Fluid Phase Equilibria 210 (2003) 319–334

An equation-of-state-based viscosity model for

non-ideal liquid mixtures夽
Ricardo Macı́as-Salinas a,∗ , Fernando Garcı́a-Sánchez b ,
Gaudencio Eliosa-Jiménez b
a
ESIQIE-Laboratorio de Termodinámica, Instituto Politécnico Nacional, UPALM—Zacatenco,
07738 México D.F., Mexico
b
Laboratorio de Termodinámica, Programa de Ingenierı́a Molecular, Instituto Mexicano del Petróleo,
Eje Central Lázaro Cárdenas 152, 07730 México D.F., Mexico
Received 4 November 2002; accepted 6 November 2002

Abstract
A viscosity model based on the Eyring’s theory and a cubic equation of state (Peng–Robinson–Stryjek–Vera) has
been applied to the correlation and prediction of experimental liquid viscosities of binary mixtures containing polar
fluids within a wide range of temperature, pressure and composition (encompassing low-pressure and compressed
liquid conditions). Highly non-idealities of the binary mixtures considered in this study were conveniently handled
via the application of the Wong–Sandler approach for the mixing rules used in the cubic equation of state. The
results obtained were highly satisfactory for various non-ideal binary mixtures over the whole composition range at
a low pressure. The predictive capabilities of the present approach were also verified in the representation of liquid
viscosities at elevated pressures preserving the same model parameters previously obtained at low pressure.
© 2003 Elsevier Science B.V. All rights reserved.
Keywords: Liquid viscosities; Viscosity models; Equation of state; Liquid mixtures

1. Introduction

A rational design or analysis of a chemical process greatly depends on the accurate representation of
the thermophysical properties of the process streams. Among these properties, liquid viscosity emerges
as one of the key transport variables needed in process design and development. Numerous attempts
have appeared in the literature dealing with the development of reliable methods for estimating liquid
viscosity of binary and multicomponent mixtures. Despite all these efforts, a theoretical description of
the viscosity of liquid mixtures nowadays remains insufficient due perhaps to the poor understanding
夽
Paper presented at the 5th International Symposium of the ESIQIE, May 29–31, 2002, Mexico City, Mexico.
∗
Corresponding author.
E-mail address: [email protected] (R. Macı́as-Salinas).

0378-3812/03/$ – see front matter © 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0378-3812(03)00169-9
320 R. Macı́as-Salinas et al. / Fluid Phase Equilibria 210 (2003) 319–334

of the liquid state itself. Consequently, most of the estimation methods for liquid viscosity appearing in
the literature are essentially empirical or semi-empirical in nature. Excellent reviews and evaluations of
the most important viscosity models so far reported for non-electrolyte pure liquids and liquid mixtures
are given by Reid et al. [1], Mehrotra [2], Monnery et al. [3], Mehrotra et al. [4] and Poling et al.
[5]. For liquid mixtures, the majority of the viscosity correlations published in the literature assume
that values of the viscosities of the pure components are available thus making the estimation method
interpolative. Other models for liquid mixture viscosity originally developed under the one-fluid basis
are extensions to multicomponent systems through the use of suitable mixing rules. As noted by Poling
et al. [5], basically all viscosity models for liquid mixtures refer to solutions of liquids below or slightly
above their normal boiling points, for example, they are restricted to reduced temperatures of the pure
fluids below about 0.7. At these temperature conditions, liquid viscosities are significantly sensitive to the
structure of the constituent molecules. For an ideal liquid mixture, there is a nearly linear behavior between
viscosity and composition. For a mixture containing alcohols and/or water, however, its corresponding
viscosity–composition function often exhibits a maximum or a minimum and sometimes both [6].
A widely used and recommended liquid mixture viscosity correlation is the Grunberg–Nissan approach
[5,6]. The calculation of the low-temperature liquid viscosity for a multicomponent system using this
approach is as follows:

c c c
1


ln ηm = xi ln ηi + xi xj Gi,j (1)
i=1
2 i=1 j=i

where Gi,i = 0. For example, for a binary mixture

ln ηm = x1 ln η1 + x2 ln η2 + x1 x2 G1,2 (2)
In the above equations, x is the liquid mole fraction and Gi,j is a binary interaction parameter which is a
function of components i and j as well as the temperature, and, in some instances, the composition. Using
regressed values of Gi,j from experimental data, Isdale [7] successfully applied Eq. (1) in the representa-
tion of liquid viscosities of non-associated mixtures and diverse mixtures containing alcohols, carboxylic
acids and ketones. Unfortunately, the Grunberg–Nissan method does not cover aqueous mixtures.
Among other approaches, the corresponding-states principle and the Eyring’s absolute rate theory have
been also extensively used to represent the viscosity of liquid mixtures. The TRAPP model developed by
Ely and Hanley [8] is so far the most popular corresponding-states viscosity model for pure liquids and
their mixtures. The TRAPP model utilizes methane as a reference fluid, however, it is only applicable to
mixtures containing non-polar fluids. On the basis of the corresponding-states principle, Teja and Rice
[9] proposed the following liquid mixture viscosity model
ωm − ω(R1)
ln(ηm εm ) = ln(ηε)(R1) + [ln(ηε)(R2) − ln(ηε)(R1) ] (3)
ω(R2) − ω(R1)
where the superscripts (R1) and (R2) stand for the reference fluids, ω is the acentric factor and the
parameter ε is defined by
2/3
Vc
ε= (4)
(Tc M)1/2
Appropriate rules for the mixture parameters ωm , Vcm , Tcm and Mm were suggested by the authors.
A temperature-independent interaction parameter ψi,j is introduced in the calculation of Tm . The value
 R. Macı́as-Salinas et al. / Fluid Phase Equilibria 210 (2003) 319–334 321

of ψi,j is of order unity which must be found from experimental data. Unlike the Grunberg–Nissan
correlation, the method of Teja and Rice gives significantly improved results for polar–polar mixtures,
particularly for aqueous solutions.
Huber and Ely [10] developed a predictive corresponding-states model for the viscosity of pure re-
frigerants and refrigerant mixtures. Their viscosity model incorporates R134a as the reference fluid and
uses equilibrium shape factors and an equivalent substance reducing ratio. Their model predictions were
considered adequate for several pure fluids and a single refrigerant mixture. Lee and Wei [11] proposed a
three-reference-fluid corresponding-states viscosity model for both non-aqueous and aqueous solutions.
The Lee–Wei approach incorporates oxygen, octane and water as the reference fluids. A set of quadratic
mixing rules was utilized by the authors in the representation of liquid viscosities for non-aqueous sys-
tems whereas for aqueous solutions the authors adopted an augmented asymmetric mixing rule for the
parameter Tcm . The authors claim satisfactory results for various binary and ternary mixtures with total
absolute average deviations (A.A.D.s) between observed and calculated viscosity data of 3.96% for 52
non-aqueous solutions and 5.23% for 16 aqueous solutions. In comparison with the Teja–Rice approach,
both methods give comparable results for binary aqueous solutions.
Recently, Cristofoli et al. [12] and Scalabrin et al. [13] presented a three-parameter corresponding-states
model for the viscosity prediction of pure hydrocarbons, halocarbons and their mixtures. Besides Tc and
Pc , the authors incorporated a new scaling factor (replacing the Pitzer’s acentric factor) into the model.
The application of their model yielded satisfactory results in the prediction of liquid viscosities of various
hydrocarbon and refrigerant mixtures. For example, they obtained an overall A.A.D. value of 3.63%
based on seven binary hydrocarbons mixtures, 10 binary refrigerant mixtures and two ternary refrigerant
mixtures [13].
Chevalier et al. [14] and Gaston-Bonhomme et al. [15] adapted and incorporated the UNIFAC activity
coefficient method into the framework of the Eyring’s absolute rate theory to predict liquid mixture
viscosities as follows:

