Optimized Use of a 50 µm Internal DiameterSecondary Column in a ComprehensiveTwo-Dimensional Gas Chromatography System

Peter Quinto Tranchida,† Giorgia Purcaro,†,‡ Lanfranco Conte,‡ Paola Dugo,† Giovanni Dugo,† andLuigi Mondello*,†,§

Dipartimento Farmaco-chimico, Facolta di Farmacia, Universita di Messina, viale Annunziata, I-98168, Messina, Italy,Dipartimento di Scienze degli Alimenti, Universita degli Studi di Udine, via Sondrio 2, 33100, Udine, Italy, andCampus-Biomedico, Via Alvaro del Portillo, 21, 00128 Roma, Italy

The focus of the present research is directed toward thedevelopment of a comprehensive two-dimensional gaschromatography (GC × GC) method, characterized by agreatly increased separation power, if compared with GC× GC approaches using classical column combinations.The analytical objective was achieved by using a 0.05 mminternal diameter (i.d.) capillary as second dimension, asplit-flow approach reported in previous research (Tran-chida, P. Q.; Casilli, A.; Dugo, P.; Dugo, G.; Mondello,L. Anal. Chem. 2007, 79, 2266-2275), and a twin-ovenGC × GC instrument. The column combination employedwas an orthogonal one: an apolar 30 m × 0.25 mm i.d.column was linked, by means of a Y-union, to a flameionization detector (FID)-connected high-resolution 1 m× 0.05 mm i.d. polar one and to a 0.20 m × 0.05 mmi.d. uncoated capillary segment; the latter was connectedto a manually operated split valve, located on top of thesecond GC. As previously shown, the generation ofoptimum gas linear velocities in both dimensions can beattained by splitting gas flows at the outlet of the firstdimension (Tranchida, P. Q.; Casilli, A.; Dugo, P.; Dugo,G.; Mondello, L. Anal. Chem. 2007, 79, 2266-2275).An optimized GC × GC method was developed andexploited for the analysis of a complex petrochemicalsample. The satisfactory results attained were directlycompared with those observed using the same instru-mentation, equipped with what can be defined as aclassical GC × GC split-flow column set: the same primarycolumn was connected to an FID-linked 1 m × 0.10 mmi.d. polar one and to a 0.30 m × 0.10 mm i.d. uncoatedcapillary. It will be herein illustrated that there is stillroom for significant progress in the GC × GC field.

Comprehensive two-dimensional gas chromatography (GC ×GC) can be considered as one of the most important inventionsin the GC field.1 Briefly, a GC × GC experiment is generallycarried out on two columns linked in series and with differing

selectivities; the use of a conventional 0.25 mm internal diameter(i.d.) apolar first-dimension column and a short 0.10 mm i.d. polarsegment is by far the most exploited combination. A transfersystem, normally a cryogenic modulator, has the fundamentalfunction of heart-cutting and reconcentrating first-dimensionbands; the latter, after the compression step, are reinjected ontothe short second dimension. Typically, the modulation processtakes between 4 and 8 s and is carried out in a continuous andsequential mode, throughout the GC × GC application. Compre-hensive two-dimensional GC has been recently described in theliterature.2,3

It has been previously demonstrated that the potential of GC× GC is only partially expressed when using the long 0.25 mmi.d. plus short 0.10 mm i.d. column option, because gas linearvelocities are generally ideal (or near to optimum) in the firstdimension and far from ideal in the second dimension.4 Thedevelopment and employment of a GC × GC system, defined as“split-flow”, enabled the generation of improved gas linear veloci-ties: a first-dimension 0.25 mm i.d. capillary was linked to a second-dimension 1 m × 0.10 mm i.d. one and to a 0.3 m × 0.10 mm i.d.uncoated capillary segment, by using a Y press fit. The second-dimension analytical column was passed through the cryogenicmodulator, whereas the uncoated column was linked to a manuallyoperated split valve. The benefits of the novel approach weredemonstrated by comparing the results attained using a single-oven GC × GC system under split-flow and non-split-flow condi-tions, in fatty acid methyl ester and petrochemical experiments.4

The present research is focused on the employment of a 1 m× 0.05 mm i.d. column as GC × GC secondary column, with theobjective of attaining optimized second-dimension separations. Inorder to achieve such an objective, a split-flow, dual-oven, GC ×GC instrument was used, characterized by a 30 m (0.25 mm i.d.)apolar primary capillary (installed in GC1) connected, through aY-union, to a 1 m × 0.05 mm i.d. polar analytical column and to a0.20 m × 0.05 mm i.d. uncoated capillary, both situated in GC2.The outlet of the uncoated column was connected to a needlesplit valve, situated on top of GC2. A series of experiments were

* To whom correspondence should be addressed. Phone: +39-090-6766536.Fax: +39-090-358220. E-mail: lmondello@unime.it.

† Universita di Messina.‡ Universita degli Studi di Udine.§ Campus-Biomedico.

(1) Liu, Z.; Phillips, J. B. J. Chromatogr. Sci. 1991, 1067, 227–231.

(2) Mondello, L.; Tranchida, P. Q.; Dugo, P.; Dugo, G. Mass Spectrom. Rev.2008, 27, 101–124.

(3) Adahchour, M.; Beens, J.; Brinkman, U. A. Th. J. Chromatogr., A 2008,1186, 67–108.

(4) Tranchida, P. Q.; Casilli, A.; Dugo, P.; Dugo, G.; Mondello, L. Anal. Chem.2007, 79, 2266–2275.

Anal. Chem. 2009, 81, 8529–8537

10.1021/ac901461y CCC: $40.75 2009 American Chemical Society 8529Analytical Chemistry, Vol. 81, No. 20, October 15, 2009Published on Web 09/17/2009

carried out on a diesel sample; the results attained were directlycompared with those attained using the same instrumentation,equipped with a classical GC × GC split-flow column set: the sameprimary column was connected to a flame ionization detector(FID)-linked 1 m × 0.10 mm i.d. polar one and to a 0.30 m × 0.10mm i.d. uncoated capillary.

EXPERIMENTAL SECTIONSamples and Sample Preparation. The diesel sample was

attained from a petrochemical supplier. The sample was injectedneat.

The C17 alkane, used for first-dimension efficiency calcula-tions, was kindly supplied by Sigma-Aldrich (Milan, Italy). Asolution, at the 100 mg/kg level, was prepared in n-hexane.

Nonanal, used for second-dimension efficiency calculations,was kindly supplied by Sigma-Aldrich.

GC × GC Analyses. All comprehensive two-dimensional GCapplications (apart from the GC × GC second-dimension efficiencycalculations) were carried out on a Shimadzu GC × GC systemconsisting of two independent GC2010 gas chromatographs andan FID (280 °C). Data were acquired using the GCsolutionsoftware (Shimadzu, Kyoto, Japan). Bidimensional chromato-grams, inallapplications,weregeneratedbyusingtheChromSquaresoftware (Chromaleont, Messina, Italy). The two GC ovens werelinked through a heated (280 °C) transfer line (Shimadzu). Theprimary GC (GC1) was equipped with an AOC-20i autoinjectorand a split-splitless injector (280 °C). The primary column(situated in GC1), an SLB-5 ms 30 m × 0.25 mm i.d., 0.25 µm filmthickness (silphenylene polymer) capillary (Supelco, Milan, Italy),was passed through the heated transfer line and connected, byusing a fixed outlet capillary column splitter (SGE, Ringwood,Victoria, Australia), to the following:

(GC × GC I setup) a Supelcowax-10 1 m × 0.10 mm i.d., 0.10µm film thickness capillary (poly(ethylene glycol)) and to a0.30 m × 0.10 mm i.d. uncoated capillary segment (bothcolumns, provided by Supelco, were located in GC2);

(GC × GC II setup) a custom-made Supelcowax-10 1 m ×0.05 mm i.d., 0.05 µm film thickness capillary and to a 0.20 m× 0.05 mm i.d. uncoated capillary segment (Supelco). Such alength was chosen because optimum flows can be generatedin both dimensions (first dimension, ∼20 cm/s; second dimen-

sion, ∼110 cm/s), when the split valve is 100% open. If differentgas flows are required, then the valve can be regulated.

