J. Phys. Chem.

1996, 100, 16779-16781 16779

Determination of the Pressure-Viscosity Coefficient of Decane by Molecular Simulation

Christopher J. Mundy* and Michael L. Klein

Department of Chemistry, UniVersity of PennsylVania, Philadelphia, PennsylVania 19104-6323

J. Ilja Siepmann
Department of Chemistry, UniVersity of Minnesota, 207 Pleasant Street SE,
Minneapolis, Minnesota 55455-0431
ReceiVed: June 27, 1996X

The shear viscosities of n-decane and 4-propylheptane have been calculated over a range of densities.
Equilibrium molecular dynamics simulations were carried out in the canonical ensemble for an empirical
force field using the united-atom representation. Quantitative agreement is obtained with the experimental
shear viscosities and the pressure-viscosity coefficient.

Introduction agreement with experiment (deviation of less than 10%) has

been found using an empirical force field which was previously
Knowledge of how the viscosity of a given molecular fluid
derived by fitting to thermodynamic properties. Here, we extend
behaves as a function of pressure and temperature is an
these calculations to different state points with the aim of testing
important parameter in the field of rheology and in the design
whether or not we are able to obtain a reliable estimate of the
of lubricants. However, despite the important technological
R coefficient for this prototype of a hydrocarbon lubricant. In
issues involved, a miscroscopic level understanding on the
the next section the force field and the simulation methodology
influence of molecular structure on transport processes is
are outlined. This is followed by a discussion of the simulation
presently not available. Recently, we have shown that pertaining
results. Particular attention is given to the relationship between
to the estimation of critical properties of complex fluids, the
the calculated viscosity and the longest relaxation time in the
field of molecular simulations has reached a state of maturity
system. We conclude that current molecular simulation meth-
that allows the successful prediction of valuable engineering
odology is able to yield qualitative resuts for properties of “real”
parameters.1,2 In this letter, we set out to test the accuracies of
fluids.
empirical force fields and molecular simulations for the predic-
tion of transport phenomena. In particular, we have utilized Simulation Details
the molecular dynamics simulation technique to estimate the The force field for the alkanes has been obtained from
pressure dependence of the shear viscosity for n-decane. This calculations of the coexistence curves and critical points of the
quantity is commonly known as the R coefficient and is defined homologous series of linear alkanes1,2 and of three heptane
by the Barus relation3 isomers11 using configurational-bias Monte Carlo in the Gibbs
ensemble. The united-atom approach, in which methylene,
η(P2) ) η(P1) exp[R(P2 - P1)] (1) methyl, and ternary CH groups are treated as one pseudoatom,
was found to yield a very satisfactory description of the fluid
where Pi is the pressure at state point i at constant temperature phase diagrams. Details of the models can be found in refs 1,
T, and η is the shear viscosity, being a function of only the 10, and 11. The simulations for the shear viscosity were
pressure. We have also calculated the shear viscosity of a performed for 64 alkane molecules in the canonical ensemble
decane isomer, 4-propylheptane, to test our approach for at T ) 480 K. Four simulations at different densities (F )
branched molecules. 0.5937, 0.6136, 0.6251, and 0.6382 g/cm3) were carried out for
To our knowledge, thus far, only one attempt to calculate n-decane, whereas 4-propylheptane was only studied at the
the R coefficient of a molecular fluid by simulation has been intermediate density of 0.6136 g/cm3. The corresponding
reported. Cummings and Varner5 carried out an extensive pressures ranged from 1 to 30 MPa (see Table 1). At each state
nonequilibrium molecular dynamics study for liquid water and point a total of three independent simulations, each of lengths
calculated the shear viscosity, as a function of shear rate (γ), 900 ps, were carried out (i.e. starting from different well-
over a range of pressures.4 In agreement with experiment, their equilibrated initial configurations). For technical details of the
simulations yield an increase in the viscosity with increasing simulations, the reader should consult ref 7. The viscosity was
pressure, but the R coefficient obtained from the simulations computed using the standard Green-Kubo (GK) formula12 that
was larger than that measured. At all densities considered, the relates the fluctuations in the stress tensor to the shear viscosity
extrapolated zero-strain viscosities deviated from experimental by

η(t) ) (V/kBT)∫0 〈Pxy(0) Pxy(t*)〉 dt*

values by a factor of approximately 2. The main part of the t
disagreement between experiments and simulations is likely
(2)
caused by inadequacies in the force field used to described the where Pxy is the xy component of the pressure tensor, V the
water molecules.8,9 volume, kB Boltzmann’s constant, and T the temperature.
In a previous paper we reported on the calculation of the shear
viscosity of n-decane at one state point.10 Rather good Results and Discussion
A plot of the viscosity of n-decane as a function of pressure
X Abstract published in AdVance ACS Abstracts, October 1, 1996. is given in Figure 1 and the numerical data can be found in
S0022-3654(96)01919-3 CCC: $12.00 © 1996 American Chemical Society
16780 J. Phys. Chem., Vol. 100, No. 42, 1996 Letters

TABLE 1: Summary of the Simulation Resultsa

τr τr
F 〈P〉 〈η〉 ηexp (eq 3) (eq 4)
molecules [g/cm3] [MPa] [10-4 -4
Pa s] [10 Pa s] [ps] [ps]
n-decane 0.5937 1.1 ( 1.0 1.41 ( 0.12 1.60 9.2 6.5
n-decane 0.6136 10.6 ( 1.0 1.90 ( 0.15 1.92 10.2 7.3
n-decane 0.6251 17.2 ( 1.0 2.31 ( 0.23 2.12 10.7 8.0
n-decane 0.6382 28.1 ( 1.2 2.30 ( 0.40 2.43 11.6 8.9
4-propyl- 0.6136 1.7 ( 0.8 1.60 ( 0.15 1.40 4.5 7.3
heptane
a
F denotes the density, 〈P〉 and 〈η〉 are the canonical ensemble
averages of pressure (including tail corrections) and viscosity, ηexp stands
for the experimental viscosity taken from refs 10 and 12, and the two
relaxation times τr are calculated from the orientational autocorrelation
function (eq 3) and the Rouse model (eq 4).

Figure 2. Orientational autocorrelation function, c(t), for liquid

n-decane (DEC) and 4-propylheptane (4PH) at F ) 0.6136 g/cm-3.

