Solute-solvent interaction free energies and retentivity of reversed phase liquid chromatography columns

MISSING IMAGE

Material Information

Title:
Solute-solvent interaction free energies and retentivity of reversed phase liquid chromatography columns
Physical Description:
ix, 130 leaves : ill. ; 28 cm.
Language:
English
Creator:
Ying, Peter Tai Yuen, 1962-
Publication Date:

Subjects

Subjects / Keywords:
Liquid chromatography   ( lcsh )
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1989.
Bibliography:
Includes bibliographical references (leaves 123-129).
Statement of Responsibility:
by Peter Tai Yuen Ying.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001521240
notis - AHD4403
oclc - 22225694
System ID:
AA00003779:00001


This item is only available as the following downloads:


Full Text














SOLUTE-SOLVENT INTERACTION FREE ENERGIES
AND RETENTIVITY OF REVERSED PHASE
LIQUID CHROMATOGRAPHY COLUMNS
















BY

PETER TAI YUEN YING


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


1989


















ACKNOWLEDGMENTS

My first acknowledgment has to go to my research director,

Professor John Dorsey. Without his patience, guidance and

encouragement, I would not have completed this work. Also, I need to

thank him for introducing me to the world of wine and Depth Charge,

although the pleasure seems to always be his. I also want to express my

thanks to Professor Ken Dill at the University of California-San

Francisco for laying the theoretical foundation of this work. The

financial support from the NIH and NSF grants for this work is greatly

appreciated.

Special thanks go to my parents and family back in Hong Kong.

Even though we are thousands of miles apart, their love and

encouragement have kept me going. Also, their financial support is

greatly appreciated.

Thanks go to the sunshine, beaches, warm weather, and all the

beautiful coeds in shorts on campus for making my four years in Florida

most enjoyable. I am going to miss all of them when I move to the Windy

city.

Lastly, I want to thank all the members in the Dorsey group for

all the wonderful times we had (I love this place). They make my four

years of graduate school experience unforgettable. I also want to thank

them for putting up with my horrible singing, ubiquitous words, and the

frequent strident noises that I make.






















TABLE OF CONTENTS


ACKNOWLEDGMENTS..........................................


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


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


ABSTRACT............................................... ...


CHAPTERS


Page
............ii


.............v


............vi


..........viii


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


Solvophobic Retention Mechanism............................2
Partiton Retention Mechanism...............................5
Soltue-Solvent Interaction Constant......................13
Stationary Phase in RPLC.................................... 14


II SOLUTE-SOLVENT INTERACTION FREE ENERGIES IN REVERSED
PHASE LIQUID CHROMATOGRAPHY.............................26


Introduction. ............................................26
Experimental Procedure...................................28
Results and Discussion...................................30


III CHARACTERIZATION OF THE RETENTIVITY OF REVERSED PHASE
LIQUID CHROMATOGRAPHY COLUMNS ............................66


Introduction.... ..................... ....................66
Experimental Section......................................68
Results and Discussion..................................... 72


IV CONCLUSIONS AND FUTURE WORK.............................88


Conclusions.......................... ....................88
Future Work ...................................... ........90


APPENDICES

A


B


CHROMATOGRAPHIC RETENTION DATA...........................94


SLOPES FROM EQUATION 3-1 AND In kw....................... 109











REFERENCES ..........................................................123

BIOGRAPHICAL SKETCH.... .......... ...................................130






































































iv




















LIST OF TABLES


Table eage

2-1. Linear regression results of equation 2-2 and results of
slopes and intercepts ......................................33

2-2. Regression results of slopes from equation 2-3 versus the
van der Waals volume, Vw, of the solutes for all the
columns....................................................52

2-3. Regression results of slopes from equation 2-3 vs the
hydrocarbonaceous surface area, HSA, of the solutes for
all the columns............................................ 54

2-4. Regression results of y-intercepts at 0B=1 from equation
2-3 versus the van der Waals volume, Vw, of the solutes
for all the columns........................................ 59

2-5. Regression results of y-intercepts at OB=l from equation
2-3 vs the hydrocarbonaceous surface area, HSA, of the
solutes for all the columns ................................61

2-6. Regression results of slopes from equation 2-3 versus the
slopes of ET(30) plots for all the columns.................64

3-1. Properties of the RPLC columns as supplied by the
manufacturers ..............................................70

3-2. List of test solutes used in this study.....................71

3-3. Regression results of graphs of slopes from equation 3-1
against van der Waals volume, Vw, of the test solutes......75

3-4. Regression results of graphs of slopes from equation 3-1
against molecular connectivity index of the test
solutes....................................................77

3-5. Regression results of slopes from equation 3-1 vs In kw
of the test solutes for all the columns....................81

3-6. Phase ratio of the columns in this study calculated using
the method presented by Sentell (1988) .....................86




















LIST OF FIGURES


Xure Page

1-1. The three steps involved in the partition process
where the transfer of solute molecule S requires the
opening of a cavity in the stationary phase C and the
closing of a cavity in mobile phase A (Dorsey and
Dill, 1989) ................................................. 7

1-2. Pair interaction potential, uXy(r), for two simple
molecules. Reversible work for bringing molecules X
and Y together to their equilibrium separation r* is
WXY (Dorsey and Dill, 1989)................................. 8

1-3. Generalized bonding scheme for the synthesis of monomeric-
bonded phase using a monochlorosilane.....................18

1-4. Early models of molecular structure and organization
of the bonded phase in RPLC: a) "picket fence"; b) "fur";
c) "stacks" (Dill 1987) ...................................21

1-5. Interphase model of the bonded phase in RPLC proposed
by Dill (1987) ............................................22

2-1. Plot of 1/OB ln(k'/kw) versus OB for the solute
4-Nitrophenol using Hypersil ODS column and
acetonitrile as modifier..................................31

2-2. Histogram of coefficient of determination, r2, for the
plots of 1/1B ln(k'/kw) versus *B for all the data
sets ......................................................32

2-3. Plot of slopes from equation 2-3 versus van der Waals
volume, Vw, of the solutes for the Sepralyte C-18
column with acetonitrile as modifier......................51

2-4. Plot of slopes from equation 2-3 versus hydrocarbonaceous
surface area, HSA, of the solutes for the Sepralyte C-18
column with acetonitrile as modifier......................53

2-5. Plot of y-intercepts at =B=1 from equation 2-3 versus
van der Waals volume, Vw, of the solutes for the
Sepralyte C-18 column with acetonitrile as modifier....... 58










2-6. Plot of y-intercepts at pB=1 from equation 2-3 versus
hydrocarbonaceous surface area, HSA, of the solutes
for the Sepralyte C-18 column with acetonitrile as
modifier..................................................60

2-7. Plot slopes from ET(30) plots versus slopes from
equation 2-3 for the Sepralyte C-18 column with
acetonitrile as modifier..................................63

3-1. Plot of slopes from equation 3-1 against the van der
Waals volume, Vw, of the test solutes for the Zorbax
TMS column using acetonitrile as modifier.................74

3-2. Plot of slopes from equation 3-1 against the molecular
connectivity index of the test solutes for the Zorbax
TMS column using acetonitrile as modifier.................76

3-3. Plot of slopes from equation 3-1 against the In kw of
the test solutes for the Zorbax TMS column using
acetonitrile as modifier..................................80

3-4. Plot of slopes from equation 3-1 against the In kw of
the test solutes for all the Zorbax columns using
acetonitrile as modifier..................................84



















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

SOLUTE-SOLVENT INTERACTION FREE ENERGIES
AND RETENTIVITY OF REVERSED PHASE
LIQUID CHROMATOGRAPHY COLUMNS

By

PETER TAI YUEN YING

August, 1989

Chair: John G. Dorsey
Major Department: Chemistry

Recently, a description of the molecular mechanism of retention in

reversed phase liquid chromatography (RPLC) was derived by using a

partition model and statistical mechanical theory. This theory relates

the capacity factor, k', of a solute in RPLC to the binary interaction

constant between the solute molecules and the mobile phase molecules.

The binary interaction constant is related to the interaction free

energy through the Boltzmann's constant and the absolute temperature.

Solute-solvent interaction free energies can be obtained from

chromatographic data if the partition theory of solute retention is

correct.

We have tested the validity of the predictions from this theory by

using more than 300 sets of retention data from the literature and our

laboratory. With only a few exceptions, the data confirm all "the

predictions, and the solute-solvent interaction free energies are found

to be solute size dependent, which was a case not considered when the


viii











theory was first developed.

Although RPLC is one of the most popular separation techniques,

the actual effect that the stationary phase has on solute retention is

still not completely known. Much research has been done on the role and

effects of the mobile phase on solute retention in RPLC, but there has

been a lack of systematic approaches in characterizing RPLC columns.

We have developed a simple method to classify RPLC columns in

accordance with their retentivity by using more than 20 solutes of very

different chemical structures. The solvent strength of the mobile

phase has to be accounted for before the retentivity of these columns

can be investigated. This is accomplished by plotting In k' of the


solute against the ET(30) polarity values of the mobile phase. A good


parameter that sums all the retention due to the stationary phase is the


capacity factor of the solute at 100% water, In kw. By utilizing In kw


and the slope from the ET(30) plot, a relative retentivity of RPLC


columns is obtained.


L




















CHAPTER I
INTRODUCTION

Reversed phase liquid chromatography (RPLC) is the most common

mode of high-performance liquid chromatography used today. RPLC is

estimated to account for more than 80% of chromatographic systems

currently employed (Melander and Horvath, 1980). The name "reversed

phase" was introduced by Howard and Martin (1950) when they used

non-polar liquid paraffin and n-octane as the stationary phase to

separate fatty acids. They used the term to distinguish it from "normal

phase" chromatography, which consists of a polar stationary phase and a

non-polar mobile phase. The use of RPLC has shown an overwhelming

increase when columns packed with chemically stable microparticulate-

bonded silica became available. The practical use of commercial bonded

phase packing in RPLC can be dated back to Kirkland and DeStefano

(1970).

These bonded phases add great advantages to RPLC, including the

fact that aqueous mobile phases having low toxicity and high optical

transparency can be adopted to achieve most separations. Also, the

stability of these bonded phases are superior to other stationary

phases. Aqueous mobile phases of pH between 7.5 and 2.5 are compatible

with these bonded phases. Moreover, because of the wide range of

polarity of the mobile phases that can be used with these bonded phases,











an enormous amount of chemical compounds can be separated by using

bonded phases in RPLC.

Despite the prevalence of RPLC in analytical separations, there is

a lack of full understanding of solute retention and selectivity. This

lack of understanding hinders the development of a retention index in

RPLC similar to the Kovats index (Kovats 1965) or Rohrschneider and

McReynolds (Rohrschneider 1965; McReynolds 1970) constants in gas

chromatography (GC). It also creates problems for practical

chromatographers in developing separation schemes and for comparison of

retention results obtained from different laboratories. Horvath and

coworkers (Horvath et al., 1976; Melander and Horvath 1980) explained

their solvophobic retention mechanism for RPLC using a solute

association model. This is an important step toward the deconvolution

of solute retention on the molecular basis in RPLC, although there are

many shortcomings in this theory such as the ignorance of the retention

effects due to the stationary phase. The deficiencies of this theory

have been demonstrated in the literature by many researchers (Lochmuller

and Wilder, 1979; Lochmuller et al., 1981; Sadek and Carr, 1984;

Berendsen et al., 1980; Sentell and Dorsey, 1989a). Dill and coworkers

(Dill 1987; Marqusee and Dill, 1986a; Dorsey and Dill, 1989) have put

forth a molecular basis retention mechanism for small solutes in RPLC

using a partition model. In the following sections, these two

significant retention mechanisms in RPLC will be discussed briefly.




Solvophobic Retention Mechanism

The solvophobic theory was promoted by Horvath and coworkers

(Horvath et al., 1976; Melander and Horvath, 1980) utilizing the











solvophobic interaction of solute association from Sinanoglu (1967). In

this theory, solute retention is viewed as a reversible association of

the solute with the hydrocarbonaceous ligands of the stationary phase.

The hydrophobic interactions between the solute and the stationary phase

are thought to come from the fact that the mobile phase in RPLC is

relative polar, and the nonpolar moiety of the solute is repelled and

forced to associate with the nonpolar stationary phase. As Melander and

Horvath (1980) stated, the solute retention in solvophobic theory

actually involves two steps. The first step is the opening of a cavity

in the solvent that has the same size and shape as the solute molecule.

The second step is the placement of the solute in the solvent cavity and

all the interactions between the encircled solute and the solvent

molecules are followed. These interactions are originated from the van

der Waals forces and the electrostatic interactions amid the solute and

the solvent molecules.

The free energy associated with the first step of the solvophobic

theory can be expressed as




AGe = Ke(r) A y N (1-1)




where N is Avogadro's number, y is the surface tension of the solvent, A

is the molecular surface area of the solute and Ke(r) is a proportional

factor for the cavity size. From equation 1-1, the free energy of the

solvent cavity is proportional to the surface tension of the solvent.

The free energy related to solute-solvent interactions is comprised of

two chemical effects and an entropic term












A Gint = A Gvdw + A Ges + RT In(RT/PV) (1-2)




The change in free energy from the first chemical effect, A Gvdw,

can be calculated by using van der Waals potential in condensed media

for the solute and solvent. The change in free energy from the second

chemical effect, A Ges, can be further divided into dipole interactions

and ionic interactions. The dipole interactions can be treated

according to the Onsager reaction field approach (1936) while the ionic

interactions can be estimated using the Debye-Huckel theory. The

entropic term in equation 1-2 is related to the "free volume" of the

solute and this free volume is the measure of the volume of a molecule

before it collides with another molecule. The solute's molar volume can

be used to compute the free volume (Melander and Horvath 1980).

