Vapor-liquid coexistence of model systems

By Sabharish Murali

Outline
Introduction Models Used Methodologies Results Conclusion

Importance of VLE
VLE gives the nature of intermolecular interactions present in the liquid and vapor phases. Useful in developing equations of state and corresponding state theories Essential for efficient process design and optimization Knowledge of saturation properties is crucial in various chemical processes such as designing distillation, chemical manufacturing etc.

Need for molecular simulation

Reliable experimental data are only available for relatively low molecular weight hydrocarbons and only over a limited temperature range because of their thermal instability issues. Molecular simulation is an alternative approach to obtain the thermo-physical properties of molecules Accuracy of the prediction depends largely on the quality of the force-field used and how well it describes the system.

Models Used
n-Octane (TraPPE united-atom model) Cyclohexane (TraPPE united-atom model) Squalane (TraPPE united-atom model) Pyrene (TraPPE all-atom model (extension)) Propanol (TraPPE united-atom model) Water (rigid SPC/E, flexible SPC) C1mim-BF4 (Zhong et al. model)

TraPPE model
TraPPE model was developed to Reproduce thermo-physical properties over a wide range of physical conditions Keep the models as transferable as possible by minimizing the number of (Pseudo) atoms needed for any particular molecule and by using the same parameters for a given (Pseudo) atom in all types of molecules

Zhong potential
Amber style force field Consists of Lennard Jones, bond stretching, torsion, angle bending and improper torsion interactions

Force Field
Bond stretching potential taken from CHARMM forcefield = ( )2 2 Harmonic potential used to describe bond angle bending = ( )2 2 Torsion potentials
= 0 + 1 1 + + 2 1 cos(2) + 3 1 + cos(3)

Force Field
Non-bonded interactions described by Lennard-jones potential and columbic interactions
= 4
12

40

Growth expanded ensemble, a scaled potential is used to describe the LJ interactions

, = 4 1 (1 )2 +
6 2

(1

)2 +

For columbic interactions Ewald summation is used.

( 0.5) , = 0.5 4 0 > 0.5 < 0.5

Alternative Approaches
Gibbs ensemble:
Chemical potential, pressure and temperature are kept constant Phase equilibrium achieved by allowing boxes to exchange volume and particles Performs poorly at low temperatures and near critical points

Configurational Bias(Siepmann and et al)

Molecule is grown atom by atom with multiple configurations examined for each atom Maintaining a reservoir is not necessary

Grand Canonical Ensemble

, volume ,temperature are kept constant. The macro variable is the number of molecules. System is coupled with a heat bath and an infinite reservoir. Hence, Energy and number of molecules fluctuate Moves performed Particle displacement Particle addition/deletion Hybrid Monte Carlo Simulations are done at relatively high temperatures to avoid sampling difficulties

Reservoir Grand Canonical

Developed by Errington and Panagiotopoulos Acceptance of Insertion/deletion moves depends on molecular architecture and size Large molecules, it is difficult to find a hole in the system If molecular architecture is complex or intramolecular potential is stiff then generating new chain to sample internal degrees of freedom can be time consuming.
Errington and Panagiotopoulos, J. Chem. Phys. (1999)

Reservoir Grand Canonical

Reservoir containing isolated molecules is created with same T as simulation box Ideal chain reservoir contains molecules with only bonded interactions Ideal gas reservoir molecules with bonded interactions and intra-molecular nonbonded interactions between atoms separated by more than 3 bonds.

Reservoir Grand Canonical

Some of the moves used in reservoir are Hybrid Monte Carlo Pivot moves Crankshaft Torsion angle rotation Depends on the internal constraints and molecular architecture. It must sample adequately all configurational space for internal degrees of freedom of molecule

Growth Expanded Ensemble

Involves stage-wise insertion/deletion of the molecules and hence facilitates transfer of large molecules in and out of the system We transfer molecules to and from system via an M-stage process Intermolecular interactions between tagged molecule and the rest of the system are scaled with stage number m.

DETERMINATION OF PHASE COEXISTENCE

Chemical potential, volume, minimum and maximum particle number is required. Number probability distribution at (T,0) is obtained using a TMMC scheme. Histogram reweighting was used to modify probability distribution to get probability distribution at coexistence
ln ; = ln ; + ln

Isothermal Isobaric Ensemble

Number of particles, temperature and pressure are kept constant System connected to a heat and volume bath allowing energy and volume fluctuations.

NPT Ensemble Moves

Volume Scaling moves
A random change in volume V is selected and new volume is VV+ V Entire simulation box is scaled and particle positions are scaled along each axis

Particle Displacement moves

Tracing coexistence curves

At vapor-liquid coexistence the probability distribution is bimodal with equal areas under liquid and vapor regions. To sample both liquid and vapor regions the system needs to traverse through intermediate states with high free energy barrier. At low temperatures, large system sizes inhomogeneous structures form at intermediate densities resulting in sampling difficulties.

Grand canonical temperature expanded ensemble

Used to determine how grand potential of a phase varies with temperature and activity Macrostate variable in bin number corresponding to . Temperature divided into a number of bins. Bin separation is selected small enough to obtain sufficient sampling across temperature range

Moves performed
Hybrid Monte Carlo Molecular growth addition/deletion Temperature expanded ensemble

Grand canonical temperature expanded ensemble

coex at 2 high temp. obtained through direct GC simulations and used to determine linear relationship between ln and . A TEE simulation along this path is performed and grand potential at one of GC simulation is used to find grand potential at other T
, = 1 , 1 ln , 1 , 1

Histogram reweighting used to find grand potential at coexistence.

, = , ln
=

; ,

; ,

Grand Canonical Temperature expanded ensemble

Use the histogram reweighting to determine the activities which gives equal values of , and , The estimates obtained are used to perform simulations again to obtain more accurate values. Simulation is stopped when the ln values obtained from two subsequent iterations differ by less than 5x10-3.

Isothermal isobaric temperature expanded ensemble

To obtain the evolution of Gibbs free energy of a particular phase with temperature and pressure Microstate variable V. In vapor phase ln V is used

Isothermal isobaric temperature expanded ensemble

Pcoex at 2 high temp. determined through direct NPT simulations and used to determine linear relationship between lnP and . Gibbs potential at one NPT simulation is used to find gibbs potential at different T.
, = , ln , ,

Histogram reweighting is used to find Gibbs potential at coexistence

, = , ln

; , 0 = = ; ,

Advantages
Separate simulations for vapor and liquid phase System does not sample intermediate states which allow one to get vapor-liquid coexistence properties at low temperatures

Hybrid Monte Carlo

Developed by Duane et. Al For nonrigid molecules, displacement moves can become inefficient because they tend to distort bond lengths and angles, it becomes extremely difficult to find intelligent moves to propagate the system MD provides a deterministic pathway for the system to evolve HMC uses short MD trajectories to generate trial configurations Better sampling of intramolecular degrees of freedom
Duane et al., Phys. Lett. B (1987)

Hybrid Monte Carlo

Three types of hybrid Monte Carlo Moves performed HMC All
MD trajectories are given to all molecules in the system

HMC One
A molecule is selected and an MD trajectory is completed with that molecule

HMC No Torsion
A HMC one move performed but the torsion potentials are neglected in MD trajectories.

Computing Probability Distribution

Wang Landau approach
Quickly produces an approximate probability distribution Non equilibrium approach which builds probability distribution via regular updates to f is the modification factor reduced in stages Poor Convergence
Wang and Landau, Phys. Rev. Lett. (2001)

Computing Probability Distribution

Transition Matrix Monte Carlo
Obtains the probability of observing a system at states along a reversible path by utilizing information about attempted transitions.

() ( ) = () ( )

Provides an accurate probability distribution with good precision

Smith and Bruce, J. Phys. A (1995); Fitzgerald, Picard and Silver, J. Stat. Phys. (2000)

Cyclohexane

System Size Dependence

n-Octane

Squalane

Reservoir Type dependence

Squalane Octane

Pyrene

Propanol

System Size Effects (Propanol)

SPC/E water

Flexible SPC Water

C1mim-BF4

Conclusion
Reservoir based grand canonical monte carlo approach provides a reasonably efficient method to compute the phase coexistence properties of complex fluids Temperature expanded ensemble provide an efficient means to trace the coexistence curve over a broad temperature range

Acknowledgements
I would like to thank
Dr.Jeffrey Errington Dr.David Kofke Vaibhaw Kumar Kaustubh S. Rane Aravindh Kamakshinadha Wenjingg

SPC/E

Critical Point determination

Using law of rectilinear diameters

+ 2

= +
1

Scaling law
=

To find the critical pressure

= +