New confinement effects on the viscosity of suspensions
Philippe Peyla, Claude Verdier
To cite this version:
Philippe Peyla, Claude Verdier. New confinement effects on the viscosity of suspensions. EPL - Europhysics Letters, European Physical Society/EDP Sciences/Società Italiana di Fisica/IOP Publishing,
2011, 94, pp.44001. hal-00567009v3
HAL Id: hal-00567009
https://hal.archives-ouvertes.fr/hal-00567009v3
Submitted on 1 Apr 2011
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
epl draft
New confinement effects on the viscosity of suspensions
P. Peyla1 and C. Verdier1
1
CNRS - Université de Grenoble I, UMR5588, Laboratoire Interdisciplinaire de Physique (LIPhy)
38041 Grenoble, France
PACS
PACS
PACS
47.50.Ef – Measurements
47.57.E- – Suspensions
47.57.Qk – Rheological aspects
Abstract. - We present rheological experiments of confined suspensions at moderate concentrations. The analysis is carried out in the framework of a previous study on particle suspensions
[Davit and Peyla, EPL, 83, 64001 (2008)] where simulations revealed the presence of unusual
effects attributed to confinement, i.e. when the gap size (h) becomes closer to the particle size
(d). Deviations from the usual viscosity trends were found. The present work investigates these
features further and confirms the important role of the confinement. Extensions are made from
the classical approach to the case of confined suspensions where the importance of the reduced
gap hd is taken into account.
Introduction. – The viscosity of solid suspensions
and pastes is important for industrial applications such as
chemical, petrol engineering, food industry as well as in
the field of environment [1] where more complex fluids are
investigated. The rheology of suspensions is an evolving
field where new measurement techniques are of interest
[2,3] but where classical rheometrical tools are still proper
devices for carrying out accurate measurements.
Even when mastering the rheology of certain fluids, it
can become quite difficult to study particular complex
flows, such as the ones possessing free surfaces or those
generated by complex geometries. An important recent
topic is the one of confined flows [4–8] where interfaces
or the presence of walls can give rise to intriguing new
regimes. This study of confinement has been addressed
previously [7] and proposed a possible way to determine
viscosity or yield stress parameters using squeezing flows
(implying smaller and smaller gaps), assuming that steady
states were obtained. But this method raised the question of such steady states in particular since clusters of
particles in complex fluids can lead to reorganization of
substructures involving longer waiting times for a steady
state [9]. Another drawback of the latter method [7] was
that a known constitutive equation was needed to derive
proper results (in fact a Herschel–Bulkley model). Most
studies on confined suspensions deal usually with concentrated systems [7, 8, 10]. Another recent approach of confinement was based on the study of fluctuations and oscillations in concentrated colloidal suspensions leading to
shear–thickening [6]. Finally, concentrated emulsions can
also behave differently in confined flows [2] and exhibit a
characteristic cooperation length which is concentration–
dependent.
In a recent study, the authors revealed the presence of
anomalous effects due to the confinement of suspensions
at moderate concentrations [11]. These effects were attributed to the presence of walls that change the nature
of hydrodynamic interactions between the particles. The
steady shear viscosity η was shown to behave differently
from theories for unconfined suspensions. In general, previous authors [12–15] have proved that for moderate concentrations, the viscosity of a non-confined suspension of
hard spherical particles behaves as
∞ 2
η = η0 (1 + a∞
1 φ + a2 φ ),
(1)
where η0 is the suspending fluid viscosity, φ is the volume concentration, and the symbol “∞“ denotes the limit
for unbounded domain (i.e. not confined). Eq. (1) is a
Taylor expansion where the linear term (a∞
1 φ) holds for
very dilute suspension and a∞
1 = 2.5 has been calculated
2
by Einstein [12, 13]. The quadratic term (a∞
2 φ ) represents the contribution of the two-body hydrodynamic interaction. Batchelor and Green [14] gave the first good
estimation of a∞
2 = 5.2 ± 0.3 for homogeneous suspensions, this term being calculated more precisely later [15]
and found close to a∞
2 = 5. But the relationship (1) seems
to fail as suspensions are confined, so that the usual con-
p-1
P. Peyla et al.
vex evolution predicted by eq. (1) from Batchelor et al.
(a∞
2 > 0) changes to become concave. This means that
the coefficient a∞
2 becomes negative while confining the
suspension. This is due to wall effects which change completely the hydrodynamic interactions at the microscale.