GEX,c GEX,r
ln ηm = xi ln(ηi Vi ) − ln Vm + + (5)
i
RT RT

where the combinatorial term GEX,c /RT is the same as in the UNIQUAC model whereas the resid-
ual term GEX,r /RT is computed using the UNIFAC method as modified by the authors. The method,
termed UNIFAC-VISCO, was successfully applied in the prediction of viscosities of mixtures containing
molecules that differ appreciably in size. For alkanes, the UNIFAC-VISCO method yielded A.A.D. val-
ues of 2.6% for 13 ternary systems and 3.6% for 4 quaternary systems. Based on the Eyring’s absolute
rate theory, Cao et al. [16] also developed a UNIFAC-based method for liquid mixture viscosities. Their
resulting group-contribution viscosity model (GC-UNIMOD) was tested against many types of systems
yielding an average mean relative standard deviation M.R.S.D. of 4.1%.
Other modeling ideas based on Eq. (5) have been recently explored by various investigators [17–20].
For example, rather than using an activity coefficient model to calculate the excess activation free energy
of flow (last two terms of Eq. (5)), Lee et al. [17,18] used the Patel–Teja equation of state not only to
calculate this quantity but also the molar volume of the mixture Vm needed in the above expression.
They successfully validated their approach in the representation of liquid viscosities of both binary
and ternary mixtures at atmospheric and elevated pressures. One shortcoming of their model is the
non-symmetrical mixing rule used for multicomponent aqueous mixtures requiring two parameters per
322 R. Macı́as-Salinas et al. / Fluid Phase Equilibria 210 (2003) 319–334

binary. More recently, on the basis of Eq. (5), Martins and Cardoso [19] and Martins et al. [20] presented
a viscosity model in which the excess activation free energy of flow was calculated via the use of the
UNIQUAC model. The overall performance of their viscosity model was good with M.R.S.D. values of
1.2% for 352 binary mixtures and 2.95% for 51 multicomponents (48 ternary and 3 quaternary mixtures)
at atmospheric pressure. As compared with the GC-UNIMOD approach, the model proposed by Martins
et al. [20] yielded improved results for various types of binary mixtures.
In an attempt to obtain highly accurate modeling of viscosity of pure fluids and mixtures from low
to elevated pressures, a novel theory was recently proposed by Quiñones-Cisneros et al. [21,22]. This
new theory, named as friction theory (f-theory) by the authors, is based on friction concepts of classical
mechanics and the van der Waals theory of fluids. The capabilities of the f-theory model were verified
using two well known cubic equations of state to reproduce the viscosity of pure n-alkanes from methane
to n-decane with excellent results. The authors also extended the application of the f-theory viscosity
model to n-alkane mixtures by using simple mixing rules for the friction parameters. They obtained
satisfactory results in the prediction of viscosities of several binary to quaternary n-alkane mixtures with
A.A.D. values varying from 0.51 to 7.4% using the five-parameter f-theory model [21], and from 1.23 to
9.67% using the one-parameter f-theory model [22].
A one-fluid viscosity model for saturated liquids developed by Macı́as-Salinas et al. [23] was extended
to the representation of experimental liquid viscosities of binary and ternary mixtures containing normal
paraffins within a wide range of temperature, pressure and composition [24]. The model is based on the
Eyring’s theory and a cubic equation of state, and uses a single binary interaction parameter to adequately
describe the whole η–T–P–x surface of interest. For the majority of the binary mixtures considered by the
authors, the performance of their model was highly satisfactory in the representation of observed liquid
viscosities with an overall A.A.D. value of 3.23% based on 1193 experimental points.
Based on the capabilities and limitations of the liquid viscosity models described above, in this paper,
we propose an Eyring-theory-based model coupled with a cubic equation of state to properly correlate
and/or predict liquid viscosity of highly non-ideal binary solutions (both non-aqueous and aqueous) over
a wide range of temperature, pressure and composition. As an important feature, the proposed model
makes use of the Wong–Sandler (WS) mixing rules in the cubic equation of state to calculate the excess
activation free energy and molar volumes needed within the Eyring’s framework. Further, the predictive
capabilities of the present approach were verified in the representation of liquid mixture viscosities at
pressures much higher than 0.1 MPa.

2. Framework of the model

The model is based on the following Eyring’s viscosity expression for liquid mixtures
 =,EX 
(ηV)ID G
ηm = exp (6)
Vm RT
where (ηV)ID is the kinematic viscosity of an ideal solution given by
 c 


(ηV)ID = exp xi ln(η0i Vi0 ) (7)
i=1
 R. Macı́as-Salinas et al. / Fluid Phase Equilibria 210 (2003) 319–334 323

The excess activation free energy of flow G=,EX in Eq. (6) depends on temperature, pressure and
composition. Wei and Rowley [25] related this quantity to the excess Gibbs free energy of the mixture
GEX through
G=,EX = −σ GEX (8)
where σ is a proportionality factor. Rather than using the above expression, the value of G=,EX can be
directly calculated from an equation of state as follows:
c

