You are on page 1of 16

Fluid Phase Equilibria 161 Ž1999.

241–256

Vapor–liquid equilibrium, fluid state, and zero-pressure solid


properties of chlorine from anisotropic interaction potential by
molecular dynamics
´ Karel Aim
Martin Lısal, )

´ Laboratory of Thermodynamics, Institute of Chemical Process Fundamentals, Academy of Sciences,


E. Hala
165 02 Prague 6, Czech Republic
Received 1 September 1998; accepted 24 March 1999

Abstract

Extensive examination of the anisotropic interaction potential of chlorine by Rodger et al. wP.M. Rodger, A.J.
Stone, D.J. Tildesley, J. Chem. Soc., Faraday Trans. 2, 83 Ž1987. 1689–1702x Žwith interaction sites located at
the positions of atoms in a molecule and the electrostatic part found by ab initio calculations. for its predictive
power has been performed. We have calculated Ži. the second virial coefficient by using a non-product
algorithm, Žii. a series of liquid-phase state points in the temperature and pressure ranges of 200 to 400 K and 0
to 6.2 MPa, respectively, by the constant pressure–constant temperature molecular dynamics simulations, Žiii.
vapor–liquid equilibrium and heat of vaporization from the triple point Ž172 K. to 300 K by the Gibbs–Duhem
integration method combined with simultaneous Žbut independent. constant pressure–constant temperature
molecular dynamics simulations of the vapor and liquid phases, and Živ. the properties of the zero-pressure
crystal structures by molecular dynamics technique due to Parinello and Rahman wM. Parrinello, A. Rahman,
Phys. Rev. Lett. 45 Ž1980. 1196–1199x. Generally, good to excellent agreement of the calculated properties with
the corresponding values for real chlorine was observed. The results obtained from the investigated interaction
potential are equivalent to Žor even better than. those reported for a more complicated potential by Wheatley and
Price wR.J. Wheatley, S.L. Price, Mol. Phys. 71 Ž1990. 1381–1404x. q 1999 Elsevier Science B.V. All rights
reserved.

Keywords: Chlorine; Intermolecular potential; Molecular simulation; Vapor–liquid equilibria; Vapor pressure; Density

1. Introduction
Prediction of thermodynamics and phase equilibria of real substances from the knowledge on
molecular interactions is one of the important goals of computer simulations with a significant

)
Corresponding author. Tel.: q42-2-2039-0300; fax: q42-2-2092-0661; e-mail: kaim@icpf.cas.cz

0378-3812r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved.
PII: S 0 3 7 8 - 3 8 1 2 Ž 9 9 . 0 0 1 8 8 - 0
242 ´ K. Aim r Fluid Phase Equilibria 161 (1999) 241–256
M. Lısal,

industrial impact w1x. Sufficiently accurate intermolecular potential model plays a key role in practical
application of such computer simulations of the thermodynamic properties and phase equilibria of real
substances. For the evaluation of complete phase diagram, the intermolecular potential model must
simultaneously reproduce the properties of vapor, liquid, and solid phases.
Two types of intermolecular potential models are in use: isotropic site–site potentials w2,3x and
anisotropic site–site potentials w4–6x. Commonly used isotropic site–site potentials represent the
repulsion–dispersion forces acting between molecules as sum of Lennard–Jones or Buckingham
potentials between particular atoms in molecules. Additional electrostatic interactions can then be
included in one of two ways: Ž i. a number of ideal multipoles is placed at the center of mass of the
molecules and moments on different molecules interact through the appropriate terms in the multipole
expansion w7x, or Ž ii. partial charges are distributed within the repulsive core of the molecules and
their position and magnitude are chosen so as to represent a number of the low non-zero electrostatic
moments of the molecule; the partial charges on different molecules interact via a Coulomb potential
w8x. Although the isotropic site–site potential provides a good description of thermodynamic and
microscopic properties w2,3x as well as vapor–liquid equilibria Ž VLE. w9x for chlorine, there are certain
features of the solid and liquid phases which appear unable to be explained by using the isotropic
site–site potentials. These are, in particular, the stability and properties of the observed crystal
structure w10x, and peaks and heights of the liquid structure factor w4,5x.
Anisotropy introduced into intermolecular potentials of chlorine by Rodger et al. w4,5x Ž RST. and
Wheatley and Price w6x Ž WP. significantly affects both the short and long range intermolecular forces
and therefore it plays an important role in determining the physical properties of chlorine. The
anisotropic site–site potentials are known to be able to overcome deficiencies of the isotropic site–site
potentials w4–6x, and thus they should in principle be able to reproduce simultaneously the properties
of all phases better than the isotropic potentials w2,3x.
In the present work, we carry out extensive test of predictive power of the anisotropic potential of
chlorine w4–6x. Chlorine was chosen because both extensive experimental data are available for the
fluid phase w11x and experimental zero-pressure properties of crystal structure w12x are known. Paper is
organized as follows. In Section 2, we describe the used potential model. Section 3 gives simulation
details. In Section 4, we first test the performance of the used potential model for the vapor and liquid
phases and subsequently we present calculation of VLE by the Gibbs–Duhem integration Ž GDI.
method w13–15x. Section 5 shows simulated properties of the zero-pressure crystal structures obtained
by Parrinello–Rahman molecular dynamics Ž MD. simulations w16,17x, and compares them with
experimental as well as available simulated data w12,6x. Finally, we give conclusions in Section 6.

