You are on page 1of 13

Fluid Phase Equilibria 376 (2014) 141–153

Contents lists available at ScienceDirect

Fluid Phase Equilibria


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

Cubic equation of state: Limit of expectations


Vitaly Abovsky ∗
63 Summit Street, Waltham, MA 02451, USA

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

Article history: The conventional cubic equation of state (EOS) is analyzed for the purpose of achieving maximum possible
Received 7 March 2014 precision of representation of thermodynamic functions and parameters of phase equilibrium for pure
Received in revised form 16 May 2014 components, as well as binary and multi-component mixtures. It is derived that the optimal version of
Accepted 22 May 2014
EOS for pure components would be 4-parameter equation with size and energy parameters as functions
Available online 2 June 2014
of temperature. The equation fits experimental data on vapor pressure, density and heat of vaporization
at the level of precision that is comparable with experimental accuracy, excluding the results at the
Keywords:
near-critical range.
Equation of state
Compressibility factor
One-fluid approximation for mixtures is analyzed on general basis of virial theorem. The correction
Phase equilibrium to conventional one-fluid approximation requires introduction of ternary parameters in the expression
Mixing rules for energy parameter of mixture, and transforms EOS for mixture from cubic to fifth degree with respect
to volume. It is shown both through formulaic derivation and by analysis of experimental data that
approaches which ignore this correction can never provide accurate calculations for multi-component
mixture on the same level as has been achieved for binary combinations of its constituent components.
© 2014 Elsevier B.V. All rights reserved.

1. Introduction This paper is an attempt to establish a cubic EOS that provides


maximum possible precision for calculating thermodynamic prop-
The undisputed popularity of the family of cubic EOS in chemical erties of fluids within the regular region of the phase diagram.
engineering is owed to their simplicity and surprising precision in
calculating vapor pressure of pure fluids as well as parameters of 2. Single component system
phase equilibrium of binary fluid mixtures.
A common problem with these calculations was large sys- 2.1. Multi-parameter cubic EOS
tematic deviations between experimental and calculated liquid
densities, and ubiquitously poor fit of thermodynamic properties in The feasibility of achievement maximum possible precision
the near-critical region of pure fluids. A radical solution of the prob- obviously has to be based on the maximum number of allowed
lem on the EOS basis was developed by Kiselev [1]. The proposed parameters. Any cubic EOS can contain three temperature depend-
“crossover” EOS connects the “scaling” expressions for thermody- ent parameters after transforming it to the cubic polynomial with
namic properties to a conventional cubic or non-cubic EOS. This respect to volume, and these parameters can be composed from
approach guarantees correct quantitative results in the close prox- more than three temperature functions. In spite of the above, most
imities of the critical point. Nevertheless, the quality of calculations of the existing proposed equations tend to exploit only one tem-
in the regular, or “classic”, region remains dependent on precision perature dependent parameter – the so called “˛-function” – that
of the conventional EOS. As can be seen from the results obtained by restricts the EOS flexibility.
Kiselev and Ely [2] the “crossover” approach based on generalized The most general expression for the cubic EOS can be written
version of Patel–Teja EOS [3] does not produce accurate quantita- as a sum of two terms that implement contributions from inter-
tive results for either liquid density of pure components or phase molecular repulsion and attraction respectively; each term can be
compositions of binary mixtures. presented in the form of Pade-approximant:

1 + b1 + b2 v/b0 A0 + A1 v + A2 v2
P = RT − (1)
v − b0 v2 + C v + D
∗ Tel.: +1 7817901279. Eq. (1) contains 8 parameters (maximum allowed in a cubic
E-mail address: vabovsky@gmail.com equation), and only 4 of them (b0 , A0 , C, D) can be temperature

http://dx.doi.org/10.1016/j.fluid.2014.05.026
0378-3812/© 2014 Elsevier B.V. All rights reserved.
142 V. Abovsky / Fluid Phase Equilibria 376 (2014) 141–153

If b2 =/ 0 then the free energy of ideal gas becomes singular.


List of symbols Eq. (1) can be rewritten now as

N Avogadro number 1 + b1 A0 + b1 v
P = RT − (4)
k Boltzmann constant v − b0 v2 + C v + D
R gas constant Eq. (4) represents the upper limit for the number of indepen-
P pressure dent parameters in the cubic EOS that is meaningful physically.
T temperature 5-Parameter EOS was proposed earlier by Jiuxun [4] in a differ-
v volume ent form – without the traditional assignment of the repulsive and
x composition attractive contributions to the specific terms of EOS. This discrep-
pc critical pressure ancy with the approach outlined here leads to a notable difference
Tc critical temperature in the final results: according to [4] parameter b1 can be temper-
zc calculated critical compressibility factor ature dependent that ignores the obvious restriction (2) above on
A parameter of attractive part of EOS the repulsive term of EOS.
b size parameter of repulsive part of EOS (covolume) The data regression analysis (it will be discussed in detail later
c size parameter of attractive part of EOS on) has shown that parameter b1 does not contribute much to
d “bridge” parameter of attractive part of EOS the precision of Eq. (4). The optimal values of this parameter for
lrij correction parameter in mixing rule for parameter the wide variety of fluids remain close to zero, but its inclusion
b increases the complexity of calculations a great deal. Therefore,
laij correction parameter in mixing rule for parameter c the rational approach would be to neglect this parameter, which
kij correction for parameter Aij results in the final EOS being a 4-parameter one:
kcij correction parameter in mixing rule for non-
conventional part of parameter Am RT A
P= − (5)
Ba contribution of intermolecular attractive forces to v−b v2 + C v + D
second virial coefficient The final goal of this current research is to obtain a reliable EOS
r intermolecular distance for multi-component fluid system. This implies well based mixing
g binary distribution function rules for all 4 parameters of Eq. (5). The more or less acceptable
h binary correlation function recommendations can be used for b and A – basing on accumulated
experience for widely used EOS of van-der-Waals type. However,
Greek letters establishing reliability of mixing rules for C and D is no less impor-
˝ dimensionless coefficient in EOS tant. The possible solution of the problem will be considered in
˛ temperature dependent part of parameter A the third part of the paper on the basis of physical meaning and
˚ pair potential function structure of parameters C and D that will be outlined below.
 reduced temperature
r reduced size parameter of ˛-function 2.2. Parameters C and D. Comparison with well known 3- and
εr reduced energy parameter of ˛-function 4-parameter EOS
 power index in the expression for temperature
dependent parameter b Eq. (5) was obtained by reducing the initial general expression
(1) on basis of three physical restrictions, and omitting the unnec-
Subscripts essary fifth parameter. The meaning of parameters C and D can be
R contribution of repulsive forces made clear by deriving Eq. (5) out of 2-parameter EOS, i.e., going in
a contribution of attractive forces the opposite direction compare to sequence Eqs. (1)–(5). Consider-
r reduced ation of the best in its class Peng–Robinson EOS (PR EOS)
i, j, l components
RT A
m mixture P= − (6)
c correction v−b v2 + bv + b(v − b)
and introduction of the third parameter c as shown below
RT A
P= − (7)
v−b v2 + c v + c(v − b)
dependent. Parameters b1 , A0 , A1 , A2 must be constants accordingly
leads to 3-parameter PR EOS (PR3 EOS):
to the following restrictions:
RT A
P= − (8)
v−b v2 + 2c v − cb
(1) Eq. (1) must satisfy the ideal gas limit (pv/RT = 1);
(2) Contribution of the repulsive term to the compressibility factor Parameter c has the dimension of volume, and contributes to
must be a function of v/b0 only if intermolecular repulsion is the attractive term of Eqs. (7) and (8) much more than covolume b
represented by hard bodies with covolume b0 ; which actually is the repulsive size parameter. Indeed, c(v − b)  cv
(3) Free energy expression based on Eq. (1) must converge at the in the liquid phase; otherwise, b  v in the vapor phase; so, the
ideal gas limit. overall influence of b on the value of the attractive term remains at
the magnitude of a small correction when compared to the contri-
bution of parameter c. Therefore parameter c can be interpreted
The above restrictions lead to the following: as the size parameter of the attractive intermolecular interac-
tion. This definition of parameter c is not applicable to Patel–Teja
b1 = A1 = const (2) EOS [3]
RT A
P= − (9)
b2 = A2 = 0 (3) v−b v2 + bv + c(v − b)
V. Abovsky / Fluid Phase Equilibria 376 (2014) 141–153 143

and to all EOS that can be reduced to it – for example, EOS proposed 0.350

Dimensionless parameter b
by Salim and Trebble [5].
d=2
Obviously, the contribution of parameter c to the attractive term 0.300
of Eq. (9) is much smaller than the contribution from parameter b in
the liquid phase, and has the same order of magnitude in the vapor 0.250
phase. So, the physical meaning of parameter c in the EOS of type
(9) is unclear, and that creates the uncertainty with the mixing rule 0.200 d=1

for this parameter.


