Introducing COSMO-based models, including F-SAC€¦ · aRenan P. Gerber and Rafael de P....

Introducing COSMO-based models, including F-SAC

Prof. Rafael de Pelegrini Soares, D.Sc.

FED. UNIV. OF RIO GRANDE DO SUL - BRAZILCHEMICAL ENGINEERING DEPARTMENT

LAB. VIRTUAL DE PREDICAO DE PROPRIEDADEShttp://www.enq.ufrgs.br/labs/lvpp

January 16, 2017

http://www.enq.ufrgs.br/labs/lvpp


IntroductionSome Challenges

Conclusions

COSMOCOSMO-RS or COSMO-SACSigma profile

Map

Prof. Rafael de Pelegrini Soares, D.Sc. Some challenges for COSMO-based models


Conclusions


Our Group: Virtual Laboratory for Property Predictions (LVPP)



Conclusions


COSMO method (solutes alone)

The COSMOa method was originally developed for thecomputation of solvation effects

The method belongs to the class of dielectric continuum models

,the cavities are discretized into segments or patches

aA Klamt and G Schuurmann. In: J. Chem. Soc., Perkin Trans. 2 (1993), pp. 799–805 Prof. Rafael de Pelegrini Soares, D.Sc. Some challenges for COSMO-based models


Conclusions


COSMO method (solutes alone)

The COSMOa method was originally developed for thecomputation of solvation effects

The method belongs to the class of dielectric continuum models,the cavities are discretized into segments or patches

aA Klamt and G Schuurmann. In: J. Chem. Soc., Perkin Trans. 2 (1993), pp. 799–805 Prof. Rafael de Pelegrini Soares, D.Sc. Some challenges for COSMO-based models

COSMO-RS – Surface contacting theory (mixtures)

In the COSMO-RSa methods we rely onCOSMO computations

Based on these pure substancecomputations, the mixture behavior ispredicted (γi )

The COSMO-SACb formulation follows thesame idea

aAndreas Klamt. In: The J. of Phys. Chem. 99.7 (1995),pp. 2224–2235

bShiang-Tai Lin and Stanley I. Sandler. In: Ind. Eng.Chem. Res. 41.5 (2002), pp. 899–913


Conclusions


COSMO-RS – Surface contacting theory

For each surface pair contact, there is an energy change

This results in different behavior for different substancesin solution

Clearly, there are many different possible contactingarranges and a statistical thermodynamics treatment isnecessary


COSMO-RS – Surface contacting theory (mixtures)

Based on these pure substance computations,the mixture behavior is predicted (γi )

“It is always desirable to express the properties of a solutionin terms that can be calculated completely from theproperties of the pure components.” – J. M. Prausnitz.Molecular thermodynamics of fluid-phase equilibria. Third.Prentice-Hall, 1999.

Chloroform/tetrahydrofuran at 30◦C


Conclusions


Sigma profile – p(σ)

For a statistical thermodynamicstreatment (without using MD),the 3D apparent surface chargesare projected into a simplehistogram

These pure compounddistributions, known as sigmaprofiles – p(σ), are the basis forcomputing the activity coefficientsin mixture

“It is always desirable to express the properties of a solution in terms that can be calculated completelyfrom the properties of the pure components.” – J. M. Prausnitz. Molecular thermodynamics offluid-phase equilibria. Third. Prentice-Hall, 1999.



Conclusions

Sigma profile databaseReliability

Sigma profile database

The quantum chemistry calculations for generating the sigmaprofiles represent the most time-consuming aspect ofCOSMO-based methods

There are several different quantum chemistry packagesimplementing the COSMO method, e.g.: Gaussian, Turbomole,MOPAC, DMol3 and GAMESS.

However, different packages lead to different sigma profiles a,requiring specific model parametrizations b

aEric Mullins et al. In: Ind. Eng. Chem. Res. 45.12 (June 2006), pp. 4389–4415.doi: 10.1021/ie060370h

bR.P. Gerber and R. de P. Soares. In: Braz. J. of Chem. Eng. 30 (1 2013),pp. 1–11


http://dx.doi.org/10.1021/ie060370h


Conclusions


VT-2005 database

The freely available database known as VT-2005 contains 1432compoundsa, mostly solvents and small molecules

The VT-2006 contains 206 pharmaceutical-related molecules

VT-databases were constructed using DMol3 (Materials Studio)which is very expensive. . .

aEric Mullins et al. In: Ind. Eng. Chem. Res. 45.12 (June 2006), pp. 4389–4415.doi: 10.1021/ie060370h


http://dx.doi.org/10.1021/ie060370h


Conclusions


Alternatives available

Alternative database with nearly 1000 compoundsa usingMOPAC (semi-empirical – could lead to poor results)

GAMESS (freely available, including source code) can be usedb

aRenan P. Gerber and Rafael de P. Soares. In: Ind. Eng. Chem. Res. 49.16 (Aug.2010), pp. 7488–7496

bShu Wang et al. In: Fluid Phase Equilib. 276.1 (2009), pp. 37–45

Suggestion

Assemble an extensive and freely available database using the rigorous methodsavailable in GAMESS



Conclusions


Reliability: The good. . .