2. Potential model
Rodger et al. w4,5x, and Wheatley and Price w6x constructed on a semi-empirical basis an effective
anisotropic site–site pair potentials for chlorine. Electrostatic part of the RST and WP potentials was
found from ab initio calculations, using a scheme called distributed multipole analysis. Repulsion–
dispersion part and many-body terms of both potentials were obtained by using ab initio calculations
and experimental data, respectively. The potentials have interaction sites located at positions of atoms.
The WP potential contains an additional interaction site at the center of mass of molecules and
possesses more complicated functional form than the RST potential. In this work we employ the RST
anisotropic potential since it is simpler and has less demand on computer time in comparison with the
´ K. Aim r Fluid Phase Equilibria 161 (1999) 241–256
M. Lısal, 243

WP anisotropic potential. In the following we show that both the potentials represent thermodynamic
properties of chlorine with equivalent accuracy.
The RST model of chlorine molecules consists of two interaction sites at a distance 2 l apart which
interact via the function:
y6
Vag s K exp ya rag y s 0 y s Ž V .
½ 5 yC rag q j 0 y s Ž V . q Ees
s Ž V . s s 110 S110 q s 2 Ž S202 q S022 . q s 3 Ž S303 q S033 .
1
Ees s 2 m 2 S112 rag
y3 y4
q 3 m Q Ž S123 q S213 . rag q 6Q 2 S224 rag
y5
Ž1.
4pe 0
In Eq. Ž1., Ees represents the electrostatic interactions arising from a point dipole m and quadrupole Q
on each atom w18x, rag is the distance between atoms a and g on the two chlorine molecules, V
denotes relative orientation of two chlorine molecules, Sl1 l 2 j are the spherical harmonic functions w19x,
and e 0 is the permittivity of free space. Expressions for the relevant S functions are listed in
Appendix A. The values of all the RST potential parameters w4,5x are summarized in Table 1. The
atomic dipole m and quadrupole Q give zero molecular dipole and molecular quadrupole Q T s 2Ž Q
y 2 l m . equal to ab initio value w4,5x of 12.060758 = 10y40 C m2 .
Potential energy of the system of N chlorine molecules can then be approximated as:
N N
Us Ý Ý Vi j Ž2.
is1 j)i

where:
2 2
Vi j s Ý Ý Vag Ž3.
as1 g s1

is the potential energy between two chlorine molecules i and j.

Table 1
Potential parameters of the Rodger et al. w4,5x anisotropic potential
2 l ŽA˚. 1.994
K ŽkJ moly1 . 1.0
C ŽkJ A ˚ 6 moly1 . 10166.7
m ŽC m. 1.228514=10y30
Q ŽC m2 . 8.480036=10y40
a ŽA ˚ y1 . 3.5100
s 0 ŽA ˚. 3.8630
j 0 ŽA˚. 0.1629
s 110 ŽA ˚. 0.0221
s 2 ŽA.˚ y0.0619
s 3 ŽA ˚. 0.3500

2 l is the bond length, K and C are the energy parameters, a , s 0 , j 0 , s 110 , s 2 and s 3 are the length parameters, m and Q
are the point dipole and quadrupole on chlorine atoms, respectively.
244 ´ K. Aim r Fluid Phase Equilibria 161 (1999) 241–256
M. Lısal,

3. Simulation details
Constant pressure–constant temperature Ž NPT. MD simulations of the fluid phase were carried out
using the Andersen method w8x and those of the solid phase were performed by the Parrinello–Rah-
man method w16,17x. The temperature was kept constant by the isokinetic scaling of translation and
angular velocities after every time step. The equations of translation and cell motions were solved by
the Gear predictor–corrector algorithm of the fifth order. The rotational motion was treated by the
method of quaternions, and it was also solved by the Gear predictor–corrector algorithm of the fifth
order w8x. To calculate forces, torques, pressure and pressure tensor, it is necessary to obtain the
analytical derivative of the RST potential. A method outlined in Ref. w20x was used and it is
summarized in Appendix B.
The minimum image convention, periodic boundary conditions, and cut-offs based on center of
mass separation were used during all MD simulations. Cut-off radius was equal to Ž i. half-box length
for fluid simulations and Ž ii. 90% of minimum half-box length for solid simulations. The long-range
correction of the potential energy, pressure, and pressure tensor was included by using angular
averages of potential and virial functions assuming that the radial distribution function is unity beyond
the cut-off radius w8x. The long range corrections were calculated for a range of cut-off radius ranging
from 3 to 50 A ˚ in steps of 0.05 A.˚ The cut-off correction required for a particular MD step was
obtained from the table by linear interpolation. We utilized the membrane mass Ms 04rm equal to
1 = 10y6 and 5 = 10y4 for the vapor and liquid phase, respectively, the box mass Wrm s 2 Ž m is the
mass of the chlorine molecule. for the solid, and the integration steps D t ranging from 2.5 to 5 fs.
During MD simulations, we also evaluated the diffusion coefficient D:
d Žt.
D s lim Ž4.
t™` 6 t
where
d Ž t . s ² < ri Ž t . y ri Ž0. < 2 : Ž5.
the translation order parameter r Ž k .:
N 2 N 2 1r2
1 1
rŽk. s ½ N
Ý cos Ž k P r i .
is1
q
N
Ý sin Ž k P r i .
is1
5 Ž6.

