Thermochimica Acta 432 (2005) 1019

Evaluation of the parameters of the Bender equation of state for

low acentric factor fluids and carbon dioxide
J. Ghazouani, O. Chouaieb, A. Bellagi
U.R.: Thermique et Thermodynamique des Procedes
Industriels,
Ecole Nationale dIngenieurs de Monastir, ENIM, Av. Ibn Jazzar, 5060 Monastir,
Tunisia
Received 26 June 2004; received in revised form 19 October 2004; accepted 1 November 2004
Available online 10 May 2005

Abstract
Volumetric properties of several low acentric factor fluids (Ar, CH4 , C2 H6 , Kr, N2 , Ne, O2 , Xe) as well as CO2 are modeled using the
Bender equation of state. This equation is a linear function of 19 adjustable parameters, which are evaluated from properties data, using a
linear numerical procedure. The validity of the EOS is tested by calculating the JouleThomson inversion curve. A simple model is in
particular used to correlate the inversion properties predicted by the Bender equation, expressed in term of reduced pressure as a function
of reduced temperatures ranging from 0.8 to 6. The simple correlation reproduces accurately the used data. We employ data on state
behaviour (P, T ) of homogeneous fluid phases, vapourliquid equilibrium, second virial coefficient and the coordinates of the critical
point.
2004 Elsevier B.V. All rights reserved.
Keywords: Equation of state; Second virial coefficient; Inversion curve; Acentric factor; Pure fluids

1. Introduction

E = a11 + a12 / T,
3

The Bender equation of state [1], with 19 adjustable parameters, constitutes a good compromise between common
simple equations and more elaborated equations. It can reproduce with high fidelity the properties of the fluids in a
wide area of the thermodynamic surface, in both homogeneous fluid regions as well as at the vapourliquid equilibrium curve. The Bender equation is generally written in
terms of the compressibility factor Z as a function of
temperature T and density :
Z = PM/RT = 1 + B/R + C2 /R + D3 /R + E4
/R + F5 /R + (G + H2 )2 /R e /a20
(1)
2

B = a1 + a2 /T + a3 /T 2 + a4 /T 3 + a5 /T 4 ,

G = a14 /T + a15 /T + a16 /T 5 ,

H = a17 /T 3 + a18 /T 4 + a19 /T 5 .

(2)

If a20 is set as usual equal to c , Eq. (1) is a linear function of the 19 adjustable fluid specific parameters which
can be evaluated from properties data using a numerical
fitting methods like that proposed by Cibulka et al. [2]. The
valid- ity of the equation is tested in the present paper
not only by comparing the calculated volumetric properties
with their data values but also by comparing the Joule
Thomson inver- sion curve predicted by the Bender
equation with published

where

C = a6 + a7 /T + a8 /T 2 ,

F = a13 / T,
4

D = a9 + a10 / T,

Corresponding author. Tel.: +216 73500244; fax: +216 73500514.

E-mail address: a.bellagi@enim.rnu.tn (A. Bellagi).

inversion data. We show the importance of incorporating a

derivative property like the second virial coefficient for the
correct evaluation of the adjustable parameters. We
consider
in this paper only the low acentric factor fluids (Ar, CH4 ,
C2 H6 , Kr, N2 , Ne, O2 , Xe) as well as CO2 . Table 1 gives
the main characteristics of the considered fluids i.e., the
critical data Tc , Pc and Zc , the Pitzer acentric factor and
the Boyle temperature TB .

0040-6031/$ see front matter 2004 Elsevier B.V. All rights

reserved. doi:10.1016/j.tca.2004.11.008

11

11

J. Ghazouani
J. Ghazouani
et al.
et al.
/ Thermochimica
/ Thermochimica
Acta
Acta
432432
(2005)
(2005)
1019
1019

Table 1
Characteristics of the considerd fluids
Fluid

Tc (K)

Ar
CH4
C 2 H6
Kr
Ne
N2
O2
Xe
CO2

Pc (kPa)

150.687
190.564
305.33
209.48
44.4918
126.192
154.581
289.734
304.1282

4863.0
4599.2
4871.8
5510.0
2678.6
3395.8
5043.0
5840.0
7377.3

2. Data sources

Zc

w [3]

TB (K)

0.2895
0.2863
0.2788
0.2918
0.3032
0.2894
0.2879
0.2893
0.2746

0.00219
0.01142
0.0993
0.001714
0.0387
0.0372
0.0222
0.00363
0.22394

408
508
747
579
119
327
405
798
722

M is the Maxwell criterion at the vapourliquid equilibrium:

The (, P, T ) properties of the pure fluids, used in the

following fitting procedure, are taken from the NIST
compi- lation [3]. In particular, we used the following state
data:
gas phase:
Z(T,, );
(Z
(a, Tsat,k

(cal)
liquid phase: Z(T,(G));
vapourliquid equilibrium: Psat (T ), (G) (T ), (L) (T
);
gasliquid critical point: (Zc , Tc , c

sat

nsat

M =

L
sat
G
sat

(Z(cal) (a, Tsat,k , )/) d

k=1

) Z
(a,
sat,kT (cal)

sat,k , (L)
sat,k ))

(7)

sat

).
The second virial coefficients are calculated either with
their
most recent equations of state as given in [4]: Ar [5], CH4
[6], C2 H6 [7], Ne [8], N2 [9], O2 [10] and CO2 [11] or
taken from the compilation in [12] (Kr, Xe).
Tables 2 and 3 give the ranges of used data groups and
the number of data points for each substance.

