Tài liệu HPLC for Pharmaceutical Scientists 2007 (Part 5) pdf

S denotes sample molecule, E denotes molecule ofstrong polar solvent, and X and Y are polar functional groups of the stationary phase.Prior to retention, the surface of stationary phase

5.2 THEORY OF RETENTION IN

NORMAL-PHASE CHROMATOGRAPHY

Unlike the more popular reversed-phase chromatographic mode, phase chromatography employs polar stationary phases, and retention is mod-ulated mainly with nonpolar eluents The stationary phase is either (a) aninorganic adsorbent like silica or alumina or (b) a polar bonded phase con-taining cyano, diol, or amino functional groups on a silica support The mobilephase is usually a nonaqueous mixture of organic solvents As the polarity ofthe mobile phase decreases, retention in normal-phase chromatography

normal-241

HPLC for Pharmaceutical Scientists, Edited by Yuri Kazakevich and Rosario LoBrutto Copyright © 2007 by John Wiley & Sons, Inc.

increases Figure 5-1 illustrates the mechanism of retention in NPC [1] tion is governed by the extent to which the analyte molecules displace theadsorbed solvent molecules on the surface of the stationary phase This reten-tion model based on adsorption was first proposed by Snyder [2–5] to describeretention on silica and alumina adsorbents and later extended to explain reten-tion on polar bonded phases, such as diol-, cyano-, and amino-bonded silica.Snyder assumed a homogeneous surface so that adsorption energies for soluteand solvent molecules are constant The stoichiometry of solute–solvent com-petition can be given by

Reten-(5-1)

m and a refer to solute (S) and solvent (E) molecules in the mobile and adsorbed phases, respectively n is the coefficient that takes into account dif-

ferent adsorption cross sections for solute and solvents; that is, adsorption of

a solute molecule displaces n solvent molecules in the adsorbed monolayer.

For a binary mobile-phase system consisting of a weak nonpolar solvent and

a strong polar solvent, adsorption of the weak solvent can be ignored fore, solute retention can be expressed by

There-(5-2)

lnk lnk A

S E

S m+nE a↔S a+nE m

Figure 5-1 Hypothetical representation of the adsorption mechanism of retention in

normal-phase chromatography S denotes sample molecule, E denotes molecule ofstrong polar solvent, and X and Y are polar functional groups of the stationary phase.Prior to retention, the surface of stationary phase is covered with a monolayer ofsolvent molecules E Retention in normal-phase chromatography is driven by theadsorption of S molecules upon the displacement of E molecules The solvent mole-cules that cover the surface of the adsorbent may or may not interact with the adsorp-tion sites, depending on the properties of the solvent (Reprinted from reference 1, withpermission.)

Here, A S is the solute cross-sectional area, A E is the molecular area of the

strong solvent, N E is the mole fraction of the strong solvent in the mobile

phase, k2is retention factor of the solute in the binary mobile-phase mixture,

and k1 is the retention factor in the strong solvent alone

Yet another adsorption-based retention model similar to that of Snyder wasproposed by Soczewinski [6] to describe the retention in NPC It assumes thatretention in NPC is the product of competitive adsorption between solute andsolvent molecules for active sites on the stationary phase surface The sta-tionary-phase surface consists of a layer of solute and/or solvent molecules,but, unlike the former, the latter model assumes an energetically heteroge-neous surface where adsorption occurs entirely at the high-energy active sites,leading to discrete, one-to-one complexes of the form

(5-3)

A* is an active surface site and q refers to the number of substituents on

a solute molecule that are capable of simultaneously interacting with the active site This equation takes into account the possibility of an analyte mol-ecule’s interaction with multiple sites Based on this model, the solute reten-tion factor can be expressed by the following equation, which is similar toSnyder’s:

(5-4)

where d is a constant Comparison of the two models reveals that both predict

a linear log k2 versus log N E plot Snyder’s model predicts that the slope ofthis line should be the ratio of the molecular areas of solute and solvent,whereas Soczewinski’s model predicts that the slope is the number of stronglyadsorbing substituent groups (number of adsorption sites on the analyte) onthe solute

In practice, it was found that equations (5-1) and (5-2) are most reliable forless polar solvents and solute molecules on alumina or silica stationary phasesonly Neither of the models is entirely satisfactory in the forms presented, par-ticularly for predicting retention behavior on bonded stationary phases Thesephases contain strongly adsorbing active sites as assumed in Soczewinski’smodel, but the solute molecular area and not just polar substituents are known

to play an important role in competitive adsorption as assumed by Snyder.Furthermore, secondary solvent effects resulting from solute–solvent interac-tions in both the mobile and adsorbed phases are not taken into considera-tion in either model These effects, such as hydrogen bonding, give rise to some

of the most useful changes in retention and often are an important source ofchromatographic selectivity [7, 8]