As far as the stationary phase effects on the retention of

solutes, Melander and Horvath (1980) considered them negligible because

the stationary phase is nonpolar and the only attraction between the

solute and the stationary phase is van der Waals forces. This van der

Waals force is insignificant compared to the van der Waals interactions

between the solute and solvent molecules. Although Melander and Horvath

(1980) acknowledged that an entropic term can be used to sum up the

restricted translational freedom of the bonded hydrocarbonaceous ligands

at the silica surface, they elected to pay no attention to this term in

their theory. In summary, the solvophobic retention mechanism takes the

approach that solute retention in RPLC is largely due to the hydrophobic

interactions of the solute and solvent molecules. The stationary phase


I











is treated as a passive entity that is forced to receive the solute from

the solvent, and its contribution to solute retention is negligible.




Partition Retention Mechanism

One of the catastrophic downfalls of the solvophobic theory is

that it disregarded the effect of the stationary phase on solute

retention. Dill and coworkers (Dill 1987; Marqusee and Dill, 1986a,

Dorsey and Dill, 1989) applied the mean-field statistical mechanical

theory, lattice theories (Hill 1986) and random-mixing approximation to

explore the retention mechanism of RPLC on the molecular level. Dill

and Dorsey (1989) used a simple partition model and took into account

the stationary phase effect on solute retention. They assumed the

stationary phase in RPLC as an "interphase" and solute retention is due

to partition between the the bulk mobile phase and the interphase. In

their model, the dominating driving force of solute transfer is from the

differences amid the chemical affinity of the mobile and stationary

phase. The solute capacity factor measured in RPLC can be expressed as




k' = K ( (1-3)




where k' is the capacity factor of the solute, K is the equilibrium

constant of the partition process and I is the phase ratio of the

chromatographic system and in RPLC, this is defined as the ratio of the

volumes of stationary and mobile phases. The equilibrium constant, K,

can be related to the free energy G(S) for the solute S


in K = -( Gosta(S) Gmoobile(S) )/kT


I


(1-4)











where Gosta(S) and Gomobile(S) are the standard-state free energies of

solute S in the stationary phase and the mobile phase respectively, and

kT is the Boltzmann's constant multiplied by absolute temperature.

The actual solute transfer of this partition process is comprised

of three steps. First, a cavity having the same size as the solute

molecule is opened in the stationary phase. Second, the solute molecule

from the mobile phase is transferred into the stationary phase cavity.

Third, the cavity left behind by the solute molecule in the mobile phase

is filled by other mobile phase molecules. These three steps are shown

in Figure 1-1, and for simplification, the mobile phase is considered to

be a single component. This will be used throughout the following

discussion of this retention mechanism. For a detailed derivation of

this retention mechanism using a binary mobile phase, the readers should

refer to Dill (1987).

The free energies involved in these three steps can be calculated

using pair interactions of molecules. The pair potential, u(r*), of

bringing a spherical molecule X from an infinite separation to within an

average equilibrium separation of spherical molecule Y can be defined as




u(r*) u(o) = u(r*) = WXY (1-5)




where wxy is the reversible work. Figure 1-2 shows a pair potential,

with short-ranged repulsion and longer-ranged attraction. The partition

process in RPLC is the result of this pair potential. The exact nature

of the pair potential is difficult to identify since the partition

coefficient of the process alone does not give enough information, but






























0< **




(iii) fill cavity


(i) open cavity


Figure 1-1.


The three steps involved in the partition process
where the transfer of solute molecule S requires the
opening of a cavity in the stationary phase C and the
closing of a cavity in mobile phase A (Dorsey and
Dill, 1989).


(ii) transfer


*0*


*0*


o*0
















Uxy(r)













Wxy = Uxy(r*)


Figure 1-2. Pair interaction potential, uxy(r), for two simple
molecules. Reversible work for bringing molecules X
and Y together to their equilibrium separation r* is
wyy (Dorsey and Dill, 1939).











it is generally believed that it is related to hydrogen bonding, dipole,

ionic and van der Waals interactions.

Dill and coworkers (Dill 1987; Marqusee and Dill, 1986a; Dorsey

and Dill, 1989) stated that from the lattice theory of liquids, the

opening of the cavity in the stationary phase leaves z sides, or z/2

contact (pair of sides), without contact. The coordination number, z,

is the number of neighbors of each molecules, or of each lattice site.

This cavity opening process in the stationary phase contributes

(-z/2)wCc to the free energy of the overall process. The same argument

can be applied to the closing of the cavity in the mobile phase, and

therefore the free energy contributed is (-z/2)wAA. The solute

molecule, S, in the mobile phase is surrounded by z nearest-neighbors of

the mobile phase molecules(solute self-association is ignored here).

The transfer process of the solute from the mobile phase to the

stationary phase involves the formation of z number of contacts in the

stationary phase while breaking z number of contacts in the mobile

phase. The free energies attributed from these two processes are zwSC


and zwSA respectively. The overall free energy of transfer is shown

below.




A Transfer = z( WSC wSA + wAA/2 WCC/2) (1-6)




It is more convenient to express this free energy of transfer in terms


of the binary interaction constant, XSC and XSA'











(A Gtransfer)/kT = XSC XSA (1-7)




where kT is Boltzmann's constant multiplied by the absolute temperature.

The binary interaction constant is defined as




XXY = z/kT ( WXY (WXX + wyy)/2 ) (1-8)




Equation 1-6 can be further simplified by using the Hilderbrand

solubility parameter concept (Hilderbrand and Scott, 1950) for small

molecules. The major interactions for small molecules are the induced

dipoles and the binary interaction constant can be approximated as a

product of factors involving unitary interaction constants, 5X and 8y,

the solubility parameters




XXY = constant ( X 5Y )2 (1-9)



The solubility parameter of a compound can be measured from the enthalpy

of vaporization of the pure compound. Although solubility parameters

have been used to model chromatographic retention (Schoenmakers et al.,

1978; 1982; 1983; Tijssen et al,. 1976; Karger et al., 1978) and the

factorization can reduce the size of the data base of constants, this

approximation is generally poor compared to the binary interaction

constant except in the case when dispersion forces are the principle

interactions (Schoenmakers et al., 1978).

The underlying molecular interactions that drive this partition

process in RPLC are very complicated. Since the capacity factor of a











solute in RPLC only gives information on the equilibrium constant of the

partition process, other experimental data are needed to fully

understand these interactions (Dorsey and Dill, 1989). For example, the

entropic dependence of these interactions can be determined by measuring

the relationship between temperature and the capacity factor of the

solute. The variation of the capacity factor with pH and salt

concentration in the mobile phase can be used to uncover the effects of

the electrostatic interactions on the partition process.

The partition model is supported by many experimental findings.

First, the capacity factors of solutes in RPLC are found to be directly

proportional to the water/octanol partition coefficients (Schantz and

Martire, 1987; Opperhuizen et al., 1987; Braumann 1986; Minick et al.,

1987; Kaliszan 1986). Second, inasmuch as the cavity formation and

closing in the stationary and mobile phases are important, the capacity

factors of solutes in RPLC are linearly dependent on the size of the

solute (Colin et al., 1983; Mockel et al., 1987; Horvath et al., 1976).

Third, the partition of solutes into the stationary phase increases when

the solubility of the solutes decreases in the mobile phase and

therefore the retention of the solutes increases. This is evidenced

when salt content in the mobile phase is found to be directly

proportional to the retention of hydrocarbons in RPLC (Tanford 1980;

Melander and Horvath 1980). Fourth, the surface tension, YA, of a pure

mobile phase can be expressed as




YA wAA/2a (1-10)











where a is the area per AA contact. Although it only describes the

cavity in the mobile phase, at situations when the stationary phase

effects on the solute retention are small (i.e., the other terms in

equation 1-6 are small) the surface tension of the mobile phase should

be proportional to the retention (Melander and Horvath, 1980; Hammer et

al., 1982; Horvath et al., 1976).

Perhaps another possibility of solute retention in RPLC is by

adsorption. An adsorption mechanism related to other types of

chromatography has been developed by Martire and Boehm (1983) and

Jaroniec (1984). Dill (1987) has examined the possibility of adsorption

mechanism and compared it with the partition theory that he proposed.

He concluded that the adsorption model is inferior to the partition

model for two reasons. First, in the adsorption model, the density of

the grafted alkyl chains on the stationary phase should have no effect

on the retention of solutes. On the other hand, in the partition model,

the partitioning of solutes should decrease if the density of the

grafted alkyl chains increases because it would be entropically

expensive for the grafted chain to uptake the solutes after a critical

bonding density (Dill 1987; Dorsey and Dill, 1989). This decrease in

solute partitioning with increase in alkyl chain density has been

reported (Claudy et al., 1985; Sentell and Dorsey, 1989a). Second, if

partition is the predominant mechanism of solute retention, the In k's

of a homologous series should be a linear function of the logarithm of

the appropriate water/octanol partition coefficient, with a slope of 1.

For the adsorption mechanism, such a plot should have a slope of 1/z.

This is because in adsorption, only one side of the solute is in contact

with the alkyl chains, and therefore the transfer free energy of this


I











process is only 1/z amount of the partition process measured by the

water/octanol partition coefficient. In the partition model, the solute

is embedded by the alkyl chains, and the whole solute is in contact with

the alkyl chains. This simulates the partition process between two bulk

liquid phases, and hence the In k' of a solute in RPLC has a one-to-one

relationship with the corresponding logarithm of the water/octanol

partition coefficient, and this has been found in many systems (Butte et

al., 1981; Braumann 1986; Kaliszan 1986).

Although Dill and coworkers (Dill 1987; Marqusee and Dill, 1986a;

Dorsey and Dill, 1989) used a few assumptions and some simple

approximations on the partition mechanism, their model correctly

accounted for most of the experimental observations that cannot be

explained sufficiently by other retention mechanisms. This model does

not treat the stationary phase as an inert part of the chromatographic

separation. The stationary phase is considered to play an active role

in the solute partition process since the free energy involved in the

cavity opening procedure in the stationary phase is of significant

influence on the total energetic of the chromatographic separation in

RPLC.




Solute-Solvent Interaction Constant

The binary interaction constant X is an important parameter for

the partition mechanism and can be used to calculate the solute-solvent

free energy as seen in equation 1-7. The binary interaction constant X

is usually obtained by measuring the free energy of transfer or vapor

pressure in the Henry's law region (Hill 1986). Dill (1987) used the

partition model and predicted that in an RPLC system having a binary











mobile phase, the capacity factor of a solute can be estimated by using

the binary interaction constants among the solutes and solvents. He

calculated the capacity factors of some solutes by using the X obtained

in vapor pressure measurement for the binary solvents and X obtained

from free energy of transfer of the solutes and found good agreement

with experimental data. The scenario can be reversed by using

chromatographic capacity factors of solutes to evaluate the

solute-solvent interaction constant. This should be very beneficial

because chromatographic data are more convenient to collect and reliable

than vapor pressure measurement data.




Stationary Phase in RPLC

The retention mechanism in RPLC has been thoroughly reviewed and

it has been pointed out by the partition mechanism that the stationary

phase in RPLC contributes a great deal to solute retention and cannot be

ignored. In order to fully understand the chromatographic process and

to forge ahead the applicability of RPLC, the structure and properties

of stationary phases used in RPLC deserve an in-depth investigation.

The stationary phase in RPLC has not been under the intense study as the

mobile phase has, but recently a few articles have been published on the

structure and properties of them (Sander et al., 1983; Gilpin 1984; Carr

and Harris, 1986; Buszewski and Suprynowicz, 1987; Nawrocki and

Buszewski, 1988).

Silica is by far the most common base material used in RPLC

stationary phases because silica is the best-known inorganic, polymeric

material (Unger 1979). Many reliable methods are available to control

the size and pore size of silica, and silica of the required size and


I











shape can be produce in large quantity. Before the introduction of

chemically bonded phases, silica was used as the support for

liquid-liquid chromatography where a thin film of liquid is coated onto

the surface of the silica and then is used as the stationary phase

(Unger 1979). These liquid stationary phases suffer from instability

and irreproducibility. The evolvement of chemically bonded phases by

chemically treating the silica surface with silanes began in the middle

of 1960s. They were debuted in gas chromatography as nonpolar

stationary phase obtained by reacting silica gel with

hexadecyltrichlorosilane (Abel et al., 1966). The employment of

chemically bonded phases in liquid chromatography was first suggested by

Stewart and Perry (1968). Chemically bonded phases have excelled to

become the most frequently used stationary phases in RPLC today due to

the advancement of synthetic technology for preparing these phases.

The underlying properties of these chemically bonded phases depend

on the physical and chemical properties of the base silica, the bonding

reactions and reagents used. The base silica constitutes the major

portion of the stationary phase, and the first property of the base

silica that influences the chromatographic process is the particle size

of the silica. The size of most silica particles employed in RPLC

stationary phases are in the range of 3-10 pm. The actual effect of the

silica particle size on RPLC is inconclusive since inconsistent results

have been reported in the literature. Dewaele and Verzele (1983) have

done a study on the influence of silica particle size distribution on

the reversed phase packing, and they found that the particle size

affects chromatographic performance of the packing minimally. Gazda et

al. (1980) also had the same conclusions, but Bristow (1978) has shown











otherwise. Pore size distribution is another important physical

property of silica particles. Verzele et al. (1985) have noticed that

the mean pore diameter of the silica used as chromatographic packing

should be in the range of 6-10 nm, and the narrower the distribution

curve the better it is. Micropores of diameter 2 nm or smaller are

undesirable because they can alter the final coverage of the bonded

phase (Berendsen et al., 1980; Engelhart et al., 1982).

Apart from the physical properties, the surface chemistry of the

silica particles endure the final characteristic of the bonded phase.

Silica surface is very complex, and it has been found to consist of

various kinds of silanols and siloxane bonds. The siloxane bonds are

usually considered as rather chemically inert and therefore do not take

part in the formation of the bonded phase. Single, geminal and vicinal

forms of silanols are found on the surface of silica (Unger 1979; Majors

1980), and chromatographers have mixed feelings toward these surface

silanols. First of all, these silanols react with silanes to form the

bonded phase, but on the other hand, these silanols are one of the major

contributions to the variability of the bonded phases. Moreover, these

silanols account for the pH of the silica. Different values of pKa of

silica have been reported in the literature (Walling 1950; Karger et

al., 1980; Majors 1980). Engelhardt and Miller (1981) first observed

the divergence of surface pH of silica packing for chromatographic

purpose. This reflects the different methods of manufacturing silicas

and connects to the different chromatographic properties of silicas

(Hansen et al., 1986).