Table 2
Ranges of data groups and number of data values for each considered
substance
Fluid

Data group

Range of T (K)

Range of P (kPa)

Number of
values

Ar

G-data
L-data
VLE-data
V-data
G-data
L-data
VLE-data
V-data
G-data
L-data
VLE-data
V-data
G-data
L-data
VLE-data
V-data
G-data
L-data
VLE-data

85700
85150
84150
84700
100600
100190
100190
100600
180625
180305
179305
180800
116800
116206
116209
116800
642000
64124
64126

8100000
800100000
70.4474734.6

3584
328
34
125
1530
469
43
101
1715
523
43
125
3896
440
46
138
9897
308
59

V-data
G-data
V-data
G-data
L-data
VLE-data
V-data
G-data
L-data

642000
25700
25700
801000
80150
80154
801000
162800
162287

CH4

3. Numerical procedure
The proposed method is based on the minimization of
the sum of the pondered quadratic residues, K , of the
individual data groups (K is: G for gas; L for liquid; VLE
for vapour liquid equilibrium; V for second virial
coefficient) as objective function of the parameters vector a = (a1 , . . . , a19
):
(a) = wG G + wL L + wVLE VLE + wV V
(3)

C2 H6

Kr

N2

8100270
2375100000
34.3764518.6
870000
100070000
74.3304837.8
8100000
800100000
74.6245443.4
8100000
800100000
14.6023364.5

where
nvap
i

i=1
nliq

(L)

[Zj
j=1
nsat

VLE =
)]2

{[(P
k=1

(a, Ti , i

)]

(4)
(c

(a, Tj , j )]2
Z (5)
j

Ne
O2

Xe
(L)
(cal)
(a, sat,k ,
sat,k

Tk ) Psat (Tk ))/Psat (Tk

+ [(P (cal) (a,

(Tk )]2

8100000
882000
80082000
30.1234930.7
8100000
800100000
(G)

391
3691
136
4818
360
66
185
3528
635

, Tk ) Psat (Tk ))/Psat

1
2
CO2

J. Ghazouani
J. Ghazouani
et al.
et al.
/ Thermochimica
/ Thermochimica
Acta
Acta
432432
(2005)
(2005)
1019
1019
289
84.8495752.4
VL
V-data
162850
Edata
G-data
2171100
162
L-data
217302

VLE-data
217304

1
2

63
8100000
1000100000
527.227355.5

140
5576
445
45

1
3

J. Ghazouani
J. Ghazouani
et al.
et al.
/ Thermochimica
/ Thermochimica
Acta
Acta
432432
(2005)
(2005)
1019
1019
M

+ 2 }

(6)

1
3

1
4

V-data

2171100

J. Ghazouani
J. Ghazouani
et al.
et al.
/ Thermochimica
/ Thermochimica
Acta
Acta
432432
(2005)
(2005)
1019
1019
179

1
4

The last term in Eq. (3) takes into account the residues of
the second virial coefficients:
nV

(cal)

[BV (Tl ) B V (a, Tl )]2 .

V =

objective function in the form

(a, ) = G wG G + L wL L + VLE wVLE VLE
+ V wV V + CP {1 f1 + 2 f2 + 3 f3 }

(8)

l=1

+B 4 f4

Group
weight
factorsamong
wi aretheused
in Eq.data
(3) for
ing mutual
weights
different
sets balanc(K =
G, L, VLE, V ). They are calculated in the same manner as
in [2].
Four constraints are imposed on the minimum of : the
three usual conditions at the critical point, constraints f1 ,
f2 and f3 :
f1 = [Z(a, Tc , c ) Zc ] = 0

By
affecting
values may
1 or be
0 toretained
the coefficients
a parK , or
ticular
data (
group
in Several
(K =1)
rejected
the calculations.
K = 0) with
tions offrom
data groups
various vectors
{K },combina(K =
G, L, VLE, V, CP, B) are
considered.
The set of linear equations can be rearranged and written
in matrix form as
X =
(16)

(9)

f2 = {P (a, T, V )/V }T,CP = 0

(10)

f3 = {2 P (a, T, V )/V 2 }T,CP = 0

(11)

and that of the Boyle temperature TB

f4 = BV (a, TB ) = 0.

(12)

with
X = {a1 , a2 , a19 , 1 , 2 , 3 , 4 }T
(17)
and

m dfm (a) = 0.