There are several very good results in the literature, for instance Eugene Paulechkaet al. In: J. Chem. Eng. Data (2015)



Conclusions


The good, the bad and the ugly. . .

There are also bad and ugly results around, e.g. Hans Grensemann andJurgen Gmehling. In: Ind. Eng. Chem. Res. 44.5 (2005), pp. 1610–1624, foracetone(1)/diisopro-pylether and perfluorohexane(1)/n-hexane, respectivelly:



Conclusions


Precision of COSMO-based models

COSMO-based models have exceptional theoretical features

However, very often empirical modifications are needed for improving theagreement with experimental data, for instance:

Corrections in the water apparent surface charge1

Empirical scaling factors2 or empirically scaled surface areas3

Special parametrizations for different families, e.g. alcohols4

. . .

1H. Grensemann and J. Gmehling. In: Ind. Eng. Chem. Res. 44.5 (2005), pp. 1610–1624.2Renan P. Gerber and Rafael de P. Soares. In: Ind. Eng. Chem. Res. 49.16 (Aug. 2010),

pp. 7488–7496.3Jens Reinisch et al. In: Fluid Phase Equilib. 310.1-2 (2011), pp. 7–10.4Robert Franke, Bernd Hannebauer, and Sebastian Jung. In: Fluid Phase Equilib. 340 (2013),

pp. 11–14.Prof. Rafael de Pelegrini Soares, D.Sc. Some challenges for COSMO-based models


Conclusions


F-SAC: Functional-Segment Activity Model

In the F-SAC model, the quantumchemistry computation is replaced byan artificial sigma-profilea

Reduced predictive power

But hopefully, with increasedresolution and with less parametersthan UNIFAC-type models

aR. de P. Soares and R. P. Gerber. In: Ind. Eng.Chem. Res. 52 (2013), pp. 11159–11171



Conclusions


F-SAC: Functional-Segment Activity Model

F-SAC and COSMO-SAC model equations are identical, the difference is in thesigma profileIn the F-SAC model there is a neutral and two charged peaks per functional groupSimply add the sigma-profiles of functional groups to get the sigma-profile ofmolecules


Why don’t just use UNIFAC instead?

According to UNIFAC (Do) revision 5 – Antje Jakob et al. In: Ind. Eng. Chem. Res.45.23 (2006), pp. 7924–7933.


Conclusions


F-SAC parameters

For the 24 grupos considered, only 152 parameterswere calibrated.

Only the hydrogen-bonding? energies are pairwise

All other parameters are for pure groups alone,reducing the total number of parameters

In order to represent the same molecules in mixtures,778 parameters are used in UNIFAC (Do).



Conclusions


VLE predictions for non-associating mixtures

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0x1, y1

0.075

0.100

0.125

0.150

0.175

0.200

0.225

0.250

Pre

ssu

re[b

ar]

Exp. 298.2 K

F-SAC

UNIFAC(Do)

COSMO-SAC

(a) Chloroform/n-heptane

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0x1, y1

0.35

0.40

0.45

0.50

0.55

0.60

0.65

0.70

0.75

0.80

0.85

0.90

0.95

Pre

ssu

re[b

ar]

Exp. 323.15 K

F-SAC

UNIFAC(Do)

COSMO-SAC

(b) Acetone/cyclohexane

R. de P. Soares and R. P. Gerber. In: Ind. Eng. Chem. Res. 52 (2013), pp. 11159–11171



Conclusions


VLE predictions for non-associating mixtures

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0x1, y1

2.00

2.25

2.50

2.75

3.00

3.25

3.50

3.75

Pre

ssu

re[b

ar]

Exp. 283.6 K

F-SAC

UNIFAC(Do)

COSMO-SAC

(c) Dimethyl ether/1-butene

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0x1, y1

0.65

0.70

0.75

0.80

0.85

0.90

0.95

1.00

1.05

1.10

Pre

ssu

re[b

ar]

Exp. 323.15 K

F-SAC

UNIFAC(Do)

COSMO-SAC

(d) Methyl acetate/1-hexene

R. de P. Soares and R. P. Gerber. In: Ind. Eng. Chem. Res. 52 (2013), pp. 11159–11171



Conclusions


F-SAC parameters for Hydrogen-Bonding mixtures (association)

HB-acceptor is red (positive), HB-donor is blue (negative)5

5R. de P. Soares, R.P. Gerber, et al. In: Ind. Eng. Chem. Res. 52 (2013), pp. 11172–11181.Prof. Rafael de Pelegrini Soares, D.Sc. Some challenges for COSMO-based models


Conclusions


VLE predictions for associating mixtures

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0x1, y1

0.25

0.30

0.35

0.40

0.45

0.50

0.55

0.60

0.65

0.70P

ress

ure

[ba

r]Exp. 298.15 K

F-SAC

UNIFAC(Do)