and the rotational order parameter P1:


1 N in
P1 s Ý eˆi P eˆi Ž7.
N is1
in order to obtain additional information on the development of the system. In Eqs. Ž 4. – Ž 7. , t is the
time, r i is the center-of-mass position of molecule i, k is the reciprocal lattice vector of the initial
lattice, and eˆiin and eˆi are the unit vectors along molecular axis of molecule i in the initial and
running configurations, respectively. If the limit Ž Eq. Ž 4.. converges rapidly to a non-zero value, the
system contains a fluid phase. The mobility of molecules in glassy or solid states is much lower and
in such cases Eq. Ž4. converges slowly or not at all, while Eq. Ž5. fluctuates around a constant value
describing a molecule vibrating in a cage w21x. The translation order parameter is of the order of unity
for a translation order structure, and positive and of the order of 1r6N for a translation disorder
structure. Similarly, the rotational order parameter is of the order of unity for a rotational order
´ K. Aim r Fluid Phase Equilibria 161 (1999) 241–256
M. Lısal, 245

structure and fluctuates around zero with amplitude O Ž1r6N . when the structure is rotationally
disordered w8x.

4. VLE
4.1. Performance of potential model for Õapor and liquid phases
Before computing VLE we judged the ability of the RST potential to represent thermodynamic
properties of single vapor and liquid phases. For this purpose, we computed the second virial
coefficient Žas a convenient property characterizing low-density region. and performed NPT MD
simulations at several state points in the liquid phase. The computed properties were compared with
experimental data w11,22x as well as with simulated data obtained by using the WP potential w6x.
The second virial coefficient, B, appears in the simple truncated virial equation of state Ž EOS. :
pVm B
s1q Ž8.
RT Vm
where p is the pressure, Vm is the molar volume, T is the temperature and R is the universal gas
constant. The simple truncated virial EOS is typically accurate for densities up to about 50% of the
critical density w23x.
Values of the second virial coefficient were computed by using a non-product algorithm w24x.
Computed and experimental values of the second virial coefficient are listed in Table 2 and they are
also shown in Fig. 1. The RST potential is an effective potential and simulates many-body

Table 2
The second virial coefficient
T ŽK. B =10 6 Žm3 moly1 .
Experimentala Calculatedb
400 y166.0 y145.6
380 y185.9 y161.3
360 y208.8 y179.3
340 y234.9 y200.2
320 y265.0 y224.9
300 y299.4 y254.2
280 y338.7 y289.8
260 y383.0 y333.7
240 y431.9 y389.1
220 y483.1 y461.3
200 y530.3 y558.7
180 – y696.6
dev. 12%
d.max. 15%
a
Experimental data were taken from Ref. w11x.
b
This work; calculation with the RST potential using a non-product algorithm; calculated values being accurate to within
; 2%.
T is the temperature and B is the second virial coefficient. Absolute mean deviation Ždev.. and maximum deviation Žd.max..
are given in the last rows.
246 ´ K. Aim r Fluid Phase Equilibria 161 (1999) 241–256
M. Lısal,

Fig. 1. Second virial coefficient B of chlorine as a function of temperature T Ž Ø experimental data w11x; ( calculation with
the RST potential; the fit by Eq. Ž15...

interactions. Hence, the second virial coefficient and other low-density properties that depend solely
on two-body interactions cannot in principle be reproduced accurately. Comparison between com-
puted and experimental second virial coefficient shows that computed values overestimate experimen-
tal values typically by about 12%. This is quite reasonable agreement that ensures good description of
the vapor phase.
The NPT MD simulations by Andersen w8x were performed on liquid chlorine at three state points
previously simulated with the WP potential: 200 K and 0 MPa, 290 K and 0.6 MPa, and 400 K and
6.2 MPa. We used 256 molecules and an fcc starting configuration in a cubic box corresponded to
experimental molar volumes. The fcc configurations were melted 50 ps by the constant volume–con-
stant temperature MD simulations. It was followed by 250 ps intermediate NPT MD simulations.
Subsequent NPT production runs took 500 ps. Results of the NPT MD simulations together with
experimental and simulated data using the WP potential are given in Table 3. One can see that both
potentials reproduce experimental molar volumes, Vm , and potential energies, U, with the same

Table 3
Simulation results for the liquid
200 K, 0 MPa 290 K, 0.6 MPa 400 K, 6.2 MPa
b c b c
Experi- WP RST Experi- WP RST Experi- WP b RST c
mentala mentala mentala
U y20.18 y21.23Ž10. y20.77Ž9. y16.25 y16.12Ž20. y16.70Ž13. y9.73 y9.69Ž30. y11.00Ž45.
ŽkJ moly1 .
Vm =10 6 42.60 42.40Ž1. 43.90Ž14. 49.98 50.55Ž3. 51.56Ž35. 74.8 77.1Ž1. 74.12Ž35.
Žm3 moly1 .
a
Experimental data were taken from Ref. w22x.
b
NPT MC simulations with the WP potential w6x.
c
This work; NPT MD simulations with the RST potential.
U is the potential energy and Vm is the molar volume. Numbers in parentheses indicate uncertainties in the last digits.
´ K. Aim r Fluid Phase Equilibria 161 (1999) 241–256
M. Lısal, 247

accuracy. Differences between simulated and experimental data are about 3% and 1% in molar
volumes and potential energies, respectively.

