See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/251552381

A new empirical viscosity model for ceramic suspensions

Article  in  Chemical Engineering Science · June 2011

DOI: 10.1016/j.ces.2011.03.040

CITATIONS READS

41 333

4 authors:

Bahman Amini Horri Ranganathan Prabhakar

University of Surrey Monash University (Australia)
39 PUBLICATIONS   875 CITATIONS    33 PUBLICATIONS   824 CITATIONS   

SEE PROFILE SEE PROFILE

Cordelia Selomulya Huanting Wang

UNSW Sydney Monash University (Australia)
218 PUBLICATIONS   5,366 CITATIONS    413 PUBLICATIONS   17,346 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

I am currently working on diagnosis parasitic diseases using molecular technique View project

GO membrane filtration process View project

All content following this page was uploaded by Ranganathan Prabhakar on 15 August 2021.

The user has requested enhancement of the downloaded file.

 Chemical Engineering Science 66 (2011) 2798–2806

Contents lists available at ScienceDirect

Chemical Engineering Science

journal homepage: www.elsevier.com/locate/ces

A new empirical viscosity model for ceramic suspensions

Bahman Amini Horri a, Prabhakar Ranganathan b, Cordelia Selomulya a, Huanting Wang a,n
a
Department of Chemical Engineering, Monash University, VIC 3800, Australia
b
Department of Mechanical Engineering, Monash University, VIC 3800, Australia

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents a new predictive viscosity model for ceramic suspensions. Starting from Einstein’s
Received 20 November 2010 model (1906), various theoretical, empirical, and phenomenological models have been proposed for
Received in revised form different suspension systems. However, there is still a lack of reliable model for ceramic suspensions
16 January 2011
used in colloidal ceramic shape-forming methods. Here, the rheological properties of ceramic
Accepted 17 March 2011
suspensions comprising NiO/YSZ (nickel oxide/yttria stabilized zirconia) as the ceramic powder, and
Available online 31 March 2011
furfuryl alcohol as the suspending media were measured over a range of shear-rates (between 1 and
Keywords: 1000 s  1) and different solid volume fractions from 0 to 0.4010. An empirical equation was then
Ceramic suspension developed for the ceramic suspensions using the mobility parameter (f/(fm  f)), which links
Relative viscosity
Einstein’s model with the more recent relative viscosity models. The proposed model was used to
Viscosity modelling
predict the relative viscosity data, showing excellent agreement to the experimental data from this
Rheology
Ceramic slurry viscosity study and with reported data in literature for other ceramic systems. The model was also used to
Suspension viscosity estimate the maximum solid volume fraction for the ceramic suspensions (fm ¼ 0.571), with better
accuracy than those estimated by existing models.
& 2011 Elsevier Ltd. All rights reserved.

1. Introduction (Moreland, 1963; Saraf and Khullar, 1975; Bhattacharya et al.,

Ceramic suspensions are used as the starting material in
colloidal-processing techniques such as slip casting, gel casting,
tape casting, spin coating, spray coating, slurry painting, and
others. Controlling and understanding the rheological behaviour
of ceramic suspensions are essential to reduce processing costs and
achieve desired ceramic microstructural properties (Lewis, 2000;
Yar et al., 2010). In the colloidal-processing methods, it is desirable
to get the highest possible ceramic loading at the lowest possible
suspension viscosity to achieve a homogeneous, dense, and defect-
free green body and lower shrinkage (Takai et al., 2006).
Ceramic suspensions especially those with higher solid volume
fraction, generally show non-Newtonian rheological behaviour
which are characterized by shear-dependent viscosities. Further-
more, the highly concentrated ceramic slurries often exhibit a time-
dependent decrease of apparent viscosity in ascending shear-rates
(upward), but the viscosity is recovered during descending shear-
rates (downward). This reversible flow characteristic is called
thixotropy (Usui et al., 2001). For a suspension containing spherical
solid particles at constant solid volume fraction, decrease in the
solid-particle size leads to an increase in the number of solid
particles and consequently increases the suspension viscosity
n
Corresponding author. Tel.: þ61 3 9905 3449; fax: þ61 3 9905 5686.
E-mail address: huanting.wang@monash.edu (H. Wang).
1998). Rheological properties of ceramic suspensions are affected
by a large number of parameters including the physical and
chemical properties of solvent, pH level, solid volume fraction,
particle size, particle size distribution, particle-solvent interactions,
temperature, and shape and surface properties of particles (Zhou
et al., 2001; Senapati et al., 2010). The prediction of suspension
viscosity is a long-standing problem in fluid mechanics and has
received much attention for years (Mewis and Wagner, 2009).
Among the different parameters, solid volume fraction (f) signifi-
cantly affects the suspension viscosity (m) and thus many empirical
and theoretical correlations have been developed to describe the m–
f relationship (Usui et al., 2001; Pabst, 2004; Stickel and Powell,
2005; Mueller et al., 2010).
So far, a variety of models have been proposed to describe the
m  f relationship (Table 1). Einstein (1906, 1911) was the first to
propose a viscosity model for dilute suspensions as
mr ¼ 1þ Bf ð1Þ
where mr is the relative viscosity, defined as a ratio of suspension
viscosity to the viscosity of the pure suspending fluid. The coeffi-
cient B was referred to as the intrinsic viscosity with a value of 2.5.
The model was developed based on the assumption of spherical and
non-interacting rigid particles in a highly dilute suspension system
(f o0.02) in Newtonian solvents without Brownian motion.
A higher order expansion of Einstein’s model was frequently used
for suspensions with spherical and rigid particles (Oliver and Ward,

0009-2509/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2011.03.040
 B.A. Horri et al. / Chemical Engineering Science 66 (2011) 2798–2806 2799

Table 1
Summary of different viscosity models for suspensions and their applications.

Model Correlation Method of approacha Model coefficients Application remarks Eq. no.

Oliver and Ward (1953) P

1 II kb Low to moderate concentration (f o 0.2) (4)
mr ¼ 1 þ ðkfÞn
n¼1
B
Roscoe (1952) mr ¼ (1 kf) II B¼ 2.5, K¼1.35 High concentration (5)
 
Mooney (1951) Bf
mr ¼ exp 1K II B¼ 2.5, 1.35o Ko 1.91 Low to high concentration (6)
f
Krieger and Dougherty (1959)  Bfm II B¼ 2.5, fm ¼ 0.6–0.7 Low to high concentration (7)
mr ¼ 1 ffm
Chong et al. (1971)  2 III K¼ 0.75, fmb High concentration (8)
mr ¼ 1 þ K 1ðff=f=fm m Þ
 n b
Dabak and Yucel (1986) mr ¼ 1 þ nðBfff m III B¼ 2.5n, fm High concentration, low to high shear-rates (9)
m  fÞ

Quemada’s Model  2 II and (K0, KN,gr)c High concentration (10)

1=2
mr ¼ 1 12 kf K0 þ K1 r g
K¼ 1=2
g
1þ r
Brouwers (2010)  ðBfm Þ=ð1fm Þ I B¼ 2.5, fmb High concentration, monosized, spherical (11)
1f
mr ¼ 1ðf=fm Þ

a
I: Theoretical, II: phenomenological (empirical observations and theoretical fundamentals), III: empirical.
b
The model coefficient(s) should be determined experimentally via curve fitting.
c
K0 and KN are intrinsic viscosities at very low and very high shear rates, and gr is the reduced shear rate and can be determined experimentally.

