Comprehensive two-dimensional gas chromatography-mass spectrometry: A review


  • Luigi Mondello,

    Corresponding author
    1. Dipartimento Farmaco-chimico, Facoltà di Farmacia, Università di Messina, viale Annunziata, 98168 Messina, Italy
    2. Campus-Biomedico, Via E. Longoni, 47 I-00155 Roma, Italy
    • Dipartimento Farmaco-chimico, Facoltà di Farmacia, Università di Messina, viale Annunziata, 98168 – Messina, Italy.
    Search for more papers by this author
  • Peter Quinto Tranchida,

    1. Dipartimento Farmaco-chimico, Facoltà di Farmacia, Università di Messina, viale Annunziata, 98168 Messina, Italy
    Search for more papers by this author
  • Paola Dugo,

    1. Dipartimento di Scienza degli Alimenti e dell'Ambiente, Facoltà di Scienze, Università di Messina, Contrada Papardo, 98166 Messina, Italy
    Search for more papers by this author
  • Giovanni Dugo

    1. Dipartimento Farmaco-chimico, Facoltà di Farmacia, Università di Messina, viale Annunziata, 98168 Messina, Italy
    Search for more papers by this author


Although comprehensive two-dimensional gas chromatography (GC × GC) has been on the scene for more than 15 years, it is still generally considered a relatively novel technique and is yet far from being fully established. The revolutionary aspect of GC × GC, with respect to classical multidimensional chromatography, is that the entire sample is subjected to two distinct analytical separations. The resulting enhanced separating capacity makes this approach a prime choice when GC analysts are challenged with highly complex mixtures. The combination of a third mass spectrometric dimension to a GC × GC system generates the most powerful analytical tool today for volatile and semi-volatile analytes. The present review is focused on the rather brief, but not scant, history of comprehensive two-dimensional GC-MS: the first experiments were carried out at the end of the 1990s and, since then, the methodology has been increasingly studied and applied. Almost all GC × GC-MS applications have been carried out by using either a time-of-flight or quadrupole mass analyzer; significant experiments relative to a variety of research fields, as well as advantages and disadvantages of the MS systems employed, are discussed. The principles, practical and theoretical aspects, and the most significant developments of GC × GC are also described. © 2008 Wiley Periodicals, Inc., Mass Spec Rev 27:101–124, 2008


The invention of the open tubular capillary column was certainly one of the most important milestones in the history of GC (Golay, 1958). Through Golay's brilliance, the high and unsuspected complexity of many real-world samples was to be revealed in the following decades. Current-day one-dimensional GC generates peak capacities [the definition peak capacity (nc) may be considered as the number of peaks that may be stacked side by side in the available one-dimensional space or elution time axis, with a specific resolution value] in the 500–1,000 range and is the most commonly applied method for the analysis of volatile analytes. Well-established GC methods provide rewarding analytical results on what may be defined generically as “simple-to-mediumly complex samples.” As will be seen, the advent of comprehensive GC has enabled a deeper insight into several matrices that require in many cases a re-location into a different, more suited category.

The complete resolution of the compounds of interest, in the minimum time, must always be the primary objective in any chromatographic separation. Ideally, it must ensured that pure analyte bands are delivered to the detector—whatever type the latter may be. The direct consequence of insufficient resolution is that many elution bands will be the summation of two or more overlapping analytes that lead to difficulties or possible errors in the identification and quantitation of target components. In several cases, co-elution might be tolerated if the detector can selectively distinguish target from interfering analytes or if it consists, in its own right, in an additional separative dimension (mass spectrometry).

In the field of GC-MS analysis, many separation scientists occupy one of the two following categories: some analysts pay scarce attention to the GC process, and prefer to circumvent insufficient chromatographic separation by directing the column effluent to a mass-analyzing second dimension: multicompound bands are transformed into a group of ions that are resolved and detected on a mass basis. Mass spectrometry can be very useful for the structural elucidation of overlapping GC peaks. On the other hand, many GC specialists devote themselves almost entirely to the optimization of the chromatographic process, and tend to under-exploit the great potential of MS detection. Such an approach is fine if the ion source receives totally separated solutes, identified commonly by using dedicated MS libraries. Problems arise when peak overlapping occurs that hence demand deeper exploitation of the MS step (for example, by using extracted ions, peak deconvolution methodologies, or knowledge of MS fragmentation processes). In truth, both analytical dimensions are complementary, and should be pushed to their full capacities.

In the last decades, it has become increasingly clear that samples characterized by several hundreds or even thousands of volatile constituents are of common occurrence. In one-dimensional experiments, volatiles are generally randomly located along the retention time axis and frequently co-elute, even when the system peak capacity is much larger than that required to accommodate all the sample components. To reduce or (sometimes) eliminate this phenomenon, the peak capacity must greatly exceed the number of sample constituents (Davis & Giddings, 1983). Taking for granted that the most adequate sample preparation process has been performed, the main question that any analyst must face is “which is the most appropriate separation step for my matrix?”. It is equally important, for apparent reasons, to avoid either a weak or excessively powerful separation procedure. If the most selective stationary phase is being used and the capillary is operated under optimum conditions, then the most obvious way to increase separation power is simply to use a longer column. It is well known, though, that resolution can be doubled for a specific compound pair if the column length is extended four times (there is a square-root dependence of resolution on the number of theoretical plates). In an 11 hr experiment, the use of a 400 m column (1.3 million theoretical plates were reported) was far from sufficient for the total separation of a fuel sample (Berger, 1996).

A more effective way to enhance resolving power is through multidimensional gas chromatography (MDGC), a method introduced more than 40 years ago (McEwen, 1964). Conventional “heart-cutting” MDGC instrumentation enables the transfer of selected bands of overlapping compounds from a primary to a secondary column, connected by means of an interface (either a switching valve or a Deans switch). The first-dimension effluent bands, subjected to re-injection, are selected through preliminary one-dimensional experiments (Lewis, 2002). The principle advantage of the technique is that the transferred fractions are subjected to a further separation on a full-length conventional column. The main drawbacks include the high time costs, the instrumentation complexity, and the operational expertise requirements. Furthermore, continuous transfers applied to the entire sample would cause considerable overlapping of compounds, previously resolved in the first dimension. The peak capacity of such a system is equal to the sum of that of the first dimension and that of the second dimension, the latter multiplied by the number (x) of transferred fractions [nc1 + (nc2 × x)]. The first-dimension peak capacity must be considered because the primary capillary separation is also exploited. In fact, apart from the bands transferred to the second dimension, the first column can be successful in fully resolving other target compounds. MDGC can be considered a prime choice whenever a limited number of heart-cuts require a second dimension analysis. Nevertheless, if the entire sample requires analysis in two different dimensions, then an alternative route must be considered.

It is rather surprising that the general concept of comprehensive chromatography was conceived before the invention of GC itself: a bidimensional planar separation, consisting of two orthogonal chromatographic migrations, was reported in the early 1940s (Consden, Gordon, & Martin, 1944). Comprehensive two-dimensional gas chromatography (GC × GC), certainly amongst the most revolutionary innovations in GC since its introduction, was first reported in 1991 (Liu & Phillips, 1991).

The separation power gap, that exists between comprehensive and one-dimensional GC, is probably even bigger than between the capillary and packed column. If the chromatographic fingerprint of many samples was redefined after the introduction of capillary GC, the same can be affirmed for GC × GC. Many so-called “well-known samples” most probably still conceal a great deal of undiscovered information. In an ideal comprehensive GC instrument, the peak capacity is equal to the product of the peak capacities (nc1 × nc2) in each dimension (usually a primary conventional column and a short secondary micro-bore capillary). Although such a value is probably too optimistic (the reasons will be discussed in the following section), the real peak capacities generated are unprecedented and make the method most suited to unravel highly complex volatile samples prior to the detection system. In general, the main advantages of comprehensive over conventional GC methods are basically three: apart from the increased resolving power, a major beneficial characteristic is enhanced sensitivity through solute band re-concentration to make the technique particularly suitable for trace-level component detection (obviously, trace-amount volatiles must be fully resolved from interfering analytes). A further advantage is the 2D chromatogram formation of chemically similar compound patterns. This aspect is of great help in the identification of unknowns, when no standard components are available, in the absence of a corresponding MS library spectrum, or when the resulting experimental MS spectra are very similar to others (e.g., homologous series of compounds, such as fatty acid methyl esters or terpenes).


Modulation Process

Among all automated comprehensive chromatographic techniques, GC × GC has been the one subjected to the greatest evolution and the most applied. Typically, comprehensive GC experiments are achieved on two capillaries (positioned together in a single, or individually in two GC ovens) connected in series, with a transfer device, defined as modulator, located at the head of the second column. The function of the modulator is to isolate, re-concentrate, and introduce heart-cuts of the primary conventional column effluent onto a short micro-bore column, continuously throughout the analysis. The time required to complete this process is defined as the modulation period (generally in the 4–8 sec range). The primary column is typically apolar and, hence, separation is essentially (although not entirely) related to differing boiling points. Each individual modulation-generated fraction undergoes a further rapid analysis, generally on a polar column: isovolatile analytes are resolved on the basis of differing polarity-based interactions (i.e., dipole–dipole, H-bonding, and polarizability effects); the apolar–polar column combination is defined as the “orthogonal” configuration. An example of the initial modulation process is illustrated in Figure 1. Worthy of note is the fact that, although modulation is today achieved through a variety of different processes, the principles remain basically the same: initially, a sharp band of first-dimension effluent (in this case, three imaginary co-eluting substances) is formed at the head of the modulator, maintained at a sufficiently low temperature (these “primordial” modulators were commonly maintained at ambient temperature) (A); plug mobilization (heart-cutting) is achieved through a rapid heating pulse, of millisecond duration (Δ1), directed to the first part of the modulator tube (B); the heart-cut, transported by the mobile phase, impacts a “cold spot” in a second portion of the modulator; meanwhile, volatiles start to accumulate at the head of the first portion, which has rapidly cooled down to the initial temperature (C); the analyte plug is re-injected onto the secondary column through another rapid heating pulse (Δ2), this time directed to the second part of the modulator tube (D); the different stationary phase selectivity produces three fully resolved compounds that reach the detector (E). As can be observed in stage D, during each short 2D analysis, elution in the first dimension continues and modulation on the following fraction is carried out. The process described is an example of dual-stage modulation, and differs from a one-stage procedure where only a single zone of the modulator is subjected to the alternate heating–cooling effect. The use of a dual-stage system is to be favored because it avoids the transfer of non-focused analytes (a phenomenon defined as “breakthrough”) during the re-injection step. The peaks contained in each fast 2D chromatogram present the same first-dimension elution-time window (expressed in min), and differing second-column retention times (expressed in sec). Ideally, all peaks must be detected before the subsequent re-injection and, hence, second-dimension component retention times must be equal or less than the modulation period. Whenever the elution time of a pulsed solute exceeds the modulation period, then a phenomenon defined as “wrap-around” occurs. Considering a 3,600 sec run-time and a 6 sec modulation period, then 600 fast 2D chromatograms will be generated. Ideally, each peak that elutes from the first column is subjected to several modulations, as illustrated in Figure 2.

Figure 1.

Schematic illustration of the modulation process. The symbols (▪ ● ▴) represent three different analytes.

Figure 2.

The ideal result of three modulations carried out on a triple-component analyte band.

Comprehensive GC analyte peak areas are attained simply by summing the areas belonging to the same compound in each modulated fraction (e.g., area peak B = area B1 + area B2 + area B3). No differences should be observed between the area of an unmodulated single peak and the summation of its pulsed peak areas, after modulation. The use of dedicated software for quantitation in GC × GC is mandatory: the general approach is, first, to integrate all second-dimension peaks through a conventional algorithm, and to sum all areas belonging to the same component. Recently, the fact that commercially available quantitation programs have been mainly sold only in combination with GC × GC hardware has been the subject of wide debate among comprehensive GC academia-users, and has stimulated many researchers to develop their own in-house software.

As could be noted in Figure 2, the modulation event generates an additional valuable characteristic: a considerable increase in sensitivity through band compression. Various opinions exist relative to the degree of signal-to-noise enhancement and, hence, it is rather difficult to provide general values, although an increase factor in the 25–50 range appears to be realistic (the sensitivity enhancement shown in the figure is considerably lower than that normally observed). This rather wide range is due to the different modulator systems today-available, which are characterized by differing performances. Moreover, the secondary column characteristics (length, internal diameter, and stationary phase), gas flow, and temperature program all have a profound influence on sensitivity. An aspect of interest that must be considered is that the chemical noise is also compressed during each modulation and is commonly visualized as a descending streak (or streaks). The latter descends across the chromatogram because its retention time in the second dimension is gradually reduced, as the oven temperature increases. The analytes in the second dimension are generally, but not always, resolved from this bleed band. If overlapping occurs between the column bleed streak and an analyte, then deeper MS investigation would be necessary. The electronic noise can be rather considerable due to the high sampling rates that are necessary for the detection of the very narrow secondary column bands (normally in the 100–500 msec range). The isolation of chemical noise is of great importance, especially in the case of MS detection, because highly pure spectra are generated.

Comprehensive GC experiments are generally carried out with a single detector and, consequently, a “raw” 2D chromatogram is simply nothing more than a stream of rapid secondary column chromatograms. The result is a very complex lining of narrow signals, the interpretation of which is totally impractical. Therefore, the “raw” data must be processed and transformed through dedicated visualization software, in a 2D contour plot (or space plane): every single fast chromatogram is positioned perpendicularly to an x-axis, and is characterized by a primary elution time, expressed in minutes. The resolved peaks in each secondary-column separation are defined by retention times, expressed in seconds, and reported along a y-axis. Resolved analytes are generally visualized as ellipse-shaped peaks, and present colors and dimensions that are related to their specific amounts. Obviously, peak shape and dimension are also dependent on the intensity scale selected. The occurrence of peak overlapping, if not nullified, is greatly reduced, because this undesirable chromatographic feature would require equal elution times on both columns. Many programs enable a 3D visualization option, where conic peaks are projected into a z-axis defined space.

The leap between conventional and comprehensive MDGC was realized through a simple and brilliant intuition: the invention of a “primordial” thermal desorption modulator (Liu & Phillips, 1991). One of the first GC × GC systems used for the analysis of pesticides in human serum (Liu et al., 1994) is illustrated in Figure 3 (modulation was carried out as previously described): the initial 16-cm segment of the secondary column was located in an isolated oven on top of the gas chromatograph, and was used as two-stage thermal desorption modulator. Externally, the column was painted with copper paint; the first stage of the modulator was double the length of the second stage to minimize sample breakthrough. Current pulses of 18 and 16 msec powered the first and second stages, respectively. The interval between the pulses was 250 msec, which was sufficient for the analytes to be focused in the second portion of the modulator (see Fig. 1C) before being launched onto the secondary column through the second heating pulse (see Fig. 1D). The entire process was repeated at a constant time interval (i.e., every 2.5 sec), correspondent to the modulation period. The low-cost system achieved comprehensive GC analyses, but was characterized by a lack of robustness. More efficient modulators, such as the thermal sweeper (Phillips et al., 1999) and the longitudinally modulated cryogenic system (LMCS; Kinghorn & Marriot, 1999), were introduced in the late 90's; both gathered several followers. Comprehensive GC modulation systems have been thoroughly detailed in the literature (Adahchour et al., 2006 and references therein). At present, the modulation process is generally carried out by using effective and robust twin-stage cryogenic modulators that necessitate high amounts of liquid nitrogen or carbon dioxide (i.e., quad-jet modulator, loop-type modulator, LMCS, etc.). Recently, a series of pneumatic modulators, requiring no cooling gases, has been developed; the pneumatic approach is certainly interesting but not widely used at the moment. Future research will necessarily be devoted to the introduction of systems with low operational costs. For example, an apparently robust electrically heated and air-cooled thermal modulator, requiring no cryogenic materials, has been recently described (Libardoni, Waite, & Sacks, 2005). It is the authors' opinion that such a direction, related to the origins of the method, appears to be one worth taking.

Figure 3.

Schematic diagram of a comprehensive GC-FID system and, an expanded view of a two-stage thermal desorption modulator (Liu et al., 1994; reproduced with written permission from the American Chemical Society).

It is widely agreed that any comprehensive GC separation must include the following points:

  • all sample components are subjected to two separations in which their displacement depends on different factors;

  • any two compounds separated in the first dimension must remain separated in the second dimension (this objective is achieved with at least three or four modulations per peak);

  • the elution profiles from both dimensions are maintained (Giddings, 1987; Murphy, Schure, & Foley, 1998; Schoenmakers, Marriott, & Beens, 2003).