Trace metal impurities found in silica is another factor which

influences the property of the bonded phase. The effect of trace metal











impurities have been ignored by many researchers, but Marshall et al.

(1986; 1987), Nawrocki (1986; 1987) and Sadek et al. (1987) have shown

that these impurities can indeed affect the chromatographic process.

Verzele et al. (1979) have pointed out that numerous transition metals

can be found in the silicas that are used for chromatographic packing.

Sadek et al. (1987) described that silica can contain metals in three

forms: surface species, internal and secluded. The surface metals can

form surface metal hydroxides which can alter the surface pH of the

silicas, while both the surface and the internal metals can participate

in coordination with the solute or solvents using their orbitals. The

secluded metals are too far away from the surface silanols and therefore

do not exert any effect on the chromatographic process. According to

Verzele and coworkers (Verzele et al., 1979; Verzele and Dewaele, 1981;

Verzele 1983) and Sadek et al. (1987), by acid treatment of the silica,

most of the metal impurities can be leached.

Chemically bonded phases are silica particles having their surface

derivatized with alkyl chains or other functionalities such as cyano or

amino groups (Dorsey and Dill, 1989). The surface of the silica

particles can be derivatized by reacting with organochlorosilanes or

organoalkyoxysilanes, and the most common synthetic scheme is called

"monomeric reaction". In this type of reaction, a single siloxane

linkage is formed between one silanol on the surface of the silica and a

monochlorosilane or monoalkoxysilane. A simplified diagram showing this

type of monomeric reaction is depicted in Figure 1-3. The popularity of

monomeric reaction is due to the fact that it gives a uniform and

well-defined single layer coverage of the surface of the silica, and

also the reproducibility is above other types of reactions. The maximum





















c'm


ce)


O
0m




+-


0
I

* -


C)
0-


CO
- C


O
4-u

0

am
C',











number of silanols on the surface of silica are predicted to be about 8

pmol/m2 and only these silanols can react with the silanes (Unger 1979).

The most common Cg and C18 containing silanes usually produce a surface

coverage of approximately 2.5-3.0 pmol/m2. This means that there is an

immense amount of residual silanols which does not react with the

silanes. These residual silanols are accessible to solutes during the

chromatographic process and can cause adsorption and mixed retention

mechanism. A second bonding procedure is usually employed by most

manufacturers to try to minimize the residual silanols. The second

bonding procedure often makes use of trimethylsilanes to react with the

remaining silanols and is termed "end-capping" (Lochmiiller and Marshall,

1982; Sadek and Carr, 1983; Evans et al., 1980; Dewaele et al., 1982;

Marshall et al., 1987). Other methods include using silica pretreatment

(Gobet and Kovats, 1984; Marshall et al., 1984), and the use of more

reactive silanes have been reported (Buszewski et al., 1987; Lork et

al., 1986).

Much research has been done on improving the synthetic means to

increase the surface coverage or bonding density of the monomeric bonded

phase (Buszewski et al., 1987; Golding et al., 1987; Khong and Simpson,

1987; Sentell et al., 1988). This striving for higher bonding density

originated as a way to suppress the residual silanols, but it was soon

noticed that there are other advantages to high bonding density. First,

by increasing the bonding density, less silanols are exposed to the

solutes and thus eliminates most of the adsorption mechanism due to the

silanols. Second, the stability of the stationary phase can be raised

by higher bonding density because the dense alkyl chains protect the

silica from being hydrolyzed.











When a di- or trifunctional silane is used instead of a

monofunctional silane in the bonding process, a polymeric coverage

frequently results. The polymeric network is a consequence of the

reaction among one or more of the leaving groups of the silane and the

silanols on the silica surface. Sander and Wise (1984) have noted that

under a controlled manner, polymeric phases can be very reproducible and

have some chromatographic advantages. Wise and Sander (1985) showed

that these polymeric phases have shape selectivity on polycyclic

aromatic hydrocarbons and are similar to liquid crystalline phases used

in GC. They proposed that polymeric phases are more ordered than

monomeric phases.

The microscopic structure of the bonded phase is very important to

the retention and selectivity behavior of the bonded phase. Many

research groups have tried many different experimental approaches to

gain better understanding of the nature and molecular organization of

the alkyl chains that have been grafted onto the silica. In the early

stage of the bonded phase, a simplified model was suggested and is shown

in Figure 1-4 (Melander and Horvath, 1980). This model is not very

realistic because it assumes that the alkyl chains are stiff rods

without any internal degrees of freedom. This is contradictory to the

corresponding disorder structure of alkane at the temperature of

interest for chromatography. Also, the "fur" and "stack" models in

Figures 1-4b and 1-4c imply that the alkyl chains are exposed to the

mobile phase, but the mobile phase used in RPLC is of very high aqueous

content and the hydrophobic effect should prohibit this total exposition

of the nonpolar alkyl chains to the polar mobile phase (Dill 1987).







21






















i







a b c










Figure 1-4. Early models of molecular structure and organization
of the bonded phase in RPLC: a) "picket fence";
b) "fur"; c) "stacks" (Dill 1987).



























mobile phase


Without Solute


With Solute


stationary phase





Figure 1-5. Interphase model of the bonded phase in RPLC proposed
by Dill (1987).











Dill and coworkers (Dill 1987; Marqusee and Dill, 1986a; Dorsey

and Dill, 1989) have proposed an interphase model for the structure of

the bonded phase. Figure 1-5 is a simple representative of this model.

The alkyl chains that are bonded onto the surface should have the same

molecular structure and organization of other similar interfacial phases

involving chain molecules such as micelles, surfactant monolayers,

bilayers and microemulsions (Dill and Flory, 1980; 1981; Dill et al.,

1984). A typical interphase has one side of the chain molecules

anchored to an interface, and the disorder of the chains increases with

the distance from the interface (Dill et al., 1988; Brown 1984; Cabane

1977). The bonded phase in RPLC has these characteristics because one

side of it is the siloxane bond between the alkyl chains and the surface

silanol. Gilpin and coworkers (Gilpin and Gangoda, 1983; 1984; Gilpin

1984) have shown by using NMR measurements of the T1 of 13C labelled

alkyl chains that segmental motion of the alkyl chains is a function of

distance from point of surface attachment. Sander et al. (1983) used

fourier transform infrared spectroscopy to study a series of

dimethyl-n-alkyl bonded phases ranging from C1 to C22. Their results

showed that these systems have disordered chains with kinks and bends.

The molecular organization of the interphase is dominated by three

factors (Dill and Flory, 1980; 1981; Dill et al., 1984; Marqusee and

Dill, 1986b). First, the molecular organization of the interphase is

restricted by the geometry of the interface and the density and length

of the chain molecules. Second, under a poor wetting agent, the

interphase will try to reject most of the hostile solvent molecules. An

example is when alkyl stationary phase of RPLC is under aqueous mobile











phase. Third, in accordance with the attempt to gain maximum entropy,

the interphase always adopts as much disorder as possible within all the

constraints.

The interphase model of the stationary phase is very appropriate

with the partition mechanism of solute retention. In theory, the

partition of a solute molecule into the interphase having a fixed

surface density will cause ordering of the alkyl chains (Marqusee and

Dill, 1986a; Dill 1987; Dorsey and Dill, 1989), and therefore solute

retention is entropically unfavorable. The partition of the solute

should be directly proportional to the surface coverage of the alkyl

chains until it reaches a maximum where interactions between individual

chains become significant. Then the partition coefficient will decrease

to zero when the maximum chain density is reached, which is roughly 8

pmol/m2. The work by Sentell and Dorsey (1989a) using bonded phases

having different bonding densities has confirmed this prediction. Since

the end of the alkyl chains on the bonded phase are found to be in rapid

motion compared to their attached ends, the solute will prefer to

partition to the free ends of the chains. White et al. (1981) observed

this type of preferential distribution when they performed neutron

scattering experiments on similar bilayer membranes.

The importance of the interphase model is that it predicts the

essential of the stationary phase effects on solute retention and

selectivity in RPLC. Dependence of solute retention on the surface

density of alkyl chains can be explained by the model. Selectivity due

to different molecular shapes have been widely noticed (Wise and Sander,

1985; Tanaka et al., 1980; Tanaka et al., 1982; Lochmuller et al.,

1985). Using the interphase model, this shape selectivity can be











considered as arising from the fact that more free energy is required to

align solutes having shapes parallel to the alkyl chains compared to

molecules that have shapes normal to the alkyl chains. Sentell and

Dorsey (1989b) measured selectivity of six four-ring polycyclic aromatic

hydrocarbons (PAH) on stationary phases having bonding densities from

1.74-4.07 pmol/m2. They found that the selectivity of the PAHs increase

swith the surface bonding density. The increase in bonding density

causes ordering of the alkyl chains and hence elevates the shape

selectivity of these PAHs.

In this study, the partition mechanism proposed by Dill and

coworkers (Dill 1987; Marquee and Dill, 1986a; Dorsey and Dill, 1989) is

tested against an extended data base. Results will be presented in

Chapter II to show that the partition mechanism is correct and the

solute-solvent free energy of the binary mobile phase and the solutes

can be obtained from chromatographic data as predicted from the theory.

The ample number of stationary phases for RPLC available today all have

distinct retention and selectivity properties. This is likely exerted

by the different stationary phase structure and properties on the solute

retention mechanism. Many different stationary phases were

investigated, and the relative retention strength of them were found to

be related to the phase ratio 4, and the functionality of the silane

used in forming the stationary phase. Results of this study will be

introduced in Chapter III along with a simple and useful method that was

derived to classify these stationary phases according to their

retentivity towards all solutes.



















CHAPTER II
SOLUTE-SOLVENT INTERACTION FREE ENERGIES
IN REVERSED PHASE LIQUID CHROMATOGRAPHY

Introduction

Binary mobile phases of water and an organic modifier such as

acetonitrile, methanol or THF are most commonly used in RPLC. From the

partition model proposed by Dill and coworkers (Dill 1987; Marqusee and

Dill, 1986a, Dorsey and Dill, 1989), the equilibrium constant, K, of a

binary mobile phase with components A and B (a case not discussed in

detail in the previous chapter) can be defined as a simple quadratic

function of the mobile phase composition and the binary interaction

constants between the mobile phase components and the solute




ln KC/AB = (XSA XSC) + 4B (XSB XSA XAB) + OB2 XAB (2-1)




where 0 : gB 1 is the fraction of the mobile phase sites occupied by B


molecules, C is the interphase stationary phase, XSA' XSB, and XSC are

the binary interaction constants between the solute and the components A

(water), B (organic modifier), and C (stationary phase) of the

chromatographic system respectively, and XAB is the binary interaction

constant between the mobile phase components A and B.

The quadratic relationship between the equilibrium constant and

the volume fraction of organic modifier has been previously reported by











Schoenmakers and his colleagues (Schoenmakers et al., 1978; 1982; 1983)

using the solubility parameter theory (Hilderbrand and Scott,1950).

Dill and coworkers (Dill 1987; Dorsey and Dill, 1989) have suggested

that instead of the quadratic equation, a more useful linear expression

can be employed to plot the dependence of retention on mobile phase

composition. Since k'/kw = K/Kw where the subscript w refers to the


mobile phase composition when B==0, equation 2-1 can be rearranged to




1/0B In (k'/kw) = (XSB XSA XAB) + PB XAB (2-2)




As stated in Chapter I, the binary interaction constant X is often

determined by measuring the free energy of transfer or vapor pressure in

the Henry's law region (Hill 1986). A careful examination of equation

2-2 reveals that a plot of 1/4B In (k'/kw) versus #B should be linear if

all the assumptions made in the partition model hold. Also, the slope

of such a plot should give the binary interaction constant XAB, and the


y-intercept at =B=1 should be equal to (XSB XSA). This predicts that

the binary interaction constants of mobile phases used in RPLC can be

obtained from chromatographic data, and by applying equation 1-7, the

solute-solvent free energies between the solute and the two components

of the mobile phase in RPLC can be calculated.

In the original derivation of the partition model (Dill 1987), the

solute was treated as having a size comparable to the mobile phase

components. In actuality, solutes in RPLC are frequently larger than

the mobile phase components. Since the cavity term is very important in

the partition model, solutes of different sizes should create cavities











of different magnitudes. If the solute molecules are larger than

molecules of the mobile phase components, equation 2-2 can be redefined

as




1/QB In (k'/kw) = n [(XSB XSA XAB) + 4B XAB] (2-3)




where n is the ratio of the size of solute to the size of the solvent

molecules. This expression results because the number of contacts of

solute with the mobile phase components and the size of the cavities are

directly proportional to the size of the solute molecule.

Theoretical derivations are not valid until they have been proven

experimentally. With this in mind, we tested the prediction of the

composition dependence expressed in equations 2-2 and 2-3 with an

extensive data base consisting of more than 300 sets of experimental

retention data of various solutes on various RPLC columns. We hope to

uncover the underlying physical nature of the slope and intercept of

this type of composition plot. In agreement with the above predictions,

linear dependence of In k' on surface area or van der Waals volume of

solute molecules, and hence the size of the cavity, has been widely

observed (Jinno and Kawasaki, 1983a; 1983b; Funasaki et al., 1986; Arai

et al., 1987; Kaliszan 1986; Mockel et al., 1987).




Experimental Procedure

We tested the above predictions with a large data base that had

been previously used for correlations of solvatochromatic solvent

polarity measurements with reversed phase retention (Johnson et al.,

1986). The data base is used here in its entirety, with the omission of











six data sets which had retention values at only three mobile phase

compositions. Additional data sets generated in our laboratory (Michels

1989) with ethanol-water and propanol-water mobile phases are also

included.

Linear regression calculations were performed using the program

CURVE FITTER (Interactive Microware, State College, PA, USA) on an Apple

II Plus 48K microcomputer. The program was used to calculate the

coefficients of correlation, r, and determination, r2, of the data.