4.2. Simulated VLE

Galassi and Tildesley w9x predicted VLE of chlorine modeled with the RST potential by Gibbs
ensemble Monte Carlo Ž GEMC. simulations w25x at high temperatures Ž between the critical tempera-
ture and temperature of 300 K.. The GEMC did not work properly for lower temperatures due to a
very low probability of successful particle transfers. On the other hand, the GDI that does not rely on
the particle transfer, can rather safely be used to predict VLE of chlorine at moderate and low
temperatures. The coexistence point at 300 K from the GEMC can be used for initiation of the GDI.
The GDI method is based on numerical solution of the Clapeyron equation which can be written
for VLE w13–15x as:
d ln p T Hv y H l
sy s f Ž 1rT , p . Ž9.
d Ž 1rT . s p Vmv y Vml
In Eq. Ž9., Hv and H l are the vapor and liquid molar enthalpies, and Vmv and Vml are the molar
volumes of vapor and liquid phases. Subscript s indicates that the derivative is taken along the
saturation line. Eq. Ž9. is a first-order non-linear differential equation of the type yX s f Ž x, y . that
prescribes how the pressure must change with the temperature for vapor and liquid phases to remain
in coexistence.
Given an initial condition, i.e., the pressure, temperature, f Ž1rT, p . , and equilibrium vapor and
liquid configuration at one coexistence point, Eq. Ž 9. can be solved numerically by a predictor–cor-
rector method w26x. The trapezoid predictor:
y kq1 s y k q h k f k Ž 10.
was used at all simulation points. In the first simulation point, the Euler corrector:
h0
y1 s y 0 q Ž f 0 q f 1 . Ž 11.
2
and in the following simulation points, the second-order corrector:
y kq1 s Ay ky1 q Cy k q E Ž D 0 f k q D 1 f kq1 . Ž 12.
were utilized. In Eqs. Ž10. – Ž12.., y stands for ln p, f k for f Ž1rTk , p k .; h k s x kq1 y x k and x s 1rT.
Coefficients of the second-order corrector w26x are A s 1rw r 2 Ž 3 q 2 r .x, C s 1 y A, D 0 s ry1 q 2 q r,
D 1 s 1 q r, and E s h krŽ3 q 2 r .; r s h ky1rh k . The quantities needed to evaluate the right-hand
side of the Clapeyron equation were obtained from simultaneous Žbut independent. NPT MD
simulations of the vapor and liquid phases.
At low temperatures, where the simple truncated virial EOS describes accurately enough the state
behavior of the vapor phase it is possible to skip simulations of the vapor phase. The molar volume of
the vapor phase, Vmv, and the heat of vaporization, D H s Hv y H l , can then be calculated by using the
simple truncated virial EOS w27x from expressions:

Vmv s
(
1 q 1 q 4 ps Br Ž RT .
Ž 13.
2 psr Ž RT .
248 ´ K. Aim r Fluid Phase Equilibria 161 (1999) 241–256
M. Lısal,

and

RT dB
D H s Hv y H l s yUl q ps Ž Vmv y Vml . q ž T yB / Ž 14.
Vmv dT

respectively. For this purpose, we fitted the calculated second virial coefficients to the equation of the
form used in Ref. w28x:

a3
½
B s a1 1 y a 2 exp ž T / y 1 5 Žm mol 3 y1
, K. Ž 15.

with parameters a1 s 60.3005 = 10y6 , a 2 s 3.4298, and a 3 s 277.0560. Deviations in the fit of Eq.
Ž15. were less than 0.5%.
The GDI was performed for chlorine between temperature of 300 K and experimental triple
temperature. We proceeded with the GDI in the following manner. Starting from the vapor–liquid
coexistence point determined by the GEMC at 300 K, the temperature was decreased and the
predictor pressure was calculated according to Eq. Ž 10.. Afterwards, the vapor and liquid configura-
tions were allowed to relax 250 ps. At the end of the relaxation period, all accumulators were set to
zero. The subsequent production run of the vapor and liquid phases took 500 ps. Ensemble averages
of vapor and liquid molar volumes, and vapor and liquid enthalpies from the production run were
used to evaluate the right-hand side of the Clapeyron equation, and thus, to calculate the corrector
pressure according to Eq. Ž11.. Then, the temperature was again decreased and the process was
repeated. The corrector pressures were evaluated according to Eq. Ž 12. . At each temperature, the NPT
MD simulations of the vapor and liquid phases were carried out simultaneously but independently.
From temperature of 225 K downwards we skipped simulations of the vapor phase and used the
simple truncated virial EOS. We decreased temperature in steps of 12.5 K for temperatures 300–200
K and in steps of 10 K for temperatures 200–180 K. The last two GDI were performed at 175 and
172 K.
Results of VLE for chlorine obtained by the GDI together with the GEMC data w9x are listed in
Table 4 for selected temperatures. One can see from Table 4 that the agreement between simulations
and experiments is excellent for vapor pressure and heat of vaporization and that is very good for
saturated densities. Simulated vapor pressures agree with experimental data within their statistical
uncertainties. Differences between experimental and simulated saturated densities and heats of
vaporization are about 3% and less than 2% Žexcept for the highest simulated temperature. ,
respectively.
We estimated the critical temperature Tc and density rc from the ordinary least-squares fit of the
law of rectilinear diameter w29x:
rv q r l
s rc q C l Ž T y Tc . Ž 16.
2