A single step completes the transition from Eqs. (8)–(5) by 0.150
introducing the fourth parameter D in Eq. (5) as
0.100
D = −dcb (10) 0.220 0.240 0.260 0.280 0.300 0.320 0.340
Critical compressibility factor
The final form of Eq. (5) comes to
RT A Fig. 1. Function br (zc ) at the different values of parameter d.
P= − (11)
v−b v2 + 2c v − dcb
The dimensionless parameter d can be designated the “bridge value of br has to be contained in the narrow range (0.29–0.32) for
parameter” as varying its value transforms Eq. (11) to cubic equa- the most fluids to provide the appropriate fit of liquid density.
tions proposed earlier. Thus, d = 0 reduces Eq. (11) to 3-paramet Fig. 1 shows that Ji–Lempe EOS satisfies this requirement in zc
modification of Redlich–Kwong EOS [6], d = 1 leads to Eq. (8), d = 2 range (0.23–0.33) that corresponds to the wide set of components –
corresponds to Ji–Lempe EOS [7], and Schmidt and Wenzel EOS [8] from water to inert gases, and Eq. (8) – does not. Of course, there are
many other factors in the EOS that influence calculations of liquid
RT A
P= − (12) density (parameters A and c, requirement of precise simultaneous
v−b v2 + ubv − wb2 fit of vapor pressure and heat of vaporization, etc.) but parameter
can be rewritten in the form (11) by substituting b remains the main factor.
Substituting the solution of Eq. (16) into Eq. (15) allows one
2c −dc
u= , w= (13) to analyze the dependence Cr (zc ) at the same fixed values of d. In
b b actuality, it is the value C = 2c as the size parameter that determines
The advantage of the form (11) of 4-parameter EOS at compari- behavior of the attractive term. Fig. 2 gives comparison of the curves
son with Eq. (12) will be shown in the third part of the paper; it is Cr (zc ) at d = 1, and d = 2.
related to the way of deriving mixing rules for EOS parameters. It follows that Cr > br at the range of zc < 0.34, and this delta
increases substantially with decreasing zc , which means that Cr
2.3. Determination of parameters at the vapor–liquid critical becomes larger for components with polar molecules – in com-
point. parison to components with van-der-Waals intermolecular forces.
This clearly confirms the validity of interpretation of value 2c in Eq.
One can use the constraint method proposed by Martin and Hou (11) as the size parameter of the attractive term – independently
[9] to express the value of parameters Ac , bc , and cc at the critical of parameter d value. More precisely, this parameter can be inter-
point in terms of Tc , pc , zc and d. This leads to the system of three preted as the size parameter of multi-particle attractive interaction
equations for dimensionless values ˝A , ˝b , and ˝c : because it contributes to all senior virial coefficients but the sec-
ond one. It is easy to see that parameter c in Patel–Teja EOS (9)
Ac pc zc d ˝c d ˝b ˝c
˝A = = + 2 + (14) cannot be interpreted in the same way as above. Both parameters
(RTc )2 ˝b zc zc2 in Eq. (9) contribute significantly to the attractive term of EOS, and
cc pc ˝ +1 c value becomes less than b in zc range (0.29–0.32) that belongs
˝c = = b − 1.5 zc (15) to the components such as hydrogen, inert gases, air components,
RTc 2
light hydrocarbons, etc. Even more importantly, substantial dif-
ference in c values for different components implies that strict
˝b3 + (2 − 0.5d − 3 zc)˝b2 + [3 zc2 − (3 − 1.5d) zc + 1 − 0.5d] ˝b
requirements for the precision of the mixing rule for this parameter
− zc3 = 0 (16) could pose a problem – taking into account its undefined physical
meaning.
where ˝b = bc pc /RTc .
The system of equations above can be solved in one of the two
ways: 2.50

2.00 d=2
Dimensionless parameter C

(1) By assigning appropriate value to parameter d like it was done


by Ji and Lempe [7]; 1.50
(2) By including determination of d into the optimization proce-
dure, which is the approach taken in this paper. 1.00

If parameter d has been chosen a priori, then Eq. (16) can be 0.50
used to illustrate the dependence of br = bc /vc on zc . This depend- P-T EOS
d=1
ence is one of the culprits that prevent a cubic EOS from adequate 0.00
simultaneous fit of liquid density at the regular and the near-critical
regions. Fig. 1 shows the behavior of the function br (zc ) at d = 1, and -0.50
0.200 0.220 0.240 0.260 0.280 0.300 0.320 0.340
d = 2; it makes clear why Ji and Lempe [7] EOS at d = 2 works better Critical compressibility factor
at the near-critical region than Eq. (8) (PR3 EOS that implies d = 1).
If parameter b is independent on temperature (i.e., b = bc ), then the Fig. 2. Function C(zc ) for different EOS.
144 V. Abovsky / Fluid Phase Equilibria 376 (2014) 141–153

2.4. Temperature dependence of EOS parameters. where  = T/Tc , r = /vc , and εr = εa /(kTc )
r = 1 + 1 (1 − ) + 2 (1 − )2 (25)
2.4.1. ˛-Function
2
It is common to express parameter A as εr () = ln(2) + ε1 (1 − ) + ε2 (1 − ) (26)

A = Ac ˛(T ) (17) Ln(2) appears in Eq. (26) to satisfy the condition ˛ = 1 at  = 1.


At the first glance, the proposed ˛-function contains 4 adjustable
where Ac is defined by Eq. (14), and ˛(Tc ) = 1. Most of the pro-
parameters, and each of them has to have the absolute value around
posed expressions for ˛(T) are empirical, and are not applicable
unity (or less) by the order of magnitude to provide fast conver-
at wide range of temperature (more than two-fold of Tc ). Never-
gence inside the range 0.5 < T/Tc < 2. However, the problem with
theless, this requirement is important for calculating properties of
convergence may not be resolved with volatile components like
the mixtures that contain components with substantially differ-
hydrogen, nitrogen, methane, etc., if the parameters are obtained
ent critical temperatures, such as natural gas and condensate, and
by fitting data on vapor pressure and some other key properties like
mixtures with volatile components (i.e. hydrogen, nitrogen, oxy-
liquid density, heat of vaporization, etc., only at  < 1. Extrapolation
gen, methane, argon, etc.) The attempt of the theoretical approach
of ˛-function for these components to ambient temperatures can
to the problem is outlined below.
generate unreasonable values due to big numbers of . The way to
By expanding the attractive term of Eq. (11) in the power series
solve (or, at least, to diminish) the problem is to replace expressions
of density and keeping the first term only one can obtain:
(25), (26) by more sophisticated ones – of Pade-approximant type.
A(T ) = −RT Ba (T ) (18) 1
r () = (27)
where Ba (T) is the contribution to the second virial coefficient 1 + 1 (1 − ) + 2 (1 − )2
from intermolecular attraction. Eq. (18) hints that the functional ln(2)
form ˛(T) can be defined using well known statistic-mechanical εr () = (28)
1 + ε1 (1 − ) + ε2 (1 − )2
expression
   No big numbers will be generated for the functions (27), (28)
1 − exp(−˚a (r)) even at  → ∞. Vice versa, both values will approach zero at high
Ba (T ) = 2N r 2 dr (19)
kT temperature limit. However, such asymptotic behavior will be
where N is Avogadro number, ˚a is the attractive branch of pair inconsistent with physical meaning of r and εr which have to be
potential function, k is Boltzmann constant, and r is a distance the finite positive numbers at any temperature. To avoid the incon-
between any two particles in the system. The direct expansion of sistency, Eqs. (27) and (28) have to be transformed to
Ba (T) in the power series of ˚a (r)/kT is not efficient here because 1 + 3 (1 − )2
convergence can be reached with no less than double digit number r (t) = (29)
1 + 1 (1 − t) + 2 (1 − )2
of terms to be retained [10]. To create the explicit expression for
A(T) with the extrapolative capability, one can consider the “effec- ln(2) + ε3 (1 − )2
tive” square well attractive potential function εr (t) = (30)
1 + ε1 (1 − ) + ε2 (1 − )2
˚sw (r, T ) = −εa (T ),  < r < L(T ), and ˚sw (r, T ) = 0, r > L(T ) Obviously, the additional terms in the numerators of both
expressions are not necessary for components within the range of
(20)
interest that does not exceed  = 2–3.
along with requirement that
2.4.2. Covolume temperature dependence
Ba [˚a (r), T ] = Ba [˚sw (r, T ), T ] (21)
So called “covolume” parameter b defines the repulsive contri-
Parameter  is the hard core diameter that can be connected to bution to the second virial coefficient at the limit of low density,
covolume b as  = [b/(2/3)]1/3 – in the case of a spherically sym- and equal to the four-fold volume of the “effective” hard sphere.
metric interaction. In the case of a non-spherical interaction the Upgrading the consideration to the level of the real soft binary
functions εa (T), L(T) play roles of average values over the angu- repulsive interaction implies a slightly decreasing temperature
lar dependency of the actual potential function. Eq. (21) is not an dependence b(T) – at least, for spherically symmetric interaction.
approximation. It implies the exact fit of data generated by Eq. (19). In spite of the fairly sophisticated result that comes from the per-
Its advantage is that parameters εa (T) and L(T) have to be slow func- turbation theory [11] the simple approximation can be used for
tions of temperature because the strong exponential behavior of b(T) due to high steepness of the repulsive branch of the potential
Ba (T) is reflected explicitly in the final expression which can be functions like Lennard–Jones, Born-Mayer, etc.:
obtained by replacing ˚a (r) by ˚sw (r, T) in Eq. (19):  T 
c
 εa (T )
 b(T ) = bc
T
(31)
Ba (T ) = (T ) 1 − exp (22)
kT where bc = b(Tc ), and
where 0<1 (32)
 2    3