In both cases, the outlet of the uncoated gap was connectedto a manually controlled valve, namely, an OSS-2 outlet splittersystem (SGE). A scheme of the split-flow, twin-oven GC × GCinstrument is illustrated in Figure 1.

The end of the primary column was used to create a doubleloop, necessary for cryogenic modulation. The latter was carriedout in GC2 and was applied every 6 s by using a dual-stage loop-type modulator (under license from Zoex Corporation, Houston,TX). The duration of the hot pulse (325 °C) was 375 ms. The FIDwas operated as follows: H2 flow, 50.0 mL/min; air flow, 400.0mL/min; make up (He), 50.0 mL/min; sampling frequency, 125and 250 Hz in the split-flow and non-split-flow GC × GCapplications, respectively.

Optimized split-flow GC × GC I (70% waste) conditions wereas follows: GC1 temperature program, 50-260 °C (10 min) at 3°C/min; GC2 temperature program, 75-260 °C (18 min) at 3 °C/min; initial H2 pressure (constant linear velocity), 61 kPa;injection volume, 1 µL; split ratio, 1:100.

Optimized split-flow GC × GC II (83.4% waste) conditions wereas follows: GC1 and GC2 temperature programs were the sameas those used for setup I; initial H2 pressure (constant linearvelocity), 180 kPa; injection volume, 1 µL; split ratio, 1:150.

Optimized non-split-flow GC × GC I conditions were as follows:GC1 temperature program, 50-260 °C (10 min) at 3 °C/min; GC2temperature program, 60-260 °C (14 min) at 3 °C/min; initial H2

pressure (constant linear velocity), 80 kPa; injection volume,1 µL; split ratio, 1:500.

Non-split-flow GC × GC II conditions were as follows: GC1 andGC2 temperature programs, 50-260 °C (10 min) at 3 °C/min;initial H2 pressure (constant linear velocity), 540 kPa; injectionvolume, 0.5 µL; split ratio, 1:500.

GC × GC first-dimension efficiency calculations: a C17 alkane100 mg/kg solution was injected at an isothermal temperatureof 155 °C, at variable H2 pressures.

GC × GC second-dimension efficiency calculations were carriedout on a Shimadzu GC-2010 gas chromatograph. The gas chro-matograph was equipped with a split/splitless injector (250 °C)and an FID (250 °C) detector which was operated as follows: H2

flow, 50.0 mL/min; air flow, 400.0 mL/min; make up (N2), 50.0

Figure 1. Scheme of the dual-oven, split-flow GC × GC setup. Abbreviations: Inj., injector; HTL, heated transfer line; Det., detector.

8530 Analytical Chemistry, Vol. 81, No. 20, October 15, 2009

mL/min; sampling rate, variable. The GC was retrofitted withan Everest longitudinally modulating cryogenic system (LMCS)

(Chromatography Concepts, Doncaster, Australia). The me-chanical stepper motor drive responsible for the movement of

Figure 2. Bidimensional chromatogram relative to a conventional GC × GC application, carried out on a diesel sample and using a 0.1 mmi.d. second-dimension column: linear velocity first dimension, ∼20 cm/s; linear velocity second dimension, ∼167 cm/s; a +10 °C temperatureoffset was applied in GC2.

Figure 3. Bidimensional chromatogram relative to a split-flow GC × GC application (70% waste), carried out on a diesel sample and using a0.1 mm i.d. second-dimension column: linear velocity first dimension, ∼20 cm/s; linear velocity second dimension, ∼110 cm/s; a +10 °Ctemperature offset was applied in GC2.