Macrorheology can reveal this microscopic behavior by
measuring the effective viscosity as a function of φ for
confined suspensions. One may argue that continuum mechanics theories only hold when the volume to be studied
is much larger than the volume occupied by a typical particle in the sample (with diameter d), therefore that the
viscosity should not depend on such effects. Still it is important for practical applications (e.g. flow of confined
suspensions through holes or tubes, like in ball bearings)
to get an idea about the relationship between the so–called
effective viscosity (measured in a rheometer) and gap size.
In this study, an attempt is made to correlate experimental results with the numerical ones obtained prevously
[11] at moderate concentrations. Rheometrical studies are
carried out in order to obtain typical low–shear viscosity
data, as the gap (h) is reduced, for a large number of concentrations up to 20%. The results are analyzed along
these lines and these results are pushed further in order to
come up with practical relationships in terms of the concentration φ and the ratio hd , which measures confinement.
In the first part, rheometrical techniques are presented,
as well as care which has been taken in order to obtain such
data. Then a discussion proposes to analyze the results
in terms of reduced parameters. Simple relationships are
obtained and the main conclusions are summarized.
Rheometrical measurements. – Rheometrical
measurements were carried out on a stress–controlled
rheometer (Bohlin, Gemini 150, Malvern) at T = 23◦ C.
A plate–plate geometry was used in order to change the
gap easily. Suspensions were prepared by slowly mixing the fluids with particles. The fluids and particles
were selected because of their closely matched densities (ρp = 1.05 g/cm3 for the polystyrene particles and
ρf = 0.978 g/cm3 for the fluid). The fluid was a Rhodorsil silicone (PDMS 47V60000, Rhodia, viscosity roughly
around 60 P a.s) and was chosen for its Newtonian behavior at the shear rates tested (typically 10s−1 ). Its
relaxation time is λ = 1 ms and is small enough so that
non–Newtonian effects do not affect the viscosity as shown
previously [16, 17]. A viscous PDMS is also a good candidate because larger viscosities slow down sedimentation,
so that these effects can be neglected here. Spherical
polystyrene particles (Dynoseeds TS80, TS140 and TS230,
Danemark) were chosen for their densities, and also because of their monodispersity (see fig. 1). Their diameters
were 80 µm, 140 µm and 230 µm respectively, in order to
be able to set final gaps in the range of the particle diameters, i.e. [100 − 300 µm]. A small gap correction was
used to avoid tolerance errors, machine precision and concentricity errors, a procedure already used before [18–20].
This led to the accurate determination of the PDMS vis-
cosity η0 = 60.6±0.9P a.s, in agreement with our previous
work [17].
For measuring the suspensions viscosity, rough solid surfaces were available (roughness 5µm) to avoid slip. Note
that it is necessary to use a small roughness since confinement effects are to be studied. Some experiments were
repeated twice and showed similar trends as the gap was
varied. This is sufficient to assert the no–slip hypothesis.
Furthermore, previous works reporting slip in suspensions
showed that such effects occur at low shear rates smaller
than 0.1s−1 typically [1] whereas our materials are tested
at a higher shear rate of 10 s−1 . Finally, slip is usually
observed at higher concentrations, not below 20% as in
this study.
Suspensions were prepared and degased; then they were
set onto the rheometer. The upper plate was gently
brought down until contact. Experiments started with
large gaps (usually 1500 µm) to obtain the asymptotic
value of the viscosity, as postulated from continuum mechanics (the gap in the case of a 230 µm diameter particle
is then around seven times higher which is good enough
to obtain this limit, as will be shown). Then the gap was
reduced to smaller values. Excess fluid was removed after each gap change. Several fresh samples were used to
cover the gap range. Thus, sedimentation was evaluated
to be negligeable. A shear rate of 10 s−1 was used (see
below). Note that due to the presence of particles, small
gap experiments usually took longer, of the order of a
few minutes. This is due to the internal reorganization of
particles under shear that may occur due to the possible
formation of particle clusters that are broken or coalesce
until equilibrium is obtained [9].
Next a series of experiments were carried out at a constant shear rate of 10 s−1 , using different particle sizes and
particle concentrations (typically between 0% and 20%).
This shear rate was used for various complementary reasons. It is a shear rate allowing steady state to be obtained
rapidly (few minutes), it is also constant as was used previously in simulations [11], and finally this value ensures
that the suspensions remain in the Newtonian regime,
and are not shear–thinning. This constant value can be
achieved thanks to a feedback loop control available with
this rheometer. Note that, in order to match the simula2
tions, low Reynolds numbers Re = ρRη0 γ̇ should be used,
which is the case here since Re is of the order 10−6 in the
experiments.