3 Expressions (31) and (32) provide slowly decreasing tempera-


(T ) = L (T ) 1− (23)
3 L(T ) ture dependence b(T) that obviously can be applicable to the fluids
The temperature functions εa (T) and (T) can be expanded in with spherical, or, quasi-spherical, particles such as argon, benzene,
power series of (1 − T/Tc ) at least within the range 0.5 < T/Tc < 2 with hydrogen sulfide, methane, nitrogen, etc. A complication arises
the expectancy of fast convergence that allows to keep only the two when considering the validity of condition (32) for fluids with sub-
first terms in each expansion. The final expression for ˛(T) then will stantially non-spherical shape of molecules. Parameter b(T) then
acquire the following form: acquires the meaning of some average value over the mutual binary
  orientations of particles. The qualitative point of view allows to
εr () assume that increasing temperature intensifies their angular oscil-
˛() = r () exp −1 (24)
 lations making the fluid structure less compact that is equivalent
V. Abovsky / Fluid Phase Equilibria 376 (2014) 141–153 145

Table 1
Parameters of selected pure components.
exp
Component 1 2 3 ε1 ε2 ε3  d zc zc

Inorganic components
Argon −0.0379 0.9474 −0.0805 0.0496 0.0054 −0.6273 0.0234 1.6742 0.3134 0.292
Hydrogen −1.3043 1.1960 0.1303 1.3265 0.8243 0.1102 −0.0201 2.5666 0.3140 0.307
Hydrogen sulfide −1.4870 0.7566 0 0.9489 1.2413 0 0.0324 1.5988 0.3063 0.287
Nitrogen −0.7782 0.8169 0.4170 0.5814 0.3497 0.0773 0.0175 1.9423 0.3065 0.288
Carbon dioxide −1.8485 1.1717 0 1.1464 1.5926 0 0.0540 1.8012 0.2941 0.277
Ammonia −1.5995 0.9125 0 0.9519 1.2064 0 −0.0390 1.5070 0.2818 0.241
Water −0.7401 0.3924 0 0.4916 −1.2368 −0.7214 0.0137 1.8767 0.2505 0.228

Organic components
Methane −1.2430 0.2698 −0.0363 0.9444 0 −0.6181 0.0065 1.8428 0.3072 0.286
Propane −0.2869 −0.9174 −0.8913 0 −0.6371 −0.7877 −0.0086 0.9677 0.3151 0.273
n-Butane −1.6358 0.9264 0 0.9278 1.2806 0 −0.0079 1.2512 0.3089 0.272
i-Butane −1.5858 0.8684 0 0.9501 1.2091 0 −0.0108 1.4414 0.3076 0.278
n-Pentane −1.6929 0.9619 0 0.9283 1.4170 0 −0.0283 1.3207 0.3059 0.271
n-Hexane −1.7023 0.9503 0 0.9423 1.5074 0 −0.0615 1.6265 0.2984 0.260
n-Octane −1.6653 0.9403 0 0.8014 1.1233 0 −0.0173 1.2950 0.2978 0.254
n-Decane −1.6630 0.9476 0 0.7422 1.0078 0 −0.0411 1.2922 0.2958 0.245
1-Hexene −1.5063 0.8367 0 0.9777 0.8635 0 −0.0107 1.7637 0.2972 0.265
Benzene −1.6261 0.8857 0 0.9619 1.2861 0 0.0043 1.5011 0.3014 0.273
Toluene −1.6213 0.9215 0 0.9080 1.1161 0 −0.0191 1.3351 0.3012 0.262
Acetone −0.3962 −0.9131 −0.9758 0 −0.7322 −0.8536 0 1.2490 0.2843 0.234
Methanol −0.6806 2.1302 2.3432 0.2669 −1.3567 −0.8906 −0.0846 1.8582 0.2626 0.224
2-Butanol −1.8879 1.2124 0 1.0743 0.5491 0 0 1.6181 0.2895 0.254

to “effective” increasing b(T) with temperature. So, the negative molecular structure. It is observable that parameter  becomes
values of  can be expected for fluids like heavy hydrocarbons, alco- negative for components with substantial angular dependency of
hols, ethers, etc. This assumption would be considered rather as an repulsive molecular interaction.
exp
explanation than as a prediction of data regression results that are The delta between zc and zc represents the upper limit of error
described in the next section of the paper. in liquid volume at the near-critical range. The data on critical tem-
Following from the above considerations, the physically consis- peratures and pressures for substituting them to Eqs. (14)–(16)
tent determination of ˛-function parameters has to be performed were taken from handbook [12].
on the basis of second virial coefficient data only – as the first Table 2 contains residual values from fitting the compila-
step of data regression procedure. The final step would be adjus- tion data on vapor pressure, density, and heat of vaporization
ting parameters b, c, and d to the data on vapor pressure, liquid for the same components as those listed in Table 1. The data
density, heat of vaporization, etc., with fixed ˛-function obtained were generated using the tables and analytical approximations
at the first step. Unfortunately, precise data on the second virial from handbooks [12,13]. Table 2 contains analogous results for
coefficients is not available for many industrially important gases, Ji–Lempe (JL) EOS [7] obtained by excluding parameter d from data
which led to ˛-function parameters being determined here concur- regression procedure and equating it to 2. Same exact structures
rently with other parameters, using data on properties mentioned
above. Therefore it could be illusionary to expect that ˛-function Table 2
obtained in this way would represent the second virial coefficient. Root mean square percentage errors of calculating the key properties along the
Nevertheless, the correspondent numbers could be pretty close saturation curves for EOS (11) with optimized parameter d, and JL EOS [7] with d = 2.
– judging from the results of fitting heat of vaporization that is Component  Rangea EOS (11) JL EOS
strongly sensitive to the saturated vapor density accordingly to
ıps ıvL ıhvap ıps ıvL ıhvap
Clausius–Clapeyron equation.
Parameters c and d are assumed to be independent on tem- Inorganic components
Argon 0.56–0.95 0.84 1.32 1.01 0.74 1.70 1.64
perature due to: (1) the absence of any physical concept of such
Hydrogen 0.42–0.98 0.72 0.78 0.86 0.89 4.32 1.79
dependency, and (2) any functionality of this kind could inter- Hydrogen sulfide 0.56–0.94 1.71 1.21 0.49 1.88 1.52 1.48
fere with A(T) bringing in noticeable uncertainty of the correlation Nitrogen 0.56–0.95 1.22 1.18 0.96 1.15 1.50 0.97
matrix. Carbon dioxide 0.64–0.97 1.81 1.52 0.81 1.78 1.58 1.10
Ammonia 0.49–0.90 1.45 0.92 1.56 2.02 2.03 1.65
Water 0.42–0.95 0.13 1.45 1.61 0.13 1.97 1.62
2.5. Fitting experimental data for pure components Organic components
Methane 0.47–0.91 0.78 0.99 0.61 1.01 1.22 0.66
A well known postulate that any conventional EOS becomes Propane 0.24–0.90 0.77 1.47 0.70 1.17 3.33 1.30
n-Butane 0.33–0.92 0.74 1.57 0.51 1.10 2.26 2.21
inapplicable in close proximity of the critical point dictates that
i-Butane 0.34–0.91 0.64 1.35 1.09 0.83 1.84 2.10
the data on density and heat of vaporization along with the coex- n-Pentane 0.30–0.92 0.74 1.31 1.07 1.22 2.33 2.54
istence curve in the range 0.95 <  < 1.0 must be excluded from n-Hexane 0.35–0.91 1.33 1.11 1.29 1.40 1.47 1.93
data regression analysis, and compressibility factor at the critical n-Octane 0.39–0.93 0.66 1.65 1.04 0.97 2.47 2.77
point is considered as an adjustable parameter. Accordingly to Eqs. n-Decane 0.39–0.91 0.29 1.48 0.85 0.56 2.59 2.54
1-Hexene 0.39–0.90 2.02 1.80 1.83 1.88 3.70 3.07
(14)–(16), this way reproduces exact experimental values of critical
Benzene 0.50–0.91 0.67 1.21 0.62 0.80 2.07 2.05
pressure and temperature, while maintaining a noticeable devia- Toluene 0.34–0.91 0.68 1.61 0.89 0.83 2.32 2.42
tion of the calculated critical volume from the experimental value Acetone 0.39–0.98 1.87 1.69 1.39 0.83 4.47 2.48
(as a rule, the calculated values are higher by 5–15%). Methanol 0.47–0.90 1.54 1.28 1.08 1.34 1.60 2.56
2-Butanol 0.53–0.90 0.37 1.39 0.67 0.36 2.40 1.22
Table 1 contains independent parameters of Eqs. (11) and
a
(29)–(31) for selected components with strongly different The upper limit for ps () range was not less than  = 0.99.
146 V. Abovsky / Fluid Phase Equilibria 376 (2014) 141–153

1.60 3.1. Mixing rules for size and “bridge” parameters.


