Title: Solvatochromic investigations of chromatographic processes
Full Citation
Permanent Link: http://ufdc.ufl.edu/UF00099466/00001
 Material Information
Title: Solvatochromic investigations of chromatographic processes
Physical Description: xv, 244 leaves : ill. ; 28 cm.
Language: English
Creator: Michels, James Joseph, 1962-
Copyright Date: 1989
Subject: Liquid chromatography   ( lcsh )
Solvation   ( lcsh )
Aluminum oxide   ( lcsh )
Chemistry thesis Ph. D
Dissertations, Academic -- Chemistry -- UF
Genre: bibliography   ( marcgt )
non-fiction   ( marcgt )
Statement of Responsibility: by James Joseph Michels.
Thesis: Thesis (Ph. D.)--University of Florida, 1989.
Bibliography: Includes bibliographical references (leaves 231-243)
General Note: Typescript.
General Note: Vita.
 Record Information
Bibliographic ID: UF00099466
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: alephbibnum - 001512962
oclc - 21924527
notis - AHC5953


This item has the following downloads:

solvatochromicin00mich ( PDF )

Full Text







This dissertation is dedicated to Beverly,
my father, mother and sister, and also my late
grandfather, John Braun, who saw potential
in me even in my pre-school days


The first person I would like to thank is my research

director, Dr. John Dorsey. He was my primary reason for

attending the University of Florida and made my stay here

quite enjoyable. I never had to worry about him looking over

my shoulder shouting commands every time I did something but

he was always available when I needed to talk about my

research (even though he did a very good Rick Yost

impersonation in the Fall of 1988). He always found ways to

fund and supply the group, whether it be by grants or gifts

from industry. He has also been a good friend and I will

surely miss events like imported beer tasting, WINO 1001,

group dinners and depth-charge tournament-action!!! I wish

him the best of luck on his new life at the University of


An important group of people who deserve recognition are

those who provided technical assistance to me over the last

four years. In the U-F chemistry department, thanks go to

Dr. Vaneica Young and Linda Volk for help with the computer

programming of the polynomial confidence interval

calculations, Jerry Grunewald for loaning the Drago group

muffle furnace and acquiring FTIR spectra of the alumina

samples, Evan House of Dr. Sam Colgate's group for his help

with some of the machining I have had to do, and Mike

Mignardi and Alicia O'Reilly for the use of the Winefordner

group spectrofluorimeter. Dr. John Cornell of the University

of Florida Statistics Department passed on useful insight for

the calculation and application of confidence interval

calculations. Dr. John Baty and Sheila Sharp of the

Ninewells Hospital in England generously provided

chromatographic retention data that were previously

unavailable. One last person is Dr. John Novak of the

Aluminum Company of America for contributing alumina samples

and arranging the donation of the IBM Spectrophotometer to

our group.

One last bit a gratitude goes to all Dorsey group

members, past and present, as well as the friends I have made

here. Without all of these people, many of the great times I

have had would have never been possible and I likely would

have lost my mind sometime before the writing of this

dissertation. I am most surely going to miss many pastimes,

such as the "Freeway to Dubs," Friday wild-card rock-blocks

on Rock 104, Saturday morning basketball at Idylwood,

pre-football tailgating, the University Golf Club, NCAA

tournament "bold calls," Cedar River crab dinners, rotisserie

baseball with the GATOR league, getting "huge" at "Let's Get

Huge," Ashley's jumbo margueritas, going to the beach, the

Cinema N' Drafthouse, numerous parties, and most important,

Friday Krystal-action!!!


ACKNOWLEDGMENTS. ............... ......................... iii

LIST OF TABLES .................. ...... ................. vii

LIST OF FIGURES ............................. .............. ix

ABSTRACT ............................................... xiii


I INTRODUCTION.......................................... 1

Solvatochromism.............. ...................... 1
Solvatochromism and Chromatography ................13
Chromatographic Estimations of Lipophilicity......19
Alumina.................. .....................28
This Work .............. ........ ..................36

HOL-WATER BINARY MOBILE PHASES ....................38

Background ............ .......................... 38
Experimental....................................... 41
Solvatochromic measurements.....................41
Retention measurements...........................42
Linear regression ............... ............... 43
Results and Discussion. ................ ............43
Solvatochromic polarity measurements ............43
Chromatographic retention measurements..........50

ET-30 SOLVATOCHROMISM. .............................69

Background ............ .......................... 69
Experimental ...... ................................ 71
Solvatochromic measurements .....................71
Retention measurements ......................... 72
Calculations. .................. ...... .......... 72

Results and Discussion ............................ 74
Solvatochromic polarity measurements for
neutral electrolyte solutions.................74
Estimation of log k', by extrapolation methods..79
Procedures for the estimation of log k'w........ 92


Background. ....................................... 105
Experimental .... ................... .............. 109
Surface polarity measurements..................109
Chromatographic retention measurements .........110
Results and Discussion ........................... 111
ET-30 spectra on alumina........................ 111
ET(30) surface polarity and pretreatment
conditions.................................... 120
ET(30) surface polarity and chromatographic
retention....... ........................... ..126

V CONCLUSIONS. ............ ....................... .131

Summary of Research. ............................. 131
Suggestions For Future Research ..................137
Lipophilicity estimations.......................137
Retention mechanisms ...........................144
Examination of alumina .........................147



III ................ .... ................... ....... 159

III .. ........................... ....... .......... 179


REFERENCES ................. .. ............. .............. 231

BIOGRAPHICAL SKETCH. ............ ........... ............ 244


Tab'0 Pace

2-1 The slopes of the log k' versus ET(30) plots
and Snyder S values for the homologous sol-
vents and acetongtrile on an Altex Ultrasphere
ODS column at 30 C.................................54

2-2 The slopes of the log k' versus ET(30) plots
and Snyder S values for the homologous sol-
vepts and acetonitrile on Zorbax TMS column at
30 C ....................................... ........64

3-1 Experimental conditions for reversed-phase
retention data taken from this work and the
literature for log k'w study ......................73

3-2 The dependence of ET(30) polarity on the con-
centration of electrolyte in neutral aqueous-
organic mixtures ................... .............. 75

3-3 Comparative figures-of-merit for log k' study
based on linear regression for both the ET(30)
and % organic models .............................. 82

3-4 Comparative figures-of-merit for log k'w study
based on linear regression for the ET(30) and %
models using methanol as modifier and second-
order polynomial regression for the % model
using ethanol and acetonitrile as modifiers
(only % model polynomial data shown for r and
RCI) ............................................. 90

3-5 Results of the correlation analysis of log k'w
versus log Pow using different binary hydro-
organic mobile phases for the ET(30) and %
models. ........................................... 94

3-6 The change in the relative confidence interval
(RCI) about log k'w as a function of the span-
ning of log k' values toward pure water. Esti-
mations were done with methanol-water mobile
phases for methylamino-5-dimethylamino-l-sul-
fonate (retention data of Lehtonen, 1984) .........98

3-7 The change in the relative confidence interval
(RCI) about log k'w as a function of the move-
ment of the ordinate centroid from pure water.
Estimations were done with methanol-water mo-
bile phases for methylamino-5-dimethylamino-l-
sulfonate (retention data of Lehtonen, 1984).
Four values of log k' are taken in each %
range, with each log k' taken every 10% organ-
ic ................. ............ ..... .......... 99

4-1 Regression coefficients for the linear corre-
lation between the logarithm of the capacity
factor and the ET(30) surface polarity of
chi-alumina using chloroform in hexane mobile
phases. ................... ...................... 128




1-1 The structure of the solvatochromic probe ET-30.....5

1-2 The structure of the solvatochromic probe pyrene.....9

1-3 The structure of the solvatochromic probe Nile
Red ... .. .................. .......... .............11

1-4 Chemical abstracts services search (January, 1989)
of papers dealing with chromatographic techniques
and physico-chemical properties of solutes..........21

1-5 The decomposition sequence of aluminum hydroxides
(transformation path a = pressure >1 atm, moist
air, heating rate >10C/min, particle size >100
micrometer; path b = 1 atm, dry air, <10C/min,
<10 pm) (taken from Wefers and Bell, 1972)..........30

1-6 The relationship between Snyder's chromatographic
activity parameter, a, and the weight percentage
of water added to X-alumina (data taken from
Snyder, 1968) .................................... 35

2-1 The ET(30) polarity of binary ethanol-water mix-
tures as a function of the volume percent of
ethanol in water.................................. 44

2-2 The ET(30) polarity of binary n-propanol-water
mixtures as a function of the volume percent of
n-propanol in water .................................45

2-3 The ET(30) polarity of binary ethanol-water mix-
tures as a function of the mole fraction of
ethanol in water. ......... ....................... 46

2-4 The ET(30) polarity of binary n-propanol-water
mixtures as a function of the mole fraction of
n-propanol in water.. ............................... 47

2-5 The retention of naphthalene on an Ultrasphere
ODS column at 30C as a function of the ET(30)
polarity of binary hydroorganic mobile phases.......52

2-6 The retention of benzylamine on an Ultrasphere
ODS column at 300C as a function of the ET(30)

polarity of binary hydroorganic mobile phases
ranging from 10 to 50% (v/v) n-propanol in water....59

2-7 The relationship between the retention behavior
of benzylamine (using the ET(30) model) on an
Ultrasphere ODS column at 300C and the carbon
number of the modifier alcohol in a binary hy-
droorganic mobile phase .............................60

2-8 The retention of naphthalene on a Zorbax TMS
column at 300C as a function of the ET(30) po-
larity of binary hydroorganic mobile phases .........63

2-9 The convergence of ET(30) retention plots for
naphthalene to the ET(30) polarity of pure water
(63.1 kcal/mole) ............. ...................... 66

2-10 The non-convergence of % organic retention
plots for naphthalene to 0% (v/v) organic mod-
ifier..................... .......................... 67

3-1 The ET(30) polarity change as a function of the
volume % of organic modifier in a mixture with
pH 7.4 0.02M MOPS/0.2%(v/v) n-decylamine buffer.....77

3-2 The ET(30) polarity change as a function of the
volume % of organic modifier in a mixture with
pH 7.4 66.6mM phosphate buffer ......................78

3-3 Frequency histogram of the distribution of lin-
ear/linear intersections between plots of log k'
versus % methanol and log k' versus % acetoni-
trile or ethanol.. ............... .................. .85

3-4 Frequency histogram of the distribution of lin-
ear/linear intersections between plots of log k'
versus ET(30) for methanol-water mixtures and
log k' versus ET(30) for acetonitrile-water or
ethanol-water mixtures. ................ ............ 86

3-5 Frequency histogram of the distribution of in-
tersections between linear plots of log k' ver-
sus % methanol and polynomial plots of log k'
versus % acetonitrile or ethanol....................91

3-6 Correlation between ET(30) estimated log k'
and octanol-water partition coefficients (Hansch
and Leo, 1979) (log k', estimations done using
retention data of Schoenmakers et al. (1981)) .......95

3-7 Relationship between log k' and % (v/v) aceto-
nitrile in water for the retention data of Baty
and Sharp (1988) ............. ..................... 101

3-8 Relationship between log k' and the ET(30) po-
larity of acetonitrile-water mixtures for the
retention data of Baty and Sharp (1988) ............102

4-1 The UV-visible diffuse reflectance spectrum of
ET-30 on z-alumina (330 mg ET-30 added per kg
alumina) ........................................... 112

4-2 The adsorption isotherm for ET-30 onto activity
grade I X-alumina using an acetonitrile carrier
solvent ............................ ..... ........ 114

4-3 Overlayed UV-visible diffuse reflectance spectra
of ET-30 on X-alumina at concentrations of 95,
191, 286, 379, 474 and 572 mg ET-30 added per kg
alumina. .......... ....... ............. ............ 116

4-4 The FTIR spectrum of X-alumina before treatment
with acetonitrile. ............ .................... 117

4-5 The FTIR spectrum of activity grade I X-alumi-
na after treatment with acetonitrile...............118

4-6 The relationship between the ET(30) polarity of
the surface of X-alumina and the activation
temperature (alumina was used directly from the
commercial container) .............................. 122