Another experimental deviation from equations (5-1) or (5-2) was mined to be due to the localization of solvent molecules onto the adsorptionsites of stationary phase resulting from silanophilic interactions When the

deter-logk2= −d qlogN E

S m+qE A a- *↔S A- *+qE m

THEORY OF RETENTION IN NORMAL-PHASE CHROMATOGRAPHY 243

polar substitution groups of a solvent molecule interact strongly with the polargroups on the surface of the column packing, they become attached or local-ized onto the stationary-phase surface An important consequence of solventlocalization is the apparent change in the solvent strength value of a polarsolvent (Solvent strength is presented by e0, which is determined empirically

by using polyaromatic hydrocarbons that do not localize but lie flat on asurface Solvent with larger value of e has stronger elution power [1].) Con-sequently, the solvent strength does not vary linearly with the concentration

of the stronger solvent for a binary mixture where one solvent is stronger thanthe other [7] There is competition between the two solvents for the active sites

of the adsorbent and the stronger solvent will preferentially adsorb, resulting

in a more concentrated adsorbed layer of the stronger solvent For instance,the dependence of solvent strength for several binary mixtures on alumina asadsorbent shows a large increase in solvent strength due to a small increase

in the concentration of a polar solvent at low concentrations But at the otherextreme, a relatively large change in the concentration of the polar solventaffects the solvent strength of the mobile phase to a lesser extent In the case

of low concentration of polar solvent before the localization on the surface ofstationary phase reaches saturation, a small change of the polar solvent con-centration can greatly affect the number of polar active sites on the columnpacking As a consequence, significant variations of analytes retention areobserved Once the polar active sites of the stationary phase are localized com-pletely, change of polar solvent concentration will have a smaller impact onanalyte retention

These deficiencies were addressed by revising Snyder’s model as follows[8] To account for the preferential adsorption of solute and solvent onto the

strong sites, empirical A S and N Evalues larger than those calculated from ecular dimensions are used based on experimental observation The revisedmodel acknowledges the tendency of polar molecules to localize on thestrongly adsorbing active site and expresses solute retention in terms of thesolvent strength as follows:

mol-(5-5)

where a′ is an adsorbent activity factor, e0 and e0 are solvent strengths for

solvent 1 and 2, and A S is the analyte cross-sectional area on the adsorbentsurface The “analyte” cross-sectional area can be predicted from moleculardimensions Secondary solute–solvent interactions are incorporated into therevised model by adding extra terms denoted by ∆ for each of the solvents asfollows:

(5-6)When a nonlocalizing, nonpolar solvent such as hexane is employed as a weaksolvent, the equation can be further simplified so that

logk2=logk1− ′a A S(ε1−ε2)+(∆2−∆1)

logk2=logk1− ′a A S(ε1−ε2)

(5-7)assuming hexane does not induce any secondary solvent effects and its solvent

strength is zero Here k his the analyte retention factor in pure hexane tion (5-7) has been found useful to understand the fundamental principles governing the retention behavior as far as solute, solvent, and bonded-phaseproperties are concerned For instance, by fitting equation (5-7) to the exper-imental NPC data, the extent of solute localization can be determined by com-

Equa-paring the slopes of a log k2versuse0 plot, provided that the molecular crosssection can be estimated accurately

5.3 EFFECT OF MOBILE PHASE ON RETENTION

Selection of suitable mobile-phase system is critical in NPC to achieve thedesired separation [4] In general, a suitable solvent should have the follow-ing properties: low viscosity, compatibility with detection system (for instance,solvent should be transparent at wavelength of detection if UV is used asdetector), available in pure state, low flammability and toxicity, highly inert,and adequate solubility for solutes Unlike RPLC, analytes become lessretained as solvent strength (solvent polarity) increases Solvent strength inNPC can be represented by e0, and values of e0for some commonly used NPCsolvents are listed in Table 5-1 for silica as column packing [1] Relative solventstrength for other NPC column packings such as alumina and polar bondedphases follow the same trend as in the table; that is, larger values of e0 areobtained for more polar solvents Ideally, the mobile-phase strength should bechosen to maintain analyte retention factor within the optimum range of 1 ≤

k′ ≤ 5 with selectivity values sufficient to reach a satisfactory resolution

In general, binary mobile phases, such as a mixture of a nonselective solventhexane with a polar solvent, are used for NPC separations If separation cannot

be achieved by adjusting mobile phase strength (change the concentration ofone of the components in a binary mixture), then variation of polar solventnature has to be pursued Snyder has developed a useful scheme to classifysolvents (nonelectrolytic solvents) nature based on their interactions withsolutes and the stationary phase [9] This approach should not be taken as con-crete rules but rather as a phenomological approach The property of a solvent