where M is the molecular mass, F the density, and η the bulk

viscosity. Values of τr calculated from eqs 3 and 4 are compared
in Table 1. At all four densities the Rouse model predicts
relaxation times approximately 30% smaller than those calcu-
lated directly from the OACFs, but this can be considered as
reasonable agreement since the validity of the Rouse model
assumes no excluded volume. The effects of entanglement
clearly can be neglected for such a short molecule.
The shear viscosity of 4-propylheptane is 1.6 × 10-4 Pa s at
the state point under investigation (see Table 1). For this
condition, we were not able to find an experimental result. For
Figure 1. Shear viscosity, η, for liquid n-decane as a function of comparison, we have extrapolated experimental results at lower
pressure at T ) 480 K. temperatures15 to the state point of interest using the Arrhenius
relation. The so-estimated experimental shear viscosity is 1.4
Table 1. Over the entire pressure (density) range, agreement × 10-4 Pa s and thus just outside of the certainty range of the
with experiment13 is quantitative (deviations of approximately simulation result. The OACF of 4-propylheptane is also shown
10%). With one exception (F ) 0.5937 g/cm3) the experimental in Figure 2. The deviation from the simulated and the calculated
results fall within the statistical accuracy of the simulations. relaxation time estimated from eq 4 is most likely attributed to
From weighted linear fits using eq 1 the R coefficient for our the fact that the side branches of 4-propylheptane are too short
model n-decane was determined as 21 ( 5 GPa-1 compared to to behave like a polymer.
the experimental value of 14 GPa-1. Most of the disagreement In summary, equilibrium molecular dynamics calculations
stems from the simulation at the lowest density which is very have been employed to determine the shear viscosities of
close to the coexistence density of our model n-decane at T ) n-decane and 4-propylheptane and the pressure-viscosity
480 K.2 coefficient of the former. We have demonstrated that an alkane
As is evident from the error bars in Figure 1, for a given model fitted to the critical parameters can be used to obtain
temperature and the same MD simulation length, the precision satisfactory values for the shear viscosity and the R-coefficient.
associated with the calculated viscosity worsens as the pressure However, the prediction of transport properties still lacks some
increases. This is to be expected, since at higher pressures the of the accuracy that can be presently achieved for experimental
longest molecular relaxation time will, in general, increase. From thermodynamic and structural properties. Nevertheless, the
calculations of various relaxation modes, we found that for agreement in the macroscopically measurable transport quantities
n-decane, at the state points studied here, the slowest relaxation is likely to be sufficient to allow insight into microscopic-level
is associated with the orientational relaxation of a decane processes occurring in lubricating fluids. The methodology
molecule, i.e., the time for a molecule to complete an end-over- employed herein can readily be extended to more complex
end rotation. This time can be monitored via the orientational systems and hence should be a useful aid in the design of
autocorrelation function (OACF) of the normalized molecular lubricants.
end-to-end vector, denoted by u, e.g., c(t) ) 〈u(0)‚u(t)〉. The
OACFs are shown in Figure 2, and the relaxation times, τr, can Acknowledgment. The work described herein was supported
be extracted by fitting the decay to a simple exponential: by the National Science Foundation and Albemarle Corp. J.I.S.
acknowledges additional support by the Camille and Henry
c(t) ) exp(-t/τr) (3) Dreyfus New Faculty Awards Program. We thank Dixie Goins,
Bill Moehle, Burney Lee, and Paul Kung for their encourage-
Knowledge of both the viscosity and the longest relaxation
ment.
time allows for an interesting comparison with the modified
Rouse model, which predicts for the relation of above quanti- References and Notes
ties7,14
(1) Siepmann, J. I.; Karaborni, S.; Smit, B. Nature 1993, 365, 330.
(2) Smit, B.; Karaborni, S.; Siepmann, J. I. J. Chem. Phys. 1995, 102,
τr ) 6ηM/π FkT 2
(4) 2126.
Letters J. Phys. Chem., Vol. 100, No. 42, 1996 16781

(3) Barus, C. Am. J. Sci. 1893, 45, 87. to a mistake in the evaluation of the tail corrections, the correct pressures
(4) The viscosity at γ ) 0 has to be obtained from an extrapolation of are reported in this letter.
the values obtained at finite (high) γ. However, the scaling relation which (11) Siepmann, J. I.; Mundy, C. J.; Klein, M. L. Mol. Phys., submitted.
connects the finite-shear viscosity to the γ ) 0 value has not been (12) Evans, D. J.; Morriss, G. P. Statistical Mechanics of Nonequilibrium
determined unequivocally and is subject to considerable uncertainty.6,7 Liquids; Academic Press: New York, 1990.
Cummings and Varner5 used the relation η ∝ γ1/2 to fit their results.
(5) Cummings, P. T.; Varner, T. L. J. Chem. Phys. 1988, 89, 6391. (13) Stephan, K., and Lucas, K., Viscosity of Dense Fluids; Plenum
(6) Chynoweth, S.; Michopoulos, Y. Mol. Phys. 1994, 81, 133. Press: New York, 1979.
(7) Daivis, P. J.; Evans, D. J., J. Chem. Phys. 1994, 100, 541. (14) Doi, M.; Edwards, S. F. The Theory of Polymer Dynamics;
(8) Smith, P. E.; van Gunsteren, W. F. Chem. Phys. Lett. 1993, 215, Clarendon Press: Oxford, 1986.
315. (15) Properties of Hydrocarbons of High Molecular Weight; American
(9) Balasubramanian, S.; Mundy, C. J.; Klein, M. L. J. Chem. Phys., Petroleum Institute, Research Project 42, Washington, 1966.
submitted.
(10) Mundy, C. J.; Siepmann, J. I.; Klein, M. L. J. Chem. Phys. 1995,
102, 3376. The pressure of the simulation reported in ref 7 is wrong due JP9619191