 
G=,EX = GEX,EoS = RT xi ln φi − ln φi0 (9)
i=1

Within this framework, the use of an equation of state is advantageous since it allows the simultaneous
determination of the fugacity coefficient of i in the mixture φi (x, T, P), the molar volume of the liquid
mixture Vm (x, T, P), the fugacity coefficient of the pure fluid φi0 (T, P) and the molar volume of the pure
liquid Vi0 (T, P). In the present work, the calculation of these properties was performed via the use of the
one-fluid Peng–Robinson–Stryjek–Vera (PRSV) [26,27] cubic equation of state CEoS
RT a(T)
P= − (10)
V − b V(V + b) + b(V − b)
with
 √
2
R2 Tc2 α(T) = 1 + κ(1 − Tr )
a(T) = 0.457235 α(T), √ (11)
Pc κ = κ0 + κ1 (1 + Tr )(0.7 − Tr )
RTc
b = 0.077796 (12)
Pc
where κ0 is a function of the acentric factor and the constant κ1 is specific for each pure component and
was used to fit low-temperature saturation pressures accurately. This version of the Peng–Robinson CEoS
was chosen since it provides accurate vapor pressure predictions, particularly at lower temperatures and
for non-hydrocarbon fluids. The application of Eq. (10) to highly non-ideal mixtures should be done by
incorporating a suitable compositional dependence of the a and b parameters into the CEoS. The mixing
rules proposed by Wong and Sandler [28] that combine excess free-energy models with a CEoS were
used for this purpose
D
am = RTQ (13a)
1−D
Q
bm = (13b)
1−D
with


 a 
Q= xi xj b − (14)
i j
RT i,j

ai GEX
γ
D= xi + (15)
bi RT CRT
324 R. Macı́as-Salinas et al. / Fluid Phase Equilibria 210 (2003) 319–334

and the combining rule

 a  1  ai   aj 
b− = bi − + bj − (1 − ki,j ) (16)
RT i,j 2 RT RT
The above mixing rules satisfy the low-density boundary condition: quadratic composition dependence
EoS = Gγ at infinite pressure
of the second virial coefficient (Eq. (14)) and the high-density condition: AEX EX

(Eq. (15)). The WS approach thus produces the desired equation of state behavior at both low and high
densities without being density dependent and allows extrapolation over wide ranges of temperature and
pressure. The simple van Laar equation was incorporated into Eq. (15) as the excess Gibbs free energy
model needed in the WS mixing rules. For a binary mixture, the van Laar GEX γ model is:

GEX
γ A1,2 x1 x2
= (17)
RT x1 A1,2 /A2,1 + x2
The binary interaction parameters A1,2 and A2,1 in the above expression were adjusted to fit experimental
liquid mixture viscosity data using Eqs. (6) and (9). Also, note that the interaction parameter for the second
virial coefficient k1,2 = k2,1 was set equal to zero in Eq. (16). The expressions for the fugacity coefficients
based on the WS approach have been already given elsewhere [28,29].
The final form of the viscosity model is thus obtained by combining Eqs. (6) and (9):
 c 
(ηV)ID

η =
cal
exp xi (ln φi − ln φi )
0
(18)
Vm i=1

The calculation of the kinematic viscosity of an ideal solution (ηV)ID , Eq. (7), requires the values of
the viscosities of the constituent pure components η0i at the temperature and pressure of interest. For all
compounds considered in this work, their experimental viscosities were available and taken from the data
sources listed in Table 1.

3. Results and discussion

Eq. (18) was first applied to the correlation of experimental liquid viscosities of binary mixtures
including non-aqueous and aqueous solutions. Table 1 shows the binary mixtures considered in this
study along with their corresponding temperature and pressure ranges, and the source of the experimental

Table 1
Experimental data
Mixture Temperature (K) Pressure (MPa) Source

Cyclohexane/ethanol 298.15 0.1 Papaloannou et al. [30]

Methanol/ethanol 340 0.1 Stephan and Hildwein [31]
Acetone/cyclohexane 298.15 0.1 Petrino et al. [32]
Acetone/toluene 298.15 0.1 Petrino et al. [32]
Water/methanol 285–330 0.1–65 Stephan and Hildwein [31]
Water/ethanol 275–350 0.1–45 Stephan and Hildwein [31]
Water/1-propanol 295–320 0.1–49 Stephan and Hildwein [31]
 R. Macı́as-Salinas et al. / Fluid Phase Equilibria 210 (2003) 319–334 325