This program was also used to extrapolate k' to the composition of pure

water, denoted kw. The In kw values are not usually reported in the

literature as they are not easily experimentally determined. We

extrapolated plots of In k' versus ET(30) polarity to the polarity value


of 100% water to obtain the In kw values used here. This method has

been shown to give more reliable values compared to other types of

extrapolation (Snyder et al., 1979; Dolan et al., 1979; Schoemakers et

al., 1979; Antle et al., 1985; Baty and Sharp, 1988) because the

calculated In kw for a given solute is consistent among methanol-water,


ethanol-water, and acetonitrile-water mobile phases when using this

method (Michels 1989). The F-values, which are used to compare the

significance level of two variances, were obtained from the StatWorks

(Heydon and Son Inc., Philadelphia, PA, USA) program on a Macintosh SE

(Apple Computer, Inc., Cupertino, CA, USA) microcomputer. The van der

Waals volume of the solutes were calculated using methods reported by

Bondi (1964), and the hydrocarbonaceous surface area was obtained from

Woodburn (1985).











Results and Discussion


The extensive data base presented here permits certain tests of

the cavity contributions to the partition retention model, and the

ability of using chromatographic data to calculate binary interaction

constants. A typical example of this type of composition plot is shown

in Figure 2-1. In the case shown, the solute is 4-nitrophenol and the

mobile phase is a series of different acetonitrile-water mixtures. From

the figure, it is obvious that this plot provides a linear

representation of the experimental data. Table 2-1 presents an

extensive compilation of the results of plotting data in this form for a

wide range of solutes, stationary phases, and various mobile phase

mixtures commonly used in RPLC. From the correlation coefficients

shown, it is clear that the degree of linearity of this type of plot is

extremely good, with only a very small number of exceptions (Figure

2-2). More than 80% of the data sets are found to have r2 of 0.9 and

higher and almost 50% with r2 of 0.99 and higher. An F-value, which is

used to compare the significance level of fitting the data sets to a

quadratic model (Anderson 1987), is calculated and expressed as a in

Table 2-1. The results of the F-test reveal that the quadratic model

does not fit the data sets significantly better (at the 90% confidence

level), except in 148 out of 346 data sets. We conclude that the linear

fitting of the data is adequate. The principal exception in the data

set is the Sepralyte C-2 column with methanol as modifier. The

coefficient of determination, r2, for this particular data set varies

from 0.4460 to 0.9162, with an average value of 0.7287. These poor

correlations are probably observed because for the short-chain

stationary phases, adsorption is expected to be the dominant retention




































0.4 0.6 0.8


Figure 2-1. Plot of 1/OB In(k'/kw) versus OB for the solute
4-Nitrophenol using Hypersil ODS column and
acetonitrile as modifier.


1/03 ln(k'/k,)





























Frequency


0.950 0.960 0.970


0.980 0.990 1.000


Figure 2-2. Histogram of coefficient of determination, r2, for the
plots of 1/#B ln(k'/kw) versus PB for all the data
sets.


























0 N NC ot o,
N0 N *no ttt N



NHIn o 4-4i4


o a- to 0 a N C! N



ON in a- N t rN O f in Oin Oi N


oooooooooooMoo
0 ON N W 0N I i


C ~ ~ m ~mOONin' toNinNt -


0 0% tl)O t t N
i~lN CO -N 0


W' 0 ,-t W ^C to0
0OT 0 into CO to





to 0 ^. in en to 0%
to) in N* i-C en in 0 in



O~msmoWW
N e 0% 0 min en e
0 0% to en in to
en in *a' 1' ^' NT


0 to to 0% 0 to> to
to (N N Ct o to in i
ij N0 in 0%0 N? N3 0000000


t.-irtt oe-4 oN WN OOOO
.toonn- -m4 Nmo WON,-en-M
en int *O NO *< r-1 *O *H D
r~~~m ~ ~ r~r


I













O t
i1












oc























0
w,


.o. oooo

0000000


0
r 4) >1
0 a .-C
0 c 0u>,

00 004 n. ) 0 0
01~ oi (U N C ^ c C acu E:

w C 4) 4) w w 2 D 0 w n
:3 N 0 >1 1 4>4 >I 0 0 0 0 t 0
M0 0 r M X -lX-l C C 0 V .:0 0
r-I I 0 4 0 0 > i i 1U 1 1 1 ) C<
>,0sQi U 000 ) >,0c cx0 -.fl0 (uiai ca

JOX: C : 0) ^ N4 = >,i4 >.),-e *N 00 0 i-C O
-4 0 C > l 0 4-i l I I I 0'-tl xC Od -
m C a, m -HC*] CL 0.0 E 0l C.u m -Z .


O-4--n-- in-o-o---o o
-4Nen~ito~o0% 4.-,-"C, ,- ,4 ~4 U


in N to toN ^ i en NO0N in
0% to 0% 0% 0% 0% in N to to N t










U





























j .=



LALO40A00





.0-04( 4 0-

d OO~~O~ (1o-n-(OA"4


rooooooo oo oooooo





ooooooooooooooo


















a .. ... 0. 0..
000 ON ( t0 0*0003000













0( O ,OOO 0 o(1( -o0-(--






tn OLA 40 000 On 40 0(10 OL















o 000 004 40 N





-^*^












-.- I

4), 1


C)
C) r-
C) C)>

C C) E X.

C) C W (a


o 0- r I x: ( a 4a^-
Cr) 4 I ,

o : 0 rV t l >4 0 0 i0. a o VI a
0.0 0. cCx ai 0o0 N >41. >4>4-.-
0. 0.O C) a a >3 03 (0-c- 0 .0. c 04


- 0 (1 n LA : ON 001 0U (1 nJ I 1t
m z~i
rO-i ~ T f~tinA~OcT'iri^t


1`00 MU "
C r- (AnLn C)
*O r C *\ D 3
Ht-4L0A -rrt^ nr



C)
















Ica
"Mmww

























On ca mA M m ao
0














































0
0
a
C)

. (1 .O a

r Ln r- 00 n^^
CM 1-4 4-4 4-) *- l- l*
0


Nq rO Lo n1(1 C)
r n 0 (r 0( LA3 .0

000000 0
\oooooo "Q]





CM r r^OD r- o
. .0 .0 .1
O 00000 m




0(100010



0 N 00 N 0
11



ooooouuo "
C)



(4-
C





t)0 00000- r
000000








>4
C)



m




.0


>4

(0
C)











0
mmmm a>uc(Uc





ICOCN~r (U V*

* ( C) ^0r)30 i4l

*..4 0 0 E 0 a
. C ) -..0 a0- 0-
.0 C..t U CD >-
E
C)









3
(1(1(1) U
> ~ n r-o wor-
i O nn r r

































n n


S'. o 00 i a In









oorooooooooooOooo









E ..... .0o -4
,., 0.







































17
---42 -4zI 0000.
In .-4 4 0 r- ,-4 In N4 to' I n 4






o c fl In N In u 00 W 4004,00 I

0000000000000000





0000000000000000


































"r r
OI 00 0o t oD N to In >I 4-
S c 0000000000000000





o In 40 0 In -. I N 4 0 I


























0
I? DZC.< .U[-C f UC i -


w









0 0 M0
01
tJ












I 0









w ooi ow oo E





r o o 2 woo o
F 0

























o
In r r o in o
00 00 o 'mf (








-I

























000

4040000 l
0004 u uu




0 4 -4 0 *0 I)

In404U0400 0
J3C^U XIh-
o~o~~unc































z: z= ::: ::2 =z-






C :2Lenr- 11 rC I1 a 0 0
w c 0 w w
-N *---L-N -INW -- -- -









C 0

.-ee in.r- [lon 0 oe n






oon 00 "H 0
o a0 a T .- aNCI N a CN -* 4.00 0 =




LAW 0 00,4 n OA
oo ooon ooINI* oj NLAaooo 0oo oo (m ooe oo














o 0









o m mo wn n N H en d a m r-
w tr' -ti-r(-r i-f r- wwr- LA
N4

















0000000000000000 0000
-4

















.j' *^ cr in rs o r a o
0 0 (. 0 .. .I
f0













WN W ( rN r --r-Wtr- -t r-;W C 0" V X
mI OwO OcCL ,"I O DO Q









































w cO',O W90 Ow\4 0 D
r 000000000000000 00000










































m > C IOI a )O 0 f I I I i ( 0 -o
C 0



















0 -4-- --- --

01 r- r0































f\ *'3 01
E--
icir rr-rrr 0w W
0 0 .1














































4 0 ,
oo






M ~
a,.0 m4 0 w Nmr- 9: *r- *a,




0









7 c
oooooooo ooooooooooooooooo





000000,0 00000000000000000






0




m
oa a o a o
0 ^ 4 0 tM 00 a' a
























3



S1
Ua
4) u
















































r a 0 czc c o 0'0I0Ia,0-o c 0w w 0
SEo












Mo o o, o
































2222 mmet St 8. %
o0 000000 0000 0000 0000000 1
mom em em mclme madH a 0
00 0 0 0 0 0 0 r-io 'r ^'o0o o o r^Ln a












fa ..















c q\0N0'0' o .'0',-4~-4- -
(U q i
> "< '^
-' c










*Oa (Uo















01 UC ~~III(U~ ~ (/3 <















































C)
co ifl c r- o \o

ON .O Ct ca 0


m ~~m 30 1r
0 OONO)NCCC


2 N r-r -a 'a 'D Ofl NN


II










S00






Il N


ON ONON
So00






o oo


0 N
o





























oJ
0


CmO NI (N OSCOONCOONO
CO~ N ON ONOWWONONN~f


.4 ONWNINr fO(ON-C
Ca nCONNoooCOC

ON .~cN~Nr-No,-
- (N~-C~N~ C,-C,


oo N OON(Coo NCOON(N(NC~d


00 0 0 0000000000



(Noo o o ooooooo-iooo
CaN Ca iC COONOOCCAONO(NCN


0 -Cfl *~C






Ca N~~ONONONNO



0 0000000000


N NN ~O N N 00NNmoom 4d


0N 0 C0 N U (rNONmONCWNNNL













Z 2 2 ZZ 22ZZ
z z z zz zzzz
U U U Uu UUU
< < < 44: < 4 4


o o E o
O -I 0 0 --< C
o >. C -1 -(--

I_- a Halu 2or rc m cm

0 0OE O C _
C OC0 I C-' 0. O i
0) UC Q 0U 0 W a -.-I-C. 0C0
S C I '-I 1 .- C -C tO
a0. -.o o -d Z I -X X-CIF



S-,. I CT C I C. 3 1 I 11 (U


' CNn eW(N'a0N CZZZZZ(NN



ON 0.-'(N W C Ca NONON0t -C(NNN'(fl
^ 0 ^- r- .< -f -1 ^- *- rt .l -tCslf( N C-C(N N(N( (N(


S = S S S


PP ~

mm o
o*o\ or

00 0


















n

















a4 a: = 3 = = = 3 = a: a= m
S41
tL 0





















































0 In0
,




























000 eeen 0-00 e n 0
(1 fl ,e n 0l l 4fl3f0en





z 4er-oD %,in40inri r f- iY00e in4 0















0 04
t r x 0 0 t CL IT 4)
r*
















400 00o0o 00
0 >


H [
oo o o o o oo o o o o "1
g mPmO ~ m~P~~














0 z UUUz OOU UU o
W <<< <:< v, 14~: ~Ul
3 n,









l- fn ^ n)--<^^ n oc c:c c o'o' -fa o
-f~( ~OEE O~~ ^-t~ ~ ~ -^>l .xam ~ c e 1 ,C?<-l-H-4tt .<-t r -- ^^-t i>
+- C flo C C O^-^^- ^<^ a >,,^3 >X <
C~ ~ E~~U OOO0~~ Ej^ -'^ c>~QQo>3-
U Kro o~~uc~ru O U^^^UU >l~^










Ir0











40
























-e 0w 4 mar- 0
a3













II
4.D




Sw
















4 00C 0nr *-
4 ro ioo(oo ooo ~oo ooo








S O a Win i-e CN- OO M O- 0- -

a0 OCO .o l






&4 -4





















o n 'ro ('4 Ce o o p- _cM i3-
OOO00000000000
0 >4






















4-j-
I i- 4 C.4 ('4.-.". 0. .00 en in














Ei E > o o
0





nr















0E
> 0.0 ^. NO .0 0 0 0














) >i o
^ :X >^ 3 --O Q *- =
*-' j *J Q.*J u 'u' c : n)ni C >,
0~~ o u ) N N a)(i c-<'4 -i-^-^c:c:ff Q
ri ee c ,E ->i> io o o o






(0P
E-*)

























ser
nU


00 .0 .rNO
~* LA N 0 0


21 m0m% 0 %rN-


0 L 0oooo 0oo-of
o oo0rn0ooo000L0oo
en oooooooooL





o- ooso^oooo oo t i
*n 0%LDW0%C'tLr o %





o OOONr-NNNN

0 000000000


vi N NNN O r c O


0






0 0 00 0


0 U U 0U U
(n : < a: < <


0 000000U000


S == 1 SS = =














10 W L 000%e

r- LA n r- flO0 0' -4








o e n rL i0 o ooon
c^ r- m so

00000000

Nooommooo

O% 0> 0%> 0o 0% 0% 0% 0


'01 0 OO0NONNNN
0% ,4 .4,LLAOOO00
0 ~ 0LAcrLA0000
0% 0% 0%0%0%'%N0%0%


00000000


0 00 0
mm0




u uo u




0 0 u 00 000


0 0 00 00 -C 0
000000 0 0 c 0
S- >0 >0 0 A




> :>14E E E E E I- E-4 El
S1 EgoEEEE0 0 0 0 1 1 0

-e. -e i f o- LM I M 0. 0.0 0 P0 C nt p r r

L NMl lNNNO' N eONNN NAN N aNMNN N



~ ~~~~~PCPP P P PP>>@@WW


>,
I















i


o


"y



oC
3






0

uBc(
ru''






















m
r(

4.-
(4

i



.c
z



a 0 =


SnoO r: o o rn o
0' 00 N 0 N N 0
40 .0TN 40 i
N4 oo 0 N 4 N ( N o
NN44 0( [-N ^ M (N 4l 44< (N 40 44





i NNNNN o N 40 n 0 4n w


-N 0 4 4 N -
I 400N' C 0 oo 40 N N 0 N' (N
40 4 N Ns-trr 40 N 40 No 40 o 0


o0r4in4 0 in )
in *o r-4 o m

4 ^040I oN 40
rllr4r4l4r r4


00 00

N m


o4 a








odood
a c0 0
a, 00 -I





00000
(N





0






0


S
a

0
t14


44 44 N



0 0\ 0

0 0 0
d oo





0 0 0
*-




o


So <-i
N' 0 44l
0n 44< 44


m
in
sr
(3N


-4 m0 00 O 0 0

(N 0 4 N' (N I 0
N' (N N' N' N' N' LI


(N 1 z = =


-4


uuuuu U 0 .


