Predicting Solubilities of

Cellulose in Ionic Liquids using COSMO-RS

Kai Masuch, Kai Leonhard, Jens Kahlen

• Tree: lignin, hemicellulose, cellulose, other stuff we don’t want (e.g. water)

• cellulose: hydrogen-bonding hinders conversion to liquid..

• ...which can be overcome when dissolved in appropriate solvents...

Why predict cell. solubilities?

• ...e.g. ionic liquids (IL’s)...

• ...which are trendy and referred as ‘green’ by marketing strategists (negligible vapor pressure).

• Cations and anion can be combined freely: numerous (!!!) candidates.

• solubility (in general) can be predicted by COSMO-RS

• fast (contrary to MD and full scale ab-initio methods for liquids)

• huge database for screening purposes

• fully predictive due to previous ab-inito calculations for each molecule (contrary to group contribution methods e.g. UNIFAC)

Why COSMO-RS?• nice colours (simplifies

understanding of interactions)

• (Luckily) many ‘weak points’ of COSMO-RS are touched:

• Combinatorial entropy for polymers

• Electrolytes in general (no long range electrostatics)

• Hydrogen bonds

• Atomic ions

COSMOOctane

Cl-

H2O

• COSMO: COnductor-like Screening MOdel

• QM-computation of surface-charge-density in a conductor

• Only electrostatic interaction energy with dielectric continuum

• Screening charge probability distribution: ‘sigma-profile’

• Chemical potential is derived from sigma-profiles with COSMO-RS

Realistic Solvation COSMO (-RS)• Chemical potential:

• Partition function:

• Residual partition function:

*

COSMO-filesBMIM+

• COSMO-database: 5000 neutral components

• Modelling Ionic Liquids:

• One COSMO-file for each IL, (conformers!)

• Combination of COSMO-files of independent ions

• IL-database: 70 cations x 30 anions = 2100 IL’s

Cl-

OAc-PF6-

Sigma-Profiles of Conformers

• Conformers might have:

• very different sigma-profiles (intramolecular h-bonds e.g. cellulose)

• very similar sigma-profiles (Alkanes/Alkenes)

Cellulose in COSMO-RS

pics on top: http://en.wikipedia.org/wiki/Cellulose

Dissolution

Structure for COSMO-file

(size and shape effects on entropy neglected)

(memory likewise)

Weighting of atoms• Only screening

surface charges of mid-unit should be used to calculate chemical potential

• screening surface charges associated to atoms of end-units are excluded

Generation of conformers

• Generation of cellotriose-conformers in gas phase

• Geometry optimization in COSMO-cavity

• Conformer clustering:

• COSMO energies (conformer distribution by Boltzmann weighting)

• Similarities in sigma-profiles

Match!

(30 clusters)

Solubility-map

Cations

Ani

ons

good

bad

Gua

nidi

um+

BM

IM+

Cl-EMIM+/OAc-

Map for 0,1% Water-Content (similar to1% and 5%)

Cations

Ani

ons

Gua

nidi

um+

BM

IM+

Anions

BMIM:+H2O:

1000ppm

exp. Data: T, DP, chryst., water % = const.(!)

Data: Vitz, et.al. 09exp.: 12% 20% 3% 2% no sol. no sol. no sol.

Anions

Hexa-methyl-guanidinium+

Acetate-

H-MeGu+H20

1000 ppm

PF6-(bad)

Cl-(good)

OAc-(very good)

‘Typical’ cationBMIM+

• Solubility mostly cation independent, probably due to homogenous charge distribution over large surface

• Solubility explanation: IL-self-associativity when sigma-profiles of + mirrors -

‘Atypical’ cation

Cl-(bad)

Guani-dinium

OAc-(bad)

G+/OAc- have strong symmetryto mirror plane: bad solubility

PF6-(good)

G+/PF6- not:interaction withcellulose morelikely

Summary/Outlook• Application of COSMO-RS as

preselection tool for IL’s dissolving cellulose

• IL’s found which are expected to solve cellulose better than commonly used IL’s...

• ...which seem to be highly intolerant to water (design has to include water effects)

• No kinetics included

• Comparison with exp. Data is difficult

• Forthcoming transfer of the model in finite dilution might improve comparability

• Still degree of polymerisation and crystallinity have to be known

• State of aggregation for IL’s should be included in the model

The End

Atomic Ions in COSMO-RS

COSMO-RSno hb

COSMO-RShb

measured(BioVT)

H2O mol/mol

activ

ity H

2O

activity of water in 1-Allyl-3-Methyl-Imidazolium-chloride

COSMO• ‘Continuum Solvation Model’

• COSMO: COnductor-like Screening MOdel (infinite permittivity)

• electron density and screening charge: iteration to self-consistency

• Dielectric solvent: scaling of screening charges by empiric function (function of permittivity)

• Only electrostatic interaction energy with dielectric continuum

O

H

H

+1.5 e/nm²

-1.5 e/nm²

H-Bonds + Dispersion

*http://www.lsbu.ac.uk/water/

(iterative)

Solubility from COSMO-RS (constant for

cellulose with given DP and crystallinity)

Basic COSMO-RSSize/shape (entropy):e.g. Flory-Huggins

COSMO-RS-comb. not suitable for polymers (Klamt)

Contact statistics• best ‘typical’: hexamethylguanidinium-acetate

• 6/30 conformers used best: weight of 83% (No.23 dE?)

• best ‘atypical’: guanidinium-tris(pentafluoroethyl) trifluorophosphate

• 23/30 conformers used best: weight of 34%

Lignin

http://de.wikipedia.org/wiki/Lignin