and the critical scaling relation w29x:


0.32
r l y r v s C2 Ž Tc y T . Ž 17.
´ K. Aim r Fluid Phase Equilibria 161 (1999) 241–256
M. Lısal, 249

Table 4
Simulation results for the vapor–liquid equilibria
T ŽK. Vmv =10 4 Žm3 moly1 . Vml =10 6 Žm3 moly1 . ps =10y6 ŽPa. D H ŽkJ moly1 .
Experimentala Simulatedb Experimentala Simulatedb Experimentala Simulatedb Experimentala Simulatedb
400 c 2.85 2.54Ž18. 75.19 72.99Ž213. 6.15 6.50Ž56. 8.109 6.210Ž973.
375c 5.08 5.32Ž90. 64.52 66.67Ž312. 4.07 4.10Ž45. 11.722 11.816Ž864.
350 c 8.62 7.87Ž56. 58.48 57.80Ž100. 2.57 2.89Ž24. 14.146 12.998Ž732.
325c 14.93 16.67Ž112. 54.35 54.94Ž91. 1.51 1.40Ž11. 16.032 15.868Ž568.
300 c 27.78 25.00Ž253. 51.28 51.71Ž52. 0.81 0.90Ž6. 17.584 17.952Ž112.
275d 54.68 51.73Ž942. 48.45 49.99Ž28. 0.39 0.43Ž8. 18.888 19.262Ž104.
250 d 126.1 120.9Ž381. 46.24 47.72Ž22. 0.16 0.19Ž6. 19.996 20.383Ž91.
225d 355.6 313.7 e 44.33 45.74Ž16. 0.052 0.059Ž9. 20.957 21.530Ž94.
200 d 1351.3 1219.5e 42.60 43.88Ž14. 0.012 0.013Ž4. 21.843 22.443Ž96.
190 d 2588.1 2405.1e 41.92 43.20Ž13. 0.0061 0.0065Ž9. 22.201 22.838Ž95.
180 d 5401.4 5138.5e 41.24 42.54Ž12. 0.0028 0.0029Ž4. 22.578 23.221Ž89.
175d 8102.5 7833.8 e 40.89 42.24Ž12. 0.0018 0.0018Ž2. 22.779 23.418Ž93.
172 d 10475.2 10200.9 e 40.67 42.01Ž11. 0.0014 0.0014Ž1. 22.898 23.552Ž89.
dev. 7.3% 2.7% 7.6% 4.3%
d.max. 12% 3.3% 19% 23%
a
Experimental data were taken from w11x.
b
Simulations with the RST potential.
c
GEMC simulations w9x.
d
This work; GDI using NPT MD simulations.
e
Evaluated from the simple truncated virial equation of state.
T is the temperature, Vmv and Vml are molar volumes of vapor and liquid phases, respectively, ps is the vapor pressure, and
D H is the heat of vaporization. Numbers in parentheses indicate uncertainties in the last digits. Absolute mean deviations
Ždev.. and maximum deviations Žd.max.. are given in the last rows.

to the simulated data. The fit of Eq. Ž 16. was performed over the whole temperature range and that of
Eq. Ž17. for T G 300 K. Typical deviations in the fit of Eqs. Ž 16. and Ž 17. were 1% and 5%,

Fig. 2. Coexistence envelope r – T of chlorine Ž experimental data w11x; ( simulation with the RST potential.. The
experimental critical point w11x is indicated by I and the critical point estimated from simulations is denoted by e.
250 ´ K. Aim r Fluid Phase Equilibria 161 (1999) 241–256
M. Lısal,

Fig. 3. Vapor pressure ps of chlorine as a function of reciprocal temperature 1r T Ž experimental data w11x; (
simulation with the RST potential.. The experimental critical point w11x is indicated by I and the critical point estimated
from simulations is denoted by e.

respectively. Estimated critical temperature, Tc , and density, rc , are 413.38 K and 589.15 kg my3,
respectively. The estimated Tc and rc are in very good agreement with experimental Tc and rc which
are equal to 416.956 K and 576.800 kg my3, respectively. To estimate the critical pressure, pc , we
have fitted the calculated vapor pressures to a Wagner-type equation w23x:
ln Ž psrpc . s Ž a1t q a 2t 1.5 q a 3t 3 q a 4t 5 . rTr Ž 18.
where Tr s TrTc and t s 1 y Tr . In the parameter estimation procedure, the vapor pressure data points
were properly weighted by employing the statistical uncertainties as given in Table 4 and pc was

