hydrostatic pressure p does not depend on z,
(b) the terms ∂uα /∂z are prevailing in the expression of the strain tensor.
Using these two assumptions, the equilibrium equations (5) can be integrated once with respect to z
 
∂uα 1 ∂p
= z + cte . (9)
∂z G ∂xα

Further integration of these relations combined with the boundary conditions (1) will provide uα and finally q
as
h3
q =− grad(p) (10)
12G
for an homogeneous material.
This equation can be interpreted as a ‘constitutive equation’ relating q to the pressure gradient grad(p), and
which combined with the flow equation (8) leads to the classical Reynolds equation
 
h3
div grad(p) = δD. (11)
12G
502 F. Auslender et al.

Application to the sphere/plane problem (4) gives


Z ∞
3GRδD 6π GR 2
p(r) = − , F = 2π p(r)r dr = − δD. (12)
h2 0 D

Strictly speaking the integral in (12) should be from 0 to r∞ , radius of the outer lateral surface, but it can be
shown that this correction is usually very small. These expressions correspond to the standard Chan and Horn
solution, classically used for the analysis of surface force results (Chan and Horn, 1985) either directly or in
a straightforward viscoelastic extension. Henceforth, this elastic formulation of the lubrication theory will be
referred to as the Incompressible Model.
Similarly application to the plane/plane problem (3) under plane strain condition directly results in

2EδD 2
p=− (a − x 2 ) (13)
h3

for a thin plane of length 2a in the x1 = x direction and of width L in the x2 direction with a  L.

2.3. The compressible case

For an isotropic compressible elastic material the constitutive equation (6), (7) has to be replaced with

σ = σkk /3 = Kεkk , (14)


sij = 2Geij , (15)

where σ is the hydrostatic stress, sij and eij the deviatoric stress and strain tensors, while K and G are the bulk
and shear elastic modulus which will be considered as constant in the homogeneous case but may also depend
on xα and z in case of heterogeneous film.
Integration of the nonvanishing volumetric strain εii then leads to a modified expression for the flow equation
Z h(xα )
εii dz = div q + δD 6= 0, (16)
0

where εii is now given by (14). Further treatment of this equation however requires a better knowledge of the
stress tensor structure. To reach this knowledge some simple problems will now be considered.

3. The plane/plane problem

Let us first consider the plane/plane problem already investigated in the incompressible case (13). A
symmetric formulation of the corresponding plane strain boundary value problem is

z = ±h/2: ux = 0, uz = ±δD/2,
(17)
x = ±a: σxx = σxz = 0.
Material compressibility effects for the squeeze of very thin films 503

3.1. Variational methods

The variational theorems in linear elasticity (Germain, 1986) give an estimation of the potential and
complementary energy of the exact solution (σij , ui ) through
 
H σbij 6 H (σij ) = K(ui ) 6 K uei . (18)

The potential energy K(uei ) of the kinematically admissible displacement field uei is
ZZZ ZZZ ZZ
 1
K uei = σeij eεij dv − fi uei dv − Tid uei dS, (19)
2 Sf

where fi and Tid respectively denote the body forces and given surface forces on Sf (here x = ±a).
Similarly the complementary energy H (σbij ) of the statically admissible stress field σbij is
ZZ ZZZ
 1
H σbij = σbij nj udi dS − σbij bεij dV , (20)
2
Su

where udi denotes the prescribed displacement on Su (here z = ±h/2).


For the plane/plane problem (17) the body forces fi and the applied surface forces Ti = (σxz , σxz ) on the
boundaries x = ±a vanish. The only non vanishing data is the displacement uz = ±δD/2 on the boundaries
z = ±h/2. Accordingly
ZZ
1 1
K(ui ) = H (σij ) = σzz δD dS = kδD 2
2 z=±h/2 2
so that the basic inequality (18) becomes

 1 1  1
H σbij = k̂δD 2 6 kδD 2 6 K uei = k̃δD 2 . (21)
2 2 2

Any kinematically [resp. statically] admissible field will provide an upper [resp. lower] bound k̃ [resp. k̂] of the
contact stiffness k introduced in (2).
The kinematically admissible displacement field
z
uex = uey = 0, ez = δD
u (22)
h

describes the “oedometric” (without lateral motion) squeeze of the film. The corresponding upper bound k̃ is
obtained from (19) as
2E(1 − ν) a 2La
k̃ = L = (λ + 2µ) = k∞ , (23)
(1 − 2ν)(1 + ν) h h
where λ, µ are the usual Lamé elastic constants.
This oedometric displacement field is not the solution of the problem because the corresponding uniform
oedometric stress state does not satisfy the boundary condition σxx = 0 for x = ±a. A statistically admissible
stress field will be obtained by using this oedometric stress state in the core on the plate, i.e. for −a1 6 x < a1
504 F. Auslender et al.