= (1 , 23 )T

(18)

23,1 23,23

(13)

The vector X (Eq. (17)) is found by solving numerically

Eq. (16) using the Harwell subroutine MA29D [14] for a
symetric matrix [15]. (Tables 35)
To test the predictive capability of the Bender equation for
the description of the vapourliquid equilibrium, we com(G)
(L)
pare the saturation properties (Psat ,cal , sat,cal , sat,cal ) at a
selected temperature T as calculated by solving the set of
the two non-linear equations with the resulting parameters
a = {ai }:

Using the Lagrangian multipliers method with the four multipliers 1 , 2 , 3 and 4 for the four constraints, we
obtain a system of 19 equations in 23 unknowns
d (a) +

1,1 1,23
.
. ,
=
.

The optimization problem is now stated as follows

min (a)
fm (a) = 0, m = 1, . . . , 4

(15)

(14)

m=1,4

The missing 4 equations are the imposed constraints

(Eqs. (9)(12)).
In order to investigate the influence of the individual
data
groups on the predictive quality of the EOS, we write the

(G)

(L)

P (a, T, sat,cal ) = P (a, T, sat,cal )

(19)

Table 3
Results of selected vectors {K } for nitrogena
i
G

G
L

All data
1
1
G-data
1
0
1
1
1
1
1
1
1
1
1
1
L-data
0
1
0
1
1
1
VLE-data
0
0
0
0

VLE

CP
3

mol1 )

VLE

CP

r Z (%)

r Z (%)

r PL (%)

r PG (%)

r BV (%)

BV (m

0.29

0.003

0.52

0.002

0.61

0.24

5.75

0.001

0
0
1
1
1
1

0
0
0
1
1
0

0
0
0
0
0
1

0
0
0
0
1
0

0.06
0.11
0.23
0.24
0.26
0.23

0.000
0.001
0.002
0.003
0.003
0.002

***
0.32
0.40
0.4
0.44
0.33

0.637
0.000
0.002
0.002
0.002
0.001

***
10.74
0.52
0.55
0.55
0.50

0.57
1.42
0.17
0.09
0.10
0.27

6.33
11.20
19.90
7.73
4.48
30.71

0.001
0.002
0.002
0.001
0.001
0.003

4.11
8.28
0.71
1.17
1.29
0

0
1
0

0
0
0

0
0
1

0
0
0

83.19
3.71
0.30

0.982
0.042
0.001

0.07
0.18
0.63

0.000
0.000
0.001

0.41
0.53
0.72

53.51
0.08
0.31

***
***
10.51

0.641
0.045
0.001

30.86
0.39
0

1
1

0
0

0
1

0
0

17.23
23.81

0.227
0.318

78.96
92.89

1.218
1.553

0.04
0.09

0.01
0.04

***
***

0.092
0.104

0.19
0

Values of deviations equal to or more than 100% are denoted by ***.

r Zc (%)

Table 4
Confidence intervals and standard error for each parameter of the Bender
EOS in the case of nitrogen for the fit with ={1, 0, 0, 0, 0, 0}
Parameters

Estimate

aa1 103
2
2
a3 104
a4 106
a5 105
a6 10
a7 104
a8
a9 109
a10 105
a11 1011
a12 108
a13 1011
a14 102
a15 104
a16 105
a17 104
a18
a19

9.66475
0.242837
12.3958
12.0759
4.83593
1.3997
6.15479
0.194958
7.60305
1.060184
2.79557
1.61611
1.50808
2.30933
10.05986
85.49781
4.30436
0.137114
0.192710

CI

4. Results
In order to specify the adequate ranges of temperature
and pressure for each pure substance, we fit the individual
data groups separately by affecting different values to the
coefficients {K } = {G , L , VLE , V , CP , B } in the
expres-

S.E.

9.60608;
9.67258
0.315011;
0.219469
12.4736; 12.1351
11.6828; 12.1928
4.89985; 4.64306
1.38824; 1.45372
8.41118; 1.135
0.200924; 0.18034
9.31186; 7.0441
0.864239; 1.15428
2.75756; 2.97717
1.75178; 1.37629
1.4038; 1.56836
2.35535; 2.18602
9.78314; 11.0648
91.8327; 83.564
5.27453; 2.77599
0.092618; 0.161308
0.2193113; 0.187131

0.00001
0.02437
8.63436
1301.01
65501.6
1.67035 107
0.00019
0.00525
5.7845 1010
7.3982 107
5.6018 1013
9.5777 1010
4.1976 1013
14.522
3269.3
210915
0.00006
0.01752
0.0250

sion
of the
objective
(a,phase
) (Eq.
For
example,
if we
considerfunction
only thegas
data,(15)).
the terms

K, K = G
are rejected from the objective function by affecting zero
to all K coefficients except for G . The adjustable parameters resulting from the fit with {K } = {1, 0, 0, 0, 0, 0} are
then used to calculate the gas phase properties which are
then
compared to their data values [3]. Similarly are treated the
ho- mogeneous liquid phase and the vapourliquid
equilibrium,
with {K } = {0, 1, 0, 0, 0, 0} and {K } = {0, 0, 1, 0, 0, 0},
respectively. The results obtained for the considered substances show a satisfactory description of the state
behaviour
( P T ) in the homogeneous fluid regions as well as at
the vapourliquid equilibrium in the specific ranges of tem-

R2 = 0.999982.

M = 0

(20)

with the saturation data. The solution of the set of Eqs. (19)
and (20) is obtained by means of a non-linear procedure
[16].

perature
and pressure
(Tableswith
614).
Several
combinations
of data groups
are considered
various
vectors
{K }. We
can so investigate the influence of fitting with an individual
data group or a combination of data groups on the
properties of the others as well as test if the incorporation of
the consid- ered data set into the calculation is necessary
for a reliable Bender equation.
The deviations of calculated properties from their data
values are expressed by means of the root of the standard
deviation RAADr and AAD, defined for a property as
RAADr /% = 100 N
N |( cal )/|i
(21)

Table 5
Parameters of the Bender equation of state of nitrogen for selected vectors
{K }
G
L
VLE
V
CP
B
a2
a1 1022
a3 10
a4 104
a5 105
a6 105
a7 103
a8
a9 109
a10 105
a11 1011
a12 108
a13 1011
a14 102
a15 104
a16 105
a17 104
a18
a19
a20

1
0
0
0
0
0
0.2428
0.9665
12.3958
12.0759
48.3594
1.3997
0.6155
0.1949
7.6030
1.0602
2.7956
1.6161
1.5081
2.3093
10.0598
85.4978
4.3043
0.1371
0.1927
c

1
1
1
0
0
0
1.7775
1.0132
3.4055
0.8036
1.9298
2.0374
8.0088
0.3743
9.3718
2.1536
4.8521
3.6394
2.4310
0.1048
0.9865
6.5933
3.191
0.1084
0.4338
c

1
1
1
0
1
0
1.3923
1.0052
3.5533
1.0208
2.2115
1.9125
9.2313
0.4509
5.8968
2.1982
4.9045
3.6235
2.3827
0.0845
1.2281
7.4913
3.9388
0.1300
0.8534
c

i=1

AAD =

| cal |i

(22)

i=1

where N is the number of data points.

In the following, we will discuss in detail the case of nitrogen as illustration of the procedure adopted. For the rest
{K } = {the
eters
1, 1,resulting
1, 1, 1, 1} as
of
theaconsidered
fluids, fit
wewith
just present
i for the general
well
as
the
corresponding
RAAD
and
AAD
(Tables
614).
paramr
An analytical estimation of the standard error as well as the
confidence
intervals of each parameter in the case of
nitrogen
0}), showsfor the fit of single phase gas ( = {1, 0, 0, 0, 0,
that at the most two figures of each parameter are significant (Table 4). Through this study, approximative compact
list of fitting parameters is given for each considered substance (Tables 514). As shown in Tables 614, the use of
the new compact list does not affect the predictive quality
of the volumetric properties.
4.1. Correlations of data groups for
nitrogen
Table 3 is structured
into several
items,vectors
where ,
selected
combinations
of data groups
for various
K =

Table 6
Parameters of the Bender EOS and statistical characteristics for a rgon for the fit with {K } ={1, 1, 1, 1, 1, 1}
Parameters
a1 = 5.1106 103 , a2 = 1.2870, a3 = 3.3752 102 , a4 =
5.2246 103 , a5 = 1.5031 105
a6 = 6.7915 106 , a7 = 1.5428 103 , a8 = 0.1344, a9 =
1.7997 109 , a10 = 2.3236 106

a11 = 5.5901 1012 ,

a12 = 3.1318 109 ,
a13 = 1.5332
1012 , a14 = 1.0057 102 , a15 = 2.1466 104
a16 = 1.1965 106 , a17 = 4.3144 104 , a18 = 0.1222, a19 =
6.3172, a20 = 5.356 102

Statistical characteristics
G-data
L-data
V-data

wG = 1.0, r Z = 0.31%, Z = 0.003

wL = 10.92, r Z = 0.44%, Z = 0.002
wV = 28.67, r BV = 12.68%, BV = 0.002

VLE-data

wVLE = 105.41,
(G)
r sat = 0.96%

r PL = 1.3%,

r PG = 0.17%,

(L)

r sat = 0.17%,

Table 7
Parameters of the Bender EOS and statistical characteristics for methane for the fit with {K } ={1, 1, 1, 1, 1, 1}
Parameters
a1 = 1.9019 102 , a2 = 6.5728, a3 = 1.5739 103 ,
a4 = 2.1534 103 , a5 = 1.4211 105
a6 = 6.5279 105 , a7 = 1.9498 102 , a8 = 2.4709,
a9 = 6.2110 108 , a10 = 1.2398 104

a11 = 6.7980 1010 ,

a12 = 5.5187 107 ,
a13 = 8.3738
1010 , a14 = 9.0544 102 , a15 = 1.8231 105
a16 = 1.4017 107 , a17 = 4.9113 102 , a18 = 2.0014 101 ,
a19 = 1.2579 103 , a20 = 1.6266 102

Statistical characteristics
G-data, wG = 1.0
L-data, wL = 3.26
V-data, wV = 15.14

r Z = 0.21%, Z = 0.002
r Z = 0.33%, Z = 0.003
r BV = 8.03%, BV = 0.001

VLE-data

wVLE = 35.58, r PL = 0.56%,

(G)
r sat = 0.97%

(L)

r PG = 0.13%,

r sat = 0.16%,

Table 8
Parameters of the Bender EOS and statistical characteristics for ethane for the fit with {K } ={1, 1, 1, 1, 1, 1}
Parameters
a1 = 2.2048 102 , a2 = 14.8680, a3 = 4.9390 102 ,
a4 = 5.6568 105 , a5 = 3.0515 107
a6 = 6.9231 106 , a7 = 1.0396 102 , a8 = 5.4227,
a9 = 1.6384 107 , a10 = 1.02 104

a11 = 1.2429 1010 ,

a12 = 2.8726 108 ,
a13 = 2.4845
1010 , a14 = 4.1160 102 , a15 = 5.2327 104
a16 = 6.6722 107 a17 = 0.1286, a18 = 6.3893 101 , a19 =
3.6942 103 , a20 = 2.070 102

Statistical characteristics
G-data
L-data
V-data

wG = 1.0, r Z = 0.54%, Z = 0.005

wL = 3.27, r Z = 0.26%, Z = 0.002
wV = 13.72, r BV = 3.04%, BV = 0.001

VLE-data

wVLE = 39.88, r PL = 0.77%,

(G)
r sat = 0.58%

r PG = 0.14%,

(L)

r sat = 0.09%,

Table 9
Parameters of the Bender EOS and statistical characteristics for krypton for the fit with {K } ={1, 1, 1, 1, 1, 1}
Parameters
a1 = 3.1860 103 , a2 = 1.3887, a3 = 2.4604 102 ,
a4 = 1.1092 104 , a5 = 3.8876 105
a6 = 1.5687 106 , a7 = 3.0953 104 , a8 = 0.1022,
a9 = 4.5100 1010 , a10 = 1.7482 107

a11 = 4.9629 1013 , a12 = 4.0074 1010 , a13 = 1.4205

1013 , a14 = 2.2258 101 , a15 = 1.9302 103
a16 = 5.2138 105 , a17 = 1.0767 104 , a18 = 3.4194 102 ,
a19 = 1.1637, a20 = 9.084 102

Statistical characteristics
G-data
L-data
V-data

wG = 1.0, r Z = 0.31%, Z = 0.003

wL = 8.85, r Z = 0.36%, Z = 0.002
wV = 28.23, r BV = 11.98%, BV = 0.002

VLE-data

wVLE = 84.69, r PL = 0.85%,

(G)
r sat = 0.57%

r PG = 0.06%,

(L)

r sat = 0.14%,

Table 10
Parameters of the Bender EOS and statistical characteristics for neon for the fit with {K } ={1, 1, 1, 1, 1, 1}
Parameters
a1 = 6.1712 103 , a2 = 0.2386, , a3 = 7.5748 101 ,
a4 = 2.2144 103 , a5 = 2.6381 104
a6 = 2.6707 106 , a7 = 1.0554 103 , a8 = 7.9161
103 , a9 = 2.9108 109 , a10 = 1.4130 106

a11
a14
a16
a19

Statistical characteristics
G-data
V-data

wG = 1.0, r Z = 0.01%, Z = 0.001

wV = 27.13, r BV = 3.47%, BV = 0.000

= 2.201 1012 , a12 = 6.2366 1010 , a13 = 2.81 1013 ,

= 2.0708 101 , a15 = 2.0279 103
= 5.0109 104 , a17 = 5.6119 106 , a18 = 7.6984 104 ,
= 3.0147 102 , a20 = 4.8191 102

Table 11
Parameters of the Bender EOS and statistical characteristics for nitrogen for the fit with {K } ={1, 1, 1, 1, 1, 1}
Parameters
a1 = 1.0809 102 , a2 = 2.6663, a3 = 2.6713 102 ,
a4 = 6.0196 103 , a5 = 1.7340 105
a6 = 1.5194 105 , a7 = 2.7664 103 , a8 = 0.2643,
a9 = 6.8534 109 , a10 = 6.2324 106

a11 = 3.7714 1011 ,

a12 = 1.8692 108 ,
a13 = 1.7616
1011 , a14 = 1.4065 102 , a15 = 2.5755 104
a16 = 1.3066 106 , a17 = 1.6227 103 , a18 = 0.4203, a19 =
1.8948 101 , a20 = 3.1330 102