H2O

Eqs. (33) and (34) are based on a simple concept that the dis-
AR
1.20 tance between the centers of spherical particles in contact with
each other is equal one half of the sum of their diameters.
Alpha(Tr)

H2 3
0.80
1/3
bi + bj
1/3

bm = xi xj (1 − lrij ), lrij = lrji (33)


2
i j
0.40
3
1/3
ci + cj
1/3

0.00 cm = xi xj (1 − laij ), laij = laji (34)


0 1 2
10 10 i j
Reduced temperature Tr = T/Tc
where lrij and laij – non-additive corrections for repulsive and
Fig. 3. Alpha-functions for selected components of different molecular structures. attractive parameters respectively. Parameter lrij can be negligi-
ble for the mixture of spherical (quasi-spherical) particles of close
size, but it is definitely useful in the case of molecules of substan-
of temperature functions A(T) and b(T) were used for the both tially different sizes and shapes. Parameter laij can be non-zero
approaches. This comparison illustrates isolated influence of the when attractive intermolecular forces of components are of differ-
“bridge” parameter on the quality of data fit. ent physical nature, and/or for molecules of different size and shape
Fig. 3 illustrates behavior of ˛-functions for selected compo- as well. It needs to be mentioned that there is no basis for use of
nents over a wide range of  (0.4–0.0). The results for hydrogen Eq. (34) for Patel–Teja EOS (9) due to unclear physical meaning of
and argon were obtained by data regression within the entire tem- parameter c contained therein.
perature range (p, v, T data were used at  > 1.0), and extrapolation While the values of bi and ci for components in a mixture can dif-
was made for water beyond the critical point. Uncommon behav- fer by the order of magnitude, the values of the “bridge” parameter
ior of hydrogen ˛-function was likely caused by implicit accounting di for the most of analyzed fluids were found in a relatively narrow
for quantum effects that play significant role at low tempera- range (1.2–1.8). Therefore the linear mixing rule is suggested for
tures. Explicit evaluation of this contribution [14], however, makes dm :
the EOS non-cubic, and goes beyond the consideration of this
paper. dm = xi di (35)
Table 3 illustrates accuracy of fitting p, v, T data at high tem- i
peratures and pressures for several volatile components. These
data were regressed simultaneously with the data along the sat- 3.2. Mixing rule for parameter Am
uration curves, and calculated with the same parameters from
Table 1. Obviously, the “bridge” parameter has a little impact on The mixing rule for Am is the most important one, due to the
data regression accuracy here. The higher values of deviations dominant role of intermolecular attraction in quantitative descrip-
for hydrogen can be referred to quantum effects that contribute tion of vapor–liquid phase transition. The so-called “conventional”
strongly at high densities, and are not taken into account by one-fluid mixing rule
Eq. (11).
Am0 = xi xj Aij (36)
i j
3. Generalization for multi-component system
can be derived under the restriction that non-ideality in mixture’s
thermodynamic behavior is caused by binary interactions only; it
One-fluid approximation will be considered below as a typical
is applicable strictly to dilute gases at the level of the second virial
application of a cubic EOS to a fluid mixture. The approach implies
coefficient. If i =
/ j then
that the EOS for mixture retains the form of expression (11) with
parameters being the certain functions of component compositions Aij = (Ai Aj )1/2 (1 − kij ) (37)
designated as “mixing rules”. However, the consideration below
based on the virial theorem eventually leads to appearance of the where kij is the empirical correction to the mean geometric value
additional term in the expression for compressibility factor that of Aij . Clearly, Eqs. (36) and (37) are not applicable to the liquid
is proportional to mixture density; thus, the final EOS for mixture phase of a mixture from the theoretical point of view, and any
turns up to be the fifth degree equation with respect to volume, attempt to use it for this purpose ends in failure (excluding mixtures
though its parameters represent combinations of pure component with closely similar components). Eq. (36) predicts the maximum
parameters and compositions as is implied by the one-fluid approx- of non-ideal contributions to thermodynamic properties of binary
imation. mixtures at the equimolar compositions what is wrong for the most
fluid systems. All the attempts to introduce “asymmetric” mixing
rules can be divided between two groups:
Table 3
Root mean square errors of calculating specific volume at high temperatures and (1) corrections to the “conventional” Eq. (36);
pressures for volatile components. (2) calculations of Am with incorporation of activity coefficient
EOS (11) JL EOS models (ACM) into EOS.
Component  Range p/pc range ıvL (%) ıvL (%)

Argon 0.67–2.67 2.08–62.5 1.22 1.53 The last approach became popular owing to high flexibility of
Nitrogen 1.11–3.97 2.95–15.2 0.76 0.99 ACM. However, it contains a principal flaw connected to calcula-
Methane 1.05–2.63 0.44–11.1 1.15 1.10 tions for multi-component systems on basis of binary parameters
Hydrogen 1.21–9.09 0.08–46.3 4.52 3.84
only. This issue will be expanded on below.
V. Abovsky / Fluid Phase Equilibria 376 (2014) 141–153 147

Correction to Eq. (36) based on thermodynamic perturbation dependency on density within higher density range. At the level of
theory (PT) was described by the author in papers [15–17]. How- cubic EOS for mixture this term has to be expressed as
ever, the applicability of PT with the system of short range repulsive
x, T
molecular interaction as a reference system to the fluids with strong Za0m = −Am0 (46)
directional attractive forces (water, alcohols, etc.) is doubtful. Nev- vRT
ertheless, the same result for “asymmetric” mixing rule can be where Am0 has been defined by the conventional expression (36).
obtained on the more general basis – without dividing the molec- To reveal the correspondence between Eqs. (41) and (43)–(45)
ular interaction on the reference one, and the perturbation. The one has to write:
derivation is shown below.   
−2N d˚a −˚(r) A(T )f (v)
Za1 = exp h(r, v, T )r 3 dr = −
3.3. Virial theorem and cubic EOS 3kT v dr kT (RT )
(47)
The consideration can be started from the general expression
where f(v) defines the non-linear density function that contributes
for compressibility factor based on the virial theorem for a pure
to attractive term Za :
component:
 v 1
2N d˚ f (v) = − (48)
Z =1− g(r, v, T )R3 dr (38) (v2 + 2c v − dcb) v
3kT v dr

where k is Boltzmann constant, v is volume per particle, ˚(r) – 3.4. Generalization for mixture
binary potential function, g(r, v, T) – binary distribution function,
and r = |r2 − r1 | where r1 , r2 – radius-vectors of centers of mass The generalization of Eq. (47) for a multi-component mixture
of particles 1, 2. Expression (38) is written for the spherically can be written as following:
symmetric binary interaction for the sake of simplicity, since gen-   
eralization for an angular-dependent interaction does not change −2N d˚aij −˚(rij )
Za1m = xi xj exp hij (rij , vm , T )rij3 drij
the final result. The potential function ˚(r) can be expressed as a 3kT vm drij kT
i j
sum of repulsive ˚R (r) and attractive ˚a (r) parts without the key
restriction on the way of this separation that is required by PT, xj Aij (x, T )fij (vm ) −Am (x, T )fm (vm )
i.e.|d˚a /dr||d˚R /dr|. Then =− xi = (49)
RT RT
i j
Z = ZR + Za (39)
The general expression of Am (x, T) can be obtained at first for a
where
 binary mixture – just for simplicity of consideration:
2N d˚R
ZR = 1 − g(r, v, T )r 3 dr (40) x12 A11 f11 + 2x1 x2 A12 f12 + x22 A22 f22
3kT v dr
Za1m = −
RT
and
 = −fm (vm )
x12 A11 f11 /fm + 2x1 x2 A12 f12 /fm + x22 A22 f22 /fm
(50)
−2N d˚a RT
Za = g(r, v, T )r 3 dr (41)
3kT v dr

The required correspondence of Eqs. (40) and (41) to cubic EOS It is important to realize that the approximation of Eq. (48) type
(11) leads to obvious expressions at the macro level of EOS is not applicable to any function fij (vm )
in Eq. (49) with exception for the pure component limits, and the
1 total mixture as well. Eq. (48) would not reflect the influence of the
ZR = (42)
1 − b/v rest of particles in the mixture on the correlation function of the
two selected ones. Therefore, Eq. (49) can be considered a general
Av
Za = − (43) definition of fij . The intermediate result for Am clearly flows from
RT (v2 + 2c v − dcb)
comparison of Eqs. (49) and (50):
The binary distribution function can be presented as
x12 A11 f11 x2 A22 f22
 −˚(r)  Am = +
2x1 x2 A12 f12
+ 2 (51)
g(r, v, T ) = exp [1 + h(r, v, T )] (44) fm fm fm
kT
If each ratio fij /fm goes to unity then Am reduces to Am0 defined
where ˚(R) is the total pair potential, and h(R, v, T) is the binary by Eq. (36), and Zam reduces to Eq. (43) with conventional Am0 (x,
correlation function [18]. Eqs. (40) and (42) cannot be considered T). This simplification can occur only for a mixture of components
as EOS of a system of “effective” hard spheres because particles are with similar potential functions ˚i (r), which results in close values
distributed here accordingly to their total potential interaction. of correlation functions hij . However, any significant difference in
It is highly relevant for further consideration that the contri- potential parameters of components even with similar potential
bution from the term associated with correlation function h(r, v, functions can cause qualitative difference in hij – as it was shown
T) to thermodynamic functions is a non-linear function of density. by Lee and Levesque [19] for the case of hard sphere system.
Obviously, this term does not contribute to the original van-der- The simplest approximations of ratios fij /fm are developed
Waals EOS. below. Any assumption about the form of binary potential func-
Substituting expression (44) into Eq. (41) leads to tions of components was not necessary for deriving Eq. (51), and it
Za = Za0 + Za1 (45) is not necessary for the following phenomenological approach.
It is clear from Eq. (49) that in the general case ratio fij /fm is a
where Za0 is the contribution of the second virial coefficient at low function of variables x, v, T. Nevertheless, neglecting dependence
density area, or the part of compressibility factor with the linear on v is unavoidable because coefficient Am can be a function of x, T
148 V. Abovsky / Fluid Phase Equilibria 376 (2014) 141–153