(a1 < a) and by correcting it near the edges x = ±a in order to satisfy equilibrium and the stress boundary
conditions (17). This can be achieved through the following Airy function
 
6q0 (a − x)3 (a − x)θh z2
φ(x, z) = − − for x ∈ [a1 , a], (24)
(θh)3 3 2 2
where θ = (a − a1 )/ h. The corresponding lower bound is after some computations obtained from (20) as
    

 ν2 

13
θ + 3
+ 2+3ν
h 6 16θ 3 2θ
k̂ = k∞ 1 − 4θ +  . (25)

 a 5(1 − ν)(1 + ν)(1 − 2ν) 

The evaluation of course depends on a1 , and the next step would be to choose θ to maximise this lower bound
and therefore to obtain the optimal size of the considered edge effect. This is however not necessary for the
present purpose which is the thin film limit h/a → 0. Indeed if θ is chosen of order of O(1), the combination
of these two estimations results in
 2E(1 − ν) La
k̂ = k∞ 1 − O(h/a) 6 k 6 k̃ = k∞ = . (26)
(1 − 2ν)(1 + ν) h
This value (23) of the contact stiffness k∞ is the thin film limit for any value ν < 1/2.

3.2. Discussion

The evaluation obtained in (26) clearly shows the validity of the oedometric approximation k∞ for very
thin films. However if this limit is obtained for any value ν < 1/2 of the Poisson ratio, it does not hold at the
incompressible limit ν → 1/2. This is for instance clear on (23) from the (1 − 2ν) term in the denominator.
These two limits, thin film and weak compressibility, do not commute. More precisely
2E(1 − ν) La
lim lim k = k∞ = ,
ν→1/2 h/a→0 (1 − 2ν)(1 + ν) h
8ELa 3
lim lim k = kinc = ,
h/a→0 ν→1/2 3h3
where the second limit has been obtained from Reynolds approximation. We have no proof for that but it is
probably true.
The above approximations can be improved through a better choice and optimisation of the kinematically
and statically admissible fields (Auslender et al., 1995; Armengaud et al., 1995). For instance the following
kinematically admissible field
  
δD x 4z2 z
ux = g 1− 2 , uy = 0, uz = δD (27)
2 a h h
combines the squeeze (22) with the horizontal displacement observed in Reynolds approximation. Here g is an
arbitrary function (with g(0) = 0) which will now be chosen in such a way to minimise the potential energy
K = W . After some computation
Z
W 1  1
= 1 + dg 02 (ξ ) + 2bg 0 (ξ ) + cg 2 (ξ ) dξ, W∞ = k∞ δD 2 , (28)
W∞ 0 2
Material compressibility effects for the squeeze of very thin films 505
2
with ξ = xa , d = 15a
8h
2 , b = 3 λ+2µ a , c = 3 λ+2µ , and where W∞ is the reference oedometric value.
2 λ h 4 µ

Minimisation of this integral with respect to the function g is a standard variational problem resulting in a
second order differential equation

cg(ξ ) − dg 00 (ξ ) = 0,
g(0) = 0, dg 0 (1) + b = 0.

The solution is readily found as


 q 
  s x a 5 (1−2ν)
sh
x ν 5 (1 − 2ν) a h 4 (1−ν)
G =−  q  . (29)
a 1 − 2ν 4 (1 − ν) a 5 (1−2ν)
ch h 4 (1−ν)

Finally, substituting this function in (28), an improved upper bound of the stiffness is obtained

  s
5 ν2 th(γ ) 5(1 − 2ν) a
k̃ = k∞ 1− , where γ= . (30)
6 (1 − ν) 2 γ (1 − ν) h

Again the singular incompressible limit (ν → 1/2) is encountered adimensional parameter 1 − 2νa/ h.
A similar and more complex analysis can also be performed to improve the lower bound (Armengaud et al.,
1995).
Asymptotic methods (Sanchez-Palencia and Sanchez-Hubert, 1992) can also be used for the thin film limit
ε = h/a → 0. However in this particular case, they do not provide any further information with respect to
the variational methods discussed above. Indeed the first order outer solution is found to coincide with the
oedometric solution. However, since this solution does not satisfy the boundary solution on x = ±a, it has to
be completed by an inner solution near these boundaries. The classical techniques used for extracting the inner
solution (Sanchez-Palencia and Sanchez-Hubert, 1992) unfortunately result in the complete Navier problem.
Asymptotic methods therefore only confirm the essential phenomenology of this problem: the oedometric
solution holds true inside the contact and has to be corrected near the edges to account for boundary effects.
The influence of these edge effects on the global response of the system however may be neglected for very
thin layer, as follows from the upper and lower bounds (26).