8531Analytical Chemistry, Vol. 81, No. 20, October 15, 2009

the cryotrap is controlled by the GCsolution software (Shi-madzu) programmed external events that by means of the

electronic controller start the motor operation, whereas thecryotrap temperature is regulated by a temperature control

Figure 4. Bidimensional chromatogram relative to a split-flow GC × GC application (70% waste), carried out on a diesel sample and using a0.1 mm i.d. second-dimension column: linear velocity first dimension, ∼20 cm/s; linear velocity second dimension, ∼110 cm/s; a +25 °Ctemperature offset was applied in GC2.

Figure 5. Bidimensional chromatogram relative to a non-split-flow GC × GC application, carried out on a diesel sample and using a 0.05 mmi.d. second-dimension column: linear velocity first dimension, ∼10 cm/s; linear velocity second dimension, ∼375 cm/s; no temperature offsetwas applied in GC2.

8532 Analytical Chemistry, Vol. 81, No. 20, October 15, 2009

system through the Comprehensive Chromatography SystemControlsCCSC software (Chromaleont). The latter enables themaintenance of a defined temperature difference between thecryotrap and the GC oven temperature. Carbon dioxide, whichwas supplied by using a pressurized CO2 cylinder,5 cools thecryotrap by its expansion, and a small internal flow of nitrogengas (approximately 10 mL/min) is applied to prevent iceformation inside the trap. A modulation time of 6 s was used,while the cryogenic trap was maintained at a temperature of130 °C below the GC oven temperature.

Applications were carried out on a 100% poly(ethylene glycol)2 m × 0.1 mm i.d., 0.1 µm df or 2 m × 0.05 mm i.d., 0.05 µm df

(Supelcowax 10) column (Supelco). The analyte was cryogeni-cally entrapped after a 1 m segment of capillary and then wasreinjected onto the remaining 1 m segment (this can beconsidered equivalent to a second dimension). Injection mode:approximately 5 mm of the end of a 10 µL SGE syringe needlewas dipped in nonanal. The needle was then inserted in theGC injector; split ratio, 1:500; H2 velocity, variable; temperature,80 °C.

RESULTS AND DISCUSSIONThe 0.25 mm i.d. plus 0.1 mm i.d. column combination has

been the most popular setup used in the GC × GC field, becauseit essentially satisfies the requirements of enlargened first-dimension bands and fast high-resolution second-dimension chro-matography. However, in open tubular column GC it is well-knownthat an i.d. reduction has a beneficial effect on column efficiency:

for capillaries with a high phase ratio, the minimum plate heightvalue can be approximated to the column internal diameter.6,7 Anevident consequence is that a 1 m × 0.05 mm i.d. column canpotentially generate a doubled theoretical plate number, withrespect to a 1 m × 0.1 mm i.d. column, if both are operated underideal conditions (∼20 000 N vs 10 000 N). However, very fewconventional GC × GC applications have been carried out using asecond-dimension 0.05 mm i.d. column; the reasons related to thelatter situation will be illustrated. On the contrary, the advantagesof employing a second-dimension 0.05 mm i.d. column, under idealconditions, in a split-flow, twin-oven system, will be demonstratedand discussed.

Split-Flow, Dual-Oven Comprehensive 2D GC Using a 0.1mm i.d. Second-Dimension Column. A primary apolar 30 m ×0.25 mm i.d. capillary was used in combination with a 1 m × 0.10mm i.d. polar column and a 0.30 m × 0.10 mm i.d. uncoatedcapillary segment; we define the latter configuration as GC × GCI setup. Two optimized methods were developed using the GC ×GC I setup, namely, non-split-flow and split-flow ones.