is characterized by the three most important parameters, which are its acceptor (Xe), proton-donor (Xd), and dipole-donor (Xn) affinity Each ofthese contributes to the overall polarity of the solvent, which in turn is related

proton-to its chromaproton-tographic strength Rohrschneider determined the values of theseparameters from distribution coefficients of test solutes such as ethanol,dioxane, and nitrobenzene [10] A medium polar solvent—such as chloroform,which has a polarity of 4.31—involves 31% proton acceptor, 35% protondonor, and 34% dipole interactions If the parameter values of the solvents areplotted on a triple coordinate system, various solvents can be grouped into

logk2=logk h− ′a A Sε2+∆2

EFFECT OF MOBILE PHASE ON RETENTION 245

eight classes (Figure 5-2) [9] Solvents within each class should show similarselectivity for a set of components, while the nature of solvents from differentclasses are quite different and may impart differences in selectivity for thesame set of components In NPC method development, replacing solventsbelonging to the same selectivity class cannot offer substantial variation inchromatographic separation Therefore, it is recommended to select solventsthat are placed close to the apices of the triangle for maximum selectivity.Common solvents in group I are isopropyl ether and MTBE, group VII sol-vents include dichloromethane and 1,2-dichloroethane, and chloroform andfluoro-alcohols constitute group VIII solvents Solvent mixtures having thesame elution strength but different selectivities are called isoelutropic mobilephases.

Binary mixtures, however, have only limited abilities for controlling phase selectivity Therefore, ternary and even quaternary mobile phases thatcontain two or more different polar solvents along with a nonpolar solvent areoften used to achieve the required selectivity If the ratio of the concentration

mobile-of two polar solvents is constant but the sum mobile-of the their concentration is beingchanged with respect to that of the nonpolar solvent, the effect on retention

is much the same as when the concentration of the single strong solvent

TABLE 5-1 NPC Solvent Strength (e 0 ) and Selectivityaof Various Solvents

Employed in HPLC

aSilica used as absorbent.

bMinimum UV wavelength; assumes that maximum baseline absorbance (100% B) is 0.5 AU.

cSolvent basicity is irrelevant for nonlocalizing solvents.

d Methyl t-butyl ether.

eDifferent selectivity due to presence of proton donor group.

Source: Reprinted from Ref 1, with permission.

changed in a binary mobile phases On the other hand, if the sum of the twopolar solvents stays constant but the ratio is variable, larger effects on theselectivity of separation are observed than in the system where the ratio isconstant This is attributable to changes in dipole–dipole and proton–donor–acceptor interactions between polar solvents and the analytes Such selectiv-ity tuning is the main purpose of using ternary mobile phases in NPC A phenomenological approach for the appropriate selection of ternary mobilemixture based on Snyder’s solvent selectivity triangle concept combined with

a statistical approach can be applied [11–15] As can be seen in Figure 5-4, aseven-run design is used A primary binary solvent mixture such as hexane-MTBE with the solvent strength that is convenient for the separation is firstselected This binary mixture represents one corner of the selectivity triangle.Two other binary mixtures, namely, hexane-dichloromethane and hexane-chloroform, having the composition with the same solvent strength, are thentested As shown in Figure 5-3, the area bound by the sides of the triangleformed by MTBE, dichloromethane, and chloroform defines the selectivitydomain in which the optimum mobile-phase composition will be found Next,separations are performed with three different ternary mobile-phase systemsproduced by mixing an equal volume of each of the binary solvents Thus, thethree experiments are set in the middle of triangle Finally, the analysis iscarried out by mixing in the three binary mixtures in equal ratio By compar-ing the seven chromatograms obtained in the above experiment, optimum

EFFECT OF MOBILE PHASE ON RETENTION 247

Figure 5-2 Snyder’s selectivity triangle for solvents (Reprinted from reference 9, with

permission.)

solvent composition for the separation can be easily identified Figure 5-4demonstrates the triangle reduction method whereby the same procedure isrepeated, starting from a smaller triangle—for instance, as defined by apices

2, 4, and 5, which corresponds to an area where the resolution is the highest—until an optimum mobile-phase mixture is determined for adequate resolution

of the separated mixtures Furthermore, optimum solvent composition can also

be obtained by regression analysis with data obtained from the seven runsexperiment [14]

Separation of acidic or basic analytes on NPC generally results in cant peak tailing due to the strong hydrogen-bonding interactions with silanolgroup on the stationary phase Therefore, acidic or basic additive such as TFA(trifluoroacetic acid) or DEA (diethylamine) are often included in the mobile-phase system to minimize the hydrogen-bonding interactions

signifi-5.4 SELECTIVITY

5.4.1 Effect of Analyte Structure

In NPC, analytes retentions generally increase in the following sequence:alkane< alkenes < aromatic hydrocarbons ≈ chloroalkanes < sulfides < ethers