4. Axisymmetric problems

4.1. Axisymmetric plane/plane problem

The results obtained in Section 3 can be extended to other more realistic geometries. As a first example let us
consider the axisymmetric squeeze of a circular plate of constant thickness h and radius a. The corresponding
boundary conditions are, in cylindrical coordinates,
 
h δD
uz r, ± =± ez , σrz (a, z) = σrr (a, z) = 0. (31)
2 2
506 F. Auslender et al.

The kinematically admissible oedometric displacement field is

δD
uer (r, z) = 0 ez (r, z) =
u z (32)
h
and the associated upper bound for the stiffness is given by

E(1 − ν) π a2
k̃ = = k∞ . (33)
(1 + ν)(1 − 2ν) h

The corresponding uniform oedometric stress state is


ν νE δD
σrra = σθθ
a
= σzza = = q0 ,
1−ν (1 + ν)(1 − 2ν) h
(34)
a
σrz = σθz
a
= σθr
a
=0
which do not satisfy the free boundary conditions (31) for r = a. A statically admissible stress field can then
be constructed by connecting this oedometric state inside the contact (r < a1 = a − θh) to the boundary r = a.
The connecting stress field σ b , which must satisfy equilibrium, the boundary conditions (31) and continuity
with σ a on r = a1 , can be taken as

σrrb = σθθ
b
= g(r), σzzb = g 00 (r) + g 0 (r)/r z2 /2, b
σrz = −g 0 (r)z, (35)

where g(r) = q20 1 − cos π(a−r)
θh
and r ∈ [a1 , a].
The associated stiffness is then
π(a − θh)2 E(1 − ν) 2
k̂ = + H (σ b ),
h (1 − 2ν)(1 + ν) δD 2
    
H σijb π θh θ 2 h2 3 h2 q02 π 4 h2 q02 π 2 (1 + 2ν)
= − 2 − 2 (4 − ν)h2 q02 + +
a2/ h E a a 4 403200θ 4 212θ 2
Z   )
h2 θ 2 h2 q 2 h2 π 2 q02 θh a πr
− 2 ν(1 − ν) 20 + 2 sin2 dr .
a π a 5120θ 2 0 a −r θh
Finally and after some computation, the correcting terms can be estimated, resulting in the following evaluation
for the stiffness k
 π a2 E(1 − ν)
k̂ = k∞ 1 − O(h/a) 6 k 6 k̃ = k∞ = (36)
h (1 − 2ν)(1 + ν)
which is an obvious extension of the result obtained in Section 3. Again the oedometric solution appears as the
appropriate approximation for thin layers and the incompressible limit is singular.

4.2. The sphere/plane problem

The sphere-plane problem is obtained from (1) with h(r) given by (4). Denoting by r∞ the lateral dimension
√ thin film limit is obtained for D/r∞ → 0. Further in order to have h = O(D), we will also assume
of film, the
r∞ = O( RD) which is the relevant characteristic lateral length for sphere-plane contact.
Material compressibility effects for the squeeze of very thin films 507

Using standard asymptotic techniques the following reduced variables are introduced
s
r z ui D
ρ=√ , t= , vi (ρ, t) = , ε= . (37)
RD D δD R

With these variables, Navier equations are obtained as


 
vr,ρ vr 
ε (λ + 2µ) vr,ρρ +
2
− 2 + ε µvz,ρt + λvr,ρt + µvr,t t = 0,
ρ ρ
    (38)
vz,ρ vr,t
ε µ vz,ρρ +
2
+ ε(λ + µ) vr,ρt + + (λ + 2µ)vz,t t = 0
ρ ρ
with boundary conditions
t = 0: vr = vz = 0,
(39)
ρ2
t =1+ = t1 (ρ): vr = 0, vz = 1,
2
 
r∞ vr
ρ=√ = ρ1 : ε (λ + 2µ)vr,ρ + λ + λvz,t = 0, εvz,ρ + vr,t = 0, on SL . (40)
RD ρ
Substituting the outer asymptotic expansion

v(ρ, t) = v 0 (ρ, t) + εv 1 (ρ, t) + ε 2 v 2 (ρ, t) + · · · (41)

in these equations directly gives the following problem at order 0


(
t = 0,
0
µvr,t
(42)
(λ + 2µ)vz,t
0
t = 0,

t = 0: vr0 = vz0 = 0,
(43)
t = t1 (ρ): vr0 = 0, vz0 = 1,
ρ = ρ1 : 0
vr,t = vz,t
0
= 0. (44)
Equation (42) with the boundary conditions (43) directly gives the oedometric solution
t
vr0 = 0, vz0 = (45)
t1 (ρ)

while the lateral boundary conditions (44) cannot be satisfied and therefore requires an inner development near
the boundaries ρ = ρ1 . The corresponding boundary value problem cannot easily be solved but its contribution
will remain small because the corresponding oedometric stress rapidly decreases with increasing ρ.
The oedometric solution
δD
uer (r, z) = 0, uez (r, z) = z (46)
h(r)
therefore appears as a reasonable approximation.
508 F. Auslender et al.

5. Oedometric Reynolds model

5.1. Basic assumptions

From a qualitative point of view, the essential features of the displacement fields is clear: oedometric squeeze
corrected by small lateral motion accounting for edge effects. To capture these features the assumptions (a)
and (b) of the Reynolds model introduced in Section 2.2 are modified in
(a1 ) The vertical stress σ = σzz does not depend on z,
(a2 ) The horizontal strain εαβ are negligible with respect to εzz ,
(b0 ) The terms ∂uα /∂z are prevailing in εαz .
Assumptions (a1 ) and (a2 ) correspond to the oedometric squeeze while (b0 ) is a mere reformulation of
assumption (b).
It should be noted that these assumptions will result in a constant σzz and εαβ within the contact, which is
consistent with static equilibrium and compatibility even in the case of an heterogeneous layered film.
Assumption (a2 ) with Hooke’s law (14) results in
   
4G 2G
σzz = σ = K + εzz , σxx = K − εzz . (47)
3 3

Assumption (a2 ) and the oedometric relations (47) now allow evaluation of the left hand side of the
compressible flow equation (16)
Z h Z h Z h σ dz
εii dz = εzz dz = .
0 0 0 K + 4G/3

Using assumption (a1 ) the modified flow equation is finally obtained as


Z h
σh 1 1 dz
− div q = δD, = . (48)
K oed K oed h 0 K + (4/3)G

Heterogeneity of the film can be taken into account by assuming K and G to depend on xα and z resulting
in K oed (xα ).

5.2. Modified flow equation

Combining assumption (b0 ) and the oedometric relation (47), the equilibrium equations require
 
∂ ∂uα ∂σxx ∂σ
G =− = −β ,
∂z ∂z ∂xα ∂xα
(49)
K(xα , z) − 23 G(xα , z)
β(xα , z) = .
K(xα , z) + 43 G(xα , z)
Integrating this equation twice with respect to z
Z Z z0 Z
∂σ z 1 00 00 0
z dz0
uα = − β(z ) dz dz + τ0 + u0 ,
∂xα 0 G(z0 ) 0 0 G(z0 )
Material compressibility effects for the squeeze of very thin films 509

where τ0 (xα ) and u0 (xα ) are two integration constants which are obtained from the boundary conditions (1). A
third integration with respect to z finally provides, after some computations

q = Roed grad(σzz ), (50)

where
(Z Z  Z Z Z  Z ) Z
h z h h dz h 1 h h z dz h dz
0 0 0 0
Roed = β(z ) dz dz − β(z ) dz dz .
0 G(z) 0 0 G(z) 0 G(z) 0 0 G(z) 0 G(z)

Combination of the two Eqs (48) and (50) gives the partial differential equation

σzz h 
− div Roed grad(σzz ) = δD. (51)
K oed
This equation defines the oedometric Reynolds model.
A much simpler form is obtained in the weakly compressible case G/K  1, ν ∼ 1/2 for which β ∼ 1 so
that (51) becomes
σh
− div(R grad σ ) = δD, (52)
K
Z h   Z h
h dz g2 zk dz
= , R = g2 − 1 , gk = , k = 0, 1.
K 0 K g0 0 G(xα , z)
This case is important for practical applications.
Similarly for an homogeneous layer, Eq. (51) simplifies into
 