only accordingly to Eq. (49). Let us analyze now each of values fij /fm had been caused by the attempting to introduce a correction that
in Eq. (51) as a function of x1 , x2 : was proportional to x12 x2 , or x1 x22 only to the second term of Eq. (56).
f11 f11 “Improvement” of this kind was equivalent to neglecting the natu-
= 1 if x2 → 0 and = 1 + kc112 if x2 → 1 (52) ral appearance of the similar corrections in the first and last terms of
fm fm
Eq. (51), and using the wrong combinatorial rule as a consequence
where kc112 can be interpreted as a measure of influence of particle – exactly opposite to what was done above.
2 on correlation function h11 at the limit x1 → 0, which makes h11 Combining Eqs. (39), (45), (46) and (48) leads to the final EOS
different from the same function for pure component 1. It is clear for a mixture:
that this influence becomes negligible at the limit x2 → 0. The sim-
plest approximation for the entire range of composition x2 is the RT Am Am − Am0
p= − + (60)
linear function: v − bm v2 + 2cm v − dm cm bm v2
f11 where Am and Am0 are defined by Eqs. (58) and (36) respectively,
= 1 + kc112 x2 (53)
fm and other parameters are represented by Eqs. (33)–(35). The last
The same consideration for f22 /fm leads to a similar result: term in Eq. (60) changes its status from a cubic EOS to an EOS of the
fifth degree respectively to volume, which means that the solution
f22 for its roots can be obtained by iterations only. This transformation
= 1 + kc221 x1 (54)
fm is unavoidable for any EOS with flexible mixing rule for parameter
The case of f12 /fm is more sophisticated. This value does not go Am that is different from the conventional one, but must reproduce
to unity at any limits of x1 , x2 : the correct second virial coefficient for mixture as a function of
compositions [24].
f12 f12
→ = 1 + kc121 if x2 → 0, Eq. (60) reduces to the conventional form if (Am − Am0 ) → 0. It
fm f11 happens either in the trivial case of the low density limit, or when
and the particles of each component in the mixture have similar pair
f12 f12 potential functions with very close energy and size parameters.
→ = 1 + kc122 if x1 → 0 Eq. (59) contains correction to the conventional mixing rule (36)
fm f22
which represents the contribution of ternary intermolecular inter-
At the same level of the linear approximation actions in the mixture. There is a substantial difference between
f12 coefficients kcij and kcijl . Parameter kcij is the measure of the influ-
= 1 + kc121 x1 + kc122 x2 (55) ence of the third particle j on the contribution of the interaction
fm
of the two similar particles (i, i), while kcijl reflects the influence
Substituting Eqs. (53)–(55) into Eq. (51) leads to the general of the third particle l on the contribution of the two different
expression for Am in the case of a binary mixture that contains four particles (i, j). Parameters of kcij type appear in the both binary
adjustable parameters even under the above simplification: and multi-component mixtures while kcijl shows up starting from
Am = x12 A11 (1 + kc112 x2 ) + 2x1 x2 A12 (1 + kc121 x1 + kc122 x2 ) ternary mixtures. It means that kcij can be determined by fitting
data only for binary mixtures, and experimental data on ternary
+ x22 A22 (1 + kc221 x1 ) (56) mixtures are needed for determination of kcijl . However, the data
on ternary combinations of components are available in very rare
Collecting terms with the same power degrees of compositions occasions compare to availability of data on binary combinations
kc112 + 2kc121 A12 kc221 + 2kc122 A12 for the given multi-component system of interest. The estimation of
kc12 = , kc21 = (57) parameter kcijl can be proposed to make Eqs. (58) and (59) correct
A11 A22
at limiting cases kcil = O(kcjl ), kcil  kcjl which guarantees avoiding
transforms Eq. (56) to the final expression for a binary mixture:
“MK syndrome”. The details are presented in Appendix.
Am = x12 A11 (1 + kc12 x2 ) + 2x1 x2 (A1 A2 )1/2 (1 − k12 ) 1/2
kcijl = sign(kcil )(kcil kcjl ) if sign (kcil ) = sign(kcjl ) and
+ x22 A22 (1 + kc21 x1 ) (58)
kcijl = 0 if sign(kcil ) =
/ sign(kcjl ) (61)
Corrections kc12 and kc21 acquire a combined meaning of influ-
ence of the third particle j on the distribution functions of ii and ij
Values of kcil are negative with rare exceptions (see Table 4)
types. Moreover, absorption of the terms that are proportional to
which is well understandable as correlation between similar pair
kc121 and kc122 makes corrections kc12 and kc21 explicit functions
of particles in mixture weakens due to the presence of particles of
of temperature. However, this functionality can be neglected due
another type.
to similar trend of Aij (T) and Aii (T) – even for binary mixtures with
No estimations of values kcij can be made on the basis of chemical
so different components like nitrogen/n-decane and methane/n-
formulas and molecular structures. The only reasonable assump-
decane as it can be seen in the next section.
tion could be expressed by inequality abs(kci j ) > abs(kcji ) if bj > bi ,
Generalization of Eq. (58) for a multi-component mixture has to
and/or Aj > Ai .
be written as
⎛ ⎞ ⎛ ⎞
3.5. Conclusive notes on the proposed mixing rule for parameter
Am = xi2 Aii ⎝1 + kcij xj ⎠ + xi xj Aij ⎝1 + kcijl xl ⎠
Am
i j=
/ i i j=
/ i l=
/ i,j

(59) 1. Mixing rule (59) reflects the simple and clear concept that the
contribution to thermodynamic properties from binary interac-
Applying Eq. (59) to a ternary mixture, and then making particles tions of the similar particles in the mixture is not the same as
j, l virtually identical transforms Eq. (59) into Eq. (58) (details are their contribution in the one-component fluid. This difference is
given in Appendix). So called “Michelsen–Kistenmacher (MK) syn- accounted for by the correction in the first term of the right side
drome” [20] that a few empirical approaches [21–23] suffer from of Eq. (59).
V. Abovsky / Fluid Phase Equilibria 376 (2014) 141–153 149

Table 4
Binary parameters for selected mixtures.
(0) (1) (2)
Comp1/Comp2 k12 k12 k12 lr12 la12 kc12 kc21

n-Butane/methanol 0.5947 −143.5 0 0 0.7151 −1.3340 −0.4491


n-Pentane/methanol 0.5252 −148.8 0 0 0.5963 −1.3976 −0.6344
2-Butanol/water 1.6768 −44.28 −0.1258 0 0.0656 −0.5176 −0.8926
Acetone/water 0.1587 −91.86 0 −0.1240 −0.0959 −0.1063 0.1690
Ammonia/water −0.3871 0 0 0 0.1500 −0.8558 −0.4280
Nitrogen/n-decane −0.2973 74.65 0 −0.4315 −0.2112 −0.753 −0.0957
Methane/n-decane −0.1097 40.79 0 −0.2620 −0.3697 −0.572 −0.0929
n-Butane/n-decane −0.1163 55.65 0 −0.0535 0 0 −0.0554
Methane/n-butane 0.0594 −14.31 0 −0.1174 −0.1412 −0.1806 0
Nitrogen/methane 0.0397 0 0 0 0 0 0

Table 5
Root square mean errors of calculating phase equilibrium data for selected binary mixtures.

Comp1/Comp2 Type of Temperature Pressure ␦p (%) ␦T (%) ␦x11 (%) ␦x12 (%) ␦y1 (%) Source of
equilibrium range (K) range (MPa) exp. data

n-Butane/methanol VLE 323.2–373.2 0.06–1.73 3.9 0 0 3.2 [26]


n-Pentane/methanol VLE 373.2–423.2 0.35–2.53 1.4 0 0 3.7 [27]
2-Butanol/water VLE 360.2–373.2 0.1013 0 0.02 0 3.0 [28]
2-Butanol/water LLE 283.2–373.2 0.1–70.9 0 0 6.6 5.6 [29]
Acetone/water VLE 373.2–523.2 0.1–6.76 2.4 0 0 2.3 [30]
Ammonia/water VLE 283.2–541.0 0.1–21.8 2.1 0 0 1.9 [31,32]
Nitrogen/n-decane VLE 310.9–410.9 2.1–34.5 3.1 0 0 0.12 [33]
Methane/n-decane VLE 310.9–410.9 0.1–29.3 2.8 0 0 0.17 [34]
n-Butane/n-decane VLE 310.9–510.9 0.1–4.8 1.3 0 0 0.7 [34]
Methane/n-butane VLE 210.9–410.9 0.1–13.1 2.8 0 0 2.4 [34]
Nitrogen/methane VLE 100.0–180.0 0.1–4.8 0.8 0 0 0.7 [35]

Table 6
Comparison of calculated and experimental data on vapor–liquid equilibrium in ternary mixture nitrogen(1)/methane(2)/n-decane(3). The compositions are in mole fractions.