1953; Rutgers, 1962; Hsueh and Becher, 2005). Efforts to find a
valid model for moderate-concentration suspensions were focused
on expanding Einstein’s model as
X
1
n ceramic suspensions was performed by Liu (2000), who proposed
mr ¼ 1 þ Bn f
n¼1
In the second order form of the equation (n¼2), B1 took the value of
2.5 while B2 took different values ranging from 7.349 to 14.1 (Guth,
1945; Manley and Mason, 1945; Vand, 1948). The main problem for
those correlations is that termination of series after second term
which means 10% or greater error at higher solid volume fractions
(f 40.15–0.2). Also, the coefficients for higher order forms of
Eq. (2) were too difficult to determine analytically. For the third
order form of the equation (n¼3), a notable work was done by
Thomas and Muthukumar (1991). They analytically derived the
values of B, B1, and B2 to equal to 2.5, 4.83, and 6.4, respectively, by
applying multiple scattering theories and considering the hydro-
dynamic interaction between three hard spheres. An early survey
summarizing more than a hundred theoretical and empirical
viscosity correlations was made by Rutgers (1962). Another com-
prehensive review for early rheological models was done by
Thomas (1965). For high concentration suspensions, several corre-
lations were proposed as closed mathematical forms. Also, since
there were less analytical solutions for high concentration suspen-
sions, empirical and phenomenological (partly empirical with
theoretical basis) approaches have been adopted to predict m–f
relationship. A phenomenological closed-form equation was pro-
posed by Mooney (1951), who derived a model based on hydro-
dynamic interaction theory, but with the model coefficients were
defined empirically. The first successful model for high-concentra-
tion suspensions was developed by Roscoe (1952) through a semi-
empirical approach. Gradually, the maximum solid fraction (fm)
was introduced in m–f correlations as a new parameter. One of the
most widely used equations was formulated by Krieger and
Dougherty (1959), which can be reduced to Einstein’s model at
the limit of infinite dilution (f-0). Chong et al. (1971) proposed an
empirical model for suspensions comprised amorphous viscoelastic
solvents and spherical particles. They typically used fm ¼0.605 for a
suspension of glass beads and found that it can be expressed as a
function of particle size distribution. Another empirical model was
suggested by Dabak and Yucel (1986), as a three-parameter model
for low and high shear rates. The experimental exponent of the
model (n in Eq. (8), Table 1) reflected the level of particle associa-
tion at the initiation of motion. Meanwhile, a phenomenological
model was developed by Quemada (1977), based on the
optimization of viscosity energy dissipation theory with a phenom-
enological constant for the model determined experimentally to be
between 2.5 and 3.7 for high concentration limit. A more recent
work to predict the relative viscosity for highly concentrated
ð2Þ
an empirical model based on mathematical modelling of experi-
mental data with the following equation:
 2
mr ¼ aðfm fÞ ð3Þ
1=2
The parameters a and fmwere determined via a single 1mr f
plot using viscosity–concentration data. In recent years, efforts to
find the m–f correlations have focused on representing new
theoretical method for spherical and non-spherical rigid particles
for low and high concentrated suspensions (Hsueh and Becher,
2005; Sherwood, 2006; Brouwers, 2010). Hsueh and Becher (2005)
applied a closed-form solution using both elastic-viscous analogy
and the analogy between polarizable-medium effective dielectric
constant and two-phase flow effective viscosity for the relative
viscosity of monosized spheres. Sherwood (2006) used cell model
theory in various boundary conditions for a suspension of spherical
particles in Newtonian fluids. Brouwers (2010) has performed the
most recent work in this field and applied geometrical considera-
tion to predict the relative viscosity of monosized particles.
In addition, in the field of thixotropy (time-dependent viscos-
ity) modelling, the basic idea was proposed by Moore (1959) to
explain rheology properties of high concentrated ceramic slurries.
Cheng and Evans (1965) then generalized the Moor’s model to
obtain conditions for thixotropy and antithixotropy. General
properties of thixotropy materials have been reviewed by
Mewis (1979) and Barnes (1997). An extensive list of available
thixotropy models have been summarized by Mujumdar et al.
(2001). Most authors use a scalar structure parameter (l) to
describe the level of slurry microstructure. The parameter l varies
between the values of zero for the completely broken-down
structure and one for the entirely built-up structure, although
values that diverge to infinity have been used in some literature
(Roussel et al., 2004). Mewis and Wagner (2009) have recently
published a remarkable review of the thixotropy fundamentals
and existing models.
Although, various models have been proposed to estimate the
suspension viscosity, however they are not sufficiently accurate
for ceramic slurries commonly used in the colloidal ceramic
shape-forming methods.
In this work, a new empirical model was proposed to predict
the viscosity of ceramic suspensions and slurries (highly
2800 B.A. Horri et al. / Chemical Engineering Science 66 (2011) 2798–2806
concentrated suspension). The model was developed based on the
rheological studies using NiO/YSZ (nickel oxide/yttria stabilized
zirconia) suspension which is an example of the ceramic powders
widely used in fabrication of anode-supported solid oxide fuel
cells through colloidal ceramic shape-forming method. To demon-
strate the model application, it was also used to predict the
relative viscosity data for a number of suspensions as reported in
literature.
2. Experimental
2.1. Raw materials
Nickel carbonate-basic hydrate, BNC (NiCO3  2Ni(OH)2  H2O,
MW¼304.12 g/mol, Sigma-Aldrich), and yttria stabilized zirco-
nium (IV) oxide, YSZ (ZrO2,Y2O3, 99.9%, MW¼123.22 g/mol,
Sigma-Aldrich) were used as the source of ceramic powders.
Furfuryl alcohol, FA (C5H6O2, MW¼98.1 g/mol, Sigma-Aldrich)
was used as suspending media. All materials were used without
further treatment.
2.2. Preparation and characterization of ceramic powders
There are various methods for preparation of nickel oxide
particles such as sol–gel, surfactant mediated, thermal decom-
position, precipitation, and polymer matrix assisted synthesis. In
this work, thermal decomposition method was used to prepare
the nickel oxide. About 300 g of BNC was decomposed at 900 1C
for 3 h. NiO/YSZ ceramic composite was prepared using NiO and
YSZ powders at the mass ratio of 60:40 and dispersed into ethanol
to make a suspension, and then mixed for 24 h by a ball-milling
technique (at 80 rpm). Then the suspension was dried at around
80 1C for 24 h. The produced ceramic powder (NiO/YSZ) was used
to prepare the suspensions using furfuryl alcohol (FA).
The prepared NiO/YSZ ceramic powder was characterized by
X-ray diffraction method (Phillips 1140 diffractometer). The mor-
phology also was observed by scanning electron microscopy (JEOL
JSM-840A, Japan). The particle size distribution of the prepared
ceramic sample was measured by the laser diffraction and
scattering method (mastersizer 2000, Malvern Instruments Ltd.).
The particle density was determined by gas pycnometric techni-
que using nitrogen (AccuPyc-1330 pycnometer, Micrometrics).
2.3. Ceramic suspension preparation and viscosity measurement
The ceramic suspension samples were prepared by suspending
a calculated amount of ceramic powder (NiO/YSZ) into FA. The
suspension was mixed by ball-milling at 80 rpm for 24 h. The
solid volume fraction of the samples was calculated based on the
ceramic powder weight fraction. The following equations were
used to compute the quantities of ingredients for each sample:
ms
w¼
ml
Vs w
|¼ ¼ ð13Þ
Vs þ Vl wþ ðrs =rl Þ
where w is the solid content, ms and ml are solid and liquid loading
mass, respectively, and rs (6.1 g/cm3) and rl (1.135 g/cm3) are
solid and liquid densities, respectively.
After ball-milling, the viscosity and shear stress of each sample
were measured by the HAAKE MARS III rheometer (Thermo Fisher
Scientific Inc.) over a wide range of shear rate between 1 and
1000 s  1 within 15 logarithmic intervals (16 points) for a time
period of 6 min. All rheological measurements were carried out at
the temperature around 22 71 1C. Initially, the samples were pre-
sheared with three shear rates of 10, 100, and 1000 s  1, respec-
tively, for 20 s at each step to get steady-state condition. The
viscosity of FA (suspending media) was measured at zero solid
loading fraction (f ¼ 0) as 5.224 mPa s.
2.4. Data analysis method
To provide the characteristic rheological curves, Tablecurve
programme and MS Excel were used to fit and plot the curves. To
relate the parameters of each model to the sample properties and
data, a stepwise multiple linear and nonlinear regression analysis
were performed based on the Levenberg–Marqurdt (L–M) algo-
rithm. For each model, the goodness-of-fit was statistically
determined using the coefficient of determination (R2) and the
Root of Mean Squared Errors (RMSE).
3. Results and discussion
3.1. Microstructure of the NiO/YSZ ceramic powder
Figs. 1 and 2 show the XRD spectrum and SEM image of the
prepared NiO/YSZ powder. Fig. 3 shows the particle size distribu-
tion of the prepared ceramic sample. The powder was dispersed
in distilled water before the particle size measurement and the
results were made in term of volume-averaged diameters for
particles of any shape. According to Fig. 3, the ceramic powder
shows a unimodal distribution in the whole particle size range
and the volumetric average particle size (d4,3), d0.1, d0.5, and d0.9,
are 1.027, 0.566, 0.975, and 1.569 mm, respectively.
3.2. Rheological results
The volume solid fraction for viscosity measurement was
varied from 0.0694 to 0.4104 (6.94 to 41.04 vol%) because the
ceramic suspensions were prone to settle below 6.94 vol%. Visual
tests were done to determine the maximum solid volume fraction
of the samples by preparing suspension with higher solid volume
fractions (f 40.41). After ball milling for 24 h, the samples were
visibly checked to have any non-flowing property, separable
particles or dry agglomerated particles in the mixture. By repeat-
ing the test two times, the value of 0.57170.001 was determined
as the maximum solid volume fraction (fm). To characterize the
rheological behaviour and also to calculate the apparent viscosity
(ms) of the ceramic suspensions, the shear-stresses (t) were
measured over a range of shear-rates (g1) from 0 to 1000 s  1 for
ð12Þ
Fig. 1. XRD pattern of the NiO/YSZ ceramic powder.
 B.A. Horri et al. / Chemical Engineering Science 66 (2011) 2798–2806
Fig. 2. SEM image of the NiO/YSZ ceramic powder.
Fig. 3. Particle size distribution of the NiO/YSZ ceramic powder.
each sample. Fig. 4 show the shear-stress versus shear-rate
rheograms of the prepared ceramic suspensions.
In Fig. 3, the rheograms of shear stress-shear rate indicated
non-Newtonian behaviour from the shape of curves, especially
pseudoplastic (shear thinning) behaviour for the NiO/YSZ-FA
suspensions. The shear thinning behaviour is a common rheolo-
gical behaviour for most ceramic suspensions. The shear thinning
behaviour has been explained as a perturbation of suspension
structure by the applied shear. At lower shear-rates, the suspen-
sion structure is close to equilibrium and exhibits an apparent
viscosity with a constant slope, since thermal motion dominates
over the viscous forces. At higher shear rates, the viscous forces
have more influence over the suspension structure, causing them
to become distorted and thus causing shear thinning or nonlinear
apparent viscosity (Yuanling et al., 2004).
The relative viscosity (mr) of the ceramic suspension was
calculated by dividing the apparent viscosity (ms) by the suspend-
ing media viscosity (mf). Table 2 shows the relative viscosity
results for typical shear-rates.
3.3. Viscosity model and analysis
Regardless of type, a predicting viscosity model must satisfy
three important conditions. Firstly, the predicted relative viscosity
must approach one as the solid volume fraction decreases to zero.
2801
Fig. 4. Rheogram of shear stress versus shear rate for NiO/YSZ-FA suspensions.
Among the available models, Liu’s model is an example that does
not satisfy this condition, and consequently cannot to be used for
low-concentration suspensions. Secondly, the relative viscosity
predicted by a model must approach infinity as the solid content
increases to the maximum solid volume fraction (Chong et al.,
1971). Thirdly, the intrinsic viscosity definition should be valid for
any viscosity model based on the particle movement theories
(Krieger and Dougherty, 1959; Powell, 1991; Stickel and Powell,
2005; Pbast et al., 2006). The conditions can be expressed as
lim mr ¼ 1 ð14Þ
f-0
lim mr ¼ 1 ð15Þ
f-fm

