Content uploaded by Alain Berthod
Author content
All content in this area was uploaded by Alain Berthod on Oct 17, 2017
Content may be subject to copyright.
Journal of Chromatography A, 1037 (2004) 3–14
Review
Determination of liquid–liquid partition coefficients
by separation methods
A. Berthoda,∗, S. Carda-Brochb
aLaboratoire des Sciences Analytiques, CNRS, Université de Lyon 1, Bat CPE-308, 43 Boulevard du 11 November 1918,
69622 Villeurbanne Cedex, France
bArea de Qu´ımica Anal´ıtica, Universidad Jaime I, Cra. Borriol s/n, 12080 Castellón de la Plana, Spain
Abstract
By essence, all kinds of chromatographic methods use the partitioning of solutes between a stationary and a mobile phase to separate them.
Not surprisingly, separation methods are useful to determine accurately the liquid–liquid distribution constants, commonly called partition
coefficient. After briefly recalling the thermodynamics of the partitioning of solutes between two liquid phases, the review lists the different
methods of measurement in which chromatography is involved. The shake-flask method is described. The ease of the HPLC method is pointed
out with its drawback: the correlation is very sensitive to congeneric effect. Microemulsion electrokinetic capillary electrophoresis has become
a fast and reliable method commonly used in industry. Counter-current chromatography (CCC) is a liquid chromatography method that uses
a liquid stationary phase. Since the CCC solute retention volumes are only depending on their partition coefficients, it is the method of choice
for partition coefficient determination with any liquid system. It is shown that Ko/w, the octanol–water partition coefficients, are obtained by
CCC within the −1<logKo/w<4 range, without any correlation or standardization using octanol as the stationary phase. Examples of
applications of the knowledge of liquid–liquid partition coefficient in the vast world of solvent extraction and hydrophobicity estimation are
presented.
© 2004 Elsevier B.V. All rights reserved.
Keywords: Reviews; Partition coefficients; Octanol–water partition coefficients; Thermodynamic parameters; Distribution constants; Hydrophobicity;
Counter-current chromatography; Alkylbenzenes; Quinolines
Contents
1. Introduction .......................................................................................................... 4
2. Partitioning of solutes between two liquid phases........................................................................ 4
2.1. Nomenclature.................................................................................................... 4
2.1.1. Distribution ratio (D)....................................................................................... 4
2.1.2. Distribution constant or partition ratio (KD).................................................................. 5
2.2. Thermodynamics................................................................................................. 5
2.3. Effect of temperature ............................................................................................. 5
2.4. Effect of chemical reactions....................................................................................... 5
3. Measurement of partition coefficients................................................................................... 6
3.1. The shake-flask method........................................................................................... 6
3.1.1. Small-scale spectroscopic methods .......................................................................... 6
3.1.2. Separation methods for phase analysis ....................................................................... 6
3.1.3. Modern variations on the shake-flask method................................................................. 7
∗Corresponding author. Tel.: +33-472431434; fax: +33-472431078.
E-mail address: alain.berthod@univ-lyon1.fr (A. Berthod).
0021-9673/$ – see front matter © 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.chroma.2004.01.001
4A. Berthod, S. Carda-Broch/J. Chromatogr. A 1037 (2004) 3–14
3.2. Liquid chromatography and correlations............................................................................ 7
3.2.1. A simple and easy method.................................................................................. 7
3.2.2. Congeneric effect .......................................................................................... 7
3.2.3. Variations on the method to improve the correlation quality ................................................... 8
3.3. Capillary electrophoresis with ordered media ....................................................................... 8
3.4. The counter-current chromatography method ....................................................................... 9
3.4.1. Direct partition coefficient measurement ..................................................................... 9
3.4.1.1. Method description................................................................................. 9
3.4.1.2. Method range and accuracy......................................................................... 9
3.4.1.3. Method variation.................................................................................. 10
3.4.2. Using the liquid nature of the stationary phase .............................................................. 10
3.4.2.1. Dual-mode CCC.................................................................................. 10
3.4.2.2. Cocurrent CCC ................................................................................... 11
4. Application: homologues and partition coefficients...................................................................... 11
4.1. Alkylbenzene partition coefficient in the heptane–methanol–water system............................................ 12
4.2. Petroleum components and waterless biphasic liquid systems ....................................................... 12
4.3. Quinoline homologues in heptane–acetonitrile–methanol systems.................................................... 12
5. Conclusion.......................................................................................................... 13
Acknowledgements ..................................................................................................... 13
References ............................................................................................................. 13
1. Introduction
If a third substance is added to a system of two immis-
cible liquids in equilibrium, the added component will dis-
tribute itself between the two liquid phases until the ratio of
its concentrations in each phase attain a certain value: the
distribution constant or partition coefficient.
Studied as soon as the end of the 19th century [1], the dis-
tribution of solutes in biphasic liquid systems has become an
essential field of study. The octanol–water distribution ratio,
Ko/w, is the accepted physicochemical property measuring
the hydrophobicity of chemicals [2].Ko/wis the most widely
employed descriptor for quantitative structure–activity rela-
tionship (QSAR) studies [3,4].
Besides pharmaceutical and biochemical industry that
uses QSAR in drug design and toxicology, the measurement
of liquid–liquid partition coefficients is extremely impor-
tant in: (1) fundamental chemistry for studying inorganic
and/or organic complex equilibria; (2) industrial chemistry
for optimization of production and waste treatment; and (3)
food chemistry for purification and extraction of sugars, fat
or caffeine [5]. This field is so important that many univer-
sities offer a special course in liquid–liquid partitioning and
solvent extraction.
Combinatorial chemistry is able to produce large num-
bers of new compounds that could be potential drugs
[6,7]. A drug has to cross four barriers associated with
absorption, distribution, metabolism and excretion referred
as ADME interface [8]. One of the core properties re-
quired to assess ADME characteristics is hydrophobicity
that is Ko/w.IftheKo/wdetermination is not to be a bot-
tleneck in the combinatorial discovery process, it must
be at least as easy as it is to synthesize the compound
[9].
Partitioning of solutes between a stationary and a mobile
phase is the fundamental principle of all kinds of chromato-
graphic methods. Separation methods are the tools of choice
to determine liquid–liquid partition coefficients. This article
will describe the different ways offered by separation meth-
ods focusing on accuracy, ease and speed of the different
methods: classical methods, high-performance liquid chro-
matography (HPLC), capillary electrophoretic (CE) meth-
ods and counter-current chromatography (CCC).
2. Partitioning of solutes between two liquid phases
2.1. Nomenclature
The term liquid–liquid partition coefficient is “not
recommended” by IUPAC in its 1993 nomenclature for
liquid–liquid distribution [10]. Since this term was widely
used, “not recommended” is read as “not forbidden” and it
continues to be preferred to the “recommended” terms that
are distribution constant or partition ratio whose symbol
should be KD. To be understood by chromatographers, the
term partition coefficient will be used throughout this arti-
cle with the correct symbol, KD, but the IUPAC definitions
are recalled here. IUPAC points the common confusion
made between distribution ratio and distribution constant
(or partition coefficient).
2.1.1. Distribution ratio (D)
The ratio of the total analytical concentration of a solute
in the liquid phase 1, regardless of its chemical form, to
its total analytical concentration in the other phase. As de-
fined, the distribution ratio varies with experimental condi-
tions (chemical reaction, precipitation, ionization). It should
A. Berthod, S. Carda-Broch/J. Chromatogr. A 1037 (2004) 3–14 5
not be confused with the distribution constant (or partition
coefficient). The distribution ratio is an experimental param-
eter and its value does not necessarily imply that distribution
equilibrium between phases has been achieved [10].
2.1.2. Distribution constant or partition ratio (KD)
The ratio of the concentration of a substance in a single
definite form, A, in the liquid phase 1 to its concentration
in the same form in the other phase (liquid phase 2) at
equilibrium:
KD=[A]1
[A]2(1)
In equations relating to aqueous/organic systems, the organic
phase concentration is, by convention, the numerator and
the aqueous phase concentration the denominator. This def-
inition clearly states that the substance should have a single
definite form. It means that the distribution constant (parti-
tion coefficient) is a true constant. If the substance evolves,
the distribution ratio (D) will change; the distribution con-
stant (KD) will not since it is, by definition, invariant. Of
course, if the solutes are not ionizable and do not change
due to chemical reaction or complexation, then D=KD.
2.2. Thermodynamics
If we consider two practically immiscible solvents, 1 and
2, they form two liquid phases of one solvent saturated in the
other. When a solute A is introduced in such a biphasic liquid
system, it distributes between the two phases. Assuming
ideal mixtures, in the solvent 1 phase, the Gibbs free energy
of A, or chemical potential, µ1A, is expressed by:
µ1A =µ0
1A +RTlnx1A (2)
where µ0
1A is the standard chemical potential of A at infinite
dilution in liquid phase 1. Similarly, in the other phase, the
chemical potential, µ2A, is:
µ2A =µ0
2A +RTlnx2A (3)
If the chemical potential is not identical in the two phases,
mass transfer of A occurs, the mole fractions xchange so that
the chemical potential of A becomes equal in both phases,
i.e. the equilibrium is reached. Then:
µ0
1A −µ0
2A =RTlnx2A
x1A (4)
in which x2A/x1A is the distribution constant, KD, expressed
by:
x2A
x1A =KD2/1=exp µ0
1A −µ0
2A
RT (5)
In case of non-ideal mixtures, the mole fractions, x, should
be replaced by activities, a=xf, in which fis the activity co-
efficient. The distribution constant or partition ratio is con-
stant only if the activity coefficients are constant which is
not true in concentrated solutions [2]. Partition coefficients
are usually expressed as molarity ratio. Molar solubilities,
[A], and mole fractions, xA, are, in first approximation and
diluted solutions, proportional and related as follows:
[A]1=xA
V1(6)
where V1is the solvent 1 molar volume (M−1).
2.3. Effect of temperature
Eqs. (4) and (5) show that the distribution constant, the
partition coefficient, is sensitive to temperature. Eq. (4) ex-
presses the free energy of transfer, G2/1:
G2/1=RTlnKD2/1(7)
Assuming the standard molar enthalpy is constant in a lim-
ited temperature range, the plot of lnKD2/1versus 1/T(clas-
sical Van ’t Hoff plots) should produce a straight line with
slope G2/1/R. Unfortunately, this is not that simple. The
mutual solubility of the two solvents is also temperature de-
pendent. At the critical solution temperature, the biphasic
system becomes monophasic [5]. As a general rule, it is pos-
sible to consider that the effect of temperature on the KD
value is not great if the solvents are not very miscible and
the temperature change is not dramatic. An average change
of 0.009logKDunit per degree, either positive or negative,
was found for a variety of biphasic systems including the
octanol–water system [2,3,11].
2.4. Effect of chemical reactions
As clearly defined in the IUPAC definition, when any
chemical reaction occurs, the concentration of a particular
species will change, the distribution ratio will change, but the
distribution constant (partition ratio or partition coefficient)
of this particular species does not change. This implies that
the concentration of this species will change in the other
phase to maintain the chemical potentials equal in the two
phases. The distribution ratio, D, can change dramatically.
The case of a compound that has a carboxylic group will
be used as an example. This compound can be represented
as AH and can ionize:
AH A−+H+(8)
Defining, respectively, K◦
D,K−
Dand Das the distribution
constants of the molecular and ionized form of AH, and its
distribution ratio, it comes:
D=[AH]2+[A−]2
[AH]1+[A−]1(9)
Using the expression of Ka, the dissociation constant of AH,
Dcan be trivially formulated as:
D=K◦
D+K−
D(Ka/[H+])
1+(Ka/[H+])(10)
6A. Berthod, S. Carda-Broch/J. Chromatogr. A 1037 (2004) 3–14
0
2500
5000
7500
10000
1357911
pH
D
pKa= 3 5 7 9
Fig. 1. Change in the distribution ration, D, of ionizable compounds of
the AHA−type with different acidity strength. KDof all molecular
forms is 10000, KDof all anionic form is 1. The dissociation constants
are indicated as pKa=−log Ka.
Fig. 1 shows the change in the partitioning of a hydropho-
bic AH compound in a biphasic liquid system, e.g. the
octanol–water system, when the pH changes. At low pH
values, AH is essentially located in the organic phase with
very high Dvalues. Because the ionized form of AH, A−,
is hydrophilic, upon increase of pH, most of the solute is
extracted in the aqueous phase in the ionized form. The pH
zone corresponding to the change of phases depends on the
solute acidity (Kavalue, Fig. 1). The distribution ratio is di-
vided by two when pH =pKa. The case of the partitioning
of ionizable solutes was fully studied recently [12]. Appli-
cations of the equations developed can also be found in the
literature [13].
3. Measurement of partition coefficients
Several review articles describing the various methods
used to determine liquid–liquid partition coefficients and
especially Ko/wappeared recently [2,9,14,15]. The methods
are briefly described focusing on accuracy and speed.
3.1. The shake-flask method
KDvalues of many solutes are directly determined us-
ing the so-called shake-flask method. The solute is simply
partitioned between the two liquid phases of the proposed
solvent system in a test tube. After equilibrium and centrifu-
gation, the relative concentration in each layer is determined
using a variety of techniques. These include spectroscopic
methods, HPLC, GC, TLC among others.
3.1.1. Small-scale spectroscopic methods
The two phases of the liquid system should be mutu-
ally well saturated before use. A suitable quantity, typ-
ically 100g, of solute is deposited in a test tube. One
milliliter of each saturated layer is added to the sample,
and the tube is capped and equilibrated for several hours
on a wrist-action shaker. When needed, a blank solution
is also prepared. The absorbance values of the equili-
brated layers and the blanks are read at a suitable wave-
length, and the partition coefficient is calculated with the
equation:
KD=[A]org −[A]blank
[A]aq −[A]blank (11)
The experiment is repeated using a lower quantity, say 50 g,
of solute to increase the method precision and accuracy.
This method is a popular method well suited for com-
pounds that distribute in the liquid system used, i.e. when
0.1<K
D<10. In this case it provides a precision higher
than 1% using 1ml phase volumes and, and can be slightly
improved by using 2ml phase volumes. The method was
used with UV spectrometers [16,17] and fluorimeters [18].
Its principal drawback is that pure compounds are needed. If
impurities are present, they will distribute also in the liquid
phases but with a different KDvalue, biasing the measured
absorbance in each phase.
3.1.2. Separation methods for phase analysis
The solute purity problem can easily be solved using a
method of separation to quantitate the amount of solute
present in each phase. It is often not possible to simply sub-
stitute a chromatographic method for spectroscopy. Indeed,
the technique restrictions apply. It is not possible to use GC
if the solute is not volatile or if the liquid phase is aque-
ous. The organic solvent of the biphasic liquid system may
not be compatible with RPLC. However, the major advan-
tage of chromatographic methods is the ability to deal with
mixtures and to determine partition coefficients of several
solutes in a single assay. This is partly offset by the effort
necessary to develop a reliable chromatographic assay for
the mixture of interest. HPLC and GC were used in the de-
termination of pesticide partition coefficients [19–21]. TLC
was also used [22].
The major point of the shake-flask method is that it
is a direct method measuring without approximation the
liquid–liquid partition coefficient of the solute in the bipha-
sic liquid system in the flask. The weak point is its limited
range that is depending on the method used to analyze the
phases. It can be roughly given as −3<logKD<+3
with chromatographic phase analysis. Partition coefficients
as big as 100000, logKD=5, need to be known accu-
rately. In this case, if 100g of the compound is introduced
in a test tube containing 2ml of each phase of a liquid
system, only 1ng will pass in one phase while 99.999g
will stay in the other phase. The very low concentration,
1ng in 2ml or 0.5g/l, obtained at equilibrium in one
phase may be below the limit of detection of many an-
alytical methods. This level was accurately determined
using radiochemical methods working with radioactive iso-
topes [23]. This method cannot be easily used by most
laboratories.
A. Berthod, S. Carda-Broch/J. Chromatogr. A 1037 (2004) 3–14 7
3.1.3. Modern variations on the shake-flask method
Flow injection analysis (FIA) was used to determine Ko/w
coefficients. The manifold comprised three pumps: one for
the aqueous phase, one for the octanol phase and one for
the sample that could be dissolved either in the aqueous
or in the octanol phase. The octanol and aqueous phases
form segments in which the injected compound of interest
can partition. A spectrophotometric detector was used to
monitor the aqueous phase absorbance. The recorded peak
areas obtained at different flow ratios allowed to extrapolate
the Ko/wsolute value [24]. A modified FIA system was
designed for rapid Ko/wdetermination adapted for com-
binatorial chemistry. The result was obtained in less than
4min consuming less than 1l of sample [25]. The FIA
method cannot determine accurately Ko/wvalues higher
than 2000 (logKo/w<3.3).
A dialysis tubing was used to contain the aqueous phase
and to separate it from the octanol phase. Ultrasonic agita-
tion shortened the equilibration time and HPLC was used
to measure the solute concentration in each phase. Ko/w
values as high as 2.0×106(logKo/w=6.3, perylene) were
accurately determined in less than 6h [26].
3.2. Liquid chromatography and correlations
3.2.1. A simple and easy method
The use of liquid chromatographic methods using stan-
dards and correlations is the most widespread way to
measure rapidly liquid–liquid partition coefficients. It is ex-
tremely simple: the logarithms of the retention factor of the
solutes are linearly correlated with the logarithm of their
partition coefficients as first described by Collander [27]:
logKD=alogk+b(12)
The aand bcorrelation coefficients are determined mea-
suring the retention factors of a set of solutes on a RPLC
column using a mobile phase that differs from the biphasic
liquid system in which the KDvalues of the test solutes are
known. This method is fast and easy (the solute concentra-
tion does not need to be known). Several compounds can be
measured simultaneously. Unfortunately, the quality of the
KDdetermination dramatically depends on the test solutes
used to determine the aand bcorrelation coefficients.
By far, this method is used with C18 or C8columns and
methanol–water mobile phases to determine the Ko/wpar-
tition coefficients in the octanol–water system [28]. Indeed,
the hydrophobicity of a compound is often quantitatively
expressed by the logarithm of its octanol–water partition
coefficient, logKo/w. The prediction of logKo/wof a com-
pound is essential in various fields such as pharmacology,
toxicology, environmental chemistry and food chemistry.
With the development of reversed phase HPLC in many
laboratories, the readily accessible retention factor of a new
compound, compared to the standard but time consuming
shake-flask method, made the method extremely popular
and widely used.
The method was extensively evaluated in 1988 in an in-
terlaboratory study involving a large set of solutes and dif-
ferent C18 columns. The conclusions of the interlaboratory
test were [29]:
(1) At least six substances should be used to prepare the
correlation line.
(2) The substances should belong to the validated list of
reference compounds.
(3) The mobile phase should contain at least 25% (v/v)
water.
(4) Extrapolation beyond the calibration range obtained
with the selected substances should only be carried out
for very hydrophobic substances (logKo/w>6.
(5) When the conditions are fulfilled, the reliability of the
method is, in the range 0 <logKo/w<6, 0.5log unit
or less of deviation from the shake-flask value.
This test was used by the Organization for Economic Co-
operation and Development (OECD) to prepare a guideline
that should be followed to produce reliable Ko/wvalues us-
ing HPLC [30]. The guideline used the five points listed
above adding to the second point that “the reference com-
pounds should be structurally related to the test substance”.
The proposed list of reference compounds are summarized
in Table 1. Three main classes of reference compounds are
distinguished: neutral, basic and acidic compounds. This list
can be adapted for particular cases [31].
3.2.2. Congeneric effect
It is interesting to comment on the five points of the con-
clusion of the interlaboratory test and the OECD text addi-
tion to the second point. The first point proposes that a min-
imum of six solutes be used to prepare a correlation line.
This comes from statistical analysis (and common sense).
The second point proposes that the six substances belong to
a validated list and the OECD addition says that the sub-
stances should be structurally related. This deals with the
congeneric effect that is always observed with the HPLC
indirect method. The correlation line is very good and the
measured logKo/wvalues are very accurate when the com-
pounds used to prepare the correlation lines belong all to the
same family as well as the unknown compounds. A counter
example is: trying to estimate logKo/wof an amine using a
correlationline prepared with aromatic acids givesdisastrous
results. Alternatively, the correlation line prepared using a
set of solutes with widely differing functionalities will have
a poor regression coefficient and poor predictive capability.
The third point: “the mobile phase should contain at least
25% (v/v) water”, is due to the residual silanols that were
always present in 1988 C18 columns. Also it makes sense
that some water be present in the mobile phase when the
partitioning stationary phase octadecyl/mobile phase is cor-
related to octanol–water. The last point said in clear that if
the logKo/wvalue of a compound is 2, or Ko/w=100, the
method can give logKo/w=1.5, or Ko/w=32 as well as
logKo/w=2.5, or Ko/w=320. It means that the estimated
8A. Berthod, S. Carda-Broch/J. Chromatogr. A 1037 (2004) 3–14
Table 1
OECD recommended reference compounds
Reference compound logKo/wClass pKa
2-Butanone 0.3 n
4-Acetylpyridine 0.5 B
Aniline 0.9 B 4.63
Acetanilide 1.0 n
Benzyl alcohol 1.1 n
4-Methoxyphenol 1.3 A 10.26
Phenoxyacetic acid 1.4 A 3.12
Phenol 1.5 A 9.92
2,4-Dinitrophenol 1.5 A 3.96
Benzonitrile 1.6 n
Phenylacetonitrile 1.6 n
4-Methylbenzyl alcohol 1.6 n
Acetophenone 1.7 n
2-Nitrophenol 1.8 A 7.17
3-Nitrobenzoic acid 1.8 A 3.47
4-Chloraniline 1.8 B 4.15
Nitrobenzene 1.9 n
Cinnamic alcohol 1.9 n
Benzoic acid 1.9 A 4.19
p-Cresol 1.9 A 10.17
cis-Cinnamic acid 2.1 A 3.89
trans-Cinnamic acid 2.1 A 4.44
Anisole 2.1 n
Methyl benzoate 2.1 n
Benzene 2.1 n
3-Methylbenzoic acid 2.4 A 4.27
4-Chlorophenol 2.4 A 9.10
Trichloroethene 2.4 n
Atrazine 2.6 B
Ethyl benzoate 2.6 n
2,6-Dichlorobenzonitrile 2.6 n
3-Chlorobenzoic acid 2.7 A 3.82
Toluene 2.7 n
1-Naphthol 2.7 A 9.34
2,3-Dichloroaniline 2.8 B 2.05
Chlorobenzene 2.8 n
Allyl phenyl ether 2.9 n
Bromobenzene 3.0 n
Ethylbenzene 3.2 n
Benzophenone 3.2 n
4-Phenylphenol 3.2 A 9.54
Thymol 3.3 n
1,4-Dichlorobenzene 3.4 n
Diphenylamine 3.4 B 0.79
Naphthalene 3.6 n
Phenyl benzoate 3.6 n
Isopropylbenzene 3.7 n
2,4,6-Trichlorophenol 3.7 A 6.0
Biphenyl 4.0 n
Benzyl benzoate 4.0 n
2,4-Nitro-6-sec-butyl phenol 4.1 n
1,2,4-Trichlorobenzene 4.2 n
Dodecanoic acid 4.2 A 4.8
Diphenyl ether 4.2 n
Phenanthrene 4.5 n
N-Butylbenzene 4.6 n
Fluoranthene 4.7 n
Dibenzyl 4.8 n
2,6-Diphenylpyridine 4.9 n
Triphenylamine 5.7 n
DDT 6.2 n
From [30]. A: acidic compounds; B: basic compounds; n: neutral com-
pounds.
Ko/wvalue may be three times two big or too low. A recent
work with deactivated modern C18 columns did not change
much the 1988 conclusions. It did show that, with a 50–50
methanol–water mobile phase, the “H-bonding atmosphere
of the C18 deactivated phase (Capcell Pak C18, a silicone
polymer-coated silica gel chemically modified with C18)is
very similar to that of octanol” [32].
3.2.3. Variations on the method to improve the correlation
quality
Several parameters were added to the logKDversus logk
relationship to improve the correlation. For example, Yam-
agami et al. proposed the equation [33]:
logKo/w=alogk+b+ρσI+sSHA (13)
with a,b,ρand sare the constants for a column and a mobile
phase, and σIand SHA an inductive electronic constant and
the proton acceptor value of the studied solute. Of course,
adding parameters does improve the correlation quality, but
it decreases the method ease since the σIand SHA values of
unknown compounds should be also estimated.
It was proposed to use logkw, the extrapolated solute re-
tention factor in pure water [34,35]. logkwis obtained by
plotting logkof the solute versus the organic modifier con-
tent in the mobile phase and extrapolating the straight line
to 0% modifier (=pure water). Then, logkwcan be used in
Eq. (12) rather than logk. The logkwparameter became a
widely used chromatographic descriptors of hydrophobic-
ity per se [36–42]. It should be noted that this extrapolated
value can yield large errors. This is proved by the fact that
the extrapolated logkwvalue, obtained with the same so-
lute and column, may differ significantly depending on the
organic modifier used: methanol or acetonitrile.
It was also proposed to use gradient elution to increase
the speed of the partition coefficient estimation. It was nec-
essary to use the Abraham linear solvation coefficients to
correlate the solute retention factors with logKo/w[43,44].
Associating methanol gradient elution with a modern sta-
tionary phase that can work within the 2–8 pH range showed
that logKo/wvalues were measured in less than 45min with
a better than 0.3log unit error in the −0.1 to 4.0 range [45].
The use of octanol- or glycerol-coated stationary phases
complicated the method and did not fully suppress the con-
generic effect observed with the HPLC methods [46,47].
Micellar liquid chromatography is another HPLC variation
that can be used for hydrophobicity measurements [48].
Taking in account the congeneric effect and accepting
the 0.5log unit possible error on the determined logKo/w
values, HPLC is now one of the tools most often used to
evaluate quickly the magnitude of the hydrophobicity of a
new synthesized compound [49–52].
3.3. Capillary electrophoresis with ordered media
Classical CE uses an electric field to sort charged solutes
by differing electrophoretic mobilities. Micellar liquid chro-
A. Berthod, S. Carda-Broch/J. Chromatogr. A 1037 (2004) 3–14 9
matography had shown that the solute micelle binding con-
stants were related to solute hydrophobicity and Ko/wco-
efficients [48]. Micelles were introduced in CE to separate
neutral solutes according to their affinity for the micelles
[53]. Herbert and Dorsey showed that the retention factors
obtained with micellar electrokinetic capillary chromatogra-
phy (MEKC) allowed to estimate the respective Ko/wsolute
values over 8.5 orders of magnitude (−2.3<logKo/w<
6.2) with a 0.5log unit or lower error [54].However,itwas
demonstrated that the CE results were also prone to the con-
generic effect [55,56].
Ishihama et al. adapted the Herbert and Dorsey’s MEKC
methodproposing to use a microemulsion insteadofmicelles
[57]. Microemulsion electrokinetic capillary chromatog-
raphy (MEEKC) with a hexane (0.82%, w/w)–1-butanol
(6.5%, w/w)–sodium dodecyl sulfate (0.05M or 1.44%,
w/w) microemulsion, the logKo/wvalue of compounds
could be predicted within the −0.5 to 4.5 range with an
error lower than 0.2log unit. The MEEKC method was
used with a cationic microemulsion to work with positively
charged solutes [58]. The method was optimized to increase
the solute throughput [59].
It was showed that a universal set of standards could
be used to calibrate the MEEKC method [60]. Cationic,
anionic and surfactant association able to form vesicles
were tried to extend the hydrophobicity windows and/or
method accuracy [61]. Today, due to its higher throughput
and accuracy, the MEEKC method is becoming the dom-
inant indirect method of estimating logKo/win industry,
supplanting HPLC wherever a CE apparatus is available
[8,9].
3.4. The counter-current chromatography method
Counter-current chromatography is a separation technique
that uses a liquid mobile phase with a stationary phase that is
also liquid. There is no solid support for the liquid stationary
phase [62]. Centrifugal fields maintain the two immiscible
liquid phase together. The only physicochemical interaction
that is responsible for solute retention in a CCC column is
liquid–liquid partitioning. The retention equation is:
VR=VM+DVS(14)
with the subscript R, M and S standing for retention, mobile
and stationary phase volumes. Since there are only liquid
phases in the CCC column, the column volume, VC, is:
VC=VM+VS(15)
The CCC machine (=column) volume is known, and then
it is necessary to measure only one volume, either VM, the
mobile phase volume, or VS, the stationary phase volume.
3.4.1. Direct partition coefficient measurement
3.4.1.1. Method description. In CCC, the retention vol-
ume of solute A is directly related, without any correlation,
Fig. 2. Direct determination of the Do/wdistribution ratio of: (1) ben-
zamide; (2) 2-acetoxy benzoic acid; (3) acetophenone; (4) benzoic acid;
(5) 2-chlorobenzoic acid; (6) chlorophenol; (7) 2-chloronitrobenzene. Ma-
chine Sanki CPC-LLN, VC=125ml, Voct =VS=22.2ml, aqueous
mobile phase at 5mlmin−1and pH 4. Adapted from [63].
to its distribution ratio, D, in the liquid system used:
D=[A]stationary phase
[A]mobile phase =VR−VM
VS=VR−VM
VC−VM(16)
Often, there is no chemical reaction, ionization or complexa-
tion in the two liquid phases, then D=KD(see Section 2.1).
Otherwise, the distribution ratio is related to the partition
coefficient taking in account the chemical changes that oc-
curred in the liquid phases [63,64].
Fig. 2 illustrates the method showing the CCC chro-
matogram obtained when a mixture of six compounds was
injected in a CCC column equilibrated with an octanol sta-
tionary phase and eluted with a buffer mobile phase [63].
Without any correlation, the solute retention volumes are
related to the Do/wratio of the solutes by Eq. (16). For ex-
ample, the retention volume of 2-chlorophenol (compound
6inFig. 2) is 3320ml (retention time of 660min or 11h)
allowing to calculate its Do/wratio as 145. Since this com-
pound is not ionized at pH 4, its Do/wratio is also its Ko/w
octanol–water partition coefficient. Fig. 2 illustrates per-
fectly the CCC direct method. Its big advantage is that it
gives the Dratio of compounds directly and in any biphasic
liquid system. The drawback is the range of the measurable
Dratios. Large Dvalues need prohibitive times and mo-
bile phase volumes to be determined. It took 11h and over
3.2l of aqueous mobile phase to measure a Dratio of 145
(logD=2.16). Orders of magnitude higher values should
be determined with hydrophobic solutes. The method was
used for biological [65], ionizable [12,13,66], or industrial
compounds [63,67]. Its use is linked to the availability of
CCC chromatographs that are difficult to obtain. This is the
main reason why this excellent method for partition coeffi-
cient determination is still little known [68].
3.4.1.2. Method range and accuracy. The error on Dwas
established as [64]:
D
D=VR
VR−VM(17)
10 A. Berthod, S. Carda-Broch/J. Chromatogr. A 1037 (2004) 3–14
In the case of 2-chlorophenol already used as an exam-
ple compound (Fig. 2), the error on reading the retention
volume is estimated to be 40ml because the peak is re-
ally broad. Then, D/D =40/(3320 −103)=0.0124
or D =1.8. This clearly shows the very high accuracy
of the CCC method compared to the HPLC method. For
2-chlorophenol, the result is D=145 ±2 or logD=
2.16 ±0.005 that should be compared to the accuracy of
the HPLC method that is commonly 0.05log unit in this
range of hydrophobicity and could be as high as 0.5log unit
with higher hydrophobic compounds and difficult extrapo-
lation (see OECD method, Section 3.2.1). Eq. (17) shows
that the accuracy may decrease as the solute retention vol-
ume decreases, especially for very low retention volumes.
It was practically estimated that the solute retention volume
should be at least 5ml higher than the mobile phase vol-
ume (=the “dead” volume) to obtain a maximum acceptable
relative error of 10% on D[64]. Then there is a minimum
measurable Dratio that is 5/VS(Eqs. (16) and (17)). With
Fig. 2,VSvalue of 22.2ml, the minimum measurable Dra-
tio is 0.22. The maximum value depends on detector sen-
sitivity to see the peaks that dramatically broaden with the
long retention times (Fig. 2). It cannot be higher than 200.
The method range is about −1<logD<2.3, that is more
than three orders of magnitude with an accuracy better than
0.1log unit.
3.4.1.3. Method variation. In CCC, the mobile phase can
be any one of the two phases of the biphasic liquid system.
Fig. 2 shows that 2-chlorophenol is highly retained by the oc-
tanol stationary phase. Then, why not use an octanol mobile
phase with the aqueous buffer being the stationary phase?
Since Eq. (16) still applies, the measured distribution ratio
would be Dw/owhich is trivially 1/Do/w. Assuming that the
CCC column made in these conditions has an aqueous sta-
tionary phase volume of 50ml, the VMoctanol phase would
be 75ml, then Eq. (14) shows that the retention volume of
2-chlorophenol would be 75 +50(1/145)=75.3ml. The
0.3ml difference between the “dead” volume and the solute
cannot be measured accurately. Clearly, the use of an octanol
mobile phase will not help to measure accurately very high
Ko/wcoefficient. It may save time in the case of Ko/wco-
efficient between 10 and 50. These compounds would have
measurable and short retention volumes. In the case of oc-
tanol, it is recalled that its viscosity is around 7cP preclud-
ing a 5mlmin−1flow rate in an analytical CCC machine
[13,63].
3.4.2. Using the liquid nature of the stationary phase
The three orders of magnitude logKDrange of the CCC
direct method is not wide enough. The unique feature of the
CCC technique is that the stationary phase is a support-free
liquid phase [62]. This allows innovative uses of the tech-
nique that cannot be conceived with any other chromato-
graphic technique with a solid or a solid support for the
stationary phase.
Fig. 3. The dual-mode CCC method. Step A: the mobile phase is the
dense hatched liquid phase flowing from left to right and eluting solutes
1 and 2 with Dvalue of 0.2 and 1, respectively. Step B: solutes 3 and 4,
with Dvalue of 10 and 20, respectively, move very slowly to the right.
Step C: the mobile phase is now the light phase flowing from right to
left, solutes 4 and 3 elute rapidly as shown by the chromatogram.
3.4.2.1. Dual-mode CCC. The solutes with very high D
values need a very large volume of mobile phase to emerge
outside the column because they move very slowly in the
column. To force them out of the CCC column, the role of
the phases can be reversed after some reasonable flowing
time as illustrated by Fig. 3. It was demonstrated that the
solute Dvalues was simply expressed by [67,69,70]:
D=V1
V2(18)
with V1the volume of phase passed in the normal way (steps
1 and 2, Fig. 3) and V2, the volume of the other phase passed
in the reversed way (step 3, Fig. 3). It cannot be simpler.
Neither the machine volume, nor the phase ratio inside the
machine or the flow rate matter, they all cancel. There is a
detection problem because the liquid phase flowing in the de-
tector during the first phase differs from the one flowing dur-
ing the reversed mode. Since the V2volume is usually much
lower than the V1volume, it is possible to reduce the flow
rate to have more time to re-equilibrate the detector. Using
the dual-mode method, Ko/wcoefficients as high as 20000
(logKo/w=4.3) were accurately determined [69,70].
In the case of the Do/wmeasurements, the distribution
ratio is simply Vaq/Voct. The error was estimated as:
dDo/w
Do/w=Vaq
Vaq −Voct
Voct (19)
A. Berthod, S. Carda-Broch/J. Chromatogr. A 1037 (2004) 3–14 11
Since this method is used to measure high Do/wvalue, the
aqueous phase volume is always much higher than the oc-
tanol phase volume. The first term of Eq. (19) is always very
small compared to the second term. It is critical to mini-
mize the error on the Voct volume using a very low flow
rate for the octanol phase. Gluck et al. recommends that the
Voct volume be at least 3ml [69]. If a maximum reasonable
Vaq volume is 20l (55h at 6mlmin−1), then the maximum
measurable Do/wis 6700 or logDo/w=3.8 with a 0.1log
unit error. Accepting the 0.5log unit error commonly ob-
served with any other methods, logDo/was high as 4.5 can
be measured (Do/w∼32000) [66,67,69].
3.4.2.2. Cocurrent CCC. This is another method that uses
the liquid nature of the stationary phase. The highly retained
solutes stay too long inside the liquid stationary phase. Then,
pushing slowly the liquid stationary phase in the same di-
rection as the mobile phase will force solutes to elute more
rapidly. The full theoretical treatment of the method was
given by Berthod [71]. It was shown that VR, the retention
volume of a solute eluted with a Faq aqueous flow rate and
a very low Foct octanol flow rate, was given by:
VR=(Faq +Foct)Vaq +Do/wVoct
Faq +Do/wFoct (20)
Eq. (20) shows that the retention volume of a lipophilic
solute will decrease dramatically with Foct. For example, a
compound with a Do/wof 1000 has a retention volume of 40 l
or a retention time of 4 days and 15h at 6mlmin−1with a
125ml machine containing only 40ml of octanol (Eq. (14)).
If a second pump is used to push slowly (0.1mlmin−1) the
octanol “stationary” phase, Eq. (20) shows that the retention
volume drops to 2.3l or 6h and 20min, a 94% reduction in
experiment duration [64].
Table 2
Octanol–water distribution ratio measured using different CCC methods [60–64,69,70]
Solute VR(ml) tR(min) Voct (ml) Do/walogDo/wLiterature value
Direct measurementsb
Benzamide 200 40 4.4 ±0.1 0.64 ±0.01 0.64 ±0.02
Benzyl alcohol 383 76.5 12.6 ±0.1 1.1 ±0.01 1.1 ±0.02
Phenol 760 152 29.5 ±0.2 1.47 ±0.01 1.5 ±0.02
2-Chlorophenol 3320 664 145 ±2 2.16 ±0.01 2.17 ±0.03
Dual-mode measurementsc
Phenol 157 42 5.15 30.5 ±0.8 1.48 ±0.02 1.5 ±0.02
2-Chlorophenol 487 103 2.8 174 ±2 2.24 ±0.05 2.17 ±0.03
Toluene 2627 536 5.5 478 ±5 2.68 ±0.02 2.71 ±0.06
Naphthalene 4795 962 1.6 3000 ±40 3.48 ±0.02 3.23 ±0.2
Cocurrent measurementsd
2-Chlorophenol 2700 337 6.8 147 ±5 2.17 ±0.01 2.17 ±0.03
Toluene 5570 697 14 490 ±10 2.69 ±0.01 2.71 ±0.06
Naphthalene 9280 1162 23.2 5100 ±100 3.70 ±0.03 3.23 ±0.2
Phenanthrene 9810 1229 24.6 20000 ±1000 4.3 ±0.1 4.4 ±0.3
aMost compounds cannot ionize, so the Do/wdistribution ratio can be read as the Po/woctanol–water partition coefficient.
bMachine volume 125ml, flow rate 5ml min−1,Voct =22.2 ml.
cMachine volume 125ml, flow rate aqueous phase 5ml min−1, octanol phase 0.5 mlmin−1in the reversed direction after switching mode.
dMachine volume 49ml, flow rate Faq =8mlmin−1,Foct =0.02mlmin−1in the same direction and simultaneously with Faq.
It was demonstrated that the minimum error (maximum
selectivity) was obtained around the Do/wrange equal to
the flow ratio Faq/Foct [71]. For example, to determine ac-
curately a Do/wvalue of 1000, an octanol flow rate of
10lmin−1should be associated with a main aqueous flow
rate of 10mlmin−1. The experiment duration will be bal-
anced with the desired accuracy. Low octanol flow rate will
produce higher accuracy on high Do/wvalues at the cost of
prohibitive retention times. Solute detection is another seri-
ous drawback of the method. As the method was designed,
two immiscible liquid phases enter and leave the CCC col-
umn. The evaporative light scattering detector can be used
to detect solid solutes. Alternately, a mixing agent such as
1-propanol can be added post-column. The UV detector will
be usable but the effluent cannot be recycled [64,70].Table 2
lists the phase volumes, retention times (experiment dura-
tion) and Do/wratio for some compounds comparing the
various CCC techniques that can be used. We recently pro-
posed a new method to extend the hydrophobicity window
using again the liquid nature of the stationary phase: the
elution-extrusion method [72]. However, this method will
not allow enhancing the accuracy of partition coefficient
measurement by CCC.
4. Application: homologues and partition coefficients
Homologues are families of compounds having the same
functionalities and differing by the length of an alkyl chain.
With such homologous series of compounds, it was shown
that the hydrophobic contribution of the alkyl chain could
be dissociated from the contribution of the functional groups
[73,74]. Starting from Eq. (7), it can be written:
RTlnKD=Gfunctional groups +nCH2GCH2(21)
12 A. Berthod, S. Carda-Broch/ J. Chromatogr. A 1037 (2004) 3–14
where nCH2is the number of methylene groups in the alkyl
chain.
4.1. Alkylbenzene partition coefficient in the
heptane–methanol–water system
The alkylbenzene partition coefficients were studied
by CCC in biphasic liquid systems made with heptane,
methanol and 20% (v/v) or less water [75]. In these biphasic
liquid systems, it was found that logKDof the alkylben-
zenes were always linearly related to their carbon number.
However, the regression lines were excellent (r2≥0.999) if
the water content was expressed in mole fraction, χ, rather
than the volume percentage. It was experimentally estab-
lished that, in this biphasic liquid system containing 20%
(v/v) or less water, for alkylbenzenes up to dodecylbenzene
and at 295K (22◦C), the heptane–hydrated methanol parti-
tion coefficients of the alkylbenzene homologues could be
expressed by:
logKD=(0.342nCH2+1.72)χ +0.080nCH2
+0.034 (nCH2<13,χ<0.361 and T=295K)
(22)
For example, the heptane–hydrated methanol partition co-
efficient of hexylbenzene in the biphasic liquid system
heptane–(methanol–water (85–15, v/v)) is 38.5, logKD=
(0.342 ×6+1.72)×0.284 +0.08 ×6+0.034 =1.585.
The CCC measured value was 40 or logKD=1.60 [75].
Trivially, the hydrated methanol–water partition coefficient
is 1/38.5=0.026 or −1.585 for the log value.
The partitioning of the alkylbenzenes in the heptane–
(methanol–water) biphasic liquid system could be related
to the similar partitioning occurring between the liquid-like
octadecylsilyl-bonded layer in RPLC with methanol rich
mobile phases. The correlation of the alkylbenzene reten-
tion factors with the water content in the mobile phase is a
common selectivity test in RPLC [73,74,76,77].
4.2. Petroleum components and waterless biphasic liquid
systems
In another study, the partition coefficients of alkylben-
zenes and polyaromatic hydrocarbons (PAH) were studied in
waterless biphasic liquid systems such as heptane–dimethyl
sulfoxide (DMSO), heptane–2-furancarboxaldehyde (fur-
fural), heptane–dimethylformamide (DMF), heptane–N-
methylpyrrolidone (NMP), all four solvents of the dipolar
and aprotic family, and heptane–methanol, a polar and pro-
tic solvent [78]. The linear relationship of Eq. (22) was
verified in aprotic solvents. It was also found that a similar
linear relationship could be established between the logKD
values of the PAHs and the number of sp2hybridized
carbons in their molecules.
The very interesting point was that, for all four aprotic
solvents studied, the solute affinity for the heptane phase
Table 3
Transfer energy for alkylated (CH2) and arylated (sp2) hydrocarbon in
waterless heptane/polar solvent systems
Parameter Solvent
NMP DMF Furfural DMSO
GCH2(kJmol−1) 0.33 0.42 0.56 0.94
Gsp2(kJmol−1)−0.56 −0.39 −0.40 −0.24
Ratio −GCH2/Gsp20.59 1.07 1.42 3.93
Data from [78]. Temperature 298K or 25◦C; NMP: N-methylpyrrolidone;
DMF: dimethylformamide; furfural: 2-furancarboxaldehyde; DMSO:
dimethylsulfoxide.
increased with the length of the alkyl chain and decreased
with the number of sp2hybridized carbons. From a ther-
modynamic point of view, the addition to a hydrocarbon
molecule of a methylene group produced a positive GCH2
contribution when the addition of a sp2hybridized carbon
produced a negative Gsp2contribution. Table 3 lists the
GCH2and Gsp2contribution for the four solvents stud-
ied. It also gives the ratio of the two contributions. This ratio
shows how many sp2hybridized carbon should be added to
the molecule of a compound to compensate for one more
methylene group so that the new compound partition coef-
ficient does not change. For example, the heptane–DMSO
GCH2/Gsp2ratio is 3.93, almost 4. In this system, the
partition coefficient of ethylbenzene is 2.12, 33% lower
than the 3.16 value for propylbenzene [78]. However, the
heptane–DMSO partition coefficient of propylnapthalene is
likely very close to 2.12 (not measured in [78]). Propyl-
naphthalene has four more sp2hybridized carbons com-
pared to propylbenzene. These four sp2hybridized carbons
in propylnaphthalene compensate for the methylene group
and ethylbenzene and propylnaphthalene should have both
similar partition coefficients in the heptane–DMSO system.
These results clearly show the interest of knowing the
partition coefficients of compounds in liquid–liquid bipha-
sic systems. Like known for decades, the study shows that
DMSO is a solvent that will selectively extract the PAHs
with little alkyl substitution from a crude oil or petroleum
fraction [5]. However, it allowed a clear view of the prop-
erties of several similar solvents. For example, NMP with
its low GCH2/Gsp2ratio (Table 3) will extract aromatic
compounds and PAHs including the alkyl-substituted ones,
leaving a raffinate rich in alkanes and saturated hydrocar-
bons.
4.3. Quinoline homologues in heptane–acetonitrile–
methanol systems
It was found that 2-alkylquinolines could be very active
against leishmaniasis, a tropical disease due to a proto-
zoan of the genus Leishmania. A French team synthesized
2-alkylquinolines with 1–16 carbon atoms in the alkyl chain
and separated them using CCC and the heptane–methanol–
acetonitrile biphasic liquid system [79]. In this waterless
A. Berthod, S. Carda-Broch/ J. Chromatogr. A 1037 (2004) 3–14 13
system they found relationship similar to Eq. (21) and
listed the slopes and intercepts of the lines in four differ-
ent biphasic compositions. The Kheptane/polar phase partition
coefficients were increasing exponentially with the car-
bon number in the alkyl chain and also with the methanol
content in the polar non-aqueous phase.
Using the published data [79], it is possible to derive
the general equation relating, Kheptane/polar phase, the partition
coefficient of the 2-alkylquinoline:
lnKheptane/polar phase =−1.3+0.3nC−0.235%MeOH
−0.145nC×%MeOH (23)
where nCcarbon number in the alkyl chain and in
the heptane/polar phase, with the polar phase made of
%MeOH (v/v) methanol percentage. For example, the
Kheptane/polar phase of 2-hexylquinoline and 2-octylquinoline
in the heptane–acetonitrile–methanol (5:1:4, v/v/v) liquid
system are calculated as 0.681 and 0.984 with nC=6 and
8, respectively, and %MeOH =0.8 or 80% of methanol in
the polar phase. The experimental values were measured
as 0.68 and 0.99 (accuracy not given) in [79]. The exact
knowledge of these partition coefficients will be very useful
in QSAR studies on the activity of these quinoline homo-
logues on the protozoan parasite. They will be used also
to separate and purify a particular member of the family
[79].
5. Conclusion
Separationmethods can be used to determineliquid–liquid
partition coefficient of solutes. The most needed liquid–
liquid partition coefficient is the octanol–water partition
coefficient. Ko/wis accepted as a good reference parameter
for solute hydrophobicity. This parameter may not be the
best one for this purpose and this could be extensively dis-
cussed. However, it is convenient to have it and it is striking
that any new method able to estimate solute hydrophobicity
systematically refers to Ko/wor logKo/wto show its effi-
cacy. Indeed, Ko/wcan be rapidly estimated using capillary
electrophoresis with a micellar or microemulsion solution
and/or RPLC. Both methods give rapidly the Ko/worder
of magnitude. Next, if needed, a more accurate value can
be obtained again using RPLC with a careful standardiza-
tion taking in account the congeneric effect. CCC working
with the octanol–water biphasic system can produce very
accurate Ko/wvalue without any approximation but within
a limited range of hydrophobicity and with an intensive
hardware. It was shown that the distribution ratio, D, of ion-
izable compounds could also be measured by CCC. Almost
any biphasic liquid system can be used in CCC allowing to
measure partition coefficients in a wide variety of solvent
environments. We are considering the new class of solvents:
the room temperature ionic liquids [80,81].
Acknowledgements
The authors gratefully acknowledge the financial support
from the European Community through the Marie Curie Fel-
lowship HPMF-CT-2000-00440 and the INTAS grant No.
00-00782. A.B. thanks the Centre National de la Recherche
Scientifique FRE2194-UMR5180 for continuous support.
References
[1] M. Berthelot, E. Jungfleish, Ann. Chim. Phys. 26 (1872) 396.
[2] J. Sangster, Octanol–Water Partition Coefficients, Fundamentals and
Physical Chemistry, Wiley, Chichester, 1997.
[3] C. Hansch, A. Leo, D. Hoelkman, Exploring QSAR: Fundamentals
and Applications in Chemistry and Biology, American Chemical
Society, Washington, DC, 1995.
[4] P.C. Taylor, in: J.C. Deardon (Ed.), Comprehensive Medicinal Chem-
istry: The Rational Design, Mechanistic Study and Therapeutic Ap-
plication of Chemical Compounds, Pergamon Press, New York, 1990,
pp. 241–294.
[5] J. Rydberg, C. Musikas, G.R. Choppin, Principle and Practices of
Solvent Extraction, Marcel Dekker, New York, 1992.
[6] A.W. Czarnik, Anal. Chem. 70 (1998) 378A.
[7] S. Borman, Chem. Eng. News November (2002) 43.
[8] S.K. Poole, S. Patel, K. Dehring, H. Workman, J. Dong, J. Chro-
matogr. B 793 (2003) 265.
[9] S.K. Poole, C.F. Poole, J. Chromatogr. B 797 (2003) 3.
[10] N.M. Rice, H.N.M.H. Irving, M.A. Leonard, Pure Appl. Chem.
65 (11) (1993) 2373.
[11] A. Leo, C. Hansch, D. Elkins, Chem. Rev. 71 (1971) 525.
[12] A. Berthod, S. Carda-Broch, M.C. Garcia-Alvarez-Coque, Anal.
Chem. 71 (1999) 879.
[13] S. Carda-Broch, A. Berthod, J. Chromatogr. A 995 (2003) 55.
[14] L.G. Danielsson, Y.H. Zhang, Trends Anal. Chem. 15 (1996) 188.
[15] K. Valko (Ed.), Separation Methods in Drug Synthesis and Purifica-
tion, Elsevier, Amsterdam, 2000.
[16] W. Szczepaniak, A. Scymanski, J. Liq. Chromatogr. Relat. Technol.
23 (2000) 1217.
[17] N. Gulyaeva, A. Zaslavsky, P. Lechner, M. Chlenov, A. Chait, B.
Zaslavsky, Eur. J. Pharm. Sci. 17 (2002) 81.
[18] B. de Castro, P. Gameiro, C. Guimaraes, J. Lima, S. Reis, Biophys.
Chem. 90 (2001) 31.
[19] M.C. Bowman, M. Beroza, Anal. Chem. 38 (1966) 1544.
[20] D. Stopher, S. McClean, J. Pharm. Pharmacol. 42 (1990) 144.
[21] J. Thomas, O. Adetchessi, F. Pehourcq, P. Dallet, C. Jarry, J. Liq.
Chromatogr. Relat. Technol. 20 (1997) 671.
[22] D.G. Martin, S.A. Mizsak, W.C. Krueger, J. Antibiot. 38 (1985) 746.
[23] J. Tools, D.T.H. Sum, Environ. Toxicol. Chem. 14 (1995) 1675.
[24] L.G. Danielsson, Y.H. Zhang, J. Pharm. Biomed. Anal. 12 (1994)
1475.
[25] K. Carlsson, B. Karlberg, Anal. Chim. Acta 423 (2000) 137.
[26] J.T. Andersson, W. Schräder, Anal. Chem. 71 (1999) 3610.
[27] R. Collander, Acta Chem. Scand. 5 (1951) 774.
[28] S.H. Hunger, P.S. Cheung, G.H. Chiang, J.R. Cook, in: W.J. Dunn,
J.H. Block, Pearlman, R.S. (Eds.), Partition Coefficient Determination
and Estimation, Pergamon Press, New York, 1986, pp. 69–82.
[29] W. Klein, H. Kordel, M. Weiss, H.J. Poremski, Chemosphere 17
(1988) 361.
[30] OECD, Guideline for Testing of Chemicals, vol. 117, 1989,
http://www.oecd.org.
[31] M. Cichna, P. Markl, J.F.K. Huber, J. Pharm. Biomed. Anal. 13
(1995) 339.
[32] C. Yamagami, K. Kawase, K. Iwaki, Chem. Pharm. Bull. 50 (2002)
1578.
14 A. Berthod, S. Carda-Broch/ J. Chromatogr. A 1037 (2004) 3–14
[33] C. Yamagami, K. Kawase, T. Fujita, Quant. Struct.–Act. Relat. 18
(1999) 26.
[34] T. Braumann, J. Chromatogr. 373 (1986) 191.
[35] C.H. Lochmüller, M. Hui, J. Chromatogr. Sci. 36 (1998) 11.
[36] R.J.E. Growls, E.W. Ackerman, H.H.M. Korsten, L.J. Hellebrekers,
D.D. Breimer, J. Chromatogr. B 694 (1997) 421.
[37] C.M. Du, K. Valko, C. Bevan, D. Reynolds, M.H. Abraham, Anal.
Chem. 70 (1998) 4228.
[38] A. Kaune, R. Brüggemann, A. Kettrup, J. Chromatogr. A 805 (1998)
119.
[39] S. Griffin, S.G. Willye, J. Marckham, J. Chromatogr. A 864 (1999)
221.
[40] L. Novotny, M. Abdel-Hamid, H. Hamza, J. Pharm. Biomed. Anal.
24 (2000) 125.
[41] Q.C. Meng, H. Zou, J.S. Johansson, R.G. Eckenhoff, Anal. Biochem.
292 (2001) 102.
[42] T.L. Djakovic-Sekulic, S.M. Petrovic, N.U. Perisic-Janjic, S.D. Petro-
vic, Chromatographia 54 (2001) 60.
[43] K. Valko, C.M. Du, C. Bevan, D.P. Reynolds, M.H. Abraham, Curr.
Med. Chem. 8 (2001) 1137.
[44] M.C. Du, K. Valko, C. Bevan, D.P. Reynolds, M.H. Abraham, J.
Liq. Chromatogr. Relat. Technol. 24 (2001) 635.
[45] N.C. Dias, M.I. Nawas, C.F. Poole, Analyst 128 (2003) 427.
[46] A. Kaune, R. Bruggemann, A. Kettrup, J. Chromatogr. A 805 (1998)
119.
[47] K. Miyake, M. Nobayasu, H. Terada, J. Chromatogr. 439 (1988)
227.
[48] A. Berthod, C. Garcia-Alvarez-Coque, Micellar Liquid Chromatogra-
phy, Chromatographic Science Series, vol. 83, Marcel Dekker, New
York, 2000.
[49] W.J. Lambert, J. Chromatogr. A 656 (1993) 469.
[50] S. Griffin, S.G. Wyllie, J. Markham, J. Chromatogr. A 864 (1999)
221.
[51] M. Nakamura, T. Suzuki, K. Amano, S. Yamada, Anal. Chim. Acta
428 (2001) 219.
[52] S.K. Sahu, G.G. Pandit, J. Liq. Chromatogr. Relat. Technol. 26
(2003) 26.
[53] S. Terabe, K. Otsuka, K. Ichikawa, T. Ando, Anal. Chem. 56 (1984)
111.
[54] B.J. Herbert, J.G. Dorsey, Anal. Chem. 67 (1995) 744.
[55] M.D. Trone, M.S. Leonard, M.G. Khaledi, Anal. Chem. 72 (2000)
1228.
[56] K.A. Kelly, T.S. Burns, M.G. Khaledi, Anal. Chem. 73 (2001) 6057.
[57] Y. Ishihama, Y. Oda, K. Uchikawa, N. Asakawa, Anal. Chem. 67
(1995) 1588.
[58] Y. Ishihama, Y. Oda, N. Asakawa, Anal. Chem. 68 (1996) 4281.
[59] S.J. Gluck, M.H. Benko, R.K. Hallberg, K.P. Steele, J. Chromatogr.
A 744 (1996) 141.
[60] W.L. Klotz, M.R. Schure, J.P. Foley, J. Chromatogr. A 930 (2001)
145.
[61] W.L. Klotz, M.R. Schure, J.P. Foley, J. Chromatogr. A 962 (2002)
207.
[62] A. Berthod, CCC, The Support-Free Liquid Stationary Phase, Com-
prehensive Analytical Chemistry, vol. 38, Elsevier, Amsterdam/New
York, 2002.
[63] A. Berthod, Y.I. Han, D.W. Armstrong, J. Liq. Chromatogr. 11 (1988)
1441.
[64] A. Berthod, in: A. Foucault (Ed.), Centrifugal Partition Chromatogra-
phy, Chromatographic Science Series, vol. 68, Marcel Dekker, New
York, 1995, Chapter 7, pp. 167–197.
[65] P. Vallat, N. El Tayar, B. Testa, I. Slacanin, A. Martson, K.
Hostettmann, J. Chromatogr. 504 (1990) 411.
[66] A. Berthod, V. Dalaine, Analusis 20 (1992) 325.
[67] S.J. Gluck, E.J. Martin, J. Liq. Chromatogr. 13 (1990) 3559.
[68] A. Finizio, M. Vighi, D. Sandroni, Chemosphere 34 (1997) 131.
[69] S.J. Gluck, E. Martin, M.H. Benko, in: A. Foucault (Ed.), Centrifugal
Partition Chromatography, Chromatographic Science Series, vol. 68,
Marcel Dekker, New York, 1995, Chapter 8, pp. 199–218.
[70] A. Berthod, R.A. Menges, D.W. Armstrong, J. Liq. Chromatogr. 15
(1992) 2769.
[71] A. Berthod, Analusis 18 (1990) 352.
[72] A. Berthod, M.J. Ruiz-Angel, S. Carda-Broch, Anal. Chem. 75 (2003)
5886.
[73] W.R. Melander, Cs. Horváth, Chromatographia 15 (1982) 86.
[74] G. Guiochon, in: Cs. Horvath (Ed.), HPLC: Advances and Perspec-
tives, vol. 2, Academic Press, New York, 1980.
[75] A. Berthod, M. Bully, Anal. Chem. 63 (1991) 2508.
[76] H. Colin, G. Guiochon, J.C. Diez-Masa, Anal. Chem. 53 (1981) 146.
[77] M. Czok, H. Engelhardt, Chromatographia 27 (1989) 5.
[78] A. Berthod, A.I. Mallet, M. Bully, Anal. Chem. 68 (1996) 431.
[79] P. Duret, M.A. Fakhfakh, C. Herrenknecht, A. Fournet, X. Franck,
B. Figadere, R. Hocquemiller, J. Chromatogr. A 1011 (2003) 55.
[80] A. Berthod, S. Carda-Broch, J. Liq. Chromatogr. Relat. Technol. 26
(2003) 1493.
[81] A. Berthod, S. Carda-Broch, D.W. Armstrong, Anal. Bioanal. Chem.
375 (2003) 191.