Initially, a conventional GC × GC method (with no flowsplitting) was optimized for the analysis of a diesel sample. First,the most appropriate first-dimension linear velocity was defined:a first-dimension Golay plot was constructed with the assumptionthat the band-broadening contribution of the 0.10 mm i.d. seconddimension, on the final N values, can be considered as negligible.The unmodulated applications carried out at ∼15, 20, 25, 30, and

(5) Mondello, L.; Casilli, A.; Tranchida, P. Q.; Dugo, P.; Costa, R.; Festa, S.;Dugo, G. J. Sep. Sci. 2004, 27, 442–450.

(6) Sandra, P.; Proot, M.; Diricks, G.; David, F. In Capillary Gas Chromatographyin Essential Oil Analysis; Sandra, P., Bicchi, C., Eds.; Huethig: Heidelberg,Germany, 1987.

(7) Donato, P.; Tranchida, P. Q.; Dugo, P.; Dugo, G.; Mondello, L. J. Sep. Sci.2007, 30, 508–526.

Figure 6. Bidimensional chromatogram relative to a split-flow GC × GC application, carried out on a diesel sample and using a 0.05 mm i.d.second-dimension column: linear velocity first dimension, ∼20 cm/s; linear velocity second dimension, ∼110 cm/s; a +25 °C temperature offsetwas applied in GC2.

8533Analytical Chemistry, Vol. 81, No. 20, October 15, 2009

35 cm/s, analyzing a C17 alkane at an isothermal temperatureof 155 °C (n ) 3), generated N values of circa 122 600, 149 600,147 600, 140 800, and 119 400, respectively. Observing this data,it is clear that the optimum first-dimension linear velocity israther “slow”, viz., 20 cm/s, and that column efficiency is verygood (∼H ) 0.201 mm); in previous GC × GC research, it hasbeen observed that the high first-dimension pressure conditions(the main pressure drop occurs in the second dimension)

reduces the analyte diffusion coefficients, and consequently,optimum velocities are far lower than in single-column GC.8

Now, if an ∼20 cm/s H2 linear velocity is generated in thefirst dimension, the second-dimension gas velocity will cor-respond approximately to ∼167 cm/s.4,8,9 The gas velocitiesgenerated appeared to be the result of a satisfactory compromiseand were used to analyze the petrochemical sample. The initialchromatogram, attained through the application of the same

Figure 7. Bidimensional chromatogram expansions derived from (a) Figure 4 and (b) Figure 6, and relative to the same analyte elution range.

8534 Analytical Chemistry, Vol. 81, No. 20, October 15, 2009

temperature program in both GCs, was characterized by slightwrap-around (data not shown). In order to fit the entire elutionband within the bidimensional space, a +10 °C offset was appliedin GC2. The optimized, non-split-flow GC × GC I result, illustratedin Figure 2, presents a very well-known GC × GC structure andwas characterized by an isothermal zone that started when theGC2 temperature gradient ended.

At this point, it was decided to exploit the split-flow valve:through the application of a 61 kPa head pressure, and diverting70% of the gas flow to waste (uncoated gap), the first-dimensionlinear velocity was essentially maintained, while the second-dimension one was reduced to ∼110 cm/s.4 Prior to the split-flow application and under the 30 (FID)/70 (waste) conditions, afirst-dimension Golay plot was constructed at ∼15, 20, 25, 30, and35 cm/s, analyzing a C17 alkane at an isothermal temperatureof 155 °C (n ) 3): N values of circa 105 600, 131 700, 135 100,132 000, and 121 600 were generated, respectively. It wasconfirmed that the optimum first-dimension linear velocity was20 cm/s and that column efficiency was still good (∼H ) 0.222mm), although slightly lower than that previously observed.The split-flow experiment, which was carried out using the

same temperature programs as those employed in the opti-mized conventional application, is shown in Figure 3.

As is evident, although resolution improved in the split-flowexperiment, extensive wrap-around also occurred. The latter waseasily resolved by increasing the positive offset of GC2 to +25°C. The optimized, split-flow GC × GC I result is illustrated inFigure 4.