Statistical characteristics
G-data
L-data
V-data

wG = 1.0, r Z = 0.29%, Z = 0.003

wL = 32.13, r Z = 0.52%, Z = 0.002
wV = 25.31, r BV = 5.75%, BV = 0.001

VLE-data

(L)

wVLE = 167.74, r PL = 0.61%, r PG = 0.24%, r sat = 0.12%,

(G)
r sat = 0.91%

Table 12
Parameters of the Bender EOS and statistical characteristics for xenon for the fit with {K } ={1, 1, 1, 1, 1, 1}
Parameters
a1 = 2.5349 103 , a2 = 1.3864, a3 = 4.9994 102 ,
a4 = 7.0781 103 , a5 = 5.6026 105
a6 = 1.2329 107 , a7 = 5.9897 104 , a8 = 0.1013,
a9 = 1.3338 109 , a10 = 7.7897 107

a11 = 7.1791 1014 , a12 = 1.0748 1010 , a13 = 3.2873

1014 , a14 = 2.7647 101 , a15 = 1.3347 103
a16 = 3.9274 106 , a17 = 1.589 104 , a18 = 6.6339 102 ,
a19 = 2.1472, a20 = 11.00 102

Statistical characteristics
G-data
L-data
V-data

wG = 1.0, r Z = 0.43%, Z = 0.004

wL = 5.55, r Z = 0.41%, Z = 0.002
wV = 25.2, r BV = 5.54%, BV = 0.002

VLE-data

G, L, VLE, V, CP, B are considered. From the fit with

=
{1, 0, 0, 0, 0, 0}, it is obvious that the volumetric properties
(Z, P) for the gas phase are mutually consistent, with an
overall RAADr lower then 0.06%, but the liquid phase region is not described correctly since the predicted values
are more than 100% apart from the data. On the contrary,
the

(L)

(G)

wVLE = 56, r PL = 0.79%, r PG = 0.04%, r sat = 0.11%, r sat =

0.55%

use of L-data alone, corresponding to = {0, 1, 0, 0, 0, 0},

results in a good description of the liquid phase properties,
but leads to significant deviations in the gas phase region
with an overall RAADr for the compressibility factor ZG
and the pressure PG
higher than 83 and 53%,
respectively. When
both of these data sets are considered, = {1, 1, 0, 0, 0, 0},

Table 13
Parameters of the Bender EOS and statistical characteristics for oxygen for the fit with {K } ={1, 1, 1, 1, 1, 1}
Parameters
a1 = 1.0286 102 , a2 = 3.735, a3 = 1.5354 102 ,
a4 = 9.1063 103 , a5 = 2.3839 105
a6 = 3.5843 106 , a7 = 2.8114 103 , a8 = 0.1379,
a9 = 2.5432 108 , a10 = 6.5938 106

a11 = 2.7781 1012 ,

a12 = 1.4746 109 ,
a13 = 2.2205
1012 , a14 = 1.0410 102 , a15 = 2.2781 104
a16 = 1.0584 106 , a17 = 6.2873 104 , a18 = 0.1909, a19 =
8.5967, a20 = 4.361 102

Statistical characteristics
G-data
L-data
V-data

wG = 1.0, r Z = 0.22%, Z = 0.002

wL = 13.38, r Z = 0.66%, Z = 0.004
wV = 26.04, r BV = 9.24, BV = 0.001

VLE-data

(L)

(G)

wVLE = 73, r PL = 1.26%, r PG = 0.15%, r sat = 0.11%, r sat =

0.71%

Table 14
Parameters of the Bender EOS and statistical characteristics for carbon dioxide for the fit with {K } ={1, 1, 1, 1, 1, 1}
Parameters
a1 = 9.2695 103 , a2 = 5.6387, a3 = 3.7608 102 ,
a4 = 2.9838 105 , a5 = 1.4837 107
a6 = 4.1742 106 , a7 = 1.5959 103 , a8 = 1.1767,
a9 = 1.1034 108 , a10 = 8.3435 106

a11 = 4.9083 1012 , a12 = 9.2096 1010 , a13 = 3.7029

1012 , a14 = 1.7365 103 , a15 = 1.1409 106
a16 = 1.8758 108 , a17 = 2.0563 104 , a18 = 1.2645, a19 =
5.3148 102 , a20 = 4.676 102

Statistical characteristics
G-data
L-data
V-data

wG = 1.0, r Z = 0.09%, Z = 0.000

wL = 12.53, r Z = 0.29%, Z = 0.001
wV = 31.15, r BV = 3.01%, BV = 0.000

VLE-data

(L)

wVLE = 123.91, r PL = 0.45%, r PG = 0.09%, r sat = 0.09%,

(G)
r sat = 0.17%

the liquid and the gas phases are correctly described but
the vapourliquid equilibrium is not accurately predicted,
with a relative deviation of about 11%. Similarly, the evaluation of the Bender equation parameters with VLE = 1, re-

the second virial coefficients data will be shown in the next

paragraph by the prediction of the JouleThomson
inversion curve.