CD IC.)C.) C


0 C

0 0000 0 C
w 00 0 0 0 0 0 0 X
.00 04 4 ( (4 ( ( 0 Cu
,aaaaao0 0 0 0 0 0 '0 0
000000 0 0 0 0 2- 0 4


Ui u U U U E E El p E n
000 0-0E-l H F- l H H L l|
N LI 40 1N) 1 C -) 0 0 N' 0 N' 0 (

(N(N(N(N( N( a(N a(n a(N a(N a(N a(n an



444444 0o 0 0 0 0 0


44


>2


I



T t


0










0
\0O in 00 o
CN4r in .( 0



















CIO.
0

44
\Lin N N N





CMm m oo mn -r
00 40D 40 4 a

0
ro
4
(0
*o
0I (N 0 -T > 4

o oo o oD En
uM o i io 0
0N4 0 -i r-l -( L(



000 40 00 00 0
( M N m ( (N

00 0 0 0 Z

r: u a) 0
oo in n o z

00 m a
CT^ C o <\ C\ (7\ '

o o( oo


.0 40 (N 0 _

0
U










0 0 0M 0
*O a\ \D n D o\




~m 1-1 o CT o





00 0 0 0 -c
>2
c






>2

040D in 40 N
mo, oo m cor-t






? ? I







= I) ( 0 0
2 in Z s z

0 U
< < o


M








; C : X: X
1 0) 1) u



in CM



0 0 0 E

C( CI) C(( CN'


) r O cri o( II
0 c>1 a I o
4 0 r0 M


oLI~o~ C.)


L 00.N a~ a~ a


Nm 0r 0 4
2 00o 0, 0 0
444 44 (N U

























4,

(H
0t
o0
:4
:4

*0
C: S : -
40


d


> I

















m



















in
0







o
CM

HL


a = = z z z =


00 o





















a00* o ooo o. o 0-
. T n oO .. i ooH-,- *









0'
r-p




























oooo ooooooooo oo o o o oooooo o
4J

. 0n . 0 o
















o o o om o o ooo
S. *. C
































0
t,






0 oo
0mp" 4 0o "o"0'O" 0a0 Ho in aiO DoN N m m m
z z vn z z z z z z z NO o i 4In No in '0 0





























000000000000 00 0 0 000000 0
CONOrmflm-4ONOCNON 0 0 (N 00 oH N40-4-4l N (




4-r
















000000000000 00 0 0 000000 0a
oHcONON ON-4NN in N 40 oi-- N0 No a





























0N0mm NN nN 0N r 000 0 0 v 0N O


ooooooooo oo o- o o o o
(NON ONO .-OfNf 40


















NeON OmmO- -NHO '0 ON ^0ON4 N 0 >







0 0 0 00 00 0 0 0) oJ 0
ON ON ON 044 ONON ON 0N ON ON

0 0 0 00 0 0 0 0 Crf 0









O00000000000 00 0 0 O 0 C


(NN((NN((NN((NN(( ((N (N (N (NN((NN(





























ON


o
o =

N0 ~N0O
*r *I(( Imuu n
40 .0o-nm--oN (N
LI 0-4In(N(N(N.I


0 ONr-o o n 0 Nor on Nn ON
ov o i ..o o





m r e N 40 ON n .n -i ON

SNNN 4( 4ON40 400
(N (N (N -4-4 ^ -4 -l -4 -4-


0000000000 P00
oN ON i-N---4-i N oN


-4ON H~4 N 00I
0000000000

9 in (iN m ON ( ON ri 4040
(N ONmONmONNNOONNOO
L(oNONO~NONoN~oNOOOO
0000000000


oooddooooo









000 000000

















0


) C0(U
m 2) C 4 2)

C r ) Q ) 4 a)C



42m() C .-(04X(>4 N
442 2 M C C 0.C 4 ) X I




4 In N EO- WrON 0 N I n
(N N ( (N(N N n I n nI
(N(N(N(N N(N ( ( ( ( (


a 0 ( ON

442
000








00
u







r-0 0 I 0 n0 0




N 00 r 00 C

.A I rH N 0

Inmmm0 0
ue
00000












ON 0 O00 N
S040 0 ( S InN
- inO nONl O C






























COO .00.0 U
aI
omrl~loo
t-<
;' 0 0) @
C C ffl 0 E
t 0) ) C f) C ^










N 0 fi 0

I ; .10 0 -










m )m no n o
I s)( C C N (J


Ii







I


4-
0
0 q
c

0









SE



0





*
dP




0
U)





























in








Ct. = = = = = = = = = =
N- NO V N ,H-O






N..C .0 .N .oe.n
t-i o~e N O N 44 O e
a.
ff S

n ~ n i ~ -iD -o (

r~~ \ \ T OO ii -io
r^o *^ 'o


I


ON en oi m O NV rO O N N
.NO en .N















oNoNoooo0o ooooooo


enONONOmmO NenmeON N
Na OI nONNONOoj 'OQOCMONO


00000000000000





( c r-sn rooddoro oo o oo tp









Som
0.











aa


N dN00 NONOONO0 00 WNONONO





O n ON 0 O m O C E 14 r
00000000000000





0









.k NNN NNNNNNN




> S N: = 33
i 000 0 000
o 0 o a a, a aj ,














C a, NX:
fl) N 00 aNa4
0a Ca N aCa, .0)




0 .: C0 : a, i N 0 >- .0 >i>. X0l N

ui 0ON a CE-.141 0 ao E
I C r -4 N r en V ON ONNO O Ne
V VV VV VV VV ONON N O O

N NNN I ~ mON NNNNN


0)



11 ra

4-


















7




















oo
a

































ra co car) c
IL








































oj o\ t-i a) co us
N o ON N n o o
an o n or- Q






ono.0mr1en 4a







S00 N N en\ 0
LOOONOOO 1a)


rOi 3 in ON en o .0
SV en enrn e












oo ON Ln en ON
0 r^- 1O1 en o




0->
-H










poONONVV




w o crm o o o
o PO oo N On ON














*-H
ON\enON0en



0


ON0 ON00 ON3




U













>4





^ a = = V


a,
>4
Ca
ffl


Olm~~0010, 1



















000Wa,




a0a0a0 ON)
N 04 a, 0 a,
SOCN N V
a,.a)0c,0Nc ,

a,4^ -4 oflio ON
*N 0 0 a, 0 4.
.a,),-4.04 M 0-H ON
C



LnLnn nn o














I0
ONO O N -
Lfulf~lf
NN N





























rO




Ca a = = a:: a = a = I = a ss s a -


*0
0
0
a:


0 0
0 0 0 00
0'- N N ^ N ON N ON N Nn ON- N
o1SCNmrNcmHnHocnmn
fl N r c *!^ 1" 1^ '1')


ooooo oOD N



000N 000000 N000







No N No H NoN N oN
00000000000000


SON O o O N N N 0' O m
0' 0 im ron in in inN N


it










































'-4









ON


I
0






U'
C


0)
0 CC r
Cl 0C Q
Q 0 0)) w N C
rC CCN CC r fl)00
COO a) 0 rC 0)0 4 E
a) 4 r (D <1 o 1-
w v m c 4) (V 1
M- a)^ M >, 4 r ^ C Q ( (D 4) 4, l
(1 ) cJ M r k a) a i) 0i c c
Hca) Co 14.0 "4*4


SC rc >- I a) 0 1 1 1 1 i
410( .0 it c tiH ~C CQH.MC QXXONE-




ON ON ON ONOLN ONON ONON ON N NN N
N NN ~~~ON NNNNNN


(U
0







S 4




00





lo





mmmmm
moNON-r--O N









0
14















v
In o N NNN
000000* Co

























In






c x




o o rio
in o o N r ~ 0'
moooooo~u c









































r- o 'an^o *
-0-























ca
N H ON N









i or o r



































U
4-4
inN N(v O 0 NC
03 o r- oo oow











oooooo
4J
C
0









.1
U


C)



a a C aVO a a
U














4'H
0.



U'
CC






WC O
C00.00.0












C~,-N~f44-
ff
d





























NNNNN 0
NNNNNo Q
N
a a
0 UC
a> J j3a) c C 0 ( ^







ct

m or o -







CNCM M(UN P'


SN N N O N N O O O
00000000000000















ON



I S an s a










47















C 00













---- -
oro






c
a c: ccL
o ^



ll o a o oo oi ooooo [ ...c M T caa
E-1
S 1




















0

,, ~~ oo~~r ~
r- m m oor
ON w .1
r(0
NO4 UONONONON00~mP

44



C
(0
S0







































o wy-rf n- n^n o^^ nu o( o 0 4 n f w o 0
CL m a) r ONr O C 0.(NC4ONONr ONN50r.
4 UN CN Q ON"n 4 o n UONCL 2































4j 4j o 4j 0 >1 w 0 1 0 a) (na o
XI
C
































2, j NC N:J t
0 zr 0000cN00 (000000000rNQ '0,0,000000 0


















a, 10, MQ) m w o I I z w o- w 0 11 1
f"N~NNm~m~~~~ -





0
aAN) o N0(0* (NN N n t cN' *N0O O.( 0 0
ONUON(N00NONON~~ONONN0 0,,NUNNNCOO 4











oooooooooooooooooooooooo 7
mON,-00.-ONU ONNP4~NN,0(NON00ON.-4 0












o' W ,'U U 0 N U ( N O N N O N'~0
Sn0))(00001()0000000000000000000m



000000000000000000000000 0














4- -l O O N0,,(O N0O O N0 0 0 I 1 1 NONU
0 .0 0N A +J 4J N ^ U U
N 4O O N N ON-ONONUNON NON.-NNON













ar


*o -
ON ONON OO ONO) ON N ON O ONNO N N ON ON ON ON ON ON O
C
000000000000000000000000
0 CN


















I 4j| ^ -tON 0ON ( ON >i ON 0 UN> 0- -4 0 ON ,- ON W OO (N' 0, 0- .4- 4N'

-4 0)l-4N :UN3>i ^INX 3>tl3Hr-
I O
WY = -: .4,4 0P~~~NNO














fOX
0r 00
o 44 44 4 44 44 454 414 0 0
(.No o ON O N ON O .-(Nt0
oo 0.



aI
0

0) 00 0) 00)0 )0 ) 0 )
0)0,-)-- )t U L C 0 000-- 4
000)0 0 0 4 0O 00)0 004.4-I0



14- 0.00-I ~ T~2 rr~Z e 0. ..41 .0oo0 ~ 4 U
I 44040>0)-(4444040(0,-04444444 9
.4 ( 0 4 4 5 1 0 4 N 5 > 4 5-( .
ZZ0-( ON Z040-4 QO Z Z
5- I 101-004 ( I)04)4 i 1 1
a co Z Z NHON0 0 0O O N 0( N 'N
I 0(0NU O N N N.(0NU O N N N(N, .4c~
0) ~ 0~ ONNNNNN~ONOOOOOOOO~ NONOONNoo 0
HEE~ (NNN(((NN N(NNN(((NNN(004N,000 UC~O

































~oo o





ooN0% *r- .1110
a ni~~~p
*N,-0'n*t)0*NO

24- i n ~


ONWI N ..s





N I I m00000 000





r^in It) 10 '00 0 Cf 10
a00 o a'I)t It ) ICT)1
0'li- In In in l rIn N M N In

000000000ua


0% 0D In- I-n N^ N 0' N In
100 N 0% Nn No In 0
N N0%'0InWI0%0%WI

ooooooooo



N- Wt ND N- In In N- (N N
S oo ooooDoo
0 TiLn[
a


0%\ WI WI N~ In WI WI In NM


- 000000000
o




In
r ,





0M 0( In 0 N 100' WI N 0

0' '* rIn In In In 10 N 1010

0


01
o




o ow^
r-I ct 2 0
o Q
M S


00
o N nCoo *C N w o 00 o o
as 0%assa.I :: *0'
.1 .. I .10WI0o0In *I)-4


* .. .* **
n.N .0 *NN .NOOOO *N







EN 0NI N 0a0:O% 100%'00 N N0! 0 o















mmmooo omo oooo3o



ooooooooooooooooo
N In N N N N oI In N N 0 N WI In N N I r
SNi o N N N N In In o cN o WI 0% o0 oo 0 0 N




N In N In In In N N N NNIn o N NNI





















In 0' InN ~N WI 0 II 00 IN II
N '0 N '00 n 1 0 NInIn'00 n 01




- = 2: = = 2 2- :


H
a)
1c
44
a) 0 i

w a w wj 0 1 a1j
z 0 r4
O r0 C a0 I) C w
0 >- 0 1 N c 0)
c 4Z .04 0 r0 W W ON
0) 00000E 01.1 00 00
0 X NONO0O C(U N m c Q3

0- Q. 01.l UNC ,00P >,i ONEOO" MV J O 00. ^-t-'Q
r_ 01.ION .IE 00.InNN14to .00 CN
C .)00)I0. 0) NN 1000 0 f0 00 N >,0 ^lMC>< ^:^
0 0 0 m 0 0 0. .0>, 41 m4 000 0 1 ON.l0 W 0 > 0 4
01= ION N :N n i0 001- .00 1.0 1.1000r1.11.
0 l C 0. in T0 04 U U U tJ X C 41 40 W1 W >
0% d;CDGQCQCOWI 0 0 000 E 0 0.0 1. 0.-) SZa..L C%0%0 O


0'I'0WIONo o' 0 N OO Wi0ONNI n'In0NWMI
OOOOOONNN NNNNN NNNN NNNN
Inn~~nnO~nn~~ In ~n n nnnmr~~nnn~~


Ei












49














S0 0




1,4

c c c P:








o a .







In mO oooaoo oo d o


So oa -m ai
i 404044 4 o 0 1


































TC) o
S000






0i A10 A 4 r0.4 i 4o00
EL ra






























































cNNh CO N r r0 w
0 0n 00000 00r C>1 r cn N c w 0
mn w u Q w mo o10o ujum




























































1,4 1004 0 4 W M0 4 >


---- mmmss mmma 8a
40-
^ N0 n 0 1 c. 1





























c c to 0 c r 0 in~n~LA40 .-e-
U asa s Kssa' i 0
> iq n '^ x n nn .,,>
c: c ** ni~O~P








'-< c:u : --l4J U C 4- Or- X:Q.fl 4 >
Ol B 4 O > <> 3 in 0 --^ ; ; )
1/ C a E3 00E 0 &CUCQtQM E C & <;S *



vr~~ ~ ~ ~ +.1~ON O~~O

y o i-ic~fl ^'n ^ r^ooCm OricNr' n^ *-roc-











mechanism rather than partitioning, which is the basis of the assumption

of equations 2-2 and 2-3. Another possibility is that the residual

silanol groups are more accessible to solutes in short-chain stationary

phases than for longer chains, and they may exert a strong effect on

retention that is not treated in the partition model. Due to the

significant curvature and the poor correlation found in this data set,

we feel that the y-intercepts at =B=1 are unreliable, and therefore are

not reported in Table 2-1. Where linearity is found to be good, it

implies that the partition model, and the few assumptions applied are

valid. This linearity is expected to be limited to the intermediate

compositions explored here, because the random-mixing approximation

assumed in the original derivation is expected to fail in the limits of

extreme mobile phase composition of a few percent of either mobile phase

component (Schoenmakers et al., 1983). In those cases, solutes or the

minor component of the mobile phase may associate to form non-random

mixtures.

According to equation 2-3, the slope of the regression of this

composition plot equals the binary interaction constant characterizing

the pair interaction between molecules of the A and B components in the

mobile phase, multiplied by the size of the cavity occupied by the

solute. Tests of this prediction of the cavity-size dependence are

shown in Figures 2-3 and 2-4. The slopes of the composition plots are

found to increase with increasing cavity size created by the solute

molecules. That is, slope of I/OB In (k'/kw) versus OB = nXAB and


slope / an = XAB


(2-5)


























Slopes from
equation 2-3


40 50 60 70 80 90 100 110 120 130


Vw (cm /mol)


Figure 2-3. Plot of slopes from equation 2-3 versus van der Waals
volume, Vw, of the solutes for the Sepralyte C-18
column with acetonitrile as modifier.
















Table 2-2.


Column

Sepralyte C-2

Sepralyte C-4

Sepralyte C-8

Sepralyte C-18

Sepralyte C-4

Sepralyte C-8

Hypersil ODS

Unisil Q C-18

Ultrasphere ODS

Ultrasphere ODS

Ultrasphere ODS

Ultrasphere ODS

Silasorb C-8

Silasorb C-8


Regression results of slopes from equation 2-3 versus the
van der Waals volume, Vw, of the solutes for all the
columns.


Modifier

Acetonitrile

Acetonitrile

Acetonitrile

Acetonitrile

Methanol

Methanol

Acetonitrile

Acetonitrile

Methanol

Ethanol

1-Propanol

Acetonitrile

Methanol

Acetonitrile


Slope(xl02

16.51

16.61

7.817

9.582

7.286

6.228

12.78

8.640

10.33

5.948

4.662

18.57

3.051

2.595


2

0.9561

0.9263

0.7897

0.9164

0.9189

0.9305

0.5499

0.4106

0.9977

0.2490

0.0495

0.8091

0.2339

0.2083


























Slopes from
equation 2-3


80 100 120 140 160 180 200


220 240 260


HSA (A2)














Figure 2-4. Plot of slopes from equation 2-3 versus
hydrocarbonaceous surface area, HSA, of the solutes
for the Sepralyte C-18 column with acetonitrile as
modifier.

















Table 2-3.


Column

Sepralyte

Sepralyte

Sepralyte

Sepralyte

Sepralyte

Sepralyte


Regression results of slopes from equation 2-3 vs the
hydrocarbonaceous surface area, HSA, of the solutes for
all the columns.


C-2

C-4

C-8

C-18

C-4

C-8


Modifier

Acetonitrile

Acetonitrile

Acetonitrile

Acetonitrile

Methanol

Methanol


Slope(xl02)

8.511

8.785

4.115

5.099

3.817

3.358


,2


0.9642

0.9465

0.8226

0.9280

0.9068

0.9520











where slope is defined as the slope of equation 2-3. Values of XAB

generated from the derivative of equation 2-5 are presented in Tables

2-2 and 2-3. Due to the limited range of solute sizes available, this

linearity is observed when either the cavity volume (Figure 2-3; slopes

in Table 2-2), as measured by the van der Waals volume of the solute

molecule or cavity area, as measured by the hydrocarbonaceous surface

area of the solute molecule (Figure 2-4; slopes in Table 2-3),

represented the size of the cavity created by the solute molecule. The

binary interaction constants have different units depending on whether

cavity size is taken to be that of the solute volume or area. The

linear relationship holds well between the slopes from equation 2-3 and

the cavity size; with most of these plots having r2 of 0.9 and better.

This linearity implies that the slopes from equation 2-3 are

proportional to the size of the cavity occupied by the solutes. The

slopes from Figures 2-2 and 2-3 represent the binary interaction

constant of the mobile phase component A with B per unit volume or area

respectively. For the Sepralyte C-18 column, shown in Figure 2-4 and

Table 2-3, the interaction free energy of the mobile phase components is

found to equal



(5.099 x 10-2)kT = (5.099 x 10-2)(592 cal/mol) = 30.2 cal/mol A2 (2-6)



for these data taken at 250C.

Several columns give poor correlations for the slope of equation

2-3 versus van der Waals volume of the solutes. For the case of the

Hypersil column with acetonitrile as modifier, most of the solutes were

nitrogenous compounds such as aniline, pyridine and amino-substituted











PAHs. It has been shown that many compounds with a nitrogenous moeity

often exhibit an unsatisfactory degree of peak tailing in RPLC, and this

effect is generally considered to come from the presence of residual

silanol groups on the surface of the bonded phase (Wahlund and

Sokolowski, 1978; Bayer and Paulus, 1987; Smith et al., 1986; Hansen et

al., 1988). The retention mechanism of these compounds is most likely a

mixed partition/adsorption process because of the interactions with the

residual silanol groups, and would not be expected to fit the theory

tested here. The Unisil Q C-18 data include many isomers of substituted

benzene. The van der Waals volumes calculated from Bondi's (1964)

method for geometric isomers are identical, but they give different

slopes for equation 2-3. It has been shown in the literature that

isomeric alkylbenzenes give relatively poor correlation with van der

Waals volume (Smith 1981, Jinno and Kawasaki, 1983a). The poor

correlation in this instance most likely arises from the

misrepresentation of the calculated van der Waals volume, rather than a

breakdown of the retention theory. The poor correlations for the

ethanol and propanol data on the Ultrasphere ODS column may be caused by

the changing stationary phase environment as these more hydrophobic

modifiers are used in the mobile phase. There is not yet any theory

published on the uptake of solvent by the stationary phase chains.

Michels (1989) has recently reported that propanol appears to saturate

the stationary phase at a very low mobile phase percentage of propanol.

The poor correlations with the Silasorb column have yet to be accounted

for.

The y-intercepts at OB=1 of equation 2-3 are predicted to equal

the free energy of transfer of the solute between pure A and pure B of











the mobile phase, multiplied by the cavity size. That is, the

y-intercepts of 1/OB In (k'/k,) versus IB at (OB=1) = n(XSB XSA) and




ay-intercepts(OB=l) / an = (XSB XSA) (2-7)




The dependence of these intercepts on the cavity volume and area

can be seen in Figures 2-5 and 2-6 respectively. The correlation

coefficients of these intercepts and the cavity size are generally

greater than 0.9, and the regression results are listed in Tables 2-4

and 2-5. The intercepts are found to increase linearly with cavity size

as expected from the theory. Figure 2-6 shows a typical case, the

Sepralyte C-18 column, with acetonitrile as component B and water as

component A of the mobile phase. The exceptions noted are the same as

those discussed above. The free energy of transfer of a solute per unit

area can be calculated by utilizing Figure 2-6 and Table 2-5




(XSB XSA)kT = -(3.689 x 10-2)(592 cal/mol) = -21.8 cal/mol A2 (2-8)



and is approximately constant for the same modifier used. The above

tests confirm the predicted dependence of solute retention on cavity

size, and the chromatographic method provides a convenient way to

experimentally determine solute-solvent interaction free energies.

According to equation 2-3, the slope of the composition plot

divided by n should give the binary interaction constant of the mobile

phase components, XAB, regardless of the nature of the solutes and

stationary phases used. This binary interaction constant should depend



























y-intercepts

at OB=1 from 1

equation 2-3
-1

-1


40 50 60


70 80 90 100 110 120 130

Vw (cm3/mol)


Figure 2-5. Plot of y-intercepts at OB=1 from equation 2-3 versus
van der Waals volume, Vw, of the solutes for the
Sepralyte C-18 column with acetonitrile as modifier.

















Table 2-4.


Column

Sepralyte C-2

Sepralyte C-4

Sepralyte C-8

Sepralyte C-18

Sepralyte C-4

Sepralyte C-8

Hypersil ODS

Unisil Q C-18

Ultrasphere ODS

Ultrasphere ODS

Ultrasphere ODS

Ultrasphere ODS

Silasorb C-8

Silasorb C-8


Regression results of y-intercepts at Bg=l from equation
2-3 versus the van der Waals volume, Vw, of the solutes
for all the columns.


Modifier

Acetonitrile

Acetonitrile

Acetonitrile

Acetonitrile

Methanol

Methanol

Acetonitrile

Acetonitrile

Methanol

Ethanol

1-Propanol

Acetonitrile

Methanol

Acetonitrile


Slope(xl02)

-8.523

-7.484

-9.793

-6.936

-12.73

-12.17

-377.2

-109.0

-22.23

10.96

95.35

-345.8

721.2

629.9


.2


0.9542

0.9344

0.9112

0.9178

0.9489

0.9136

0.7089

0.5035

0.9993

0.2696

0.1945

0.5705

0.2434

0.2139



























y-intercepts
at ~B=1 from
equation 2-3





















Figure 2-6.


80 100 120 140


160 180 200 220 240 260


HSA (A2)


Plot of y-intercepts at B=' from equation 2-3 versus
hydrocarbonaceous surface area, HSA, of the solutes
for the Sepralyte C-18 column with acetonitrile as
modifier.

















Table 2-5.


Column

Sepralyte

Sepralyte

Sepralyte

Sepralyte

Sepralyte

Sepralyte


Regression results of y-intercepts at 03=1 from equation
2-3 vs the hydrocarbonaceous surface area, HSA, of the
solutes for all the columns.


C-2

C-4

C-8

C-18

C-4

C-8


Modifier

Acetonitrile

Acetonitrile

Acetonitrile

Acetonitrile

Methanol

Methanol


Slope(xl02)

-4.383

-3.957

-5.118

-3.689

-6.752

-6.378


r'

0.9658

0.9389

0.9052

0.9282

0.9460

0.9489











solely on the nature of the organic modifier in the mobile phase of

RPLC. By comparing the values of the slopes of Table 2-3, they are

found to be independent of the solutes, but not the stationary phase.

There is a factor of 2 differentiating the slopes from a short-chain

stationary phase to a long-chain stationary phase, such as Sepralyte C-2

to the Sepralyte C-18 column. This variation can be attributed to two

factors. First, in short-chain stationary phases, surface silanols are

more accessible to solutes, and the interactions between them are much

larger than in long-chain phases. Second, an alternate retention

mechanism, adsorption (Dill 1987), should dominate in short-chain

stationary phases. Both of these factors are neglected in the original

partition model (Dill 1987).

Furthermore, the y-intercept at =B=1 divided by n should only be

dependent on the nature of the solute and organic modifier. The data of

the Sepralyte columns from Table 2-4 confirm this prediction since the

same solutes are used for the Sepralyte columns. The large differences

of the values in column 3 of Table 2-4 are mainly due to the differences

in solutes employed with different stationary phases.

Johnson and coworkers (Johnson et al., 1986; Dorsey and Johnson,

1987) have shown in an earlier study that a solvatochromic dye molecule,

referred to as ET(30), can be used to probe the chemical nature of the

mobile phase and its strength as used in RPLC. The visible spectral

shift of the dye is found to be linearly proportional to the binary

interaction constant XAB of the mobile phase components (Figure 2-7).

This provides a justification for the employment of ET(30) as a probe

for the study of the mobile phase, since it appears to directly measure



























Sloces from
E-(30) plots 500--
x 10
400-


300-0I I -I I
9 10 11 12 13 14 15 16

Slopes from equation 2-3















Figure 2-7. Plot slopes from ET(30) plots versus slopes from
equation 2-3 for the Sepralyte C-18 column with
acetonitrile as modifier.

















Table 2-6.


Column

Sepralyte C-2

Sepralyte C-4

Sepralyte C-8

Sepralyte C-18

Sepralyte C-2

Sepralyte C-4

Sepralyte C-8

Silasorb C-8

Silasorb C-8

Unisil Q C-18

Hypersil ODS

Ultrasphere ODS

Ultrasphere ODS

Ultrasphere ODS

Ultrasphere ODS


Regression results of slopes from equation 2-3 versus the
slopes of ET(30) plots for all the columns.


Modifier

Acetonitrile

Acetonitrile

Acetonitrile

Acetonitrile

Methanol

Methanol

Methanol

Methanol

Acetonitrile

Acetonitrile

Acetonitrile

Methanol

Ethanol

1-Propanol

Acetonitrile


Slepe(xl021

3.238

2.699

5.396

4.047

14.67

9.315

11.03

7.342

5.137

4.913

3.671

8.055

6.166

2.467

4.066


X2

0.9996

0.9909

0.8796

0.9997

0.9956

0.9820

0.9955

0.8369

0.9987

0.9303

0.9187

0.9997

0.8827

0.2630

0.8702











the free energy of contact between components A and B of the mobile

phase.

For the Sepralyte C-18 column with acetonitrile as modifier, the

slope of these two parameters are found to be



(4.047 x 10-2) x (592 cal/mol) = 23.96 units(cal/mol) (2-9)




Other columns are observed to have the same linearity with other organic

modifiers, and the linear regression results between these two

parameters are listed in Table 2-6.

Even though the experimental data base used to test the

theoretical predictions is relatively large, the range of mobile phase

compositions is still limited to within 30 50% variation in solvent

composition. The linear expression proposed by Dill (1987) and Dorsey

and Dill (1989) tested in this work is found to hold well with the data

base used. If the partition model and other assumptions are incorrect,

nonlinearity should be observed. Since good linearity is found, the

slopes and y-intercepts generated by plotting the composition plot of

equation 2-3 should reflect, at least approximately, the binary

interaction constants of the solutes and the solvents, and can be used

to calculate the solute-solvent interaction free energies in RPLC.





















CHAPTER III
CHARACTERIZATION OF THE RETENTIVITY OF
REVERSED PHASE LIQUID CHROMATOGRAPHY COLUMNS

Introduction

There is an immense number of RPLC columns available on the market

today. The majority of them are made from Cg or C18 functional groups,

but the use of columns having other functionalities, such as phenyl and

cyano, has been on the rise. Recent studies done on commercial RPLC

columns have found that they all showed significant differences in

absolute retention and selectivity, a, for the same solute and mobile

phase, even when they are all labelled as C18 columns (Goldberg 1982;

Sander and Wise, 1988). This variability of RPLC columns is largely due

to the difference in the starting silica as well as the bonding reaction

(Dorsey and Dill, 1989). Since the stationary phase in RPLC has been

shown to have ample effects on solute retention and selectivity (Sander

1988; Dill 1987; Sentell and Dorsey, 1989a; Sentell and Dorsey, 1989b),

the variability has caused practical chromatographers many difficulties

in choosing the best column to develop optimal separations. Developed

methods are hard to transfer unless the brand and manufacturer of the

column are specified.

Many studies have been performed by investigators to characterize

the retention behavior of RPLC columns. Smith (1982a and 1982b)

employed a homologous series of alkylarylketones to develop a retention











index with a set of reference compounds to define the retention

performance of RPLC columns made from different manufacturers and

functionalities. A different set of constants for each column is

obtained for every different mobile phase composition used. These

column retention constants are not very useful since new calibration on

the column has to be done when mobile phase composition is changed.

Antle and coworkers (Antle and Snyder, 1985; Antle et al., 1985) used

gradient elution theory to characterize RPLC columns according to their

solvophobic retention. They examined columns produced from the same

base silica but having different bonded functionalities such as C18, C8,


phenyl, C1, or cyano groups. A large variety of test solutes having

very diverse chemical structure were used in their study. They combined

the volume phase ratio, 0, and the polarity of a column into an

effective phase ratio, J', which gives the retentivity of a column. A

reference and a standard column were used to acquire the relative J'

value of all columns studied. Their results revealed that the J' value

of the columns are in the order of C18 > Cg > phenyl > C1 > cyano. They

concluded that the contribution of the polarity of a column to the

retentivity is small compared to the phase ratio of the column. Cooper

and Lin (1986) have looked at the retention behavior of C8, phenyl and

cyano columns using three solutes chosen from Snyder's selectivity

triangle (Snyder 1974; 1978). They found that the polarity of these

columns is in the order of phenyl > cyano > C8, and the overall

retentivity of these columns is dominated by the phase ratio of the

columns. Other researchers have used chemometric and factor analyses to

characterize commercial columns (Delaney et al., 1987; Chretien et al.,











1986; Walczak et al., 1987), but only qualitative results were obtained

in these studies. Walczak et al. (1987) concluded that carbon loading,

nature of the organic ligands, and the accessibility of the surface

silanol groups are the main factors governing the retentivity and

selectivity of the columns. Delaney et al. (1987) summarized that their

classification of RPLC columns from chemometric results agreed well with

a qualitative scheme developed by a liquid chromatography specialist.

The present study provides a simple method to characterize the

retentivity of commercial RPLC columns. Binary mobile phases of water

and an organic modifier under isocratic conditions were used throughout

the study to avoid lengthy equation derivation, and to keep experimental

parameters simple. No reference or standard column was needed, and

solutes of many different types were employed so that the results of the

study should be applicable to all solutes.



Experimental Section

All the retention measurements were obtained either with a

Spectra-Physics SP8700 ternary proportioning LC system (Spectra-Physics,

San Jose CA, USA) or with two Spectroflow 400 pumps (ABI Analytical,

Kratos Division, Ramsey, NJ, USA). A Valco liquid chromatography

injection valve (Valco Instrument Company, Houston, TX, USA) and a 20 pl

sample loop were used to inject the solutes. The detection system was

either a Spectroflow 783 absorbance detector/gradient controller (ABI

Analytical, Kratos Division, Ramsey, NJ, USA) operated at 254 nm or a

Spectroflow 757 absorbance detector (Kratos Analytical Instruments,

Ramsey, NJ, USA) also operated at 254 nm. When the two Spectroflow 400

pumps were used, a Rainin dynamax dual chamber mixer (Rainin Instrument











Company, Inc., Woburn, MA, USA) was placed before the injection valve

for better mixing of the binary solvents. All the Zorbax columns were

commercially available from Du Pont (E. I. du Pont de Nemours and

Company, Wilmington, DE, USA), the B & J columns from Burdick & Jackson

Laboratories, Inc. (Baxter Healthcare Corp., Burdick & Jackson Division,

Muskegon, MI, USA), and the Ultrasphere column from Beckman (Beckman

Instruments, Inc., Fullerton, CA, USA). The high density column was

synthesized and packed in our laboratories (Sentell 1987). Some of the

properties of these columns as supplied by the manufacturer are listed

in Table 3-1. A Fisher Recordall Series 5000 (Fisher Scientific,

Pittsburgh, PA, USA) strip chart recorder was used to record all the

chromatographic peaks. The solvents used were HPLC grade from Fisher

Scientific (Fisher Scientific, Pittsburgh, PA, USA), and the water was

filtered through a Barnstead NANOpure II system (Barnstead Company,

Dubuque, IA, USA) before being used. The columns were maintained at

300C by a water jacket and a Haake D1 circulator (Haake, Dieselstrasse,

West Germany).

A total of 26 solutes was used and they are listed in Table 3-2.

They were from various suppliers and were used without further

purification. The solutes were grouped into mixtures containing no more

than 5 solutes based on their retention data when they were injected

individually at the beginning of the study. Peak assignments were

accomplished by identifying the peak area and elution order. The void

volume of each column was determined by the injection of water at mobile

phase composition of 75% water and 25% organic modifier. The flow rate

was maintained at 1 ml/min except for the Burdick & Jackson C-18 column

which was used with a flow rate of 1.5 ml/min. All the retention data

















Table 3-1.






Column


Zorbax ODS


Zorbax C-8


Zorbax phenyl


Zorbax TMS


Zorbax cyano


B & J C-8


B & J C-18


Properties of the RPLC columns as supplied by the
manufacturers.


Ligand


C18


Surface area(m2/Lg % Carbon loading


340 20


C8


Phenylpropyl


C1

Cyanopropyl


C8


C18


Ultrasphere C-8 Cg
















List of test solutes used in this study.


Methyl paraben

Propyl paraben

Cortisone

Toluene

Benzyl Alcohol

Propachlor

Tri-p-tolyl phosphate

Fluorobenzene

Dimethyl phthalate

Butyl benzyl phthalate

1-Methyl phenanthrene

p-Terphenyl

1,1,4,4,-Teraphenyl-1,3-butadiene


Ethyl paraben

Butyl paraben

Corticosterone

Hexylfluorobenzene

o-Nitrophenol

1-Methyl naphthalene

Methyl benzyl amine

Chloropropham

Diethyl phthalate

Dibutyl phthalate

Dioctyl phthalate

Chrysene

Octadecanophenone


Table 3-2.











were obtained from averaging at least two measurements of a solute. The

ET(30) polarity values for the different mobile phase compositions were

calculated from the quadratic relationship between percent organic

modifier and the ET(30) values reported by Dorsey and Johnson (1987).

The regression calculations were all done by using the StatWorks

(Heydon and Son Inc., Philadelphia, PA, USA) program on a Macintosh SE

(Apple Computer, Inc., Cupertino, CA, USA) microcomputer.



Results and Discussion

Since the composition of the mobile phase plays a primary role in

the retention of solutes in RPLC, it is important to have a method to

account for the solute retention contributed by the mobile phase before

the retentivity of the stationary phase can be explored. Many workers

have used solvatochromic measurements to measure the polarity of the

mobile phase used in RPLC, such as the "Z" scale (Kosower 1958), the x*

multiparameter scale (Taft and Kamlet, 1976; Kamlet et al., 1983; Taft

et al., 1985; Sadek et al., 1985), and the ET(30) scale (Johnson et al.,

1986; Dorsey and Johnson, 1987). Johnson and coworkers (Johnson et al.,

1986; Dorsey and Johnson, 1987) have shown that the In k' of a solute in

RPLC is linearly associated with the ET(30) polarity scale and has

better linearity than the volume percent of organic modifier. Their

relationship can be expressed as


In k' = m ( ET(30) ) + c


(3-1)











where ET(30) is the polarity value in kcal/mol of the mobile phase


measured by the ET(30) scale, and m and c are the slope and y-intercept


of the linear regression. The ET(30) values are found to have an

excellent quadratic relationship with most common binary mobile phases

used in RPLC (Dorsey and Johnson 1987). Moreover, the slope m has been

shown in Chapter II to be directly proportional to the size of the

solutes and the binary interaction constant of the mobile phase. We

selected to employ the slope m from equation 3-1 as the descriptor for

the mobile phase contribution to solute retention.

Molecular Descriptor Approach

Although the mobile phase effect can be accounted for by using the

slope in equation 3-1, the cavity created in the stationary phase also

is a major contribution to solute retention according to the partition

model. The first approach we took in this study was to search for a

molecular descriptor of the test solutes that can delineate the cavity

in the stationary phase, and use it to plot against slope m. The

resulting slope of this plot should only be a function of the

retentivity of the column. We first predicted that a linear

relationship should occur between the slopes from equation 3-1 and a

molecular size descriptor. It has been shown in the literature that a

linear dependence exists between In k' of a solute and some molecular

size descriptors of the solute such as the van der Waals volume, Vw,

(Hanai and Hubert, 1984a; Smith 1981; Jinno and Kawasaki, 1983a) and the

molecular connectivity index, X1 (Wells et al., 1982; 1986). Slopes

from equation 3-1 were plotted against both the van der Waals volume and

the molecular connectivity index (Figures 3-1 and 3-2) and the




























Slopes from
equation 3-1


0 100 200 300

Vw (cm3/mol)


Figure 3-1.


Plot of slopes from equation 3-1 against the van der
Waals volume, Vw, of the test solutes for the Zorbax
TMS column using acetonitrile as modifier.
















Table 3-3.


Regression results of graphs of slopes from equation 3-1
against van der Waals volume, Vw, of the test solutes.


Slopex(1031


Zorbax

Zorbax

Zorbax

Zorbax

Zorbax


ODS

C-8

phenyl

TMS

cyano


ACN

ACN

ACN

ACN

ACN


4.08

3.82

4.27

9.25

6.98


y-inter.


0.845

1.03

0.593

0.336

0.310


0.533

0.534

0.484

0.861

0.472



























Slopes from
equation 3-1


0 2 4 6 8 10 12

connectivity index


Figure 3-2. Plot of slopes from equation 3-1 against the molecular
connectivity index of the test solutes for the Zorbax
TMS column using acetonitrile as modifier.
















Table 3-4.


Column



Zorbax

Zorbax

Zorbax

Zorbax


Regression results of graphs of slopes from equation 3-1
against molecular connectivity index of the test solutes.


ODS

C-8

phenyl

TMS


Zorbax cyano


ACN

ACN

ACN

ACN

ACN


6.21

6.54

7.14

16.7

11.5


0.992

1.14

0.723

0.563

0.531


0.401

0.417

0.379

0.829

0.354











regression results of these graphs on five Zorbax columns are shown in

Table 3-3 and 3-4 respectively.

The coefficients of determination, r2, of these regressions are

all well below 0.9 meaning that there is no significant correlation

between the slopes from equation 3-1 and the two molecular size

descriptors of the solutes that we have chosen. One of the reasons for

the breakdown of our initial hypothesis is probably because the

molecular size descriptors do not account for all the interactions among

the solutes and the stationary phase. Despite Chapter II showing that

the cavity opened in the stationary phase is proportional to the size of

the solute, there are other chemical interactions such as polarity,

dipole moment, and hydrogen bonding ability between the solute and the

stationary phase. These interactions cannot be completely summed by one

single molecular size descriptor.

In kw Approach

The next approach we took was to seek a parameter that will

accurately include all the interactions between the solute and the

stationary phase. The logarithmic capacity factor of a solute at 100%

water, In kw, has been shown to have great correlation to the logarithm


of the water/octanol partition coefficient, log Po/w, of a solute

(Miyake et al., 1988; Braumann 1986; Braumann et al., 1987; Minick et

al., 1987). The linear relationship shows that both In kw and log Po/w


are measuring a similar partition process. This suggests that In kw

measures the driving force of the partition of a solute into the

stationary phase, and is the parameter that measures all the

interactions between the solute and the stationary phase.