σh βh3
− div grad σ = δD. (53)
K + 4G/3 12G

6. Applications

6.1. The plane/plane problem

The plane/plane problem under plane strain condition, which has been analysed in Section 3 is now
easily solved. Equation (53) becomes an ordinary differential equation which, with the boundary conditions
σ (±a) = 0 is readily solved as
 q 
3 (1−2ν)
ch 2x (1 − ν)E δD δD
σzz = σ0 1 − 
h 2 n
q , where σ0 = = (λ + 2µ) , (54)
2a 3 (1−2ν) (1 − 2ν)(1 + ν) h h
h 2 ν

where G and K have been replaced by their expression in terms of E and ν. The corresponding stiffness is
found as
 q 
2(1 − ν)E a th 2a 3 1−2ν
L 1 − 
h 2 ν
k= q (55)
(1 − 2ν)(1 + ν) h 2a 3 1−2ν
h 2 ν
510 F. Auslender et al.

Figure 2. Plane/plane problem. Evolution of the reduced stiffness k/k∞ as a function of ν for h/a = 10, 30 and 100.

which obviously tends to the oedometric limit (23) for h/a → 0.


It can also be verified that for any fixed value x this solution tends to the incompressible solution (13) when
ν → 1/2. Again these two limits: incompressibility ν → 1/2 and thin film h/a → 0, do not commute. The
appropriate adimensional number for this problem is

α = a 1 − 2ν/ h (56)

which combines the film thinness with its compressibility. These results are exemplified in figures 2 and 3.
As expected, this solution corresponds to the oedometric stress field inside the film corrected by a boundary
layer near the edge
s
  h 2 ν
− (a−x) 0
σ ∼ σ0 1 − e l0 , where l = . (57)
2 3 (1 − 2ν)
This oedometric model is consistent with all the information gathered in Sections 3 and 4 about the solution
of some simple problems. It therefore appears reasonable. It remains however empirical and requires further
validation, in particular concerning the boundary layers which is the essential new feature it adds with respect
to the simple oedometric approximation.
In order to check this validity some finite elements simulations have been performed. These simulations have
been restricted to the boundary layers, i.e. on

 = [a1 = a − θh, a] × [−h/2, +h/2]


Material compressibility effects for the squeeze of very thin films 511

(a)

(b)
Figure 3. Plane/plane problem. Evolution of the stress distribution for different values of ν.

with the standard boundary conditions on z = ±h/2 and x = a while the oedometric solution is imposed on
x = a − θh. The value of θ has to be chosen large enough with respect with the boundary layer width, which
as shown by (57) increases as ν → 1/2. Practically θ has been chosen as 5 for ν = 0.3 and 10 for ν = 0.48.
The corresponding results for the σxx stress at different z values are represented in figure 4. (Remember that
512 F. Auslender et al.

(a)

(b)
Figure 4. Stresses σxx (x) resulting from a finite element calculation for three different values of z and from the oedometric.
Material compressibility effects for the squeeze of very thin films 513

Figure 5. Sphere/plane problem. Compressible correction f (α) as function of the dimensionless parameter α = GR/KD.

the boundary layer effect essentially results from this σxx component which must decrease from its oedometric
value inside the contact to zero.)
A correct agreement is obtained except in the vicinity of the corner (z = h/2, x = a) where a stress
singularity exists. The approximation also appears better for larger value of ν.
The axisymmetric plane/plane problem of Section 4.1 could be treated in the same way but no analytical
solution appears available.

6.2. The sphere/plane problem

The axisymmetric problem now has to be solved with h(r) given by (4). An adimensional formulation is
obtained by setting

√ 1−ν h ρ2 D
r= RDρ, δD = DU , h̄ = =1+ , σ zz (ρ) = σzz (r). (58)
ν D 2 RUG
The differential equation (53) is
 
1 d h̄3 dσ zz 1+ν G R
α h̄σ zz − ρ = 1, where α= (59)
ρ dρ 12 dρ 3ν K D

which has to be solved with the boundary conditions σ → 0, ρ → ρ∞ or ρ → ∞.


The structure of (59) implies that the solution σ (ρ) only depends on α. The normal force associated to the
normal stress σ = σzz is then given by
Z
6π GR 2 δD (1 − ν) 1 ρ∞
F= f (α), where f (α) = ρσ (ρ, α) dρ. (60)
D ν 3 0
514 F. Auslender et al.


