Molecular Modeling and Simulation of Thermodynamic ... · Molecular Modeling and Simulation of Thermodynamic Properties of Fluids for Industrial Applications Hans Hasse1 and Jadran

PROF. DR.-ING. HABIL. JADRAN VRABEC THERMODYNAMICS AND ENERGY TECHNOLOGY

MECHANICAL ENGINEERING DEPARTMENT

ThEt

PPEPPD, Suzhou, May 17, 2010 Molecular Modeling and Simulation of Thermodynamic Properties of Fluids for Industrial Applications Hans Hasse1 and Jadran Vrabec2 1Laboratory of Engineering Thermodynamics, TU Kaiserslautern 2Thermodynamics and Energy Technology, University of Paderborn

Laboratory of Engineering Thermodynamics Prof. Dr.-Ing. H. Hasse



ThEt

Scales relevant to thermodynamics



ThEt

Molecular dynamics (MD) deterministic system static and dynamic

properties straightforwardly applicable

to non-equilibria

Monte-Carlo (MC) statistical approach energetic acceptance criteria only static properties

Molecular simulation



ThEt

Phosgene group

C6H5-Cl C6H6 CCl2O o-C6H4-Cl2 HCl C6H5-CH3

Industrially important hazardous systems

Ethylene Oxide group

C2H4O H2O HO(CH2)2OH

High economic interest Difficult experiments Few reliable data Need for predictive modeling and simulation

Excellent test cases for molecular modeling and simulation



ThEt

Molecular model type, e.g. for Ethylene Oxide

Rigid, non-polarizable Multicenter Lennard-Jones + electrostatic sites United atom approach Development based on quantum chemical and VLE data



ThEt

Geometry Hartree-Fock with small basis set (e.g. 6-31G) or DFT methods

Electrostatics from electron density distribution Møller-Plesset2 with medium, polarizable basis set

(e.g. 6-311+G**) Embedded in a dielectric cavity (COSMO) to account

for a liquid (dense) phase Dispersion and repulsion

At least dimers have to be regarded CCSD(T) or Møller-Plesset2 with large basis set

(TZV or QZV) Large computational effort due to large basis set

and accurate electron correlation

Better to be optimized to VLE data

Molecular properties from quantum chemistry



ThEt

Density [mol/l]0 5 10 15 20 25

Tem

pera

ture

[K]

200

300

400

500

Temperature [K]200 300 400 500

Pres

sure

[MPa

]

0

2

4

6

8

Optimization to experimental VLE data Lennard-Jones parameters optimized to

saturated liquid density and vapor pressure

Optimization result:

correlations of exp. data simulation

— ● 0.4%′δρ =

p 1.5%δ =

temperaure / K density / mol/l

tem

pera

ure

/ K

pres

sure

/ M

Pa



ThEt

Deviations from reference values

uncertainty of reference new model

deviation from experiment / %-40 -20 0 20 40 60

sat. vap. thermal conductivitysat. liq. thermal conductivity

sat. vapor shear viscositysat. liquid shear viscosity

surface tensionsat. vap. isoth. compressib.

sat. liq. isoth. compressib.sat. vapor isob. heat capacitysat. liquid isob. heat capacity

critical temperaturecritical density

normal boiling temperatureenthalpy of vaporization

vapor pressure2nd virial coefficient

sat. vapor densitysat. liquid density

deviation from experiment / %-40 -20 0 20 40 60

sat. vap. thermal conductivitysat. liq. thermal conductivity

sat. vapor shear viscositysat. liquid shear viscosity

surface tensionsat. vap. isoth. compressib.

sat. liq. isoth. compressib.sat. vapor isob. heat capacitysat. liquid isob. heat capacity

critical temperaturecritical density

normal boiling temperatureenthalpy of vaporization

vapor pressure2nd virial coefficient

sat. vapor densitysat. liquid density

deviation from reference / % Industrial Fluid Properties Simulation Challenge 2007



ThEt

Phosgene group: pure component models

C6H5-Cl C6H6 CCl2O o-C6H4-Cl2 HCl C6H5-CH3

Phosgene Benzene Hydrogen Chloride

Toluene Chloro-Benzene

o-Dichloro-Benzene



ThEt T / K

0.002 0.003 0.004 0.005 0.006 0.007

ln(p

/ M

Pa)

-15

-10

-5

0HCl

Benzene

Cl-Benzene

Phosgene

Toluene

oCl2-Benzene

Vapor pressure

● Simulation ── DIPPR correlations



ThEt ρ / mol/l

0 10 20 30

T / K

200

300

400

500

600

700

HCl

Phosgene

oCl2-Benzene

Cl-Benzene

Toluene

Benzene

Saturated densities ● Simulation ── DIPPR correlations



ThEt

A A

B B

σA, εA

σB, εB

σAB, εAB

Unlike interaction A-B: Electrostatics fully predictive Lennard-Jones parameters from combination rules

Molecular modeling of mixtures

( )AB A B+= / 2σ σ σ

or

State-independent parameter ξ fitted to one experimental data point p(T,x) oder H(T)

ξ = 1 Predictions

Modified Lorentz-Berthelot ξ ⋅AB A B=ε ε ε



ThEt

Vapor-liquid equilibria of mixtures Applied to >360 binaries, >35 ternaries and one quaternary



ThEt

Phosgene group: studied binary mixtures

Phosgene + Benzol Phosgene + Toluol

Phosgene + Cl-Benzene

Phosgene + o-Cl2-Benzene

HCl + Benzene HCl + Toluene HCl + Cl-Benzene HCl + o-Cl2-Benzene

Phosgene + HCl



ThEt xHCl / mol mol-1

0.0 0.2 0.4 0.6 0.8 1.0

p / M

Pa

0

5

10

15

283.15 K

423.15 K

393.15 K0.05 0.10

0.0

0.1

0.2

0.3

Vapor-liquid equilibrium of HCl + Chlorobenzene

ξ = 1.02

Case: good predictions both by molecular simulation and EOS deviations upon approaching critical point

+ Experiment ● Simulation ── Peng-Robinson EOS



ThEt

Ethylene Oxide + Water

Ethylene Oxide + Ethylene Glycol

Water + Ethylene Glycol

Ethylene Oxide group: studied binary mixtures



ThEt

0.0 0.2 0.4 0.6 0.8 1.0

350

4000.4428 MPa

T / K

xEO / mol mol-1

kij = -0.1ξ = 1.200

Vapor-liquid equilibrium of Ethylene Oxide + Water


Case: good predictions by molecular simulation wrong prediction by EOS



ThEt xMeOH / mol mol-1

0.0 0.2 0.4 0.6 0.8 1.0

η / 1

0-4 P

a s

0

10

20

30

Shear viscosity of Methanol + Water

o Experiment ● Simulation

TIP4P/2005 Water model by Vega et al.

Methanol model, this work

5 °C

25 °C

0.1 MPa



ThEt xMeOH / mol mol-1

0.0 0.5 1.0

Di

/ 10-9

m2 s

-1

0.0

0.5

1.0

1.5

2.0

2.5

0.0 0.5 1.0

Self-diffusion coefficients of Methanol + Water TIP4P/2005 Water model by Vega et al.

Methanol model, this work

5 °C

25 °C

5 °C

25 °C

Water Methanol

o Experiment ● Simulation

0.1 MPa



ThEt

ms2: molecular simulation tool Molecular dynamics / Monte Carlo Arbitrary mixtures of rigid molecules Several ensembles Grand Equilibrium method for VLE calculations

Many static properties (thermal, caloric, entropic) Transport properties (Green-Kubo)

Consistent FORTRAN90 code Reasonably object oriented Distributed memory parallelization by MPI All relevant loops vectorized Interface to 2,5D (OpenGL) and 3D Virtual Reality visualization



ThEt

Monte Carlo embarrassing parallelization

Instead of performing one consecutive

probabilistic simulation over K loops

Efficient, as hardly any communication is necessary

M ensembles (on M

processors) can be

performed over K / M loops



ThEt number of processes

20 40 60 80 100 120

spee

d-up

20

40

60

80

100

120

Monte Carlo speed-up – without equilibriation

1372 particles

640.000 loops



ThEt

Summary

Molecular force fields are comprehensive models for thermodynamic properties having strong predictive power

They allow for an efficient description of mixture properties

Molecular simulation is a versatile tool to investigate the behavior of fluids which is about to be transferred to industry

Using parallel simulation codes and appropriate computing equipment, acceptable response times may be achieved

Molecular modeling and simulation can be used to investigate a large variety of topics



ThEt xPhosgene / mol mol-1

0.0 0.2 0.4 0.6 0.8 1.0

p / M

Pa

0

2

4

308.15 K

423.15 K

448.15 K

0.00 0.05 0.100.00

0.02

0.04

Vapor-liquid equilibrium of Phosgene + Toluene

ξ = 0.99

Case: Molecular simulation and EOS agree well




ThEt

Binary VLE of C2H6 + R22

+ experiment

PR-EOS, kij adjusted

simulation, ξ adjusted

simulation, ξ = 1

simulation, ξ = 0,95 to 1,05



ThEt

Grand Equilibrium method

( ) ( ) ( )µ ≈ µ + ⋅ −l l li i 0 i 0p p v p p

Pseudo grand canonical simulation (at , V, T) ( )µl

i p

Specified: T, x

p, y

Liquid Vapor

NpT calculation of chemical potentials and partial molar volumes

Documents

Molecular Modeling and Simulation of Thermodynamic ... · Molecular Modeling and Simulation of Thermodynamic Properties of Fluids for Industrial Applications Hans Hasse1 and Jadran