The results attained in these initial applications deserve to bediscussed: in the previous split-flow GC × GC research,4 carriedout with a single GC oven, it was shown that the possibility toregulate the gas flows was a very attractive option. In the presentstudy, an additional GC oven was employed; two distinct GC ×GC experiments were carried out with approximately the samefirst-dimension linear velocities. On the other hand, two differentsecond-dimension gas velocities were applied and were evaluatedin terms of efficiency: calculations were carried out using nonanalat a temperature of 80 °C (n ) 3); as described in the ExperimentalSection, second-dimension efficiency measurements were carriedout using an LMCS system. The efficiency values attained can beconsidered only as an indication because (a) second-dimensionretention times were approximate ones, due to the difficulty tocalculate analyte release times, after entrapment; (b) second-dimension bandwidths were characterized by a certain degree ofvariability. At a 110 cm/s H2 velocity, the 1 m × 0.1 mm polarcolumn generated approximately 9650 theoretical plates (H )

(8) Beens, J.; Janssen, H.-G.; Adahchor, M.; Brinkman, U. A. T. J. Chromatogr.,A 2005, 1086, 141–150.

(9) Shellie, R.; Marriott, P.; Morrison, P.; Mondello, L. J. Sep. Sci. 2004, 27,503–512.

Figure 8. Four untransformed high-speed second-dimension chromatograms, relative to single modulations, using GC × GC I and II columncombinations. For the sake of comparison and discussion in the text, a series of peaks have been defined by numbers and by the symbol *.

8535Analytical Chemistry, Vol. 81, No. 20, October 15, 2009

0.104 mm), whereas at a 175 cm/s H2 velocity (approximatelythe same as in the non-split-flow GC × GC I experiment), thesame column generated approximately 7050 theoretical plates(H ) 0.142 mm). Hence, it can be concluded that a moderatelyhigher-than-ideal linear velocity was used in the conventionalapplication, whereas an optimum one was used in the split-flow experiment. Now, considering the GC2 temperatureprograms, positive offsets of +10 °C and +25 °C were appliedin the non-split-flow and split-flow applications, respectively. Byobserving Figures 2 and 4, it can be concluded that no evidentresolution differences were present considering each dimension.Consequently, when using a 0.25 mm i.d. plus 0.1 mm i.d. columncombination, the availability of either independent GC ovens orflow regulation appears to be sufficient to attain satisfactoryanalytical results. In truth, the availability of both technologicaloptions would appear to be more than required to achieveoptimized GC × GC using the previously described columncombination.

Split-Flow, Dual-Oven Comprehensive 2D GC Using a0.05 mm i.d. Second-Dimension Column. A primary apolar30 m × 0.25 mm i.d. capillary was used in combination with a 1 m× 0.05 mm i.d. polar column and a 0.20 m × 0.05 mm i.d. uncoatedcapillary segment; we define the latter configuration as GC × GCII setup. Two methods were developed using the GC × GC IIsetup, namely, non-split-flow and split-flow ones.

The reasons why very few GC × GC applications have beencarried using a second-dimension 0.05 mm i.d. column arebasically two: (I) the reduced sample capacity of a 0.05 mm i.d.capillary. It is obvious that reduced sample amounts must beinjected onto the column. However, because band broadening isgreatly reduced by decreasing the column i.d., then sensitivityshould be essentially conserved; (II) the second reason, the mostimportant one, is that very high (far from optimum) gas velocitiesare generated in the second dimension. Considering non-split-flow GC × GC conditions, and using the GC × GC II setup, a first-dimension H2 velocity of circa 10 cm/s will correspond to asecond-dimension linear velocity of about 375 cm/s.4,8,9 Hence,a very slow first-dimension separation will be combined with anexcessively fast one in the second dimension. In truth, improvedsecond-dimension linear velocity values can be attained byreducing the first-dimension gas velocity; however, such a choicewould appear to be an unwise one, because of the excessivelylong analysis times. The generation of constant first-dimensionlinear velocities, exceeding 12.5 cm/s, was not possible becauseof instrumental pressure restrictions; in such a case, however,the second-dimension gas velocities would have been higher than400 cm/s.