Fig. 4. Heat of vaporization D H of chlorine as a function of temperature T Ž experimental data w11x; ( simulation
with the RST potential..
´ K. Aim r Fluid Phase Equilibria 161 (1999) 241–256
M. Lısal, 251

treated as an adjustable parameter. With constants a1 s y6.5811, a 2 s 1.2308, a 3 s 0.94637,


a 4 s y7.3486, and pc s 7.96 MPa, Eq. Ž 18. represents the simulated vapor pressures well within the
estimated uncertainties. The critical pressure value obtained by this extrapolation is only 0.4% lower
than the experimental value of 7.9914 MPa, which means certainly excellent agreement.
Finally, in Figs. 2–4 we display comparison of experimental and simulated orthobaric densities,
vapor pressures, and heats of vaporization of chlorine.

5. Zero-pressure solids

The solid was simulated at four temperature values Ž 22, 55, 100 and 160 K. and zero pressure by
using the Parrinello–Rahman method w16,17x. Experimental data at these conditions w12x and Monte
Carlo ŽMC. simulations with the WP potential at 55, 100 and 160 K and zero pressure w6x are

Table 5
Simulation results for the solid at zero pressure
Experimentala WP b RST c Experimentala WP b RST c
22 K 55 K
a=10 10 Žm. 6.1453Ž2. – 6.108Ž5. 6.1804Ž2. 6.187Ž1. 6.153Ž3.
b=10 10 Žm. 4.3954Ž1. – 4.494Ž2. 4.4174Ž1. 4.663Ž1. 4.542Ž4.
c=10 10 Žm. 8.1537Ž2. – 8.330Ž3. 8.1711Ž2. 7.849Ž1. 8.303Ž8.
Ž a,b . Ž8. 90 – 90.00Ž3. 90 89.9Ž1. 89.99Ž5.
Ž a,c . Ž8. 90 – 90.00Ž1. 90 90.01Ž1. 90.00Ž2.
Ž b,c . Ž8. 90 – 90.00Ž1. 90 88.99Ž1. 90.00Ž2.
rŽk. 1.00 – 0.99 1.00 – 0.99
P1 1.00 – 0.97Ž3. 1.00 1.00 0.89Ž7.
U ŽkJ moly1 . – – y31.37Ž1. – – y30.66Ž1.
Vm =10 6 Žm3 moly1 . 33.16 – 34.42Ž1. 33.59 34.09Ž1. 34.94Ž1.

100 K 160 K
a=10 10 Žm. 6.2235Ž2. 6.358Ž1. 6.228Ž7. 6.2929Ž3. 6.513Ž1. 6.670Ž25.
b=10 10 Žm. 4.4561Ž1. 4.847Ž1. 4.626Ž10. 4.5361Ž2. 4.988Ž1. 5.078Ž25.
c=10 10 Žm. 8.1785Ž2. 7.521Ž1. 8.247Ž21. 8.1617Ž3. 7.433Ž2. 7.380Ž66.
Ž a,b . Ž8. 90 89.9Ž1. 89.99Ž9. 90 90.1Ž1. 90.01Ž9.
Ž a,c . Ž8. 90 90.07Ž2. 90.00Ž4. 90 89.98Ž3. 90.00Ž11.
Ž b,c . Ž8. 90 90.01Ž1. 90.00Ž3. 90 89.96Ž1. 89.96Ž99.
rŽk. 1.00 – 0.97 1.00 – 0.93Ž1.
P1 1.00 0.99 0.96Ž3. 1.00 0.99 0.95Ž2.
U ŽkJ moly1 . – – y29.55Ž2. – – y27.05Ž1.
Vm =10 6 Žm3 moly1 . 34.15 34.89Ž1. 35.77Ž2. 35.08 36.35Ž1. 37.61Ž9.
a
Experimental data were taken from Ref. w12x.
b
NPT MC simulations with the WP potential w6x.
c
This work; NPT MD simulations with the RST potential.
a, b and c are the lattice parameters, Ž a,b ., Ž a,c . and Ž b,c . are the angles between appropriate lattice parameters, r Ž k . is
the translation order parameter, P1 is the rotational order parameter, U is the potential energy, and Vm is the molar volume.
Numbers in parentheses indicate uncertainties in the last digits.
252 ´ K. Aim r Fluid Phase Equilibria 161 (1999) 241–256
M. Lısal,

Fig. 5. Lattice parameters a, b and c of the zero-pressure crystal structures of chlorine as a function of temperature T. The
lines are drawn through experimental data of Powell et al. w12x Ž experimental values of a; - - - experimental values
of b; P P P experimental values of c; ( this work, calculation with the RST potential; e calculation with the WP potential
w6x..

available. The simulation box contained 240 molecules. Starting configurations were experimentally
observed structures. The experimentally observed structures are isostructural, C-centered orthorhom-
bic Žspace group Cmca., with two chlorine molecules in the primitive unit cell. The molecules are
stacked in a layer structure with the intramolecular bonds lying in planes parallel to the bc plane w12x.
The starting configurations were equilibrated 25 ps and following production runs took 250 ps.