Liquid comps Pressure Vapor compositions

T (K) x(1) x(2) exp calc ıp y(1) ıy(1) y(2) ıy(2)


MPa % exp calc % exp calc %

310.9 0.0110 0.2430 6.896 7.097 2.9 0.094 0.095 1.1 0.906 0.904 −0.2
310.9 0.0975 0.2250 13.79 13.65 −1.0 0.448 0.489 9.2 0.552 0.510 −7.6
310.9 0.2864 0.0438 27.58 26.68 −3.3 0.934 0.936 0.2 0.065 0.062 −4.6
310.9 0.3300 0.0520 34.47 32.90 −4.5 0.932 0.932 0.0 0.065 0.066 1.5

344.3 0.0490 0.1405 6.896 7.423 7.6 0.453 0.441 −2.6 0.546 0.558 2.2
344.3 0.1701 0.0419 13.79 14.25 3.3 0.907 0.901 −0.7 0.092 0.097 5.4
344.3 0.2779 0.0631 27.58 26.13 −5.3 0.901 0.908 0.8 0.095 0.090 −5.3
344.3 0.1110 0.5830 34.47 35.13 1.9 0.222 0.210 −5.4 0.769 0.765 −0.5

377.6 0.0523 0.1167 6.896 7.075 2.6 0.483 0.484 0.2 0.514 0.511 −0.6
377.6 0.1132 0.1958 13.790 14.690 6.5 0.532 0.529 −0.6 0.463 0.464 0.2
377.6 0.1720 0.3310 27.58 26.58 −3.6 0.474 0.466 −1.7 0.518 0.519 0.2
377.6 0.2640 0.2780 34.47 34.47 0.0 0.615 0.609 −1.0 0.375 0.375 0.0
377.6 0.3210 0.1350 34.47 33.74 −2.1 0.810 0.808 −0.2 0.180 0.181 0.6

410.9 0.0505 0.1200 6.896 7.012 1.7 0.440 0.441 0.2 0.553 0.545 −1.4
410.9 0.0939 0.0216 6.896 6.762 −1.9 0.892 0.886 −0.7 0.101 0.102 1.0
410.9 0.1047 0.1943 13.79 13.65 −1.0 0.506 0.488 −3.6 0.484 0.497 2.7
410.9 0.1925 0.2875 27.58 25.50 −7.5 0.517 0.513 −0.8 0.468 0.464 −0.9
410.9 0.3730 0.0670 34.47 34.13 −1.0 0.881 0.892 1.2 0.102 0.090 −11.8

Average relative deviations, %: 3.3 1.6 2.6

2. Moreover, the contribution from binary interactions of the two the parabolic one, can lead to unnecessary complications as the
different particles to properties of multi-component mixture is simplest approach provides adequate fit of experimental data.
not the same as the contribution in their binary mixture. This 4. If the correction in the last term of Eq. (59) is significant it means
difference is accounted by the correction in the last term of the following: any method that ignores such contribution (either
Eq. (59). The hierarchy of this kind can be extended infinitely, obtained from data or estimated) cannot aspire to the same level
but restricting it by ternary contributions already provides the of calculation accuracy for multi-component mixture as data fit for
level of accuracy that satisfies requirements of most engineering binary systems provides.
applications as it is shown below. The statement above is relevant for all EOS with conventional
3. The restriction by ternary parameters in Eq. (59) is the straight mixing rules, and activity coefficient models (ACM), and also very
result of the assumed linear dependency on compositions of popular various mixing rules for EOS that are based on ACM–EOS
ratios fij /fm in Eq. (51). The more sophisticated dependency, say, connection as well. The methods which are associated with ACM
150 V. Abovsky / Fluid Phase Equilibria 376 (2014) 141–153

Table 7
Comparison of AAD in pressure and constants of phase equilibrium in ternary mixture nitrogen(1)/methane(2)/n-decane(3) calculated by Eq. (60) and PR EOS [31].

Liquid comps Pressure (MPa) ıp (%) ıK(1) (%) ıK(2) (%) ıp (%) ıK(1) (%) ıK(2) (%)

T (K) x(1) x(2) Eq. (60) PR EOS

310.9 0.0110 0.2430 6.896 2.9 1.1 −0.2 −0.3 −8.2 −0.6
310.9 0.0975 0.2250 13.79 −1.0 9.2 −7.6 −2.1 −12.4 −8.0
310.9 0.2864 0.0438 27.58 −3.3 0.2 −4.6 −11.0 −10.3 −7.1
310.9 0.3300 0.0520 34.47 −4.5 0.0 1.5 −5.5 −5.0 −0.2

344.3 0.0490 0.1405 6.896 7.6 −2.6 2.2 −5.1 −1.9 −6.6
344.3 0.1701 0.0419 13.79 3.3 −0.7 5.4 −0.6 −0.5 −10.1
344.3 0.2779 0.0631 27.58 −5.3 0.8 −5.3 −5.1 −4.7 −10.6
344.3 0.1110 0.5830 34.47 1.9 −5.4 −0.5 −5.1 −0.8 −2.7

377.6 0.0523 0.1167 6.896 2.6 0.2 −0.6 −2.1 −3.2 −0.8
377.6 0.1132 0.1958 13.790 6.5 −0.6 0.2 −4.4 −1.6 −6.0
377.6 0.1720 0.3310 27.58 −3.6 −1.7 0.2 −2.5 −4.9 −1.4
377.6 0.2640 0.2780 34.47 0.0 −1.0 0.0 −0.7 −2.5 −5.1
377.6 0.3210 0.1350 34.47 −2.1 −0.2 0.6 −3.5 −4.1 −7.1

410.9 0.0505 0.1200 6.896 1.7 0.2 −1.4 −4.6 −5.6 −3.6
410.9 0.0939 0.0216 6.896 −1.9 −0.7 1.0 −10.7 −10.7 −4.7
410.9 0.1047 0.1943 13.79 −1.0 −3.6 2.7 −5.0 −9.0 −1.1
410.9 0.1925 0.2875 27.58 −7.5 −0.8 −0.9 −8.5 −8.6 −2.9
410.9 0.3730 0.0670 34.47 −1.0 1.2 −11.8 −5.6 −3.3 −6.4

Average absolute deviations, %: 3.2 1.6 2.6 4.6 5.4 4.7

do not even leave room for modification of Eq. (59) type. Only of excess properties for binary mixtures only. Applying them for
a rare occasion of mutual cancelation of errors can provide a multi-component mixtures with strongly different components
satisfactory level of accuracy here. brings the calculations to the uncertain level of inaccuracy that
5. The final result (59) is obtained by phenomenological transfor- could be illustrated by results that have been analyzed in paper
mation of general expression (41) based on the virial theorem [25]. Another confirmation of this statement can be seen by anal-
that is derived under assumption of the pair additivity of micro- ysis of calculations for ternary mixture nitrogen/methane/decane
scopic potential function of the system. It means that retaining below.
micro level of the consideration up to the final result for thermo-
dynamic functions (numerical methods, integral equations for 3.6. Data regression analysis for binary systems; predictions for
the binary distribution function, etc.) does not need any data for ternary mixtures.
determination of micro parameters beyond data on pure com-
ponents and their binary mixtures. So, the appearance of ternary Binary systems for data fit and comparison of experimental and
parameters in Eq. (59) is the unavoidable compensation for the calculated results were chosen accordingly to the two principles:
phenomenological transition (47) from micro to macro level of
consideration. Estimation (61) allows to avoid analysis of data on 1. Sophisticated phase diagram.
ternary mixtures but its reliability is obvious only in the limits 2. Availability of experimental data for ternary systems containing
kcil = O(kcjl ). and kcil  kcjl as it was mentioned above. the corresponding components.

The basic assumption of ACM, or, EOS–ACM approaches about For all binary systems being analyzed only parameter k12 had
pair additivity of parameters of excess Gibbs energy has noth- to be made temperature dependent for some systems. The simple
ing in common with the similar assumption for the microscopic functionality was used:
potential function. The macro parameters of Wilson, NRTL, and (1)
ef (0) k12 (2)
UNIQUAC models can be used, strictly speaking, for calculations k12 = k12 + + k12 ln(T ) (62)
T

20.0 30.0

25.0
T=422.6 K
15.0
T= 373.16 K
Pressure, atm

20.0
Pressure, atm

LIQUID 15.0
10.0 LIQUID
VAPOR VAPOR
10.0 T=372.7 K
5.0
T= 323.16 K 5.0

0.0 0.0
0.00 0.20 0.40 0.60 0.80 1.00 0.00 0.20 0.40 0.60 0.80 1.00
Mole fraction n-Butane Mole fraction n-Pentane

Fig. 4. Phase equilibrium diagram in n-butane/methanol system. The lines are cal- Fig. 5. Phase equilibrium diagram in n-pentane/methanol system. The lines are
culated results, and points represent experimental data [26]. calculated results, and points represent experimental data [27].
V. Abovsky / Fluid Phase Equilibria 376 (2014) 141–153 151

Table 8
Comparison of calculated and experimental data on vapor–liquid equilibrium in ternary mixture methane(1)/n-butane(2)/n-decane(3). The compositions are in mole fractions.

Liquid comps Pressure Vapor compositions

T (K) x(1) x(2) exp calc ıp y(1) ıy(1) y(2) ıy(2)

MPa % exp calc % exp calc %

344.3 0.2471 0.6019 6.896 6.796 −1.5 0.827 0.816 −1.3 0.172 0.183 6.4
344.3 0.4445 0.3333 13.79 13.71 −0.6 0.887 0.877 −1.1 0.107 0.117 9.3
344.3 0.5640 0.0870 20.68 20.36 −1.5 0.960 0.957 −0.3 0.029 0.030 3.4
344.3 0.5830 0.1670 20.68 20.77 0.5 0.917 0.917 0.0 0.067 0.068 1.5

410.9 0.1059 0.1790 3.448 3.535 2.5 0.837 0.835 −0.2 0.146 0.146 0.0
410.9 0.2130 0.1580 6.896 6.943 0.2 0.898 0.898 0.0 0.087 0.085 −2.3
410.9 0.4050 0.0800 13.79 14.06 2.0 0.944 0.941 −0.3 0.038 0.035 −7.9
410.9 0.3910 0.1230 13.79 13.52 −2.0 0.918 0.921 0.3 0.063 0.055 −12.7

Average absolute deviations: 1.4 0.4 5.4

Table 9
Excess volumes Vex , and deviations of calculated molar volumes of liquid paraffins binary mixtures ıV at equimolar compositions from experimental values [34].

Comp1/comp2 T (K) P (MPa) Vex (%) ıV (%) T (K) P (MPa) Vex (%) ıV (%)

Methane/n-decane 310.9 27.57 −15.0 0.6 410.9 27.57 −20.5 −1.5


41.35 −7.6 1.0 41.35 −12.0 −1.2
55.14 −4.7 1.5 55.14 −7.6 0.8
68.92 −3.1 2.3 68.92 −5.3 −0.3

Methane/n-butane 310.9 13.79 −47.2 3.3 410.9 13.79 −14.6 5.6


27.57 −14.5 1.4 27.57 −12.4 −2.9
41.35 −5.7 0.8 41.35 −5.6 −1.0
55.14 −3.4 0.7 55.14 −3.3 −0.1
68.92 −2.6 0.7 68.92 −3.6 −0.6

n-Butane/n-decane 310.9 13.79 −0.8 −0.3 410.9 13.79 −3.4 −1.8


27.57 −0.6 −0.2 27.57 −1.7 −2.2
41.35 −0.5 0.1 41.35 −1.1 −2.2
55.14 −1.5 0.4 55.14 −2.7 −2.0
68.92 −1.1 0.9 68.92 −0.8 −1.2

Table 4 contains results for parameters obtained exclusively by 374

fitting data on phase equilibrium. 372


LIQUID
VAPOR
As can be seen nitrogen/methane system requires only small
correction k12 which supports adequate applicability of mixing 370
Temperature, K

rules (33)–(37) to the binary mixture of quasi-spherical particles 368


of close sizes. The systems with more sophisticated intermolec-
ular interaction require almost the entire arsenal of parameters 366

introduced by Eqs. (33)–(56). Correlation matrix did not show any 364
sensitivity to parameters equal to zero in Table 4.
Table 5 illustrates average relative deviations for pressure (tem- 362

perature) and vapor/liquid 2 compositions within p, T ranges for the 360


binary systems listed in Table 4 (the ordinary least squares (OLS) 0.00 0.20 0.40 0.60 0.80 1.00

method was used here). Mole fraction SBA


Figs. 4 and 5 show results for n-butane/methanol and
Fig. 6. VLE in SBA/water system at 1 atm. The points represent experimental data
n-pentane/methanol respectively. In spite of all correction param- [28].
eters for these binaries being fairly close values (that is
understandable due to similar molecular structure of hydrocar-
bon components) the shapes of the phase diagrams are noticeably
different. It means that the shape of a phase diagram is mostly 1 atm
380.0 300 atm
dependent on the differences in fundamental parameters of com- 700 atm
ponents – such as critical ones, etc.
Temperature, K

Figs. 6 and 7 present results for 2-butanol/water mixture that


forms flat azeotrope at vapor–liquid equilibrium, and closed loops 340.0

with noticeable dependence on pressure at liquid–liquid splitting.


Only data at atmospheric pressure [28,29] were used to determine
parameters for this system. The comparison of experimental [29] 300.0
and calculated data at high pressures for this system illustrates the
extrapolation capability of the proposed mixing rules.
Table 6 represents deviations of calculated results for pres- 260.0
20 40 60 80
sure and vapor compositions in ternary system nitrogen(1)/
SBA fraction, Mass %
methane(2)/n-decane(3) from experimental results [33]. This mix-
ture is an example of substantial influence of the big particles of one Fig. 7. LLE in SBA/water system. The points represent experimental data [29].
component (n-decane) on correlation functions of small particles of
152 V. Abovsky / Fluid Phase Equilibria 376 (2014) 141–153

Table 10 in [36] for the same properties for all three binary combinations
Calculated liquid specific volumes of ternary mixture methane(1)/n-butane(2)/n-
below are pretty close to the values shown in Table 5:
decane(3) at comparison with experimental data [34].
For nitrogen/n-decane: ıp = 4.6%, ıy(1) = 0.16%;
Mole compositions T (K) P (MPa) Vcalc Vexp ıV (%) For methane/n-decane: ıp = 2.9%, ıy(1) = 0.22%;
X(1) X(2) X(3) cm3 /g For nitrogen/n-decane: ıp = 1.1%, ıy(1) = 2.8%;
It means that the results of data fit cannot be improved sub-
0.0906 0.0909 0.8185 310.90 13.78 1.414 1.405 0.67
27.57 1.399 1.384 1.08 stantially for binary systems with any approach, and the large
41.35 1.387 1.366 1.54 differences for the ternary mixture have to be attributed to the rea-
55.14 1.377 1.351 1.92 son mentioned above, i.e., impossibility to maintain the same level
68.92 1.369 1.336 2.47 of precision for binary and multi-component systems on basis of
410.90 13.78 1.545 1.565 −1.28 binary macro parameters only.
27.57 1.505 1.522 −1.12 Table 8 illustrates the similar results as Table 6 for ternary sys-
41.35 1.477 1.488 −0.74
tem methane(1)/n-butane(2)/n-decane(3). This system was chosen
55.14 1.456 1.461 −0.34
68.92 1.440 1.440 0.00
not just due to substantial differences in size of its components but
also due to availability of consistent data on specific volumes for
0.2307 0.0769 0.6924 310.90 13.78 1.458 1.445 0.90
the total mixture, and for its components and their binary mix-
27.57 1.440 1.422 1.27
41.35 1.426 1.401 1.78 tures as well [34]. Data on specific volumes for binary and ternary
55.14 1.414 1.384 2.17 systems were used only for comparison with results of calculations
68.92 1.405 1.370 2.55 that were performed using parameters obtained on basis of data
410.90 13.78 1.615 1.640 −1.52 on vapor pressure, density, and heat of vaporization for pure com-
27.57 1.563 1.579 −1.01 ponents (Table 1), and phase equilibrium data for binary systems
41.35 1.529 1.540 −0.71 (Table 4).
55.14 1.504 1.509 −0.33
Table 9 illustrates deviations in molar volume values for
68.92 1.486 1.486 0.00
equimolar binary systems. It also represents experimental excess
0.0525 0.4738 0.4737 310.90 13.78 1.478 1.481 −0.20 molar volumes at the same compositions where they reach their
27.57 1.456 1.455 0.07
maximum values. It can be seen that the average deviations for pure
41.35 1.438 1.434 0.28
55.14 1.424 1.413 0.78 components (Table 2) and their maximum values for binary mix-
68.92 1.414 1.398 1.14 tures have the same order of magnitude. The deviations remain at
410.90 13.78 1.653 1.680 −1.61
the same level for the ternary system that is obvious from Table 10.
27.57 1.589 1.623 −2.09
41.35 1.548 1.579 −1.96
55.14 1.519 1.547 −1.81 4. Conclusion
68.92 1.497 1.516 −1.25
(1) A new conventional EOS for pure fluids has been proposed
0.3333 0.3333 0.3334 310.90 13.78 1.622 1.616 0.37
27.57 1.584 1.571 0.83 that provides maximum possible precision of calculating ther-
41.35 1.557 1.537 1.30 modynamic properties within the family of cubic EOS. The
55.14 1.536 1.511 1.65 restrictions imposed on the EOS by fundamental thermody-
68.92 1.527 1.486 2.76
namic laws and its sensitivity to data fit made it 4-parameter
410.90 13.78 1.925 1.938 −0.67 EOS with 2 temperature dependent parameters.
27.57 1.792 1.810 −0.99 (2) A new alpha-function as an exponent with parameters in the
41.35 1.719 1.739 −1.15
form of Pade-approximants provides applicability of the EOS in
55.14 1.670 1.689 −1.12
68.92 1.643 1.646 −0.18 the wide range of temperature including two- and one-phase
state.
0.2309 0.6922 0.0769 310.90 13.78 1.795 1.789 0.34
27.57 1.724 1.730 −0.35
(3) Generalization of EOS for multi-component fluid system has
41.35 1.678 1.684 −0.36 been derived on basis of virial theorem. The correction to
55.14 1.645 1.642 0.18 the conventional one-fluid approximation transforms the cubic
68.92 1.634 1.617 1.05 EOS for pure components to the fifth degree with respect to
410.90 13.78 2.356 2.315 1.77 volume EOS for their mixtures.
27.57 2.046 2.065 −0.92 (4) The proposed approach is different from the most of the
41.35 1.911 1.944 −1.70 previous publications about mixing rules containing ternary
55.14 1.829 1.871 −2.24
68.92 1.789 1.811 −1.21
parameters in the following aspects:
theoretical foundation of the current approach;
clear physical meaning of adjustable parameters;
method of estimation of ternary parameters that guaran-
two others. As it was mentioned in the previous section this influ-
tees avoiding “MK syndrome”;
ence is reflected by the numerical values of parameters kc12 , kc21
correct concentration dependency of the second virial coef-
in Table 4 which are both negative that means noticeable dimin-
ficient.
ishing of small particles contributions to thermodynamic behavior
(5) Extensive comparison of calculated and experimental data on
of the total mixture. Parameter kc123 was estimated accordingly to
thermodynamic properties for pure components, binary and
empirical recommendation (61); as can be seen the deviations for
ternary mixtures have shown discrepancies mostly at the level
the ternary system have the same order of magnitude as for binary
comparable to experimental errors.
combinations of its components.
Table 7 illustrates comparison of deviations from experimental
data [33] for pressure and constants of phase equilibrium in mix- Appendix. Check for Michelsen–Kistenmacher syndrome
ture nitrogen(1)/methane(2)/n-decane(3) calculated by Eq. (60)
and PR EOS. The results for PR EOS are taken from handbook [36]. MK syndrome is a deficiency of mixing rule for a parame-
It is important to make accent that the results that are reported ter that contains corrections proportional to the cubic degree of
V. Abovsky / Fluid Phase Equilibria 376 (2014) 141–153 153

compositions. The deficiency can be detected by reducing a ternary x12 A1 (1 + kc12 x2 ) x12 A1 (1 + kc12 x2 )
 kc21 x1
system to a binary one. This can be achieved by making the assump- Am = + + x22 A2 1 +
4 4 2
tion that two components of the ternary system are identical. The 
non-zero difference in values of the parameter for the reduced kc21 x1 x12 A1 (1 + kc12 x2 )
+ + + x1 x2 A12 + x1 x2 A12
binary system and for the original one means that MK syndrome 2 2
is present. The relevant test of mixing rule (59) is outlined below.
= x12 A1 (1 + kc12 x2 ) + x22 A2 (1 + kc12 x2 ) + 2x1 x2 A12 (A-6)
Eq. (59) for a ternary mixture can be written in a transparent
form as follows:
which is nothing other than Eq. (58).
Am = x12 A1 (1 + kc12 x2 + kc13 x3 ) + x22 A2 (1 + kc21x1 + kc23 x3 )

+ x32 A2 (1 + kc31 x1 + kc32 x2 ) + 2x1 x2 A12 (1 + kc123 x3 ) References

+ 2x1 x3 A13 (1 + kc132 x2 ) + 2x2 x3 A23 (1 + kc231 x1 ) (A-1) [1] S.B. Kiselev, Fluid Phase Equilib. 128 (1997) 1.
[2] S.B. Kiselev, J.F. Ely, J. Chem. Phys. 119 (2003) 8645.
[3] N.C. Patel, A.S. Teja, Chem. Eng. Sci. 37 (1982) 463.
Making components 1, 2 identical implies that a set of symbols [4] S. Jiuxun, Fluid Phase Equilib. 193 (2002) 1.
[5] P.H. Salim, M.A. Trebble, Fluid Phase Equilib. 65 (1991) 59.
has to be renamed with superscripts (3) and (2) denoting values
[6] O. Redlich, J.N.S. Kwong, Chem. Rev. 44 (1949) 233.
for the original ternary system, and the reduced binary one respec- [7] W.R. Ji, D.A. Lempe, Fluid Phase Equilib. 147 (1998) 85.
tively. It is expedient to consider equal amounts of compositions [8] G. Schmidt, H. Wenzel, Chem. Eng. Sci. 35 (1980) 1503.
(3) (3) [9] J.J. Martin, Y.C. Hou, AIChE J. 1 (1955) 142.
x1 and x2 in the original ternary system. All actions of renaming [10] J.O. Hirschfelder, C.F. Curtiss, R.B. Bird, Molecular Theory of Gases and Liquids,
are identified by arrows below: John Wiley, New York, 1954.
[11] L. Verlet, J. Weis, Phys. Rev. A5 (1972) 939.
(2) (2)
(3) x1 (3) x1 (3) (2) [12] B.E. Poling, J.M. Prausnitz, J.P. O’Connell, The Properties of Gases and Liquids,
x1 → ; x2 → ; x3 → x2 (A-2) McGraw-Hill, 2001.
2 2 [13] R.H. Perry, D.W. Green, J.O. Maloney, Perry’s Chemical Engineers’ Handbook,
(3) (2) (3) (2) (3) (2) (3) (2) (3) (2) McGraw-Hill, 1998.
A1 → A1 ; A2 → A1 ; A3 → A1 ; A12 → A1 ; A13 → A12 ;
[14] L.D. Landau, E.M. Lifshitz, Statistical Physics, Pergamon Press, 1969.
(A-3)
(3) (2) [15] V. Abovsky, AIChE Symp. Ser. 298 (1994) 188.
A23 → A12 [16] V. Abovsky, Fluid Phase Equilib. 116 (1996) 170.
[17] N.C. Patel, V. Abovsky, S. Watanasiri, Fluid Phase Equilib. 152 (1998) 219.
Accordingly to Eq. (52) kcij = 0 if i = j which is taken into account [18] H.N.V. Temperley, J.S. Rowlinson, G.S. Rushbrook, Physics of Simple Liquids,
below: vol.1, North-Holland Publishing Company, Amsterdam, 1968.
[19] L. Lee, D. Levesque, Mol. Phys. 26 (1973) 1351.
(3) (2) (3) (2) (3) (2) (3) (2) (3)
kc12 → kc11 = 0; kc12 → kc11 ; kc21 → kc11 = 0; kc23 → kc12 ; kc31 [20] M.L. Michelsen, H. Kistenmacher, Fluid Phase Equilib. 58 (1990) 229.
[21] A.J. Panagiotopoulos, R.C. Reid, ACS Symp. Ser. 300 (1986) 571.
(2) (3) (2) [22] R. Stryjek, J.H. Vera, Can. J. Chem. Eng. 64 (1986) 323.
→ kc21 ; kc32 → kc21 (A-4)
[23] J. Schwartzentruber, H. Renon, S. Watanasiri, Fluid Phase Equilib. 52 (1989)
127.
[24] V. Abovsky, S. Watanasiri, Int. J. Thermophys. 19 (1998) 1429.
Accordingly to Eq. (61) and (A-4) [25] N.C. Patel, V. Abovsky, S. Watanasiri, Fluid Phase Equilib. 185 (2001) 397.
[26] A. Leu, C. Chen, D.B. Robinson, AICHE Symp. Ser. 85 (271) (1989) 11.
(3) (2) (2) 1/2 (2) (2) 1/2 (2) (3)
kc123 = [kc13 kc23 ] → [kc12 kc12 ] = kc12 ; kc132 [27] R. Wilsak, S. Campbell, G. Thodos, Fluid Phase Equilib. 33 (1987) 157.
[28] A.I. Altsybeeva, V.P. Belousov, N.V. Owtrakht, Zh. Fiz. Khim. 38 (1964) 1242.
[29] T. Moriyoshi, S. Kaneshina, K. Aihara, K. Yabumoto, J. Chem. Thermodyn. 7
(2) (2) 1/2 (2) (2) 1/2 (3)
= [kc12 kc32 ] → [kc11 kc21 ] = 0; kc231 (1975) 537.
[30] J. Griswold, S.Y. Wong, Chem. Eng. Progr. Symp. Ser. 18 (1952) 48.
(2) (2) 1/2 (2) (2) 1/2 [31] S. Rizvi, R.A. Heidemann, J. Chem. Eng. Data 32 (1987) 183.
= [kc21 kc31 ] → [kc11 kc21 ] =0 (A-5) [32] C.L. Sassen, et al., J. Chem. Eng. Data 35 (1990) 140.
[33] A. Azarnoosh, J.J. Mcketta, J. Chem. Eng. Data 8 (1963) 494, 513.
[34] (a) H.H. Reamer, B.H. Sage, W.N. Lacey, Ind. Eng. Chem. 43 (1951) 1436;
Replacing the values in Eq. (A-1) by their equivalents from the (b) H.H. Reamer, B.H. Sage, W.N. Lacey, Ind. Eng. Chem. 44 (1952) 1671.
right sides of expressions (A-2)–(A-5) yields the result for the two- [35] A.F. Kidnay, Cryogenics 15 (1975) 531.
[36] J. Sorenson, W. Artl, Chemistry Data Series, vol. V, Part 1, Dechema, 1979.
component limit of the initial ternary system:

You might also like