The reduced viscosity ηη0 − 1 was investigated first. All
curves are shown on the same graph to start with and one
can notice an important dispersion. Indeed fig. 2 shows
the raw data at several concentrations (5−7.5−10−12.5−
15 − 17.5 − 20%) and for three typical particle diameters
(80 µm, 140 µm and 230 µm). Note that in some cases, a
few data points seem to be more dispersed at small h/d
or high concentrations, which does not seem to be satisfactory. These data points were shown to be attributed
to the difficulties inherent to the small gap measurements.
p-2
Confined suspensions
Table 1: Values of parameters m1 , m2 , m3 used as a model fit
for the data in fig. 3.
Concentration
2.5%
5%
7.5%
10%
12.5%
15%
17.5%
20%
m1
0.058
0.13
0.21
0.29
0.42
0.55
0.70
0.88
m2
0.58
0.73
1.0
1.29
1.40
1.60
1.50
1.55
m3
0.88
0.65
0.65
0.55
0.54
0.50
0.40
0.39
As mentioned before, for small gaps, plate parallelism is
important as well as tools concentricity effects or the fact
that some particles may be in contact with the wall or
touch one another. This leads to reduced accuracy as the
viscosity rises rapidly, especially for high concentrations.
These data points were kept in the next figures and only
concern a few measurements for each concentration, particularly when h < 1.5d.
It is then proposed to explain this data in terms of the
reduced gap h/d, as shown in fig. 3. Data points superpose
quite well together for all particle diameters. Note that the
limiting values of the viscosities are obtained for large h/d,
usually when h/d > 10 roughly. This whole set of data is
the main set of results which is used throughout the rest
of the paper.
A typical trend is obtained showing a plateau for the
reduced viscosity at large reduced gaps h/d, then as h/d
becomes smaller, the reduced viscosity ηη0 − 1 increases
similarly for all concentrations. It was found that a simple
exponential function of the type ηη0 −1 = m1 +m2 e−m3 h/d
fits the data quite well. This leads to the following set of
parameters in table. 1; in particular m1 represents the
value of ηη0 − 1 for large reduced gaps h/d.
Together with this limit for large h/d, other values of
η
η0 − 1 at various values of h/d can now be plotted as a
function of φ.
Discussion. – In order to see the relevance of these
results in view of previous studies, it is now suggested to
look at the dependence of the reduced viscosity in terms
of φ as a truncated series, while keeping only the first two
terms in φ, as in eq. (2).
η
− 1 = a1 (h/d) φ + a2 (h/d) φ2
η0
(2)
The exponential fit used in fig. 3 worked very well. It
was required for replotting the data for all values of h/d
with a very good accuracy (see fig. 4). On the other
hand, eq. (2) is more likely to be compared to previous works on suspensions [11, 14, 15]. As it is shown in
fig. 4, the asymptotic expansion should only be valid up
to medium range concentration as in the case of large h/d,
where Cichoki’s correction is shown to work only when
0 < φ . 0.12. As a comparison, the Krieger–Dougherty
model [21], ( ηη0 )KD = (1 − φ/φm )−2.5φm , works well for
all concentrations in the range studied. In any case, it
is clear that a curvature inversion is observed for confined suspensions as can be seen in fig. 4, especially when
h/d < 2. Then we used eq. (2) to predict the values of the
coefficients a1 (h/d) and a2 (h/d), functions of the reduced
gap h/d, when φ . 0.12. Good correlations are obtained
within measurement uncertainty.
In fig. 5, values of such parameters a1 (h/d) and a2 (h/d)
used in the asymptotic expansion are shown as a function
of confinement. Several comments can be made. First it
is demonstrated that as gap size decreases, a1 increases,
therefore moves away from the constant Einstein’s value
of 2.5. The latter value is obtained only for large values
of h/d. The behavior of the second coefficient a2 is more
complex and is important because it contains the hydrodynamic interactions between particle pairs. This term a2
starts with values close to 5, as predicted earlier [14, 15],
and decreases very sharply as confinement increases. This
result is similar to the previous numerical study [11] and
to a recent work using a semi–analytical treatment [22].
Thus it can be concluded that hydrodynamic interactions
in the presence of walls decrease the second term in the
expansion until it becomes very small and possibly negative.
The discussion of these results can be related to several previous works. Confinement is believed to affect the
hydrodynamic interactions between particles [23]. Goyon
et al. [2] showed that cooperative effects can take place
through the introduction of a correlation length in the
flow of confined concentrated emulsions whereas Frydel
et al. [24] found velocity correlations over long distances
and long times. Thus our results provide another justification of the viscosity change due to the presence of walls
in the vicinity of the particles. Note that the effective
viscosity was also measured in capillary tubes (through
pressure determination) in the case of concentrated suspensions of red blood cells by Pries et al. [25] and that
changes were also observed for high confinements. But
this study is the first one to analyze rheometrical data
in the context of the well–known Batchelor’s eq. (1) for
semi–dilute suspensions, which is largely affected by confinement. First, the parameters of the classical formulas
[14, 15] are not valid any longer as one moves away from
the continuum mechanics limit, in our case this limit is
met when 7 < h/d < 10; then unusual effects appear
with higher confinements, as already suggested previously
[11, 26]. Finally simple formulas can be proposed to investigate the dependence of the effective viscosity of the
type ηη0 − 1 = a1 (h/d) φ + a2 (h/d) φ2 , where a1 (h/d) and
a2 (h/d) have been measured experimentally, and correspond to simulations [11]. Note that we remain in the
range of small concentrations such as φ . 0.12 in order
to look for a reduced series expansion in terms of φ. It
p-3
P. Peyla et al.
is possible that larger concentrations may be used from
this work up to φ ∼ 0.2, but then other relations may
be needed by modification of the empirical fits by Krieger
and Dougherty [21]. Note that for higher concentrations,
it may also be necessary to consider the more complex rheology of concentrated suspensions and use other models
based on the shear–rate dependence of the viscosity, like
shear–dependent [8] or yield–stress fluids [7]. Such fractal
models or Herschel–Bulkley laws exist and seem to provide
satisfactory results for unconfined suspensions [9, 27] but
fail in confined concentrated systems [2]. Indeed a cooperative length scale defines the transition [2, 28], but it is
very small already below 40%. Thus this study precisely
falls in the range where no controlled–length phenomena
exist. Still it leads to unusual effects. Finally, other effects like ordering transitions may also appear at higher
concentrations [10] leading to a different rheology. But
this was not the major purpose of this study, for already
surprising results were found to arise only by studying the
zero–shear viscosity of confined suspensions.
Conclusions. – In this analysis, rheometrical data
has been obtained for suspensions as the gap becomes
close to the diameter of suspended particles, for moderate concentrations. The results found lead to an extension of Batchelor’s work [14] in terms of the reduced gap
parameter h/d, and simple formulas were derived from experiments. The parameters in eq. (2) were determined so
that the relationship between the effective reduced viscosity ηη0 − 1 and φ showed a concavity inversion at small
reduced gaps h/d. This model allows one to apply viscosity corrections when confinement effects become important. It may be realistic to use it in the case of special
processes, such as moulding, forming or extrusion of materials which involve suspensions in geometries whose sizes
become comparable to particle diameters.
[8] Brown E., Zhang H., Forman N. A., Maynor B. W.,
Betts D. E., DeSimone J. M. and Jaeger H. M., J.
Rheol. , 54 (2010) 1023.
[9] Quemada D., Eur. Phys. J. AP , 1 (1998) 119.
[10] Yeo K. and Maxey M. R., Phy. Rev. E , 81 (2010)
051502.
[11] Davit Y. and Peyla P., Europhys. Letters , 83 (2008)
64001.
[12] Einstein A., Annals der Physik , 19 (1906) 289.
[13] Einstein A., Annals der Physik , 34 (1911) 591.
[14] Batchelor G. K. and Green J. T., J. Fluid Mech. , 56
(1972) 401.
[15] Cichoki B. and Felderhof B. U., J. Chem. Phys. , 89
(1988) 1049.
[16] Elkissi N., Piau J. M., Attané P. and Turrel G.,
Rheol. Acta , 32 (1993) 293.
[17] Verdier C. and Brizard M., Rheol. Acta , 41 (2002)
514.
[18] Binding D. M. and Walters K., J. Non-Newtonian
Fluid Mech. , 1 (1976) 277.
[19] Connelly R. W. and Greener J., J. Rheol. , 29 (1985)
209.
[20] Kramer J., Uhl J. T. and Prud’homme R. K., Polym.
Eng. Sci. , 27 (1987) 598.
[21] Krieger I. M. and Dougherty T. J., Trans. Soc. Rheol.
I, II (1959) 137.
[22] Sangani A., Acrivos A. and Peyla P., submitted ,
(2011) .
[23] Diamant H., J. Phys. Soc. Jpn. , 78 (2009) 041002.
[24] Frydel D. and Diamant H., Phy. Rev. Letters , 104
(2010) 248302.
[25] Pries A. R., Neuhaus D. and Gaehtgens P., Am. J.
Physiol. Heart Circ. Physiol. , 263 (1992) 1770.