Fig. 6. Molar volume Vm and potential energy U of the zero-pressure crystal structures of chlorine as a function of
temperature T. The full line is drawn through experimental data of Powell et al. w12x Ž experimental values of Vm ;
( this work, calculation of Vm with the RST potential; e calculation of Vm with the WP potential w6x; Ø , - - -, this work,
calculation of U with the RST potential..
´ K. Aim r Fluid Phase Equilibria 161 (1999) 241–256
M. Lısal, 253

During the whole production runs, the structures remained stable and the predicted space group was
still Cmca. Results of our solid simulations together with experimental and MC simulated data using
the WP potential are summarized in Table 5. Fig. 5 shows comparison of experimental and simulated
values of lattice parameters a, b and c. One can see from Fig. 5 that the RST potential reproduces
better the experimental values than the WP potential. However, the lattice parameters remain within
10% of the experimental values for the both potentials. Differences between the experimental and
simulated values are larger at the highest investigated temperature of 160 K which is close to the
experimental triple temperature. Fig. 6 displays comparison between experimental and simulated
values of potential energies and molar volumes. The simulated molar volumes are systematically by
about 5% larger than experimental values.

6. Conclusions

The main observations based on our calculation results for the RST potential may be summarized
as follows: Ži. in view of the fact that the RST potential is an effective multibody potential Ž i.e., not
primarily designed to reproduce properties resulting from two-body interactions. , good agreement
with experimental second virial coefficients is obtained; Ž ii. in the temperature and pressure ranges of
200 to 400 K and 0 to 6.2 MPa, respectively, the RST potential reproduces the liquid molar volume
and internal energy data within 3% and 1%, respectively, which is the same accuracy as obtained by a
more complicated Wheatley and Price potential; Žiii. the prediction of orthobaric Ž saturated liquid and
vapor. densities is very good Žwithin 3% of their experimental values.; Ž iv. the prediction of saturated
vapor pressures is excellent, practically within experimental uncertainties; Ž v. the critical temperature
and density are estimated from the simulated VLE data Žby using the rectilinear diameter law and
critical scaling relation. within 1% and 2% of their experimental values, respectively; Ž vi. the critical
pressure is extrapolated from the predicted vapor pressure data by Wagner-type equation within 0.4%
of its experimental value; and Ž vii. regarding the zero-pressure properties of the crystal structure of
solid chlorine, the RST potential predicts the lattice parameters more accurately than the WP
potential; the simulated solid-phase molar volumes are about 5% larger than the corresponding
experimental values.

Acknowledgements

The authors acknowledge partial support of the Grant Agency of the Academy of Sciences of the
Czech Republic ŽGrant No. A-4072712..

Appendix A. Spherical harmonic functions

The S functions needed to evaluate the potential energy between atoms a and g on two chlorine
molecules can be expressed in terms of four dot products containing vectors rag , ea and eg
254 ´ K. Aim r Fluid Phase Equilibria 161 (1999) 241–256
M. Lısal,

Fig. 7. The geometry of a pair of chlorine molecules; i and j are the centers of mass of the molecules and a and g are the
positions of anisotropic sites in each molecule Ž r i j s r i y r j , rag s ra y rg s r i j q ea y eg ..

Žsee Fig. 7.: D s rag P rag , Da s ea P rag , Dg s yeg P rag , and Dag s ea P eg w20x. The relevant
expressions for the S functions are:
S110 s ly2 Dag
1
S202 s
2
Ž 3ly2 Dy1 Da2 y 1 .
1
S022 s ž 3l y2
Dy1 Dg2 y 1 /
2
1
S303 s
2
Ž 5ly3 Dy3r2 Da3 y 3ly1 Dy1r2 Da .
1
S033 s ž 5l y3
Dy3r2 Dg3 y 3ly1 Dy1r2 Dg /
2
1
S112 s ly2 Dag q 3 Dy1 Da Dg
ž /
2
1
S123 s y3
ž 5l Dy3r2 Da Dg2 q 2 ly3 Dy1r2 Dg Dag y ly1 Dy1r2 Da /
2
1
S213 s ž 5l y3
Dy3r2 Da2 Dg q 2 ly3 Dy1r2 Da Dag y ly1 Dy1r2 Dg /
2
1
S224 s 35ly4 Dy2 Da2 Dg2 q 20 ly4 Dy1 Da Dg Dag y 5ly2 Dy1 Da2 q Dg2 q 2 ly4 Dag
ž 2
q1 /
8
´ K. Aim r Fluid Phase Equilibria 161 (1999) 241–256
M. Lısal, 255

Appendix B. Forces and torques

The force acting on molecule i from the site–site potential Vi j can be written as a sum over three
dot products Ždefined in Appendix A.:
E Vi j EVi j EVi j
fi s Fi  D 4 q Fi  Da 4 q Fi  Dg 4
ED EDa EDg

The force operator Fi can be expressed using vectors rag , ea and eg Žsee Fig. 7. as:
Fi  D 4 s y2 rag

Fi  Da 4 s yea
Fi  Dg 4 s eg

The force on molecule j is f j s yf i .