4-7 The relationship between the ET(30) polarity of
the surface of X-alumina and the activation
temperature (alumina was saturated with water
for 24 hours and vacuum-filter dried before use)...123

4-8 The relationship between the ET(30) polarity of
the surface of X-alumina and the weight % of
water added. ....................................... 125

4-9 The relationship between the logarithm of the
capacity factor of 3,5-dinitroaniline and the
ET(30) polarity of the surface of X-alumina.
The polarity was calculated by equation 4-1 from
the % (w/w) of water added to the alumina..........127

5-1 The correlation between the EF(NR) polarity scale
and the ET(30) polarity scale for pure solvents
ranging from chloroform to water ...................138

5-2 The correlation between the EF(NR) polarity scale
and the ET(30) polarity scale for binary aqueous-
organic mixtures. The organic modifiers were
methanol and acetonitrile and aqueous portions
consisted of pure water, 0.01 M citrate buffer,
0.01 M MOPS buffer and 66.6 mM phosphate buffer.
All buffers were at a pH of 7.4. ....................139

5-3 The change of EF(NR) polarity as a function of
the concentration of citric acid in a buffer of
pH 3 ................... .............. .............141

5-4 The structure of the solvatochromic probe 4AMP.....142

5-5 Fluorescence spectrum of Nile Red adsorbed onto
the surface of ODS silica using a mobile phase
of 50% methanol in water. Excitation was at 550
nm and the emission collected at a resolution of
1 nm ................. ............... ............ 146

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy




August, 1989

Chair: John G. Dorsey
Major Department: Chemistry

The phenomenon of solvatochromism has been used to

investigate the mechanism of retention in reversed-phase

liquid chromatography (RPLC), improve the estimation of

solute lipophilicity by chromatographic methods and develop a

convenient and continuous scale for examining changes in the

surface activity of alumina. The solvatochromic method used

in this work is the popular ET(30) scale, which is based on

the visible absorption of the molecule 2,6-diphenyl-4-

(2,4,6-triphenyl-N-pyridinio)phenolate (ET-30).

The dynamics of the solvation layer of the RPLC

stationary phase have been studied from the perspective of

the ET(30) polarity of the mobile phase using an homologous

series of normal alcohols (methanol, ethanol, n-propanol) as

the modifiers. The results imply that a systematic change in

the extent of the solvation of an octadecyl-derivatized

stationary phase (C18) occurs with respect to the size of the


organic modifier, but this relationship exists only under

specific eluent conditions. Similar work with a

trimethyl-derivatized phase (Cl) provided no conclusive

judgments, however, because of unpredictable chemistry at the

Cl/mobile phase interface.

The estimation of the RPLC lipophilicity parameter log

k'w using solvatochromism has been examined. It was found

that plots of log k' versus ET(30) for solutes using

different organic modifiers intersected near the ET(30)

polarity of pure water and could prove to be useful

predictors of log k'w. Figures-of-merit (FOM) were

calculated for 204 sets of retention data and results showed

the ET(30) model is reliable for the estimation of log k'w

because of excellent statistical confidence and an accurate

description of the RPLC partitioning process. Correlating

ET(30)-estimated log k'w values to octanol-water partition

coefficients (log Pow) expressed a very sensitive scale of

estimating lipophilicity.

The ET(30) polarity scale has also been used to examine

the surface polarity of alumina with the intent of developing

a convenient and continuous scale to characterize the surface

activity of industrial adsorbents and catalysts. UV-visible

diffuse reflectance spectra and adsorption isotherms for the

dye on X-alumina were obtained. Surface polarity was found

to vary with the amount of water deactivation as well as

thermal activation of the alumina. Normal-phase retention


measurements showed a relationship between solute retention

and pre-chromatographic surface polarity.



The phenomenon of solvatochromism was first labeled in

1922 during studies of the ultraviolet-visible absorption

properties of triphenylcarbinols and dibenzalketones in

various solvents (Hantzsch, 1922). It could be inferred that

the term stems from chrome, meaning color, and solvato,

meaning solvent. An operational definition is the change in

the electro-magnetic (E-M) spectrum of a probe molecule with

respect to changes in the solvent or going from the gas-phase

to solution (Reichardt, 1988). This change in the E-M

spectrum could be in the form of intensity, width or

position. When the intensity of a band increases in

magnitude going from one solvent to another, it is called a

hyperchromic effect and likewise an intensity decrease is

called a hypochromic effect. For movement of spectral

maxima, a bathochromic or red shift is the shifting to longer

wavelengths and a hypsochromic or blue shift is toward

shorter wavelengths.

Comparisons can be made between the polarities of

different solvents and changes in a probe spectrum. Polarity

can be defined as the sum of all molecular properties

responsible for all interaction forces between solvent and

solute molecules (Kovats, 1968). If a probe is sensitive to

certain intermolecular interactions in solution, the solvent

contribution to the overall chemical environment can be

determined. When a probe spectrum undergoes a bathochromic

shift with increasing solvent polarity (i.e., if going from

hexane to water) it is classified as experiencing positive

solvatochromism and conversely a hypsochromic shift with

increasing polarity is categorized as negative


Differences in molecular spectra between solvents arise

because of variations in the types of intermolecular

interactions present in each solvent (Suzuki, 1967; Reichardt

1988). The extent of the spectral difference depends on the

strength of the interaction between the solute and solvent so

that the weaker the interaction, the smaller the spectral

change. Interactions occurring in solution are usually a

result of electrostatic (ion-dipole, dipole-dipole,

dipole-induced-dipole), dispersion, hydrogen-bonding,

charge-transfer and repulsive forces and these forces

influence the energy of the excited state of a molecule.

Intermolecular forces can be further broken down into van der

Waals forces, which are physical and nonspecific, and

chemical forces, which are specific for particular molecular

functionalities. Dispersion, electrostatic and repulsive

forces can be considered physical, while hydrogen-bonding and

charge-transfer are chemical forces. Specific interactions

are generally stronger because their energy of association is

greater. For example, hydrogen bonds have energies ranging

from 5 to 25 kJ/mole (Brown and LeMay, 1981) while attractive

van der Waals forces are less than 1 to 2 kJ/mole (Kaliszan,

1987b). Nonpolar and nonpolarizable solvents such as

saturated hydrocarbons primarily inflict weak, nonspecific

dispersion forces on solutes to produce spectra similar to

those taken in the gas phase. Polar solvents like water and

methanol include many complicated and specific intermolecular

forces and yield spectra broader and less intense than in the

vapor state (Mataga and Kubota, 1970).

Many types of solvatochromic polarity techniques exist

using ultraviolet-visible (UV-VIS), infrared (IR),

fluorescence, electron spin resonance (ESR) and nuclear

magnetic resonance (NMR) spectroscopies (Reichardt, 1982;

Pytela, 1988). The most popular method is measurement of the

UV-VIS molar transition energy, ET, of an

interaction-specific probe. The ET of a UV-VIS band is a

function of the dipolar symmetries of the ground and excited

states (Kosower, 1961). During solvation when the dipoles of

the solute and solvent molecules align with each other, the

energy separation between the ground and excited states is

either increased or decreased. When the interactions become

stronger and more complicated, for instance adding

hydrogen-bonding to dispersion interactions, the energy

difference between the two states becomes larger or smaller

even more. The direction of a UV-VIS shift ultimately relies

on the polarizability of the excited state with respect to

the around state (Suzuki, 1967), but solvation effects are

quite complicated and make band shift predictions difficult.

Bayliss and McRae (1954) distinguished four limiting cases

for predicting solvent effects based on the type of

intermolecular interaction and refractive index, but the

discovery of negatively solvatochromic molecules in the late

1950s disproved the theory. Dielectric properties of

solvents can also be useful predictors of polarity but they

often do not correlate with spectral shifts (Drago, 1977).

Because solvent effects on E-M spectra are rather arduous to

define and quantify, solvatochromic methods are referred to

as being empirical scales of solvent polarity.

The most noteworthy of the UV-VIS polarity scales is

ET(30) (Dimroth et al., 1963). It is based on the negatively

solvatochromic charge-transfer complex formed between the

solvent and the molecule 2,6-diphenyl-4-(2,4,6-triphenyl-N-

pyridinio) phenolate (ET-30) (Figure 1-1). ET(30) refers to

the value of the polarity measurement and ET-30 is the

acronym for the compound's name. The polarity of an

individual solvent is computed by the equation:

ET(30) (kcal/mole) = 28592/xmax(nm) (1-1)

where Xmax is the maximum absorption wavelength and 28592 is

a product of the speed of light, Planck's constant and

Avagadro's number. ET-30 possesses a large permanent dipole

Figure 1-1. The structure of the solvatochromic probe ET-30.

moment (pD = 15D) and a 44 n-electron network that makes it

sensitive to solvent dipolarity/polarizability changes. The

phenolate functional group is also an excellent hydrogen bond

acceptor. Combination of these characteristics produces the

largest solvatochromic shift known to date, going from 453 nm

in water to 810 nm in diphenyl ether (Reichardt, 1988) and

yields an extremely sensitive method of characterizing

solvent polarity. This sensitivity was originally exploited

to decipher the effects of the solvent on chemical reaction

rates by the relationship (Reichardt, 1988):

k = ko + A[ET(30)] (1-2)

Here k stands for a rate constant, ko is the y-intercept and

A the slope of the regression between k and ET(30) An

interesting aspect of ET-30 is the charge-transfer complex it

forms with the solvent absorbs in the visible region of the

E-M spectrum and proper adjustment of the solvent polarity

can express all seven principle colors. A solution of the dye

is red in methanol, violet in ethanol, sky-blue in

acetonitrile and green in acetone.

Because the ET(30) scale does not fit into the framework

of SI units, the ETN scale was devised. The ETN scale is the

normalization of the ET(30) scale relative to the nonpolar

solvent tetramethylsilane by the equation:

ETN = [ET(30)i ET(30)TMS]/[ET(30)w-ET(30)TMS] (1-3)

where ET(30)i denotes the polarity of the solvent of

interest, ET(30)w the polarity of water and ET(30)TMS the

polarity of tetramethylsilane. This manipulation creates a

range of polarities going from 0.00 in TMS to 1.00 in water.

A major disadvantage of performing this normalization is the

large sensitivity to polarity changes is lost. In all

correlations with polarity, whatever the scale, it is best to

use the original solvatochromic parameters (Abboud et al.,


Three other UV-VIS scales are based on a principle known

as solvatochromic comparison. The 7* scale (Kamlet et al.,

1977) is in units of absorption wave numbers and the polarity

of an individual solvent is denoted by the average of the

individual absorption wave numbers of nine primary

nitroaromatic compounds in that solvent. The a (Taft and

Kamlet, 1976) and P (Kamlet and Taft, 1976) scales calculate

specific solvent polarity in terms of the absorption wave

number difference between two standard compounds. The wave

number difference between 4-nitroaniline and

N,N-diethylaniline determines a and 3 is found by the

difference between 4-nitroanisole and ET-30. A disadvantage

of these statistical-based scales is they are ill-defined and

do not express a discrete intermolecular event. Kamlet and

co-workers have published forty-one papers (Kamlet et al.,

1987) on these scales with each paper being an adjustment to

"constant" solvent parameters. Also, in order to expand the

database of each scale to new solvents, numerous spectral

measurements must be made for a single solvent.

Two fluorescent polarity scales will be mentioned.

Solvent effects on fluorescence spectra occur when the

dipolarity of the excited state of the molecule with respect

to the ground state changes upon interaction with the solvent

(Abe, 1988). The Py scale (Nakajima, 1971) is based on

changes in the intensity of UV emission bands for pyrene

(Figure 1-2). Pyrene possesses 5 characteristic fluorescence

bands between 370 and 400 nm labeled in progressive order

(the highest energy band labeled I and the lowest energy band

V). By taking the ratio of the intensities of the first and

third bands the Py polarity of the solvent can be obtained:

Py = I/IIII (1-4)

Emission bands I and III show intensity changes with polarity

changes but the exact mechanism of how this process works is

not well understood. Disadvantages of Py are the difficulty

in precisely measuring small differences in emission

intensity (Street and Acree, 1986) and a lack of functional

groups sensitive to hydrogen bonding interactions. Despite

these pitfalls, pyrene has been a useful probe for studying

solutions of organized media (Kalanasundaram, 1978).