Table 2
Optimized parameters at 0.1 MPa, non-aqueous systems
Mixture N T (K) A1,2 A2,1 A.A.D. η (%)

Cyclohexane/ethanol 10 298.15 −0.4079 −1.4835 0.974

Methanol/ethanol 9 340 0.3267 0.4093 0.007
Acetone/cyclohexane 10 298.15 −1.1860 −0.6697 0.871
Acetone/toluene 5 298.15 0.3302 0.4911 0.125
Overall 34 0.563

viscosity data. A non-linear regression analysis of the data was performed to obtain the mixture parameters
A1,2 and A2,1 at each temperature and 0.1 MPa. The minimization of the following objective function served
for this purpose
N
 2

ηcal
j
min f = 1 − expt (19)
j=1
ηj

where the summation is over the whole composition range; N is the number of experimental points, ηexpt
and ηcal are experimental and calculated viscosities of the binary mixture, respectively. The correlating
results obtained at low pressure (0.1 MPa) are summarized in Tables 2 and 3 for four non-aqueous systems
and three aqueous systems, respectively. These results include the correlating temperature, the regressed
van Laar interaction parameters A1,2 and A2,1 , and the resulting percent of A.A.D.s between calculated
and observed viscosity data defined by
 
100
 
N
ηcal
j 
A.A.D. η =  1 − expt  (20)
N j=1  ηj 

Table 3
Optimized parameters at 0.1 MPa, aqueous systems
Mixture N T (K) A1,2 A2,1 A.A.D. η [%]
Water/methanol 36 285 2.6974 10.6466 0.598
300 2.5899 8.3518 0.671
315 2.5031 6.6189 0.989
330 2.4529 5.2366 1.567
Water/ethanol 45 275 3.2546 37.3072 1.795
285 3.0490 28.1108 1.269
300 2.7755 20.3027 0.812
330 2.4301 15.4644 0.370
350 2.0941 11.0591 0.304
Water/1-propanol 36 295 3.0029 28.0747 0.521
300 2.9410 25.5320 0.660
310 2.8147 21.9733 0.843
320 2.6866 19.7062 0.869
Overall 117 0.867
326 R. Macı́as-Salinas et al. / Fluid Phase Equilibria 210 (2003) 319–334

Fig. 1. Representation of observed liquid viscosities at 0.1 MPa, non-aqueous systems.

3.1. Non-aqueous systems

It can be seen from Table 2 that the ability of the present approach in reproducing the experimental
viscosity data is excellent with low A.A.D. values for the four systems: cyclohexane/ethanol at 298.15 K,
methanol/ethanol at 340 K, acetone/cyclohexane at 298.15 K and acetone/toluene at 298.15 K. In fact,
the overall performance of the model gives an A.A.D. value of only 0.563% based on 34 experimental
points. Fig. 1 graphically depicts the comparison between calculated and observed viscosity data for
the four systems over the whole composition range. As shown by Fig. 1, the mixtures methanol/ethanol
and acetone/toluene both exhibit ideal solution behavior as evidenced by the nearly linear functionality
between viscosity and composition. The curvature of the viscosity–composition function shown by the
other two mixtures (cyclohexane/ethanol and acetone/cyclohexane) reveals a significant departure from
ideal behavior; in fact, the cyclohexane/ethanol mixture exhibits a minimum. In all cases, the model does
an excellent job in correlating the experimental viscosity data, including the highly non-ideal viscosity
behavior displayed by the cyclohexane/ethanol system.

