Temperature Pulsing for Controlling Chromatographic Resolution inCapillary Liquid ChromatographyTim J. Causon,† Hernan J. Cortes,†,‡ Robert A. Shellie,† and Emily F. Hilder*,†

†Australian Centre for Research on Separation Science, School of Chemistry, University of Tasmania, Private Bag 75, Hobart,Tasmania, Australia, 7001‡HJ Cortes Consulting LLC, Midland, Michigan 48642, United States

*S Supporting Information

ABSTRACT: In this study we introduce the implementationof rapid temperature pulses for selectivity tuning in capillaryliquid chromatography. Short temperature pulses improvedresolution in discrete sections of chromatograms, demon-strated for ion-exchange chromatography (IC) and hydrophilicinteraction chromatography (HILIC) modes. Using aresistively heated column module capable of accurate andrapid temperature changes, this concept is first illustrated withseparations of small anions by IC using a packed capillary column as well as a series of nucleobases and nucleosides by HILICusing a silica monolithic column with zwitterionic functionality (ZIC-HILIC). Both positive (increasing temperature) andnegative temperature pulses are demonstrated to produce significant changes in selectivity and are useful approaches forimproving resolution between coeluted compounds. The approach was shown to be reproducible over a large number ofreplicates. Finally, the use of temperature gradients as well as other complex temperature profiles was also examined for both ICand HILIC separations.

The role of temperature in controlling selectivity for liquidchromatography (LC) has been historically viewed as less

useful than variation of the eluent strength (e.g., the amount of

large, singly charged polarizable ions with strongly exothermicretention behavior (e.g., iodide, thiocyanate, perchlorate).These large differences in retention behavior allow the elutionorder to be adjusted by isothermal variation of temperature,which is in good agreement with results from studies ofionizable solutes in other modes of LC.4,20

The broad range of retention processes that can take placeusing hydrophilic interaction chromatography (HILIC) sta-tionary phases are also known to be strongly influenced by thetemperature.20,21 Stationary phases used for HILIC aregenerally unmodified silica or silica derivatized with groupssuch as amine, amide, cyano, diol, or zwitterionic (ZIC)functionality. With a large percentage of organic modifier in themobile phase it is thought that a layer of water-rich mobilephase is partially immobilized on the stationary phase surface.Solutes can then partition between this layer and the bulkmobile phase allowing increased retention of polar and ionicsolutes compared to reversed-phase LC.22 However, the detailsof retention mechanisms involved in this mode are open tosome debate as polar−polar interactions (i.e., dipole−dipole,hydrogen bonding, and charge−dipole) are also noted to have astrong influence on separation behavior.23 Generally speaking,when increased temperature is applied, the retention of solutesis expected to decrease as the polarity of the organic solvent isreduced (i.e., exothermic retention behavior) as is frequentlythe case for reversed-phase LC. However, some studies20,21,24

have shown that both endothermic and exothermic retentionbehavior are observed in this mode and that temperature can bea useful parameter for altering selectivity by exploiting the polarand ionic characteristics of solutes, mobile phase components,and the column itself. Furthermore, as retention in HILIC isparticularly sensitive to the amount of organic modifier across anarrow range of mobile phase compositions (typically 75−97%v/v acetonitrile), temperature changes might provide a moreconvenient approach to adjusting chromatographic selectivity.As an alternative to previous approaches, the use of rapid

temperature changes (e.g., pulsing) during the separationprocess is attractive for any mode of LC as the resolution maybe manipulated independently for discrete sections of thechromatogram. This offers significant advantages whenresolution may be insufficient between particular componentsbut changes in selectivity throughout the remainder of thechromatogram would be undesirable. A rapid temperature“pulse” yields a transient change in apparent elution strengthwhich cannot be achieved practically with a conventionalsolvent gradient. Furthermore, in comparison to conventionalisothermal high-temperature LC, applying brief pulses of hightemperatures reduces the risk of damaging the stationary phasematerial.In this study, we introduce the use of rapid temperature

pulses to alter selectivity and improve resolution for the LCanalyses of some small molecules with illustrative examplesusing both IC and HILIC columns. These modes of LC werechosen because the separation mechanism means that it wouldbe expected that temperature is a useful tool for manipulatingselectivity and also because they are both modes of LC forwhich significant changes in separation selectivity can bedifficult to achieve, in particular if individual solutes are to betargeted. Thus temperature pulsing should offer significantbenefits. We also investigate the use of both simple andcomplex temperature gradients as a complement to conven-tional alteration of solvent strength for both systems.