< ketones ≈ aldehyde ≈ esters < alcohols < amides << phenols, amines, and carboxylic acids [16] The retention also depends to some extent on the

Figure 5-3 Selected solvents for mobile-phase optimization in NPC (Reprinted from

reference 11, with permission.)

hydrocarbon part of the solutes Unlike RPLC, however, analytes become lessretained as the size of alkyl chains increases Furthermore, the separation inhomologous series is less satisfactory than in RPLC According to Soczewin-ski’ model, analyte can have multiple interaction sites simultaneously whenthe adsorption sites interacts with a specific steric position of functional groups

in the solute molecules with multiple functional groups On the other hand,molecules with other positions of functional groups may have weaker orabsent multiple sites interaction with the stationary phase (e.g., ortho versusmeta versus para positions on an aromatic solute) This feature makes the use

of NPC very suitable for the separation of positional isomers In addition, ference in the retention and selectivities of molecules of similar polarities, butdifferent shapes, such as rigid planar, rod-like, or of a flexible chain structure,are often observed in NPC

dif-5.4.2 Types of Stationary Phases

In order to accomplish the desired separation, the selection of appropriate tionary phase and eluent system is imperative The most commonly used sta-tionary phases in normal-phase chromatography are either (a) inorganicadsorbents such as silica and alumina or (b) moderately polar chemicallybonded phases having functional groups such as aminopropyl, cyanopropyl,nitrophenyl, and diol that are chemically bonded on the silica gel support [16].Other phases that are designed for particular types of analytes have also

Figure 5-4 Procedure for selectivity optimization in NPC based on mixtures with

hexane of nonlocalizing solvent (CH2Cl2), a basic-localizing solvent (MTBE), and anonbasic localizing solvent (ACN or ethyl acetate) All mobile phases are of equalstrength (Reprinted from reference 1, with permission.)

proved to be successful These include modified alumina [17], titania [18], andzirconia [19–21].

Since the stationary phase in normal phase chromatography is more polarthan the mobile phase, analyte retention is enhanced as the relative polarity

of the stationary phase increases and the polarity of the mobile phasedecreases Retention also increases with increasing polarity and number ofadsorption sites in the column This means that retention is stronger on adsor-bents with larger specific surface areas (surface area divided by the mass ofadsorbents) Generally, the strength of interaction with analytes increases inthe following order: cyanopropyl < diol < aminopropyl << silica ≈ alumina sta-tionary phases However, strong selective interactions may change this order.The use of silica columns is less convenient for analytical applications.However, isomer and preparative separation favors the use of unmodifiedsilica Basic analytes are generally very strongly retained by the silanol groups

in silica gel, and acidic compounds show increased affinities to aminopropylsilica columns Aminopropyl and diol-bonded stationary phases prefer com-pounds with proton–acceptor or proton–donor functional groups as in alco-hols, esters, ethers, and ketones, whereas dipolar compounds are usually morestrongly retained on cyanopropyl silica than on aminopropyl or diol silica.Alumina phase has unique application in the separation of compounds withdifferent numbers or spacing of unsaturated bonds This is because aluminafavors interaction with π electrons and often yields better selectivity than silica[16]

Despite the many desirable properties of silica, its limited pH stability(between 2 and 7.5) is also a major issue in NPC when strong acidic or basicmobile-phase additives are used to minimize interactions Hence, other inor-ganic materials such as alumina, titania, and zirconia, which not only have thedesired physical properties of silica but also are stable over a wide pH range,have been studied Recently, Unger and co-workers [22] have chosen a com-pletely new approach where they use mesoporous particles based not only onsilica but also on titania, alumina, zirconia, and alumosilicates These materi-als have been used by the authors to analyze and separate different classes ofaromatic amines, phenols, and PAHs (polyaromatic hydrocarbons)

Bonded stationary phases for NPC are becoming increasingly popular inrecent years owing to their virtues of faster column equilibration and beingless prone to contamination by water The use of iso-hydric (same water con-centration) solvents is not needed to obtain reproducible results However,predicting solute retention on bonded stationary phases is more difficult thanwhen silica is used This is largely because of the complexity of associationspossible between solvent molecules and the chemically and physically het-erogeneous bonded phase surface Several models of retention on bondedphases have been advocated, but their validity, particularly when mixedsolvent systems are used as mobile phase, can be questioned The most com-monly accepted retention mechanism is Snyder’s model, which assumes thecompetitive adsorption between solutes and solvent molecules on active sites