3.2. Aqueous systems

The correlative capabilities of the model were also tested against three aqueous solutions (water/
methanol, water/ethanol and water/1-propanol) at different temperatures and at 0.1 MPa. Unlike most
non-aqueous solutions, aqueous systems often exhibit a maximum as well as a strongly asymmetric
 R. Macı́as-Salinas et al. / Fluid Phase Equilibria 210 (2003) 319–334 327

Fig. 2. Representation of observed liquid viscosities at 300 K and 0.1 MPa, aqueous systems.

behavior in their viscosity isotherms with respect to composition. This peculiar behavior is mainly at-
tributed to hydrogen-bond formation. As shown in Table 3, the performance of Eq. (19) in reproducing
liquid viscosities of aqueous solutions was also remarkably good with A.A.D. values between 0.304 and
1.795%. These values fall within the experimental uncertainty reported by Stephan and Hildwein [31]. The
overall performance of the model yielded an A.A.D. value of 0.867% based on 117 experimental points.
In order to better illustrate the goodness-of-fit, in Fig. 2 we have plotted the calculated and experimental
values of the mixture viscosity for the three water/alcohol systems at 300 K and at 0.1 MPa. In all cases,
the present approach provides an excellent fit to experimental results over the entire range of composition.
More importantly, the model is able to accurately reproduce the maximum and the highly asymmetric
behavior of viscosity exhibited by the three aqueous systems. Similar plots were prepared for each wa-
ter/alcohol system at different temperatures and at 0.1 MPa. Accordingly, Figs. 3–5 graphically show
the correlating results obtained for the mixtures water/methanol, water/ethanol and water/1-propanol, re-
spectively. For each mixture, as the temperature increases, the experimental viscosity isotherms broaden
with approximately the same degree of asymmetry indicating that non-ideal viscosity behavior is more
significant at lower temperatures. Again, as evidenced by Figs. 3–5, the model accurately captures this
typical behavior for the three mixtures at all temperatures, except for the water/ethanol mixture (Fig. 4)
for which the model underpredicts the maximum point at the lowest temperature (275 K) and at higher
concentrations of water (a mole fraction of about 0.8). However, for such a non-ideal aqueous system,
the general fit should be considered good.
328 R. Macı́as-Salinas et al. / Fluid Phase Equilibria 210 (2003) 319–334

Fig. 3. Reproducibility of observed liquid viscosities at 0.1 MPa and at different temperatures, water/methanol system.

There is a correlation between the numerical values of the model parameters A1,2 and A2,1 , and the
degree of non-ideality of the fitted viscosity data. If the value of A2,1 /A1,2 significantly departs from unity,
then the mixture is expect to exhibit a highly non-ideal viscosity behavior as demonstrated for the case
of water/ethanol at 275 K (Table 3). For a nearly ideal viscosity behavior, the value of A2,1 /A1,2 should
approach unity as in the case of the methanol/ethanol system at 340 K (Table 2).

3.3. Prediction of viscosity data at elevated pressures

An attractive feature of the WS approach is that it permits to perform extrapolations of vapor–liquid

phase behavior to higher temperatures and pressures with good accuracy. In an attempt to verify the
predictive capabilities of the WS approach within the present framework, Eq. (18) was applied to predict
the viscosity behavior of non-ideal mixtures at higher pressures using the same van Laar parameters A1,2
and A2,1 previously obtained at low pressure. Extrapolation to higher temperatures was not possible since
the model parameters were found to be temperature-dependent. Table 4 gives the results of the predictive
capabilities of the model in terms of A.A.D. values between extrapolated and experimental viscosity
data for the three aqueous systems originally considered in this study over a wide pressure range from
10 to 60 MPa and at different temperatures. As shown in Table 4, the model predicts liquid viscosities
of aqueous systems at pressures much higher than 0.1 MPa with A.A.D. values varying from 0.64 to
7.229%. The overall predictive performance of the model should be regarded as highly satisfactory yield-
ing an A.A.D. value of 2.969% based on 381 data points. Figs. 6–8 give a graphical comparison between
 R. Macı́as-Salinas et al. / Fluid Phase Equilibria 210 (2003) 319–334 329

Fig. 4. Reproducibility of observed liquid viscosities at 0.1 MPa and at different temperatures, water/ethanol system.

observed and predicted viscosity data for three aqueous systems at constant temperature. For the mixtures
water/methanol at 285 K (Fig. 6) and water/ethanol at 350 K (Fig. 7) the model moderately underpredicts
and/or overpredicts the experimental viscosity isobars over the entire composition range. In both cases,
however, the correct non-ideal viscosity behavior is predicted by the model (at least qualitatively) at
higher pressures. For the case of the water/1-propanol system at 295 K (Fig. 8) model predictions are

Table 4
Predicted results for aqueous systems at elevated pressures
Mixture N T (K) P range (MPa) A.A.D. η (%)

Water/methanol 216 285 10–60 1.782

300 2.144
315 3.149
330 3.862
Water/ethanol 135 300 10–45 2.327
330 7.229
350 1.911
Water/1-propanol 30 295 10–49 0.640
320 0.996
Overall 381 2.969
Fig. 5. Reproducibility of observed liquid viscosities at 0.1 MPa and at different temperatures, water/1-propanol system.