[26] Cui B., Diamant H., Lin B. and Rice S. A., Phys. Rev.
Letters , 92 (2004) 258301.
[27] Iordan A., Duperray A. and Verdier C., Phys. Rev.
E , 77 (2008) 011911.
[28] Bonnoit C., Lanuza J., Lindner A. and Clement E.,
Phys. Rev. Letters , 105 (2010) 108302.
Acknowledgments. – We are grateful to Dr. H. Galliard (Laboratoire de Rhéologie, Grenoble) for providing
the PDMS fluid and helpful advice regarding the experiments.
REFERENCES
[1] Coussot P., Rheometry of pastes, suspensions and granular materials (Wiley–Interscience) 2005.
[2] Goyon J., Colin A., Ovarlez G., Ajdari A. and Bocquet L., Nature , 454 (2008) 84.
[3] Ovarlez G., Rodts S., Chateau X. and Coussot P.,
Rheol. Acta , 48 (2009) 831.
[4] Pesché R. and Ngele G., Phys. Rev. E , 62 (2000) 5432.
[5] Bhattacharya S., Blawzdziewicz J. and Wajnryb E.,
J. Fluid Mech. , 541 (2005) 263.
[6] Isa L., Besseling R., Morozov A. N. and Poon W.
C. K., Phys. Rev. Letters , 102 (2009) 058302.
[7] Yan Y., Zhang Z., Cheneler D., Stokes J. R. and
Adams M. J., Rheol. Acta , 49 (2010) 255.
p-4
Confined suspensions
Fig. 1: Microscopy images of the spherical monodisperse polystyrene particles of respective diameters 80µm, 140µm, 230µm,
from left to right. Scale bar is 100 µm on all three images.
2.5
2.5% (80µm)
5% (80µm)
7.5% (80µm)
10% (80µm)
12.5% (80µm)
15% (80µm)
17.5% (80µm)
20% (80µm)
η/η 0 − 1
2.0
2.5% (140µm)
5% (140µm)
7.5% (140µm)
10% (140µm)
12.5% (140µm)
15% (140µm)
17.5% (140µm)
20% (140µm)
2.5% (230µm)
5% (230µm)
7.5% (230µm)
10% (230µm)
12.5% (230µm)
15% (230µm)
17.5% (230µm)
20% (230µm)
1.5
1.0
0.5
0.0
0
200
400
600
800
1 000
1 200
1 400
1 600
h (µm)
Fig. 2: Reduced viscosity
η
η0
vs. gap h, at different concentrations φ, and for different particle diameters.
2.5
2.5%
5%
7.5%
10%
η/η 0 − 1
2
12.5%
15%
17.5%
20%
1.5
η/η0 − 1 = m1 + m2 exp (-m3 h/d)
1
0.5
0
0
5
10
15
20
h/d
Fig. 3:
Master curves showing
η
η0
− 1 vs.
80µm, 140µm, 230µm. Fits of the type
m3 can be found in table. 1.
η
η0
h/d at different concentrations and corresponding to particle diameters
− 1 = m1 + m2 e−m3 h/d are also shown, the values of parameters m1 , m2 , and
p-5
P. Peyla et al.
2
h/d=1.4
h/d=1.6
h/d=2
h/d=2.8
h/d=3.4
h/d=5
h/d=10
h/d=20
η /η 0 − 1
1.5
1
KD model
KD model
1+2.5 φ + 5 φ2
1+2.5 φ
0.5
0
0
0.05
0.1
0.15
0.2
0.25
φ
Fig. 4: Values of the reduced viscosity ηη0 − 1 vs. φ at different reduced gaps h/d. Fits using Krieger-Dougherty ( ηη0 )KD =
(1 − φ/φm )−2.5φm with φm = 0.5 and 0.55 are shown to work well at any concentration for high values of h/d (respectively for
h/d = 10 and 20). The classical Einstein law [12,13] ηη0 = 1 + 2.5φ and the correction [14,15] ηη0 = 1 + 2.5φ + 5φ2 are also shown
for comparison. The latter one works well for 0 < φ . 0.12. Lines to guide the eye for lower gaps (h/d = 1.4 and h/d = 1.6)
show the curvature inversion.
a1(num)
a2(num)
a1(exp)
a2(exp)
20
10
a1, a2
10
20
0
0
-10
-10
0
5
10
15
h/d
Fig. 5: Values of the parameters a1 and a2 used in the asymptotic expansion of eq. (2), in terms of the confinement h/d.
Parameters a1 and a2 have been adjusted from fig. 4 for φ . 0.12. The lines correspond to the numerical simulations of Davit
et al. [11].
p-6