Studies on thermionic ionisation detection in comprehensive two-dimensional gas chromatography

J. Sep. Sci. 2006, 29, 2375 –2382 D. Ryan et al. 2375

Danielle RyanPhilip Marriott

School of Applied Sciences, RMITUniversity, Melbourne, Victoria,Australia

Original Paper

Studies on thermionic ionisation detection incomprehensive two-dimensional gaschromatography

This study explores the application of specific thermionic ionisation detection incomprehensive 2-D GC (GC6GC) and represents the first report of GC6GC withnitrogen phosphorus detection (GC6GC-NPD). Of particular interest is the perfor-mance of the NPD with respect to peak parameters of asymmetry and sensitivity.Since GC6GC produces much narrower peaks than obtained with fast GC (e. g.100 ms vs. a1 s) the effect of detector response time and any lack of symmetry aris-ing from the detection step is important if peak separation (resolution) is to bemaintained. It was observed that detector gas flows had a significant impact onpeak asymmetry and peak magnitude, and that optimisation of the detector was cri-tical, particularly for complex sample analysis by GC6GC-NPD. Peak asymmetriesranging from As = 1.8 to 8.0 were observed under different conditions of detectorgas flows. Comparison of GC6GC-NPD with GC6GC-flame ionisation detection(FID) showed the former to be approximately 20 times more sensitive for the detec-tion of nitrogen-containing methoxypyrazines analytes, and GC6GC-NPD had a lar-ger linear detection range compared to GC6GC-FID. Furthermore, comparison ofGC6GC-NPD and GC6GC-TOFMS chromatograms for the analysis of coffee head-space demonstrated the benefits of selective detection, ultimately realised in a com-paratively simplified contour plot.

Keywords: Comprehensive two-dimensional gas chromatography (GC6GC) / Fast gas chromato-graphy / Nitrogen phosphorus detection / Peak asymmetry / Selective detection / Sensitivity /

Received: November 9, 2004; revised: August 21, 2005; accepted: August 22, 2005

DOI 10.1002/jssc.200400033

1 Introduction

In the development of technology surrounding GC6GC,perhaps no other aspect of implementing the procedurehas been as problematic as the detector. Notwithstand-ing the important function of the modulator, it was notlong after the introduction of GC6GC that the cryogenicmodulation approach [1] was demonstrated and hasbeen the predecessor of most of the commercial modula-tors in use today. Likewise, there is little that differenti-ates the injector used in GC6GC from that used in con-ventional capillary GC; all injection methods should beequally applicable. Thus the detector stands apart asbeing the instrumental aspect of continuing develop-ment. Today, it is axiomatic that detector developmentsare still lagging the demands for suitable transducers,optimised for GC6GC.

Fundamentally, selective detection in GC provides twobenefits: (i) isolation of the detection signal of specificcompounds from the net signal pertaining to interferingor unresolved background matrix and (ii) enhanced sen-sitivity for target analytes over the commonly employed,universal flame ionisation detector (FID). The formeracknowledges that the capillary GC method may beincapable of providing adequate separation power to reli-ably separate the target analyte(s) from others (matrix) inthe sample. Thus for overlapping peaks, the selectivedetector is able to give sufficient signal selectivity to per-mit measurement of the target, with the matrix ideallybeing transparent to the detector. Point (ii) is not critical;however, it is often clearly advantageous that the selec-tive detector does have improved sensitivity, particularlyfor trace analyses.

It is apparent that points (i) and (ii) may be considered asinherent benefits of the GC6GC technique. Where theselective detector is capable of isolating response signalsfor improved analysis (i. e. response vs. no response for tar-get and matrix, respectively), GC6GC provides separa-

Correspondence: Professor Philip Marriott, School of AppliedSciences, RMIT University, GPO Box 2476V, Melbourne, Victoria3001, Australia.E-mail: [email protected]: +61-3-96391321.