(e) diethyl-ether/chloroform

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0x1, y1

0.25

0.30

0.35

0.40

0.45

0.50

0.55

0.60

0.65

0.70

0.75

0.80

0.85

0.90

0.95

Pre

ssu

re[b

ar]

Exp. 380.15 K

Exp. 368.15 K

F-SAC

UNIFAC(Do)

(f) toluene/3-methyl,1-butanol

R. de P. Soares, R.P. Gerber, et al. In: Ind. Eng. Chem. Res. 52 (2013), pp. 11172–11181



Conclusions


IDAC comparison for associating mixtures (water excluded)

-2 -1 0 1 2 3 4 5 6Logarithm of Experimental IDAC

-2

-1

0

1

2

3

4

5

6

Log

arith

m o

f M

odel

ID

AC

(g) F-SAC

-2 -1 0 1 2 3 4 5 6Logarithm of Experimental IDAC

-2

-1

0

1

2

3

4

5

6

Log

arith

m o

f M

odel

ID

AC

(h) UNIFAC (Do)




Conclusions


IDAC comparison for mixtures with water

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20Logarithm of Experimental IDAC

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

Log

arith

m o

f M

odel

ID

AC

(i) F-SAC

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20Logarithm of Experimental IDAC

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

Log

arith

m o

f M

odel

ID

AC

(j) UNIFAC (Do)




Conclusions


FSAC: Hydrocarbon-water mutual solubilities

275 300 325 350 375 400 425 450 475

Temperature [K]

0.00001

0.0001

0.001

0.01

0.1M

ola

r S

olu

bili

tyExp. S1

Exp. S2

F-SAC*

UNIFAC(PSRK)

UNIFAC(Do)

UNIFAC-LLE

(k) n-Hexane-Water

275 300 325 350 375 400 425 450 475 500 525 550 575

Temperature [K]

0.0001

0.001

0.01

0.1

Mo

lar

So

lub

ility

Exp. S1

Exp. S2

F-SAC*

UNIFAC(PSRK)

UNIFAC-LLE

(l) Toluene-Water

L.F.K. Possani et al. In: Fluid Phase Equilib. 384 (2014), pp. 122–133


Predicting high-pressure VLE with SCMR

The F-SAC can be combined with SRK with Mathias-Copeman α function andthe Self-Consistent Mixing Rule6 (SCMR) for propane-benzene, no parameter isadjusted for the mixture effects:

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0x1, y1

0

5

1 0

1 5

2 0

2 5

3 0

3 5

4 0

4 5

5 0

5 5

6 0

6 5

Pre

ssu

re[b

ar]

Exp. 310.93 K

Exp. 344.26 K

Exp. 377.59 K

Exp. 477.59 K

SRK(MC)+SCMR(FSAC)

6Paula B. Staudt and Rafael de P. Soares. In: Fluid Phase Equilibria 334 (2012), pp. 76–88.


Conclusions


Extension for electrolytes: eF-SAC

Ions considered spherical with a given charge (+1,-1, +2, -2, etc.)

Then the only parameter is the surface area (waterparameters from previous works)

To avoid direct contact between ions, theirinteraction energy is assumed high



Conclusions


eF-SAC σ-profiles

Group Q+k (A

2) Q−

k (A2) σ+

k (eA−2

)

Li+ 0 36,96 -0,027Na+ 0 29,05 -0,034K+ 0 47,13 -0,021Rb+ 0 40,17 -0,025Cs+ 0 41,64 -0,024F− 27,12 0 0,037Cl− 39,19 0 0,026Br− 30,11 0 0,033I− 38,69 0 0,026



Conclusions


Results: σ-Profile for eF-SAC



Conclusions


Results: γ±

NaFaq



Conclusions


Results: γ±

LiIaq



Conclusions


Results: γ±

NaBraq



Conclusions


Results: γ±

LiBraq



Conclusions


Results: γ±

CsIaq



Conclusions

ConclusionsLinks and more info

Conclusions

COSMO-based models have exceptional theoretical features

An extensible, open, and rigorous sigma-profile database needs to be developed

Currently, results can be considered semi-quantitative

Agreement with experimental data is usually obtained by means of empiricalcorrections

The variation known as F-SAC usually produces better results but depends onfitted parameters

There are certanly many other opportunities for improvement and new fields forapplication



Conclusions

ConclusionsLinks and more info

Thank you!

Molecules database at http://code.google.com/p/jcosmo/

Download the F-SAC demonstration code athttp://www.enq.ufrgs.br/labs/lvpp

F-SAC is already available in the iiSE process simulator and being integrated intoDWSim (open source) process simulator

Contact: [email protected]


http://code.google.com/p/jcosmo/


[email protected]

Introducing COSMO-based models, including F-SAC€¦ · aRenan P. Gerber and Rafael de P....

Documents

Transcript of Introducing COSMO-based models, including F-SAC€¦ · aRenan P. Gerber and Rafael de P....