The torque on molecule i can be written as a sum over four dot products:
EVi j EVi j EVi j EVi j
ti s Ti  D 4 q Ti  Da 4 q Ti  Dg 4 q Ti  Dag 4
ED EDa EDg EDag

The torque operator Ti can be expressed using vectors rag , ea and eg as:
Ti  D 4 s 2 Ž rag = ea .

Ti  Da 4 s rag = ea

Ti  Dg 4 s ea = eg

Ti  Dag 4 s y Ž ea = eg .

For a pair of molecules i and j the torques t i and t j are not equal and opposite Ž unlike the forces
f i and f j . but it is valid that t i q t j q r i = f i q r j = f j s 0. Hence, the torque on the molecule j must
be expressed, like t i , as:
EVi j EVi j EVi j EVi j
tjs Tj  D 4 q Tj  Da 4 q Tj  Dg 4 q Tj  Dag 4
ED EDa EDg EDag

The torque operator Tj , like Ti , is expressed by using vectors rag , ea and eg as:
Tj  D 4 s y2 Ž rag = eg .

Tj  Da 4 s y Ž ea = eg .

Tj  Dg 4 s y Ž rag = eg .

Tj  Dag 4 s ea = eg
256 ´ K. Aim r Fluid Phase Equilibria 161 (1999) 241–256
M. Lısal,

References

w1x K.E. Gubbins, Fluid Phase Equilibria 83 Ž1993. 1–14.


w2x K. Singer, A. Taylor, J.V.L. Singer, Mol. Phys. 33 Ž1977. 1765–1795.
w3x S. Romano, K. Singer, Mol. Phys. 37 Ž1979. 1765–1772.
w4x P.M. Rodger, A.J. Stone, D.J. Tildesley, J. Chem. Soc., Faraday Trans. 2 83 Ž1987. 1689–1702.
w5x P.M. Rodger, A.J. Stone, D.J. Tildesley, Mol. Phys. 63 Ž1988. 173–188.
w6x R.J. Wheatley, S.L. Price, Mol. Phys. 71 Ž1990. 1381–1404.
w7x C.G. Gray, K.E. Gubbins, Theory of Molecular Fluids: Vol. 1. Fundamentals, Clarendon Press, Oxford, 1984.
w8x M.P. Allen, D.J. Tildesley, Computer Simulation of Liquids, Clarendon Press, Oxford, 1987.
w9x G. Galassi, D.J. Tildesley, Mol. Simul. 13 Ž1994. 11–24.
w10x S.L. Price, Mol. Phys. 62 Ž1987. 45–63.
w11x S. Angus, B. Armstrong, K.M. De Reuck ŽEds.., Chlorine, International Thermodynamic Tables of the Fluid State, Vol.
8, Chemical Data Series, No. 31, Blackwell, Oxford, 1984.
w12x B.M. Powell, K.M. Heal, B.H. Torrie, Mol. Phys. 53 Ž1984. 929–939.
w13x D.A. Kofke, J. Chem. Phys. 98 Ž1993. 4149–4162.
w14x M. Kofke, Mol. Phys. 78 Ž1993. 1331–1336.
w15x ´ V. Vacek, Mol. Simul. 17 Ž1996. 27–39.
M. Lısal,
w16x M. Parrinello, A. Rahman, Phys. Rev. Lett. 45 Ž1980. 1196–1199.
w17x ´ M.L. Klein, J. Chem. Phys. 78 Ž1983. 6928–6939.
S. Nose,
w18x S.L. Price, A.J. Stone, M. Alderton, Mol. Phys. 52 Ž1984. 987–1001.
w19x A.J. Stone, Mol. Phys. 36 Ž1978. 241–256.
w20x P.M. Rodger, A.J. Stone, D.J. Tildesley, Mol. Simul. 8 Ž1992. 145–164.
w21x I. Nezbeda, J. Kolafa, Mol. Simul. 14 Ž1995. 153–161.
w22x J.D. Sullivan, P.A. Egelstaff, Chem. Phys. 82 Ž1983. 479–488.
w23x R.C. Reid, J.M. Prausnitz, B.E. Poling, The Properties of Gases and Liquids, McGraw-Hill, New York, 1987.
w24x S. Murad, Quantum Chemistry Program Exchange 12, No. 357, Indiana University, 1978.
w25x A.Z. Panagiotopoulos, Mol. Simul. 9 Ž1992. 1–23.
w26x F.A. Escobedo, J.J. de Pablo, J. Chem. Phys. 106 Ž1997. 2911–2923.
w27x S. Gupta, J. Yang, N.R. Kestner, J. Chem. Phys. 89 Ž1988. 3733–3741.
w28x A.R.H. Goodwin, M.R. Moldover, J. Chem. Phys. 93 Ž1990. 2741–2753.
w29x J.V. Sengers, J.M.H. Levelt Sengers, Progress in Liquid Physics, C.A. Croxton ŽEd.., Wiley, New York, 1978, pp.
103–174.

You might also like