|Table of Contents|
Table of Contents
Chapter 1. Introduction
Chapter 2. Syntheses of silica-based RP stationary phases
Chapter 3. Syntheses of controlled pore glass-based RP stationary phases
Chapter 4. Correlations between chromatographic retention and octadecyl bonding density
Chapter 5. Correlations between chromatographic selectivity and octadecyl bonding density
Chapter 6. Conclusions
INTERPHASE SOLUBILITY AND CHROMATOGRAPHIC RETENTION
KAREN BELINDA SENTELL
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
This dissertation is dedicated to my fiance, Daniel Coffman.
I could never have done
this without you.
There are many people whom I wish to acknowledge for
their assistance with this work. Thanks are extended to Mel
Courtney for performing the numerous elemental analyses of my packings and to the technical staff in the departmental machine shop and glassblowing shop for their courteous assistance and advice. I am grateful to Dr. John Gerdes for suggesting the use of 4-dimethylaminopyridine (4-DMAP) in the bonded phase syntheses, to Nepera, Inc. for providing
the 4-DMAP to Dr. John Novak of the Aluminum Corporation of America for providing scanning electron micrographs of the silica and controlled pore glass supports and to Dr. Lane Sander of the National Bureau of Standards (NBS) for providing the NBS column evaluation test mixture.
I would like to express my gratitude to the Society for Analytical Chemists of Pittsburgh for funding my summer American Chemical Society (ACS) Analytical Division Graduate Fellowship and to Procter and Gamble for funding my fullyear ACS Analytical Division Graduate Fellowship.
Thanks are also due to my fellow Dorsey group members (both present and former) for their friendship, advice and support. The camaraderie within our group will be one of my warmest memories of graduate school. I look forward to i ii
continued association with them during iny postdoctoral tenure.
The love and moral support from my parents, Bobby and Ruth Sentell, have helped to sustain me throughout my education. They are responsible for instilling in me a love of reading, a respect for education and an unquenchable thirst for knowledge. I am grateful to them and to my sister, Michelle, for their encouragement during the
My deepest gratitude is extended to my graduate
research advisor, Dr. John G. Dorsey, for his advice and guidance. He is the epitome of what a research advisor should aspire to be and has served as an inspiration to me both as a research scientist and as a teacher. I have greatly enjoyed our conversations and I look forward to our continued professional interaction over the next year of my postdoctoral appointment. I also thank him for encouraging my oenophilic tendencies; after all, everyone needs to develop a new vice now and then.
Lastly, I want to thank my fiance, Daniel Coffman. His love, patience and support have sustained me even when I was discouraged and disheartened; without his help I could never have completed this work. In addition to his moral support, I would also like to thank him for his expert drafting and technical assistance as well as for accompanying me on my numerous midnight sorties to check on my reactions. He deserves my heartfelt gratitude now and forever for always being there when I need him.
TABLE OF CONTENTS
P ag e
ABSTRACT ................. . . ...................................... vii
I INTRODUCTION.. . . ..................................................... 1
Models of Reversed Phase Liquid Chromnatographic
Retention.. ............................................ ............1
Theories of Retention in RPLC ... .. .. .. .. .. .. .. .. .. ...17
II SYNTHESES OF SILICA-BASED RP STATIONARY PHASES ...... 29
Experimental Considerations in the Synthesis
of RP Stationary Phases...................................... 29
Experimental Procedure............................... 34
Syntheses Utilizing Ultrasonic Waves...............41
Effect of Subambient Temperature on the
Ultrasound Reaction ..............................44
The Use of 4-Dimethylaminopyridine *t~;* as. the'
Acid-Acceptor Catalyst.. ...................................... 45
III SYNTHESES OF CONTROLLED PORE GLASS-BASED RP STATIONARY PHASES.............................................. 52
Comparison of Controlled Pore Glass and Silica
as Supports for RP Stationary Phases ...............52
Experimental Procedure ............................56
Comparison of Silica and CPG'BondingDensities
via Reflux and Ultrasonic Syntheses .. .. .. .. .. .. .. ..60
IV CORRELATIONS BETWEEN CHROMATOGRAPHIC RETENTION
AND OCTADECYL BONDINGUDENSITY...................... 65
Chromatographic Determination of Thermodynamic
Partition Coefficients .............................65
Experimental Procedure ...............................76
V CORRELATIONS BETWEEN CHROMATOGRAPHIC SELECTIVITY
AND OCTADECYL BONDING DENSITY .................... 105
Introduction ....................................... 105
Experimental Procedure ............................. 110
Results and Conclusions ............................ 113
VI CONCLUSIONS ........................................ 135
Syntheses of RP Stationary Phases .................. 135
Validity of Chromatographic Partition
Coefficient Measurements ......................... 138
The Effect of Octadecyl Bonding Density on the
the Chromatographic Partition Coefficient ........ 149
Suggestions for Future Work ........................ 154
REFERENCES ............................................... 164
BIOGRAPHICAL SKETCH ...................................... 174
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
INTERPHASE SOLUBILITY AND CHROMATOGRAPHIC RETENTION
KAREN BELINDA SENTELL
Chai rm an: John G. Dorsey
Major Department: Chemistry
The retention and selectivity behavior of small solutes on silica and controlled pore glass (CPG) reversed phase liquid chromatographic (RPLC) stationary phases was studied as a function of stationary phase alkyl bonding density. These monomeric octadecyl phases were synthesized by both reflux and ultrasound methods; high alkyl bonding densities (3.60 ipmol/m2) were obtained via low temperature ultrasound reactions using 4-dimethylaminopyridine as the acid-acceptor catalyst. Using an improved method for the calculation of the stationary phase volume, the chromatographic capacity factors for the solutes were divided by the volume phase ratio (stationary/mobile) to obtain the thermodynamic partition coefficients; their behavior as a function of stationary phase octaciecyl bonding density was examined in vi i
55/45 methanol/water and 85/15 acetonitrile/water mobile phase systems.
For the silica packings in both mobile phase systems, the partition coefficients linearly increased until a critical bonding density of about 3.1 imol/m2 was reached; after this point the partition coefficients began decreasing with increasing bonding density. This behavior supports a partitioning retention mechanism for RPLC. In this model, the driving force for retention is the creation of a solute sized cavity in the stationary phase interphase structure. Beyond the critical density, increased alkyl bonding density results in enhanced interphase chain packing constraints which increase the energy necessary for solute cavity formation, resulting in decreased chromatographic partition coefficients.
Methylene and phenyl selectivity were also examined as a function of octadecyl bonding density. Methylene selectivity was approximately constant, but phenyl selectivity increased linearly with bonding density. This further supports the partitioning theory; methylene selectivity is not expected to be affected by chain ordering but phenyl selectivity for the linear solutes used should increase as the interphase packing structure becomes more ordered. Identical selectivity and retention studies on the CPG bonded phases garnered inconclusive results, as no obvious trends were discernable in either study.
This work supports a partitioning mechanism for RPLC retention and as such gives insight into the retention v ii i
process on a molecular level. This theory is predictive without adjustable parameters and is relevant to partitioning behavior in organized assemblies, including micelles, membranes and vesicles.
Models of Reversed Phase Liquid Chromatographic Retention Retention Indices Based on Solute Descriptors
Reversed phase liquid chromatography (RPLC) is one of
the most popular and powerful analytical separation methods. In RPLC, the stationary phase support consists of silica particles which are typically 3 to 10 pm in diameter; alkyl chains, usually 8 or 18 carbons in length, are attached to oxygen atoms on the silica surface via covalent bonds, resulting in a nonpolar surface. The mobile phase consists of water and an organic modifier such as methanol, acetonitrile or tetrahydrofuran. Thus the mobile phase is much more polar than the stationary phase; mobile phase polarity is adjusted by varying the volume ratios of water and organic modifier. It has been estimated that 80-90% of the high performance liquid chromatography (HPLC) systems currently in use are reversed phase systems (Melander and Horvath, 1980). Yet many practitioners of RPLC view this technique as "black magic" because its retention mechanism is not well understood, especially at the molecular level. This makes the prediction of retention for new compounds of interest extremely difficult. A basic understanding of RPLC retention at the molecular level is a necessity for the
formulation of a predictive retention index system. Such a retention index system would allow accurate interlaboratory comparison of RPLC retention data.
Chromatographic retention is most often quantified by the capacity factor, k'. Thermodynamically, the capacity factor for a chromatographic solute is the ratio of the number of moles of solute in the stationary and mobile phases. The capacity factor is also a measure of solute retention which normalizes for the mobile phase flow rate and the physical dimensions of the chromatographic column, since k' = (Vr Vm)/Vm, where Vr is the solute retention volume and Vm is the retention volume of an unretained solute, often called the dead volume. Many different investigators have attempted to correlate RPLC retention data with topological, geometric and/or calculated physical property descriptors of chromatographic solutes in order to predict RP retention. Topological descriptors include molecular connectivity, molecular complexity and correlation factor; van der Waals volume, molecular surface area and length/breadth parameters are geometric decriptors. Physical property descriptors include hydrophobic substituent constants, UNIFAC models of activity coefficients, and octanol/water partition coefficients (D'Amboise and Bertrand, 1986; Funasaki et al., 1986; Jinno and Kawasaki, 1984a, 1984b and 1984c; Petrovic et al., 1985).
Topological descriptors such as molecular connectivity indices are used to correlate chromatographic retention with molecular structure. These indices are numerical values which quantitatively describe carbonaceous adjacency relationships in the molecular structure of a solute (Lehtonen, 1984). Molecular connectivity indices have been shown to be proportional to the cavity surface area of a molecule. When a nonpolar hydrocarbon solute is introduced into an aqueous or hydroorganic environment, a large negative entropy of solution results. It has been suggested
that this negative entropy is a result of structural ordering around the hydrocarbon molecule (Karger et al., 1976). This ordering comes about from the formation of a cavity of water molecules around the hydrocarbon molecule. To overcome the entropy loss, nonpolar molecule segments will try to remove themselves from the aqueous medium and/or they will group together. The term "hydrophobic effects" is used to decribe these two phenomena of cavity formation and nonpolar clustering. The calculated surface area of the water cavity is significant because it can be related to the solubility of hydrocarbons in water (Karger et al., 1976). Likewise molecular connectivity, since it is also proportional to the cavity surface area, has also been correlated to non-electrolyte water solubility; however in contrast to cavity surface area, simple first order molecular connectivity indices are quite easy to calculate. Since in some cases the logarithm of the aqueous solubility
of a solute is proportional to the logarithm of its capacity factor k', it was expected that molecular connectivity would also be proportional to k', allowing prediction of retention to be made from these molecular connectivity calculations (Karger et al., 1976).
In their comparisons of experimental capacity factors to those predicted via simple molecular connectivity calculations, Karger et al. (1976) found very good agreement for para-substituted phenols and primary alcohols. However, the predicted k' values were uniformly high for secondary alcohols. This is because the experimental log k' values for two of the primary alcohols were used to determine the slope and intercept in the presumed linear relationship between log k' and molecular connectivity index for both types of alcohols. The high predicted k' values for the secondary alcohols reflect that their steric environment is different from that of the primary alcohols, showing that simple molecular connectivity indices can only be used to predict relative retention for compounds with the same functional group as those used for standards (Karger et al., 1976) .
Lehtonen (1984) used molecular connectivity indices of different orders, which correct for complex branching as well as for the nature of atoms other than carbon which make up the solute framework, to predict retention behavior for 16 dansylamides. Predicted and experimental k' values could
be correlated very well by combinations of different order
connectivity indices. However, a computer program had to be used in order to find the best combination of indices to obtain good correlation and these combinations were often nonlinear, involving combinations of connectivity indices raised to powers ranging from -2 to +2. The choice of indices also varied according to which organic modifier was used in the mobile phase as well as its percent composition. Although Lehtonen obtained good correlations between predicted and experimental retention, his method requires extensive computer calculations as no general connectivity index combination was applicable even within the same class of solutes (dansylamides). These two types of examples point out the shortcomings in the use of molecular connectivity indices to predict retention: the capacity factors for at least two members of the same class of compounds must be determined in order to find the proportionality constant between capacity factor and molecular connectivity, these predictions are only valid for compounds of the same functional group as the standards, for complex molecules a computer program must be used in order to find the best combination of indices to predict retention and this combination for a particular type of molecule may change if the mobile phase composition is altered. This method is also unable to distinguish between geometric isomers (Funasaki et al., 1986).
Molecular complexity is a topological descriptor and
the general index of molecular complexity (GIMC) is an index
whose applicability is genera] and which does not require the use of experimental or empirical data (D'Amboise and Bertrand, 1986). The GIMC is derived from combinations of graph theory and statistical information theory. It is so named because it considers all of the features which make a molecule more or less complex such as size, symmetry, branching, ring structures, multiple bonds and atomic heterogeneity. Molecules are represented by their skeletal molecular graph whose complexity is determined from a statistical information theory derived formula. Since any observable behavior related to a molecule's complexity is a function of the GIMO, chromatographic retention should also correlate with the GIMC. O'Amboise and Bertrand (1986) point out that GIMG is able to make retention predictions for solutes such as alcohols or fatty acids, which are not well correlated with hyrophobicity. They poi nt out that GIMC is a structure sensitive parameter representing the various reactive attributes of a molecule; therefore it should be related to the interaction mechanism in retention. However, plots of log k' versus GIMC for alcohols show distinct curvature, especially for alcohols with five or less carbons. Other difficulties exist as well. The index does not seem to be well applicable to molecules with different heteroatoms that are similarly bonded or for molecules belonging to nonhomologous series. Correlation of data between different stationary phases has also proven to be a problem (D'Amboise and Bertrand, 1986).
Geometric descriptors are in general fairly easy to
calculate. Van der Waals volume and surface area are both calculated from the van der Waals radii of the atoms from which the molecule is composed (Jinno and Kawasaki, 1984a). Length to breadth ratio (L/B) is a shape parameter based on the rectangle with minimum area which could envelop a molecule (Wise et al., 1981). However, all of these descriptors considered alone or in combination were found to have poor direct correlation with RPLC retention for substituted benzene derivatives. dinno and Kawasaki (1984a) concluded that this indicates that molecular size and shape were not the dominant forces controlling retention for these molecules. However, they noted that size and shape are important contributors to retention for alkylbenzenes and polycyclic aromatic hydrocarbons (PAHs) (Jinno and Okamoto, 1984). Wise et al. (1981) have also found that L/B is useful for predicting PAH elution order; however this parameter is useless for establishing general PAH retention indices since these indices vary according to the type of octadecyl bonded phase column used, necessitating the determination of a retention index equation for each different octadecy] column.
Physical property descriptors have been the most successful for predicting RPLC solute retention. The physical properties on which these descriptors are based come about from the solute's solution behavior in the mobile and stationary phases (Karger et al., 1976). Since
retention is controlled by the thermodynamic equilibrium of the solute between the mobile phase and stationary phase, retention could theoretically be predicted from standard Gibbs free energies since AG0 = -RT ln K, where AGO is the standard Gibbs free energy, R is the gas constant, T the absolute temperature and K the equilibrium distribution constant for the solute between the stationary and mobile phases. Since k' =K(Vs/Vm) where Vs and Vm are the stationary and mobile phase volumes, the capacity factor for a solute (and thus its normalized retention) could easily be calculated were AG0, V5 and Vm known. Since experimental AG0 values for these systems are unavailable, they are often estimated using liquid mixture models with readily available physical parameters, such as Hildebrand solubility parameters and group contribution concepts. But Hildebrand solubility parameters are only useful for qualitative descriptions of chromatographic behavior; therefore the group contribution concept is often used, since it was developed to predict activity coefficients in nonelectrolyte liquid mixtures (Petrovic et al., 1985). This concept is the basis for the UNIFAC model for chromatographic retention, which combines a model based on extension of quasi-chemical theory of liquid mixtures (UNIQUAC) with the concept of functional group solubility. In this method solute activity coefficients in the mobile and stationary phases are calculated via structural and binary parameters characterizing the mutual interaction energy of the
functional groups in the system. The solute activity coefficient is the product of a combinatorial and residual contribution. The combinatorial contribution is dependent on the size and shape of the molecules in the system; the residual depends on the interaction energy of functional group pairs, as well as the fraction of the surfaces on these groups which are available for mutual interactions (Petrovic et al., 1985).
Assuming infinite dilution, the relationship between the capacity factor of a solute i and its activity coefficient (fi) in the mobile and stationary phases can be written as
ln ki' = In Ki + ln (VS/Vm)
therefore in ki' =In fim ln fijs + In (Vs/Vm). Petrovic et al. (1985) assume that fsand Vs/Vm are constants; therefore ln ki and in fim are linearly related with a slope of one. If the activity coefficients of the solute i in the chromatographic system and Vs/Vm are known, retention can then be predicted. Petrovic et al. (1985) calculated infinite dilution activity coefficients of solutes in the mobile phase from experimental gas-liquid chromatographic data and then correlated them with their RP retention values to see how well the UNIFAC method could predict chromatographic behavior. They assumed that Vs/Vm would be the same for any octadecyl RP column at any methanol/water mobile phase composition (an assumption that will be thoroughly disputed in Chapter IV of this tome) and that
in fis was zero (since stationary phase interactions were assumed to be very weak and nonselective). They found that the predicted values of k' were at best a rough estimate of actual chromatographic retention and concluded that therefore the solute interaction with the stationary phase could not be ignored in the prediction of RP retention. However they found the UNIFAC method to be useful for predicting changes in relative solute retention with varying mobile phase composition, since the solute activity coefficients in the mobile phase could be accurately calculated using UNIFAC parameters (Petrovic et al., 1985).
Hydrophobicity is the physical descriptor which has most accurately been used to predict RPLC retention. Hydrophobic effects between solutes and hydroorganic mobile phases were described earlier in this chapter. Solute hydrophobicity is usually described in terms of the pi scale developed by H-ansch and Leo (1979). By evaluation of solute partition coefficients between n-octanol and water (P) they were able to establish substituent hydrophobicity parameters. The logarithm of the partition coefficient is determined for both a compound containing the substituent group and the parent compound; the difference in these two log P values is pi, the hydrophobicity parameter for the substituent (Melander and Horvath, 1980). Jinno (1982) and dinno and Kawasaki (1984a and 1984c) use the descriptors pi, HA and HD, where HA is the number of electron acceptor groups and H-D is the number of electron donor groups, to
correlate with RPLC retention for substituted benzenes (excluding phenol). They found a very good correlation between ln k' and pi, and poor correlations between in k' and molecular connectivity, correlation factor (number of double bonds plus number of primary and secondary carbons minus 0.5 for nonaromatic rings) and van der Waals volume and surface area. They interpret this to mean that the size and shape of these molecules were not dominant for controlling their retention. However, when the size and shape of the solute molecules are a dominant retention force such as for PAHs and hydroaromatics, the correlation factor
(F) has been found to correlate quite well with log k' (Hurtubise et al., 1982). If a linear combination of HA, HD and pi descriptors was used, they found an even better correlation between predicted and experimental k' values. For phenol solutes, hydrophobicity alone was not an adequate retention descriptor so they added Hammett's acidity parameters (sigma) to account for the strong hydrogen bonding ability of the phenols (Jinno and Kawasaki, 1984a and 1984c). Linear combinations of pi and sigma for the phenol solutes resulted in excellent correlations with retention behavior. However, for both sets of solutes the parameters had to be multiplied by certain proportionality constants in order to correctly predict retention and these constants were different not only for different organic modifiers in the mobile phase but also for each volume composition. This requires that at every different mobile
phase composition a least squares fitting for a large data set must take place in order to determine these proportionality constants, requiring a tremendous amount of data both in terms of pi (and possibly sigma) values for each solute and in terms of retention values for these solutes at each mobile phase composition. Jinno and Kawasaki (1984a and 1984c) also have not shown that this model is applicable for larger more complex molecules.
Funasaki et al. (1986) examined retention behavior for alcohols and ethers with positional and geometric isomers and examined the correlations between the solute's log k' and molecular cavity surface area (S), the logarithm of the aqueous solubility (log Cw) and the logarithm of the octanol-water partition coefficient (log P). As previously mentioned in this chapter, it was found that S and k' were very well correlated, even in the case of conformational isomers. This is because S for a molecule in water is defined as the area of the surface traced out by the center of a water molecule rolling over the van der Waals surface of the solute molecule (Funasaki et al., 1986). However S is very difficult to accurately calculate, requiring the construction of solute molecular models for rigorous work, as well as detailed knowledge of the molecular conformation of the molecule of interest. If S is calculated from group surface areas, an easier but less rigorous approach, the molecular surface area is usually overestimated for very crowded molecules. Using S to predict retention does have
one very strong merit--since the slope of a log k' versus S plot is related to interfacial tension, the dependence of log k' on the organic modifier content of the mobile phase can be predicted (Funasaki et al., 1986).
Funasaki et al. (1986) found the correlation between log k' and the logarithm of solute aqueous solubility, Cw, to be rough at best. The main drawbacks to this method are that the extent of correlation will depend on whether Cw was measured for the compound in the gas, liquid or solid state, that some compounds are infinitely water soluble and that the correlation is particularly poor for branched solutes. In contrast, they found that the correlation between the logarithm of the solute octanol-water partition coefficient (log P) and log k' was particularly good for the solutes examined (alcohols and ethers) in methanol/water mobile phases. Braumann (1986) reports that other workers have found good correlations between log k' and log P for a variety of compounds including PAHs, alkylbenzenes, substituted benzenes, pesticides, phenols and barbiturates; again correlations were much better with methanolic mobile phases than with those containing acetonitrile, due to methanol's hydrogen bonding properties. One drawback to this method is that octanol-water partition coefficients are measured via shake-flask methods, which are very time consuming. However, P values are tabulated for many compounds and additivity rules using Hansch's pi parameters may be used to estimate P for other compounds although
estimated P values do not in general correlate as well with retention as measured ones (Braumann, 1986; Funasaki et a]., 1986). This approach is very much like that of Jinno and Kawasaki (1984a and 1984c) previously discussed except that Jinno and Kawasaki examined substituent group hydrophobicity (pi parameters) in relation to retention whereas Funasaki et al. (1986) examined its relation with overall molecular hydrophobicity (log P).
Funasaki et al. (1986) also examined the effect of
temperature and mobile phase composition on the degree of correlation between log k' and log P. Not surprisingly, they found that they were better correlated when both were measured at the same temperature than when they were measured at different temperatures. They also found that the log k' value estimated from extrapolation to zero percent methanol (i.e. totally aqueous) mobile phase gave a better correlation with log P than those obtained with hydroorganic mobile phases. As in all other cases previously discussed, the extent of correlation will be dependent upon the test solutes, the chromatographic column and the experimental conditions used (Funasaki et al.,
In summary, most of the retention indices based on
solute descriptors tend to accurately predict retention for certain small sets of similar compounds; they are by no means universal. Additionally, some of them require extensive calculations and/or experimental data in order to
predict retention. At present, none of these solute descriptor index systems is adequate for reliable prediction of RPLC retention.
Empirical Prediction of Retention
Jandera and coworkers (Colin et al., 1983a; Jandera, 1986; Jandera et al., 1982; Jandera and Spacek, 1986) have developed an empirical model to predict absolute or relative retention. They assumed that the stationary phase contribution to retention is very small compared to that of the mobile phase and that nonpolar interactions between the solute, stationary phase and mobile phase cancel each other. If this is true, the energy of transfer of the solute from the mobile to the stationary phase will depend on the interaction energy (energy of cohesion) between mobile phase molecules and the interaction energy between the mobile phase and solute. They defined an interaction index, determined from retention data in hydroorganic mobile phase systems, which describes polar interactions between solute molecules and the mobile phase components.
The interaction index for a solute (Ix) can be
determined if the volume of interaction (Vx) for the solute and the column phase ratio (Vs/Vm) are known, since in this model,
log (kx'/Vx) log ((Vs/VmT)/Vx) = A BIx (1)
where A and B are constants which depend on the stationary and mobile phases used. Jandera et al. (1982) plotted (log kx' log (Vs/Vm))/Vx versus solute polarity (based on
Snyder's polarity index) to determine the value of Ix for "standard" solutes with different functional groups, and then found the average value of these Ix's for a given solute in many different mobile and stationary phases. They then assumed that IX will be constant for compounds which undergo the same type of interactions with the mobile phase as the "standard" solute does. Therefore if VX for a similar compound is known, its retention can be predicted from equation 1, since A, B and Ix are already known from the standard data Wandera et al., 1982).
Although this model is useful in a practical sense
since it is based on empirical considerations, it cannot be completely justified in a theoretical sense. Although the solute experiences stronger interactions with the mobile phase than with the stationary phase, Jandera et al. (1982) do not prove that stationary phase interactions can be completely ignored. They also state that the standard compounds must be chosen "correctly" or else the predictive power of the model fails; "correct" solutes will be those with little or no specific interactions (such as hydrogen bonding) between the solutes and mobile phase components. This severely limits the types of compounds whose retention can be predicted from this model. Finally, at present the model cannot give very accurate (about 5 to 20% accuracy) retention predictions (Jandera et al., 1982).
Theories of Retention in RPLC
Sol vophobi c Theory
In order to truly understand the retention process in RPLC and thereby be able to predict solute retention, the retention process must be examined at the molecular level. At present there are two main schools of thought on the retention mechanism of RPLC at the molecular level. The solvophobic theory espoused by Melander and Horvath (1980) states that RP retention comes about from solute binding onto the stationary phase fromn the mobile phase and is mainly due to hydrophobic interactions between the solute and the mobile phase. Other workers have utilized statistical mechanical analysis based on mean field lattice theory to show that RPLC solute retention is due to solute partitioning from the mobile phase into the bonded stati onary phase chains (Dill, 1987a and 1987b; Marqusee and Dill, 1986; Martire and Boehm, 1983). The main tenets of both of these proposed theories will be outlined below.
Melander and Horvath's (1980) solvophobic theory
assumes that the mobile phase plays the dominant role in the RPLC retention process. This is because the stationary phase is nonpolar; therefore the only attractive forces occurring between the stationary phase and a nonpolar solute will be van der Waals forces, which are weak and nonspecific. They attribute the interactions between the solute and the mobile phase to a type of hydrophobic effect, which was discussed earlier in this chapter. In the
specialized RPLC environment, they have adopted a variation of this effect, termed the "solvophobic" theory, since the hydrophobic theory assumes a totally aqueous environment and RPLC mobile phases are generally a mixture of aqueous and organic components. Solvophobic theory is based on a theory of solvent effects on chemical equilibria developed by Sinanoglu (1968). The theory states that chromatographic retention is based on the free energy change as the solute is transferred from a hypothetical gas phase at atmospheric pressure to the mobile phase. The energy involved in this process is calculated in two steps. In the first, a cavity of the proper shape and size for the solute molecule is formed in the solvent. In the second, the solute enters the cavity and interacts with the surrounding solvent molecules via van der Waals and electrostatic interactions (Melander and Horvath, 1980).
The free energy change accompanying the mobile phase
cavity formation comes about from the fact that the solvent surface area will increase by the molecular surface area of the solute (Melander and Horvath, 1980). Therefore the mobile phase free energy will increase by an amount proportional to the solvent surface tension and the increase in area. The change in free energy due to the interaction of the solute with the surrounding solvent molecules will be due to chemical and entropic effects. The chemical effects are van der Waals interactions and electrostatic effects. The van der Waals interaction energies are a function of the
polarizability and ionization potentials of the solute and solvent species. Electrostatic interactions consider both dipole and ionic effects; dipole effects are calculated from the solute dipole moment, polarizability and molecular radius as well as from the dielectric constant of the solvent. Ionic effects are calculated from conventional electrostatic theories such as Debye-Huckel treatments. The entropic term is a measure of solute "free volume," which is the volume that the molecule encounters before colliding with another molecule. The "free volume" is assumed to be proportional to the solute molar volume (Melander and Horvath, 1980).
Although the solvophobic theory outlined above is
pertinent to a liquid-liquid system, the bonded stationary phase has not been considered in this process. Melander and Horvath (1980) regard the change in free energy for retention to be a combination of the mobile phase effects just described and a small contribution from the adsorption of the solute onto the stationary phase surface. This
adsorption is viewed as a reversible reaction between the solute and stationary phase to form an associated complex. The free energy of adsorption is quantified as the van der Waals interaction energy between the solute and stationary phase in the absence of solvent molecules. Althouyh Melander and Horvath (1980) mention that an entropic term should be introduced to account for the restricted translational freedom of the bonded chains at the silica
surface, they ignore this effect because they feel its contribution is negligible. In summary, in the so] vophobic theory of RPLC retention, retention is mainly dependent upon the free energy of creation of a solute sized cavity in the mobile phase; stationary phase effects are considered to be weak and therefore rather negligible. Partitioning Theory
One of the most severe drawbacks to Melander and
Horvath's (1980) solvophobic theory is that it is based on a one phase model--that of the mobile phase. But RPLG involves two phases, the stationary and mobile phases; therefore a one phase model is not completely applicable for such a system. Melander and Horvath (1980) view the retention process as if there is no true transfer of the solute from the mobile phase to the stationary phase; the solute is merely associated with the stationary phase through weak adsorptive effects. Melander and Horvath (1980) also account for the stationary and mobile phase
interactions by viewing them as bulk phases with homogeneous properties throughout. In reality, the stationary phase/mobile phase boundary is a highly heterogeneous area consisting of the core silica particles, the alkyl chains bonded to the silica surface, residual silanol groups remaining on the silica surface and the mobile phase solvating these silanols and the bonded chains. In such a heterogeneous system, it is highly unlikely that bulk phase thermodynamic considerations based on ideal solution behavior are applicable (Marqusee and Dill, 1986).
In their adsorption model, Melander and Horvath (1980) use very simple descriptions of the stationary phase surface which contain two gross oversimplifications. They imply first of all that the bonded chains are rigid rods containing no internal degrees of freedom. But at the temperatures commonly used in RPLC, the bonded alkyl chains are quite disordered (Dill, 1987a). Melander and Horvath (1980) also use stationary phase models wherein the bonded alkyl chains are fully exposed to the mobile phase. However, in a hydroorganic mobile phase system the chains cannot be fully exposed to such a highly aqueous environment; such a configuration would be prohibitively expensive in free energy terms (Dill, 1987a).
Dill (1987a) has proposed an alternative model of the RPLC stationary phase surface which regards the grafted phase as an organized "interphase" similar to those found in surfactant aggregates such as monolayers, bilayers, micelles and microemulsions (Marqusee and Dill, 1986). Interphases are composed of alkyl chains that have one end anchored at an interface; their thickness is on the order of a few molecular dimensions. The anchored chain density (i.e. the number of chains anchored per unit silica surface area) is sufficiently high as to cause severe configurational constraints. Two important properties distinguish this system as an interphase: its surface area/volume ratio is high and its properties vary with the distance from the anchored end. The relationship between orientational order
and distance from the silica surface is termed a "disorder gradient"; the bonded chains have much greater orientational order at the anchored ends and this order decreases with the distance from the attached end. This is in contrast to bulk matter phases in which by definition properties are invariant with spatial position (Dill, 1987a).
In Dill's (1987a) retention theory, the nature of the retention process is dependent upon the nature of the molecular organization within the interphase. There are three factors which determine this organization: the first is those constraints imposed by the surface density and chain lengths of the alkyl groups bonded to the surface and by the surface's geometry; the second requirement is that in highly aqueous mobile phases the interphase region must largely exclude the solvent due to hydrophobic effects; the interphase volume will be filled by chain segments and solute molecules. The final requirement is that the chains adopt as much disorder as is consistent with the other two constraints in order to conform to the second law of thermodynamics. This approach allows the consideration of any possible geometry of the silica surface onto which the chains are bonded. Dill (1987a) assumes that the silica surface is approximately planar; in terms of molecular dimensions this should be a good approximation for chromatographic silicas, which commonly have pore diameters of 60 to 100 Angstroms or more; the effects of curvature will be small unless the radii of curvature are a few molecular chain lengths or less (Marqusee and Dill, 1986).
Dill (1987a) considers partition and adsorption
separately as alternative RPLC retention mechanisms. In both cases a lattice interphase mode] is used for the bonded phase surface and statistical thermodynamic calculations are used to predict solute retention in the system (Dill, 1987a; Marqusee and Dill, 1986; Martire and Boehm, 1983). Dill (1987a) predicts the equilibrium partition coefficient for a solute from the mobile phase to the stationary phase from the chemical potentials of the solute in the mobile phase system and in the bonded chain interphase. These calculations include the entropy of mixing of the solute and solvent in the mobile phase, the decrease in configurational entropy of the bonded chains when the solute is inserted within the interphase and the total contact free energy of the system, which will be due to intermolecular interactions of the molecules with their neighbors. After careful consideration of both the partitioning and adsorption retention mechanisms in conjunction with his interphase model and available experimental evidence, Dill (1987a) concludes that the principal retention mechanism in RPLC is partitioning due to two lines of evidence. Partitioning will be affected by the surface density of the bonded alkyl chains; adsorption will not. Therefore if partitioning is dominant, after a certain critical bonding density, solute retention should decrease with increasing alkyl chain surface density. It has been observed that there is less solute retention in bonded alkyl phase stationary phase
systems than in the corresponding bulk alkane systems (Colin et al., 1983b). Although Melander and Horvath (1980) have interpreted this as favoring an adsorption mechanism, Dill (1987a) interprets this in terms of partial chain ordering in the stationary phase, leading to less retention than in the completely disordered bulk alkane. This is supported by the work of Lochmuller and Wilder (1979) since solute methylene selectivities should be unaffected by the molecular organization of the interphase (Dill, 1987a).
The other line of evidence is that ln k' for congeneric sets of molecules can be linearly correlated with ln P (octanol-water partition coefficient) with a slope of one. A slope of one is expected for the partitioning mechanism, since all of the solute surface area would be available for partitioning within the interphase. The slope for the adsorption mechanism is expected to be considerably less (about 1/6) because only a small fraction of the solute surface area would contact the hydrocarbon chains, giving a smaller driving force for retention (Dill, 1987a). Experimental evidence has resulted in linear plots of ln k' versus ln P with a slope of one (Melander and Horvath, 1980) .
Based on the partitioning mechanism of retention, Dill (1987b) predicts that at low surface densities (less than 2.1 micromoles of bonded alkyl chains per square meter of silica surface) nonpolar solute retention will increase linearly with increasing surface density. At these low
densities, chain configurational constraints are very small and interphase chain packing will have no effect on solute retention; solute retention will increase as the volume of chains increases since there will be more alkyl phase for the solutes to partition into. At a chain density of zero (bare silica) the nonpolar solute will be unretained (Dill, 19 8 7b) .
Once a critical bonding density (predicted to be about 2.7 wPmol/m2) is reached, the bonding density is high enough for severe configurational constraints to result. In the Dill (1987a) retention model, the free energy involved in retention is determined by the differences in the free energy between the creation of a solute sized cavity in the interphase region and the destruction of a solute sized cavity in the mobile phase. As alkyl bonding density is increased past the critical bonding density, the chain packing constraints become more and more severe, requiring larger amounts of energy to create a solute cavity in the interphase structure. Thus in the high density region solute partitioning is predicted to decrease with increasing alkyl bonding density due to the increasingly prohibitive expenditure necessary for interphase cavity creation to accomodate the solute. Dill (1987a) predicts that at 8.1 Pmol/m2, the maximum achievable bonding density if every surface hydroxyl group on the silica is derivatized with an alkyl ligand, solute retention would be zero and the solute would be completely excluded from the interphase chain packing structure.
Some researchers have examined solute retention as a function of increasing alkyl chain length of the bonded ligands. Colin et al. (1983b) and Jinno and Kawasaki (1984c) found that log k' increased with increasing bonded chain length, while Melander and Horvath (1980) have stated that more or less contradictory results have appeared in the literature on the influence of chain length. Spacek et al. (1980) and Berendsen and de Galan (1978b) have also noted a general trend of increased retention with increasing chain length of the bonded ligand, but their results were less conclusive than those of Jinno and Kawasaki (1984c). It is not clear from any of these studies whether this trend is due to the actual increased partitioning of the solute into the longer alkyl chains or whether this trend is an artifact of the retention parameter measured. This is because in all of these studies, it is the capacity factor, k', that is used to quantitate retention. But k' = K(Vs/Vm) where K is the chromatographic partition coefficient and Vs/Vm is the volume phase ratio of the stationary and mobile phases. It is obvious that as the chain length of the bonded alkyl ligand is increased, a corresponding increase in V5 will occur, as pointed out by Colin et al. (1983b). Therefore it is unclear as to whether solute retention increased because of increasing partition coefficient or merely because of the phase ratio increase. In order to reliably determine the cause of changes in actual solute retention as stationary phase parameters are changed, the changes in the solute partition coefficient must be examined.
In this study, wie examined the effects of octadecyl
alkyl chain bonding density on both retention and selectvity of small nonpolar salutes. We were particularly interested in experimental verification of the molecular mechanisms of RPLC retention proposed by Dill (1987a and 1987b). Although Sander and Wise (1984a and 1984b; Wise and Sander, 1985) and Wise and May (1983) have extensively examined the effect of alkyl bonding density on retention and selectivity for PAHs, they have mainly examined polymeric stationary phases, which are not as well structurally characterized as the monomeric stationary phases used in our study. Another problem with their work is that they examined capacity factor (k') behavior in their retention studies, which fails to account for phase ratio changes. Additionally, PA~s are not ideal salutes for such a study, since their large sizes and unusual shapes are not typical of most chromatographic s o1 ut es .
In order to determine the effect of interphase chain packing on solute partitioning, the behavior of the chromatographic partition coefficient was examined as a function of octadecyl bonding density. Chromatographic selectivity was also studied as a function of bonding density for salutes of different sizes and shapes. Novel synthetic methods utilizing ultrasound as a reaction driving force were devised to obtain stationary phases with high bonding densities. In this manner, we were able to see if Dill 's (1987a and 1987b) proposed RPLC retention mechanism
was verified by actual chromatographic behavior. These experiments were carried out using both silica and controlled pore glass substrates in order to compare the two materials as stationary phase supports.
SYNTHESES OF SILICA-BASED RP STATIONARY PHASES
Experimental Considerations in the Synthesis of RP
Stati onary Phases
Reversed phase bonded silicas are the most popular packings used in high performance liquid chromatography (HPLC). Although the role of the mobile phase in chromatographic retention and selectivity has been extensively studied, that of the stationary phase has only come under intense scrutiny recently and as a result the effects of the stationary phase on these chromatographic properties is not yet fully understood. One reason for this dearth of knowledge is the lack of precise and reliable methods for determining bonded phase characteristics such as the density, homogeneity and topographical distribution of the bonded alkyl ligands and the residual hydroxyl groups on the support surface. These properties are a direct consequence of the bulk silica medium and the reagent and reaction conditions for the silanization process (Kinkel and Unger, 1984). In order to obtain reversed phase packings with reproducible surface characteristics, the silanization reaction conditions must be painstakingly controlled.
In the preparation of reversed phase packings, one
objective is the modification of as many surface hydroxyl
groups on the silica as possible, especially the highly acidic isolated silanols. These residual isolated silanol groups have been shown to be the main cause of tailing of chromatographic peaks for basic compounds, of mechanical instability of the packing, and of low sample capacity of the column (Kohler et al., 1986; Kohler and Kirkland, 1987). Di- or trireactive alkylsilanes had previously found favor over monoreactive silanes because of their greater reactivity and the possibility of reacting simultaneously with two or three hydroxyl groups. However, any unreacted sites on the bonded functional groups will be hydrolyzed upon contact with water (i.e. from the mobile phase), forming additional undesirable silanol groups (Berendsen and de Galan, 1978b; Snyder and Kirkland, 1979). Di- and trireactive silane reagents also often result in nonreproducible stationary phases since the degree of polymerization is highly dependent on the residual water content of the silica and the reagents used in the bonding reaction (Snyder and Kirkland, 1979). Another drawback of polymeric stationary phases is their lower chromatographic efficiency, which results from poor solute mass transfer in these relatively thick stationary phases. Therefore many investigators now advocate the use of monofunctional silanes for the silica derivatization reaction, since this results in a reproducible and well defined chemically bonded phase. Additionally, monomeric stationary phases generally exhibit superior column performance to polymeric phases due to their
faster solute mass transfer kinetics (Cooke and Olsen, 1980). For octadecyldimethylchlorosilane, the most commonly used monoreactive silane, the resulting bonding reaction is depicted in Figure 2-1.
Kinkel and Unger (1984) have studied the roles of the solvent and the base in these monofunctional bonding reactions and have found their choice to be crucial. When alkylhalosilanes are reacted with silica, a base is added to serve as the acid-acceptor catalyst, binding the haloacid formed during the reaction and driving the equilibrium to the product side. In addition, the base favorably affects the kinetics of the silanization reaction. Mechanistic studies of these types of reactions (Corriu and Guerin, 1980) have shown that two molecules of base attack one molecule of silane, activating the Si-X bond such that a reactive intermediate and a hydrohalide are formed. Formation of this reactive intermediate greatly increases the kinetics of the bonding reaction; indeed, the addition of the acid-acceptor catalyst results in approximately 90% of the total conversion taking place within the first hour of the reaction. In their study, Kinkel and Unger (1984) found that the two most effective acid-acceptor catalysts for organohalosilanes were imidazole and 2,6-lutidine.
The reaction solvent must also be carefully chosen.
The solvent can interact specifically with the silane, the base and the surface silanol groups on the silica. When the solvent interacts with a silanol group, there is a
Si -0-H + CI-SH-CH)FCH3 --o Si-0-SCH)rTCH3 + HCI
Figure 2-1. Bonding reaction for monomeric octadecyl
reversed phase packings.
considerable effect on the strength of the bond between the silicon and oxygen atoms. Solvents which have both a pronounced Lewis acid and Lewis base character cause the Si-O bond strength to be weakened and facilitate the bonding reaction. The solvent can also activate the silicon atom of the organohalosilane by forming a pentacoordinated intermediate through nucleophilic attack. The resultant bond lengthening causes nucleophilic activation to occur, favoring attack by a second nucleophile (such as the base). The solvent may influence the base as well, as it is known that in aprotic polar solvents the nucleophilic character of reactants is more pronounced. All of these considerations may have a synergistic relationship as well. Based on their experimental work with organohalosilanes, Kinkel and Unger (1984) found that methylene chloride and N,N-dimethylformamide were the most effective solvents for the bonding reaction.
Many organic reactions have been shown to be enhanced
by ultrasound (Boudjouk, 1986; Bremner, 1986; Clough et al., 1986; Han and Boudjouk, 1982 and 1983; Suslick, 1986). Boudjouk and Han (1981) have shown that in the presence of ultrasonic waves both alkyl and aryl chlorosilanes could be coupled over lithium wire; without ultrasonification, this reaction occurred to no appreciable extent. Reactions at solid-liquid interfaces are also particularly enhanced by ultrasound (Bremner, 1986; Suslick, 1986). It is then reasonable to assume that reversed phase bonding reactions
might be facilitated under ultrasonification. The use of ultrasound has two distinct advantages over traditional reflux methods. The ultrasonic waves serve as a driving force which is controlled independently of temperature, allowing reaction temperatures to be varied over any desired range. Secondly, the power of the ultrasonic driving force can be varied by using a variable power ultrasonic probe.
We have investigated the effect of ultrasound on the
silane bonding reaction, including the effects of subambient and superambient temperatures on the ultrasonic reaction. In addition to these investigations, a novel base, 4-dimethylaminopyridine, was utilized as the acid-acceptor catalyst, in hopes that it might prove superior to 2,6-1utidine.
All of the organic solvents used were supplied by Fisher Scientific (Fairlawn, NJ). Water was deionized, passed through a Barnstead Nanopure (Boston, MA) purification system, irradiated in a Photronix Model 816 HPLC reservoir with a UV source (Photronix Corp., Medway, MA) for at least 48 hours, and filtered through a 0.45 im Nylon 66 membrane (Rainin, Woburn, MA). The methanol used was HPLC grade; the chloroform, methylene chloride, and diethyl ether were reagent grade. Methylene chloride was dried by stirring over phosphorus pentoxide (Fisher Scientific) for 24 hours, followed by distillation under a dry nitrogen atmosphere.
Dimethyloctadecylchlorosi lane, n-octyldimethylchlorosilane and trimethylchlorosilane (99.9%) were used as received from Petrarch Systems (Bristol, PA). The 2,6-lutidine (Sigma Chemical Co., St. Louis, MO) was stirred for 24 hours over barium oxide (Fisher Scientific) prior to distillation under dry nitrogen atmosphere; 4-dimethylaminopyridine (4-DMAP; Nepera Inc., Harriman, NY) was oven dried at 80 C for 24 hours before use.
The chromatographic silica was from a single lot of Davisil (W. R. Grace, Baltimore, MD) synthetic amorphous silica, grade 641LCOX1823. The silica had an average pore diameter of 147 Angstroms, an absolute surface area (SBET, as measured by BET analysis) of 300 m2/g, a particle size range of 20-30 pm with an 80% distribution of 23 + 10 pm and a nitrogen pore volume of 1.10 cm3/g (Grace, 1984). As recommended by Snyder and Kirkland (1979) the silica was acid leached in 0.1 M nitric acid at 90 C for 24 hours in order to fully hydroxylate the surface and to remove any metal contaminants remaining from the manufacturing process. The silica was then washed thoroughly with water until all traces of the nitric acid had been removed and dried under vacuum at 240 'C for 24 hours prior to use in order to remove physically adsorbed water from the surface (Unger, 1979). Scanning electron micrographs of the acid-leached silica (Figures 2-2 and 2-3) show that this type of chromatographic silica exhibits an irregular shape as well as an irregular surface. This surface irregularity is reflected in the silica's high surface area.
Figure 2-2. Scanning electron micrograph of acid-leached Davisil
silica; 500X magnification.
Figure 2-3. Scanning electron micrograph of acid-leached Davisil
silica; 3000X magnification.
Silane Bonding Reaction
It is essential that the silane bonding reaction be carried out under scrupulously dry conditions in order to prevent the water-initiated dimerization of the silane reagent. Glassware used in the derivatization reaction was presilanized by etching the surface with a 10% (v/v) hydrofluoric acid (Fisher Scientific) solution, drying, and then soaking the glassware for an hour in a 5% (v/v) trimethylchlorosilane in chloroform solution. Immediately prior to use, the glassware was oven dried at 125 'C for at least 4 hours in order to remove trace moisture and allowed to cool in a dry box under nitrogen atmosphere. The reagents were mixed together in the dry box and the reaction flasks kept under dry nitrogen atmosphere at all times. Based on Kinkel and Unger's (1984) estimation of a maximum of five micromoles of reactive hydroxyl sites per square meter of silica surface, a twofold excess of the silane reagent was added to achieve exhaustive derivatization of the silica surface. A fourfold excess of the base (2,6-lutidine or 4-DMAP) was added both to serve as an acid-acceptor catalyst for the HCI produced in the reaction and to act as a reactive intermediate at the silica-solution interface. Dry methylene chloride was used as the reaction solvent, using a ratio of 10 ml of inethylene chloride per gram of base
The reaction flasks were sonicated by immersion to the flask neck in an ultrasonic cleaning bath (Bransonic model
B-2200R-1, Branson Cleaning Equipment Co., Shelton, CT) with a power rating of 100 W and a frequency of 55 kHz. Stirring of the reagents within the flasks was accomplished by rotating a magnetic bar submerged in the bath adjacent to the reaction flask, resulting in the corresponding rotation of a magnetic stirring bar within the flask. Temperature control of the ultrasonic bath was accomplished by passing a thermostatted solution of ethylene glycol and water through coiled copper tubing lining the inner perimeter of the bath. The solution was thermostatted by an Endocal or Exacal water bath (Neslab Instruments, Portsmouth, NH). Refluxed reactions were carried out at 50 'C using an oil bath and magnetic stirrer. Control reactions were carried out by stirring the reaction mixture at room temperature.
Refluxed reactions utilizing n-octyldimethylchlorosilane as the reactive silane, 2,6-lutidine as the acidacceptor catalyst and methylene chloride as the reaction solvent were carried out for reaction times of 24, 36 and 48 hours in order to ascertain whether the differences in reaction time would make a statistically significant difference in the reaction yield. Nine replicate reactions were performed for each reaction time. Student's t (tcalc) was calculated from the pooled standard deviation of percent carbon for all 27 reaction yields, the differences in the mean percent carbon at each reaction time, and the number of replicates at each time in order to determine if the mean yield for each reaction time was statistically different
from those at the other reaction times. In the comparison of the 24 hour reactions to the 36 hour reactions, tcalc was
0.984; for the 24 hour reaction time versus the 48 hour reaction time, tcalc was 0.148 and for the 36 hour reaction time versus the 48 hour reaction time, tcalc was 0.835. Since all of the calculated t values are less than the critical t value (tcrit) at the 80% confidence level (tcrit,80%= 1.34 for 16 degrees of freedom), there is no significant statistical difference in the reaction yields for reaction times of 24, 36 and 48 hours at the 80 confidence level (Peters et a]., 1974). Therefore, reactions were carried out for 24 hours (as also recommended by Kinkel and Unger (1984)) unless otherwise noted.
Once the reaction time was complete, the product was washed in order to remove excess reagents. Each bonded phase product was washed three times with each solvent using the rinse sequence methylene chloride, methanol, 50/50 (v/v) methanol/water, methanol and diethyl ether. After the ether was allowed to evaporate from the product, the derivatized silica was dried in a vacuum oven at 125 'C for 16-24 hours. (Caution: It is imperative that the ether be completely evaporated from the product prior to drying the product in the vacuum oven in order to avoid a possible explosion.) Products were analyzed by in-house elemental analysis performed at least in duplicate for each sample. Reliability of the elemental analysis was confirmed by repeated submission of a standard packing material over a
two year period; for 66 measurements the resultant standard deviation was + 0.20% carbon.
Syntheses Utilizing Ultrasonic Waves Use of Ultrasound as a Reaction Catalyst
The region of frequencies above 16 kHz is beyond the sensitivity of the human ear; it is therefore termed the ultrasound region. The first reported use of ultrasound in organic chemistry was in 1938, but it was not until the late 1970s that ultrasound was used to speed reactions in nonaqueous media; indeed the use of ultrasound as a reaction catalyst is still in its infancy (Boudjouk, 1986; Bremner, 1986). Ultrasonic radiation can be introduced to the reaction medium either by immersion of the reaction vessel into the liquid of a common laboratory ultrasonic cleaning bath or by introduction of an ultrasonic generating probe directly into the reaction medium.
Ultrasonic frequencies span the range of 20 kHz to 10 MHz, with associated acoustic wavelengths of 7.6 to 0.015 cm. Therefore sonochemistry cannot be accounted for in terms of direct coupling of the acoustic field with chemical species on a molecular level (Suslick, 1986). However, the effects of ultrasound can be attributed to three different phenomena. The variation of sonic pressure causes the rapid movement (oscillation) of fluids, subjecting them to compression and rarefaction. Negative pressure in the rarefaction region gives rise to cavitation, the formation and collapse of microbubbles. The violent implosion of
these microbubbles generates powerful shock waves with a considerable energy output (Boudjouk, 1986; Bremnner, 1986; Suslick, 1986). Pressures in the kilobar range and temperatures of 2000-3000 'C have been estimated in the region of the collapsing bubble for time periods in the nanosecond range (Sehgal et al., 1980). The third contributing phenomenon is microstreaming, where a large amount of vibrational energy is put into small volumes with little heating (Bremner, 1986). The extremes of temperature and pressure generated by ultrasonic waves cause the generation of free radicals and ions, the dispersion of chemical layers and the promotion of intimate contact between reactants. Emulsification of immiscible liquids and enhanced mass transfer at solid-liquid interfaces are secondary effects of ultrasonification. All of these effects can contribute to the promotion of chemical reactions (Bremner, 1986; Suslick, 1986). Effect of Ultrasound on the Bonding Reaction
In order to define the surface coverage of the bonded silica in an unambiguous and pertinent manner, the surface coverage should be expressed as the number of silane molecules attached to the surface, usually as micromoles of bonded silane molecules per square meter of silica surface, taking into account the increase in weight of the silica after the bonding reaction. These surface coverages are calculated from the percentage of carbon as obtained from elemental analysis of the bonded phase (Berendsen and de
Galan, 1978b). This calculation is quite straightforward for monoreactive silanes and for monochlorosilanes (the most commonly used monoreactive reagents) can be expressed by = (%C) (106) (1)
(12.011) (nc) (S) (lO0-L(%C/(12.011)(nc)](M-36.5)) where a is the surface coverage (pmoles/m2); %C is grams carbon per 100 grams bonded silica, as obtained from elemental analysis; nc is the number of carbon atoms per mole silane; M is the molecular weight of the silane; and S is the surface area of the native silica in m2/g. Although
the typical value for the average surface hydroxyl concentration of amorphous silica is 8 imol/m2 (Cheng and McCown, 1985), in practice octadecyl bonded phase coverages are limited to about 3 Pmol/m2 due to steric considerations (Berendsen et al. 1980a; Berendsen and de Galan, 1978a and 1978b; Cheng and McCown, 1985; Snyder and Kirkland, 1979).
Three sets of experiments were compared in order to determine the effect of ultrasound on the silica bonding reaction. In all cases dimethyloctadecylchlorosilane was the reactive silane, methylene chloride was the reaction solvent and 2,6-lutidine the acid-acceptor catalyst; all reaction mixtures were stirred during the reaction time period (24 hours). In the first set of experiments, the reaction mixture was stirred at ambient temperature (22.0 'C); in the second set the reaction mixture was refluxed at 50.0 'C. The third set of experiments was performed at 28.5 'C, but the reaction vessels were immersed in an ultrasonic
cleaning bath. The refluxed stationary phases had an average bonding density (+ one standard deviation over three trials) of 2.82 + 0.02 pmol/m2. The room temperature reaction resulted in a bonding density of 2.69 pmol/m2 with a range of + 0.03 Pmol/m2 over two trials; the ultrasound reaction gave a bonded phase (over two trials) with an average bonding density of 2.71 + 0.01 umol/m2. The small bonding density difference between the stirred reaction at ambient temperature and the one at reflux temperature is not surprising as Lork et al. (1986) have shown that the bonding density increases slightly and in a linear fashion with increasing reaction temperature when monochlorosilanes are used as the silanizing reagent. These experimental results show that ultrasound is indeed a viable method for the bonded phase synthesis, giving results which are comparable to those obtained using traditional reflux techniques.
Effect of Subambient Temperature on the Ultrasound Reaction
Two sets of experiments were performed using ultrasound in conjunction with subambient reaction temperatures. In achieving high bonding densities one of the greatest obstacles is increasing steric hindrance at the silica surface as more and more bulky dimethyloctadecylsilyl groups are bonded to the surface. It is possible that at low temperatures the bonding density might be enhanced due to the increased order (decreased entropy) in a lower temperature system. It is here that the ultrasound reactions are most unique, as they allow the temperature of the
reaction to be controlled independently of the ultrasonic driving force. Additionally, low reaction temperatures have often been found to enhance reaction yields for ultrasonically catalysed chemical reactions. One explanation for this phenomenon is that low temperatures cause the vapor pressures of the reactants to be decreased, enabling increased efficiency of ultrasonically produced cavitation (Boudjouk, 1986; Bremner, 1986; Suslick, 1986). In order to overcome the slower kinetics expected at lower temperatures, reaction times were increased beyond the usual 24 hour time period.
In the first set of experiments, two reaction vessels were sonicated and stirred at 15.0 'C for 48 hours with a resultant average bonding density (+ the range) of 2.74 +
0.00 Pmol/m2. Since this result was little different from that at room temperature, it was decided to increase the reaction time as well as to decrease the reaction temperature. In this set of experiments, two reaction flasks were sonicated and stirred at 8.5 0C for 101 hours with a resultant average bonding density of 2.84 + 0.01 pmol/m2, a slightly higher value than for those ultrasound reactions run at higher temperatures. These preliminary results indicated that subambient temperatures could indeed enhance the ultrasonic silica bonding reaction. The Use of 4-Dimethylaminopyridine as the Acid-Acceptor Catalyst
There are also advantages in the use of 4-dimethylaminopyridine (4-DMAP) as the acid-acceptor catalyst. The
presence of the dimethylamino group should serve to make this base better at forming a reactive intermediate with the silane than 2,6-lutidine. In addition it has a relatively high melting point (108-110 'C) which allows it to be oven-dried rather than necessitating distillation to remove adsorbed water. The odor of 4-DMAP is also quite mild in comparison to that of 2,6-lutidine.
The first set of experiments using 4-DMAP as the acid-acceptor catalyst was performed using methylene chloride as the solvent and dimethyloctadecylchlorosilane as the reactive silane. The reaction mixture was refluxed and stirred at 50.0 0C for 24 hours. The bonded phase product had an average bonding density (+ the range for two trials) of 3.44 + 0.02 imol/m2, much higher than that achieved in our previous syntheses using 2,6-lutidine (2.82 + 0.02 Pmol/m2). This bonding density is also greater than that (3.34 imol/m2) achieved by Kinkel and Unger (1984) under reflux conditions using methylene chloride and 2,6-lutidine. This is especially significant because the silane in our experiments was used as received from a commercial source; Kinkel and Unger synthesized and then distilled their silane under reduced pressure in order to obtain a reactive silane of the utmost purity. Silane purity has been shown to be a very important factor in obtaining high bonding densities (Kinkel and Unger, 1984).
The second set of experiments using 4-DMAP as the
acid-acceptor catalyst was run with the same reagents as
described above; the reaction mixture was immersed in the ultrasonic bath and stirred at a temperature of 31.0 'C for 24 hours. The average bonding density of the resultant bonded phases (+ the range for two trials) was 3.35 + 0.05 i mol/m2, again much higher than that achieved under similar circumstances using 2,6-1 uti dine as the aci d-acceptor catalyst (2.71 + 0.01 p.mol/m2).
A low temperature ultrasound reaction was then carried out under the same conditions as stated above, with a reaction temperature of 4.0 'C for a duration of 97 hours. For the two trials, an average bonding density of 3.24 +
0.01 ijmol/m2 was obtained. A second set of low temperature ultrasound reactions was performed under analogous conditions with a reaction temperature of 3.0 'C for 144 hours. The bonded phase resulting from this experiment had a higher bonding density than achieved in any of our previous attempts; the average + the range for the set was
3.60 + 0.01 ,imol/m2. To our knowledge, this bonding density is higher than any previously reported in the literature using dimethyloctadecylchlorosilane as the reactive silane (Berendsen et al., 1980a; Cheng and McCown, 1985; Kinkel and Unger, 1984).
To ensure that the high bonding density in this second low temperature experiment was a result of the ultrasonic driving force as well as the lengthy reaction time, two other reactions were carried out. In one, the reaction was performed exactly as above (at 3.0 0C for 144 hours with
stirring) except that the reaction flask was not sonicated. In the other, the reaction was stirred for 144 hours, but the reaction mixture was refluxed at 50.0 'C rather than sonicated. From duplicate elemental analyses of each of these two materials, the average bonding density (+ the range) for the silica stirred (but not sonicated) at 3.0 C for 144 hours was 3.48 + 0.00 Pmol/m2; that for the silica refluxed and stirred at 50.0 'C for 144 hours was 3.44 + 0.03 Pmol/m2. Since the absolute error in the elemental analysis is + 0.20% carbon, which corresponds to + 0.03 Pmol/m2 for the octadecyl packings, the differences in bonding density between these two materials and the silica which was sonicated at 3.0 0C for 144 hours (3.60 pmol/m2) is both real and significant. Therefore it can be concluded that subambient ultrasound reactions are especially efficacious for synthesizing stationary phases with very high alkyl bonding densities.
In order to investigate the effect of superambient
temperatures on the ultrasonic bonding reaction, two types of experiments were performed. In the first, the reaction was carried out by stirring with the same reagents as previously described for a reaction time of 24 hours and with the ultrasonic bath maintained at a temperature of 50.0 'C. For two trials, the average + the range was 3.34 +
0.04 Pmol/m2, virtually identical to that achieved under ambient ultrasonic conditions (3.35 + 0.05 pmol/m2). In the second experiment, the reagents were stirred and sonicated
at 31.0 'C for 1 hour and then refluxed and stirred at 50.0 C for an additional 23 hours, in hopes that the preliminary sonication of the reagents would permit greater accessibility of the reactive silane to silanols located deep within the silica pores. The resulting bonding density (for two trials) of 3.42 + 0.03 jmol/m2 is comparable to that achieved under reflux conditions alone (3.44 + 0.02 Pmol/m2). These results indicate that silica bonding reactions performed in an ultrasonic bath are not affected by superambient temperatures; this is in contrast to those performed at subambient temperatures, which were found to give an increasing yield as the temperature was decreased.
Experiments were also carried out in the ultrasonic bath at 28.0 0C for 24 hours using trimethylchlorosilane (TMCS) as the reactive silane, 4-DMAP as the base and methylene chloride as the reaction solvent. TMCS is a much smaller molecule than the octadecyl silane and therefore should approximate the maximum bonding density obtainable in these reactions when steric hindrance is minimized. The average bonding density achieved in the two trials was 3.51 + 0.01 Pmol/m2; the octadecyl bonding densities achieved in the above reactions show that the TMCS bonding density at ambient temperatures can be exceeded under subambient conditions even with bulky octadecyl reagents. The results of these 4-DMAP experiments, as summarized in Table 2-1, demonstrate that it is indeed a superior acid-acceptor catalyst to 2,6-lutidine for reversed phase bonding reactions.
Table 2-1. Comparison of silica octadecyl bonding densities using
4-DMAP and 2,6-lutidine as acid-acceptor catalysts.
Reaction Reaction C18 Bonding Density (pmol/m2)
Conditions Temperature (oC) Time (h) 4-DMAP 2,6-lutidine
Refluxed 50.0 24 3.44 2.82
Ultrasound 28.0 24 2.71
Ultrasound 31.0 24 3.35
Ultrasound 8.5 101 2.84
Ultrasound 4.0 97 3.24
Ultrasound 3.0 144 3.60
Stirred Only 3.0 144 3.48
Refluxed 50.0 144 3.44
The use of ultrasound as a driving force for reversed
phase syntheses has been shown to be a viable synthetic procedure. Ultrasonic syntheses performed at subambient temperatures have proven to be especially effective for the production of high alkyl bonding density stationary phases. The use of 4-dimethylaminopyridine as the acid-acceptor catalyst is recommended due to its ease of use and the resulting high bonding densities. The small ranges of the bonding density for duplicate syntheses show that reversed phase packings with reproducible bonding densities can be synthesized by these methods.
SYNTHESES OF CONTROLLED PORE GLASS-BASED RP STATIONARY PHASES
Comparison of Controlled Pore Glass and Silica as Supports for RP Stationary Phases
The most commonly used column packing materials for reversed phase liquid chromatography (RPLC) are based on microparticulate silica. As has been described in Chapter II, this material is modified by chemically bonding alkyl chains of the desired length onto the silica surface. The use of such siliceous supports is widespread due to their high reactivity and relatively low cost. S ilIica-based RP bonded phases also exhibit good column stability within the pH range of 2.5 to 7.5 (Melander and Horvath, 1980).
However, such materials are not without problems. The surface of silica gel is very porous in nature and there is a wide distribution in the size of these pores. This can affect chromatographic selectivity by causing size exclusion effects. This broad pore size distribution is one of the contributors to the problem of inhoinogeneous energies of transfer between the stationary and mobile phases for solutes in RPLC, leading to distorted peak shapes and decreasing chromatographic efficiency. The effects of pore size and structure have received much attention in size exclusion chromatography, but these parameters have garnered 52
little attention in reversed phase systems (Sander and Wise, 1984b). At present, there is no satisfactory explanation of the effect of pore size distribution on the properties of hydrocarbonaceous bonded phases; however differences in the pore structure of the support material may account for some of the differences observed in the RPLC behavior of various commercial bonded phases having the same alkyl chain length but different siliceous substrates (Melander and Horvath,
Controlled pore glass (CPGY) offers an ideal medium for investigating the effects of pore size and structure on RP retention and selectivity. CPG, which has mainly been used in size exclusion chromatography, consists of nearly pure quartz glass with pores of uniform size. In CPG, the pore diameter is the same at the surface as it is in the interior of the particle; 80% of the pores show a deviation of less than + 10% from the nominal pore diameter (Fluka). Chemical modification of the surface of CPG is accomplished by reacting surface silanol groups with the appropriate reactive silane, as has been described for silica in Chapter II. Although such reactions have been performed to prepare CPG RPLC bonded phases (Dawidowicz et al., 1983; Dawidowicz and Rayss, 1985; Rayss et al., 1983; Suprynowicz et al., 1978 and 1985) commercial bonded phases based on CPG are currently impractical due to its much greater expense compared to that of silica.
The differences in the pore structures of silica and CPG come about from the differences in their chemical compositions and in their manufacturing processes. The manufacture of chromatographic silica is described in detail by Unger (1979). The starting materials in the manufacture of porous silica are soluble silicates such as sodium silicates, silicon tetrachloride or tetraalkoxysilanes. By adjusting the pH of an aqueous solution of the starting material within a range of 8 to 9, silica sols are made. In the sol, polysilicic acids are formed by polycondensation and polymerization, growing into colloidal particles ranging from 1 to 100 nm. The sol consists of spherically shaped, nonporous and amorphous discrete silica particles. Unless stabilized, the discrete particles in the sol aggregate, mainly due to gelling. The particles become linked together to eventually form a three dimensional packing of silica particles that is a gelatinous mass called silica hydrogel. The hydrogel is washed and water is then removed by heating. This dehydration results in shrinkage from the partial collapse of the globular hydrogel structure; the resultant xerogel consists of hard porous grains. The silica particles are also cemented together by dissolutiondeposition processes. The conversion of the hydrogel to the xerogel is the origin of the porosity of chromatographic silica. This porosity comes about from compaction of the dispersed silica in the hydrogel; the pore space is made up of the interparticle interstices and voids. This results in
a totally porous structure; moreover these pores are quite irregularly shaped. Factors which can be varied to control
the final pore structure include changes in the sol and/or hydrogel pH, changes in the duration of the hydrogel ripening (via stabilization of the sol), variation of pH during washing of the hydrogel, and substitution of an organic liquid wash for the hydrogel rather than an aqueous one (Unger, 1979). The chemical composition of the final amorphous silica can be exemplified by the composition of the Davisil silica used in our experiments; it consists of 99.60% by weight of SiO2 and 0.10% Na20 with the remaining
0.30% made up of other metal oxides (Grace, 1984).
The procedure used to produce controlled pore glasses was first reported by Wolfgang Haller in 1965 (1965a and 1965b). The starting material consists of a \lycor type glass consisting of 7% Na2O, 23% B203 and 70% Si02. The glass is crushed, fractionated to the desired particle size distribution by sieving, and then heated at approximately 600 'C for the desired number of hours. This heating period causes fusion to take place within the glass, resulting in the formation of microheterogeneous regions in the continuous silica network. This alkali borate-rich microphase is then removed from the glass by a series of acidic and basic leachings, resulting in a finished material which is porous throughout its entire volume (Dawidowicz et al., 1983). The diameter of the pores is determined by the length and temperature of the heat treatment (Haller,
1965a). The chemical composition of controlled pore glass is also different from that of amorphous silica; the composition of the finished product is typically 96% by weight of Si02, 3% B203, less than 1% Na20, and a trace amount of other metal oxides (Electro-nucleonics Inc., 1987).
All of the reagents used in the preparation of the
reversed phase CPG packings were as described in Chapter II, with the exception of CPG being used as the support material instead of silica. All CPG was manufactured by Electronucleonics Inc. (Fairfield, NJ). The CPG denoted as CPG-86 was from a single lot of CPG-10-75A (Fluka Chemical Corp,; Hauppauge, NY) and had a mean pore diameter of 86 Angstroms with a pore size distribution of + 9.8%. The absolute surface area (SBET, as measured by BET analysis) was 153.1 m2/g; the particle size range was 37-74 jim and the nitrogen pore volume was 0.48 cm3/g. The CPG denoted as CPG-167 was from a single lot of PG-170-400 (Sigma Chemical Co., St. Louis, MO) with a mean pore diameter of 167 Angstroms and a pore size distribution of + 9.6%. The absolute surface area was 161 m2/g, the particle size distribution was 37-74 pm and the nitrogen pore volume was 1.0 cm3/g. Bonded Phase Preparation
Both CPG's were acid leached, dried and reacted with the appropriate silane reagents for derivatization as
described in Chapter II. The only difference in the procedure between the CPG and silica was in t'he method of agitation. CPG is more mechanically fragile than silica; therefore direct stirring via a magnetic stirring bar is inadvisable. Agitation was accomplished by a rotary evaporator; a nitrogen gas line was attached to what is normally the vacuum outlet in order to maintain a dry atmosphere. An evacuated glass Dewar-type condenser (fabricated in-house) was used to join the reaction flask to the rotary evaporator; this piece of glassware was necessary in order to prevent the escape of the reaction solvent, especially under reflux conditions.
Reactions were performed under ambient as well as
reflux conditions; ultrasonic reactions were also carried out. The reaction products were washed and vacuum dried as described in Chapter II. Evaluation of the bonding procedure was performed via in-house elemental analysis. Scanning electron micrographs of the acid-leached CPG-86 (Figures 3-1 and 3-2) show that like the silica used in previous reactions, the CPG also exhibits an irregular particle shape as well as an irregular surface. This irregularity is not surprising considering that CPG particles of the desired size range are obtained by mechanically crushing and sieving bulk Vycor glass (Haller,
19 6 5 a ) .
Figure 3-1. Scanning electron micrograph of acid-leached CPG-86;
Figure 3-2. Scanning electron micrograph of acid-leached CPG-86;
Comparison of Silica and CPG Bonding Densities via Reflux and Ultrasonic Syntheses
The controlled pore glasses of both pore sizes (denoted CPG-86 and CPG-167) were derivatized under the same types of reaction conditions as for the Davisil silica. Two aspects in particular were to be examined by these experiments; the first was to determine the reactivity of CPG compared to that of amorphous silica and the second was to determine the effect of pore size on CPG reactivity. From pore shape indices, it has been found that the pore shape of siliceous pores is not usually cylindrical; therefore the cylindrical pore model is not a good approximation for silica (Nikolov, 1986). Since CPG has a very uniform pore diameter compared to that of silica and the CPG pores are much larger than the molecular dimensions of the reactive silane (the average molecular cross-sectional area for the octadecylchlorosilane is 50.95 (Angstroms)2 per molecule and its length is 24.72 Angstroms (Cheng and McCown, 1985)) it was expected that the CPG's would exhibit higher reactivity and therefore result in alkyl bonding densities higher than those achieved with silica. In addition, it was expected that CPG-167 would be more reactive than CPG-86 due to its larger pore size. Other workers have found that octadecyl bonding density increases with increasing pore size for silica substrates (Engelhardt et al., 1982; Sander and Wise, 1984b; Sands et al., 1986; Staroverov et al., 1986).
In the first set of reactions, both CPG-86 and CPG-167 were rotated at ambient temperature (26.0 'C) for 24 hours
using dimethyloctadecylchlorosilane as the reactive silane, 2,6-lutidine as the acid-acceptor catalyst and methylene chloride as the reaction solvent. All bonding densities are calculated from duplicate elemental analyses as described in Chapter II and the mean value + the range is reported in all cases. The bonding densities for the CPG-86 and the CPG-167 were 2.56 + 0.03 Pmol/m2 and 2.07 + 0.02 imol/m2 respectively. In the second set of experiments, the same reagents and reaction time as above were used to react both CPG's, but the reactions were performed under reflux conditions at a temperature of 50.0 0C. The CPG-86 bonding density was 2.63 + 0.00 pmol/m2 and that for the CPG-167 was
2.28 + 0.00 Pmol/m2. As expected from previous work with silica (Chapter II), reflux temperatures resulted in a higher bonding density than ambient temperatures.
The next three sets of experiments were run for 24
hours with rotation under ultrasound conditions as described in Chapter II. The first set of experiments used the same reagents as described above at a temperature of 28.0 C. The CPG-86 had a resultant bonding density of 2.55 + 0.01 Pmol/m2; that for the CPG-167 was 2.04 + 0.03 pmol/m2. As in the case for the silica, bonding densities achieved under ultrasonic conditions with 2,6-lutidine as the acid-acceptor catalyst were comparable to those achieved at ambient temperatures.
The second set of ultrasound experiments was run under the same conditions as the first set and at a temperature of
28.5 'C, except that 4-dimethylaminopyridine (4-DMAP) was used as the acid-acceptor catalyst instead of 2,6-lutidine. The bonding densities for CPG-86 and CPG-167 were 3.30 +
0.02 and 3.04 + 0.08 Pmol/m2 respectively. The 4-DMAP had again proven to be a superior acid-acceptor catalyst to the 2,6-lutidine as had been the case for silica. The third set of ultrasound experiments was performed at 28.0 C using 4-DMAP as the acid-acceptor catalyst, methylene chloride as the reaction solvent and trimethylchlorosilane (TMCS) as the reactive silane. TMCS was used in order to approximate the maximum bonding density achievable under minimum steric hindrance conditions, as explained in Chapter II. The TMCS bonding density was 4.19 + 0.02 pmol/m2 for CPG-86 and 5.15 + 0.10 Pmol/m2 for CPG-167. As expected, use of a less bulky silane reagent resulted in a higher alkyl bonding density, since steric hindrances are minimized.
A comparison of the bonding densities of the 147
Angstrom pore size silica, 86 Angstrom CPG and 167 Angstrom CPG achieved under all sets of conditions is tabulated in Table 3-1. As seen from these results, neither of our expectations was realized. In all of the octadecyl silane reactions, reactivity of the amorphous silica was greater than that of either CPG. Even more puzzling, the smaller pore CPG (CPG-86) exhibited greater reactivity than the wider pore CPG (CPG-167). In the case where TMCS was the reactive silane, these trends were reversed. This seems to indicate that the reactivity trends for the silane bonding
Table 3-1. Comparison of octadecyl bonding densities for 147 Angstrom (pore size)
silica, 86 Angstrom controlled pore glass (CPG-86) and 167 Angstrom
controlled pore glass (CPG-167).
Reaction 1 C18 Bonding Density (pmol/m2)
Conditions Temperature _(C) silica CPG-86 CPG-167
Refluxed/ 50.0 2.82 2.63 2.28
Ultrasound/ 28.0 2.71 2.55 2.04
Ultrasound/ 28.5 3.35 3.30 3.04
Ambient/ 26.0 2.69 2.56 2.07
C silane2/ 28.0 3.51 4.19 5.15
1 Reaction method/acid-acceptor catalyst.
2 C1 silane used instead of C18 silane in order to estimate bonding density
achievable using a less bulky silane reagent.
reactions could be due to steric problems, since TMCS is much smaller than dimethyloctadecylchlorosilane, but the lower reactivity of the CPG-167 compared to that of CPG-86 for the octadecyl reaction contradicts this theory. The chemical composition of the CPG's may be implicated in this problem, since CPG contains 3% B203 and silica contains no more than trace amounts. Other workers have found that heating CPG to a temperature of 700 0C for 5 to 100 hours results in migration of boron atoms to the glass surface and consequent enrichment of boron on the surface of the porous glass. They further found that this surface boron enrichment resulted in higher octadecyl bonding densities than achieved on untreated CPG (Dawidowicz et al., 1983; Dawidowicz et al., 1986; Dawidowicz and Rayss, 1985; Rayss et al., 1983; Rayss and Dawidowicz, 1986; Suprynowicz et al., 1985). At present, we are unable to explain the anomalous octadecyl bonding behavior of the controlled pore
CORRELATIONS BETWEEN CHROMATOGRAPHIC RETENTION AND OCTADECYL BONDING DENSITY
Chromatographic Determination of Thermodynamic Partition
C o e f f i c i e n t s
Retention in any chromatographic process occurs when
the solute of interest is transferred from the mobile phase to the stationary phase. The process of transfer of the solute between the mobile phase and the stationary phase is characterized by the thermodynamic distribution or partition coefficient, K, which is the ratio of the concentration of the solute in the stationary and mobile phase. Chromatographic retention is a function of the distribution coefficient and the volumes of the respective phases, and is most often described by the capacity factor, k', the ratio of the number of moles of solute in the stationary phase and in the mobile phase. The capacity factor also expresses the ratio between the amount of time the solute spends in the stationary phase and in the mobile phase. It is the commonly used measure to describe retention because it accounts for differences in column dimensions and mobile phase flow rates; it is also easy to measure as k' = (Vr Vm)/Vm where Vr is the solute retention volume and Vm is the mobile phase void volume. Measurement of the capacity factor provides valuable thermodynamic information
about solute retention in a particular chromatographic system since retention is related to the thermodynamic distribution coefficient through the volume phase ratio Vs/Vm (stationary/mobile) in that k' = K(VS/Vm). Therefore for rigorous theoretical treatment of chromatographic retention, the phase ratio must be accurately known in order to determine the partition coefficient. Measurement of the Mobile Phase Volunme in RPLC
The determination of the mobile phase volume, Vm, in liquid chromatographic systems is a problem that has generated great interest as well as considerable controversy. Its value is an essential component for the calculation of the capacity factor k' as well as for the thermodynamic distribution coefficient K. Many workers have addressed this dilemma, yet there is little consensus on a generally applicable convention for measuring Vm (Berendsen et al. 1980b; Engelhardt et al., 1984; Gutnikov and Hung, 1984; Knox and Kaliszan, 1985; Le Ha et al., 1982; McCormick and Karger, 1980; Melander et al., 1983a and 1983b; Slaats et al., 1981; Wainwright et al., 1985; Wells and Clark, 1981). The problem is especially complex for RPLC, since preferential sorption of mobile phase components by the stationary phase results in the formation of a solvation layer on this surface. The thickness and composition of the solvation layer varies with the bulk composition of the mobile phase and the local concentration of organic modifier in the solvation layer may be greater than that in the bulk
mobile phase due to the hydrophobicity of the stationary phase. The composition of the solvation layer also varies with its distance from the anchored ends of the RP chains; therefore its presence results in an ill-defined boundary between the stationary and mobile phases (Berendsen et al. 1980b; Gutnikov and Hung, 1984; Le Ha et al., 1982, Knox and Kaliszan, 1985).
There are three general categories of procedures used to determine Vm: the use of unretained compounds, the linearization of the net retention time for a homologous series and static methods (Berendsen et al., 1980b). The choice of an "unretained5 compound for V. measurements in RPLC systems is a difficult one. In any case, neither its heat of sorption nor its size should differ from those of the mobile phase components. For this reason, mobile phase constituents and especially their deuterated analogs are often used (Engelhardt et al., 1984). However, this choice is not without its problems. Since these compounds are transparent in the UV region, sensitive detection of them requires a refractive index detector. The use of deuterated organic modifier as an unretained compound is only valid when there is a large amount of organic modifier present in the mobile phase since in organic-lean mobile phases the marker will be slightly retained, especially in the case of methanol/water mobile phase systems (Engelhardt et al., 1984). The same is true for 020 in organic-rich mobile phase systems; this phenomenon is attributed to D20
adsorption onto the residual silanol groups on the stationary phase surface (McCormick and Karger, 1980). Melander et al. (1983a, p. 213) suggest that . the most weakly bound solvent component is not present in the salvation layer." and that this component should be used for the determination of Vm. They concur with McCormick and Karger (1980) that D20 is a useful probe for mobile phase volume determination unless the mobile phase is water-lean. As secondary probes of Vm, fructose and urea have been suggested for all compositions of methanol/water mobile phases and for acetonitrile volume fractions from 0 to 0.75 for acetonitrile/water mobile phase (Melander et al., 1983a). Gutnikov and Hung (1984) have also proposed the use of oxalohydroxamic acid as a UV-detectable Vm probe.
The use of UV-active inorganic salts such as nitrates has also been recommended for determination of RPLC dead volumes; however in unbuffered mobile phases the dead volumes obtained increase with increasing amount of salt injected. Nitrate is also prevented from penetrating the stationary phase pores by the Donnan potential which comes about from the negatively charged silicate ions present on the stationary phase surface (Berendsen et al., 1980b; Engelhardt et al., 1984; Wells and Clark, 1981). Knox and Kaliszan (1985) suggest using a volume fraction-weighted average of the retention volumes of isotopically labelled forms of the mobile phase components. However, besides the aforementioned problems associated with the detection of
these species, this method adopts the convention that Vm is the total volume of all mobile phase components within the column bed; the solvation layer adsorbed on the stationary phase is thereby included in the mobile phase volume.
The linearization of a homologous series of compounds
in order to find the column dead volume has been widely used in gas chromatography; therefore it is not surprising that this method has also been applied for LC systems. This method assumes that there is a linear relationship between the logarithm of the net retention time tr and the carbon number of a homologous series (Berendsen et al., 1980b; Laub and Madden, 1985; Wainwright et al., 1985). The mobile phase volume can be obtained by comparing the retention times for two consecutive homologs, termed n and n+1. For a
homologous series the ratio of capacity factors for consecutive homologs is assumed to be constant and
trn+l = A(trn) (A I)t0.
By plotting tr,n+1 versus tr,n, the slope A can be obtained and to can be determined from the intercept and multiplied by the mobile phase volume flow rate to obtain Vm. The results obtained can be precise to within 1% if alkylbenzenes are the homologous series used (Berendsen et al., 1980b); however the use of homologous aromatic alcohols gives inconsistent data due to their interactions with silanol groups on the stationary phase (Laub and Madden, 1985) .
The linearization method is not without criticism. It assumes that the relationship between the logarithm of the
retention times for a homologous series is a linear function of the number of carbons in the series and therefore that the change in free energy of partitioning per methylene group is constant. This implies that as the number of methylene groups increase, the rest of the molecule has a constant effect on stationary phase interactions. Yet in some cases the relationship between the logarithm of retention and carbon number is not linear, implying that this assumption is invalid. The linearization method is also very time consuming, since the retention time measurements must be determined very precisely in order to obtain precision in the Vm value (Knox and Kaliszan, 1985).
Determination of Vm by the static method is a
gravimetric procedure. A thermostatted packed LC column is filled successively with two pure liquids with greatly different densities and weighed. From the differences in the column masses, w, and the density of each liquid, d, the total mobile phase volume can be calculated, since m=W a Wbt
(Berendsen et al., 1980b; Knox and Kaliszan, 1985; McCormick and Karger, 1980). The mobile phase volume as determined by this method is the maximum volume within the column that is accessible to a molecule comparable in size to those used in the procedure (McCormick and Karger, 1980). However, this method ignores the possibility of a salvation layer on the stationary phase and therefore can overestimate the value of a dynamic Vm by as much as 15% for a pure methanol mobile
phase (Berendsen et al., 1980b). The volume of the mobile phase as determined by this method is useful for a reference point, both in terms of whether or not a compound experiences retention in a chromatographic system and in understanding the changes in the salvation layer which occur when the bulk mobile phase composition is changed (McCormick and Karger, 1980). Knox and Kaliszan (1985) even argue that in the theoretical treatment of thermodynamic aspects of chromatography that the maximum Vm value is the pertinent one. They argue that since the thickness of the boundary between the bulk stationary phase and bulk mobile phase is on the order of one nanometer that its position cannot be sufficiently well defined to give an accurate measure of the two volumes and that calculational methods for the volume of the solvation layer are very arbitrary. Because the gravimetric procedure gives a precise and reproducible value for Vm that is unambiguous and convenient to measure, this convention was chosen to determine the mobile phase volume for the chromatographic thermodynamic distribution coefficients calculated in this work. Measurement of the Stationary Phase Volume in RPLC
Although much work has been done on measuring the
volume of the mobile phase, measurement of Vs,' the volume of the stationary phase has not been as thoroughly investigated (Berendsen et al., 1980b; Jandera et al, 1982;, McCormick and Karger, 1980; Melander et al., 1980; Sander and Field, 1980; Slaats et al., 1981). In determining a method for the
measurement of Vs, a convention for defining V. must be chosen, since the choice of the phase ratio must be compatible with the definition of K that is in agreement with the molecular mechanism of retention. Jandera et al. (1982) have defined the stationary phase volume as that fraction of the column volume that is not occupied by the mobile phase. While this choice is certainly convenient and can be readily determined, it is at best a crude measure, as similar (or even identical) values of VS would be obtained for stationary phases made from the same bulk silica but with different bonding densities of alkyl chains, or possibly even of different chain lengths. Any determination of stationary phase volume based solely on mobile phase volume measurements is doomed to failure, as such a measurement cannot be sufficiently sensitive to ascertain bonding density or small chain length differences.
Melander and Horvath (1980) have suggested defining the phase ratio as the ratio of the surface area of the adsorbent (m2) divided by the column dead volume (cm3). While this approach is an improvement in definition, it again fails to account for certain variations in the structure of the bonded phase and it implies that adsorption is the sole mechanism in RPLC retention. The major drawback to this proposed phase ratio convention, however, lies in the accurate measurement of the two parameters involved. As previously mentioned, chromatographers have been unable to embrace any one of the commonly used methods for determining
column dead volumes as being accurate and consistent enough for precise work (Engelhardt et al., 1984; Melander et al., 1982; Smith et al., 1986). Melander and Horvath (1980) state that a small relative error in the determination of the column dead volume results in a commensurate relative error in calculating both the capacity factor and the Gibbs free energy of the solute transfer.
Other problems exist as well. The surface area of the adsorbent is usually found by use of the BET analysis method. It should be noted that the surface area of the adsorbent must be determined after derivatization with the alkyl ligand, as the surface area of the derivatized silica will be significantly different from that of the underivatized support. Although use of the BET method for surface area determination is widespread, this method is inappropriate in assessing that surface area of derivatized silica packings which is chromatographically significant. The BET method measures the area of surface that is accessible to a small molecular probe such as nitrogen. Yet in an irregular surface such as porous silica, there may exist many pores which are large enough to allow nitrogen in, but which are too small to allow the passage of any larger molecules of chromatographic interest. Chromatographic support surface area data based on BET analysis is usually overestimated, and the amount of overestimation is by no means a constant, depending on the base silica structure and the derivatization method.
Melander and Horvath (1980, p. 270) state that any estimation of "stationary phase volume" on the basis of BET surface area of the support is likely to be inaccurate." Due to the errors in determining both adsorbent surface area and column dead volume, there will consequently be a large error propagated in the subsequent calculation of the phase ratio if Melander and Horvath's convention is used.
Sander and Field (1980) have estimated the phase ratio by constructing physical models of the bonded phase using manufacturer's data regarding silanol surface coverage and percent carbon loading. This approach is quite reasonable from a theoretical standpoint, as it accounts for variation in bonding density and alkyl chain length. However because it is based on models it can only be an estimate of VS; the construction of such physical models is also time consuming.
In determining the stationary phase volume, the
pertinent volume should be the volume of the alkyl chains bonded to the silica surface. Dill (1987a) has performed statistical mechanical calculations based on a lattice interphase model of RPLC stationary phases which describe chromatographic retention in reversed phase systems. These calculations have shown that in a well endcapped column, chain interactions with solutes are the most important stationary phase contribution to solute retention. Therefore the calculation of Vs should give only the actual volume of the alkyl chains bonded to the support surface. The assumption here is that all of the bonded stationary
phase volume is accessible to the solute.
A simple method for calculating the stationary phase volume has been devised in our laboratory. The only measurements necessary are the carbon load of the packing and the actual weight of packing contained in the chromatographic column. Wise and May (1983) proposed that the surface density of a bonded alkyl ligand (in micromoles of alkyl ligand per square meter of packing surface) can be calculated by
Cs : %C (106) (4-1)
1200 nc SBET
where Cs is the bonding density (pmol/m2), %C is the carbon loading of the packing as determined from elemental analysis or by a gravimetric procedure (Cheng, 1985), nc is the number of carbons in the alkyl ligand, and SBET is the surface area of the derivatized packing as determined by BET analysis. But the volume of the stationary phase, Vs, can be expressed as
V = (Cs)(SBET)(v)(Wp)(10-6) (4-2)
where v is the molar volume of the bonded alkyl group in cm3/mole and W is the weight of the bonded packing contained in the chromatographic column. Molar volume, v,
v = M/d (4-3)
where M is the weight of the bonded phase alkyl group and d is the density of the bonded alkyl group. Cheng (1985) has experimentally determined the pertinent densities of commonly used bonded silanes and reported values of 0.8607
g/cm3, 0.8625 g/cm3, and 0.8638 g/cm3 respectively for the octadecyldimethylsilyl, octyldimethylsilyl, and trimethylsilyl bonded groups. Substitution of equations 4-1 and 4-3 into equation 4-2 results in the volume of the stationary phase, Vs (in cm3), as expressed by the following formula
(10%C M) d) (4-4).
This method provides a much more accurate calculation of Vs than has been previously possible. A principal advantage of this method is that the surface area of the packing is not used in determining Vs, which eliminates the errors associated with this measurement. The stationary phase volume that is calculated in this method is the volume that is important in the chromatographic process, i.e. the actual volume of the bonded alkyl chains themselves. The precision is limited only by the carbon loading determination (+ 0.20 %C for our departmental elemental analysis) and by the measurement of the mass of packing in the column (+ 0.1 mg on any analytical balance). Calculation of the volume of the stationary phase by this method provides the means for a more accurate and uniform determination of the phase ratio.
Preparation of Bonded Phases of Varied Bonding Densities
All of the reagents used in the preparation of the
silica and CPG bonded phases are described in Chapters II and III. Silica bonded phases with octadecyl bonding
densities greater than or equal to 2.75 pmol/m2 were prepared under traditional reflux conditions as well as under ambient, subambient and superambient ultrasonic conditions, using 2,6-1utidine or 4-DMAP as the acid-acceptor catalyst as described in Chapter II. Controlled pore glass bonded phases with octadecyl bonding densities of 3.30 and 2.63 pmol/m2 were prepared from the 86 Angstrom CPG (CPG-86) under ambient ultrasonic conditions using 4-DMAP as the acid-acceptor catalyst or under reflux conditions using 2,6-1utidine as the acid-acceptor catalyst respectively as described in Chapter III. Table 4-1 lists the experimental conditions, acid-acceptor catalyst and resultant octadecyl bonding density for each of these stationary phases.
In order to prepare bonded phases with octadecyl
bonding densities less than 2.6 pmol/m2, the experimental conditions of the bonding reaction must be altered so that a less than maximal bonding density is achieved. One strategy that can be used to accomplish this is to use a less than stoichiometric amount of the reactive silane. Another approach is to partially cover some of the reactive silanols with trimethylsilane before exhaustive derivatization with the octadecyl silane reagent (Marshall et al., 1984 and 1986). By varying the amount of trimethylsilane precoverage and then reacting the precovered silicas with an excess of the octadecylsilane, lower coverage octadecyl bonded phases of varying bonding densities can be synthesized.
Table 4-1. High octadecyl bonding density reversed phase packings. Packing Base Reaction Temperature Reaction Acid-Acceptor C18 Bonding Density
Identifier Packing Conditions (oC) Time (h) Catalyst (pmol/m2)
C18-2 silica reflux 50.0 24 2,6-lutidine 2.75
LT2 silica ultrasound 8.5 101 2,6-1utidine 2.84
DMAP1 silica ultrasound 28.0 24 4-DMAP 3.06
DMAP1/31 silica ultrasound 28.0/4.0 24/97 4-DMAP 3.15
DMAP3 silica ultrasound 4.0 97 4-DMAP 3.24
US/refl silica ultrasound 50.0 24 4-DMAP 3.34
ref/DMAP silica reflux 50.0 24 4-DMAP 3.43
rederDMAPl2 silica ultrasound 28.0 24 4-DMAP 3.56
DMAP5 silica ultrasound 3.0 144 4-DMNAP 3.60
CPG2 CPG-86 reflux 50.0 24 2,6-1utidine 2.68
CPG4 CPG-86 ultrasound 28.5 24 4-DMAP 3.30
1 Packing DMAP1/3 is a 50/50 (weight/weight) mixture of packings DMAPI and DMAP3.
2 Packing rederDMAPl was obtained by reacting packing DMAPI with the octadecyl silane
under identical conditions as for DMAP1.
Based on Kinkel and Unger's (1984) estimation of a
maximum of five micromoles of reactive hydroxyl sites per square meter of silica surface, amounts of trimethylchlorosilane (TMCS) corresponding to approximately 5%, 10%, 15%, 30% and 40% coverage of these hydroxyl sites were reacted as described in Chapters II and III under ambient conditions (at a temperature of 26.5 OC) for 24 hours to partially precover the silica and controlled pore glass supports. The reaction solvent was dry methylene chloride and 2,6-lutidine was used as the acid-acceptor catalyst. It should be noted that it is unlikely that exactly 5%, 10%, 15%, 30% or 40% of the surface hydroxyl groups on the supports were reacted; the amounts of TMCS used merely represent some fraction of the amount necessary for total coverage of the surface (Marshall et al., 1984). After TMCS precoverage, the supports were washed and dried as described in Chapters II and III. The precovered supports were then reacted with a twofold excess of octadecyldimethylchlorosilane and a fourfold excess of 2,6-lutidine with methylene chloride at a temperature of 26.5 'C for 24 hours and washed and dried as described in Chapters II and III. The resultant C1 and C18 bonding densities for these precovered bonded phases are summarized in Tables 4-2 and 4-3. HPLC Column Packing Procedures
HPLC columns were assembled from 15 cm lengths of 1/4" outer diameter and 4.6 mm inner diameter seamless precision bore polished HPLC tubing (Alltech Associates, Inc.,
Table 4-2. Bonding densities for precovered silica reversed
Packing C1 Bonding density C18 Bonding density
Identifier (2mol/m ) (mol/m )
5% 0.63 2.07
10% 0.98 2.09
15% 1.50 1.98
30% 1.38 1.74
40% 1.90 1.60
Table 4-3. Bonding densities for precovered 86 Angstrom
controlled pore glass reversed phase packings. Packing C1 Bonding Pensity C18 Bonding ensity
Identifier (Pmol/m ) (umol/m )
5% CPG 0.55 3.21
10% CPG 1.08 2.83
15% CPG 2.25 1.70
30% CPG 1.27 2.72
40% CPG 1.34 2.59
Deerfield, IL) and from Swagelok 1/4" to 1/16" zero dead volume reducing union chromatographic end fittings (Crawford Fitting Company, Solon, OH) which had been fitted with 2 pm passivated 316 stainless steel frits (Alltech Associates, Inc., Deerfield, IL). The column tubing and end fittings were made of 316 stainless steel and were passivated prior to use by ultrasonication for 30 minutes in 3 M nitric acid, followed by an aqueous and a methanol rinse.
Silica columns were packed using a Shandon high
pressure HPLC column packer with a 33 ml slurry reservoir (Shandon Southern Instruments, Inc., Sewickley, PA). Approximately 1.5 grams of derivatized silica were slurried in 30 ml of chloroform and sonicated for 10 minutes. The column was then packed at a packing pressure of 6000 psi in the downward position using a sequence of 150 ml each of 50/50 (v/v) chloroform/methanol, methanol and 50/50 methanol/water. The column was then removed from the packer fittings, the packing leveled with a spatula and the remaining column end fitting installed. Controlled pore glass columns were packed in an identical manner except for the type of packer used. Since CPG is mechanically fragile and brittle (Fluka), it cannot be packed using a high pressure packer. Therefore a Beckman Model 100A HPLC pump(Beckman Instruments Inc., San Ramon, CA) running at a flow rate of 9.9 ml/min was used in conjunction with continuous mechanical vibration to pack the CPG columns. Packing in such a manner generated a packing pressure of
200-300 psi, preserving the integrity of the CPG packings and resulting in a stable packing bed. All columns were equilibrated with the desired chromatographic mobile phase by passing approximately 125 ml of the mobile phase through the column immediately prior to use. Chromatographic Measurements
The liquid chromatographic system used for
chromatographic measurements consisted of a Valco C6W injector (Valco Instrument Company, Inc., Houston, TX) with a 10 microliter sample loop, a Beckman Model 100A isocratic HPLC pump, a Beckman Model 153 fixed wavelength 254 nm UV detector and a Fisher Recordall chart recorder (Fisher Scientific, Fairlawn, NJ). Samples were loaded in the injection loop via a Hamilton 705 SNR syringe (Reno, NV). Temperature control of the chromatographic column was accomplished by passing thermostatted water through a water jacket fitted around the column. Superambient temperatures were controlled using a Lauda Model MT heater/circulator (Brinkmann Instruments Company, Westbury, NY). Subambient temperatures were brought about by a Neslab Endocal 800 water bath (Neslab Instruments, Portsmouth, NH). Mobile phases were made of HPLC grade methanol or acetonitrile (Fisher Scientific, Fairlawn, NJ) mixed with water which had been prepared as described in Chapter II; mobile phase compositions are designated by volume ratios of organic modifier to water. Mobile phases were premixed by adding the appropriate volume of organic modifier to the
appropriate volume of water; these mobile phases were then mixed well and placed in an ultrasonic bath for 15-30 minutes in order to degas them. The flow rate of the mobile phase in all cases was 1.5 ml/min. Naphthalene (Eastman Organic Chemicals, Rochester, NY) was chosen as a small nonpolar test solute; standards were made up in HPLC grade methanol for use in the retention studies. Solute retention and column holdup volumes were measured from the chart recorder tracings. The solvent disturbance peak was used to determine column mobile phase volumes for the calculation of capacity factors; since this disturbance comes about from the methanol in which the test solute is dissolved, its choice for V. falls under the category of using an unretained compound to determine Vm. This convention was chosen in order to account for variances in the individual column dead volumes resulting from differences in the stationary phase packing density within the chromatographic
c o 1 u m n s .
Measurement of V,,, and V
The gravimetric procedure previously described was used to calculate Vm. While an ambient temperature of 25.0 'C was maintained, 150 ml of methylene chloride was passed through a 15 cm LC column packed with either silica or CPG. The column was then capped and weighed on an analytical balance. The procedure was duplicated using methanol as the mobile phase and by dividing the difference in column masses by the difference in the solvent densities at 25.0 OC (1.318
and 0.7866 g/cm3 for methanol and methylene chloride respectively) the silica column dead volume was determined to be 1.805 ml while Vm for the CPG column was 1.752 ml. It was assumed that the mobile phase volumes measured by this procedure will be constant (within experimental error) for any of the silica or CPG reversed phase columns used in this work, since in each case the packings were based on the same starting material.
The stationary phase volume for each LC column was
calculated using Equation 4-4. For the precovered bonded phases the total volume of both the trimethyl- and octadecylsilyl alkyl groups was used for Vs. Percent carbon for each column packing was obtained from in-house elemental analysis and the densities used for the trimethylsilyl and octadecylsilyl groups were 0.8638 and 0.8607 g/cin3 respectively as reported by Cheng (1985). The weight of the packing contained in the chromatographic column was determined by weighing an empty chromatographic column, packing it as described previously in this chapter and drying it at 100 0C to constant weight in a gas chromatograph with a constant helium flow through the column. From the mass differences in the two weighings the weight of the column packing was determined to be 1.1705 g for the silica columns and 1.2898 for the CPG columns. The capacity factors for the naphthalene solute were determined by triplicate injections onto each of the chromatographic columns of different alkyl bonding density using mobile
phases consisting of 55/45 methanol/water and 85/15 acetonitrile/water. Thermodynamic partition coefficients for the transfer of the naphthalene solute from the stationary phase to the mobile phase were calculated by dividing the capacity factor by the volume phase ratio.
Re s ult s
Silica-based Stationary Phases
Thermodynamic partition coefficients for the
naphthalene solute as a function of silica stationary phase octadecyl bonding density are listed in Tables 4-4, 4-5 and 4-6 respectively for the 55/45 methanol/water mobile phase system at 20.0 CC and 35.0 0C and for the 85/15 acetonitrile/water mobile phase system at 35.0 'C. Graphical representations of these data are shown in Figures 4-1, 4-2 and 4-3.
In all cases the partition coefficient increased as a linear function of bonding density until a bonding density of 3.1 pmol/m2 was reached. The best fit line for this linear region of each plot was calculated using least squares linear regression. For 55/45 methanol/water at 20.0 OC the slope and y-intercept for the best fit line were 30.1 and 12.3 respectively with a coefficient of correlation of 0.989. At 35.0 'C for 55/45 methanol/water the slope was 21.0, the y-intercept 12.7 and the coefficient of correlation 0.973. For the 85/15 acetonitrile/water mobile phase at 35.0 'C the slope was 1.20, the y-intercept 1.54 and the coefficient of correlation 0.982. In all cases the
Table 4-4. Naphthalene thermodynamic partition
coefficients at 20.0 0C as a function of
silica octadecyl bonding density for
55/45 methanol/water mobile phase.
C18 Bonding ensity Naphthalene Thermodynamic Partition
(hirol/m )Coefficient at 20.0 0C
1 .60 57 .6
1 .74 66.7
1 .98 69 .6
2 .07 78.6
2 .09 75.8
2 .75 94.3
2 .84 96.2
3 .15 97 .2
3 .24 93.4
3 .34 89 .6
3.43 87 .2
3 .56 86.1
3 .60 85.9
Table 4-5. Naphthalene thermodynamic partition
coefficients at 35.0 'C as a function of
silica octadecyl bonding density for
55/45 methanol/water mobile phase.
C18 Bonding ensity Naphthalene Thermodynamic Partition
jjMo jm Ceficient at 35.0 0C
1 .60 43.4
1 .74 50.7
1 .98 51 .3
2 .07 60.5
2 .16 59 .5
2 .75 69 .8
2 .84 71 .3
3 .15 70 .9
3.43 65 .0
Table 4-6. Naphthalene thermodynamic partition
coefficients at 35.0 'C as a function of
silica octadecyl bonding density for
85/15 acetonitrile/water mobile phase.
C18 Bonding ensity Naphthalene Thermodynamic Partition
(4mol/m ) Coefficient at 35.0 0C
1 .74 3.69
147 A Silica at 20 C 120
I I I I I I I I0.5 1D 1.5 2.0 2.5 3.0 3.5 4.0
Cme Bonding Density (xmol/ n2)
Figure 4-1. Naphthalene thermodynamic partition coefficient at 20.0 oC
as a function of silica octadecyl bonding density for
55/45 methanol/water mobile phase.
147 A Silica at 35 C
I I I I I I I I
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Ci Bonding Density (jpmrnol/ rn2)
Figure 4-2. Naphthalene thermodynamic partition coefficient at 35.0 oC
as a function of silica octadecyl bonding density for
55/45 methanol/water mobile phase.