A non-split-flow GC × GC analysis was carried out at a first-dimension H2 velocity of circa 10 cm/s and with a singletemperature program. The result, illustrated in Figure 5, speaksfor itself: there is hardly any second-dimension resolution.

The second-dimension 375 cm/s gas velocity was evaluatedas previously described: circa 1770 theoretical plates (H ) 0.565mm) were generated by the 1 m × 0.05 mm i.d. polar column,elucidating the reason for such poor chromatography. Modifica-tions of the GC2 initial temperature (negative offsets were tested)were of little help, causing only the drift of the main analyte band(data not shown).

At this point, various split-flow GC × GC experiments werecarried out by manually regulating the split valve at differentstages. The best operational conditions were attained with the splitvalve completely opened (83.4% waste): linear velocities of circa20 and 110 cm/s were calculated in the first and seconddimension, respectively.4 Such gas velocity values can be consid-ered as optimum: unmodulated applications were carried out at10, 15, 20, 25, and 30 cm/s, under the same previously describedconditions; the calculated N values were circa 74 100, 117 500,124 300, 105 100, and 81 700, respectively. Observing this data, itis clear that the highest column efficiency (H ) 0.241 mm),attained at a 20 cm/s gas velocity, was slightly lower than thatobserved in the previously described split-flow GC × GC Iexperiment (N ) 135 100). With regards to the 110 cm/s second-dimension linear velocity, this was again evaluated: ∼20 650theoretical plates (H ) 0.048 mm) were generated by the 0.05mm i.d. capillary.

First, a split-flow GC × GC experiment was carried out usingthe same temperature program in both ovens (data not shown).A great improvement was attained in terms of the chromatographyperformance; however, wrap-around again occurred. Conse-quently, a series of positive temperature offsets were tested inGC2, with the best result attained through the application of +25°C. The bidimensional chromatogram relative to the optimizedsplit-flow GC × GC II application is shown in Figure 6.

As can be seen, fine-tuning of the GC × GC temperatureprograms and gas linear velocities has provided a satisfactoryresult. However, in order to evaluate the performance of theanalysis illustrated in Figure 6, a direct comparison must be madewith the split-flow GC × GC I experiment. With regards to thefirst-dimension separation, and on the basis of the Golay plotresults, it was anticipated that 0.25 mm i.d. plus 0.1 mm i.d.combination would have supplied the best result. As expected,the GC × GC I unmodulated experiment was very similar, butslightly more detailed than that relative to GC × GC II (data notshown). It must be noted that the GC × GC II analysis was slightlyslower than GC × GC I, on average 0.75-1 min; hence, peakseluted from the first dimension at a 2-3 °C higher temperatureduring the GC × GC II analysis. Such slight imperfections in thegas flow equations (due to nonideal gas behavior) have been notedpreviously.4 With regards to the second-dimension separation, twochromatogram expansions, derived from Figures 4 and 6, arereported in parts a and b of Figure 7, respectively. A visualcomparison of the expansions, which report the same analyteelution range, is sufficient to appreciate the different degree ofsecond-dimension resolution attained in the two experiments.

As well-known, bidimensional chromatograms are software-generated and derive from a monodimensional chromatogram.Although peak location can be seen, there is obviously no z-axisin a contour plot and, hence, no projection of peaks in height.Depending on the signal intensity at which a GC × GC chromato-gram is visualized, vicinal peaks can be seen at different “apparent”degrees of resolution; it is therefore much easier and more correctto measure resolution using the untransformed monodimensionalchromatogram. A direct comparison was made between twountransformed cuts, relative to the split-flow GC × GC I and IIexperiments. If the cut (and three peaks) defined by the vertical

8536 Analytical Chemistry, Vol. 81, No. 20, October 15, 2009

line in Figure 7a (split-flow GC × GC I) is considered, thenresolution appears to be satisfactory. However, observing theuntransformed second-dimension chromatogram, illustrated in theupper left-hand expansion in Figure 8, the true degree of resolutionis lower than what it appeared to be, with three peaks character-ized by severe overlapping; the same peaks were practicallybaseline-separated in the GC × GC II analysis (upper right-handchromatogram), while the hint of a fourth peak (located by thesymbol * in the figure) also appeared. The lower left-handchromatogram shows five more-or-less resolved GC × GC I peakswith a resolution value of 5.7 calculated for peaks 1 and 5. In theGC × GC II experiment (lower right-hand expansion), theseparation quality between the same five peaks was muchimproved, with a resolution value of 10.1 calculated for peaks 1and 5, accompanied by the appearance of two additional peaks(located by the symbol * in the figure). In this last case, in truth,a resolution value of approximately 8.1 was expected becausetheory dictates that resolution is proportional to the square rootof the N value: if the latter is doubled in the second dimension,then resolution should be increased by a factor of approximately1.4. Generally, slightly higher-than-expected second-dimensionresolution values were obtained in the GC × GC II analysis. Thelatter aspect, which could be related to slightly higher first-dimension elution temperatures, will be subjected to furtherresearch.

On the basis of the experiences made, the optimized employ-ment of a 50 µm i.d. secondary column can be achieved by usingthe following:• a single GC oven and an external split valve;• two GC ovens and a fixed retention gap segment for flow-splitting.

In truth, the second option appears to be the best choice, dueto the greater flexibility and easier use of the system. However,it is important to use a suitable retention gap segment length forthe second-dimension analytical column. For example, using theGC × GC II column set, if approximately 15% of the gas flow isdirected to the FID, then ideal gas velocities in both dimensionsare generated if the right head pressure is applied. At this point,all that needs to be optimized are the temperature programs ineach GC oven. If only a single GC oven is available, using the GC

× GC II column set, then gas flow regulation by using a split valveis highly advisable.

Finally, it must be added that in the initial works of Phillipsand co-workers1,10 the use of a flow splitter prior to modulationwas described: a tee union connected the two analytical columnsto a short capillary segment, enabling the diversion of about athird of the gas flow. That initial approach has been revisited againin the present paper and can be considered as an evolution of theprevious split-flow GC × GC research.4

CONCLUSIONSA series of conclusions can be made in relation to the GC ×

GC experiments described in the present contribution, the firstbeing that there is still room for considerable evolution in theGC × GC field. Additionally, even if the 0.25 mm i.d. plus 0.1 mmi.d. column combination has been by far the most popular, it doesnot appear to be the best choice. In previous years, severalalternative column combinations have also been suggested: 0.25mm i.d. plus 0.15 mm i.d., 0.25 mm i.d. plus 0.25 mm i.d., etc.However, it is our opinion that the considerable advantages ofusing the split-flow GC × GC II column set have been hereindemonstrated. At this point, further development could be directedto the improvement of the first-dimension separation, almostinvariably carried out with a 0.25 mm i.d. column. In fact, futureresearch will be devoted to the following:• an increase in the first-dimension resolving power; forexample, a 0.15 mm i.d. plus 0.05 mm i.d. combination wouldappear to be of high interest;• the development of enhanced-resolution GC × GC/MSmethods.

ACKNOWLEDGMENTThe authors gratefully acknowledge Shimadzu Corporation and

Supelco Corporation for the continuous support.

Received for review July 3, 2009. Accepted September 3,2009.

AC901461Y

(10) Phillips, J. B.; Venkatramani, C. J. J. Microcolumn Sep. 1993, 5, 511–516.

8537Analytical Chemistry, Vol. 81, No. 20, October 15, 2009