sults
in a goodindescription
of the
as illustrated
Table 3. The
fit vapourliquid
with = {1, 1,equilibrium
1, 0, 0, 0}
leads to an overall good agreement with the basic data for
the homogeneous phases as well as the saturation properties
with an overall RAADr ranging from 0.2% to 0.5%. We can
then conclude on the necessity of incorporating each data
group for the obtention of a reliable equation of state describing accurately the state behaviour of the homogeneous
fluid regions as well as the vapourliquid equilibrium. A
good description of the second virial coefficient is obtained
from the fit of single phase gas data ( = {1, 0, 0, 0, 0, 0}),
which is not the case of the other combinations of data

4.2. JouleThomson inversion curve from the

Bender
EOS

groups
to the second
virial coefficients
data
(Vwithout
= B reference
= 0). However
the inclusion
of virial
coefficients
1, 1, 1, 1, 1}for the fit with = {1, 1, 1, 1, 0, 1} or = {1,
leads to improved description of this property with an overall RAADr ranging from 4.5% to 5.7%. Large deviation at
the critical point is observed (r Zc = 30.7%) for the fit with
= {0, 1, 0, 0, 0, 0}. On the other hand, the inclusion of the
coordinates of the critical point with CP = 1 does not
affect
significantly the predictive capability of the equation for the
other
data groups
but leads
to a1,good
critical point.
The fit with
= {1,
1, 0,description
1, 0} leads at
to the
a good
description of both single phase regions, the vapourliquid
equilibrium and the critical point, with a low relative
average deviation (00.5%) but the second virial coefficient
descrip- tion is worsened. We see in Table 5 that the
parameters ai
obtained from the fit with = {1, 1, 1, 0, 0, 0} do not differ
significantly from the fit with inclusion of the coordinates
of
the critical point. The case of the last data set, the second
virial coefficients, is worthy of closer consideration. From
Table 3 it is seen that equal good prediction of the PVT
properties and the critical point are performed with the
Bender equation,
with reference, = {1, 1, 1, 1, 1, 1}, and without reference,
= {1, 1, 1, 0, 1, 0}, to the second virial coefficients,
which
is
case
of for
virial
with
relative deviationnot
of the
about
31%
V coefficients,
= B = 0. The
reala importance
of

The JouleThomson inversion curve (JTIC) is the locus

of states of the thermodynamic surface where the fluid temperature is invariant upon isenthalpic expansion:
(T/P )H = 0
(23)
This criterion can be written in several alternative forms, in
particular in the following form suitable for the Bender
EOS
Tr (Zr /Tr )r r (Zr /r )Tr = 0
(24)
JTICs are usually represented in the reduced temperature (Tr := T/ Tc ) and reduced pressure (Pr := P/Pc ) plane,
where Tc and Pc are, respectively, the critical temperature
and pressure. The JTIC passes through a maximum
pressure
at intermediate tempearture and goes to zero at a maximum
inversion temperature.
It is known that the prediction of the JouleThomson inversion curve constitues a rather severe test of an equation
of state.
Fig. 1 shows the nitrogen inversion curves for two
selected fits in comparison to published data. Curve c1
results when
using all data groups, = {1, 1, 1, 1, 1, 1}, while curve c2
Table 15
Characteristics of the JouleThomson inversion curve for each considered
substance
Fluid

Tr

Pr

Tr,max

Ar
CH4
C2 H6
Kr
Ne
N2
O2
Xe
CO2

2.26
2.26
2.16
2.24
2.38
2.22
2.22
2.20
1.92

11.7373
11.6574
12.1326
11.5088
11.137
11.5354
11.474
11.3985
12.4181

5.04
5.0
4.72
5.22
4.9
4.9
5.1
5.16
4.5

Fig. 1. Comparison between calculated and published inversion data for nitrogen.

is
predicted
the fit with
=Up
{1,to1,T1,=0,3.5,
1, 0}
without
the
second
virialby
coefficients
data.
a good
agreer
ment between both curves is observed. But the behaviour of
c2 becomes erratic for larger values of the reduced temperature. On the contrary, inversion curve c1 agrees with all the
published inversion data [4,9,1719] of the fluid on the
whole reduced temperature range. A good prediction of the
Tr,max is also observed.
We have calculated the inversion curves c1 and c2 for all
considered fluids and made the same observation: in each
case the inversion curve predicted by the Bender equation
with reference to the second virial coefficient is more
realis- tic than the one calculated without reference to
these data. A further illustration of the good prediction of
the inver- sion curve c1 is seen in Fig. 2 for the case of
carbon diox- ide. In Table 15 the coordinates of the peak of
the inversion curves and the maximum inversion
temperature Tr,max as pre-

dicted by the Bender equation are given for the considered

substances.
5. Correlation of the inversion data of low acentric
factor uids
It is known that the inversion effect obeys the
correspond- ing state principle. In order to establish a
corresponding state correlation for the seven low acentric
factor fluids consid- ered here (Ar, CH4 , Kr, Ne, N2 , O2 ,
Xe), we applied the structural optimisation technique
developed by Wagner that we implemented earlier [21] in
Mathematica (version 4.1). The simple correlation
obtained:
Pr = 31.541 58.389 eTr 6.194Tr
(25)
is very similar to the universal equation correlating most of
the published inversion data for low acentric factor gases

Fig. 2. JTIC of carbon dioxide predicted by: ( ) the Bender EOS; ( ) experimental data [20]; ( ) compressibility factor [20] and () the span Wagner
EOS [20].

