PROF. DR.-ING. HABIL. JADRAN VRABEC ThET

Fully automatized determination of

Fundamental Equations of State based on

molecular simulations in the cloud

CAPE Forum

Paderborn, 29.04.2015

Andreas Köster

Tao Jiang

Gábor Rutkai

Colin W. Glass

Jadran Vrabec

PROF. DR.-ING. HABIL. JADRAN VRABEC ThET

R. Span, “Multiparameter Equations of State”, Springer, Berlin (2000)

accurate knowledge in entire fluid region : ~10 substances

For pure chemical substances…

satisfactory knowledge : <100 substances

For mixtures…

the knowledge is much worse

Problem: scarce experimental data

PROF. DR.-ING. HABIL. JADRAN VRABEC ThET

Fundamental equation of state (FEOS) correlation

, ,E N V S

, ,H N p S , ,F N V T , ,G N p T

/ ,1 / ,F T N T V

E

H

F

G

S

V

p

T

N

: internal energy

: enthalpy

: Helmholtz energy

: Gibbs energy

: entropy

: volume

: pressure

: temperature

: particle number

= “explicit function in terms of N, (1/T), and V”

Every thermodynamic property is

a combination of

/

1 /

xyz

x y z

F T

T V N

PROF. DR.-ING. HABIL. JADRAN VRABEC ThET

/1 /

1 /

mnm n

mnm n

F RTT A

T

pressure

internal energy

enthalpy

01

pA

RT

10

EA

RT

10 01

HA A

RT

Thermodynamic properties from FEOS

( / )N V

20

vc

AR isochoric heat cap.

PROF. DR.-ING. HABIL. JADRAN VRABEC ThET

2

01 11

20

01 022

pA Ac

AR A A

2

201 11

01 02

20

2A AMw

A ART A

01 02 11

2

01 11 20 01 022

A A AR

A A A A A

speed of sound

Joule-Thomson coeff.

isobaric heat cap.

Thermodynamic properties from FEOS

02 01

1

2T

RTA A

isothermal comp.

PROF. DR.-ING. HABIL. JADRAN VRABEC ThET

Molecular Simulation

Molecular dynamics (MD) Monte Carlo (MC)

PROF. DR.-ING. HABIL. JADRAN VRABEC ThET

Simulation framework

R. Lustig., Mol. Phys., 110, 3041 (2012).

R. Lustig., Mol. Sim., 37, 457 (2011).

R. Lustig., J. Chem. Phys., 100, 3060 (1994).

A single MD or MC

simulation run per state

point yields:

ms2

C. W. Glass et al., Comp. Phys. Comm., 185, 3302 (2014).

/

1 /

mn

mnm n

F TA

T

p

cv

www.ms-2.de

cp ,w

↓

f(A01, A20, A11, A02)

PROF. DR.-ING. HABIL. JADRAN VRABEC ThET

Molecular simulation as engineering mainstream tool* (for equilibrium thermodynamic property generation)

*E. J. Maginn, AIChE J., 55, 1304 (2009).

Advantages: - much cheaper than experiment

- faster than experiment (once source code is available)

- several thermodynamic properties simultaneously

- applicable under all conditions

Problems: - dependent on the quality of molecular models

- code development and usage require considerable

expertise

PROF. DR.-ING. HABIL. JADRAN VRABEC ThET

i i i i i

N Mt d t d cRe s

i i

i 1 j k 1

, a a exp ...

Cloud-based automatized FEOS fitting

Goals: - target substances/fluid regions for which there are no data

available

- perform the simulation and obtain a FEOS within 24 h

- little human interaction

How: - use simulation data only

- simple fitting approach (linear fit)

(everything else is taken from generalized FEOS)

A

PROF. DR.-ING. HABIL. JADRAN VRABEC ThET

User‘s task: - select state points and molecular model

Cloud-based automatized FEOS fitting

PROF. DR.-ING. HABIL. JADRAN VRABEC ThET

Cloud-based automatized FEOS fitting

Automatized deployment:

- Distribution of simulations to

available High Performance

Computers (HPC) in the

Cloud

PROF. DR.-ING. HABIL. JADRAN VRABEC ThET

User‘s task: - select fitting options and check results

Cloud-based automatized FEOS fitting

PROF. DR.-ING. HABIL. JADRAN VRABEC ThET

~ 300 state points x 9 derivatives (= 2700 measurements)

Cloud-based automatized FEOS fitting

Example: Phosgene (extremely toxic, yet millions of tons produced annually)

PROF. DR.-ING. HABIL. JADRAN VRABEC ThET

Quality of the FEOS from Simulation (Phosgene)

FEOS : G. Rutkai, J. Vrabec, J. Chem. Eng. Data, submitted.

PROF. DR.-ING. HABIL. JADRAN VRABEC ThET

Quality of the FEOS from Simulation (Phosgene)

FEOS : G. Rutkai, J. Vrabec, J. Chem. Eng. Data, submitted.

Experiment

Present FEOS

VLE curve from present FEOS

PROF. DR.-ING. HABIL. JADRAN VRABEC ThET

Quality of the FEOS from Simulation (Ethylene Oxide)

100(

Data

–

FE

OS)

/

Data

100(w

Da

ta –

wF

EO

S)

/ w

Da

ta

p / MPa p / MPa

Experiments

Molecular simulation based FEOS (top plots)

FEOS : A. Köster, T. Jiang, G. Rutkai, Colin W. Glass, J. Vrabec, to be submitted.

Experiment based FEOS, advanced fitting approach (bottom plots)

,

PROF. DR.-ING. HABIL. JADRAN VRABEC ThET

Molecular simulation data sets

• Useful for FEOS development (cost effective, fast)

• Applicable under all conditions

Cloud-based FEOS fitting

• Enables non-expert users to create FEOS in a matter of days

• Contribution to making molecular simulation more accessible

Outlook

• Creation of force fields may be automatized extending this workflow

Summary

PROF. DR.-ING. HABIL. JADRAN VRABEC ThET

Thank you for listening!

This project was funded by the European Union