Fig. 6. Prediction of liquid viscosities at 285 K and at high pressures, water/methanol system.
 Fig. 7. Prediction of liquid viscosities at 350 K and at high pressures, water/ethanol system.

Fig. 8. Prediction of liquid viscosities at 295 K and at high pressures, water/1-propanol system.
332 R. Macı́as-Salinas et al. / Fluid Phase Equilibria 210 (2003) 319–334

exceptionally good, thus confirming the suitability of the present approach in predicting compressed
liquid viscosities at different temperatures and within a wide pressure range.

4. Conclusions

On the basis of the Eyring theory for liquid mixtures, a new viscosity model was developed for the
calculation of liquid mixtures viscosities of highly non-ideal solutions. The model adequately describes
the viscosity–composition function in terms of equilibrium properties easily determined via the use of
a well known cubic equation of state coupled with WS mixing rules. The following conclusions can be
drawn from this work:
• For all the binary mixtures considered in this study, the reproducibility of low-pressure experimental liq-
uid viscosities was excellent using the present approach with A.A.D. values of 0.56% for non-aqueous
solutions and 0.87% for aqueous solutions.
• The model parameters A1,2 and A2,1 were found to be temperature-dependent. For systems exhibiting
highly non-ideal viscosity behavior, the value of A2,1 /A1,2 significantly departed from unity. For a nearly
ideal viscosity behavior: A2,1 /A1,2 ≈ 1.
• The predictive capabilities of the proposed approach were satisfactorily verified in the representation
of observed liquid viscosities of aqueous systems at elevated pressures, from 10 to 60 MPa.
• Based on the present results, the use of a cubic equation of state with WS mixing rules proved to
be highly promising within the Eyring’s framework to represent liquid viscosity of highly non-ideal
mixtures over a wide range of pressure and temperature, and over the entire composition range.

List of symbols
a attractive parameter in the PRSV CEoS
A Helmholtz free energy, van Laar interaction parameter
b co-volume parameter in the PRSV CEoS
c number of components
D as defined in Eq. (15)
f objective function
G Gibbs free energy
k interaction parameter in the PRSV CEoS
M molecular weight
N number of data points
P pressure
Q as defined in Eq. (14)
R ideal gas constant
T temperature
V molar volume
x liquid mole fraction

Greek letters
α alpha function in the PRSV CEoS
ε as defined in Eq. (4)
 R. Macı́as-Salinas et al. / Fluid Phase Equilibria 210 (2003) 319–334 333

φ fugacity coefficient
η dynamic viscosity
κ0 , κ1 parameters in the PRSV CEoS
σ proportionality factor
ω acentric factor

Subscripts
c critical property
EoS equation of state
i, j i–j pair interaction
ID ideal solution
m mixture property
r reduced property
γ activity-coefficient based model

Superscripts
c combinatorial contribution
cal calculated value
expt experimental value
EX excess property
r residual contribution
0 pure substance
= activation state

Acknowledgements

One of the authors (R. M.-S.) gratefully acknowledges to the Instituto Mexicano del Petróleo and
Instituto Politécnico Nacional for providing financial support for this work.

References

[1] R.C. Reid, J.M. Prausnitz, B.E. Poling, The Properties of Gases and Liquids, fourth ed., McGraw-Hill, New York, 1987.
[2] A.K. Mehrotra, Ind. Eng. Chem. Res. 30 (1991) 420–427;
A.K. Mehrotra, Ind. Eng. Chem. Res. 30 (1991) 1367–1372.
[3] W.D. Monnery, W.Y. Svrcek, A.K. Mehrotra, Can. J. Chem. Eng. 73 (1995) 3–40.
[4] A.K. Mehrotra, W.D. Monnery, W.Y. Svrcek, Fluid Phase Equilib. 117 (1996) 344–355.
[5] B.E. Poling, J.M. Prausnitz, J.P. O’Connell, The Properties of Gases and Liquids, fifth ed., McGraw-Hill, New York, 2001.
[6] J.B. Irving, Viscosities of Binary Liquid Mixtures: The Effectiveness of Mixture Equations, Natl. Eng. Lab., Rept. 63, East
Kilbride, Glasgow, Scotland, 1977.
[7] J.D. Isdale, in: Proceedings of Symposium on Transp. Prop. Fluids and Fluid Mixtures, Natl. Eng. Lab., East Kilbride,
Glasgow, Scotland, 1979.
[8] J.F. Ely, H.J.M. Hanley, Ind. Eng. Chem. Fundam. 20 (1981) 323–332.
[9] A.S. Teja, P. Rice, Ind. Eng. Chem. Fundam. 20 (1981) 77–81.
[10] M.L. Huber, J.F. Ely, Fluid Phase Equilib. 80 (1992) 239–248.
[11] M.-J. Lee, M.-C. Wei, J. Chem. Eng. Japn. 26 (1993) 159–165.
334 R. Macı́as-Salinas et al. / Fluid Phase Equilibria 210 (2003) 319–334

[12] G. Cristofoli, M. Grigiante, G. Scalabrin, High Temp. High Press. 33 (2001) 83–94.
[13] G. Scalabrin, G. Cristofoli, M. Grigiante, Int. J. Energy Res. 26 (2002) 1–26.
[14] J.L. Chevalier, P. Petrino, Y. Gaston-Bonhomme, Chem. Eng. Sci. 43 (1988) 1303–1309.
[15] Y. Gaston-Bonhomme, P. Petrino, J.L. Chevalier, Chem. Eng. Sci. 49 (1994) 1799–1806.
[16] W. Cao, K. Knudsen, A. Fredenslund, P. Rasmussen, Ind. Eng. Chem. Res. 32 (1993) 2087–2092.
[17] M.-J. Lee, T.-K. Lin, S.-M. Hwang, J. Chin. Inst. Chem. Eng. 29 (1998) 73–83.
[18] M.-J. Lee, J.-Y. Chiu, S.-M. Hwang, H.-M. Lin, Ind. Eng. Chem. Res. 38 (1999) 2867–2876.
[19] R.J. Martins, M.J.E.M. Cardoso, O.E. Barcia, Ind. Eng. Chem. Res. 39 (2000) 849–854.
[20] R.J. Martins, M.J.E.M. Cardoso, O.E. Barcia, Ind. Eng. Chem. Res. 40 (2001) 1271–1275.
[21] S.E. Quiñones-Cisneros, C.K. Zéberg-Mikkelsen, E.H. Stenby, Fluid Phase Equilib. 169 (2000) 249–276.
[22] S.E. Quiñones-Cisneros, C.K. Zéberg-Mikkelsen, E.H. Stenby, Fluid Phase Equilib. 178 (2001) 1–16.
[23] R. Macı́as-Salinas, F. Garcı́a-Sánchez, O. Hernández Garduza, AIChE J. 49 (2003) 799–804.
[24] R. Macı́as-Salinas, F. Garcı́a-Sánchez, G. Eliosa-Jiménez, in: Paper Presented at the 2001 AIChE Annual Meeting, Reno,
Nevada, 2001.
[25] I.C. Wei, R.L. Rowley, Chem. Eng. Sci. 40 (1985) 401–408.
[26] D.-Y. Peng, D.B. Robinson, Ind. Eng. Chem. Fundam. 15 (1976) 59–64.
[27] R. Stryjek, J.H. Vera, Can. J. Chem. Eng. 64 (1986) 323–333;
R. Stryjek, J.H. Vera, Can. J. Chem. Eng. 64 (1986) 820–826.
[28] D.S.H. Wong, S.I. Sandler, AIChE J. 38 (1992) 671–680.
[29] H. Orbey, S.I. Sandler, Modeling Vapor–Liquid Equilibria, Cambridge University Press, New York, 1998.
[30] D. Papaloannou, T. Evangelou, C. Panaylotou, J. Chem. Eng. Data 36 (1991) 43–46.
[31] K. Stephan, H. Hildwein, Recommended Data of Selected Compounds and Binary Mixtures, DECHEMA Chemistry Data
Series, 1987.
[32] P.J. Petrino, Y.H. Gaston-Bonhomme, J.L.E. Chevalier, J. Chem. Eng. Data 40 (1995) 136–140.