Fig. 3. Comparison between the JTIC predicted by the correlation of Eq. (25) and the published model of Eq. (26) as well as the used data.

established by our research group recently [21]:

Pr = 32.771 60.859 e

T r

6.535Tr

(26)

As shown in Fig. 3, the simple correlation of Eq. (25) describes adequately the used data for the set of considered
fluids as well as the inversion data predicted by the model
of
Eq. (26).

R
S
S.E.
T
V
Z
w

universal gas constant

entropy
standard error
temperature
volume
compressibility factor
weighing factor

Greek letters
6. Conclusion
The evaluation of the Bender equation parameters yields
satisfactory results of the description of the state behaviour
of homogeneous fluid regions as well as the vapourliquid
equilibrium curve. The validity of the equation is tested by
predicting the JouleThomson inversion curve. A
substantial result of this study is to show the importance of
incorporating a derivative property like the second virial
coefficient for the correct evaluation of the EOS parameters.
The inversion data predicted by the Bender EOS for several
low acentric factor fluids (Ar, CH4 , Kr, Ne, N2 , O2 , Xe)
are correlated with a simple model in terms of reduced
pressure as a function of reduced temperature.
Nomenclature
ai
BV
CI
f
H
M
N,
P

adjustable parameters of the Bender equation

second virial coefficient
confidence intervals
fugacity
enthalpy
molar mass
n number of values
pressure

mass density
objective function
Lagrangian multipliers
partial derivative
acentric factor

Subscripts
B
cal
CP, c
G, vap
L, liq
r
sat
VLE
V

Boyle
calculated
critical point
gas phase
liquid phase
relative
at saturation
vapourliquid equilibrium
second virial coefficient

Superscripts
cal
f,
g,

calculated
L liquid phase
G gas phase

References
[1] E. Bender, Cryogenics 15 (11) (1975) 667673.
[2] I. Cibulka, J. Kovac ikova, L. Hnedkovsky , J.P. Novak, Fluid
Phase
Equilib. 180 (2001) 2740.
[3] NIST-Server http://webbook. nist. gov/Chemistry/.
[4] R. Tillner-Roth, Fundamental Equations of State, Shaker Verlag,
Aachen, Germany, 1998.
[5] C. Tegler, R. Span, W. Wagner, Eine neue Fundamentalgleichung
fu r das fluide Zustandsgebiet von Argon fu r Temperaturen von der
Schmel- zlinie bis 700 K und Dru cke bis 1000 MPa, Fortschr. -Ber.
VDI Reihe
3, Nr. 480, VDI-Verlag, Du sseldorf, 1997.
[6] U. Setzmann, W. Wagner, J. Phys. Chem. Ref. Data 20 (1991) 1061.
[7] D.G. Friend, H. Ingham, J.F. Ely, J. Phys. Chem. Ref. Data 20 (1991)
275.
[8] R.S. Katti, R.T. Jacobsen, R.B. Stewart, M. Jahangiri, Adv. Cryo.
Eng.
31 (1986) 1189.
[9] R.T. Jacobsen, R.B. Stewart, M. Jahangiri, J. Phys. Chem. Ref. Data
15 (1986) 735.
[10] R. Schmidt, W. Wagner, Fluid Phase Equilibria 19 (1985) 175.
[11] R. Span, W. Wagner, J. Phys. Chem. Ref. Data 25 (1996)
1509.

[12] H.V. Kehiaian, Virial coefficients of selected gases, Handbook of

Chemistry Physics, CRCnet Base, 1999.
[14] Harwell Subroutine Library, HSL Archive Package, UK.
[15] R. Fletcher, J. Linear Algebra Appl. 14 (3) (1976).
[16] M. Daroux, Analyse Numerique Appliquee, LSGC, ENSIC,
Nancy, France, 1991.
[17] R.H. Perry, C.H. Chilton, Chemical Engineers Handbook, 5th ed.,
McGraw-Hill, 1973, pp. 8288.
[18] J.R. Roebuck, H. Osterberg, Phys. Rev. (1934) 46.
[19] W.E. Deming, L.W. Deming, Phys. Rev. 46 (1935) 448.
[20] C.M. Colina, M. Lsal, F.R. Sipersrein, K.E. Gubbins, Fluid Phase
Equilib. 202 (2002) 253262.
[21] K. Fnena, Kh. Mejbri, A. Bellagi, Journal de la societe chimique de
Tunisie 5 (1) (2003) 95106.

Further reading
[13] J. Vidal, Thermodynamique: Application au genie chimique et a`
lindustrie petrolie`re, Technip, France, 1997.