■ EXPERIMENTAL SECTION

Reagents. All chemicals used were of analytical reagentgrade and were used as supplied by Sigma-Aldrich (Sydney,Australia) unless stated otherwise. Eluents were prepared usingdeionized 18.2 MΩ water from a Millipore Milli-Q waterpurification system (Bedford, MA, U.S.A.), sodium chloride(Merck Chemicals, Victoria, Australia), ammonium formate,HPLC grade acetonitrile, and formic acid. Probe solutes wereprepared in deionized water (IC) or acetonitrile (HILIC) tomake 1000 mg/L stock solutions. IC solutes (benzoate, iodate,iodide, fumarate, molybdate, nitrate, nitrite, and thiocyanate)were prepared from their respective sodium salts. Solutes usedfor HILIC separations were adenosine, cytosine, cytidine,thymidine, uridine, 2′-deoxyadenosine, 2′-deoxycytosine, 2′-deoxyguanosine, 2-hydroxybenzoic acid (salicylic acid), 2,3-dihydroxybenzoic acid (2-pyrocatechuric acid), 2,4-dihyrox-ybenzoic acid (β-resorcylic acid), 2,5-dihydroxybenzoic acid(gentisic acid), and 3,5-dihydroxybenzoic acid (α-resorcylicacid). Solutes for HILIC were further diluted in mobile phaseprior to analysis.

Chromatography. Analyses were performed on a DionexUltimate 3000 HPLC system (Thermo Scientific, Lane Cove,Australia), equipped with a ternary low-pressure-mixinggradient pump including a membrane degasser unit, atemperature-controlled column oven with a flow manager,and an autosampler. IC separations were carried out usingcolumns containing Dionex IonPac AS11 polymeric anion-exchange particles packed into 41 cm × 100 μm i.d. (360 μmo.d.) fused-silica capillary (Polymicro, AZ, U.S.A.). The HILICseparations were carried out using a silica monolith columnfunctionalized with zwitterionic groups (150 mm × 0.1 mmi.d.) obtained from Merck SeQuant (Umea,̊ Sweden). Theeluent used for IC analysis was an aqueous solution of sodiumchloride; for the HILIC analyses a mixture of acetonitrile in anaqueous 25 mM ammonium formate buffer titrated to pH 3.8using formic acid was used. All eluents were filtered with 0.20μm nylon membranes (Grace Davison Discovery Sciences,Rowville, Australia) prior to analysis. Photometric detectionwas performed at 214, 220, and 254 nm using a VWD-3000series UV detector equipped with a 45 nL flow cell. A 1.0 μLsample loop was used in partial loop mode for all analyses.Chromatographic data were collected at 2.5 Hz, and chromato-grams were processed using the Chromeleon software.A low-thermal-mass (LTM) module (Agilent Technologies,

Mulgrave, Australia) was used for temperature control of thecolumn. It was composed of a nickel resistive wire, resistivetemperature detector, an insulating fiber, and an LTM A68controller. This assembly was also equipped with a smalldesktop computer fan operated by the LTM controller toexpedite cooling, which was necessary for rapid temperaturechanges.25 Capillary columns were bundled tightly with theresistive heating wire, temperature sensor, and insulating fiber.The complete bundle was wrapped with aluminum foil, andcolumns were joined to the HPLC system via polyether etherketone (PEEK) unions and fittings.For experiments with the shorter HILIC column, a glass

press fit (Restek, Victoria, Australia) was used to connect thecolumn to a 10 cm length of 25 μm i.d. fused-silica capillary toincrease the total length of column wrapped in the heatingmodule.

Column Packing Procedure. IC columns were packed byfollowing a slurry packing technique described previously.26

Analytical Chemistry Article

dx.doi.org/10.1021/ac300161b | Anal. Chem. 2012, 84, 3362−33683363

Briefly, a 50 cm length of 100 μm i.d. fused-silica capillary wasconnected using a piece of 0.4 mm i.d. PEEK tubing to anHPLC in-line filter containing a metal frit. The other end of thecapillary was connected to a stainless steel slurry reservoir (100mm × 2 mm i.d.). A frit was prepared by packing a 50 mg/mLslurry of Develosil 5 μm porous silica particles (NomuraChemical Co., Japan) in acetone at 6000 psi using a Haskel40102 air-driven fluid pump (Haskel, Brisbane, Australia) withwater as the driving liquid. A custom heating element was usedto sinter the silica particles while flushing the column withwater at 4500 psi. Excess silica particles were then flushed fromthe column with water.After preparing the frit, a 150 mg/mL slurry of the ion-

exchange particles was prepared in high-purity water andpacked into the column at a pressure of 3000 psi. After packing,the column was allowed to decompress for 2 h. The columnwas then flushed with water for 30 min, and sections notcontaining stationary phase particles were removed prior toHPLC experiments. The final column length was 41 cm.

■ RESULTS AND DISCUSSIONInfluence of Isothermal Separation Temperature. To

examine the fundamental influence of temperature on LCseparations using fused-silica capillary columns, initial studiesexamined retention with the resistively heated column moduleunder isothermal and isocratic conditions. The influence oftemperature on retention was first studied with a series of UV-absorbing anionic solutes by IC (Figure 1a). Under theseconditions linear retention behavior was observed in goodagreement with a previous study using the same stationaryphase material.16 The large, singly charged and polarizablesolutes (iodide and thiocyanate) exhibited strongly exothermicretention behavior in contrast to the endothermic behavior ofmost other solutes (both singly and multiply charged).Improvements in efficiency at elevated flow rates were alsoobtained for all probe solutes, and the best theoretical platenumber (N) values recorded for iodide correlated well withexpected values for 13 μm particles assuming an optimum plateheight of twice the particle diameter (26 μm corresponds to N= 15 800 for a 41 cm column, see Table S1 in the SupportingInformation).Further isothermal studies of isocratic separations of a series

of nucleobases and nucleosides were performed with the HILICcolumn across a similar temperature range (Figure 1b). Using aconventional column oven, the retention behavior of probesolutes was found to be linear across the temperature rangestudied for this example with most solutes exhibitingexothermic retention behavior, except for 2′-deoxyguanosine.Retention behavior observed when the column was placedoutside the oven and wrapped in the resistively heated columnmodule was also found to be linear, but not identical to thatobtained in the instrument column oven at above ambienttemperatures. This is likely to have occurred as the ends of thecolumn could not be wrapped completely in the resistivelyheated bundle leaving “cold spots” at both ends of the column.This difference was found to be more significant when usingthe shorter HILIC column.Rapid Temperature Pulsing. Having established some

isocratic conditions for separations of probe solutes using ICand HILIC columns, the possibility of rapid temperaturevariations during the separation process was examined. As theresistive heating system used in this study was described asbeing capable of both rapid heating (up to ∼1800 °C/min) and

cooling (∼200 °C/min),27 we first considered the use of a rapidtemperature “pulse” during the analysis to influence selectivitybetween poorly resolved solutes. The term “pulse” in used inthis study to refer to a programmed, rapid (≤1 min) deviationfrom the isothermal operating temperature during theseparation process. The IC analysis performed at 23 °C yieldeda coelution of fumarate and iodide anions (Figure 1a) which

Figure 1. (a) Isothermal separations of (i) nitrate, (ii) benzoate, (iii)iodide, (iv) fumarate, (v) thiocyanate, and (vi) molybdate performedin the IC mode using the AS11 column. Eluent was 64 mM NaCl, andflow rate was 1.00 μL/min. The detection wavelength was 214 nm. (b)Isothermal separations of (1) thymidine, (2) adenine, (3) adenosine,(4) uridine, (5) cytosine, (6) 2′-deoxycytidine, (7) cytidine, and (8) 2′-deoxyguanosine performed in the HILIC mode using the ZIC-HILICcolumn. Eluent was 95% v/v acetonitrile and 5% v/v 25 mMammonium formate (pH 3.8). The flow rate was 0.755 μL/min. Thedetection wavelength was 254 nm.

Analytical Chemistry Article

dx.doi.org/10.1021/ac300161b | Anal. Chem. 2012, 84, 3362−33683364

represents a typical elution problem that might be solved byadjusting the solvent composition or optimization of theisothermal separation temperature. In the case of capillary LCanalysis, changing the solvent composition can be a slowprocess due the large dwell volumes and long equilibrationtimes. An isothermal temperature change would be sufficient tosolve this particular coelution problem, but a pulsed change ismore attractive to avoid loss of resolution in other sections ofthe chromatogram caused by an absolute change in temper-ature. As a first example, resolution was manipulated betweenthe coeluted peaks by incorporating a positive temperaturepulse a few minutes prior to their elution (Figure 2a). It wasfound that resolution between these peaks could be furtherimproved by increasing the height of the temperature pulse(i.e., the maximum temperature), but an increase in the UVbaseline level was observed when using a maximum temper-ature of 100 °C. To maintain baseline stability, a maximumtemperature of 90 °C (pulse height of 67 °C) was used forremaining IC experiments. This maximum was sufficient toeffect an adequate change in retention time of the iodide andfumarate ions (−4.8% and +0.81%, respectively) allowing theresolution between these peaks to be increased to 1.2. Clearlythe retention of compounds already eluted from the column isunaffected by this pulse, while increasingly smaller changes inretention times were observed for compounds eluted after these(e.g., −3.5% for thiocyanate). The magnitude of the change inretention time should be directly related to the respectivethermodynamic retention properties. In this instance, thestrongly exothermic retention behavior of iodide (ΔHR ∼ −7.7kJ/mol) compared to the weakly endothermic retentionbehavior of fumarate (ΔHR ∼ 4.7 kJ/mol) is the primaryreason that a short (a length of 0 s applied by the heatingdevice), pulsed change in temperature (ΔT = 67 °C) wassufficient to exert such a noticeable change in retention.It was also observed that the timing of the pulse had a strong

influence on the improvement in resolution observed. Applyingthe pulse increasingly early yielded progressively broader peakswhile applying the pulse too late (i.e., when the compoundswere close to or at the end of the column) yielded a sharpdecrease in resolution as the peaks were no longer separated.The potential of negative temperature (cooling) pulses was

also examined by pulsing the temperature from 58 °C down toa minimum of 23 °C in an analogous manner (Figure 2b). Itwas clear from this example that the cooling rate in thistemperature range provided a less efficient means to improveresolution as a pulse length of 60 s was required to gain suitableresolution between these peaks. However, this example ispossibly limited by the smaller change in temperature (ΔT = 37°C) utilized as well as the proximity of the operatingtemperature range to ambient conditions.To assess the general applicability of this concept by

considering another mode of LC, we also studied temperaturepulses using a HILIC column (Figure 3). In this example theisothermal (40 °C) coelution of 2′-deoxyguanosine and cytidineis improved by incorporation of a positive temperature pulse upto 120 °C during the separation. The difference in molarretention enthalpies was smaller than in the previous examplefor IC (ΔHR ∼ −3.9 kJ/mol for cytosine and ΔHR ∼ 3.1 kJ/mol for 2′-deoxyguanosine), which rendered the temperatureeffect less influential. However, this could be countered byincreasing the pulse height and width. It was also found thattemperature regulation was more difficult using a shortercolumn (15 cm in this example) as the end fittings on the

column now led to significant portions of the columnremaining as “cold spots” (i.e., at ambient temperature).

Using a longer column would also improve the efficiency,

which would further aid in enhancing resolution between these

Figure 2. (a) Influence of a positive temperature pulse on ICseparation of iodide and fumarate anions performed at 23 °C.Chromatograms shown illustrate the effect of the pulse height whichoccurred at 6.5 min with a heating/cooling rate of ±500 °C/min heldfor 0 s. Peak identities and other conditions are the same as Figure 1a.The timing and width of pulse is shown by the gray, dashed line whichwas typical for all temperature pulsing experiments. (b) Influence of anegative temperature (23 °C) pulse on separation of fumarate andthiocyanate performed at 58 °C. The pulse length was varied from 15to 60 s at 8.5 min with a heating/cooling rate of ±500 °C/min. Acomparison to holding the temperature at 23 °C (step gradient) is alsoshown. Peak identities and other conditions are the same as Figure 1a.

Analytical Chemistry Article

dx.doi.org/10.1021/ac300161b | Anal. Chem. 2012, 84, 3362−33683365

peaks and reducing the fraction of the column which is notcorrectly temperature regulated. However, we found that,despite these issues, the reproducibility of this approach wasstill very good by performing an identical temperature pulsedanalysis repeatedly and comparing the resultant chromatograms(Figure 4).From a practical perspective, over the course of this study we

did not observe significant changes in either the particle packedIC column or monolithic HILIC column despite the numerous

high-temperature pulses that were applied. This offers a distinctadvantage over typical approaches to the use of high operatingtemperatures in LC in which column (and solute) stabilityremains a significant limitation.

Use of Gradient and Complex Temperature Profiles.Although temperature can be manipulated to optimizeselectivity for LC, numerous authors have noted that thiseffect is generally smaller than the influence exerted by thesolvent strength.3,4,11,15,28 Despite this, the use of simple7 andcomplex27 temperature gradients to improve resolution hasbeen demonstrated by a number of authors primarily forreversed-phase separations. One significant advantage of using aresistively heated capillary column module is that it allowstemperature gradients of almost any profile to be utilized toimprove resolution. It is interesting to note that many solutes inthe IC mode exhibit endothermic retention behavior, soreversed temperature gradients might also find application inconditions where retention and efficiency of weakly retainedcompounds need to be increased (Figure 5a). This type ofinverse temperature gradient implementation has beenpreviously demonstrated for reversed-phase LC separations ofpoly(ethylene glycol) oligomers.29

When using the strong anion-exchange column with asolvent gradient elution of 5−95 mM sodium chloride, the ion-exchange coefficient was found to exert a much strongerinfluence on retention, particularly for polyvalent solutes (e.g.,molybdate, chromate). The use of a steep salt gradient thusreduced the usefulness of temperature pulsing for improvingresolution between coeluted peaks (see Figure S1 in theSupporting Information).The use of temperature changes using the HILIC column

was found to be particularly useful as any small variation in theorganic modifier concentration had a strong effect on theretention of probe solutes. Conversely, an isocratic analysis withtemperature pulsing or any form of temperature programmingmight be an easier approach for optimizing separations whenusing this mode for separations where appreciable differences invan’t Hoff plots are observed. Even for a sample containing aseries of mono- and dihydroxybenzoic acids, marked differencesin retention behavior were observed due to the differencesbetween pKa of the stationary phase, mobile phase, and thesolutes.20 Use of positive and negative temperature gradients inthis example resulted in improved resolution between somesolutes and decreased the total analysis time simultaneously(Figure 5b). Of particular note is that 3,5-dihydroxybenzoicacid (α-resorcylic acid) exhibits strongly exothermic behaviorand is the solute with the highest pKa value. Thus, temperatureprogramming can be useful for improving resolution even forsolutes within a homologous series when used with ionizablesolutes and suitable mobile phase pH conditions.In general, employment of linear temperature gradients for

both modes of LC studied was found to have less effect onretention than conventional solvent gradients. This isparticularly noticeable for band compression effects observedwhen using a solvent gradient elution (see Figure S1 in theSupporting Information). However, the combination ofisocratic analysis and temperature gradients can still be a usefultool for resolution optimization with some notable advantagesover solvent gradient programming for capillary LC. First, thechange in eluent strength is applied almost instantly to theentire column when using a resistive heating module, unlike insolvent programming where a substantial dwell volume delaysthis change (the gradient delay volume of the system used in

Figure 3. Examples of temperature pulsing for a HILIC separation ofnucleobases and nucleosides. The temperature pulse from 40 to 120°C occurred at 24 min with a heating/cooling rate of ±500 °C/minheld for 15 s Peak identities and other conditions are the same asFigure 1b.

Figure 4. Overlay of 10 consecutive temperature-pulsed chromato-grams. Chromatographic conditions are as described in Figure 3.Chromatograms shown are normalized according to the height of thelargest peak to account for variation in the injection volume.

Analytical Chemistry Article

dx.doi.org/10.1021/ac300161b | Anal. Chem. 2012, 84, 3362−33683366