The In kw of a solute is difficult to obtain experimentally in

RPLC due to long retention times and poor peak shapes arising from slow

mass transfer with pure water as the mobile phase. There is a popular

belief that the logarithm of the capacity factor, In k', of a solute is

linearly related to the percent by volume of organic modifier in the

mobile phase (Snyder et al., 1979; Dolan et al., 1979; Schoemakers et

al., 1979; Antle et al., 1985; Baty and Sharp, 1988). Hence, In kw is

often estimated by extrapolation from a linear regression of In k'

versus percent organic modifier back to 0% organic modifier (Reymond et

al., 1987; Braumann 1986; Braumann et al., 1987; Baty and Sharp, 1988).




In kw = S (%organic) + c' (3-2)




Unfortunately, this kind of extrapolation has been found to be not as

reliable as using the ET(30) polarity scale (Michels 1989). The ET(30)


polarity value of pure water is 63.11 kcal/mol, and therefore In kw of a

solute can be approximated by substituting 63.11 kcal/mol into equation

3-1.

Since the slopes from equation 3-1 and the In kw of the solutes

summarize the retention effects due to the mobile phase and the solute

interactions with the stationary phase respectively, the slope of a plot

of these two parameters for a column should be independent of the two

effects, and unveil the retentivity of the column (Figure 3-3). The

regression results of these plots for the columns studied are presented

in Table 3-5. The 95% confidence interval of the slopes of these




















3-




2

Slopes from
equation 3-1






0-
0 10 20 30
In kw















Figure 3-3. Plot of slopes from equation 3-1 against the In kw of
the test solutes for the Zorbax TMS column using
acetonitrile as modifier.
















Table 3-5.


Regression results of slopes from equation 3-1 vs In kw
of the test solutes for all the columns.


Column


Zorbax ODS

Zorbax C-8

Zorbax Phenyl

Zorbax TMS

Zorbax Cyano

Zorbax ODS

Zorbax C-8

Zorbax Phenyl

Zorbax TMS

Zorbax Cyano

B & J C-8

B & J C-18

Ultrasphere C-8

High Density


ACN

ACN

ACN

ACN

ACN

MeOH

MeOH

MeOH

MeOH

MeOH

MeOH

MeOH

MeOH

MeOH


0.0835

0.069

0.104

0.110

0.126

0.120

0.124

0.129

0.134

0.130

0.134

0.114

0.122

0.115


0.525

0.727

0.293

0.349

0.241

0.320

0.310

0.271

0.346

0.362

0.209

0.326

0.307

0.314


0.894

0.834

0.98

0.982

0.99

0.977

0.987

0.993

0.979

0.96

0.987

0.945

0.99

0.923


a)95% confidence interval of the slopes of the regressions


CI-@t0 .05.


0.004

0.035

0.005

0.007

0.005

0.003

0.004

0.003

0.006

0.006

0.009

0.012

0.004

0.013


y-inter. L2


Modifier Slope











regressions are calculated (Sharaf et al., 1986; Anderson 1987), and

they are listed in the last column of Table 3-5. The confidence

interval shows that almost all the slopes of the regressions are

statistically different from one another. Good linearity appears

between these two parameters with the r2 of all the columns tested above

0.9, except in the case of the Zorbax ODS and C-8 columns with

acetonitrile as modifier. The high r2 confirms that the slopes from

equation 3-1 and In kw of the test solutes are highly correlated. This


type of slope versus In kw plot is not new. Several groups (Braumann et

al., 1983; Hammer et al., 1982) have shown that plots of the slope, S,

and In kw from equation 3-2 are linear. They found the slopes and

y-intercepts of these plots formed two empirical parameters that can be

employed to classify solutes into different groups. These solute groups

can be used to recommend the use of In kw for the estimation of log


Po/w- Schoenmakers et al. (1979) found good linearity when methanol is

used as mobile phase modifier but not with acetonitrile or THF. They

have shown that when linearity is observed, the slopes of these plots

can be used to determine the shape of the gradient program. Moreover,

these slopes can be employed to predict isocratic capacity factors from

a simple gradient analysis (Schoenmakers et al., 1981). Baty and Sharp

(1988) observed good correlation between S and In kw for methanol,

acetonitrile and THF using structurally similar solutes. They tried to

use the slopes and y-intercepts of these plots to predict capacity

factors of the solutes, but they found that the slopes and y-intercepts

of these plots vary with the nature of the organic modifiers and











columns. Therefore, the chromatographic system and the solute group

have to be defined before the capacity factors can be predicted.

The general retentivity of the columns in this study for a given

organic modifier is found to be inversely proportional to the slope of

the slopes versus In kw plot. The usefulness of this relative

retentivity scale can be seen by plotting the linear regressions of the

five Zorbax columns using acetonitrile as modifier on one graph (Figure

3-4). As shown in Chapter II, the slopes from equation 3-1 are found to

be proportional to the size of the cavity created by the solutes;

therefore, at a given solute cavity size, the linear regression having

the smallest slope will have the largest In kw and hence, the largest

retentivity. From the data shown in Figure 3-4, the Zorbax cyano column

has the largest slope followed by the TMS column, phenyl column, ODS

column and C-8 column; retentivity of these column is in the order of

C-8 > ODS > phenyl > TMS > cyano. The Zorbax C-8 column seems to have a

higher retentivity than the Zorbax ODS column when acetonitrile is used

as modifier. A careful examination of the confidence interval of these

two slopes demonstrates that they actually have no statistical

difference, and the apparent higher retentivity of the Zorbax C-8 column

could very well be an experimental artifact. The apparent stronger

retentivity of the Zorbax cyano column over the TMS column with methanol

as modifier can be attributed to the hydrogen bonding between the

surface silanols and the methanol. Cyano columns have been shown to

have more accessible silanols than other columns (Cooper and Lin, 1986;

Smith and Miller, 1989). Since methanol can form hydrogen bonds with

these silanols, the surface of cyano columns is significantly modified
























Slopes from
equation 3-1


0 10 20 30


In kw of solutes


Figure 3-4.


Plot of slopes from equation 3-1 against the In kw of
the test solutes for all the Zorbax columns using
acetonitrile as modifier.











by the methanol, and hence its apparent retentivity is stronger than the

TMS column. When acetonitrile is used as modifier, due to its lack of

hydrogen bonding ability, the retentivity of the TMS column is found to

be stronger than the cyano column. Other commercial columns can be

tested using the same procedures and compared with the retentivity of

the Zorbax columns. With methanol as modifier, the B & J C-18 column is

found to have retentivity larger than any of the Zorbax columns, and the

B & J C-8 column is found to have retentivity approximately the same as

the Zorbax TMS column, while the Ultrasphere C-8 column is shown to have

retentivity between the Zorbax ODS and C-8 columns.

One interesting point raised by many researchers is the importance

of phase ratio of the column on the total retentivity of the column.

The significance of phase ratio on retentivity was also investigated in

this study. Phase ratio in RPLC can be calculated using the volume

ratio of the stationary phase to the mobile phase. Although much work

has been done on measuring the volume of the mobile phase, Vm, (Melander

et al., 1983; Smith et al., 1986; Engelhardt et al., 1984; Hennion and

Rosset, 1988) measurement of the volume of the stationary phase, Vs, is

more ambiguous (Jandera et al., 1982; Melander et al., 1980; Sander and

Field, 1980; Slaats et al., 1981; McCormick and Karger, 1980a;

Berendsen et al., 1980). By employing the Vs and Vm calculation method

presented by Sentell (1987), the phase ratios of the columns studied

were obtained and are listed in Table 3-6. In the case of the phenyl

and cyano columns, the density of the ligands are not available and the

Vs of the columns cannot be calculated. The phase ratio of these

columns are therefore not reported in Table 3-6. With the exception of

















Table 3-6.


Phase ratio of the columns in this study calculated using
the method presented by Sentell (1988).


Column

Zorbax ODS

Zorbax C-8

Zorbax phenyl

Zorbax TMS

Zorbax cyano

B & J C-8

B & J C-18

Ultrasphere C-8

High density


Phase ratio

0.367

0.226

unknown

0.172

unknown

0.0619

0.110

0.380

0.465











the two B & J columns, the phase ratio of the columns displays the same

order as the relative retentivity scale of the columns. The high

density column, which has the largest phase ratio, is found to be close

to having the highest retentivity. This result points out that the

phase ratio of a column indeed has a vital contribution to the overall

retentivity of the column.

This work exhibits a simple method that can be used to classify

commercial RPLC columns according to their overall retentivity. The

isocratic approach used has kept the data interpretation and

manipulation simple. The use of In kw in the procedure is justified and

is found to be very appropriate. The phase ratio of a column is shown

to dominate the total retentivity of the column. Since such a large

variety of test solutes were employed in this work, the relative

retentivity scale based on this study should be applicable to almost all

solutes.



















CHAPTER IV
CONCLUSIONS AND FUTURE WORK

Conclusions

The research that has been discussed in the previous chapters was

performed to confirm and apply the partition retention mechanism of

RPLC. The recognition of the partition mechanism as the correct

retention model in RPLC is very important because it can help

chromatographers to understand the processes happening in RPLC. Also,

with a sound fundamental understanding of solute retention, more robust

separation methods can be developed.

In Chapter II, we have tested the partition model with a large

data base. We found that the relationships between solute capacity

factors and mobile phase compositions are in good agreement with the

prediction from the partition model (Dill 1987; Dorsey and Dill, 1989)

with only very minor discrepancies. The partition model proposed by

Dill (1987) is based on simple lattice models and the interphase model

of the stationary phase, but the results of our tests are still very

good. With a more explicit treatment of the simple assumptions, the

theory may be able to account for the small discrepencies. The

molecular origins of these composition plots are also clarified in our

tests. The slopes of these plots are found to be a function of the

binary interaction parameter of the mobile phase components, and the

y-intercepts at OB=I are proportional to the free energy of transfer of

the solutes. The solute size plots provide a method to calculate the











solute-solvent interaction free energies, and an explanation to the

widely observed relationship between capacity factors and solute sizes

in RPLC (Hanai and Hubert, 1984b; Jinno and Kawasaki, 1983a; 1983b; Feng

et al., 1988; Mockel et al., 1987). Moreover, the slopes and

y-intercepts at OB=1 of the composition plots predicted by the theory

give a simple and reliable method to estimate solute-solvent interaction

free energies. These free energies are difficult to obtain

experimentally and are seldom reported in the literature. With this new

approach, these interaction free energies should be much easier to

access due to the availability of chromatographic data.

The partition model is put to practice in Chapter III. We

employed the partiton model as our basis to develop a relative

retentivity scale for RPLC columns. In the partiton model, the entire

stationary phase contribution to retention is summed in the In kw term

of the solutes. Although this assumption is an oversimplification of

the stationary phase effect, it has been shown that In kw is a very

effective parameter to estimate the solute partition coefficient (Miyake

et al., 1988; Braumann 1986; Braumann et al., 1987; Minick et al.,

1989). We used the In kw of the solutes and the slopes from the In k'


versus ET(30) values plots to obtain the relative retentivity values of

the RPLC columns. The correlation between these two parameters is shown

to be good, and this further supports the partition model as the

dominant solute retention mechanism. The retentivity scale is found to

be in good agreement with the literature (Antle and Snyder, 1985; Antle

et al., 1985). Also, the experimental procedures are kept relatively

simple, and therefore this retentivity scale should be useful in











classifying the retentivity of commercial RPLC columns, and should help

practical chromatographers to select the best column for their

applications. Since test solutes of various chemical structures are

employed in our study, the retentivity scale should be applicable to all

solutes. The phase ratio of the columns is found to play a major role

in determining the retentivity of the column.



Future Work

Although the data base that is used in Chapter II to test the

predictions from the partition model is rather large, it lacks a wide

range in mobile phase compositions especially at the extreme

compositions. Most of the data sets have mobile phase composition range

between 30 50% organic modifier. A test of the predictions at extreme

mobile phase compositions such as mobile phases having a few percent of

organic modifier would be interesting, since it will unveil how well the

theory holds under extreme mobile phase composition. The stationary

phase has been reported to take on a different structure as the

composition of the mobile phase changes (Yonker et al., 1982a; 1982b;

McCormick and Karger, 1980a; 1980b; Carr and Harris, 1986; 1987; McNally

and Rogers, 1985). The changing in the structure of the stationary

phase implies that the retention mechanism may switch from partition to

other forces. By testing the theory at extreme compositions, the mobile

phase composition that can significantly change the stationary phase

structure may be observed as the relationship becomes nonlinear.

Another test that is useful in testing the validity of the

partition model is to use retention data from stationary phases having

functionalities other than alkyl chains, such as phenyl and cyano











groups. The data from these stationary phases should give information

on the retention mechanism in these different phases. If linearity

exists in these columns, partition should be the dominate force. If a

nonlinear relationship occurs, a retention mechanism other than

partition should be considered for these columns.

The experimental procedures for obtaining the relative

retentiviity scale in Chapter III is simple; nevertheless it is rather

time consuming since isocratic data of more than 20 solutes have to be

collected. A possible time saving step is to find a few representative

solutes so as to eliminate the useage of all 26 test solues. The Snyder

selectivity triangle (Snyder 1974; 1978) provides a good starting place

to look for representative solutes.

Moreover, the present study only involved monomeric coverage

columns. In order to expand the practical utility of the retentivity

scale, stationary phases other than monomeric coverage should also be

tested to find out if they fit in the retentivity scales. Since

polymeric phases have been shown to have better selectivity for PAH's

than monomeric phases (Sander and Wise, 1984), and more and more

polymeric phases are available, there is a need to classify and compare

them with the monomeric phases. By using polymeric columns, the

applicability of our retentivity scale can be tested. Also, since the

retentivity scale is based on the partition model, a failure in

classifying polymeric columns with the retentivity scale may suggest

that a retention mechanism other than partition is the dominate factor

in polymeric phases. The surface coverage of polymeric phases is

usually higher than monomeric phases. It has been pointed out that

these reported values may not give the true surface coverage for




Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E2C19I9DD_14NN7Q INGEST_TIME 2012-02-07T17:33:06Z PACKAGE AA00003779_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES