Anal. Chem.

2001, 73, 1336-1344

Concepts and Preliminary Observations on the

Triple-Dimensional Analysis of Complex Volatile
Samples by Using GC×GC-TOFMS
Robert Shellie, Philip Marriott,* and Paul Morrison

Chromatography and Molecular Separations Group, Department of Applied Chemistry,

Royal Melbourne Institute of Technology, GPO Box 25476V, Melbourne, Victoria 3001, Australia

The high-resolution two-dimensional comprehensive gas ment of peak identity. More recently, MS/MS techniques have
chromatography (GC×GC) separation of a complex sample also provided improved component identification and sensitivity,
of an essential oil is reported, with tentative identification especially for the limited spectral fragmentation information
of selected separated components provided by time-of- produced with soft ionization methods; liquid chromatography-
flight mass spectrometry (TOFMS). The GC×GC tech- electrospray ionization MS and related techniques benefit from
nique allows orthogonal separation mechanisms on the such approaches.2,3 The use of the mass spectrometer to provide
two columns to achieve separation of components that the unique identity of components that overlap on the GC column
otherwise are unresolved on a single column, as is and, hence, neither can be quantitated nor their structures
demonstrated for the pairs of components borneol and assigned in the absence of mass spectral data highlights the
terpinen-4-ol, and cis-caryophyllene and β-farnesene. limitation of single-column capillary gas chromatography, espe-
Peak compression and a short second column used in cially when analyzing complex mixtures. Davis4 has described the
GC×GC lead to generation of fast second-dimension GC GC separation problem on a statistical basis, in which the
peaks and higher detection sensitivity, by about 25 times, probability of overlapping peaks will be common in all but the
as compared to conventional GC elution. This allows many simplest mixtures, and has recently shown in subsequent work
more compounds to be recognized when using the GC×GC that the need for a two-dimensional separation exists5 if improved
approach. Additionally, rapid mass spectral methods are component separation is required. Giddings has defined the
required if accurate data and reliable searchable spectra problem in general terms of dimensionality in the separation
are to be obtained for the fast peaks; this is achieved with process,6 especially concluding that a method that comprises two
TOFMS. This leads to a three-dimensional analytical separation dimensions that allow orthogonal separation (where
technique. Application of the technique to the complex the chemical processes occurring in each dimension and, hence,
essential oil sample containing a range of chemical the separation achieved between sample components is based on
compound classes shows that superior separation and different interaction mechanisms) will provide superior component
more accurate peak assignment results. Once peaks are discrimination. It can be said that the capacity of the separation
identified within the two-dimensional separation space, system has been increased.
it is conceivable that mass spectrometry might no longer It is widely acknowledged that multidimensional gas chroma-
be required for the routine analysis of such samples, tography can deliver improved component separation. There have
instead relying on the precision of flame ionization detec- been numerous reviews and books devoted to the subject of
tion to give quantitative analysis; however, the support of multidimensional gas chromatography (MDGC). Schomburg has
mass spectral characterization will be invaluable in vali- demonstrated the technical implementation of MDGC.7 As recently
dating the GC×GC approach. as 1999, Bertsch8 commented that even though the power of
MDGC had been appreciated and commercial systems were
available the realization of useful and readily employed MDGC
1. INTRODUCTION systems was lacking. The crux of the problem is that conventional
Contemporary high-resolution separation of volatile and semi- MDGC applies the two-column separation advantage only to
volatile compounds normally employs the technique of capillary limited regions of a sample, that is, where the “heartcuts” are
gas chromatography/mass spectrometry (GC/MS).1 The basic
method has not changed greatly over the last 20 years, although (2) Kienhuis, P. G. M.; Geerdink, R. B. Trends Anal. Chem. 2000, 19, 460.
(3) Volmer, D. A.; Vollmer, D. L.; Wilkes, J. G. LC-GC Asia Pac. 1998, 1, 44.
there have been improvements in some of the mass spectral (4) Davis, J. M. In Statistical Theories of Peak Overlap in Chromatography; Brown,
approaches, such as automated peak detection and library assign- P. R., Grushka, E., Eds., Chromatogr. Sci. Series, Vol. 34; Marcel Dekker:
New York,1974.
* To whom correspondence should be addressed: E-mail: philip.marriott@ (5) Davis, J. M.; Samuel, C. J. High Resolut. Chromatogr. 2000, 23, 235.
rmit.edu.au. (6) Giddings, J. C. J. Chromatogr. 1995, 705, 3.
(1) Ragunathan, N.; Krock, K. A.; Klawin, C.; Saaki T. A.; Wilkins, C. L. J. (7) Schomburg, G. J. Chromatogr. 1995, 703, 309.
Chromatogr. A 1999, 856, 349. (8) Bertsch, W. J. High Resolut. Chromatogr. 1999, 22, 647.

1336 Analytical Chemistry, Vol. 73, No. 6, March 15, 2001 10.1021/ac000987n CCC: $20.00 © 2001 American Chemical Society
Published on Web 02/10/2001
conducted. The hyphenated technique of MDGC/MS will again instruments possess the necessary speed of spectral acquisition
provide more identification of components, (see, for example, the to give g50 spectra/second, and thus, ∼10 spectra/peak. Mass
analysis of chiral components of a fruit extract9), but the overall spectrometry using a quadrupole instrument has been reported
separation improvement is still limited to the relatively small for GC×GC,21 but the analysis had to be slowed excessively to
capacity increase of the MDGC process over that of the single get just one reasonable scan across the peaks.
column analysis. The general attributes of TOFMS have been reported by
For a sample of greater complexity that requires MDGC to Guilhaus.22,23 Although this technology is not new and is well-
be applied over the total sample elution, a new approach to the established for instruments such as the MALDI-TOF method, the
MDGC experiment is demanded. The answer is reasonably logical, coupling of GC methods to TOFMS instruments is only very
but took some years to come to fruition. Jorgensen and Bushey recent. Guilhaus has reviewed the limited prior work in this area,24
showed10 that the direct combination of different high performance although the need to adequately match the mass spectral capabili-
liquid chromatography (HPLC) modes (e.g., size exclusion and ties with the chromatographic performance has been recognized
reversed-phase) in which the second dimension separation is for some time,25 and in that review, the limitation of speed in the
conducted on a much faster time scale than that of the primary TOFMS detection was attributed to data handling considerations.
dimension allows a comprehensive separation, with solutes spread The primary applications proposed appear to be for fast GC, for
across the total separation space. The key to the GC application which the number of peaks produced per time is increased over
of a comprehensive chromatography analysis was the ability to conventional GC methods.26 This allows faster turnaround of
rapidly pulse segments of effluent from column 1 to column 2.11 sample analysis, but can only be realized in a practical sense if
Phillips proposed a heated zone between the two columns,12 data analysis is also performed fast. Hence, automated reporting,
subsequently using a rotating heated sweeper to compress and peak deconvolution (fast elution methods would normally decrease
deliver zones of solutes to the second, fast analysis column.13 the solute separation), library searching, and other data processing
Recently, an issue of the Journal of High-Resolution Chromatog- functions must be completed within the time scale of the GC/
raphy [Vol 23 (2000)] was dedicated to the memory of Phillips MS run. Recently, Sacks and co-workers reported the use of fast
and the contributions he made to this area, covering many areas GC/MS with pressure tuning to increase the separation of target
of comprehensive gas chromatography technology. The present solutes in various synthetic mixtures.27 This same technology was
group14-16 applied a cryogenic method to achieve the same effect. applied to a test mixture of gasoline-range hydrocarbons.28 Also
The two most complex sample types reported to date using at Pittcon 2000, two papers on the three-dimensional technique
GC×GC are petroleum-derived samples17 and vetiver essential of GC×GC-TOFMS were presented by the Centres for Disease
oil,18 although the advantages of GC×GC still enhance the analysis Control that discussed fundamental considerations29 and its
of mixtures such as semi-volatile aromatics19 and sterols.20 application to endocrine disruptor analysis.30 In the present study,
While GC×GC provides unsurpassed resolution capability, the reduction in total analysis time is not an object of the study.
most pressing need is to prove that the implied complexity of a However it should be recognized that the greater peak selectivity
mixture is represented by the multitude of peaks spread around of the GC×GC experiment could only be achieved (if indeed it
the 2-D space. This may be viewed as validation of the power of could be achieved at all) in a single-column GC analysis of
the technique. If each separated peak can be positively identified significantly greater time. Hence, with respect to equivalent peak
and, thus, should be reported, quantitated, and taken into capacity, the present analysis is a fast analysis, although the
consideration in the analysis, then the technique can be justified primary column is essentially operated under conventional condi-
to analysts. If, moreover, the technique is simple to implement, tions of 3 °C/min temperature programming with a 30m × 0.25
then all laboratories should be able to enjoy the advantages of mm ID column. An alternative view of the GC×GC method here
the technique. Thus mass spectrometry, the most recognized is that the second column is, in effect, acting as a detection step
spectroscopic tool for identification of GC separated components, for the solutes eluting from the first column. Its detection
is required for the detection step in GC×GC. Because pulsed mechanism is merely a separation process, which happens to be
peaks may be of the order of 150 ms wide at half-width (depending selected according to orthogonal separation principles, and the
upon the column length of the second column and the carrier TOFMS is a mass-selective response as a third dimension.
gas flow velocity and temperature used), conventional quadrupole
instruments are unsuited to the task. Time-of-flight (TOFMS) (21) Frysinger, G. S.; Gaines, R. B. J. High Resolut. Chromatogr. 1999, 22, 251.
(22) Coles, J. N.; Guilhaus, M. Trends Anal. Chem. 1993, 12, 203.
(9) Full, G.; Winterhalter, P.; Schmidt, G.; Herion, P.; Schreier, P. J. High Resolut. (23) Guilhaus, M. J. Mass Spectrom. 1995, 30, 1519.
Chromatogr. 1993, 16, 642. (24) Guilhaus, M.; Selby, D.; Mlynski, V. Mass Spectrom. Rev. 2000, 19, 65.
(10) Bushey, M. M.; Jorgensen, J. W. Anal. Chem. 1990, 62, 161. (25) Holland, J. F.; Enke, C. G.; Allison, J.; Stults, J. T.; Pinkston, J. D.; Newcomoe,
(11) Phillips, J. B.; Beens, J. J. Chromatogr. A 1999, 856, 331. B.; Watson, J. T. Anal. Chem. 1983, 55, 997A.
(12) Phillips, J. B.; Ledford, E. B. Field Anal. Chem. Tech. 1996, 1, 23. (26) van Ysacker, P.; Guilhaus, M.; Roach, L.; Mlynsky, V.; Janssen, H.-G.;
(13) Phillips, J. B.; Venkatramani, C. J. J. Microcolumn Sep. 1993, 5, 511. Leclercq, P. A.; Cramers, C. A. Proceedings of the 18th International
(14) Marriott, P. J.; Kinghorn, R. M. Anal. Chem. 1997, 69, 2582. Symposium of Capillary Chromatography, Riva del Garda, Italy, May 20-
(15) Kinghorn, R. M.; Marriott, P. J. J. High Resolut. Chromatogr. 1999, 22, 235. 24, 1996; p 1496.
(16) Kinghorn, R. M. PhD. Dissertation, Royal Melbourne Institute of Technology, (27) Roberts, G.; Sacks, R. Abstract No. 265, Pittcon 2000, New Orleans, U.S.A.,
2000. March 12-17, 2000; Veriotti, T.; Sacks, R. Abstract No. 268, Pittcon 2000,
(17) Beens, J. PhD. Dissertation, University of Amsterdam, 1998. New Orleans, March 12-17, 2000.
(18) Marriott, P.; Shellie, R.; Fergeus, J.; Ong, R.; Morrison, P. Flavour Fragr. J., (28) Veriotti, T.; Sacks, R. Anal. Chem. 2000, 72, 3063.
2000, 15, 225. (29) Dimandja, J.-M.; Grainger, J.; Patterson, D. Abstract No. 267, Pittcon 2000,
(19) Marriott, P. J.; Kinghorn, R. M.; Ong, R.; Morrison, P.; Haglund, P.; Harju, New Orleans, March 12-17, 2000.
M. J. High Resolut. Chromatogr. 2000, 23, 253. (30) Dimandja, J.-M.; Clouden, G. C.; Colon, I.; Grainger, J.; Patterson, D. Abstract
(20) Truong, T.; Marriott, P.; Porter, N. J. A.O.A.C. Int. 2001, in press. No. 1192, Pittcon 2000, New Orleans, March 12-17, 2000.

Analytical Chemistry, Vol. 73, No. 6, March 15, 2001 1337

 This paper reports the three-dimensional GC×GC-TOFMS
analysis of a lavender sample to demonstrate the analytical power
of the technique to provide improved separation and identification
of major and minor components of the essential oil.

2. EXPERIMENTAL SECTION
2.1 Gas Chromatography System. A model 6890GC (Agilent
Technologies, Burwood, Australia) with Chemstation data system
was used in these studies, either as a stand-alone unit with flame
ionization detection (where the FID was operated at a data
collection frequency of 100 Hz) or interfaced to the mass
spectrometer. The Chemstation event control is used to instruct Figure 1. Schematic diagram of the GC×GC-TOFMS system. GC
the modulation control system to commence modulation at a components: 1, autoinjector; 2, autosampler tray; 3, modulation
device; 4, moveable cryotrap; D1, first-dimension gas chromatography
precise time.
column; D2, second-dimension gas chromatography column; 5,
2.2 Column Set. The same column combination found to be heated reentrant interface region. TOFMS components: 6, ionizing
suited to essential oil work previously31 was used for the present filament, electron flux, and collector; 7, push plate; 8, ion optics; 9,
study. In the GC/MS setup, the second column was inserted beam of ionized molecules; 10, reflectron; 11, multichannel plate
through the heated interface, held at 250 °C, with ∼ 27 cm of the detector.
column located in this transfer region. The primary column was
a 30 m × 0.25 mm ID × 0.25 µm film thickness (df) BPX5-coated plotting. This served to make data processing faster. This
column (5% phenyl; low polarity), directly coupled to a 2.0 m × instrument is capable of unit m/z resolution only. The NIST library
0.1 mm ID × 0.1 µm df BP20-coated column (poly(ethylene provided with the instrument was used for spectral searching.
glycol); polar column) second dimension. Both columns were from 2.5 Data Conversion. Either Chemstation data (100 Hz), or
SGE International (Ringwood, Australia). The GC was operated TOFMS data (50 Hz) may be exported in ascii file format (*.csv
under temperature program conditions, from 60 to 240 °C at a files) for subsequent data display. They comprise 2 or 3 columns
temperature program rate of 3.0 °C min-1. These conditions were of data for GC-FID and GC-TOFMS, respectively. To generate
chosen as standard for most of our studies to allow direct the 2-D separation space display, a conversion program has been
comparison with literature data.32 Split injection was employed, written to take the desired time-response data and generate a
with a split ratio of ∼50:1. matrix based on the modulation frequency and sampling rate. The
2.3 Cryogenic Modulator. The GC×GC experiment is two sets of data (original exported file and the matrix file) were
achieved here by use of a cryogenic modulator as described read into either Origin (Microcal Software, Northampton, MA)
elsewhere,33 with the pneumatic system outlined therein replaced or Transform (Fortner Research), respectively, to present “normal”
by a motor drive. The first demonstration of the principles of or 2-D chromatographic traces. Mass spectral data may be
cryogenic modulation was reported in 1997,14 and a summary of exported in a variety of data types (total ion counts or selected
its unique capabilities was recently outlined.34 The present unit, ion response) to give different chromatographic interpretations
the LMCS Everest model, was from Chromatography Concepts of the results.
(Doncaster, Australia). The cryogenic system was retrofitted to 2.6 Samples. The whole essential oil was provided by
the 6890GC. It was operated under modulation timing of 4-s cycle Australian Botanical Products (Hallam, Victoria, Australia) and
time and 0.5-s hold time in the release position. The CO2 cryogen was used as received. GC×GC-FID results for a number of
coolant maintained the temperature of the modulation trap to at essential oil samples have been reported recently.18,31
least 100 °C below the prevailing oven temperature. Figure 1
illustrates a schematic illustration of the instrumental setup. 3. RESULTS AND DISCUSSION
2.4 Time-of-Flight Mass Spectrometry. A Pegasus II TOFMS 3.1 Description of the Three-Dimensional Analysis Sys-
instrument (LECO Corp., St Josephs, MI) was used to acquire tem. Conventional multidimensional GC separations (effectively
mass spectral data, using 70 eV electron impact ionization. The limited to two separation dimensions) involve heartcutting sections
transfer interface temperature was 245 °C. The mass range of a first-dimension separation to a second column, where in a
collected was from 45 to 250 m/z, with 50 spectra/s transferred subsequent elution step, greater resolution of previously poorly
to the data station. Although 5000 spectra/s are acquired, these resolved components is achieved. The standard approach is to
are processed by sum averaging 100 transients into one spectrum use two different phase types to effect this condition where the
prior to submitting to file. Data were recorded and analyzed using retention mechanism of each column involves different solute-
the software provided with the Pegasus instrument. Most often stationary phase interactions.
the m/z 93 ion data, diagnostic for many of the terpene compounds Maximizing the multidimensional separation advantage over
of interest in these samples, were exported for chromatogram a complete chromatographic analysis is elegantly invoked in the
comprehensive GC×GC technique. Because the rapid, repetitive
(31) Shellie, R.; Marriott, P.; Cornwell, C.; Morrison, P. J. High Resolut.
Chromatogr. 2000, 23, 554. heartcut process is immediately followed by fast GC analysis at
(32) Adams, R. P. Identification of Essential Oils by Ion Trap Mass Spectrometry; the prevailing oven temperature, the orthogonality of the two col-
Academic Press: New York, 1989. umns’ separation mechanisms should be ensured, and the second
(33) Kinghorn, R. M.; Marriott, P. J.; Dawes, P. A. J. High Resolut. Chromatogr.
2000, 23, 245. column’s ability to provide resolution of components, therefore,
(34) Marriott, P.; Kinghorn, R. M. LC-GC Eur. 2000, 13, 428. maximized. Figure 2 illustrates the rapid 2-D analysis of pulsed
1338 Analytical Chemistry, Vol. 73, No. 6, March 15, 2001
 nents resolved to the greatest extent and rapid scanning of the
mass spectrometer, yielding sufficient spectra to define the peak
contour when data are converted to matrix format. The ultimate
aim will clearly be to temporally resolve all components, with the
spectrometer only required for its identification capabilities. This
is a significant change to the role of the mass spectrometer when
used in a single-dimensional separation GC/MS system, where
the MS both gives identification and also, importantly, provides a
degree of specificity to allow overlapping components to be
separately recognized. Thus, in a sample such as kerosene there
will be considerable overlap of different classes of compounds,
Figure 2. Schematic diagram illustrating the procedure for acquisi- but extracted ion chromatograms allow, for instance, the branched
tion of comprehensive gas chromatography data. The first dimension alkanes and mono- and diaromatics to be drawn out of the spectral
(1D) is temperature programmed 50 to 240 °C; the modulator results
data by choosing appropriate diagnostic ions. The mass spectrum
in rapid delivery of pulsed bands to the second dimension (2-D) every
4 s. Thus, the second dimension acts as a fast analyzer of the peaks at any particular scan will still clearly be a composite of the
arriving at the end of the first dimension. The oven temperature unresolved components at that point, and the use of off-peak
increments are ∼0.2 °C during the 4-s interval of elution on the second background subtraction will not completely correct for the
column. overlapping components. This means that mass spectra must be
compromised to some extent, although its wide acceptance
packets of solute as the primary separation (1-D) proceeds. The testifies to the value inherent in the GC/MS method.
temperature increment during the 2-D analysis will be about 0.2 By contrast, in the GC×GC-TOFMS system, there will be
°C, over a 4-s elution time at an oven temperature program rate many more completely resolved components, and in these cases
of 3 °C/min. The first column acts precisely as a conventional there is less of a requirement to resort to methods that can give
capillary GC column. Over the 63-min analysis time, there will be a screen or deconvolution of overlapping components. Addition-
a maximum of some 945 second dimension-pulsed analyses. ally, background subtraction will be a simpler process, because
The system used in the present study comprises two indepen- in many cases, spectral interferences do not have to be subtracted
dent gas chromatographic dimensions (D1 and D2, Figure 1) from overlapping peaks, but it is the spectrometer background
based on a comprehensive gas chromatography process. The third which is subtracted. In a study of atmospheric organic analysis
dimension is provided by mass spectral acquisition of spectra on by using GC×GC-FID,35 it was demonstrated that the chemical
a time frame compatible with the fast GC elution on D2. TOFMS baseline causing an elevated detector response above the true
is the most viable technology available to give this rapid spectrum- baseline can lead to considerable uncertainty in estimation of total
acquisition capability. A multidimensional system provides ad- chemical load of a sample in single-column analysis of complex
ditional data where the dimensions are orthogonal; that is, the mixtures. The high resolution power and 2-D separation space
dimensions must give independent response information. means that it is relatively easy to locate a background that just
Although GC/MS is accepted as a typical example of a two- constitutes detector baseline. This will mean also that spectra
dimensional instrumental analysis system, it has been demon- should be “cleaner” and more precisely represent the true
strated that the combination of GC with GC in the comprehensive component, and this should mean fundamentally better library
separation step prior to MS is capable of independent separation matching.
mechanisms such that the GC×GC column coupling does 3.2 Sequential Pulses of Modulated Components, and
constitute orthogonal separate dimensions. Thus, the orthogonal Resolution of Peaks in 2-D. The GC×GC technique relies on
two-dimensional GC/MS experiment can be extended to a three- pulsing the effluent from the first column to the second column,
dimensional experiment of GC×GC-MS in which a two- with usually about 5 pulses/peak as the desired frequency.
dimensional GC process is finished with mass spectral signal Coeluting peaks from the first column are simultaneously collected
acquisition of components which have been better separated on and pulsed together to the second column, and provided the
the two columns. Such a system has a number of advantages, selectivity of the second column permits, they will then be
primarily for complex mixture analyses when multiple overlapping separated as very sharp individual peaks. Given that the second
solutes will reduce the effectiveness of mass spectral analysis or column is operated under fast elution conditions, the pulses appear
quality of mass spectra if only a single column is employed in at the end of the second column with peak half-widths of the order
GC/MS. The MS identification capacity can also provide a more of a few tenths of a second. For FID, a detection frequency of
informative interpretation of the identities of components where 50-100 Hz is typically used. To obtain reliable peak profiles with
they are separated in the second dimension and, therefore, MS detection, a similar sampling frequency is required. Such fast
provides a basis for making conclusions on the system perfor- spectral acquisition is available with TOF technology.
mance of the coupled column experiment. The data report from the GC×GC-TOFMS experiment reveals
The three-dimensional analysis system, comprising two- groups of pulsed peaks with the same library identification,
dimensional separation and a subsequent third dimension of corresponding to the series of pulses for the one component.
spectral identification, is most suitably developed in the form of These will be separated by a time interval of 4 s, due to the 0.25
comprehensive gas chromatography with time-of-flight mass
spectrometry. Every solute is subjected to the dual separation (35) Lewis, A. C.; Carslaw, N.; Marriott, P. J.; Kinghorn, R. M.; Morrison, P.;
mechanisms of the coupled columns with neighboring compo- Lee, A. L.; Bartle, K. D.; Pilling, M. J. Nature 2000, 405, 778.

Analytical Chemistry, Vol. 73, No. 6, March 15, 2001 1339

 should identify the individual pulses as their appropriate identities
at the respective retention times. Thus, for Figure 3B, nine
separate peaks identified by the characteristic mass spectra of
compound X and nine for compound Y are expected. The actual
number reported will be determined by sensitivity considerations,
especially for the minor pulsed peaks at the extremities of the
band. Table 1 illustrates a summary of the data report for the
analysis of a lavender essential oil sample, in which peaks
identified as the same component are grouped together. Some
compounds were either not identified in the library matching
process, or were not included in the automatic report. The original
data stream may be converted following procedures outlined
elsewhere18 into matrix format to generate a 2-D separation space.
The 2-D plot will be constructed with axes of first-dimension
retention (1tR) and second-dimension peak position (2tR). The latter
of these might not strictly be retention on the second column,
depending on the phenomenon of peak “wrap-around”, where
peaks take longer than the modulation time to traverse the second
column. Figure 4A is a chromatogram of the lavender pulsed-
data stream from which the data in Table 1 were generated.
Converting the pulsed data from Figure 4A into matrix format
produces the 2-D separation space seen in Figure 5. The
components at a retention time of 1270-1290 s (Table 1;
components 21, 22) coelute on the first column. The 2-D space
shows that their contour peaks are resolved in the second
dimension; they are identified by library matching as borneol and
terpinen-4-ol, with the former eluting slightly before the latter on
the first column. Figure 4B,C is an expansion of selected regions
of interest of Figure 4A. The smaller magnitude of the peak pulses
Figure 3. A: Pulsed series of peaks obtained at the end of the at the periphery of the chromatographic peak means that library-
second dimension for a single first-dimension chromatographic peak, match quality deteriorates for these peaks, but the major pulse at
denoted compound X. B: Two overlapping peaks, X and Y, from the the peak maximum is better matched. Because the GC×GC
first column form an envelope profile of pulsed peaks at the end of
the second column. Provided the stationary phase shows adequate technique gives greater peak sensitivity than the corresponding
selectivity to separate the two pulsed components, we get resolution normal capillary GC result, then for a given injected amount, the
of the two species as shown, and hence, interweaving of the two match quality will be better for the GC×GC result. Indeed, it is
resolved pulsed bands. possible to get matches in the GC×GC method where no peak
was seen in the normal GC analysis.
s-1 modulation frequency. Where overlapping peaks from the first More abundant components consist of a greater number of
column occur, then the result will be interleaved pulses of peaks, pulses above the nominal detection or identification limit of the
provided they are resolved on D2, alternatively identifiable by their mass spectrometer. This will also depend on the relative mass
library matches. Figure 3A illustrates the hypothetical interpreta- spectral properties (ion abundances) that affect the library search
tion of such situations for a single peak; in this case, there are quality. Thus, column 2 in Table 1 shows that different compo-
nine separate pulses shown. It is possible to overlay onto the nents have a different number of individual peaks that have
pulsed peak profile the nonmodulated peak distribution (shown separately listed peak maxima. For instance, minor component 6
by the dotted line in Figure 3B), which in the present case is gave only one identified peak at 801.83 s, whereas 1,8-cineole
simply obtained by not turning on the cryofluid and not operating (component 10) gave three peaks at 856.59-, 860.71-, and 864.55-s
the modulating trap. Generally, there is excellent correspondence retention time (differing by the 4-s modulation frequency) that
of these two profiles, although the pulsed peaks have substantially were identified by the library search routine as being this
greater response heights due to the zone compression effect of compound. The major component linalyl acetate (component 25)
the modulator. The total peak area is conserved for the nonmodu- was positively matched in 6 sequential pulses, being at retention
lated single peak and the pulsed series of peaks. A case of a times of 1426.4, 1430.3, 1434.1, 1438.2, 1442.3 and 1446.2 s. Again,
multiple (two) peak overlap (Figure 3B) shows that two compo- an approximately 4-s interval is seen between each successive
nents which would be poorly resolved on the first dimension linalyl acetate pulse, which corresponds to the modulation
(shown as the broad broken lines) are pulsed into slices and frequency. It can also be seen in Figure 5 that the major
completely resolved on the second column with interleaved peak component has a broader peak representation in the 2-D plot,
pulses. The two components are denoted X and Y, respectively. because the contours are drawn at a selected peak height response
3.3 GC×GC-TOFMS of Essential Oil. In the GC×GC- and the more abundant component will give a wider peak at a
TOFMS experiment, a data report at the conclusion of the analysis given response height.
1340 Analytical Chemistry, Vol. 73, No. 6, March 15, 2001
Table 1. Library Matches for Pulsed Chromatographic Result Shown in Figure 5.a

peakb retention time (s)c identityd matche

1 619.25, 623.23 camphene 899
2 671.67, 675.63 R-thujene 851
3 691.39, 695.41, 699.97, 704.01 R-pinene 883
4 763.77, 767.63, 771.57 sabinene 867
5 787.91, 791.87 carene 785
6 801.83 3,7,7-trimethyl-1,3,5-cycloheptatrienef 868
7 817.73, 821.57 β-pinene 861
8 824.41, 828.53, 832.47 limonene 895
9 836.19, 840.11, 844.19 β-phellandrene 857
10 856.59, 860.71, 864.55 1,8-cineole 887
11 906.37, 910.23 unknown 1
12 945.77, 949.73 linalool oxide 862
13 954.87, 958.77 cis-sabinene hydrate 815
14 958.59, 962.39 p-cymene 904
15 970.71, 974.49 γ-terpinene 785
16 980.55, 984.59 terpinolene 568
17 986.79, 990.55, 994.35 3,7,7-trimethyl-bicyclo[4.1.0]hept-4-ene-3-oelf 872
18 1025.9, 1029.7, 1033.9, 1037.3 linalool 890
19 1095.6, 1099.3 1,2,3,4-tetramethyl benzenef 680
20 1201.4, 1205.2 camphor 912
21 1273.5, 1277.5, 1281.2, 1285.2 borneolg 846
22 1279.1, 1282.8, 1286.8, 1290.9 terpinen-4-olg 864
23 1327.2, 1331.1, 1335.0 R-terpineol 830
24 1412.1, 1415.9 naphthalene 770
25 1426.4, 1430.3, 1434.1, 1438.2, 1442.3, 1446.2 linalyl acetate 880
26 1522.6, 1526.3, 1530.4, 1534.4 unknown 2
27 1554, 1557.8, 1561.9 bornyl acetate 828
28 1731.7, 1735.4, 1739.1 neryl acetate 864
29 1787.2, 1791.3 geranyl acetate 889
30 1794.4, 1798.3 germacrene D 759
31 1894.7, 1898.7, 1902.6, 1906.7 1,7-dimethyl-7-(4-methyl-3-pentenyl)- 852
tricyclo[2.1.1.0(2,6)]heptanef
32 1915.3, 1919.1, 1923.1 cis-caryophylleneg 863
33 1926.7, 1930.7, 1934.5 β-farneseneg 885
34 1963.1, 1967.3, 1971.1, 1975.0, 1979.0 R-farnesene 899
35 2075.7, 2079.6, 2083.5 germacrene A 855
36 2151.9, 2155.7, 2159.6 germacrene B 846
37 2342.6, 2346.9, 2351.0 caryophyllene oxide 825
38 2478, 2481.5 cadinol 762
a Identifications were based on the NIST library matches and retention time data derived from known elution patterns of the essential oil
components, as reported in Adams’ text.32 Because each chromatographic peak consists of a series of pulses, the library will give a series of
matches for these related pulses. The number of successful matches will be determined in part by the intensities of the series of pulses that are
above a nominal threshold allowing adequate library identification. Note that a comprehensive terpenes library was not available for this study, and
so this limits the number of positive identifications b Peak numbering identification as in Figure 5. c Retention time as given in the peak table
listing in the post-analysis report. d Peak identified by library and reference data. e Peak match is a similarity index given by the mass spectrometry
software, and is reported for the best match found from among the pulsed peaks for the compound. f Compounds were given by the library match
only, and could not be confirmed by Adams’ data. g Compounds are pairs of overlapping components.

The GC-TOFMS analysis of the same sample at the same respect, to the GC×GC-TOFMS analysis. The test of the method
concentration is shown in Figure 6. In this case, the same column will clearly occur where multiple minor components overlap with
set was used, but the cryofluid was not supplied to the modulator. a major component, as seen previously for vetiver oil18 such that
The result indicates that the sensitivity is much worse for the the mass spectral interpretation will be compromised in the
nonmodulated case, as is well-established now for comprehensive absence of complete peak resolution.
GC that incorporates zone compression between the two dimen- Instances of coelution on the first column can be noted. For
sions. Thus, component 32 (cis-caryophyllene) has a maximum
example, there is a small peak (unknown; peak 11) which would
relative response of 50 000 in the GC×GC experiment, but only
overlap peak 10 (1,8-cineole; refer to Figure 5) and peak 17
2000 in the normal GC experiment. This analysis is able to provide
abutting peak 15, which would overlap with peak 16 (terpinolene)
neither the degree of component sensitivity provided by GC×GC-
in a single separation dimension analysis using the column set
TOFMS nor the quality of spectral matching for the same amount
of sample introduced to the column. The expanded peak inset employed here. Likewise, limonene, β-phellandrene, and β-pinene
indicates that peak widths are about 5 s, as compared to <1 s in (peaks 7, 8, and 9) may overlap to varying extents in single-column
Figure 4A. Presumably, the low sensitivity and high background analysis, depending on their relative amounts, but can be
means that it is more likely that peak apexes are recorded, which independently found in the 2-D space. This region often shows
decreases the apparent peak widths in the normal GC mode. variable degrees of separation on different columns, such that not
The normal GC analysis in Figure 6 exhibits far fewer peaks, all can be effectively resolved, which can confound their identifica-
and so its GC-TOFMS analysis will be inferior, at least in this tion.
Analytical Chemistry, Vol. 73, No. 6, March 15, 2001 1341
 Figure 5. Two-dimensional separation space of data presented in
matrix form from the analysis shown in Figure 4A. Peaks are identified
by reference numbers according to Table 1 components.

Figure 6. Normal GC-TOFMS gas chromatographic analysis of

French lavender essential oil. In this case, the cryofluid was not
introduced to the cryotrap. Note the much lower sensitivity (less signal-
to-noise) and fewer total identifiable peaks in this chromatogram, as
compared to Figure 4A. The inset of expanded data from 1tR ) 1880-
2000 s is the region that is equivalent to that seen in Figure 4B for
the zone-compressed pulsed data.

mentally acquired spectra (Figure 7B,C) in the present study. The

improved correlation of Figure 7A with 7B is evident, especially
regarding most of the patterns for the major peak clusters. The
extra mass flux of the peak when it enters the source, again arising
from zone compression, will be an advantageous feature of the
GC×GC process here. Interestingly, both Figure 7B and 7C differ
Figure 4. A: GC×GC-TOFMS chromatogram of the extracted ion
qualitatively from the library spectrum. It is not clear whether
chromatogram of m/z 93 ion for French lavender essential oil, with
an inset of expanded peaks from 1tR ) 2060-2130 s. B: Expanded mass discrimination differences between TOFMS and the library
region of 1tR ) 1878-2000 s of the chromatogram shown in part A, spectrum method are the reason for the differences, but ion
with two components, 32 and 33, overlapping on the first column. C: statistics uncertainties (due to low ion abundance) for Figure 7C
Further expansions of the chromatogram shown in part A. In part iii, will account for a degree of the difference. A more systematic
a selection of related pulses of peaks that are close to detection limit
study with the TOF analyzer will be required to allow a better
(as defined by low signal-to-noise ratio) is given. Note that peaks 15
and 18 will overlap with unidentified components [labeled with an appreciation of the spectral comparison.
asterisk (*)] on the first column because their peak pulses are As a comparison with the GC×GC-TOFMS result for the
interleaved. lavender sample, Figure 8 shows the equivalent GC×GC-FID
2-D chromatogram. Direct correspondence of the two results
The quality of mass spectral comparisons can be indicated by apparently is not achieved here; however, it should be realized
choosing the major component, linalyl acetate (component 25) that the effect of the vacuum draw on the short second column
using both comprehensive and normal capillary GC-TOFMS. does not necessarily give an equivalence of retention on this
Figure 7A compares the library spectrum with the two experi- column, and so results may be affected. Further experimental
1342 Analytical Chemistry, Vol. 73, No. 6, March 15, 2001
 GC×GC-TOFMS experiment), and the uncertainty of the effect
of the vacuum on retention on this short column section may be
a factor. Many compounds are still effectively resolved in the
GC×GC experiment, and it allows their positive identification,
which can be translated back to the GC×GC-FID separation. The
different distribution of peaks within the 2-D space arises from
the different retentions on the second column. Because the data
are transformed to a matrix on the basis of the modulation
frequency, it is possible that some peaks elute later than the 4-s
period (wrap-around) and, therefore, they are potentially plotted
at different positions within the space. The lower sensitivity of
TOFMS here compared to the FID result means that TOFMS does
not reveal the small peaks that are found in the FID result. In
addition, to get adequate sensitivity, larger amounts of the lavender
oil were injected in the GC×GC-TOFMS experiment. This
appears to affect peak shape (possibly due to nonlinear chroma-
tography effects) in some cases. Although TOFMS is expected
to be more sensitive than FID and quadrupole MS, this was not
the case here. This was interpreted as the TOFMS being poorly
optimized in the arrangement employed; however, it was not
possible to address system sensitivity at the time of this study.
3.4 The Role of GC×GC-TOFMS and GC×GC-FID for
Quantitation. For complex sample analysis, for example, kero-
Figure 7. Mass spectra for linalyl acetate. A, library spectrum; B, sene, the only reliable way to quantitate minor components such
spectrum obtained for the largest peak pulse for component 25 (Table as mono- and diaromatics in single column analysis is by using
1; Figure 5) using GC×GC-TOFMS; C, spectrum obtained when appropriate spectroscopic detection, such as a GC/MS experi-
using single column GC-TOFMS.
ment, that allows selection of their diagnostic ions in an extracted
ion chromatogram or selected ion monitoring analysis. Provided
these ions are specific to the target components, then no
interference will result for the poorly resolved chromatographic
peaks. In such cases, a FID result will not be able to provide
quantitative data for the aromatic compounds. If in GC×GC
complete component separation of the saturated components from
the aromatics is obtained, as has been demonstrated elsewhere
for GC×GC, then FID quantitation becomes a real possibility. The
well-characterized FID response may then be used as the basis
for quantitative analysis. It has been shown elsewhere that the
GC×GC experiment can give quantitative data at least as good
as that in normal capillary GC,20 and the peak pulsing process
does not adversely affect data quality. The extra benefits of
increased peak sensitivity arising from zone compression in the
modulation system will be a further benefit. If GC×GC-FID
analysis is chosen for the above reasons, the primary role of the
GC×GC-TOFMS instrument will be to validate the separation
Figure 8. Example of GC×GC-FID data for the same sample of quality through the identification of components. Once these are
lavender oil. This 2-D chromatogram indicates that the peaks in the identified, then provided the method gives equivalent retentions
GC×GC-FID experiment are apparently more efficient (the peaks with both FID and TOFMS detection, direct comparison will serve
are narrower) than those in the TOFMS experiment. There is clearly
the purpose of qualitative and quantitative analysis. The mass
scope for improved performance in the TOFMS instrument. Correla-
tion of the two experiments' results will not be undertaken here. spectral identity of saturated versus aromatic compounds is a best-
case scenario. In essential oils, many mass spectra are not so
unique, and so the second variable of 2-D relative retention,
optimization may be required to quantify the differences of the combined with TOFMS spectra, will be a valuable additional factor
two arrangements. However, a number of comparisons should in positive identification of components. This is especially so for
be made. The first column retention is (should be) essentially multiple, severely overlapping minor components with major
the same, and so the correlation in this respect (e.g., 1tR times) components in a single column analysis. Ideally complete resolu-
should be sound. Second dimension peaks appear to be broader tion would be preferred in order to have reliable identification.
in the TOFMS detection case. Differences in column lengths (a 3.5 Conclusion. The TOFMS detection of narrow peaks
longer second-dimension column length was employed in the generated under comprehensive two-dimensional gas chromatog-
Analytical Chemistry, Vol. 73, No. 6, March 15, 2001 1343
raphy has been shown to produce greater detection sensitivity peak separation or capacity cannot be realized in a single column
than that achieved in normal GC analysis. This arises from greater analysis.
component mass flux into the spectrometer (by about 25-fold), ACKNOWLEDGMENT
which can be directly translated into larger signal-to-noise and The authors acknowledge the loan of the TOFMS system from
improved spectral ion statistics. Each component is pulsed into LECO Australia, which allowed this work to proceed, and SGE
the mass spectrometer at the modulation rate used to generate International (Ringwood, Australia) for provision of capillary
the GC×GC experiment, and the number of recorded pulses for columns. The assistance of Australian Botanical Products in
a component depends in the first instance upon its abundance, helping to initiate some of the essential oil work, and assisting
and hence, its magnitude above the noise level. In the lavender with some identification of components is appreciated.
sample studied, 38 completely resolved components were re- Received for review August 17, 2000. Accepted December
ported, although in the GC×GC-FID analysis of the same sample 30, 2000.
more than 150 individual resolved peaks were seen. An equivalent AC000987N

1344 Analytical Chemistry, Vol. 73, No. 6, March 15, 2001