i 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.jss-journal.com

2376 D. Ryan et al. J. Sep. Sci. 2006, 29, 2375 –2382

tion selectivity through enhanced temporal separationof analytes, as a direct result of increased peak capacity.It may therefore be stated that separation selectivity inthe second column in GC6GC is somewhat equivalent tosignal selectivity in specific detection in conventionalGC. Point (ii) has been the object of some opinion andalso controversy [2], and the question remains as towhether GC6GC offers better sensitivity compared toconventional GC. Regardless of the basic mass flux

,sensitivity’ performance of the detector, the zone com-pression effect in GC6GC with cryogenic modulationensures comparatively narrower peaks and so better sen-sitivity (based on peak height), all other factors beingconstant. However, signal also depends on detector dataacquisition rate; noise increases with increased detectionacquisition rate. This does not address method detectionlimit or considerations of linear range. So if GC6GC ismore sensitive, then for an FID detector it may be possi-ble to reach limits conventionally available to more selec-tive, sensitive detection methods in conventional GC.This was shown to be the case in a study of sterol GC andGC6GC [3]. Additionally, this implies that a selectivedetector with GC6GC should have further advantageswith respect to detection sensitivity beyond those classi-cally achieved with single column GC with selectivedetection or its GC6GC-FID counterpart.

In GC6GC to date, detection techniques employedinclude FID, sulphur chemiluminescence detection(SCD), atomic emission detection (AED), electron capturedetection (ECD), nitrogen chemiluminescence detection(NCD) and MS detection (both TOFMS and quadrupole(qMS)) (see Table 1). In order to conduct GC6GC, the scanspeed of the detector is critical, however Table 1 showsthat less than optimum data acquisition rates have beenused in some instances. Scan speed cannot be directlytaken as an indication of a detector’s applicability or suit-ability for GC6GC analysis. For example, the AgilentmicroECD (lECD) is designed for use with fast GC, butmore correctly we now appreciate this to be conven-

tional speed GC. The acquisition rate of 50 Hz is just onthe limit of useful acquisition speed for GC6GC peaks ofbasewidth 100 ms. Modulated peaks of widths 150 mswith the FID produce peaks of width 300–400 ms in thelECD when the second-dimension column (2D) exit isswapped from the FID to the ECD. To attain narrowerpeaks in keeping with those demanded of GC6GC, thedetector make-up flow rate must be operated at veryhigh settings, outside the recommended range. This stillgives peaks broader than those in the FID, but also com-promises sensitivity due to a dilution effect. Scan speedlimitations are also recognised for GC6GC-qMS, result-ing in a reduced number of data points than normallyrequired for peak quantitation; however, identificationprovided by qMS has proved valuable [4]. Thus, it is appar-ent that each detector used for GC6GC must be criticallyevaluated with respect to operational considerationsthat may limit or affect performance. In the future, theconclusion may be that detectors will require redesign tobe truly compatible with GC6GC.

In the present work, the thermionic ionisation detector(TID), specifically the nitrogen phosphorus detectionmode (NPD) in GC6GC, is reported for the first time.Aspects of its optimisation, and factors such as peakasymmetry and response provided by operational param-eters of gas flow rates and bead voltage are described. Atypical application is shown (without interpretation).

2 Experimental

2.1 Standards

Standard reference compounds 2-isopropyl-3-methoxy-pyrazine (IPMP), 2-secbutyl-3-methoxypyrazine (SBMP)and 2-isobutyl-3-methoxypyrazine (IBMP) were pur-chased from Aldrich Chemical (Milwaukee, USA). The for-mer was 97% pure, while the latter methoxypyrazineswere 99% pure. All standard solutions were prepared inethanol (200 mg/L), suitably diluted with hexane (Ajax


Table 1. Detectors used in GC6GC and selected applications

Detector Application Data acquisition rate (Hz)

FID Plant volatile emissions [9]Trans-resveratrol in wine [10]

100100

lECD Chlorinated dibenzodioxins and dibenzofurans [11]Toxaphene [12]Polychlorinated biphenyls [13]

505050

AED Petrochemical analysis [14] 10SCD Diesel oils [15] 100NCD Diesel fuel [16] 100qMS Ginseng [4] 20TOFMS Petroleum [17]

Cigarette smoke condensates [18, 19]5050

J. Sep. Sci. 2006, 29, 2375 –2382 GC6GC with NPD 2377

Finechem, Seven Hills, Australia) and stored in brownglass bottles at –188C.

2.2 GC6GC

Injections were facilitated by the CombiPAL Focus system(CTC Analytics AG, Zwingen, Switzerland), which wasserially connected to an Optic 3 high performance injec-tor (ATAS GL International, Veldhoven, The Netherlands)and an Agilent 6890 GC (Agilent Technologies, Burwood,Australia). In order to conduct GC6GC analyses, the6890 GC was retrofitted with a longitudinally modulatedcryogenic system (LMCS) from Chromatography Con-cepts (Doncaster, Australia). ChemStation software (Agi-lent) was used to instruct the electronic modulator con-trol to commence modulation at a predefined time. TheCombiPAL Focus system and Optic 3 injector were con-trolled using Cycle Composure version 1.5.2 and ATASEvolution Workstation version 1.2 software, respectively.

The first dimension of the GC6GC column set was aBPX5 (5% phenyl polysilphenylene-siloxane) column(30 m60.25 mm60.25 lm; SGE International, Austra-lia) with a BP20 PEG second dimension (1 m60.1mm60.1 lm; SGE International). Injections were per-formed in the splitless mode (vent closed for 1 min) usingthe Optic 3 injector where the temperature was rampedfrom 50 to 2508C at 108C/s. The GC oven temperature washeld at 508C for 2 min, then ramped at 258C/min to1008C (held 15 min), followed by a final ramping at 258C/min to 2508C. Detection was achieved using an AgilentNPD, and FID (Agilent) for comparative purposes. The FIDwas operated at 2508C with gas flows of 1, 30 and300 mL/min for hydrogen carrier, nitrogen make-up andair, respectively. The NPD was also operated at 2508Cwith optimised detector gas flows of 5, 7.5 and 100 mL/min for hydrogen, nitrogen make-up and air, respec-tively. Column flow (hydrogen) was maintained at 1 mL/min for both detectors and an Adjust Offset of 50 pA wasused (bead sensitivity) with the NPD. All data wereacquired at 100 Hz for GC6GC analyses. Modulationusing the LMCS was performed from 10 to 18 min with a4 s modulation period (PM), ensuring at least three modu-lations per first dimension peak. GC6GC data were trans-formed using an in-house program and visualised as con-tour plots using Transform software (Fortner Research,Virginia, USA). For data analysis, total peak height andarea data were generated using a Matlab program (L. Xie).Statistical analysis was conducted using Minitab 14.

2.3 Analysis of coffee bean headspace byGC6GC-NPD

Arabica (Coffea arabica) headspace was extracted usingheadspace solid-phase microextraction (HS-SPME) asdescribed previously [5]. The SPME fiber was thermallydesorbed in the Optic 3 injector, which was operated in

splitless mode at 2508C for 2 min, after which the splitratio was 20:1. The same column set used for theGC6GC-NPD optimisation was used, with temperatureprogrammed analysis. The GC oven temperature com-menced at 608C (5 min) and was raised to 2308C at 1.58C/min, before final ramping to 2508C (2 min) at 508C/min.The NPD was operated at the optimised gas flow condi-tions described above. Column flow was maintained at1 mL/min hydrogen and an Adjust Offset of 50 pA wasused (bead sensitivity). A PM of 5 s was used and data wereacquired at 100 Hz.

3 Results and discussion

3.1 NPD operation

Modern TIDs are composed of a solid glass or ceramicbead sensor, which has been doped with an alkali metalsalt. The bead is attached to an electrically heated wire,which serves as the thermionic source. Carrier gas ismixed with hydrogen at the base of the detector andflows through a jet where it is then mixed with air. Thehydrogen and air detector flows are insufficient to estab-lish a flame, but rather a plasma located around thethermionic source is maintained. A collector electrodeeither surrounds or is positioned above the thermionicsource, and a voltage is set between the collector elec-trode and the jet tip. This allows the collection of nega-tive ions and the ion current is measured by an electro-meter. The detector output current is proportional to thenumber of ions collected.

The ionisation mechanism of the detector is generallythought to arise due to surface ionisation effects, wherealkali metal atoms are ionised by collision with plasmaparticles. The response of the detector is dependent uponthe composition of the gas environment immediatelysurrounding the thermionic surface, and when air ismixed with a low hydrogen flow, the plasma providesselectivity for nitrogen- and phosphorus-containing com-pounds. Detector selectivity and sensitivity are depen-dent on experimental variables, namely source heatingcurrent, source location, jet potential, air and hydrogenflow rates and the choice of carrier gas.

3.2 Data acquisition rate and experimental design

Since NPD has not been previously used in GC6GC anal-yses, it was important to determine whether the dataacquisition rate of the detector was adequately fast.GC6GC analyses of a standard solution of methoxypyra-zines at 0.5 mg/L was conducted at frequencies of 20, 50and 100 Hz. Analysis of the data showed that 1200, 3000and 6000 data points per minute were generated, respec-tively, indicating the accurate performance of the detec-tor at each condition and the suitability of the NPD forfast GC6GC analyses.



It was then necessary to optimise the performance of theNPD in terms of peak symmetry and magnitude. This wasachieved by varying the ratio of hydrogen, make-up andair flows through the detector. A standard Box–Behnkenstatistical design was applied to this task, comprisingthree factors; hydrogen flow, air flow and nitrogen flow.Agilent recommend a maximum flow of 5 mL/minhydrogen through the NPD and an air flow of 60 mL/min. In this experiment, these values were exceeded inorder to test if the recommended flows produced opti-mal GC6GC-NPD performance, and if settings beyondthe normally advised range might be of use withGC6GC, since it is likely that manufacturers do not testdetector response under stringent fast GC6GC condi-tions. In the present case, detector hydrogen flows of 1, 3and 5 mL/min were used, together with air flows of 40,70 and 100 mL/min and make-up of 5, 7.5 and 10 mL/min. Triplicate centre replicates were performed at 3, 70and 7.5 mL/min hydrogen, air and make-up, respectively.Table 2 shows the results gained in terms of analyte peakheight and second dimension peak asymmetry (2As). Theperformance reproducibility of the detector was verifiedbased on the centre replicate results, which yielded RSDvalues for peak height of 1.8, 1.9 and 2.5% for IPMP, SBMPand IBMP, respectively, and 1.8% RSD for averaged 2As.

3.3 Influence of NPD detector flows on peakasymmetry (As)

It can be assumed that peaks eluted from 2D are approxi-mately symmetric (Gaussian) peaks, since this was foundwith prior FID studies. Furthermore, the amountsinjected gave linear (nonoverloaded) peaks with GC6GC-FID. Nevertheless, the variation in 2As found in thisexperiment with NPD is perhaps rather surprising with a

minimum and maximum average 2As of 1.6 and 8.0,respectively, indicative of ,source tailing’ [6]. Conse-quently, average 2D peak widths also varied and weredirectly correlated with 2As (Table 2). Figure 1 shows theeffects of peak 2As on the use of the 2-D separations space,and clearly shows that when 2As is high, resolution in thesecond dimension is compromised due to peak tailing.From total peak width considerations, the first compo-nent in Fig. 1A encompasses about 5% of the total 2-D sep-aration window, whereas in Fig. 1B, it is about 17%. This


Table 2. Box–Behnken statistical design showing the optimisation of detector gas flow rates using GC6 GC-NPD. Bold textshows experimental centre point replicates

Hydrogenflow (mL/min)

Air flow(mL/min)

Nitrogen flow(mL/min)

Peak heightIPMP(pA)

Peak heightSBMP(pA)

Peak heightIBMP(pA)

Averagepeak 2As

2D peak widthat half height(ms)

1 70 5 76.63 54.72 66.82 2.8 1973 70 7.5 221.18 178.33 185.55 3.2 1405 40 7.5 4.30 0 0 1.9 753 40 5 151.38 142.42 157.99 3.5 4175 100 7.5 457.85 397.35 339.00 1.8 1105 70 10 373.03 327.80 291.02 3.0 1253 40 10 116.95 201.43 219.95 8.0 1883 100 5 270.41 207.70 237.60 2.0 1185 70 5 385.60 333.94 369.30 3.4 1811 100 7.5 78.53 63.27 67.71 1.9 1103 100 10 220.57 220.73 228.10 1.6 983 70 7.5 219.86 178.76 187.63 3.3 1391 40 7.5 123.82 105.36 141.34 5.8 3433 70 7.5 213.99 172.86 178.72 3.3 1381 70 10 123.29 105.20 104.68 2.8 126

Figure 1. Influence of various detector gas flow settings onmethoxypyrazine peak symmetries using GC6GC-NPD. (A)and (B) correspond to detector gas flows that give low andhigh 2As, respectively. Contour settings were from 20 to 600at intervals of 30. Refer to Table 2 for gas flow details.


will pose problems in the analysis of complex samples, inthat peak tailing will exacerbate the possibility of peakcoelution in the second dimension, especially where aminor component closely follows the elution of a majorcomponent.

Using a higher hydrogen flow rate can sometimesimprove tailing [7]; however, statistical analysis of theexperimental design results showed that in the presentcase, 2As was primarily dependent on air flow, and best2As was achieved when air flow was maximum (100 mL/min) (Fig. 2; Table 2). Hydrogen did not significantlyaffect 2As, and no significant interaction between hydro-gen and air flow was observed. This eliminates any syner-gistic relationship between air and hydrogen flows withrespect to 2As.

High peak asymmetry suggests either unswept volumein the detector or processes that cause slow desorptionfrom or slowly reversible reaction at the detector. This isreiterated by Poole [7] who has stated peak tailing mayarise when sample decomposition products remain tem-porarily absorbed on the thermionic source. Addition-ally, Draper [6] suggests that surface decomposition reac-tions may be the rate-limiting step in thermionic ionisa-tion. Unfortunately, the detection mechanism of theNPD is not well understood [6, 7]; however, the above-mentioned theories suggest that it may be difficult toimprove peak symmetry beyond a limiting value definedby decomposition reaction kinetics. One may supposethat the narrower the plasma boundary layer at the sur-face of the bead, the faster the desorption process andthe sharper (less asymmetric) the peak. Potentially, thismay be achieved by increased detector housing tempera-

ture. Nevertheless, testing this theory will require detailsof how the flow ratios and total flow affects the plasmaformation, temperature and geometry.

3.4 Influence of NPD detector flows on peakmagnitude

Statistical analysis of the peak magnitude data showedthat in general, highest peak responses were achievedwhen hydrogen and air flows were maximised (Fig. 3;Table 2). Best detection was achieved for IPMP and SBMPat H2, 5 mL/min; air, 100 mL/min; make-up (N2) 7.5 mL/min, whilst for IBMP, flows of H2, 5 mL/min; air, 70 mL/min; make-up (N2) 5 mL/min produced best peak magni-tude. As observed for 2As data, nitrogen make-up did nothave any significant effect on peak magnitude. Over thetest range, peak magnitude varied from 76.63 to457.85 pA for IPMP (Table 2), equivalent to a six-fold dif-ference in detector response, and highlights the need forcareful optimisation of the NPD.

The NPD response is a function of how well the N (and P)atoms provide perturbation of the electron flux (ionisa-tion) in the plasma. The process, as stated above, is notwell understood, but it seems logical to believe that itdepends on the mass flow of carrier/detector gases andthe effect of the detector gases on the plasma formation.It is possible that these effects are synergistic – a highflow may dilute the N-atom flux and at the same timecould adversely affect the plasma geometry. However,there may be an opposite effect. A smaller plasma at thesurface arising from high gas flow may enhance theatom interaction with the plasma, at the same time asthe high mass flow has a dilution effect on N in the gas


Figure 2. Response surface plot showing peak asymmetry(2As) versus detector hydrogen and air flows.

Figure 3. Response surface plot showing peak responseversus detector hydrogen and air flows.


stream. Coupled with this will be the need to appreciatethe effect of the gas flow geometry (e. g. H2 is premixedwith carrier, air is provided as a sheath flow).

Data presented in Table 2 show that there is not a directcoincidence of best flow settings for 2As and those forresponse. The best settings to use will thus be a decisiontaken by the analyst based on experimental needs for theparticular study. If 2As is most critical to achieve separa-tion (increased 2As will reduce resolution, Rs, in the sec-ond dimension) then settings will be chosen to give an2As closest to unity, irrespective of the effect on peak mag-nitude, since deviations from unity may adversely affectquantitative capability. Conversely, if peak magnitude iscritical to achieve best detection limits for trace compo-nents then sacrifice of peak shape may be acceptable atthe expense of increased peak response. Optimised NPDcondition in terms of peak magnitude was achievedwhen flows of 5 and 100 were used for hydrogen, air.Make-up flow was held at 7.5 mL/min. This condition pro-duced an 2As value very similar to the minimum observed2As of 1.6 (Table 2). It should be noted that air flow wasfurther increased to 150 mL/min in a subsequent anal-ysis with hydrogen and make-up held at 5 and 7.5 mL/min, respectively. An average 2As of 1.4 was achieved;however, detector sensitivity was reduced by approxi-mately 13 % with respect to analyte peak heights, com-pared to that achieved using optimised detector flows.

3.5 Influence of detector output frequency on peakresponse

The sensitivity of the NPD can be further adjusted usingthe Adjust Offset option in the ChemStation software.The bead voltage will then be adjusted until the detectoroutput frequency has stabilised and is equal to the setpoint. An Adjust Offset of 30 pA is recommended; how-ever, increased sensitivity can be achieved when a valueof greater or equal to 50 pA is used. Settings of 30 and50 pA were compared with the latter resulting in an aver-age of 36% improved sensitivity (peak height values) forthe three analytes. No change in 2As was observedbetween the two conditions.

Operating at higher Adjust Offset settings will reducebead life; however, the significant gains in sensitivity

(between 30 and 40%) justify the higher setting. It isknown that the NPD bead has a limited lifetime, andwith time, the performance of the bead will deteriorate.Draper [6] observed severe source tailing with source age-ing. Recoating the NPD source, however, rectified thisproblem. The tailing observed in the present study can-not be attributed to source deterioration, since this studywas the first time the NPD had been used, and triplicateanalyses presented in Table 2 indicate the reproduciblefunctioning of the NPD in terms of peak magnitude and2As.

3.6 Detection limits of GC6GC-NPD versusGC6GC-FID

For comparison purposes, standard solutions ranging inconcentration from 0.005 to 100 mg/L were analysed byGC6GC-FID and GC6GC-NPD. Both detectors were oper-ated at 100 Hz, and injection and column conditionswere kept constant, so as the only variable would be thedetector used. As expected, the NPD gave significantlygreater peak responses compared to those obtained byFID, and subsequently, detection limits were higher inthe latter case. Detection limits were calculated based ona minimum detectable signal of three times the SD of thebackground noise. Noise was determined using theChemStation software, over a retention time range priorto the elution of the methoxypyrazines standards. Table3 shows the comparative results achieved using thedifferent detectors. Calibration curves were linear up to25 mg/L with R2 values of at least 0.99 achieved for eachanalyte using both detectors; however, deviations fromlinearity were observed beyond 25 mg/L, resulting in alinear range of l0.0001–25 and l0.002–25 mg/L for NPDand FID, respectively. Most significantly, NPD was foundto be approximately 15–20 times more sensitive for thedetection of all three methoxypyrazines analytes, withrespect to detection limits, demonstrating the enhancedperformance of NPD for the characterisation of thesenitrogen-containing compounds.

3.7 Application of GC6GC-NPD to complexsample analysis

As a demonstration of the use of the NPD for GC6GCdetection, coffee bean headspace was analysed under


Table 3. Detection of methoxypyrazines standards using GC6 GC-NPD and GC6GC-FID

GC6GCdetector

r backgroundnoise

Minimumdetectablesignala), pA

Linearity range,mg/L

Detection limitIPMP, mg/L

Detection limitSBMP, mg/L

Detection limitIBMP, mg/L

NPD 0.024 0.072 0.0001–25 0.00013 0.00016 0.00016FID 0.031 0.093 0.002–25 0.002 0.003 0.003

a) Calculated as 3r background noise


optimised detector gas flow conditions (H2, 5 mL/min;air, 100 mL/min; make-up (N2) 7.5 mL/min). Coffee waschosen because it contains a wealth of N-containing com-pounds (pyrazines, pyridines, pyrroles) and has recentlybeen analysed by this group using GC6GC-TOFMS [5]. Inthat study, the pyrazines were of main interest, since theauthentic standards were available in a separate com-parative study and methylpyrazines are known to signifi-cantly contribute to coffee aroma [8]. Two different col-umn-set combinations were used [5] with a polar/apolarset yielding the best separation. In the present analysis,an apolar/polar column set was used, ensuring signifi-cant component overlap and low elution temperatures(Te) of polar analytes, as observed previously [5]. Thus inFig. 4C, the nonselective detector (TOFMS-TIC) (TIC–totalion chromatogram) demonstrates significant peak den-sity early in the chromatogram which is greatly reducedwith the NPD selective detector (Fig. 4B). In this way, theselectivity of the detector to only nitrogen (and by infer-ence phosphorus containing) analytes can be fully tested.The GC6GC-NPD result shows a dramatic simplification

with respect to peak density compared to the TICobtained using GC6GC-TOFMS (Fig. 4), where in the lat-ter, all volatile analytes between a mass range of 41–415 U were monitored. Peaks eluting early in the chro-matogram are broad (Fig. 4A and B) due to the splitlesssampling procedure, where peak trapping is not veryeffective in the absence of solvent effect or supplemen-tary cryotrapping at the head of the column. Large, abun-dant peaks were overloaded, and whilst split injectionshould improve peak shape, this would be at the expenseof peak magnitude for smaller abundance components.

Nonoverloaded peaks were found to have an 2As ofapproximately 1.3, which is improved compared to thatachieved for the methoxypyrazines analytes under thesame detector conditions, and is most likely attributedto the temperature programmed conditions used for theGC6GC-NPD analysis of the coffee headspace. However,without optimisation of the detector gas flow param-eters, significant peak coelution in the second dimensionwould occur especially for minor components following


Figure 4. Comparison of GC6GC results using NPD and TOFMS detection for coffee bean headspace. (A) and (B) correspondto the GC6GC-NPD contour plot and chromatogram, respectively, whilst (C) is the GC6GC-TOFMS chromatogram. Note thatthe time scale for (B) and (C) do not exactly compare. Contour settings for (A) were from 100 to 6000 at intervals of 150. See textfor reference to the boxed section.


major compounds in 2D. Figure 4A shows a boxed reten-tion region in which second dimension resolution maybe significantly compromised due to poor peak symme-try and peak broadening, and reinforces the need forcareful optimisation of the NPD for complex sampleanalysis.

4 Concluding remarksThe performance of the NPD when used for GC6GC anal-yses was reported for the first time. The data acquisitionrate of the NPD was found to be adequately fast forGC6GC analyses and could be operated reliably at100 Hz. Peak magnitude (height) was found to be depen-dent upon detector hydrogen and air flows, and maxi-mum peak heights were observed when both gases weremaximised. Additionally, improvements in peak heightswere observed when bead sensitivity was increased. Con-versely, 2As was dependent primarily on air flow and bestpeak symmetry was achieved when air flow was maxi-mised. However, this can adversely affect the peak height.

Achieving best resolution and detection of analytes usingGC6GC-NPD requires careful optimisation of the detec-tor gas flows. This is particularly so for the analysis ofcomplex samples, where peak coelution may occur, andwill be exacerbated if detector conditions result in poorpeak symmetry. However, once optimised, GC6GC-NPDwas found to be highly sensitive, selective and reproduci-ble, compared to its GC6GC-FID and GC6 GC-TOFMScounterparts. GC6GC-NPD was found to be approxi-mately 20 times more sensitive than GC6GC-FID for thedetection of methoxypyrazines, and dramatically simpli-fied the characterisation of nitrogen-containing com-pounds in coffee bean headspace, despite the complexityof the volatile fraction evidenced in GC6GC-TOFMS anal-yses. Future investigations will extend to the validationand application of GC6GC-NPD to phosphorus-contain-ing volatile organic analytes. This research establishesGC6GC-NPD as a selective and highly sensitive meansfor the characterisation of nitrogen (and by inferencephosphorus)-containing compounds, which should findwidespread application for trace detection of volatilenitrogen species.

The present study used NPD data acquisition of 100 Hz.Whilst 200 Hz may be also used, this will not affect detec-tor-based asymmetry where it arises from either largedetector internal volume or slow response kinetics(chemistry) in the detector. The experience with conven-

tional GC6GC conditions used here can be extended toeven faster GC6GC operation (2D peaks widths a 50 ms),where we conclude that the narrower 2D will result inthe extent of peak asymmetry being even greater. It isanticipated that greater attention will be placed onimproved detector performance, consistent with thatrequired of GC6GC, in the future.

The authors wish to thank Rod Minett, Agilent Technologies, forthe gift of the NPD detector, and Dr. Jeff Hughes for assistancewith statistical analyses.

5 References[1] Marriott, P. J., Kinghorn, R. M., Anal. Chem. 1997, 69,

2582–2588.

[2] Harris, C. M., Anal. Chem. 2002, 74, 410A.

[3] Truong, T. T., Marriott, P. J., Porter, N. A., Leeming, R., J.Chromatogr. A 2003, 1019, 197 –210.

[4] Shellie, R. A., Marriott, P. J., Huie, C. W., J. Sep. Sci. 2003,26, 1185–1192.

[5] Ryan, D., Shellie, R., Tranchida, P., Casilli, A., Mondello,L., Marriott, P., J. Chromatogr. A 2004, 1054, 57 –65.

[6] Draper, W. M., J. Agric. Food Chem. 1995, 43, 2077 –2082.

[7] Poole, C. F., The Essence of Chromatography, Elsevier,Amsterdam 2003.

[8] Buffo, R. A., Cardelli-Freire, C., Flavour Frag. J. 2004, 19,99–104.

[9] Perera, R. M. M., Marriott, P. J., Galbally, I. E., Analyst2002, 127, 1601–1607.

[10] Shao, Y., Marriott, P., H�gel, H., Chromatographia 2003,57, S349–S353.

[11] de Geus, H.-J., de Boer, J., Brinkman, U. A. Th., Chromato-graphia 2002, 55, 339 –344.

[12] Koryt�r, P., van Stee, L. L. P., Leonards, P. E. G., de Boer,J., Brinkman, U. A. Th., J. Chromatogr. A 2003, 994, 179 –189.

[13] Koryt�r, P., Leonards, P. E. G., de Boer, J., Brinkman, U.A. Th., J. Chromatogr. A 2002, 958, 203 –218.

[14] van Stee, L. L. P., Beens, J., Vreuls, R. J. J., Brinkman, U. A.Th., J. Chromatogr. A 2003, 1019, 89 –99.

[15] Hua, R., Li, Y., Liu, W., Zheng, J., Wei, H., Wang, J., Lu, X.et al., J. Chromatogr. A 2003, 1019, 101 –109.

[16] Wang, F. C.-Y., Robbins, W. K., Greaey, M. A., J. Sep. Sci.2004, 27, 468 –472.

[17] van Deursen, M., Beens, J., Reijenga, J., Lipman, P., Cra-mers, C., J. High Resol. Chromatogr. 2000, 23, 507 –510.

[18] Lu, X., Zhao, M., Kong, H., Cai, J., Wu, J., Wu, M., Hua, R.et al., J. Sep. Sci. 2004, 27, 101 –109.

[19] Lu, X., Cai, J., Kong, H., Wu, M., Hua, R., Zhao, M., Liu, J.,Xu, G., Anal. Chem. 2003, 75, 4441 –4451.


Studies on thermionic ionisation detection in comprehensive two-dimensional gas chromatography

Documents

Transcript of Studies on thermionic ionisation detection in comprehensive two-dimensional gas chromatography