mr 1
lim ¼B ð16Þ
f-0 f
The parameter B has been defined as the intrinsic viscosity, which
takes the value of 2.5 for spherical particles (Einstein, 1906). For
non-spherical particles, B takes different values between 2.5 and
10.008 when the aspect ratio varies from 1.05 to 100 (Leal and
Hinch, 1971). Another work shows that B could be close to 2.5 for
the prolate and oblate spheroids when the ellipticity varies from
0 to 0.5 (Jeffery, 1922; Mueller et al., 2010).
All suspensions contain both solid powder and suspending
medium. Using the definition of void fraction (voidage) in particle
dynamics for rigid particles suspension systems as the computa-
tional basis for viscosity and supposing each part of the suspen-
sion exerts its own viscosity, the following equation can be
expressed:
ms ¼ emf þ ð1eÞmres ð17Þ
where e is the suspension void fraction which is filled by the
suspending medium (fluid), ms and mf are the suspension viscosity
and fluid viscosity, respectively, and mres is referred to as the
residual viscosity and defined as the exerted viscosity due to
immersed particles in fluids. Combining the void fraction law in
particle dynamics for solid–liquid systems (e þ f ¼ 1) and dividing
both side of Eq. (17) by mf will result in
mr ¼ ð1fÞ þ fmres
r ð18Þ
res
where mr is the relative viscosity of suspension and m is here r
defined as relative-residual viscosity. Eq. (18) can be rewritten to
2802 B.A. Horri et al. / Chemical Engineering Science 66 (2011) 2798–2806

Table 2
Relative viscosity data for NiO/YSZ-FA suspensions at different solid volume fractions.a

Shear rate, c1 (s  1) / ¼ 6.94% / ¼11.33% / ¼16.59% / ¼ 22.98% / ¼ 30.91% / ¼ 41.04%

25.12 3.58 9.46 26.80 93.80 344.58 1387.89

39.81 3.12 7.33 19.14 65.09 250.78 1196.46
63.09 2.76 5.82 14.80 45.94 174.20 915.05
100 2.45 4.75 11.24 32.54 126.35 744.67
251.2 2.05 3.41 6.97 17.99 71.56 432.49
630.9 1.79 2.64 4.73 10.68 38.29 239.29

a
Average error of 7 0.001 for reported f.

give an expression for the relative-residual viscosity as

mr ð1fÞ
mres
r ¼ ð19Þ
f
To find out a suitable correlation for the relative-residual viscos-
ity, Eq. (19) was used to calculate the values of mres
r based on the
experimental data reported in Table 2 (relative viscosity versus
solid volume fraction). A non-dimensional form of solid volume
fraction in the form of (f/(fm  f)) was used as the independent
variable. This non-dimensional term of (f/(fm  f)) was referred
to as the mobility parameter by Dabak and Yucel (1986) and is a
significant physical parameter affecting the suspension viscosity.
Fig. 5 shows the plot of the experimental values of relative-
residual viscosity (mres
r ) versus mobility parameter ((f/(fm  f)))
at different shear rates. Curve fitting analysis was used to
correlate the residual-relative viscosity with the mobility para-
meter using the experimental data in Fig. 5, the same method was
also applied to model the data from literature (as referenced in
Table 5). The following equation was found to satisfy the experi-
mental data from this work and other works in literature:
X1  2n
1 f
mres
r ¼ aþb ð20Þ
n¼1
2n! fm f
Fig. 5. Plot of experimental residual-relative viscosity versus mobility parameter
where a and b are the model coefficients and n is the equation (f/(fm  f)) at different shear rates (both axes are dimensionless).
order. By substituting Eq. (20) into Eq. (18), the following
equation can be derived for the relative viscosity:
Eq. (23) reconciles Einstein’s model with the more recent models
X1  2n
1 f using the mobility parameter. The parameters K and fm can be
mr ¼ 1 þ ða1Þf þ bf ð21Þ
n¼1
2n! fm f determined from experimental data by curve fitting analysis.
Eq. (23) was used to predict the relative viscosity measured
Eq. (20) must satisfy the general conditions of viscosity predictive experimentally in this work as shown in Table 3, which also
models for suspensions, i.e. Eqs. (14)–(16). Eq. (20) satisfies the shows the comparative results of other models.
first and second conditions, because as the solid volume fraction The proposed model has the highest values of coefficient of
(f) reduces to zero, mr approaches 1, while mr approaches infinity determination (R2) and also the lowest values of the root mean
when f increases to fm. The third condition, Eq. (16), can be square error (RMSE) for the shear-rate range observed here
applied using the following equation: (Table 3). By comparison, although Liu’s model has reasonably
 (  2n )
mr 1 X1
1 f good R2, their calculated RMSE values are relatively high to predict
lim ¼ lim ða1Þ þb ¼B ð22Þ the experimental data with acceptable accuracy, particularly for
f-0 f f-0
n¼1
2n! fm f
low solid volume fractions. Dabak and Yucel’s model shows
which implies that the coefficient a takes the value of B þ1. From relatively low R2 and high RMSE, while Chong et al. model could
the SEM micrograph (Fig. 2), it can be observed that the ceramic only predict the viscosity of the ceramic suspension at higher
particles of this work are more close to spheroid such that the shear rates ( 4251.2 s  1). Other models such as Brouwer’s model,
value of B of 2.5 can be used. Thus, the value of a can be estimated Krieger & Dougherty’s model, Mooney’ model and Roscoe’s model
to be 3.5. It is obvious that to determine the exact value of a, the were not successful in predicting the experimental data.
aspect ratio of particles should be determined accurately. The Since both the proposed model and Liu’s model showed
parameter n in Eq. (21), can be assumed to be any positive integer acceptable values of R2 at overall shear-rate range (R2 40.999
number (1, 2, 3, 4, etc.), although in this case, the relatively good for the proposed model and R2 40.989 for Liu’s model), the
fit of Eq. (20) to the experimental data can be achieved when n predicting ability of the proposed model and Liu’s model were
takes the value of 1 even for highly concentrated suspensions and studied using the RMSE as the basic analysing parameter. The
higher shear rates. Thus, with the assumptions of n¼ 1, a ¼3.5 and RMSE analysis was carried out by comparing the observed and
substituting K ¼b/2, Eq. (21) becomes estimated relative viscosities at the same solid volume fractions
 2 (Fig. 6A and B). As shown, the estimated values of mr by the
f proposed model (Eq. (23)) are very close to the observed values
mr ¼ 1 þ2:5f þ K f ð23Þ
fm f (Fig. 6A), while for the Liu’s model, a considerable deviation can
 B.A. Horri et al. / Chemical Engineering Science 66 (2011) 2798–2806 2803

Table 3
Model coefficients and goodness-of-fit data for different viscosity models.

Model Model coefficients and goodness of fit c1 (s  1)

25.12 39.18 63.09 100 251.2 630.9

Proposed Model (Eq. 23) K 1996.8 934.53 533.62 311.76 163.92 80.468
/m 0.7260 0.6430 0.6114 0.5808 0.5727 0.5636
2
R 0.9999 0.9999 0.9999 0.9999 0.9999 0.9999
RMSE 3.9674 1.8255 1.0216 1.1462 0.8255 0.7101

Liu’s Model (Eq. (8)) a 0.3369 0.4124 0.5030 0.6050 0.8059 1.0876
/m 0.4900 0.4805 0.4761 0.4709 0.4701 0.4698
R2 0.9897 0.9933 0.9951 0.9965 0.9969 0.9979
RMSE 61.538 42.720 28.030 19.408 10.578 4.8584

Dabak & Yucel Model (Eq. (7)) n 175.24 160.745 208.40 230.92 2.1E13 1.8E13
/m 0.4766 0.4781 0.4817 0.4846 0.4939 0.5050
R2 0.9175 0.9431 0.9544 0.9554 0.9703 0.9766
RMSE 174.36 125.25 85.838 61.013 32.828 16.038

Chong et al. Model (Eq. (6)) /m – – – – 0.4259 0.3917

R2 0.9712 0.9861
RMSE 28.847 11.076

Fig. 7. Experimental data and relative viscosity predicted by different model

versus solid volume fraction at shear-rate 251.2 s  1 (for Mooney’s, Krieger &
Dougherty’s, and Brouwers’s model the experimental value of fm was used,
fm ¼0.571; both axes are dimensionless).

be seen between the estimated and observed values especially at

Fig. 6. RMSE analysis for the proposed model (A) and Liu’s model (B) based on
calculated coefficients in Table 3 (both axes are dimensionless).
lower relative viscosities (Fig. 6B).
The modelling plots are shown in Fig. 7 for the shear rate of
251 s  1 showing that the proposed model can predict the relative
viscosity of the ceramic powder with an acceptable deviation at
the entire range of the solid volume fraction. The logarithmic
scale in Fig. 7 was selected to show the contrast between different
models. Compared to the other models, the proposed model is
more accurate in predicting the viscosity values especially at
lower f. This can be associated with using the intrinsic viscosity
(the coefficient of 2.5 in Eq. (23)) and mobility term (the term of
K(f/(fm  f))2 in Eq. (23)) as the controlling parameters of the
relative viscosity values at lower and higher solid volume frac-
tions, respectively. Also, the intrinsic viscosity strongly controls
the slope of curve especially at lower solid volume fractions.
The relative impact of both coefficients on Eq. (23) can be expressed
by dividing the intrinsic coefficient by the mobility term
(2.5/Kðf=ðfm fÞÞ2 ) as shown in Fig. 8. Depending on the shear-
rate values, the range of f between 0.025 and 0.085 is showing the
2804 B.A. Horri et al. / Chemical Engineering Science 66 (2011) 2798–2806

same impact for the mobility term and the intrinsic coefficient,
referred to here as the equal-impact line (Fig. 8). Below the equal-
impact line (for example f o0.08 for g1¼630.9 s  1), the intrinsic
coefficient is the main controlling parameter by the ratio of 1 to
around 100. On the contrary, above this criterion (for example
f 40.025 for g1¼ 25.12 s  1), the mobility term is the main con-
trolling parameter by the ratio of 1 to around 1000.
Yucel’s models, the predicted values for fm by the proposed
model have a wider variation, with overestimated values at the
lower shear rates ( o100 s  1). However the average errors overall
were lower, and the model can predict fm more accurately at
higher shear rates.
3.5. Application

3.4. Maximum solid volume fraction
The maximum solid volume fraction is an important physical
property of a suspension that strongly depends on the particle
size, shape and particle size distribution for rigid particles.
Theoretical approaches for prediction of fm have been usually
developed for spherical particles (random packing) with consid-
erable inconsistency between calculated and measured values
(Dabak and Yucel, 1986). In this work, by introducing the mobility
parameter ((f/(fm  f))) in the viscosity model, a better predic-
tion of fm can be achieved. Table 4 shows the predicted values of
fm by different models with the average errors from the different
model. Krieger & Dougherty’s and Brouwers’s models could not be
used to predict physically realistic values for fm at the range of
shear rates of the present work. Compared with Liu’s and Dabak &
To evaluate the model predictive capability for different
ceramic suspension systems, various data in the literature were
used to test its validity. Data were selected by considering the
range of solid volume fractions (low to high concentration), wide
range of shear rates (1–1000 s  1), and different suspension
systems, temperatures, and measuring equipments. Despite the
scattering in the reported data, the proposed model could predict
the experimental data relatively well. Table 5 summarized the
goodness of fit and the predicted coefficients for different ceramic
and non-ceramic systems. In most cases, the tabulated data for
relative viscosity versus solid volume fractions were used. One
difficulty on using existing data was that the maximum solid
volume fraction was often not reported. Based on literature, the
measured fm for alumina/xylene, silica/ethanol, silica/glycerol,
and coal/glycerin are 0.468, 0.652, 0.640, and 0.520, respectively
(Dabak and Yucel, 1986). Comparing the reported data for the
maximum solid volume fraction with the predicted values in
Table 5 (0.4646, 0.6403, 0.6494, and 0.5328, respectively) demon-
strates the relatively good predictive capability of the proposed
model for the maximum solid fraction of these systems.

4. Conclusions

In the current work, a general model has been introduced to

Fig. 8. Relative impact of the intrinsic viscosity and the mobility term on Eq. (23)
at different shear-rates (both axis are dimensionless).
predict the viscosity of ceramic suspensions. The proposed model
uses the mobility term ((f/(fm  f))) in accordance with
Einstein’s model. For validation, a set of experiments was con-
ducted by preparing suspensions comprising different amounts of
NiO/YSZ as the ceramic powder and FA (furfuryl alcohol) as the
suspending medium. For the prepared ceramic suspensions,
the solid volume fraction was in the range of 0 to 0.4104 with
the maximum solid volume fraction of 0.571. The prepared
samples were measured to determine the shear-stress versus
shear-rate data between 1 and 1000 s  1. The relative viscosities
of the system were modelled using the proposed equation with
good fit. The model could also be used to predict the maximum
solid volume fraction (fm) for ceramic suspensions with much
greater accuracy especially at high shear rates compared to

Table 4
Predicted values for fm by different modes.a

Model Parameter c1 (s  1) Average error (%)

25.12 39.18 63.09 100 251.2 630.9

Proposed model (Eq. (23)) Predicted fm 0.7260 0.6430 0.6114 0.5808 0.5727 0.5636 7.92
Error (%) 27.14 12.61 7.07 1.72 0.297  1.31

Liu’s Model (Eq. (8)) Predicted fm 0.4900 0.4805 0.4761 0.4709 0.4701 0.4698  16.60
Error (%)  14.18  15.85  16.62  17.53  17.67  17.72

Dabak and Yucel’s Model (Eq. (7)) Predicted fm 0.4766 0.4781 0.4817 0.4846 0.4939 0.5050  14.77
Error (%)  16.53  16.27  15.64  15.13  13.50  11.56

Chong et al. Model (Eq. (6)) Predicted fm – – – – 0.9712 0.9861 71.40

Error (%) 70.08 72.70

Brouwers’ Model (Eq. (9)) Predicted fm  0.156  0.285  0.489  1.083  32.24 –  1299.7
Error (%)  127.3  149.9  185.6  289.6  5746

a
The experimental value of the maximum solid volume fraction is 0.571 for samples.
 B.A. Horri et al. / Chemical Engineering Science 66 (2011) 2798–2806 2805

existing models. Its application for other ceramic and non-

26.234

2.9679

1.3385

2.2133
0.2953

1.3015

0.3343

5.1025

0.1782

0.8611

0.4556

0.8285

0.1449
2.3607

0.2599
0.0598
RMSE ceramic suspensions was verified using various experimental
data available in literature.
0.9986

0.9999
0.9999

0.9994

0.9999
0.9999
0.9999
0.9998

0.9971

0.9975

0.9987
0.9999
0.9999
0.9958
0.9985

0.9935
Acknowledgements
R2

The authors acknowledge the assistance of Dr. D.A. Nguyen in

0.6494
0.7828

0.4864
0.5452

0.4646

0.5328

0.8937

0.7559
0.7734
0.6418

0.7553
0.6940

0.6403

0.8055

0.7803
0.7490
the molecular (rheology group) for help with rheology measure-
/m

ment. Also, the technical support of MCEM is gratefully acknowl-

edged. This work was supported by the Australian Research
12.975
39.384

4.4515

5.2372

1.7319
3.6985

14.766
9.9501

208.68

0.3295

8.1064

10.766

3.7308

1.0841
1.0018

0.8047
Council through a Discovery Grant (DP0877305).

Solid content is a mixture of tungsten carbide and cobalt alloy, organic solvent is mixture of 20% paraffin wax, 20% ethylene acrylic acid, 10% polyethylene glycol, and 50% LDPE.
K

References
Solid/solid ratio

Barnes, H.A., 1997. Thixotropy—a review. Journal of Non-Newtonian Fluid

85/15 Mechanics 70, 1–33.
90/10

80/20
100/0

Bhattacharya, I.N., Panda, D., Bandopadhyay, P., 1998. Rheological behaviour of

95/5
95/5

nickel laterite suspensions. International Journal of Mineral Processing 53,

–
–

–
–

– 251–263.
Brouwers, H.J.H., 2010. Viscosity of concentrated suspension of rigid monosized
Shear rate (s  1)

particles. Physical Review 81, 041402-1–041402-10.

Cheng, D.C.H., Evans, F., 1965. Phenomenological characterization of the rheolo-
gical behaviour of inelastic reversible thixotropic and antithixotropic fluids.
British Journal of Applied Physics 16, 1599–1617.
700–800

Chong, J.S., Christiansen, E.B., Baer, A.D., 1971. Rheology of concentrated suspen-
1000

1000

143
108

108

sions. Journal of Applied Polymer Science 15, 2007–2021.

10

80
1
1
1

Dabak, T., Yucel, O., 1986. Shear viscosity behavior of highly concentrated
suspensions at low and high shear-rates. Rheological Acta 25, 527–533.
Einstein, A., 1906. Eine neue Bestimmung der Molekuldimensionen. Analen der
/ Range

0–0.45

0–0.65

0–0.45

0–0.46

0–0.64

0–0.67

Physik 324, 289–306.

0–0.6
0–0.5

0–0.5

0–0.7

0–0.4

Einstein, A., 1911. Berichtigung zu meiner Arbeit: eine neue Bestimmung der
Molekuldimensionen. Analen der Physik 339, 591–592.
Organic solvent is a mixture of poly propylene, microcrystalline wax, and octadecanoic acid in the weight ratio 6:2:1.

Ewais, E., Zaman, A.A., Sigmond, W., 2002. Temperature induced forming of
zirconia from aqueous slurries: mechanism and rheology. Journal of European
3.58/7.77: (0.2–25)/(0.4–100)

Ceramic Society 22, 2805–2812.

Guth, E., 1945. Theory of filler reinforcement. Journal of Applied Physics 16, 20–25.
Average PS:PSDa (lm)

Honek, T., Hausenerova, B., Saha, P., 2005. Relative viscosity models and their
application to capillary flow data of highly filled hard-metal carbide powder
13.51: 4.74–34.55

compound. Polymer Composites 26, 29–36.

Hsueh, C.H., Becher, P.F., 2005. Effective viscosity of spheres. Journal of the
0.4: 0.25–0.95

3.58: 0.2–25

American Ceramic Society 88, 1046–1049.

1 : 0.2–10

0.7: 0.2–5

Jeffery, G.B., 1922. The motion of ellipsoidal particles immersed in a viscose fluid.
–: 0.5–5

Proceedings of the Royal Society, A: Mathematica, Physical and Engineering

Science 102, 161–179.
Model evaluation and curve fitting results for different ceramic and non-ceramic suspensions.

Kao, S.V., Nielsen, L.E., Hill, C.T., 1975. Rheology of concentrated suspensions
–
–

–
–

spheres, I: effect of the liquid–solid interface. Journal of Colloid and Interface

Science. 53, 358–366.
Temp. (1C)

Krieger, I.M., Dougherty, T.J., 1959. A mechanism for non-Newtonian flow in

suspension of rigid spheres. Transactions of the Society of Rheology 3,
137–152.
225

140
200

25
25

Leal, L.G., Hinch, E.J., 1971. The effect of weak Brownian rotation on particles in
–
–

–
–

shear flow. Journal of Fluid Mechanics 46, 685–703.

Lewis, J.A., 2000. Colloidal processing of ceramics. Journal of the American Ceramic
Silicon powder-silicon carbide/a735

Society 83, 2341–2359.

Average PS (particle size) and PSD (particle size distribution).