Another fluorescent scale is the positively

solvatochromic EF(NR) scale (Klessinger and Ldttke, 1966).

This scale measures the visible energy of fluorescence (in

Figure 1-2. The structure of the solvatochromic probe

kcal/mole) of the molecule 9-diethylamino-5H-

benzo[a]phenoxazine-5-one (Nile Red) (Figure 1-3). Nile Red

forms a charge-transfer complex with the solvent similar to

ET-30 and can visually express polarity changes like ET-30.

The solution is deep blue in ethylene glycol, violet in

methanol, red in acetonitrile and orange in 1,4-dioxane.

Because the phenoxazone group is not as strong a hydrogen

bond acceptor as the phenoxide group of ET-30 and the entire

molecule has only half the i-electrons, Nile Red is not as

sensitive a solvent polarity probe as ET-30. Nile Red has,

however, been useful as a fluorescent probe of hydrophobic

proteins (Sackett and Wolff, 1987) and a stain for

intracellular lipids (Greenspan and Fowler, 1985) because of

its strong fluorescent intensity.

All of the previously mentioned scales are called

single-parameter polarity scales and can be used individually

or in conjunction with others in a multi-parameter scale. A

single-parameter scale bases a polarity measurement on one

type of reference probe compound. It is believed that by

linearly combining scales sensitive to different

interactions, the total polarity of a solvent can be better

estimated. One such method has been developed that combines

1*, the measure of solvent dipolarity/polarizability, a,

which estimates the hydrogen bond acidity of a solvent, and

p, the solvent hydrogen bond basicity descriptor (Kamlet et

al., 1987). This type of scale is based on the assumption of

linear solvation relationships (LSR) and describes a property


Figure 1-3. The structure of the solvatochromic probe Nile

of the solvent by the equation:

XYZ = XYZo + AlK* + BC + CP (1-5)

where XYZ is a measurable physical quantity dependent on

solvent polarity, XYZo is the intercept and A, B and C are

the respective regression coefficients for the multiple

relationship. Another multi-parameter approach associates

ET(30) and the donor number of the solvent, DN (Krygowski and

Fawcett, 1975)

XYZ = XYZo + A[ET(30)] + B*DN (1-6)

where again A and B are the regression coefficients. The

donor number is a nuclear magnetic resonance method that

measures the hydrogen bond basicity of a solvent (Mayer et

al., 1975). Unification of the two parameters into one

equation clarifies the primary solvent interactions.

Two problems exist with multi-parameter polarity scales.

One is, as stated earlier for the work of Kamlet et

al.(1987), the parameters TK*, a and P are ill-defined,

non-discrete and rigorous to determine. Another problem with

any multi-parameter scale is the illusion that it discloses

"fundamental" information about the effects of solvents. An

interesting perspective on the "meaning" of single- and

multi-parameter solvent scales is found in a paper by

Sjostrom and Wold (1981) and a reply by Kamlet and Taft

(1985). Sjbstrom and Wold argue that instead of the

classical interpretation of LSRs expressing a combination of

"fundamental" effects, they should be viewed strictly as

locally valid linearizations of complicated relationships.

It is perhaps best to say, whether with single- or

multi-parameter scales, they are merely providing a

convenient linearization of very complicated phenomena.

Since both methods are empirical, the less rigorous

single-parameter approach should be more desirable to use

because it is less complicated, gives comparable results and

is easier to expand to new solvents.

Solvatochromism and Chromatoorachy

Many theories exist examining the mechanisms of retention

in most forms of chromatography (Dill, 1987; Horvath et al.,

1976; Martire and Boehm, 1983; Jaroniec and Martire, 1986),

but it is agreed in all that intermolecular interactions are

predominant driving forces in these processes. While bulk

physical parameters such as temperature, viscosity and the

nature of the support materials are important variables,

deciphering the effects of microscopic phenomena appears to

be the link to complete understanding of these complicated

events. Employing empirical solvatochromic methods for

studying fundamental processes in chromatography can aid in

the interpretation of the role of key intermolecular

interactions. This enhanced understanding can improve

separation optimization and physical property estimations.

Chromatographic systems have been studied from the

viewpoint of both the mobile and stationary phases with the

most extensive work being done on reversed-phase liquid

chromatography (RPLC). Lochmuller et al. (1983; 1984; 1985)

have observed the fluorescence solvatochromism of pyrene

chemically bound to RPLC surfaces. This work of Lochmuller's

presented an idea known as the "Micro-Droplet" theory which

states that surface silanols on microparticulate silica do

not exist homogeneously, but rather in concentrated clusters.

An emission band attributed to an "eximer-like" complex was

their evidence for high density silanol pools. The complex

formed at low % carbon and was believed to result from

bimolecular ground state associations of the

pyrene-derivatives that were within a critical interaction

distance necessary for eximer formation. Since the complex

occurred at low carbon load, they rationalized the silanols

involved in bonding were in close proximity to each other.

In contrast to chemically bonding pyrene or other probes

to silica, other projects have physically adsorbed the probe

to alkyl-derivatized silica and observed its spectrum. This

allows the examination of a surface actually used in practice

instead of one specially synthesized for the probe. Carr and

Harris (1986) sorbed pyrene to RPLC silica under

RPLC-mimicking conditions. A flow cell was constructed using

a quartz column enclosed in a stainless steel jacket that

would allow the derivatized-silica to be packed under

high-pressure. By pumping pyrene-containing mobile phase

through the column, Py values for the surface-sorbed species

were taken at different organic compositions. They found an

inverse relationship between the polarity of the C18 surface

and that of the mobile phase over large ranges of solvent

composition. Problems with light-scattering by the silica

reduced the resolution of the experiment, however.

Two problems exist with the pyrene adsorption approach to

studying RPLC surfaces. The first one is that only limited

ranges of organic modifier can be used. For the case of high

% organic, too much pyrene remains in the mobile phase and

dwarfs the signal from the surface. Contrary to high %, at

low % organic the surface can become overloaded with probe to

produce self-association effects. A second problem is that

the Py polarity scale is limited in differentiating the

polarities of different solvents. For example, the Py

polarity of acetonitrile and water are the same (Stahlberg

and Almgren, 1985). What is needed is a probe that has a

strong affinity for the hydrophobic surface, has an easily

detectable signal and is sensitive to more intermolecular

interactions than just dispersion and polarizability effects

(i.e., hydrogen bonding).

The most significant work exploiting solvatochromism as a

tool for studying chromatographic processes has been done

with the popular UV-VIS-based polarity scales.

Chromatographic applications of these techniques have

recently been reviewed (Dorsey, 1987). Pioneering work

describing mobile phase polarity effects on retention in RPLC

was done by Johnson et al. (1986) using the ET(30) scale.

The "strength" of the mobile phase was described by the bulk

polarity measured independently of the column, in contrast to

the commonly used volume percent organic modifier, % (Snyder

et al., 1979). Dorsey and Johnson (1987) presented universal

equations for determining the ET(30) polarity of aqueous

methanol, acetonitrile and tetrahydrofuran binary mobile

phases from the original volume % of the organic modifier.

Through the analysis of 332 sets of chromatographic

retention, the logarithm of the capacity factor for a solute,

log k', was found to be better correlated to ET(30) by the


log k' = m[ET(30)] + b (1-7)

than the volume % of organic modifier in the eluent by the


log k' = S% + log k'w (1-8)

where m and S are the slopes and b and log k', are the

intercepts. Log k', is also called the capacity factor using

pure water as mobile phase. The slope and intercept of log

k' versus ET(30) plots increase as a function of the size of

the solute. Comparisons of solute methylene selectivity, log

a, and mobile phase ET(30) polarity were also examined

(Johnson, 1986) but no significant improvements over the %

strength model were discovered.

RPLC retention has also been modeled by the Kamlet and

Taft multi-parameter solvatochromic comparison method.

Instead of relating solute retention to the polarity of the

mobile phase, retention was predicted by the polarity of the

solute itself from the equation (Sadek et al., 1985):

log k' = log k'0 + mV2/100 + s7*2 + bP2 (1-9)

where V2 is the molar volume of the solute. This approach

was designed to describe the retention process from the

viewpoint of the characteristic chemical interactions of the

solute on any column under any mobile phase condition. The

molar volume term accounts for the size of the cavity taken

up by the solute in the mobile phase. The 7* and 0 terms

account for the dominant chemical interactions contributed by

the solute. A training set of calculated and experimentally

measured log k' values was constructed for a wide range of

solute types under specific chromatographic conditions.

While Sadek et al. (1985) found reasonable agreement

between calculated and experimental log k' values, this

method has many problems. Training sets must be constructed

for any new column and this would be a tremendous amount of

work to do whenever a new column is purchased or made. The

values of it* and P are not easily accessible for many

compounds and cannot be measured for solutes that are solids

at room temperature. Equation 1-9 also does not account for

the hydrogen bond acidity of the solute which should not be

neglected since solutes containing hydroxyl substituents can

donate hydrogen bonds.

To counter the above-mentioned setback of overlooked

solute hydrogen bond acidity, Carr et al. (1986) and Park et

al. (1988) added an a term to equation 1-9 to produce

log k' = log k'0 + mV2/100 + sc*2 + a12 + bP2 (1-10)

This adjusted view of retention was used to observe

dependence on the mobile and stationary phases and

temperature. Even though a more thorough description was

made, the method is still too tedious and complicated to be

of any practical use.

Cheong and Carr (1988) have begun work attempting to

describe retention from the standpoint of the Kamlet-Taft

multi-parameter polarity of the mobile phase. This has

probably been taken on as a result of the success of the

single-parameter ET(30) method. 7t* and 3 parameters were

measured for binary mixtures of water with methanol,

2-propanol and tetrahydrofuran but no multiple correlations

with retention data have yet been reported.

Supercritical fluid chromatography (SFC) systems have

also been explored with UV-VIS scales. Yonker et al. (1986)

measured the K* polarity of supercritical carbon dioxide as a

function of the density of the supercritical fluid and

related that polarity to SFC retention. A biphasic character

in log k' versus 7* plots for nitrobenzene derivatives was

observed. This behavior was attributed to changes in the

solvating ability of the supercritical carbon dioxide at high

and low densities.

Chromatoqraphic Estimations of LiDoohilicity

A growing application of reversed-phase liquid

chromatography (RPLC) is the determination of the

physico-chemical properties of chemical compounds, also known

as quantitative structure-retention relationships (QSRR).

QSRR constitutes a large subset of quantitative

structure-activity relationships (QSAR) and is the

correlation of the chromatographic retention of a compound

with a specific molecular property. One such

physico-chemical property that has many uses in the

environmental and biological sciences is lipophilicity, a

descriptor of the hydrophobic partitioning character of a

compound (Niralakhandan and Speece, 1988; Hansch, 1978). A

recent Chemical Abstracts Service (CAS) database search done

at the University of Florida revealed that since 1975 the

number of papers published regarding chromatography and QSAR

has increased steadily. In 1975, the percentage of those

papers concerning HPLC was approximately 12% and as of 1988

that portion has expanded to about 25%. These CAS trends are

illustrated in Figure 1-4.

The lipophilic/hydrophobic character of compounds has

been examined by experimental and theoretical

non-chromatographic means. The first experimental

undertaking of measuring hydrophobicity was the development

of the static log P shake-flask system (Hansch, 1969). The

log P method calculates the partition coefficient, P, for a

compound distributing between immisible organic and aqueous

phases by the equation:

P = [Organicl/[Water] (1-11)

where [Organic] is the equilibrium concentration of the

analyte in the organic phase and [Water] is the analyte

concentration in an aqueous phase. This partition

coefficient is then correlated to some form of biologically

measured activity or a structural change in the molecule.

The most common organic phase referred to is n-octanol

because of its similarity to lipids in structure (hydrophobic

tail and polar head-group) and low solubility in water.

Other organic phases such as cyclohexane, diethyl ether,

olive oil and organic membranes have also been used but there

is no standard approach. Shake-flask methods of

lipophilicity measurement are disadvantageous because they

exhibit poor reproducibility, are slow, require large amounts

o All Techniques
* HPLC only

1978 1981 1984 1987

Figure 1-4. Chemical abstracts services search (January,
1989) of papers dealing with chromatographic
techniques and physico-chemical properties of

of Papers




of pure sample and need two separate analyses done for every

individual trial.

To avoid tedious lab work for new solutes, dry-lab

methods governed by the assumption of linear free energy

relationships (LFER) have been devised to calculate

octanol-water partition coefficients (log Pow) with only a

knowledge of the compound's structure. LFERs computate

lipophilicity by adding together hydrophobicity values

associated with structural portions of a compound. These

methods include Hansch and Leo's t hydrophobicity parameters

(1979), Rekker's fragmental f constants (1977) and Randic's

molecular connectivity index, X (1975). Octanol-water

partition coefficients have also been calculated by the

solvatochromic comparison method (Kamlet et al., 1988) and

statistical thermodynamic approaches (Kasai, 1988; Schantz

and Martire, 1987). All of these methods are continually

being updated to account for deviations arising from intra-

and intermolecular interactions and this can make

standardization almost impossible. The "rules" for

assignment of hydrophobicity identities have become so

complicated that microcomputers are necessary for

calculations, but no commercial software is available.

Recent reviews have covered research studying the use of

RPLC retention data as a measure of lipophilicity (Braumann,

1986; Carney, 1985; Kaliszan, 1981; 1986; 1987a; 1987b) It

was first mentioned in the landmark paper by Martin (1950)

that a substituent changes the partition coefficient of a

solute by a factor dependent on the nature of the substituent

and the mobile and stationary phases. Fifteen years later,

Iwasa et al. (1965) first conjectured the usefulness of

chromatographic data for QSAR. Melander and Horvath (1980)

noted that relating RPLC retention data to log Pow values is

theoretically valid. The major advantages of using RPLC over

shake-flask methods are that it is faster, shows a larger

dynamic range than log P, is more convenient to perform

experimentally, is extremely reproducible and is automatable.

Gas, paper and thin-layer chromatographies can also provide

partitioning data but all three are limited with regard to

applicable samples. RPLC also works well over shake-flasks

because precise peak height or area quantitation is not

necessary and the sample need not be 100% pure.

Lipophilicity determinations by RPLC are done by

measuring the capacity factor of the solute. The log of the

capacity factor, log k', is described by the relationship

(Scott and Kucera, 1977):

log k' = log K + log D (1-12)

and experimentally measured by the equation:

k' = (tr to)/to (1-13)

where K is the thermodynamic distribution coefficient, D is

the ratio of the volumes of the stationary and mobile phases

for the chromatographic system (Vs/Vm), tr is the retention

time for the solute and to is the dead time of the system. P

is assumed to be proportioanl to K and since k' is directly

proportional to K, capacity factors have been correlated with

P by the equation:

log k' = m[log P] + b (1-14)

where m and b are the respective slope and intercept for the

linear regression. Since the type of interactions present in

RPLC, static shake-flask and real biological systems are

quite different, chromatographic and static techniques are

merely estimations of hydrophobicity, not direct measures.

Various approaches have been devised to obtain RPLC

capacity factors for lipophilicity information. One approach

attempts to mimic octanol-water shake-flask systems by

saturating both the mobile and stationary phases with

n-octanol. Mirlees et al. (1976) obtained excellent

correlations between log k' and log Pow with a system

consisting of a Cl-silanized glass tube coated with n-octanol

and octanol-saturated buffer as eluent. By using an

octanol-coated commercial RPLC column and an

octanol-saturated buffer as solvent, Unger and Chiang (1981)

also produced good log k'/log Pow correlations. Two problems

with this experiment are the instability of the octanol

coating on the surface and a limited range of determinable

log Po, values. Very hydrophobic solutes cannot elute from

this system because of the poor thermodynamic and kinetic

characteristics of solute distribution across the

liquid/liquid interface region. No modifiers could be added

to the mobile phase to speed up the analysis, however, or

else the mimicking of the shake-flask experiment was lost.

Another plan has been to perform a conventional RPLC

analysis and correlate isocratic retention to lipophility.

Brodsky and Ballschmitter (1988) calculated retention indices

for polychlorobiphenyls (PCB's) on a C18 column using a 55%

acetonitrile in water mobile phase. The indices were

favorably compared with the number of chlorine atoms, log Pow

and water solubility. Instead of a silica-based C18 column,

Kaliszan et al. (1988) adsorbed polybutadiene onto y-alumina

and eluted analytes with 50% methanol in aqueous buffer.

While this alumina-based column provides a wide range of pH

to work with, it is not as rugged as conventional RPLC

columns and has inherent reproducibility problems with the

polymerization of the butadiene onto the alumina.

A serious problem in general with isocratic log k'

correlations is the effect of the organic modifier on the

hydrophobic expression of the solute. Miyake et al. (1988)

noticed distinct effects on log k' for low (20%) in

comparison to high (85%) compositions of methanol in water.

Peak inversion is common for different solutes at unspecified

compositions and can lead to erroneous assignments of

lipophilic character. Minick et al. (1988) observed a switch

in the elution order from a C18 column for anisole and

3-(trifluoromethyl)phenol occurring at 45% methanol in water.

The only compounds that cannot invert at different mobile

phase compositions are members of a class of molecules

varying only in the number of substituents, as with the PCB

study, or size, as with methylene analogs of an homologous

series of alkylbenzenes. Specific solvation effects due to

the organic modifier can take place between various solute

classes as well as with the stationary phase surface to

over-complicate the analysis.

An RPLC technique that has recently received much

attention is the parameter log k'w, the logarithm of the

capacity factor using only an aqueous phase as the eluent

(Hammers et al., 1982; Braumann et al., 1987; Gaspari and

Bonati, 1987; Minick et al., 1988). The beauty of using log

k'w is that it is independent of any organic modifier

effects, reflects polar/nonpolar partitioning in a manner

similar to shake-flask methods (Braumann et al., 1983;

Schantz and Martire, 1987) and is dependent on the solute's

structure and polar functionalities (Snyder et al., 1987).

This parameter, however, is difficult to measure directly in

an RPLC experiment because of poor partitioning kinetics

across the stationary phase/mobile phase interface and

prohibitively long retention times. It is also arduous to

both detect and locate a peak centroid because of the skewing

of the peaks.

Log k'w can be estimated, however, by the intercept of

equation 1-8 (Snyder et al., 1979). The retentivity of the

chromatographic system with respect to changes in mobile

phase "strength," denoted here by %, is inferred by the slope

S. It has been suggested that log k', be a direct measure of

lipophilicity for neutral solutes because it minimizes

hydrogen bonding effects (Miyake et al., 1988) and has been

found to be reproducible between C18 columns used (Braumann

et al., 1987). These columns must be from the same

manufacturer, however, since log k'w should vary from column

to column because of bulk silica, bonding density and 0


Despite the utility of log k', as a lipophilicity

descriptor, a problem exists in the fundamental detail of its

estimation. While equation 1-8 holds on a qualitative basis,

curvature exists in these plots (Schoenmakers et al., 1978;

Bor6wko et al., 1987; Jandera, 1984) and could lead to

erroneous extrapolation results. Snyder and Quarry (1987)

proposed a statistical method to counteract the curvature,

but that approach appears to be best applicable to mobile

phase optimization for method development.

Another remedy to this problem of mobile phase strength

characterization for QSRR has been to describe retention data

with the equation (Reymond et al., 1987; Baty and Sharp,


log k' = A%2 + B% + log k', (1-15)

where A and B are the first and second coefficients of the

second-order polynomial regression. While a second order

polynomial will theoretically give a better fit to the data

(Dill, 1987; Schoenmakers et al., 1978), additional

uncertainty in an extrapolation occurs due to the A%2 term

and that uncertainty can only be reduced by measuring extra

data points closer to the point of extrapolation. For some

solutes, though, measuring extra log k' values at lower %

organic mixtures can increase the total estimation time

considerably. Equations 1-8 and 1-15 also only cite changes

in bulk organic modifier composition. They do not, however,

account for any microscopic differences in the chemical

interactions characteristic of those varying composition

eluents nor is there any differentiation made between the

interactions present in eluents using different modifers

(i.e., acetonitrile instead of methanol).


The name "alumina" has been used to describe a broad

range of sustances derived from aluminum hydroxides and

products of their thermal decomposition. Their chemical

nature and final usage are a function of the structure and

temperature response of the hydroxides. Alumina is one of

the largest volume pure inorganic chemicals produced

worldwide with production amounting to roughly 40 million

metric tons per year (Misra, 1986). Final applications range

from pure aluminum metals to ceramics, catalytic supports and

industrial adsorbents.

The type of alumina chemical of interest in

chromatography is activated aluminum oxide. Activated

aluminas are products of the thermal decomposition of the

aluminum hydroxide Gibbsite which is precipitated from the

raw ore Bauxite by the Bayer refining process. Once

isolated, Gibbsite is transformed into useful entities by

varied processes concerning temperature, pressure, atmosphere

and particle size (Wefers and Bell, 1972). Figure 1-5 shows

the thermal decomposition sequence of aluminum hydroxides

from Gibbsite and another hydroxide, Bayerite. The

decomposition products from the hydroxides are usually

labeled with Greek letters to denote differences in their

crystal structures detected by X-ray diffraction analysis and

the amount of hydration.

Low temperature aluminas (X,yand T) are used as catalysts

and catalyst supports as well as solid phases for analytical

and large-scale adsorption liquid chromatography (ALC).

Their surfaces contain a variety of sites at which selective

adsorption of organic compounds can occur. The surface

consists of negative-field oxide ions (0-2, ) ,

positive-field aluminum ions (Al+3), aluminols (Al-OH),

valence holes and adsorbed water (Benesi and Winquist, 1978)

with the composition of the individual sites dependent on the

pretreatment conditions. Adding water to a low temperature

activated alumina sets off an exothermic reaction between the











0 100 200 300 400 500 600 700

800 900 1000 1100 1200


Figure 1-5.

The decomposition sequence of aluminum hydroxides (transformation path a =
pressure >1 atm, moist air, heating rate >10C/min, particle size >100
micrometer; path b = 1 atm, dry air, <10C/min, <10 micrometer) (taken from
Wefers and Bell, 1972).

water and coordinated aluminum and oxygen sites to form

aluminols (Hendriksen et al., 1972). Heating alumina to

3000C or more drives off most of the adsorbed water, leaving

behind aluminols that can be later removed at 800C or

higher. These aluminols have been hypothesized to exist in

five distinct configurations and their existence was verified

by IR spectroscopy (Knizinger and Ratnasmy, 1978) An

activated surface can undergo specific intermolecular

interactions through hydrogen-bonding aluminols and the

positive- and negative-field sites acting as Lewis acids and

bases, respectively. A chromatographic or catalytic surface

possessing unique properties is produced by these various


Liquid chromatographic applications of alumina have been

devised for the separation of various organic molecules.

When used in normal-phase mode (NPLC) with nonpolar solvents

like n-hexane and chloroform, Lewis acid sites can dominate

the retention process and produce large adsorption energies

for strong electron-donating solutes like amines and amides

(Snyder, 1966). Surface hydrogen bonding basicity also

contributes heavily because it was also found that hydrogen

bond acidic compounds such as alcohols, carboxylic acids and

mercaptos expressed the largest energies of adsorption.

For the analysis of basic drugs, the alumina can act as

an ion-exchanger. By using an aqueous buffer as a mobile

phase, Billiet et al. (1985) were able to alter the

amphoteric behavior of the surface to affect the retention of

heroin and opium. The normal isoelectric point of alumina

(pH=9) can be shifted lower by different buffers as well.

Since alumina has a wide pH stability range, it has also been

used for the elution of proteins that also have wide ranges

in isoelectric behavior (Laurent, et al., 1984).

Activated aluminas are useful as active catalysts and as

catalyst supports. Fluorinated alumina has been shown to be

an effective catalyst in reactions of acetylene (Allenger et

al., 1987). The impregnation of a modifier like fluoride

into the aluminum oxide matrix changes the Lewis acidity and

hydrogen bond acidity of the surface to make it more

reactive. Metal catalysts can also be immobilized onto

alumina by impregnation or coprecipitation to improve the

physical stability of the catalyst. An example of this is

the catalytic converter of most automobiles that oxidizes

hydrocarbons into carbon monoxide, carbon dioxide and water

(Misra, 1986). Platinum, palladium and rhodium are

impregnated onto an alumina support for convenience of

construction and have improved thermal and physical stability

on this substrate.

One of the problems associated with using alumina in

catalysis and chromatography is the lack of a continuous and

convenient method for characterizing its surface activity.

Catalysis chemists believe that activity stems from the

concentration of acidic aluminols on the surface and

therefore attempt to quantitate them. The most common

approach is to titrate the surface with a basic probe

molecule like pyridine (Healy et al., 1989). Using infrared

spectroscopy, the intensity of a complex formed between the

pyridine and aluminol is correlated to the amount of pyridine

added and used to calculate the number of active sites.

Another technique quantitates the number of sites from a

high-frequency infrared hydroxyl stretch due to a surface

aluminol (Van Veen, 1988).

Chromatographic methods have also been devised to

characterize alumina activity. Based on classical

thermodynamics, Snyder (1968) described retention in

adsorption liquid chromatography (ALC) by the equation

log Ko = log Va + af(X,S) (1-16)

where Ko is the chromatographic partition coefficient of the

solute, Va is the volume occupied by the active surface, a is

the "activity" of the adsorbent surface and f(X,S) is an

energy function that is said to remain constant for a given

solute in a given solvent. The term f(X,S) was determined to

be the difference in the adsorption energies of the solute

and solvent

f(X,S) = SO AsE0 (1-17)

where S is the adsorption energy of the solute, As is the

surface area of the solute and E0 is the adsorption energy of

a pure solvent, also known as the solvent eluotropic

strength. Snyder's model allowed the forces contributing to

ALC retention to be separated into measurable independent

terms. The surface activity, a,could be used to characterize

the retentivity of the adsorbent as a function of the

activation temperature and the amount of water added to

deactivate it. Figure 1-6 shows the change in the surface

activity as the amount of added water is varied. As the

weight percentage of water exceeds 10%, the activity

calculated by this method becomes constant.

As important as Snyder's theory has been to the

development of adsorption chromatography, many problems exist

that limit its practical utility for determining the activity

of alumina. Many retention measurements must be made and

tedious calculations done to obtain the value of the

partition coefficient. Considerable error could be incurred

if the measured partition coefficient is either too small or

too large. The activity scale is unitless and relative to an

arbitrary activity of 1.0 for a nondeactivated adsorbent.

Also the adsorption energies of the solute and solvent in the

f(X,S) term of equation 1-16 are assumed to remain constant

from adsorbent to adsorbent, but this is improbable since the

nature of the surface is changed by the addition or removal

of hydration. Once f(X,S) is determined on undeactivated

alumina of a = 1.0, it is then used to calculate other values

of a. Not all undeactivated alumina will be the same,

however, since different activation conditions produce

varying activities.





0.6- B

0 3 6 9 12 15
% (w/w) H20/A1203

Figure 1-6.

The relationship between Snyder's
chromatographic activity parameter, a, and the
weight percentage of water added to X-alumina
(data taken from Snyder, 1968).

Another chromatographic method developed by Brockmann and

Schodder (1941) based alumina activity on the elution

behavior of five standard dyes. This procedure creates five

activity categories over the range from 0 to 15% by weight

added water. A sample of alumina is placed into a respective

category depending on which of the five dyes elutes from the

top of the column using a carbon tetrachloride solvent. This

technique, as was the problem with Snyder's method, requires

many hours of tedious sample elution and it is not a

continuous method since 15% is spread over only five

categories. Some of the standard dyes are also not

commercially available and some are known or suspected

carcinogens. Despite these pitfalls, all alumina sold for

column chromatography is still described by its Brockmann


This Work

This dissertation will present and discuss attempts to

use the phenomenon of solvatochromism to improve or simplify

existing methodology regarding retention mechanisms,

estimations of solute lipophilicity and characterization of

alumina surface activity. Chapter II will show how the use

of homologous alcohols and the ET(30) scale has obtained

ancillary information about the dynamics of the RPLC

stationary/mobile phase interface. In Chapter III, the

estimation of the RPLC parameter, log k' has been improved


by using extrapolations of log k' versus ET(30) plots. And

finally, Chapter IV describes a new method of delineating the

activity of chromatographic alumina by determination of the

empirical ET(30) polarity of the surface.



Many approaches have been taken to study the effects of

the mobile phase in reversed-phase liquid chromatography

(RPLC) (Carr et al., 1986; Horvith et al., 1976; Jandera et

al., 1982; Karger et al., 1978; Martire and Boehm, 1983;

Schoenmakers et al., 1982). The most commonly used mobile

phases in RPLC are binary solutions of water with an organic

solvent modifier such as methanol, acetonitrile or

tetrahydrofuran. Retention in RPLC is primarily controlled

by the chromatographic strength of the mobile phase, with the

strength frequently denoted as the volume percent of the

organic modifier in the binary aqueous solution (Gant et al.,

1979). It has been shown both experimentally (Schoenmakers

et al., 1978) and theoretically (Dill, 1987) that a quadratic

function best describes plots of log k' versus percent

organic modifier (equation 1-15).

A model of RPLC retention relating the strength of the

mobile phase to the polarity of the solvent has been

developed (Johnson et al., 1986). An independent examination

of the effect of changing mobile phase polarity on

chromatographic retention was performed using ET(30)

solvatochromic solvent polarity measurements. The regression

analysis of 332 data sets revealed a higher degree of

linearity for the ET(30) model (equation 1-7) over the

percent organic modifier model (equation 1-8). If the ET(30)

polarity is truly a good descriptor of mobile phase strength,

then by correlating retention versus the ET(30) polarity of

the solvent for a given solute on a given column, the

retention behavior of the solute in all solvent systems on

that particular column should be the same. Retention

behavior is symbolized by the slope of equation 1-7 and

defined here as the sensitivity of the change in retention of

a solute to changes in the mobile phase ET(30) polarity.

This is similar to Snyder's "S" value, which is the slope of

equation 1-8 (Snyder et al., 1979). Inspection of the

regression coefficients for 89 of the 332 data sets revealed

that the retention behavior of these systems using equation

1-7 are not normalized as expected. Data taken from the

literature on columns ranging in chain length from C2 to C18

show that for a given solute and column, the methanol slope

was greater than the acetonitrile slope by an average ratio

of 1.4.

One interpretation of this 1.4 methanol-acetonitrile

slope ratio for ET(30) plots is the active role of the

stationary phase in RPLC. If the stationary phase were a

passive entity, as is said to be true by the solvophobic

theory of Horvath et al. (1976), the slopes of log k' versus

ET(30) plots should be the same for a solute in any solvent

system. The ET(30) scale has previously been shown to

accurately measure solution properties, as evidenced by

correlations of ET(30) polarity with reaction rate constants

(Elias et al., 1981) and heats of solution at infinite

dilution (Ilic et al., 1984). Retention is a result of the

free energy change as a solute transfers between the mobile

and stationary phases. As iso-ET(30) values of two mobile

phases suggest that they are energetically equivalent, at

least as seen by the ET-30 molecule, then the different

slopes indicate the solute is experiencing a different

environment in the stationary phase with methanol-water as

compared to acetonitrile-water. While it can be argued that

the local environment of the ET-30 molecule is very different

between methanol and acetonitrile, it will be shown later in

this chapter that slope ratio differences are seen between

methanol and ethanol, where the local environment of the

ET-30 molecule would be more similar. This must primarily

result from differences in the extent of solvation of the

alkyl chains bonded to the silica as the organic modifier is

changed. These differences in the extent of solvation have

also been shown by others (McCormick and Karger, 1980a;

McCormick and Karger, 1980b; Yonker et al., 1982a; Yonker et

al., 1982b).

To further clarify the meaning of the slope of equation

1-7, it was attempted to induce systematic changes in the

retention behavior for a given system by using a homologous

series of organic modifiers. The series chosen was that of

the n-alcohols (methanol, ethanol and n-propanol) because

they are readily available, non-toxic, have low wavelength UV

cutoffs and their distribution behavior in an RPLC system has

been previously characterized (Scott and Simpson, 1980). The

extent of solvation of the stationary phase was expected to

change in direct proportion to the molecular size of the

modifier, resulting in a systematic change in the slopes of

equation 1-7 for each solute. Therefore, the aim of this

study was to provide further evidence that the change in the

slopes of log k' versus ET(30) plots are indeed due to

changes in the nature of the stationary phase and to also

characterize ethanol and n-propanol mobile phases for RPLC by

the ET(30) solvatochromic solvent polarity scale.


Solvatochromic measurements

All solvatochromic measurements were made using ET-30

(Reichardt's Dye) (Aldrich Chemical, Milwaukee, Wisconsin).

Binary solvents were prepared by mixing additive volumes of

ET-30 in pure organic solvent, pure organic solvent and water

to the desired solvent compositions with the final

concentration of ET-30 being approximately 200 mg/L. Samples

were placed into a 1 cm path length quartz cell and spectra

obtained with a Hewlett-Packard (Palo Alto, California) Model

8450A diode array spectrophotometer or an IBM Instruments

(Danbury, Connecticut) Model 9420/9430 UV-Visible

Spectrophotometer. Maximum absorbance wavelengths were

determined using a peak-picking algorithm on each instrument.

Three spectra were acquired for each sample and the ET(30)

values for each sample were averaged. The ET(30) data were

taken every 10% organic and fit to the appropriate degree

polynomial using the Interactive Microware (State College,

Pennsylvania) program CURVE FITTER run on an Apple

(Cupertino, California) II Plus 48K microcomputer. Any

unmeasured ET(30) values were determined by interpolation.

Retention measurements

All retention measurements were obtained with a Beckman

(San Ramon, California) Model 100A isocratic LC pump, a

Beckman Model 153 fixed wavelength (254 nm) UV detector, a

Valco (Houston, Texas) C6W injector with a 10 (L sample loop,

a Fisher (Austin, Texas) Recordall Series 5000 strip chart

recorder and a Hamilton (Reno, Nevada) 705 SNR LC syringe. A

Beckman Ultrasphere ODS (5 iLm), 15 cm X 4.6 mm column and a

DuPont (Wilmington, Delaware) Zorbax TMS (6 gm), 15 cm X 4.6

m-n column were used. The columns and solvents were

zhermostated at 300C with a Brinkmann Lauda (Westbury, New

York) Model MT heater/circulator. Fisher (Austin, Texas)

HPLC grade methanol and acetonitrile, certified 1-propanol

and Florida Distillers (Lake Alfred, Florida) Absolute Ethyl

Alcohol (200 proof) were used as received. Water was first

purified with a Barnstead (Milford, Massachusetts) Nanopure

system, irradiated with UV light in a Photronix (Medway,

Massachusetts) Model 816 HPLC reservoir for at least 24 hours

and then filtered through a Rainin (Woburn, Massachusetts)

0.45 pm Nylon-66 membrane filter prior to use. Pure solutes

were used as received and stock solutions made in HPLC grade

methanol: Eastman Kodak (Rochester, New York) reagent ACS

spectro grade toluene, butylbenzene, naphthalene,

p-nizroanisole and benzylamine, Fisher certified ethylbenzene

and nitrobenzene, Mallinckrodt (St. Louis, Missouri) organic

reagent benzophenone, MCB (Norwood, Ohio) p-nitrophenol and

Alfa (Danvers, Massachusetts) n-propylbenzene. Retention

times were determined manually and the breakthrough time (to)

used to calculate capacity factors found by the elution of an

injection of HPLC grade methanol.

Linear regression

Regression calculations were done by using the

Interactive Microware program CURVE FITTER run on an Apple

(Cupertino, California) II Plus 48K microcomputer.

Results and Discussion

Solvatochromic polarity measurements

The results of the ET(30) solvent polarity measurements

for binary aqueous solutions of ethanol and of n-propanol are

illustrated in Figures 2-1 to 2-4. Figures 2-1 and 2-2 show


0 20 40 60
% (v/v) Ethanol

Figure 2-1.

The ET(30) polarity of binary ethanol-water
mixtures as a function of the volume percent of
ethanol in water.

80 100




ET(30) 58-



0 20 40 60 80 100
% (v/v) n-Propanol

Figure 2-2.

The ET(30) polarity of binary n-propanol-water
mixtures as a function of the volume percent of
n-propanol in water.




ET(30) 58



0.0 0.2 0.4 0.6 0.8
Mole Fraction of Ethanol in Water

Figure 2-3.

The ET(30) polarity of binary ethanol-water
mixtures as a function of the mole fraction of
ethanol in water.




ET(30) 58
(kcal/mole) 56
56 -


52 -

0.0 0.2 0.4 0.6 0.8
Mole Fraction of n-Propanol in Water

Figure 2-4.

The ET(30) polarity of binary n-propanol-water
mixtures as a function of the mole fraction of
n-propanol in water.

the change in ET(30) values for ethanol and n-propanol versus

volume percent and Figures 2-3 and 2-4 versus mole fraction.

ET(30) measurements for methanol and acetonitrile aqueous

mobile phases have previously been discussed (Dorsey and

Johnson, 1987; Johnson, 1986). As previously evidenced with

methanol-water and acetonitrile-water eluents, the ET(30)

polarity values for aqueous ethanol and n-propanol solutions

show non-ideal behavior when related to either mole fraction

or volume percent of organic modifier. For any binary

solvent mixture to be ideal, the free energy change upon

mixing must be linearly related to the mole fraction of one

of the components (Lewis and Randall, 1961):

Fl Fo1 = RT In X1 (2-1)

where F is the free energy of the solution, R is the ideal

gas constant, T is the temperature in Kelvin and X is the

mole fraction of the component of interest. Any bulk

physical property that is linearly related to the free

energy, like partial molar volume, dielectric constant or

viscosity, can be substituted into equation 2-1. The

observation that the described solvent mixtures form

non-ideal solutions is not surprising since there are many

complicated intermolecular interactions present. Nonpolar

and nonpolarizable solvents such as benzene and toluene can

form ideal solutions because only weak dispersion forces are


As has been previously discussed about the organic-rich

region of the acetonitrile-water system (Balakrishnan and

Easteal, 1981; Johnson, 1986), the ET-30 probe molecule may

be sensing a breakdown in the hydrogen-bonding network of

the solutions in the high and low organic content ranges. At

roughly 80% organic, the ET(30) polarity decreases more

significantly than at the higher water content mixtures.

This is believed to occur because the association of the

negatively charged phenoxide group of ET-30 and water is a

more staunch interaction than that between the dye and

organic solvents. As the aqueous portion is almost

completely removed, the measured polarity plunges drastically

toward that of the pure organic solvent because less hydrogen

bonding is sensed. Unlike the acetonitrile-water system,

however, the ET(30) changes at high organic shown here for

ethanol and n-propanol aren't as great because these alcohols

are stronger hydrogen-bond donors than acetonitrile.

Methanol does not exhibit this behavior because its hydrogen

bond acidity is even stronger than the other three solvents

(Kamlet et al., 1983).

For high water content mixtures (less than 0.2 mole

fraction), the ET(30) polarity increases substantially from

the medium-range mole fractions. From entropy and enthalpy

of mixing data it has been hypothesized for dilute aqueous

solutions of a non-electrolyte that a collective

stabilization of the hydrogen-bond lattice of water occurs

due to an increase either in the energy of water-water

hydrogen-bonds or in their number (Naberuklin and Rogov,

1971). Therefore, it could be that a change in stabilization

of the hydrogen-bonding network of the solution is being

sensed by the ET-30 probe in the dilute alcohol concentration

region. Furthermore, the random mixing approximation is not

expected to be generally viable in the limits of extreme

composition (Schoenmakers et al., 1978), less than a few

percent of either mixture component at both high and low

organic. In these regions, the ET-30 probe molecules or the

minor component of the binary solvent may associate to form

non-random mixtures.

Chromatoqranhic retention measurements

Retention data were gathered for ten solutes on an

octadecylsilane (C18) reversed-phase column using aqueous

methanol, ethanol, n-propanol and acetonitrile solutions as

the mobile phases. The test solutes were chosen so that a

variety of compound types would be used. Acetonitrile was

used as a reference organic modifier from which the

homologous alcohols could be compared in the form of log k'

vs. ET(30) slope ratios.

It is expected that upon going from methanol to ethanol

to n-propanol mobile phases, the slopes of the log k' vs.

ET(30) plots for any of the test solutes would decrease by an

amount linearly related to the carbon number of the alcohol

modifier. This was believed since it has been shown for

alcohol-water mobile phases on a C18 column that the extent

of solvation of the surface by the alcohol over the water

increases with carbon chain length of the alcohol (Scott and

Simpson, 1980). Yonker et al. (1982b) have noticed

differences in solute selectivity and in the phase ratio of

the column between modifiers of varying solvating ability.

Carr et al. (1986) and Dill (1987) discussed the importance

of the formation of solute-sized cavities in the stationary

phase but both ignored the possible consequences of eluent

solvation of the surface. Ying (1989), however, has shown

the binary interaction constant for water and the organic

modifier calculated by Dill's theory, XAB' must be corrected

for the size of the solute with respect to the organic


An example of the data generated in this study is the log

k' versus ET(30) plots for naphthalene in all four solvent

systems (Figure 2-5). The average correlation coefficient

(r) for a total of forty different data sets (similar to

Figure 2-5 but not illustrated) is 0.997 0.002. By

observing the positioning of each of the data sets along the

ET(30) axis, it is obvious that not all of the polarities for

the different solvent mixtures overlap. The weakest

n-propanol solvent used (30%) was about 1050 cal/mole less

polar than the strongest methanol solvent used (90%). It is

also obvious that there are no iso-ET(30) values in

reasonable isocratic retention ranges among the four solvent

systems. In other words, there are no points of identical

retention at one particular ET(30) value. If the ET(30)

log k'

-0.50 1 1', I
52.5 54.0 55.5 57.0 58.5

Figure 2-5.

The retention of naphthalene on an Ultrasphere
ODS column at 300C as a function of the ET(30)
polarity of binary hydroorganic mobile phases.

o Methanol
* Ethanol
n Acetonitrile
o n-Propanol

values are giving a useful measure of mobile phase strength,

then one possible explanation is a continued change in the

solvation structure of the stationary phase even for organic

modifiers of the same functionality differing only in size.

From Figure 2-5 and the slopes of equations 1-7 and 1-8

for all ten solutes (Table 2-1), a few trends can be

observed. First, as seen previously with aqueous methanol

and acetonitrile mobile phases (Johnson, 1986), the slopes of

the log k' versus ET(30) plots for the ethanol and n-propanol

mobile phases increase with increasing solute size. The

configurational entropy of the stationary phase, and thus

retention itself, depends heavily on the size of the solute

(Dill, 1987). For this work, the addition of methylene

groups in the alkylbenzene homologous series (toluene to

n-butylbenzene) produces a linear relationship with respect

to the log k' versus ET(30) slopes (r = 0.998 for ethanol and

0.999 for n-propanol) Second, the log k' versus ET(30)

plots for methanol, ethanol and acetonitrile appear to extend

to a common point. The average intersection point between

the retention plots of a solute using two different modifiers

came to an average ET(30) value of 65.65 11.60 kcal/mole (n

= 30), which is near the ET(30) value for pure water of 63.1

kcal/mole (Reichardt, 1988). The n-propanol plots, however,

do not approach the other plots near 63.11 kcal/mole at all,

which suggests that a different mechanism of retention may be

occurring with the aqueous n-propanol mobile phases in

contrast to the other three. Third, the expected results of

Table 2-1.
The slopes of the log k' versus ET(30) plots and Snyder S
on an Altex Ultrasphere ODS column at 300C.

values for the homologous solvents and aoetonitrile


Solute Er(30) Slope -S


ET(30) Slope -S


ET(30) Slope -S


E-p(30) Slope -S






































2.19 0.298 1.66














































0.274 1.88

Benzylamine 0.321

a decrease in the log k' versus ET(30) slopes with increasing

carbon number of the alcohol solvent modifiers did not occur.

As seen in Table 2-1, the ET(30) slopes for each test solute

decreased when going from methanol-water to ethanol-water,

but upon going from ethanol-water to n-propanol-water the

slopes increase. None of the solutes, however, showed

intersections near 0% modifier when volume % was used as the

eluent strength descriptor.

If the slope of equation 1-7 for a test solute in an

alcohol-modified solvent is ratioed to that in

acetonitrile-modified solvent for all three alcohols, the

retention behavior for each alcohol-water system can be

compared to the others for all 10 solutes. The average

methanol-acetonitrile ratio for all ten solutes came to be

1.13 0.09, for ethanol-acetonitrile 0.84 0.05 and for

n-propanol-acetonitrile 1.14 0.06. The

methanol-acetonitrile slope ratio was within experimental

error of the previously determined average value of 1.18

0.05 for C18 at 400C (Johnson, 1986). The larger the slope

or slope ratio, the greater the change in retention per unit

of polarity change. Methanol slope ratios are greater than

those for ethanol due to the fact that in a binary

alcohol-water mobile phase, methanol does not selectively

solvate the stationary phase to as large a degree as ethanol.

When comparing one chromatographic system using methanol to

another using ethanol, the stationary phase will be less

polar (contain less alcohol) for the methanol system when the

mobile phases for both are equal in alcohol composition. If

an equivalent increase in mobile phase polarity as measured

by ET(30) is performed in each system, retention will change

to a greater extent in the methanol system than the ethanol

system due to the greater difference in the polarity of the

mobile and stationary phases. For n-propanol-water mobile

phases, however, the stationary phase may be saturated by

n-propanol when using solutions of 30% or more. It was

calculated that when using a 9% n-propanol-water solvent, the

C18 chains on the surface already contain up to 95% of the

maximum uptake by n-propanol (Scott and Simpson, 1980).

Therefore, for the concentrations of n-propanol used (30 to

70%), increasing the amount of n-propanol in the solvent may

not produce changes at the stationary phase in the same

manner as with methanol and ethanol.

Another consideration is the chemical information

obtained from ET(30) measurements. The ET-30 molecule is

sensitive to solvent dipolarity/polarizability and hydrogen

bond acidity. According to Kamlet et al. (1983), the ET(30)

polarity of a given solvent can be related to the

dipolarity/polarizability (I*) and hydrogen bond acidity (a)

scales by the regression equation

ET(30) = 28.21 + 12.407* + 14.40a (2-2)

For water, methanol, ethanol, n-propanol and acetonitrile,

respectively, the values of K* are 1.09, 0.60, 0.54, 0.52 and

0.75 and the values of a are 1.17, 0.93, 0.83, 0.78 and 0.19.

These t* and a values show only small differences among the

three pure alcohols. If the trends seen in the log k' versus

ET(30) plots among the various organic modifiers were not due

to differences in the solvation structure of the stationary

phase, then the trends should be due to the solvent

parameters measured by t* and a. Cheong and Carr (1988) have

examined the i*, aand p properties of aqueous solutions of

methanol, 2-propanol and tetrahydrofuran and attribute

decreases in hydrogen bond acidity for methanol-water

mixtures from that of the pure solvents to the formation of

less hydrogen bond acidic complexes. These effects between

the different mobile phases, however, are normalized by the

ET(30) scale and should not show up as differences in the log

k' versus ET(30) plots.

It was necessary to see if the hypothesis of n-propanol

saturation at the C18 surface was possibly occurring.

Retention measurements were made for benzylamine and

p-nitrophenol, the two least retained solutes, at n-propanol

in water compositions below 30% (10, 15, 20, 25%). The

previously determined n-propanol-acetonitrile slope ratio for

benzylamine was 1.09 while that for p-nitrophenol was 1.15.

When the retention measurements were observed from 35 to 10%

n-propanol, the new slope ratios for n-propanol-acetonitrile

came to 0.57 for benzylamine and 0.23 for p-nitrophenol. It

is interesting to note that these slope ratios are less than

the average ethanol-acetonitrile value of 0.84 0.05. The

p-nitrophenol value is questionable because the log k' versus

ET(30) plot was not as linear as desired (r = 0.969). Figure

2-6 shows the log k' versus ET(30) plot for benzylamine from

10 to 50% n-propanol in water. The linear region of the plot

extends from 10 to about 35% with an apparent break occurring

between 35 and 40%. It is also interesting to note that when

the point of intersection of the 10 to 35% n-propanol plots

with the other three modifiers was calculated for both

benzylamine and p-nitrophenol, the average E (30)

intersection point was 58.70 1.06 kcal/mole (n = 6), which

is close to the ET(30) value of pure water. Based on the

resulting n-propanol-acetonitrile slope ratios and the

intersection of the n-propanol plots with those of the other

three modifiers, it would appear that for n-propanol

compositions at or below 30% the mechanism of retention is

similar to that when using the other modifiers. Above 30%

n-propanol, the C18 stationary phase may be saturated by

n-propanol and a different mechanism of retention is

operative. While only two data sets were measured due to the

extremely long retention times and poor peak shapes in this

mobile phase range, the implications are encouraging.

If the slopes of equation 1-7 for benzylamine are plotted

against the carbon number of the alcohol modifier, a linear

relationship is found (r = 0.999), as illustrated in Figure

2-7. The same trend is seen for p-nitrophenol but the

linearity is slightly worse (r = 0.979). When the n-propanol

slope of the 30 to 50% solvents was used, no linear



log k' 0.55



53 55 57 59

Figure 2-6.

The retention of benzylamine on an Ultrasphere
ODS column at 300C as a function of the ET(30)
polarity of binary hydroorganic mobile phases
ranging from 10 to 50% (v/v) n-propanol in


log k' vs. ET(30)



2 3
Alcohol Carbon Number

Figure 2-7.

The relationship between the retention behavior
of benzylamine (using the ET(30) model) on an
Ultrasphere ODS column at 300C and the carbon
number of the modifier alcohol in a binary
hydroorganic mobile phase.

correlation was found between the ET(30) slope and alcohol

carbon number. Thus, a systematic change in retention may

occur when using modifier compositions in a range where no

surface saturation occurs.

In order to see if the solvation differences of the

stationary phase between solvent systems could be minimized,

retention measurements for the ten test solutes were made on

a trimethylsilane (Cl) column. By using a Cl column, it was

anticipated that specific solvation interactions between the

solvent modifier and the bonded phase would be minimized.

Normalization of the retention behavior of a solute in

various solvent systems should then be reflected in all

alcohol-acetonitrile ratios of log k' versus ET(30) slopes

approaching unity.

No conclusive results could be found in the Cl retention

data, however. One of the problems encountered when using Cl

columns in RPLC is the impreciseness of the chemistry of the

surface. The polarity of the reversed-phase surface

increases as the bonded carbon chain is shortened since the

residual silanol groups are left less shielded from the

solute and mobile phase and are freer to interact in the

retention process (Antle et al., 1985). Silanol interactions

can affect results in the form of irreproducible retention

times and tailed peak shapes. Furthermore, a Cl column

likely exhibits an adsorption type mechanism rather than

partitioning because of the shallow "depth" of the stationary

phase. A representative plot of the data gathered is shown

for naphthalene in Figure 2-8. The linearity for all data

sets deteriorated in contrast to the C18 data as expressed in

a lower and more scattered average r value of 0.991 0.016

(n = 40). A slight degree of curvature was evident with most

of the plots. No surprising trends are noticable from the

ET(30) slopes and S values shown in Table 2-2. When slope

ratios were calculated, methanol-acetonitrile remained about

the same (1.16 0.15), n-propanol-acetonitrile increased

slightly (1.22 0.20), and ethanol-acetonitrile came to

unity (1.00 0.10). The uncertainty in each average slope

ratio, though, is too great to make any generalizations with

respect to the solvation behavior of each solvent system on

the C1 surface.

A beneficial aspect of this work is the intersection of

the log k' versus ET(30) plots for different modifiers on a

C18 column at the ET(30) value of pure water. This

intersection is further evidence that the ET(30) model of

solvent strength usefully monitors the RPLC partitioning

processes under different equilibrium conditions. A problem

often associated with the evaluation of solute lipophilicity

by HPLC is the extrapolation of equation 1-8 to 0% organic

concentration in the mobile phase (Garst and Wilson, 1984).

The reproducibility of the retention value in pure water (log

k'w) is affected by the magnitude and direction of curvature

in a log k' versus % plot (Garst, 1984). Reymond et al.

(1987) have recently shown that linear extrapolations of

methanol-water retention data yielded practically identical

log k'


-0.50 -z

E Methanol
* Ethanol
* Acetonitrile
o n-Propanol

55.0 56.5 58.0 59.5


Figure 2-8. The retention of naphthalene on a Zorbax TMS
column at 30C as a function of the ET(30)
polarity of binary hydroorganic mobile phases.

Table 2-2.
The slopes of the log k' versus BE(30) plots
on Zorbax TIMS column at 30C.


Solute ET(30) Slope -S

and Snyder S values for the hinologous solvents and acetonitrile


EB(30) Slope -S


ET(30) Slope -S


ET(30) Slope -S



























































































log k', values as parabolic extrapolations of

acetonitrile-water mixtures. Schoenmakers et al. (1983),

however, reported that the quadratic fit of log k' versus %

organic modifier does not hold at >90% water, and they

recommend adding an empirical (% modifier)1/2 term for that


Since more than one binary solvent system can be used to

determine the log k', value, it would be desirable to have as

low of an uncertainty between log k'w values as possible. An

average difference between log k', values for each test

solute in the different mobile phases was determined. The

data sets used were those for methanol, ethanol and

acetonitrile. Due to the previously discussed problem of

stationary phase saturation, the n-propanol data of this work

was not considered. For thirty data sets, calculation of log

k'w by extrapolation of % organic plots to 0% gave an average

difference of 0.39. Subsequent calculation of log k'w at

63.11 kcal/mole (the ET(30) polarity of pure water) for

ET(30) plots gave an average difference of 0.24; an

improvement of about 40%. In addition, log k' versus ET(30)

plots for different modifiers converge toward the ET(30)

value of water, as seen in Figure 2-9. Plots of log k'

versus % for different modifiers do not converge to 0%

organic at all but diverge (Figure 2-10). The use of log k'

values from ET(30) plots has also been demonstrated for the

successful calculation of solvent-solvent contact free

energies of binary organic/water mixtures (Ying, 1989).

log k' 2.0- o Methanol
1.0 a Acetonitrile


53 55 57 59 61 63

Figure 2-9. The convergence of ET(30) retention plots for
naphthalene to the ET(30) polarity of pure water
(63.1 kcal/mole).

log k' 0 Methanol
10 Ethanol
1.0 n Acetonitrile


0 20 40 60 80 100
% (v/v) Organic Modifier

Figure 2-10. The non-convergence of % organic retention plots
for naphthalene to 0% (v/v) organic modifier.


Since the scatter of the values appears to be decreased from

scatter found using the % model and the ET(30) plots for

different solvents converge to a common point, the ET(30)

solvatochromic polarity scale should be a reliable means of

estimating log k', values for the evaluation of solute




As was stated in Chapter I, to date there has not been a

completely adequate measure of RPLC eluent strength for

lipophilicity studies, but one possible solution to the

problem lies in empirical solvatochromic solvent polarity

scales. These methods are useful for this purpose because

they quantify some of the significant intermolecular

interactions experienced by a solute in the chromatographic

system. By normalizing the solvation interactions of the

mobile phase by a polarity scale, the bulk hydrophobic

interactions experienced by the solute between aqueous and

nonpolar phases can be estimated by the parameter log k'w.

The ET(30) scale is attractive for use in RPLC lipophilicity

studies because the ET-30 molecule is readily soluble in RPLC

solvents and is extremely sensitive to small changes in

organic modifier composition.

Because the ET(30) solvent strength model provides a more

linear description of the retention process than equation 1-8

(Johnson, 1986), a more accurate estimation of log k', should

result upon extrapolation to the polarity of mobile phase

containing no organic modifier. It has been reported

(Braumann et al., 1983) that values of log k', estimated by

equation 1-8 are dependent on the modifier used. Also, since

the ET(30) polarity scale shows a linear relationship with

log k', equation 1-7 should produce results rivaling or

surpassing those from using equation 1-15.

Two important conclusions made in Chapter II can commence

procedural changes that could enhance the ability to estimate

lipophilicity by RPLC. When the retention data for solutes

were plotted with both equations 1-7 and 1-8 using methanol,

ethanol and acetonitrile as single organic modifiers, i) the

average difference between estimated log k'w values between

two different modifiers was improved by 40% with the ET(30)

model over % and ii) ET(30) plots for all three modifiers

converged toward the polarity of water while volume % plots

diverged at 0% modifier. These conclusions imply that a

solvatochromic model of eluent strength, like ET(30), may

yield reliable values of log k'w in more than just one


These types of convenient linearizations, however, are

likely the best application of solvatochromic measurements in

liquid chromatography as it is not clear that any type of

true fundamental information can be obtained from them. This

chapter compares the estimation of log k'w by the ET(30) and

% models of solute retention through the calculation of

comparative figures-of-merit and correlations with

octanol-water partition coefficients. A discussion of

recommended execution of lipophilicity experiments using

estimated log k'w values is also presented.


Solvatochromic measurements

Solvatochromic solvent polarity measurements were made on

binary solutions of pure organic solvent and aqueous buffer

using ET-30 (Reichardt's Dye) (Aldrich Chemical, Milwaukee,

Wisconsin). The organic solvents were Fisher (Austin, Texas)

HPLC grade methanol and acetonitrile and all water used was

purified using a Barnstead (Newton, Massachusetts)

purification system. One buffer was composed of 0.02 M Kodak

(Rochester, New York) 3-morpholinopropanesulfonic acid (MOPS)

and 0.2% (v/v) Aldrich n-decylamine and another made up of

66.6 mM Fisher ACS certified sodium phosphate monobasic.

Both buffers were adjusted to a pH of 7.4 with aqueous sodium

hydroxide. Binary organic-buffer solutions were prepared by

mixing additive volumes of pure organic solvent and buffer

solution in increments ranging from 0 to 100% organic

modifier for MOPS buffer and from 0 to 70% for phosphate

buffer. ET-30 was added to the simulated mobile phases to a

final concentration of approximately 100 mg/L. Samples were

placed into a Fisher 5 cm path length glass cell and spectra

obtained with an IBM Instruments (Danbury, Connecticut) Model

9420/9430 UV-Visible Spectrophotometer. Six spectra were

acquired for each sample and the ET(30) values averaged.

Maximum absorbance wavelengths were determined using a first

derivative algorithm on the instrument. The ET(30) data were

taken every 10% organic and fit to an appropriate degree

polynomial using the Crickett Software (Philadelphia,

Pennsylvania) program STATWORKS run on an Apple (Cupertino,

California) Macintosh SE Microcomputer. Any unmeasured ET(30)

values (ie. 45% methanol in MOPS buffer) were determined by

interpolation. ET(30) polarity values for methanol-water,

acetonitrile-water mobile phases were the same as those used

by Johnson (1986).

Retention measurements

RPLC retention measurements were taken from Chapter II of

this work and the literature (Baty and Sharp, 1988; Johnson,

1986; Lehtonen, 1984; Reymond et al., 1987; Schoenmakers et

al., 1981). Table 3-1 summarizes the pertinent experimental

details of each set of retention data.


All computations were done on the Apple Macintosh SE

computer with the exception of the polynomial confidence

intervals, which were done on an Apple II 48K microcomputer

using the program POLYCONFINT, written for this work. Linear

and polynomial regression were performed with STATWORKS and

all other calculations accomplished with the Microsoft

(Redmond, Washington) spreadsheet EXCEL.

Table 3-1.
Experimental conditions for reversed-phase retention data taken from this work and the

References Columns Modifiers to Methods

This Work Ultrasphere ODS Methanol Solvent Elution

Reymond et al.,

Johnson, 1986

Lehtonen, 1984

et al., 1981

LiChrosorb RP-18

Ultrasphere ODS

Spherisorb S5 ODS2

Nucleosil 10-RP18






Solvent Elution

Solvent Elution

NaNo3 Elution

Set to = 125 sec

literature for log k'w



Nicotinate Esters


1-Sulfonyl Derivatives

Nitro-, Amino-, Alkyl-,
Alcohols, Heterocyclics

amorpholino-propane-sulfonic acid

Results and Discussion

Solvatochromic polarity measureme-ts for neutral electrolyte

Studies have been previously performed investigating the

solvatochromic polarity of surfactant and electrolyte

solutions. One study of ionic surfactant solutions

(Zachariasse et al., 1981) showed that the ET(30) polarity of

a surfactant solution was dependent on the presence of added

buffer salts. Using NMR data, it was shown (Zachariasse et

al., 1981; Plienlinger and Baumgartel, 1983) that the charged

groups of ET-30 are coulombically influenced and align with

the other oppositely charged groups of the surfactants.

Another study using Kosower's Z Scale (Mohaamed and Kosower,

1970), which uses a probe similar to ET-30, has also shown

solvatochromic polarity increasing with the addition of


Table 3-2 presents the results of this work showing the

effects of increasing the amount of electrolyte on ET(30)

polarity. ET(30) values were measured for mixtures of

organic modifier with neutral buffers and sodium chloride

solutions. Within an experimental error of about 0.10

kcal/mole, an increase in polarity is observed with increase

in electrolyte concentration. ET(30) values of electrolyte

solutions, however, should be viewed with caution when

directly compared to values taken in nonelectrolyte solutions

because of the added coulombic interactions between the

Table 3-2.
The dependence of ET(30) polarity on the concentration of electrolyte in
neutral aqueous-organic mixtures.

Solution ET(30) (kcal/mole)

Water 58.30
50% (v/v) Methanol

1 rmM NaCla 58.03
50% (v/v) Methanol

10 mM NaCla 58.03
50% (v/v) Methanol

100 nM NaCla 58.25
50% (v/v) Methanol

Water 57.46
40% (v/v) Acetonitrile

31 nM Phosphateb 57.40
40% (v/v) Acetonitrile

66 nmM Phosphateb 57.48
40% (v/v) Acetonitrile

99 rM Phosphateb 57.76
40% (v/v) Acetonitrile

Water 57.46
60% (v/v) Methanol

10 nim MOPSc 58.08
0.2% (v/v) DAd
60% (v/v) Methanol

20 mM MOPSc 58.73
0.2% (v/v) DAd
60% (v/v) Methanol

31 iMi MOPSc 59.29
0.2% (v/v) DAd
60% (v/v) Methanol

sodium chloride
sodium phosphate monobasic
Cmorpholino-propane-sulfonic acid

charged functional groups of the dye and the electrolytes.

This is evident when comparing the water-organic to

electrolyte-organic mixtures.

The ET(30) polarity of a solution is also influenced by

acidity. Langals (1987) has shown that ET(30) values cannot

be measured in acidic electrolyte solutions because the

phenoxide group of ET-30 becomes protonated and the

charge-transfer complex is destroyed. There are no problems

of complete protonation of the dye with the work presented

here on account of ET-30 spectra being obtained in mobile

phases buffered at pH 7.4. The pKa of the dye has been

reported to be 8.4 in 50% methanol-water (Zachariasse et al.,

1981). Based on its pKa and the pH of the solutions used,

ET-30 can sense acidic interactions. Reichardt (1988) has

defined polarity as the total solvating power of the solvent

and thus these forces can be included in the ET(30)

"polarity" of basic solutions.

The relationships between the ET(30) polarity and the

volume percent of methanol and acetonitrile added to the

buffer solutions are shown in Figures 3-1 and 3-2. The

methanol-MOPS buffer curve in Figure 3-1 expressed nonlinear

behavior from 0 to 100%. One explanation of the shape of

this plot could be that the probe is more specifically

solvated by the MOPS and n-decylamine molecules at the high

and low % methanol regions. Furthermore, the random mixing

approximation is likely to fail at these composition

extremes. Mobile phases of compositions between 30 and 70%

0 20 40 60 80 100
% (v/v) Organic Modifier

Figure 3-1.

The ET(30) polarity change as a function of the
volume % of organic modifier in a mixture with
pH 7.4 0.02M MOPS/0.2%(v/v) n-decylamine buffer.


o Methanol
* Acetonitrile

(kcal/mole) 60



20 40 60
% (v/v) Organic Modifier

Figure 3-2.

The ET(30) polarity change as a function of the
volume % of organic modifier in a mixture with
pH 7.4 66.6mM phosphate buffer.

0 Methanol
* Acetonitrile

methanol give almost a linear polarity response with respect

to %. At high and low % methanol, specific solvation of the

probe is most likely different than for the middle % ranges

to produce the nonlinear behavior. The acetonitrile-MOPS

buffer curve in Figure 3-1 shows the polarity decreasing

nonlinearly from 0 to 80% acetonitrile and rapidly decreasing

from 80 to 100%. This behavior is similar to what has been

seen with acetonitrile-water mixtures (Johnson, 1986). The

phenoxide group of the ET-30 molecule is preferentially

solvated by water and once the composition is higher than 80%

acetonitrile, the ET-30 molecule senses primarily solvent

dipolarity and no hydrogen-bonding.

Mixtures of phosphate buffer with both methanol and

acetonitrile were studied and their polarity profiles

presented in Figure 3-2. The ET(30) polarity for these

solutions also exhibited nonlinear behavior in the range of %

organic compositions studied for both solvent systems. Due

to problems with solubility of the phosphate buffer salt with

the organic modifiers, the maximum mobile phase compositions

allowed were 70% methanol and 60% acetonitrile, so no

polarity measurements could be made at organic compositions

greater than these.

Estimation of loo k' by extrapolation methods

Values of log k'w were estimated by extrapolating linear

regression plots of log k' versus % organic for each solute

to 0% organic modifier and plots of log k' versus ET(30) to

the polarity of pure water. If a buffer solution was used as

the aqueous component of the mobile phase, the ET(30) plots

were extrapolated to the measured ET(30) polarity of the pure

buffer (63.36 kcal/mole for the MOPS/decylamine buffer).

Data sets from this work and the literature (Johnson, 1986;

Lehtonen, 1984; Reymond et al., 1987; Schoenmakers et al.,

1981) were used that had reported log k' versus % organic on

a C18 column for both methanol and a second solvent modifier,

such as acetonitrile or ethanol. A data set refers to a

collection of log k' values for one solute taken at different

mobile phase conditions using one modifier.

Three restrictions were placed on each set of log k' data

used. The first restriction was to not use log k' values

above 90% methanol or 80% of the second modifier. It has

previously been discussed (Johnson, 1986) that a limitation

of the ET(30) polarity scale used as a measure of mobile

phase "strength" for RPLC is the occurrence of specific

solvation effects between the probe molecule and the solvent

components at high percentages of organic modifier. Changes

in the solvent polarity above these % values were found to

not relate linearly to log k'; this is no problem, however,

because log k' values measured with more aqueous-rich mobile

phases should be used to extrapolate to log k'w. This

reasoning will be explained later in the discussion dealing

with confidence intervals. A second restriction was to not

use log k' values much less than -0.30 to try to minimize

measurement errors associated with small values of log k'.

The last restriction is directly related to the first two

such that if a data set contained less than four log k'

values, it was not used. This will also be explained by the

confidence interval discussion.

In order to compare the reliability of the log k', values

estimated by both the % and ET(30) models, four figures of

merit (FOM) were calculated using a total of 204 data sets.

Final results for each reference based entirely on linear

extrapolations are summarized in Table 3-3. The first FOM

was the correlation coefficient, r, which is a descriptor of

the "goodness-of-fit" of a linear model to a set of data. It

would be desirable to use a model that gives the most linear

description to the data to minimize errors associated with

forcing a line through data points that do not express linear

behavior. In the previous study of linearity of log k'

versus ET(30) (Johnson, 1986) only ten data sets on a C18

column were obtained, so r was monitored in this study as

well to check if similar linearity improvements over the %

model are obtained.

From Table 3-3, it can be seen that the average

correlation coefficient was greater using the ET(30) model

with the only exception being where the average r for both

models was the same. The overall apparent correlation

improvement using ET(30) over % may only be considered

marginal, however, since all of the average r values are

greater than 0.99. It was found that retention plots using

methanol as the modifier were comparable between the two

Table 3-3.
Comparative Figures-of-Merit for log k'w study based on linear regression for both the

Delta Log k'wb

% ET(30)


% ET(30)
(v/v) (kcal)

Ey(30) and % organic


% ET(30)

This Work

et al., 1987

Johnson, 1986

Lehtonen, 1984

et al., 1981


































+ 0.78

+ 0.28



correlation coefficient
difference between log k'w values estimated using two different modifiers
c% or ET(30) intersection point between regressions using two different modofiers
relative confidence interval about log k'w



% ET(30)

+ 7.50


+ 5.72

+ 3.79

+ 8.02

+ 8.69

+ 4.01

+ 4.78

+ 3.17

+ 1.47

+ 7.74

+ 6.91

models but acetonitrile and ethanol showed much better

linearity with the ET(30) model as compared to volume %. An

F--est (Anderson, 1987) of 95% confidence performed on the

variances of the averaged correlation coefficients for both

models, however, determined the two average r values (n=204)

to be significantly different. These results solidify

previous findings on linearity improvements and support the

observation that a solvatochromic polarity scale such as

ET(30) provides a more linear description of the strength of

the mobile phase than the bulk organic modifier composition.

A second FOM was Alog k',, the difference between the log

k'w values estimated by two different modifiers. Since it

would often be useful to employ another modifier, the

difference between estimations by the two modifiers should be

minimized. Furthermore, agreement of log k'w values from two

(or more) modifiers lends confidence to the accuracy of the

extrapolated value. Chapter II of this work has shown that

for 30 data sets a 40% decrease in the Alog k'w occurred when

using the ET(30) model over %. Table 3-3 shows a Alog k'w

improvement of roughly 21% when using the ET(30) model for

all reference data sets (N = 117) (Alog k'w=0.4089 for ET(30)

and 0.5144 for %).

The third FOM is the extrapolated intersection point of

linear regressions for a solute using two different

modifiers. This intersection determines if two different

retention plots converge toward a point indicating an

unmodified aqueous mobile phase. If a given retention model

describes the strength of the mobile phase in a useful

manner, then % plots should converge to 0% organic or the

ET(30) plots to the polarity of unmodified aqueous solvent.

It will be defined here that % intersections less than 50%

organic and Em(30) intersections greater than a polarity of

58 kcal/mole (the approximate polarity of 50% organic mobile

phase) will be considered converging toward log k'w. From

Table 3-3 it can be seen that all of the average

intersections of the ET(30) data sets converged toward the

polarity of water while it is difficult to infer the same

about the % data. Some of the average % intersections were

negative, some were positive, and the standard deviations of

those averages were quite large.

In order to extract more meaning out of the intersection

results, frequency histograms were constructed to demonstrate

the distribution of these values. Figure 3-3 shows the

spread of the % intersections covering a range from -200 to

500% organic for 97 of the 114 total data sets. Some

intersections could not be calculated (expressed in Table 3-3

as a NULL result) because the two retention plots of interest

were parallel. It was found that 70% of those intersections

occurred at values greater than 50% organic, with 41% of the

total sets being in the range between 100 and 200% organic.

Figure 3-4 illustrates the spread of the ET(30) intersections

in the range from 55 to 70 kcal/mole for 94 of the 117 total

data sets. It was found that 78% of these intersections were

greater than 58 kcal/mole, with 42% of the total sets


1 .................................

-200 -100 0 100 200 300 400
% (v/v) Organic Intersections

Figure 3-3.

Frequency histogram of the distribution of
linear/linear intersections between plots of log
k' versus % methanol and log k' versus %
acetonitrile or ethanol.

University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs