Microsoft Word - FinalVersion.docxLiquid-liquid Chromatography
in der Flüssig-flüssig Chromatographie
Erlangen-Nürnberg zur Erlangung des Grades
DOKTOR-INGENIEUR
Friedrich-Alexander-Universität Erlangen-Nürnberg Tag der mündlichen Prüfung: 04.10.2017 Vorsitzender des Promotionsorgans: Prof. Dr.-Ing. Reinhard Lerch Gutachter/in: Prof. Dr.-Ing. Wolfgang Arlt Prof. Dr. Mirjana Minceva
Parts of this thesis have been published in the following journals:
1.) E. Hopmann, A. Frey, M. Minceva, J. Chromatogr. A, 1238 (2012) 68.
2) J. Goll, A. Frey, M. Minceva, J. Chromatogr. A, 1284 (2013) 59.
3) A. Frey, E. Hopmann, M. Minceva, Chem. Eng. Technol., 37 (2014) 1663.
Parts of this thesis have been presented at the following conferences:
1) A. Frey, E. Sponsel, J. Völkl, M. Minceva, L. Mokrushina, W. Arlt: COSMO-RS as a
valuable a priori prediction tool in the field of process synthesis, 9th World Congress of
Chemical Engineering (WCCE9), Seoul, Korea, 2013. (Oral presentation)
2) A. Frey, J. Goll, E. Hopmann, M. Minceva: From molecule to process: Development of
liquid-liquid chromatographic separation processes, 9th World Congress of Chemical
Engineering (WCCE9), Seoul, Korea, 2013. (Oral presentation)
3) A. Frey, M. Minceva, W.Arlt: Liquid-Liquid Chromatography: Tailoring a biphasic liquid
system for the separation of a given feedstock, Dechema Fachausschuss Extraktion, Baden-
Baden, Deutschland, 2013. (Oral presentation)
4) A. Frey, M. Minceva: Liquid-Liquid Chromatography: Application of a Predictive
Thermodynamic Model for the Selection of the Stationary and Mobile Phase, AICHE,
Pittsburgh, USA, 2012. (Oral presentation)
5) A. Frey, M. Minceva, W. Arlt: Flüssig-Flüssig Chromatographie: Systematische
Lösungsmittelauswahl für ein spezifisches Trennproblem, ProcessNet, Karlsruhe, Germany,
2012. (Oral presentation)
6) A. Frey, M. Minceva: Design of tailor-made stationary and mobile phase in liquid-liquid
chromatography, International Symposium on Preparative and Industrial Chromatography,
Brussels, Belgium, 2012. (Oral presentation)
7) A. Frey, Johannes Goll, M. Minceva: Design of a Continous Centrifugal Partition
Chromatographic Process, 7th International Conference on Countercurrent
Chromatography, Hangzhou, China, 2012. (Oral presentation)
8) A. Frey, M. Minceva: Selection of the mobile and stationary phase in support free liquid-
liquid chromatography applying a predictive thermodynamic model, 7th International
Conference on Countercurrent Chromatography, Hangzhou, China, 2012. (Oral
presentation)
9) A. Frey, E. Hopmann, M. Minceva: Liquid - Liquid Chromatography: tailoring the mobile
and stationary phase, 8th Congress of Chemical Engineering together with ProcessNet-
Annual Meeting, Berlin, Germany, 2011. (Oral presentation)
10) E. Hopmann, A. Frey, M. Minceva, W. Arlt: A-priori selection of the mobile and
stationary phase in centrifugal partition chromatography, First poster price, PREP 2011
"Preparative and Process Chromatography", Boston, MA, USA, 2011. (Poster)
The following supervised student theses contributed to this work. Every student gave
permission to be cited:
1) S. Röhrer, Anwendung von auf Ionischen Flüssigkeiten basierenden wässrigen
Zweiphasensystemen in der Flüssig-Flüssig Chromatographie, Bachelorarbeit, FAU
Erlangen, Erlangen, 2012.
2) L. Stumpf, Bestimmung der Verteilungskoeffizienten von Proteinen, in auf ionischen
Flüssigkeiten basierenden wässrigen Zweiphasensystemen, Bachelorarbeit, FAU Erlangen,
Erlangen, 2012.
3) L. Kempf, Experimentelle Validierung eines Ansatzes zur Auswahl der Lösungsmittel in
der Flüssig-Flüssig Chromatographie, Bachelorarbeit, FAU Erlangen, Erlangen, 2013.
4) A. Kosider, Vorhersage von Flüssig-Flüssig Gleichgewichten mittels COSMO-RS,
Bachelorarbeit, FAU Erlangen, Erlangen, 2013.
5) Lisa Britting, Einfluss physikalischer Größen auf die Trennleistung der Flüssig-flüssig
Chromatographie, Bachelorarbeit, FAU Erlangen, Erlangen, 2012.
Danksagung
Diese Arbeit ist das Ergebnis meiner wissenschaftlichen Arbeiten am Lehrstuhl für Thermische
Verfahrenstechnik der Friedrich Alexander-Universität Erlangen-Nürnberg.
Mein besonderer Dank gehört meinem Doktorvater Prof. Wolfgang Arlt sowie meiner
„Doktormutter“ Prof. Mirjana Minceva. Beide gaben mir die Möglichkeit ein selbstgewähltes Thema
zu bearbeiten und frei zu gestalten. Für das entgegengebrachte Vertrauen meine Resultate auf
nationalen und internationalen Konferenzen zu präsentieren bedanke ich mich herzlich.
Mein Dank gilt auch der übrigen Prüfungskommission. Für die Bereitschaft den Vorsitz zu
übernehmen bedanke ich mich bei Prof. Malte Kaspereit mit dem ich auch in meiner Zeit am
Lehrstuhl so manche lebhafte Diskussion führen durfte. Herzlichen Dank an Prof. Geoffrey Lee für
die Bereitschaft als weiteres Prüfungsmitglied mitzuwirken.
Großer Dank gebührt vor allem jedem einzelnen meiner Helfer und Unterstützer:
Das „TVT Labor“ um Petra Kiefer, Petra Koch, Roswitha Eckstein, Lisa Schnaus und Andreas
Postatny haben meine Arbeit mit zahlreichen Messungen unterstützt und standen mir bei der
Interpretation der Ergebnisse immer zur Seite. Nicht vergessen darf hier Edelgard Schuhmann
werden, von der ich eine Menge gelernt habe. Für die große Unterstützung von Prof. König möchte
ich mich an dieser Stelle ebenfalls bedanken.
Vera Cremers und Elisabeth Wenzel gehört mein Dank für die Organisation und der Hilfe bei kleinen
und großen Problemen.
Für die spaßigen und lehrreichen Jahre danke ich dem kompletten Lehrstuhl für Thermische
Verfahrenstechnik.
Für die fachliche und private Unterstützung möchte ich hier vor Allem Katharina Stark, Johannes
Hartmann, Johannes Goll, Irma Goll, Alexander Günther, Peter Hausmann, Alexander Buchele und
Karsten Müller hervorheben.
Für Ratschläge im Bereich Thermodynamik bedanke ich mich bei Liudmilla Mokrushina.
Meinen Studenten Simon Röhrer, Nora Rottmann, Lisa Stumpf, Lara Kempf, Axel Uwe Kosider und
Lisa Britting bin ich zu tiefem Dank verpflichtet, da ihr unermüdlicher Einsatz diese Arbeit überhaupt
erst möglich machten.
Der größte Dank gebührt meinen Eltern, Großeltern und meinem Bruder, die mich auch nach kleinen
und großen Rückschlägen immer wieder auffingen und auffangen.
Meinen Freunden bin ich ebenfalls für Ihre Unterstützung dankbar. Ohne sie wäre all das nicht
möglich gewesen.
Wichtig ist es mir, mich an dieser Stelle auch bei Dora zu bedanken, die mich unermüdlich in der
Endphase unterstützt hat.
„Die Theorie liefert viel, aber dem Geheimnis des Alten bringt sie uns kaum näher.
Jedenfalls bin ich überzeugt, dass der nicht würfelt.“ (Albert Einstein)
Table of Contents
Goals and structure of the thesis at hand ............................................................................ 5
Theoretical basis ......................................................................................................................... 6
Liquid-liquid chromatography ............................................................................................ 6
2.1.4 Solvent selection and typical solvents in liquid-liquid chromatography .................. 18
2.1.5 Challenges and problems in liquid-liquid chromatography ...................................... 23
Thermodynamics ............................................................................................................... 24
2.2.3 Liquid-liquid equilibrium .......................................................................................... 25
2.2.4 Partition coefficient ................................................................................................... 26
2.2.5 Solid-liquid equilibria ............................................................................................... 27
2.2.7 Conductor-like Screening Model for Real Solvents (COSMO-RS) ......................... 29
2.2.8 The UNIFAC model ................................................................................................. 31
Material and Methods ............................................................................................................... 33
Experimental Methods ...................................................................................................... 38
3.3.2 Determination of the stationary phase retention ....................................................... 38
3.3.3 Determination of the enthalpy of fusion and the melting temperature ..................... 38
3.3.4 Determination of the partition coefficient ................................................................. 39
Theoretical Methods ......................................................................................................... 41
3.4.1 Prediction of activity coefficients using COSMO-RS and UNIFAC ....................... 41
3.4.2 Prediction of the solubility ........................................................................................ 43
3.4.3 Prediction of the partition coefficient ....................................................................... 44
3.4.4 Partition coefficient screening of the target compound(s) with COSMO-RS .......... 44
3.4.5 Prediction of the LLE data with COSMO-RS and UNIFAC .................................... 45
3.4.6 Accuracy of the LLE prediction ............................................................................... 46
Results and Discussion ............................................................................................................. 47
LLE prediction .................................................................................................................. 53
Screening of the partition coefficient in different ARIZONA solvent systems using
thermodynamic models ................................................................................................................. 62
4.3.1 Study cases for the applicability of LLE data in different prediction models .......... 63
Table of Contents
iii
4.3.2 Selection of a biphasic solvent system using COSMO-RS and different sources of
LLE data .................................................................................................................................. 69
Biphasic solvent system selection procedure .................................................................... 76
4.4.1 Selection of a pool of solvents .................................................................................. 78
4.4.2 Gathering information about the solubility for target/s and identification of two
phase systems ............................................................................................................................ 78
4.4.3 Screening of potential biphasic liquid systems ......................................................... 79
4.4.4 Final solvent selection of the most appropriate biphasic solvent system ................. 79
Case studies for the evaluation of the developed solvent selection pathway ................... 80
4.5.1 Separation of a mixture of coumarin and piperine .................................................... 80
4.5.2 Separation of a mixture of methyl-, ethyl-, propyl-, and butylparaben .................... 92
Influence of the physical properties on the stationary phase retention ............................. 96
4.6.1 Stationary phase retention as a function of the density difference ........................... 97
4.6.2 Stationary phase retention as function of the viscosity of the mobile phase ............ 98
4.6.3 Stationary phase retention as a function of the interfacial tension !" 100
4.6.4 Stationary phase retention as function of the stability ............................................ 103
Evaluation of the applicability of ionic liquid based Aequous Two Phase System (ATPS)
in liquid-liquid chromatography ................................................................................................. 105
Letter Description Unit # surface area (Eq. 2.40) Å% # component A -
' activity - B component B -
C capacity - c concentration + ,-. cp isobaric heat capacity 0 ,-.1 c QSPR parameter - d diameter ,, F flow ,. ,45 G molar Gibbs free energy 0 ,-. f fugacity 8 ,% g gravitational acceleration , :% h molar enthalpy 0 ,-. i compound - j compound - K distribution coefficient - K calibration constant
viscosimeter (Eq. 3.1)
?,² :% k UNIFAC group - k component (Eq.2.10 - Eq.2.12) - L length ,, N number of stages - n number of moles - m number of values - P partition coefficient / GH =
JK L
JK M -
P pressure (Eq.2.10, Eq.2.13) N'O p charge density P #%
Rs resolution - R gas constant 0 ,-.1 r radius ,, !" stationary phase retention % s molar entropy 0 ,-. T temperature °V t time : q UNIFAC parameter - V volume ,. v molar volume ,[ \,-. v frequency (Eq. 2.44) 1 : w peak width ,, x molar fraction - z coordination number (Eq.
2.40)
Greek symbols
Letter Description Unit a chemical potential \0 ,-. b phase b - b separation factor - c phase c - d activity coefficient - d interfacial tension 8/, difference - g COSMO adjustable parameter h surface charge density P 5,% i volume fraction - j surface fraction - Γ activity coefficient of a group - l dynamic viscosity ,G': m dielectric constant / relative
permittivity
Superscripts
Letter Description ∞ infinitely diluted 0 reference state b phase b c phase c b activity coefficient
?-,N combinatorial part t excess Puv experimental w lower phase . liquid x upper phase OP: residual part y mobile phase ! stationary phase : solid zO triple point
Subscripts
vi
Subscripts
Letter Description 0 pure substance 4 component i
ideal ideal w lower phase P excess y mobile phase { retention ! stationary phase S solvent x upper phase # component A | component B v isobaric
Abbreviations
vii
Abbreviations
heptane/ethyl acetate/methanol/water ATPS aqueous two phase system CCC counter-current chromatography ChMWat chloroform/methanol/water COSMO-RS conductor like screening model for real solvents COSMO-SAC conductor like screening model segment activity
coefficient CPC centrifugal partition chromatography CPE centrifugal partition extractor DFT density functional theory etoac ethyl acetate exp experimental GC gas chromatography GUESS generally useful estimation of solvent systems HemWat hexane/ethyl acetate/methanol/water HPLC high performance liquid chromatography IL ionic liquid LL liquid-liquid LLC liquid-liquid chromatography LLE liquid-liquid equilibrium/a LSC liquid-solid chromatography MeOH methanol mibk methyl isobutyl ketone mtbe methyl tertiary butyl ether NRTL non-random two liquids PEG polyethylene glycol ReSS2 2-dimensional reciprocal shifted symmetry RMSE root mean square error Rpm rounds per minute SLE solid-liquid equilibrium/a TZVP triple zeta valence polarization UNIFAC universal quasi-chemical functional group activity
coefficient model UNIQUAC universal quasi chemical
Abstract
Abstract
The thesis at hand focuses on the development and the enhancement of a solvent selection
procedure for liquid-liquid chromatographic separations. This procedure simplifies the
design and the execution of a liquid-liquid chromatographic separation process due to the
replacement of measurements by thermodynamic predictions. The developed pathway
guides the user through the design of the solvent system for any liquid-liquid pulse injection-
chromatographic separation and includes the a-priori determination of the separation
performance characteristics on basis of thermodynamics as well as short cut methods to
judge about hydrodynamic performance characteristics of the particular solvent system. The
procedure is adaptable to any production scale requirements including productivity, costs
and safety.
In the designed solvent selection pathway, a connection between liquid-liquid
chromatography and thermodynamics is drawn. The a priori COnductor like Screening
MOdel for Real Solutions (COSMO-RS) and the structure interpolating method UNIversal
quasichemical Functional group Activity Coefficients (UNIFAC) are successfully applied
for screening purposes for the solvent selection in liquid-liquid chromatography.
The LLE data of potential solvent combinations as well as partition coefficients of model
compounds are predicted successfully using COSMO-RS and UNIFAC. It is shown that the
solubility can be estimated qualitatively in any solvent using COSMO-RS. The application
of thermodynamic models results in a reduction of the number of experiments carried out in
the solvent selection procedure and hence will safe costs and time.
The so designed solvent selection procedure, i.e. the decision-making steps are demonstrated
step by step for two model mixtures. One is the artificial feed of the compounds coumarin
and piperine, the other model mixture contains methyl-, ethyl-, propyl- and butyl paraben.
Other essential properties of the solvent systems are the characteristics of hydrodynamic. In
fact, all physical properties have a direct or indirect impact on the stationary phase retention.
A high stationary phase retention is very important for a separation with liquid-liquid
chromatography. Therefore, general guidelines for a suitable solvent system to achieve high
stationary phase retention in a dynamic CCC are developed.
Kurzfassung
Kurzfassung
Die vorliegende Arbeit zeigt die Entwicklung und die Erweiterung einer
Lösungsmittelauswahlmethode für Flüssig-flüssig chromatographische Trennprozesse.
thermodynamische Modelle anstatt von Messungen zum Einsatz kommen. Die so
entwickelte Methode führt den Anwender Schritt für Schritt durch den
Lösungsmittelauswahlprozess für jede Art von Pulsinjektionsexperiment in der Flüssig-
flüssig Chromatographie. In dem hier entwickelten Lösungsmittelauswahlprozess werden
auf Basis der Thermodynamik die essentiellen Trennparameter vorhergesagt, während die
Hydrodynamik auf Basis der physikalischen Eigenschaften des Lösungsmittelsystems
bewertet wird. Diese Methode kann an jegliche Anforderungen an ein Lösungsmittelsystem
einschließlich Kosten, Produktivität und Sicherheit angepasst werden.
Die Lösungsmittelauswahlprozedur, spannt den Bogen zwischen Chromatographie und
Thermodynamik. Hierfür wurden das thermodynamische a priori Model COnductor like
Screening MOdel for Real Solutions (COSMO-RS) und das strukturinterpolierende
UNIversal quasichemical Functional group Activity Coefficients (UNIFAC) Model
angewendet. Die Phasenzusammensetzungen (LLE) potentieller zweiphasiger
Lösungsmittelsysteme, sowie die für die Lösungsmittelauswahl unverzichtbaren
Verteilungskoeffizienten von Modelsubstanzen wurden mit COMSO-RS und UNIFAC
vorhergesagt. Es konnte zudem gezeigt werden, dass die Löslichkeit von Stoffen mit
COSMO-RS qualitativ abgeschätzt werden kann. Die Anwendung dieser Modelle bewirkt
somit eine deutliche Reduktion an Experimenten und spart so Kosten und Zeit.
Schritt für Schritt wird die hier entwickelte Prozedur an Modelmischungen gezeigt. Eine
dieser Mischungen ist ein künstlicher Feed aus Cumarin und Piperine, die Andere enthält
Methyl-, Ethyl-, Propyl- und Butyl paraben.
Ebenfalls essentiell ist das hydrodynamische Verhalten des Lösungsmittelsystems. Faktisch
haben alle physikalischen Eigenschaften einen direkten oder indirekten Einfluss auf den
stationären Phasenrückhalt. Ein hoher stationärer Phasenrückhalt ist wichtig für eine
effiziente Trennung mit Flüssig-flüssig Chromatographie. Hierzu werden Richtlinien
abgeleitet, um einen hohen Phasenrückhalt in einer dynamischen CCC zu erzielen.
Introduction
1
Introduction
State of the art
Chromatography invented by Mikhail Semjonowitsch Tswett one century ago is a highly
selective method frequently used in down-streaming processing [1]. Applying chromatography
complex mixtures of very similar compounds (like enantiomers) can be separated and product
purities higher than 99.9 % can be obtained. A chromatographic separation is realized as the
result of the different distribution of the substances present in the mixture to be separated
between two phases. One phase, called stationary phase, is held in place while the other one is
driven through the stationary phase and hence, called mobile phase. In terms of their state, the
mobile phase can be any fluid and the stationary phase can be solid or liquid [2].
In support-free liquid-liquid chromatography, the phases of a biphasic liquid system are used
as mobile and stationary phase. This technique is more popularly known as Counter-Current-
Chromatography (CCC) or also referred to as Centrifugal Partition Chromatography (CPC). It
was invented by Yoichiro Ito in 1966, trying to separate lymphocytes in a continuous extraction
process [3]. In Liquid- liquid chromatography one of the two liquid phases is kept stationary by
centrifugal force in a special designed rotating column while the other phase is pumped through.
This concept is similar to liquid-solid chromatography. However, the separation mechanism is
the partitioning of a compound between the two liquid phases and hence identical to the
separation mechanism of liquid-liquid extraction. In liquid-liquid chromatographic tasks
partition coefficients around one are favorable, while in liquid-liquid extraction partition
coefficients higher than one are desired [4]. In contrast to liquid-liquid extraction in liquid-
liquid chromatography the separated compounds leave the column again dissolved in the mobile
phase.
The main application of liquid-liquid chromatography at the moment is the purification of
natural products at lab scale. The natural products are among others plant extracts, flavonoids
and alkaloids [5,6].
At the beginning of the development of this technology only small-scale apparatus were
available. However, starting 1970 a continuous improvement of the moving parts creating the
centrifugal field has held on. Until the 1980s the main drawback of this technology has been
the long separation time which was between several hours and days. In the 1980s the invention
of the “J” type centrifuge marked the application of this technology. It had a far bigger
stationary phase retention at higher flow rates due to the new column design leading to a shorter
Introduction
2
hydrostatic and hydrodynamic machines took place. Recently both types of machines emerged
industrial scale. Even a maxi scale centrifuge with a volume of 20 liters (hydrodynamic
machine) is now commercially available [7-10].
The unique feature of liquid-liquid chromatography (CCC/CPC) is that the user prepares not
only the mobile phase, as in liquid-solid chromatography, but also the stationary phase. This is
done by simply mixing certain portions of preselected solvents that form a biphasic liquid
system. The compositions of the phases of the biphasic liquid system in equilibrium correspond
to the composition of the mobile and the stationary phase. The choice of an almost unlimited
number of biphasic liquid systems makes this technique versatile and allows for a tailor-made
system. However, this is also the reason why it is challenging and time consuming to find the
best suited solvent system for a particular separation task. Currently, only few solvents are
considered for liquid-liquid chromatographic separations. The solvents mostly used are n-
hexane, ethyl acetate, methanol and water. In addition to that also n-butanol, chloroform,
acetonitrile, ethanol, methyl-tert-butyl ether and n-heptane are applied in smaller extend
[11,12]. Presently, the partition coefficient is applied as the screening parameter for the
selection of a biphasic solvent system for a particular separation task. The goal is to find solvent
systems in which the target compound has a partition coefficient between 0.4 and 2.5, while a
partition coefficient of 1 marks the optimum [4,13,14]. The selection of the biphasic solvent
systems in liquid-liquid chromatography is mostly done using previously published literature
data of similar separations or by trial and error experimental screening procedures of multi-
solvent systems. These multi-solvent systems contain different portions of the same solvents
and are organized in tables according to the overall polarity of the systems. Therefore, they are
called solvent system families [13-15]. The most frequent employed solvent system families
are: hexane/ethyl acetate/butanol/methanol/water (“HemWat”-family) [16],
heptane/acetonitrile/butanol/water [6,8], chloroform/ methanol/water (“ChMWat”-family) [17]
and heptane/ethyl acetate/methanol/water (“ARIZONA”-family) [5]. Normally the solvent
selection procedure includes the experimental determination of the partition coefficient of the
target component/s and the related impurities in preselected biphasic solvent systems. The
concentration of the compounds in the phases in equilibrium of the solute is commonly
determined with UV-Vis spectroscope, GC (gas chromatography) or HPLC (high performance
liquid chromatography) analysis. These steps take a tremendous amount of work and also have
the disadvantage of not exploring all possibilities of this technology (i.e. tailor-made systems
through the unlimited number of solvents). However, combining a robotic liquid handling
Introduction
3
system and an analytical system (HPLC or GC) the trial and error experimental procedure can
be fully automated and applied to perform high-throughput screening [18].
At the moment, the most systematic solvent selection method is the “best solvent” approach,
proposed by Foucault [13]. The first step in this approach is to find a solvent in which the
sample is highly soluble. After that, two other mutually poorly miscible solvents are selected.
The “best solvent” should be soluble in these two solvents and a two-phase system must be
formed. It is expected that the “best solvent” and the sample would partition between the two
liquid phases in a similar way. Hence, the tie lines of the ternary diagram of the selected ternary
solvent system can be used to anticipate the solute partitioning. This approach, even though
systematic, uses only the sample solubility as a relevant parameter for the selection of the
biphasic liquid system and requires experimental work for the determination of the partition
coefficient [13].
A more advanced option is to calculate the solute partition coefficient using well established
models for the description of the thermodynamic equilibrium of multicomponent liquid–liquid
systems. Using this approach, the experimental effort and cost needed to select a biphasic liquid
system and to tailor its composition can be significantly reduced, if not completely eliminated.
In previous works [19-22], the applicability of the fully predictive Conductor like Screening
MOdel for Real Solvents (COSMO-RS) implemented in the program COSMOtherm as a tool
for screening and selection of biphasic liquid systems in various applications has been shown.
Even the predictive LLE determination is possible [21-25]. COSMO-RS was firstly applied in
chemical engineering by Maassen et al. [26]. Further applications have been shown in [27-32].
The model is frequently used as screening tool [31] and has been proven to be capable in the
prediction of octanol/water partition coefficients of complex mixtures [33,34]. The only
information needed for the prediction of the partition coefficient is the molecular structure of
the solutes and solvents as well as the composition of the phases of the biphasic solvent system.
On basis of the idea of the “best solvent” approach with the objective of minimizing the
experimental work for the selection of the biphasic system for a separation task at the Institute
of Separation Science & Technology in Erlangen (TVT) an attempt for a systematic method for
the solvent selection using thermodynamic models was designed [19,20]. It follows the well-
known and commonly used approaches for the selection of the mobile and stationary phase in
liquid–solid chromatography, which were adapted to meet the particularities of the liquid–
liquid chromatography related to the liquid nature of the phases.
Following the published works from the TVT another research group took up the idea on the
simplification of the solvent selection process in liquid-liquid chromatography. In their methods
Introduction
4
for the selection of solvent systems the COSMO-SAC model is used instead of COSMO-RS
[35].
Obviously, the efficiency of a separation is not only determined by thermodynamics (partition
coefficients), but also by the hydrodynamic behavior of the solvent systems. These are related
to the column design, operating parameters and physical properties of the phases, as for example
viscosity, interfacial tension and density. There are already some considerations made in this
field [36-41] which can be included in the solvent selection pathway. At the moment, the
physical properties of the phases are not determined very frequently. The only property which
is sometimes measured is the density difference which should be preferentially higher than
0.1 g/cm3 [42]. Usually the chromatographers use a simple short cut method to judge about the
suitability of a biphasic solvent systems in terms of the physical properties of the solvent
system. This short cut method is the determination of the settling time. The settling time is the
time the phases of the biphasic solvent system in equilibrium need to split after a vigorous
shake. If the settling time is less than 60: a system is considered suitable [42].
For separations of biomolecules, as proteins and nucleic acids, the classical solvent systems
cannot be used. In such separations the so-called Aqueous Two-Phase Systems (ATPS), which
have a high water content in each phase leading to mild conditions are applied. The separation
using ATPS though is challenging since the stationary phase retention within the column is
rather low. In addition to that a constant loss also referred to as bleeding of stationary phase has
been observed. The reason for these problems lies in the high content of water in both phases
which leads to low density differences and interfacial tensions [43-46]. Several approaches to
improve the designs (new forms of geometry) are done at the moment in order to achieve a
better performance with ATPS [47-49]. Besides the improvements in the design of the columns
it is also tried to tune the physical properties of the phases applying alternative ATPS [50].