Liu, D.M., 2000. Particle packing and rheological property of highly-concentrated

ceramic suspensions: jm determination and viscosity prediction. Journal of
Material Science 35, 5503–5507.
Silicon nitride/stearic acid

Manley St., R., Mason, S. G., J., 1945. The viscosity of suspensions of spheres: a note
Carbides (BI)/org sol.c

on the particle interaction coefficient. Canadian Journal of Chemistry 32,

Zirconia-yttria/water
Suspension system

(poluethylene wax)

Glass beads/lithene

763–767.
Alumina/org. sol.b

Mewis, J., 1979. Thixotropy—a general review. Journal of Non-Newtonian Fluid

Alumina/xylene
Alumina/water

Silica/glycerol

Mechanics 6, 1–20.
Silica/ethanol

Coal/glycerin

Mewis, J., Wagner, N.J., 2009. Current trends in suspension rheology. Journal of
Non-Newtonian Fluid Mechanics 157, 147–150.
Mooney, M., 1951. The viscosity of concentrated suspension of spherical particles.
Journal of Colloid Science 6, 162–170.
Moore, F., 1959. The rheology of ceramic slips and bodies. Transactions of the
British Ceramic Society 58, 470–494.
Moreland, C., 1963. Viscosity of suspension of coal in mineral oil. Canadian Journal
Stedman and Evans (1990)

Tsao and Danforth (1993)

of Chemical Engineering 41, 24–28.

Zhang and Evans (1989)

Dabak and Yucel (1986)

Mueller, S., Llewellin, E.W., Mader, H.M., 2010. The rheology of suspensions of
Wright et al. (1990)

solid particles. Proceedings of the Royal Society, A: Mathematica, Physical and

Honek et al. (2005)

Ewais et al. (2002)

Engineering Science 466, 1201–1228.

Kao et al. (1975)

Mujumdar, A., Beris, A., Metzner, A., 2001. Transient phenomena in thixotropic
systems. Journal of Non-Newtonian Fluid Mechanics 2072, 1–22.
Oliver, D.R., Ward, S.G., 1953. Relationship between relative viscosity and volume
concentration of stable suspensions of spherical particles. Nature 171,
Table 5

396–397.
Ref.

b
a

Pabst, W., 2004. Fundamentals considerations on suspension rheology. Journal

Ceramics-Silikaty 48, 6–13.
2806 B.A. Horri et al. / Chemical Engineering Science 66 (2011) 2798–2806
Pbast, W., Gregorova, E., Berthold, C., 2006. Particle shape and suspension rheology
of short-fiber systems. Journal of the European Ceramic Society 26, 149–160.
Powell, R.L., 1991. Rheology of suspensions of rodlike particles. Journal of
Statistical Physics 62, 1073–1094.
Quemada, D., 1977. Rheology of concentrated disperse systems and minimum
energy disspiation principle. Rheological Acta 16, 82–91.
Roscoe, R., 1952. The viscosity of suspensions of rigid spheres. British Journal of
Applied Physics 3, 267–269.
Roussel, N., Roy, R., Le, Coussot, P., 2004. Thixotropy modelling at local and
macroscopic scales. Journal of Non-Newtonian Fluid Mechanics 117, 85–95.
Rutgers, Ir.R., 1962. Relative viscosity and concentration. Rheological Acta 2,
305–348.
Saraf, D.N., Khullar, S.D., 1975. Some studies on the viscosity of settling suspen-
sions. Canadian Journal of Chemical Engineering 53, 449–452.
Senapati, P.K., Mishra, B.K., Parida, A., 2010. Modeling of viscosity for power plant
ash slurry at higher concentrations: effect of solids volume fraction, particle
size and hydrodynamic interactions. Powder Technology 197, 1–8.
Sherwood, J.D., 2006. Cell models for suspension viscosity. Chemical Engineering
Science 61, 6727–6731.
Stedman, S.J., Evans, J.R.G., 1990. Rheology of composite ceramic injection
moulding suspensions. Journal of Material Science 25, 1833–1841.
Stickel, J.J., Powell, R.L., 2005. Fluid mechanics and rheology of dense suspensions.
Annual Review of Fluid Mechanics 37, 129–149.
Takai, C., Tsukamoto, M., Fuji, M., Takahashi, M., 2006. Control of high solid
content yttria slurry with low viscosity for gelcasting. Journal of Alloys and
Compounds 408–412, 533–537.
View publication stats
Thomas, C.U., Muthukumar, M., 1991. Three-body hydrodynamic effects on viscosity
of suspension of spheres. The Journal of Chemical Physics 94, 5180–5199.
Thomas, D.G., 1965. Transport characteristics of suspension, VIII: a note on the
viscosity of Newtonian suspensions of uniform spherical particles. Journal of
Colloid Science 20, 267–277.
Tsao, I., Danforth, S.C., 1993. Rheological behavior of injection-mouldable silicon
powder-silicon carbide whisker formulations. Journal of the American Ceramic
Society 76, 2977–2984.
Usui, H., Kishimoto, K., Suzuki, H., 2001. Non-Newtonian viscosity of dense slurries
prepared by spherical particles. Chemical Engineering Science 56, 2979–2989.
Vand, V., 1948. Viscosity of solutions and suspensions, I: theory. Journal of
Physical Chemistry 52, 277–299.
Wright, J.K., Edirisinghe, M.J., Zhang, J.G., Evans, J.R.G., 1990. Particle packing in
ceramic injection molding. Journal of the American Ceramic Society 73,
2653–2658.
Yar, Y., Acar, F.Y., Yurtsever, E., Akinc, M., 2010. Reduction of viscosity of alumina
nanopowder aqueous suspensions by the addition of polyalcohols and sac-
charides. Journal of the American Ceramic Society 93, 2630–2636.
Yuanling, S., Xiaolin, L., Jianfeng, C., 2004. The maximum solid loading and
viscosity estimation of ultrafine BaTiO3 aqueous suspension. Colloids and
Surfaces A: Physicochemical and Engineering Aspects 247, 27–34.
Zhang, T., Evans, J.R.G., 1989. Predicting the viscosity of ceramic injection
moulding suspension. Journal of the European Ceramic Society 5, 165–172.
Zhou, Z., Scales, P.J., Boger, D.V., 2001. Chemical and Physical control of the
rheology of concentrated metal oxide suspension. Chemical Engineering
Science 56, 2901–2920.