Figure 6. Sphere/plane problem. Reduced stress distribution σ h2/GR as a function of ρ = r DR for α = 0, 0.8 and 2.0.

This expression for f has been normalised from the incompressible case (12) which corresponds to α = 0
with f (0) = 1. The relevant dimensionless parameter α clearly shows that compressibility cannot be neglected,
even for nearly incompressible materials, i.e. for high values of the ratio K/G. Indeed this ratio appears in α
as multiplied by the ratio D/R which, for thin layers, may be very small. The function f (α) is plotted in
figure 5. Typically a reduction by half of the force is obtained for α = 0.8, i.e. if for instance ν = 0.4999 for
D/R = 1.6 10−4 . It should be noted that in the mentioned experiments (Tonck et al., 1988; Monfort et al., 1991;
Georges et al., 1993) D/R typically is of order 10−6 .
In fact the limiting case of large α is probably more relevant. This corresponds to the oedometric
approximation (46).
The corresponding adimensional stress distribution has been represented in figure 6 for different values of α.

7. Conclusion

The influence of material compressibility has been investigated in the case of thin linear elastic films
squeezed between two rigid bodies. Variational and asymptotic methods have provided estimations of the exact
solution and understanding of the oedometric deformation mechanism. This essentially justifies the developed
Oedometric Model approximation which appears as a reasonable description of the squeeze of very thin films.
In this paper attention has been limited to the homogeneous elastic case but it should be noted that the developed
Compressible Model is valid for heterogeneous layer and can easily be extended to the viscoelastic harmonic
case by using complex quantities. This is now in progress for application to real tribological situations (Sidoroff
and Auslender, 1996; Auslender, 1996).
Another important result of this analysis is the fact that for the squeeze problem of weakly compressible
thin films the relevant adimensional parameter are not the compressibility ratio G/K or 1 − 2ν and the film
aspect ratio h/a or D/R but rather their combination (56) or (59). This clearly explains the singularity of the
incompressible limit in the mechanics of thin films. The combination of these two limits (thin film and small
Material compressibility effects for the squeeze of very thin films 515

compressibility) provides an interesting mathematical problem which, although encountered in other physical
situations (like for instance layered metal-polymer composites), has apparently received little attention.

References
Armengaud G., Auslender F., Ségura J.M., Sidoroff F., 1995. Méthodes énergétiques appliquées à l’étude du comportement des couches minces
élastiques. In: Actes du 12ème Congrès Français de Mécanique, Strasbourg, II, pp. 125–128.
Auslender F., 1996. Nanorhéologie des couches minces confinées. Thèse de l’Ecole Centrale de Lyon, no 96-42.
Auslender F., Armengaud G., Sidoroff F., Ségura J.M., 1995. Rôle de la compressibilité dans l’écrasement des couches minces. In: Actes du 12ème
Congrès Français de Mécanique, Strasbourg, I, pp. 221–224.
Cameron A., 1967. The Principle of Lubrication. Wiley, New York.
Chan D.Y., Horn R.G., 1985. Drainage of thin liquid films. J. Chem. Phys. 83, 5311–5324.
Georges J.M., Millot S., Loubet J.L., Tonck A., Drainage of thin liquid films between relatively smooth surfaces. J. Chem. Phys. 98, 7345–7360.
Germain P., 1986. Mécanique. Ellipses, Paris.
Israelachvili J.N., 1992. Intermolecular and Surfaces Forces. Academic Press, New York.
Monfort J.P., Hadziioannou G., 1988. Equilibrium and dynamic behaviour of thin films of a perfluorinated polyether. J. Chem. Phys. 88, 7187.
Monfort J.P., Tonck A., Loubet J.L., Georges J.M., 1991. Microrheology of high polymer solutions. J. Polymer Sci. 29, 677–682.
Sanchez-Palencia E., Sanchez-Hubert J., 1992. Introduction aux méthodes asymptotiques et à l’homogénéisation. Masson, Paris.
Sidoroff F., Auslender F., 1996. Nanorheological behaviour of confined liquid layers for normal contact, The third body concept. In: Dawson D.
et al. (Eds), Interpretation and Tribological Phenomena, Elsevier, Amsterdam, pp. 195–203.
Tonck A., Georges J.M., Loubet J.L., 1988. Measurements of intermolecular forces and the rheologie of Dodecane between Alumina Surfaces.
J. Colloid Interface Sci. 126 (1), 1540–1563.

You might also like