Recenty the applicability of ATPS based on ionic liquids was studied [51].
Introduction
5
Goals and structure of the thesis at hand
The goal of this thesis is to develop, enhance and experimentally validate a solvent selection
pathway for liquid-liquid chromatographic separations. In the long run this pathway should
replace trial and error methods for the solvent selection in liquid-liquid chromatography and
therefore reduce the number of experiments needed to a minimum. In addition, an attempt was
done to apply a new class of solvents, i.e. aqueous two-phase systems based on ionic liquids.
Hopmann showed in [19] and [20] the possibility to apply thermodynamics for the prediction
of physical properties. Hopmann showed how the application of thermodynamics can
significantly reduce the experimental effort involved in the solvent selection procedure for a
liquid-liquid chromatographic separation. In particular she predicted and screened partition
coefficients with COSMO-RS in biphasic solvent systems.
Based on these results, we demonstrated how COSMO-RS can be used in a consecutive solvent
selection pathway for liquid-liquid chromatographic separation tasks in a previous cooperative
work [52].
The thesis at hand builds on this previous research by extending and refining the pathway
presented in [52]. A modified systematic pathway for the solvent selection is developed, taking
also the possibility of calculating the solubility solely on thermodynamics into account.
The generation of LLE data applying predictive thermodynamic models is demonstrated using
COSMO-RS and UNIFAC as representative models.
Further also the option to apply the predictive models COSMO-RS and UNIFAC for the
prediction of partition coefficients using different sources of LLE data is highlighted and
evaluated in detail for the selection of a suitable biphasic solvent system. The idea is to simplify
the solvent selection process drastically. The most challenging and time-consuming
experiments such as the accurate measurements of LLE shall therefore be replaced by fast
thermodynamic predictions.
In order to fully design an optimum solvent system also ways to estimate the stationary phase
retention for each mode (i.e. ascending or descending) with quick measurements of physical
properties, are demonstrated. These measurements can replace tedious test measurements.
Following this pathway which consists of a sequence of several decision-making steps, the
whole potential of this technology, i.e. the unlimited number of possible solvent systems can
be explored.
The first part of section 2 deals with the principles and set-ups of liquid-liquid chromatography.
In this part, also the main challenges of designing a successful separation process in liquid-
liquid chromatography are named.
6
In part 2 of section 2 the thermodynamic basis of the applied methods is explained. The
outrageous importance of the activity coefficient for the solvent selection pathway developed
in this work is carved out.
In section 3 the chemicals used and the equipment applied is explained. In addition, the methods
applied for the determination of solubility and stationary phase retention are shown. Further the
measurement of the enthalpy of fusion, melting temperature, and partition coefficient is
described. The theoretical methods to determine activity coefficients as well as LLE and the
prediction of the solubility is explained in detail.
Section 4 is the core of the thesis. It combines the single steps namely, potential solvent
collection, solubility estimation, liquid-liquid equilibria determination and prediction of
partition coefficients, to a complete solvent selection procedure which is applied for model
separations.
The last section 5 deals with potential ATPS based on ionic liquids adopted from liquid-liquid
extraction. It carves out the problems and challenges of this solvent class and ATPS in
combination with liquid-liquid chromatography.
This chapter comprises the concepts and fundamentals of liquid-liquid chromatography
followed by the thermodynamic principles and models needed for the solvent selection
pathway.
Liquid-liquid chromatography The principle of liquid-liquid chromatography (LLC) is demonstrated in Figure 2.1 The
chromatographic column is represented as a series of ideally mixed cells, (reservoirs) similar to
the representation of a plug flow tubular reactor in the field of reaction engineering [53].
Theoretical basis
Figure 2.1: Illustration of liquid-liquid chromatography [3,7,9,53]
In Figure 2.1 the separation of a binary mixture represented by the letters A and B. The partition
coefficient GH}~ is defined as the ratio of the molar concentration of compound 4 in the lower
phase to the concentration in the upper phase (definition in Eq.2.2 and Eq.2.18). The partition
coefficients of compounds A and B are qualitatively expressed as G}~ > GÅ }~. The separation
of both compounds A and B is shown as a sequence of four consecutive steps. The stationary
phase is presented in dark grey while the mobile phase (upper phase) is lighter colored. At the
beginning of the separation in step 1 each of these cells is filled with defined amounts of
stationary (lower) phase lower phase. The feed mixture dissolved in the mobile phase (upper
phase) is given into the first vessel.
Then the vessel is vigorously mixed and equilibrated. In vessel 1, the compounds # and |
distribute between the two liquid phases according to their partition coefficients, i.e. most of
# goes in the lower phase while the main amount of | stays in the upper phase. After that in
step 2, the mobile phase (upper phase) and its dissolved components of each vessel are
transferred to the next vessel while fresh mobile phase is introduced to the first vessel. Then the
vessels are shaken vigorously and equilibrated again. In each vessel which contains # and |, #
and | will distribute according to their partition coefficient (most of # in the stationary phase).
Theoretical basis
8
This procedure is repeated as presented in step 3 until the compounds, first | than # leave the
last vessel as pure presented in step 4 [53].
In LLC, the concept of separation presented in Figure 2.1 is realized as a continuous process by
keeping one of the two phases stationary by means of centrifugal force while the other phase is
pumped through. This is similar to liquid-solid chromatography in which a solid phase is used
as stationary. However, there are some significant differences between these two technologies.
Due to the liquid state of the stationary phase the mobile and the stationary phase cannot be
selected independently. The mobile and the stationary phase consist of a biphasic system
obtained by mixing of two or more solvents. The separation mechanism, i.e. different
partitioning of the feed mixture components between two liquid phases is identical to the
separation mechanism of liquid-liquid extraction. In contrast to liquid-liquid extraction there is
no counter current movement and the compounds enter and leave the column dissolved in the
same phase i.e. the mobile phase. In Table 2.1 the three separation techniques, liquid-solid
chromatography, liquid-liquid extraction and liquid-liquid chromatography are compared. For
this comparison, typical modes of operation for the three technologies are considered.
Theoretical basis
Table 2.1: Comparison of the key features of liquid-solid chromatography, liquid-liquid extraction and liquid-liquid chromatography
Unit operation Batch liquid-solid chromatography
(HPLC)
First auxiliary phase
force Second auxiliary
Target compound/ s
Dissolved in second
Target compound/ s
Dissolved in first auxiliary
Separation mechanism Mostly adsorption
Partition coefficient
range of the adsorption equilibrium (isotherm)
No specific range
equilibrium, but also possible with higher
concentrations
* --state 2015
In Figure 2.2 the basic LLC setup is illustrated. It is constructed like a typical chromatographic
set-up with the modification that instead of a column packed with solid stationary phase a
specially designed liquid-liquid chromatography column (see Sections: 2.1.1/2.1.2) is used. In
this column, also referred to as machine or apparatus one of the liquid phases (light) is held
stationary by means of centrifugal forces while the mobile phase (dark) is pumped through.
Theoretical basis
Figure 2.2: Setup of the liquid-liquid chromatographic separation unit
Before the separation run, the biphasic solvent system i.e. the mobile and stationary phase are
prepared by mixing pre-defined solvent portions. The phases are equilibrated at a constant
temperature (identical to the one used during the separation) and separated afterwards into two
reservoirs. At first the column is filled with the phase which is designated to be the stationary
phase. After that, the column is brought into rotation, which causes the centrifugal force. As
soon as the preset rotational speed is reached, the mobile phase flow into the column is started
and starts replacing a part of the stationary phase from the column. This occurs until the
hydrodynamic equilibrium is reached. After this moment, no more stationary phase leaves the
column, but only the mobile phase. As soon as only the mobile phase leaves the column, the
mixture to be separated (dissolved in the mobile, stationary or both phases) can be injected. The
volume of the columns ranges between 20 ml and recently 20 l [9,45]. There are two basic
designs of LLC columns: the hydrodynamic and the hydrostatic column. The column itself
together with the periphery which consists of the auxiliary parts for the rotation are often
referred to as machine or apparatus.
2.1.1 Hydrodynamic apparatus
In Figure 2.3 the principle of operation of a hydrodynamic column, also known as counter
current chromatographic column (CCC) is shown. Figure 2.3 shows the most common
hydrodynamic apparatus the so-called J type centrifuge.
Theoretical basis
11
Figure 2.3: Design of a hydrodynamic column, reprinted from [12] by permission from Elsevier
A hydrodynamic column is a flexible tube wounded around a drum (bobbin). The drum rotates
about its own axis and revolves around the axis of the centrifuge at the same angular velocity
p in the same direction. These two different rotations cause a centrifugal force with a variable
direction and magnitude along the column creating mixing and settling zones through the whole
length of the column. The mixing zones are where the centrifugal force is the lowest while the
settling zones can be found where the centrifugal force is the highest. The mixing and de-mixing
zones along the column are presented in Figure 2.4.
Theoretical basis
12
a)
b)
Figure 2.4: Mixing and de-mixing zones within a hydrodynamic LLC column, reprinted from [12] by permission from Elsevier
In Figure 2.4 a) a view over various column positions during one column rotation around the
axis of the centrifuge is given. The mixing zones are close to the center of the axis of the
centrifuge (illustrated in black color) and the settling zones (presented in white).
In Figure 2.4 b) the distribution of the mixing and de-mixing zones along the column (presented
as unwounded tube) is illustrated. It can be seen from position I to IV that a mixing zone travels
through the stretched column [12].
2.1.2 Hydrostatic apparatus
In Figure 2.5 the design of a hydrostatic liquid-liquid apparatus which is also called Centrifugal
Partition Chromatograph (CPC) is shown. In a hydrostatic column, several identical disks are
mounted one above the other. Each disk has a series of cells interconnected by ducts, see Figure
2.5 a).
Between two adjacent discs a polytetrafluorethylene plate is placed. The volume of a column
can be tuned changing the number of disks. The column is mounted on the axis of a rotor of a
centrifuge. The mobile phase flows through the rotary seals as presented in Figure 2.5 b).
Theoretical basis
13
a)
b)
Figure 2.5: a) Disk containing cells connected by ducts b) hydrostatic column consisting of stacked discs [54]
In hydrostatic columns, the mixing and de-mixing zones occur within the cells while in the
ducts only mobile phase is flowing. The mixing and demixing of the phases in the cells is a
result of the centrifugal force and the flow of the mobile phase. The mixing zones are at the
beginning of the cells where the mobile phase enters it and the demixing zones are close to the
outlet of the cell. Detailed visualization of these hydrostatic flow regimes have been analyzed
using transparent discs and camera recording [40,55,56].
2.1.3 Features of liquid-liquid chromatography
The most important feature of liquid-liquid chromatography is that the stationary phase is liquid
and the whole volume of the stationary phase is hence accessible to the solutes. Moreover, each
of the phases, i.e. upper or lower phase of the biphasic liquid system can be used as stationary
Theoretical basis
14
phase. The mode of operation is called descending mode, also referred to as head to tail mode
or reversed phase mode if the upper phase is used as stationary phase. The mode is called
ascending mode, also referred to as tail-to-head mode or normal phase mode if the lower phase
is used as stationary phase. The volume of the phases in the column is represented by a
parameter the so called stationary phase retention which is the fraction of the stationary phase
of the biphasic liquid system held back within the column:
!" = n} nÇ
(2.1)
where n} stands for the volume of the stationary phase in the column in hydrodynamic
equilibrium and nÇ the volume of the column. The volume of the stationary phase depends on
the characteristic geometrical parameters of the column, the operation parameters (the mobile
phase flow rate and rotational speed) and the physical properties of the biphasic solvent system
[41]. The distribution of a compound between the two phases at equilibrium can be expressed
by the partition coefficient. The partition coefficient GH of a compound 4 is the ratio of its
concentration in the stationary phase ?H} and the mobile phase ?H~, in thermodynamic
equilibrium:
GH = ?H }
?H ~ (2.2)
The partition coefficient is a function of the concentration of the solute in the system.
In a low concentration range the partition coefficient is constant and similar to the Henry
constant in liquid-solid chromatography, which describes the linear range of the partition
equilibrium [57].
In the linear range of the partition equilibrium, the retention volume nÉH of compound 4 is
definded as:
nÉH = n~ + GH }~n} (2.3)
where n~ is the volume of the mobile phase, n} is the volume occupied by stationary phase and
GH is the partition coefficient of the compound 4. The volume of the stationary phase has a direct
influence on the retention volume nÉH and hence, on the retention time of a compound as well
as on the separation performance.
The separation factor of a binary mixture consisting of compound 4 and Ö is defined as the ratio
of their partition coefficients of GH and GÜ in a particular biphasic system:
Theoretical basis
bHÜ = GH }~
GÜ }~ (2.4)
In this work in the calculation of bHÜ the bigger partition coefficient is in the numerator while
the lower partition coefficient is placed in the denominator. Further due to long names of
compounds, only the capital letter is used as a subscript of bHÜ.
The resolution {} is defined in the same way as in liquid-solid chromatography:
{} = nÉH − nÉÜ àH + àÜ
2
(2.5)
where nÉH and nÉÜ are the retention volumes of compounds 4 and Ö, while àH and àÜ are the
peak widths at the base line. In separation technology, to describe the separation the separation
unit is divided in consecutive theoretical plates. A theoretical plate is also referred to as ideal
plate and describes a region in which full phase equilibrium is reached. For symmetrical elution
peaks the Gaussian distribution can be assumed and hence the number of theoretical plates 8H,
i.e. the efficiency of the column for compound 4 can be calculated:
8H = äãK åK
çK
% (2.6)
Combining equation Eq.2.3, Eq.2.4 and Eq.2.5 together with the assumption of an equal number
of theoretical plates for compound 4 and Ö, i.e. 8H = 8Ü = 8, the resolution can be expressed by
equation Eq.2.7.
{} = 1 4
8 (GH
2
(2.7)
where 8 is the average number of theoretical plates of compound 4 and Ö, n~ the volume of the
mobile phase, n} the volume of the stationary phase, GH and GÜ are the distribution constants
(partition coefficients) of compound 4 and Ö, respectively.
Three factors have an influence on the separation resolution in liquid-liquid chromatography:
the number of theoretical plates, the stationary phase retention and the partition coefficients of
the compounds to be separated. The number of theoretical plates in conventional hydrodynamic
and hydrostatic columns is with only several hundred to a few thousand significantly lower than
in liquid-solid chromatography. Therefore, the impact of this parameter on the resolution ({})
is not as significant as it is in liquid-solid chromatography. On the other hand, the ratio of the
Theoretical basis
16
volume of the mobile and stationary phase in the column which is directly related to the
stationary phase retention (see Eq.2.1), has a significant influence on the resolution. The lower äë äí
(i.e. the higher the stationary phase retention) the better is the resolution. In general, the
stationary phase retention lies within the range of 60 % and 80 %. It is also possible to operate
with lower stationary phase retentions, but a stationary phase retention which is less than 30%
normally leads to a poor peak resolution [42].
In Figure 2.6 the resolution {ì of a binary separation with a separation factor b of 1.4 assuming
1000 theoretical plates is plotted versus the partition coefficient GH}~ of the first eluting
compound 4 for different ratios of äë äí
, i.e. between 1 and 0.25. A ratio äë äí
= 1 corresponds to
= 0.25 corresponds to a stationary phase
retention !" = 0.8. The preferred “optimum working” region (0.4 < GH}~ < 2.5) is identified
by two vertical lines and localized by two arrows.
Figure 2.6: Resolution of a binary mixture as a function of the partition coefficient of the first eluting compound GH
}~ for different of äë äí
rations. (8 = 1000, b = 1.4. The arrows and vertical lines mark the preferred partition coefficient region [54]
In Figure 2.7 it can be seen that the difference of the partition coefficients, i.e. the separation
factor b of the compounds which are separated has a significant influence on the resolution.
The higher b is the bigger is the resolution and vice versa. However independent of b the
resolution is low (or rather drops significantly) if GH < 0.4 while it does not change significantly
if GH > 2.5.
17
Figure 2.7: Resolution of a binary mixture as a function of the partition coefficient of the first eluting compound 4 for different separation factors b. (8 = 1000,
äë äí = 0.66 (!" = 0.6)). The
arrows define the preferred region of the partition coefficient [54]
In liquid–liquid chromatography, the partition coefficient of the target component/s is the
parameter used for the selection of the mobile and stationary phase, i.e. the biphasic solvent
system. This parameter is not only important in terms of the achievable resolution {}, but also
determines the duration of the separation run. This can be seen in Figure 2.8 in which a
chromatogram of a pulse injection of a mixture containing solutes with different partition
coefficients is presented. For the sake of clarity, the region of interested is detached with dashed
lines and the peaks colored in dark.
Figure 2.8: Chromatogram of a pulse injection liquid-liquid separation [58] reprinted by permission of Taylor & Francis LLC
Analyzing the peak shape in the chromatogram having in mind the presented Eq.2.3 and Eq.2.5
it can be understood that for partition coefficients GH < 0.4 the peak resolution is poor. On the
Theoretical basis
18
other hand, partition coefficients GH > 2.5 lead to broad peaks and therefore highly diluted
products. In addition to that the separation will “take long” which leads to a high solvent (i.e.
mobile phase) consumption. Therefore, the region of a partition coefficient 0.4 < GH < 2.5 is
recommended and called “optimum working region”. This region is advantageous since the
result is the same no matter which phase is used as mobile and which as stationary phase due
to the definition of the partition coefficient. If the separation is done in ascending mode, the
upper phase is mobile while in descending mode the lower phase is mobile. Hence the partition
coefficient of one mode is the reciprocal of the other mode. However, this defined optimal
working region of the partition coefficient should not be regarded as a strict range, since the
achievable resolution as discussed (Eq.2.7) is not only a function of the absolute values of the
partition coefficients, but also depends on the number of theoretical plates and the stationary
phase retention.
2.1.4 Solvent selection and typical solvents in liquid-liquid chromatography
The limitless choice of solvents theoretically gives the user the possibility to actually tailor the
biphasic solvent system for a specific separation problem. At the moment, the selection of the
biphasic solvent system is almost exclusively done using the partition coefficient as a screening
parameter, while costs and environmental issues are not really considered. This is due to the
fact that the technology is mostly used for the isolation and identification of active compounds
from plants without the objective of a direct transfer to production scale [6]. Therefore the
selection of the biphasic solvent systems in liquid-liquid chromatography is mostly done using
previously published literature data for similar separation problems or by an experimental
screening of predefined limited number biphasic multi-solvent systems and system
compositions organized in tables according to the overall polarity of the system, so called
solvent system families [13-15]. As it is shown in [11] 56 % of the solvent systems used in LLC
consist of 4 solvents, only 37 % consist of 3 solvents, 5 % of the solvent systems consists of 2
solvents while solvent systems with 5 solvents are rare with about 2 %. The most frequently
used biphasic CCC solvent system families are: hexane/ethyl acetate/butanol/methanol/water
[16], heptane/acetonitrile/ butanol/water [13,16], chloroform/methanol/water [17] and
heptane/ethyl acetate/ methanol/water (“ARIZONA” family) [15]. As an example, for a solvent
system family, the compositions of the systems from the ARIZONA solvent system families
are listed in a table and presented in a 3-dimensional (pyramidal) plot in Figure 2.9.
Theoretical basis
System Heptane Ethyl
acetate Methanol Water 6 A 0 1 0 1 7 B 1 19 1 19 8 C 1 9 1 9 9 D 1 6 1 6 10 F 1 5 1 5 11 G 1 4 1 4 12 H 1 3 1 3 13 J 2 5 2 5 14 K 1 2 1 2 15 L 2 3 2 3 16 M 5 6 5 6 17 N 1 1 1 1 18 P 6 5 6 5 19 Q 3 2 3 2 20 R 2 1 2 1 21 S 5 2 5 2 22 T 3 1 3 1 23 U 4 1 4 1 24 V 5 1 5 1 25 W 6 1 6 1 26 X 9 1 9 1 27 Y 19 1 19 1 28 Z 1 0 1 0
Figure 2.9: Heptane/ethyl acetate/methanol/water (“ARIZONA”) system global composition (v/v/v/v) [13,59]
Ito proposes two different strategies to select an appropriate solvent system in [12]. The first
attempt must always be a thoroughly literature search for similar components. If such
components cannot be found a trial and error method is recommended. To find a suitable solvent
system Ito proposes to organize a solvent system family (as e.g. ARIZONA) according to a
decreasing order of the hydrophobicity of their organic phase and to start the trial and error
Theoretical basis
20
procedure with a system providing a medium hydrophobicity. In case of the ARIZONA solvent
system family this would be system 17 or also called system N from Figure 2.9. Based on the
experimental determined partition coefficient of the target and impurities in this system the user
can decide if a more (system with higher number) or less hydrophobic (system with smaller
number) system is needed in order to achieve the preferred partition coefficient. All of the
solvent systems used in liquid-liquid chromatography at the moment take up this idea and are
therefore organized in tables with solvent system global compositions sorted according to their
polarity.
For the selection of a biphasic solvent system for separations of natural products, Friesen and
Pauli [58] proposed a mixture of over 20 commercially available compounds covering a large
range of different polarities, functional groups, and structures. This mixture is called
“GUESSmix” (Generally Useful Estimation of Solvent Systems). In this mixture, the
chromatographer has to identify compounds similar to the target compound. A new potential
solvent system is evaluated determining the partition coefficients of this mixture via thin layer
chromatography [58] or as recently proposed using polarity and selectivity assessed by
2-dimensional reciprocal shifted symmetry (ReSS2) plots which basically compare the
selectivity and the polarity of solvent systems [60]. According to Friesen and Pauli in the most
cases the so found systems or systems nearby will be suitable for the desired target compound.
The “best solvent” approach, proposed by Foucault [13,61], is the first systematic pathway for
finding an appropriate biphasic solvent system. The first step in this approach is to find a solvent
in which the sample is highly soluble. After that, two other preferably poorly miscible solvents
are selected. If the miscibility is unknown, it is recommended to use two solvents having
different polarity defined using the Reichardt’s index. The Reichardt’s index is a relative
measure of polarity normalized to the value of water which is set as the most polar solvent. It
is assumed that a rather high polar solvent and a rather low polar solvent will lead to a biphasic
liquid system [13]. The “best solvent” should be soluble in these two solvents and a two-phase
system should be formed. It is expected that the “best solvent” and the sample would partition
between the two liquid phases in a similar way. Therefore, the tie lines of the ternary diagram
of the selected ternary solvent system can be used to anticipate the solute partitioning. This
approach, even though systematic, uses only the sample solubility and solvents relative polarity
as parameters for the selection of the biphasic liquid system. After the selection of the system
the partition coefficients are experimentally determined and the composition of the system is
further tuned if required.
21
Taking up the idea of the “best solvent” approach the Chair of Separation Science and
Technology of the Friedrich-Alexander-University Erlangen-Nuremberg enhanced the solvent
selection process in liquid-liquid chromatography using predictive thermodynamic models
[20,52,54]. The predictive thermodynamic models used are COSMO-RS and UNIFAC. While
predictive thermodynamic models are not depending on experimental data, non-predictive
(correlative) models need experimental determined data. However, UNIFAC in contrast to
COSMO-RS is a binary parameter based method so the interaction parameters between two
groups derived from experiments need to be available. Based on the idea of implementing
thermodynamic models in the solvent selection procedure there are several research groups
working on solvent selection procedures, using either predictive [23] or non-predictive, i.e.
correlative thermodynamic models [35,62,63].
All solvent selection procedures rarely consider the physical properties of the solvent system.
Instead of measuring them the settling time is taken as a fast measure of the hydrodynamic
behavior of a biphasic solvent system. According to Ito a settling time less than 20s will lead
to stationary phase retentions of at least 50% [12]. However as rough estimation parameter the
settling time of a solvent system should not exceed the time of 60s to be suitable for a liquid-
liquid separation task. Hence a short settling time should lead to a high stationary phase
retention and vice versa [12,42,64].
It is not possible to use organic solvents for the separation of proteins with LLC because in the
environment provided by classical solvents the proteins will denature. For the purpose of the
separation of very sensitive compounds the so called Aqueous Two-Phase Systems (ATPS) are
used. An aqueous two-phase system is a biphasic system obtained by mixing an aqueous
solution of two polymers or a polymer and a salt. They are very gentle systems for sensitive
compounds because of their high content of water in both phases and their low interfacial
tension. The first application of ATPS was reported in 1956 in the field of extraction [65]. Since
the beginning of 1980 there are several studies and attempts to use ATPS in the field of liquid-
liquid chromatography. All applications though showed the same problem, namely the
emulsification of the phases which leads to a loss of stationary phase. This is due to the low
interfacial tension and the high viscosity of the phases. In such cases it was shown that the
hydrostatic type of liquid-liquid chromatographic machines is superior in terms of the stationary
phase retention [66]. However mostly the same test solutes and the same solvent systems are
used since presently the main focus lies in the improvement of the performance. The test system
consists of PEG 1000, water and K2HPO4, while the solutes are lysozyme, myoglobin and
Theoretical basis
22
cytochrome. Berthod also investigated PEG with higher molecular weight and ATPS in which
PEG is replaced by ionic liquids [67-69]. Currently two approaches are followed to overcome
the drawback of the low stationary phase retention of ATPS in a liquid-liquid chromatographic
apparatus. One is the attempt to tune the physical properties of the phases. Variing the molecular
weight of PEG from PEG 1000 to PEG 4000 it is tried to gain a faster phase separation and
hence a better stationary phase retention. For that matter at the TU Dortmund, Germany in the
group of Prof. Schembecker it is tried to improve the separation performance developing a
guideline for the selection of ATPS in CPC [50]. In the work at hand ATPS based on ionic
liquids are further investigated. An ionic liquid is a molten salt (melting point less than 100°C)
which has a very low vapor pressure and a high polarity. It is composed of an organic cation
and an inorganic or organic anion [70]. A combination of different anions and cations gives the
engineer an almost unlimited number of combination possibilities. Varying the cation and anion
can change many properties such as viscosity, density, solubility etc. and allows for a tailor
made solvent [71]. The number of proposed IL based solvent systems published in literature
such as [72-74] has increased tremendously in the last couple years. However, few were tested
for the separation of proteins such as [75-78]. Dreyer et al. introduced ATPS based on ionic
liquids in the field of liquid-liquid extraction and successfully used them to separate proteins
[79,80]. These systems might be promising alternative systems in liquid-liquid extraction for
successful separations of proteins which might be transferable to liquid-liquid chromatography.
The high amount of interest in ionic liquids lies also in their “greenness” [81] however this is
discussed controversial.
At TU Munich in the group of Prof. Minceva recently this idea was picked up and the
application of so-called non conventional aqueous two phase systems was further investigated.
Non-conventional ATPS as they are defined in [51] differ from the convential ATPS which are
based on PEG or on a polymer since they are either based on ionic liquids or contain ionic
liquids as modifiers. In [51] the researchers showed that it is possible to use different ILs in
small amounts in PEG 600/phosphate salt mixture/water sytems in order to alter the partition
coefficients of the proteins. Moreover for ATPS based on ionic liquids the proposal was made
to use instead of conventional pulse injection liquid-liquid chromatographic separations rather
capturing modes to concentrate the target proteins in the IL-rich phase. Afterwards the proteins
can be recovered using ultra- and/ or nanofiltration [51]. The second attempt is to improve the
geometry of the cells. This is done by the group of Prof. Schembecker, Dortmund which
designed different cell geometries. The separation characteristics are determined and the flow
pattern in these cells is visulaized with a camera [82,83]. Recently Armen Instrumens brought
Theoretical basis
23
a new cell design with bigger cell size to market which show significantly higher stationary
phase retentions [84].
As described previously, the biggest challenge in developing a liquid-liquid chromatographic
separation process is the selection of the appropriate biphasic solvent system for a particular
separation task. The solvent system needs to provide partition coefficients in a specific range,
acceptable separation factors, high stationary phase retention and in order to be productive a
high solubility for the feed compounds.
In the field of protein separation all chromatographers which try to work with ATPS in liquid-
liquid chromatography face the same problems. Namely, classical ATPS consist either of
polymer/polymer/water or polymer/salt/water. In these cases, the viscosity of at least one of the
resulting phases is very high. Due to the high amount of water in both phases the interfacial
tension and the density difference are very low at the same time. These physical properties
cause low stationary phase retention !" in commercially available columns, i.e. in the
hydrodynamic and hydrostatic machines, which leads to a loss of the resolution [66].
Furthermore, a constant loss of stationary phase called “bleeding” was observed [43,44]. There
are two possible solutions to overcome this problem. One possibility would be to adapt the
geometry of the apparatus in order to achieve the desired performance of the unit. There is a
new generation of CPC disks which have bigger cells and thus a lower number of cells for the
same column volume. A hydrostatic apparatus containing such columns is called centrifugal
partition extractor (CPE). Such CPEs offer a higher stationary phase retention !" for ATPS
compared to the centrifugal partition chromatographs, but at the same time a lower resolution
due to the lower number of cells [84]. They can be applied for extractive purposes and target
component purification, i.e. every case in which the separation factor is high. Another
possibility which is also investigated in the work at hand is to actually evaluate new solvent
systems which might have similar physical properties as organic solvents. Therefore similar to
[67,85], the applicability of ionic liquid based ATPS based on their physical properties has been
evaluated. The high potential in the usage of ionic liquids lies in the fact that ionic liquids can
be designed combining different cations and anions and allow for a tailor-made solvent. ATPS
on basis of ionic liquids have already been introduced to the field of extraction [79,86]. In 2004
Berthod studied the suitability of room temperature ionic liquids for liquid-liquid
chromatography. Even though the ionic liquid based ATPS were better retained than the
Theoretical basis
24
classical PEG based ATPS, the stationary phase retention in the hydrostatic column was rather
low [85].
In this section, the description of the liquid-liquid and solid-liquid equilibrium by means of
thermodynamics is presented. For this matter, also the theory behind the two predictive models,
COSMO-RS and UNIFAC which are applied in this work is shown.
2.2.1 Thermodynamic equilibrium
The thermodynamic equilibrium of a system consisting of 5 phases and \ components is
achieved when thermal, mechanical and chemical equilibria are reached. The equilibrium
conditions are shown in the three equations below:
öõ = öõõ = = öù
thermal equilibrium (2.8)
Gõ = Gõõ = = Gù
mechanical equilibrium (2.9)
aH õ = aH
õõ = = aH ù chemical equilibrium (i=1…k) (2.10)
where aH is the chemical potential of compound 4. The chemical potential of a compound 4 in a
mixture can be expressed as the partial molar quantity of the Gibbs energy û.
aH = ü†
üùK °,¢,ù£§K , (i=1…k) (2.11)
where 5H is the number of moles of compound 4, at constant ö and G while 5Ü ≠ 5H.
Applying the concept of the fugacity ¶H of a compound 4, derived by G.N. Lewis in 1901 [57],
the phase equilibrium can be described with the so-called iso-fugacity criterion:
¶H õ = ¶H
õõ, (i=1…k) (2.12)
Further details about the derivations of fugacities can be found in [57,87].
2.2.2 Activity and activity coefficients
For the description of liquid phases in thermodynamics the activity is usually applied instead
of the concentration. The activity of compound 4, 'H at temperature ö, pressure G and the
Theoretical basis
25
composition uH is defined as the ratio of the fugacity in this state, to the fugacity of the pure
substance ¶Hß at pressure Gß at the same temperature ö:
'H(ö, G, uH) ≡ ¶H(ö, G, uH) ¶H ß(ö, Gß)
(2.13)
Using the activity 'H, the activity coefficient γ™ of the compound 4 is defined as:
dH = 'H uH
(2.14)
where uH is a handy measure of the concentration of compound 4 (mostly molar fraction). The
meaning of the activity coefficient dH can be interpreted from Eq.2.14. Figuratively spoken the
higher the activity coefficient of a compound 4 is, the higher are the repulsion forces and hence
the lower is the solubility/concentration of compound 4. Therefore dH can be taken as a
parameter that represents the interaction of a compound 4 with surrounding compounds [57,87].
2.2.3 Liquid-liquid equilibrium
Based on the Eq.2.12 and Eq.2.14 the liquid-liquid equilibrium of non-ideal mixtures can be
described using the concept of fugacity. However, the fugacity is not a directly measurable
quantity. Hence fugacity coefficients or activity coefficients need to be applied in order to
account for the real behavior of a mixture. As it is shown in [87] it is more convenient to
describe liquid phases applying the concept of activity coefficients dH rather than using the
present available equations of state. This is due to the fact that until now it is hardly possible to
describe the fugacity in the liquid phase over a large range of pressure and temperature with
equations of state especially in polar systems. Since polar solvents are very frequently used in
liquid-liquid chromatography this is an important matter of this work. Therefore, the fugacities
in the iso-fugacity criterion (see Eq.2.15) are expressed via activity coefficients as follows:
uHdH¶H ß ´ = uHdH¶H
ß ¨ (2.15)
where ¶Hßis the fugacity of pure compound 4. Assuming the same standard state of compound 4
for both phases it simplifies to:
(uHdH)´ = (uHdH)¨ (2.16)
Eq. 2.16 can also be expressed with the activity 'H of compound 4 in both phases:
'H ´ = 'H ¨ (2.17)
2.2.4 Partition coefficient
The partition coefficient G of a compound 4 between two liquid phases is defined as the ratio
of its concentration in the phases, at a constant pressure and constant temperature:
GH ´¨ =
?H ´
?H ¨ (2.18)
where ?H´ represents the concentration ?H of compound 4 in the phase b and ?H ¨ is the
concentration of compound 4 in the phase c. Applying the definition of the partition coefficient
in liquid-liquid chromatography (Eq 2.2) b stands for the stationary phase and β for the mobile
phase of the chromatographic system. The partition coefficient can also be expressed with the
concentration of the solute 4 given in mole fractions. In this case the partition coefficient is
expressed with the letter 1 instead of G:
1H ´¨ =
uH ´
uH ¨ (2.19)
Using Eq. 2.16 in Eq. 2.19, 1H ´¨ can also be expressed as a ratio of the activity coefficient dH of
the solute 4 in each phase.
1H ´¨ =
dH ¨
dH ´ (2.20)
Using the definition of the mole fraction uH and the molar concentration ?H Eq.2.18 can be
rearranged as follows:
≠´ (2.21)
where ≠´and ≠¨are the molar volumes of the phases [34]. The molar volume of the phases can
be calculated as the weighted sum of the molar volume of the solvents and the excess volume
of mixing ≠Æ.
≠Ø∞±≤ = ≠H≥∞±≤ + ≠Æ = uH≠ßH + ≠Æ (2.22)
The partition coefficient is not a constant value. Besides on temperature and pressure it also
depends on the concentration of the solute in the biphasic system. In this work, it is assumed
that the composition of the phases of the biphasic liquid system does not change after the
addition of the solutes. This condition is fulfilled in the linear range of the solute partition
equilibrium at low concentrations and can be described using the activity coefficients of the
Theoretical basis
27
solute infinitely diluted in the upper and lower phase. This means that the concentration of the
solute in both phases is approaching to zero (uH → 0) and consequently lim ∂K→ß
dH = dH ∑. Under
GH ´¨ = 1H
Similar to liquid-liquid equilibria, the iso-fugacity criterion is used as starting point for the
description of solid-liquid equilibria.
≤ (2.24)
where ¶Hì is the fugacity of compound 4 in the the solid phase : and ¶H≤ is the fugacity of
compound 4 in the liquid phase .. The fugacities of compound 4 in each phase can be expressed
as a function of the molar fraction uH, the activity coefficient dH and the standard state fugacity
of compound 4 in the corresponding phase ¶ßHì for the solid phase and ¶ßH≤ for the liquid phase.
The standard fugacities are a reference value and hence arbitrary:
uH ìdH
ì¶ßH ì = uH≤dH≤¶ßH≤ (2.25)
dH ì is the activity coefficient of compound 4 in the solid state. Assuming that solute 4 crystallizes
pure it follows:
¶H ì = ¶ßH
ì = uH ìdH
ì=1 (2.26)
Rearranging Eq.2.25 the molar solubility uH≤ of compound 4 in the liquid phase . can be
expressed as:
≤ (2.27)
The standard state and hence the standard fugacity ¶ßH≤ is defined as pure subcooled liquid at a
temperature ö which is lower than the triple point temperature öH∏Øin order to be independent of
the solvent:
The main challenge in the calculation of the molar solubility uH≤ with Eq.2.27 is the description
of the ratio π∫K ª
π∫K º at a temperature ö < ößH
∏Ø. However, this problem can be overcome relating the
fugacity to the overall change in the molar Gibbs free energy +ßH as shown in Eq.2.28:
Theoretical basis
+ßH = +ßH ≤ − +ßH
ì = {ö.5 ¶ßH ≤
¶ßH ì (2.28)
The change in the molar Gibbs energy +H is also related to the corresponding enthalpy ßH
and change in entropy :H of compound 4 at the system temperature ö:
+ßH = ßH − ö:ßH (2.29)
For the calculation of +ßH a hypothetical thermodynamic cycle which comprises the transition
from a pure solid phase to a supercooled liquid via a detour using the independence of state
functions of the path is used. More details of this matter can be found elsewhere [57,87]. The
solution for the molar liquid solubility uH≤ gives:
uH ≤ =
1
ößH ì≤ (2.30)
In Eq.2.30 the solubility depends on the activity coefficient of compound 4 in the solvent and
on the solvent independent quantities æßHì≤ , T and ößHì≤. Therefore, the capacity of a solvent for
a compound 4, VH can be used as a qualitative measure of the solute solubility in a given solvent.
The capacity VH is defined as a reciprocal value of the activity coefficient of a solute 4 infinitely
diluted in a solvent dH∑.
VH(:-.≠P5z) = 1
This relationship has already been successfully applied for solvent selection purposes in the
field of extraction, distillation and recently liquid-liquid chromatography [20,88,89] and is an
appropriate parameter if relative values are needed only.
2.2.6 Prediction of activity coefficients
Activity coefficients can be recalculated from suitable experiments or calculated using
thermodynamic models. In general, there are several types of models that can be used for that
purpose. One type of model is correlative. It correlates the activity coefficients for the system
of interest to experimentally determined data (liquid-liquid-equilibrium, vapor- liquid
equilibrium). An example for a correlative model is NRTL (Non-Random-Two-Liquid) which
is frequently used in process design due to its accuracy. Information about correlative models
can be found elsewhere [87,90]. In this work models belonging to the other group, i.e. structure
interpolating and predictive models are applied. The main difference to correlative models is
Theoretical basis
calculated without specific model parameters for the particular solvent system. Therefore, it can
always be applied without doing any new parameterization from experimental measurements
for the system of interest even if experimental data of the system of interest are not available at
all. In this work two models of this kind are evaluated: The Universal Quasichemical Functional
Group Activity Coefficients (UNIFAC) and the COnductor like Screening MOdel for Real
Solvents (COSMO-RS). In case of UNIFAC interpolating parameters, which have to be
available for the calculations while in case of COSMO-RS the calculations are done via
quantum chemical descriptions of the structures of the molecules involved.
It has to be mentioned at this point that the prediction of LLE data for the solvent systems is
always done on a solute free basis. It is assumed that the concentration of a solute present in the
solvent system is so low that it can be seen as infinitely diluted. Hence the predictions of
partition coefficients are done under the assumption that the solute does not change the
composition of the solvent system.
2.2.7 Conductor-like Screening Model for Real Solvents (COSMO-RS)
The Conductor-like Screening MOdel for Real Solvents (in literature also referred to as
Conductor-like Screening MOdel for Realistic Solvations), COSMO-RS has been developed
by Klamt [32] to predict thermo physical data of pure compounds and mixtures. The model
combines a quantum mechanical treatment of solutes and solvents with methods of statistical
thermodynamics for the description of molecular surface interaction. The methods of statistical
thermodynamics are applied instead of the time consuming and tedious molecular simulation.
In COSMO-RS the chemical potential of a compound can be predicted and therefore the
thermodynamic properties that can be derived from the chemical potential can be calculated.
For predictions in COSMO-RS the only information needed is the molecular structure of the
compounds of the system, i.e. all solvents and solutes. However up to now it is only possible to
predict interactions with molecules having a molecular weight below 2000 g/mol [91].
COSMO-RS belongs to the continuum solvation models. Continuum solvation models describe
electrostatic interactions of a molecule to its surrounding environment, i.e. the solvent or
solution. In the COSMO-RS model the molecule is treated as a molecule dissolved in an ideal
infinitely broad conductor having a dielectric constant ¡ = ∞. Around each molecule a cover
the cavity which determines the border between molecule and its surrounding environment
(continuum) is constructed. For this cover the radii of the atoms are used. For accurate
descriptions of the behavior of the molecules specially optimized radii have to be used for the
Theoretical basis
30
construction of the cavity. These radii exist for common atoms, but not for transition metals
such as Cobalt (Co). As a rough estimation for the atomic radius 1.17 times the van-der-Waals-
Radius can be taken [92,93].
The constructed cavity is divided into a number of segments (¬). Each segment ¬ has an area
'√ and the charge density h√. The chemical potential of a segment a}(h) is calculated
accounting for the electrostatic, the Van-der-Waals interactions as well as hydrogen bonding.
Since every segment has a distinctive charge density h√, the ensemble is characterized by the
frequency function of the charge density G(h). This distribution is called the h-profile. The h-
profile of a pure substance is directly calculated using the density functional theory [94]. A
combinatorial term aHJøƒ≈ is added which originates from different shapes and sizes of the
solute and solvent molecules. In the current COSMO-RS parameterization this term is
complemented with the Staverman-Guggenheim [95] term leading to the following expression:
aH Jøƒ≈ = −{ö g.5#} + 1 −
#H #} + .5
#H #}
(2.32)
where #} is the average solvent surface area and #H is the total surface area of compound 4, nH is
the molecular volume of compound 4, n} is the average molecular volume of the solvent and nß
is the partial volume of the surrounding solvent. ¬ is the coordination number.
The chemical potential aH is then calculated as the following sum:
aH = aH Jøƒ≈ + v h a} h qh (2.33)
In this work in COSMO-RS activity coefficients of compounds in mixtures are predicted.
Activity coefficients are needed for the calculation of the liquid-liquid equilibrium, the solid-
liquid equilibrium (i.e. solubility), the capacity and the partition coefficients. The activity
coefficient dH is calculated from the chemical potential aH (Eq. 2.33) according to the following
equation:
(2.34)
where (aH − aHß) is the difference in the chemical potential of a substance between the state
which is calculated aH and the COSMO standard state aHß while ö is the system temperature
and { is the universal gas constant.
As a short summary and illustration of COSMO-RS the most important calculation steps are
visualized in Figure 2.10:
Figure 2.10: h profile calculation for water using COSMO-RS
In Figure 2.10 the steps in the calculation of the h -profile with COSMO-RS are illustrated.
First the molecule is composed then its cavity is constructed. The cavity is basically a cover
over the molecule which is roughly 20% bigger than the van-der-Waal radii of the atoms. The
cavity is then divided in segments from which the h -profile is created via statistical methods.
The h -profile represents a frequency distribution of a certain charge density.
More information about the model COSMO-RS itself and its applicability can be found in
[32,92,93,95-99].
2.2.8 The UNIFAC model
Another predictive model for the determination of activity coefficients is the Universal
Quasichemical Functional Group Activity Coefficients (UNIFAC) model. UNIFAC is a group-
contribution method and deduced from the UNIQAC (UNIversal QuAsi Chemical) model. In a
group-contribution model a mixture of compounds is regarded as a mixture of functional groups
that interact with each other as illustrated in the next Figure 2.11 [100,101].
Figure 2.11: Division of molecules in functional groups and the visualization of the group- interaction parameter table
In the left part of Figure 2.11 water and ethanol are shown. Both molecules are dived in their
“UNIFAC” functional groups, which are not always identical to the chemical functional groups.
For the later calculations, all group binary interactions have to be taken into account. On the
right side the binary interaction of different functional groups is symbolized. It has to be pointed
Theoretical basis
32
out that the mixture is incompressible and the space of the molecules (light blue) is only inserted
for the sake of visibility. In the real liquid, the molecules touch each other. These interactions
are described with group specific parameters in the UNIFAC equations (2.35), (2.36) and
(2.37).
The activity coefficient is defined as a sum of a combinatorial term dHÇ and residual term