In truth, in many comprehensive GC separations the second point is not always met: at least three modulations per peak are required to preserve the first-dimension separation. As mentioned above, modulation periods are nearly always in the 4–8 sec range; that timing means that first-dimension peaks must be characterized by a width in the 15–25 sec range. Obviously, such a condition is rarely met by all peaks across a chromatogram and, hence, some compounds might undergo 1 or 2 modulations and be re-mixed with a vicinal analyte band. Such an event is altogether acceptable if the overall GC × GC separation attained satisfies all the analytical objectives.

Instrumental Parameters: Method Optimization

Comprehensive GC method optimization is certainly one of the most painstaking issues and is, in the opinion of many, one of the several reasons behind the rather limited use of the technique (another is probably a natural affection towards habitual methods). In comparison with conventional GC, the scenario is considerably more complex because the two dimensions are intimately related. Acquired experience in the field of conventional, classical MD, very fast micro-bore column, and high-speed mega-bore column vacuum outlet GC is of great help. Apart from modulation parameters, the main operational conditions that must be considered are the stationary phase chemistries, column dimensions, gas flow, the temperature program(s), outlet pressure conditions, and the detector settings.

In gas chromatography, it can be affirmed that, irrespective of the liquid stationary phase, analyte vapor pressure plays a major role in all GC separations. No two liquid stationary phases present entirely independent separation mechanisms (defined as orthogonal) and, therefore, partial correlation between the dimensions will exist, whatever type of column combination is employed. For this reason, the nc1 × nc2 peak capacity value is an over-estimation of the real value, because specific 2D regions might remain inaccessible: a wide variety of chromatograms reported in the literature are characterized by fan-shaped bands of closely eluting (or overlapping) compounds that extend across different lengths of the space plane. There is normally a lot of empty chromatographic space that surrounds these compound locations. Due to the lack of complete orthogonality, it is rather rare that an early eluting first-dimension compound will be highly retained on the secondary column, and vica-versa. Studies concerning a wide variety of GC phases have highlighted the lack of entirely different selectivities and the need for the development of new GC phases (Poole & Poole, 2007, in press).

During trial-and-error method optimization, the primary apolar and secondary polar combination is usually the first tested. It must be emphasized that the “orthogonal” option is not always the most suitable choice, and other column combinations might produce a better result. Although the separation of the highest number of components per unit of time, and/or the isolation of target analytes is the main objective in any GC method, other aspects must be considered in comprehensive GC. That is, the formation of ordered patterns of structurally related compounds and the avoidance of analyte wrap-around. With regards to the latter aspect, excessive second dimension retention might occur, especially for compounds with highly polar functionalities (acids, alcohols, etc.), when using a polar column. Wrap-around is a common issue in GC × GC experiments and can be tolerated, as long as the objectives of the application are not compromised.

With regards to column dimensions, the most common choice for the primary capillary is the conventional 30 m × 0.25 mm I.D. × 0.25 µm column; the second dimension consists usually in a 1–2 m × 0.1 mm I.D. × 0.1 µm micro-bore column. The main reason for such a combination is related to the necessity of a relatively “slow” analysis in the first dimension and a very fast high-resolution secondary separation. It is important, though, that comprehensive GC applications are performed under as-near-as-possible optimized conditions. However, if column efficiency is considered, then most GC × GC separations are carried out in an obliged compromise situation at a gas linear velocity that is optimum (or slightly less than ideal) in the first dimension and far-from-optimum in the second dimension (Shellie et al., 2004a; Beens et al., 2005). Apart from the lack of orthogonality, the generation of very high second-dimension gas velocities is an additional cause of the low exploitation of the available bidimensional space. Secondary column linear velocities of ≥300 cm/sec are commonly observed in GC × GC applications. A possible solution which would enable both capillaries to be operated under ideal conditions would be to use a wider-bore 0.15–0.18 mm I.D. secondary column. The price to pay to use such an approach is a reduced resolving power in comparison with a 0.1 mm I.D. column. In recent research, a GC × GC system, operated at improved gas linear velocities, was described (Tranchida et al., 2007): a primary column was linked to an FID-connected secondary analytical column, which passed through the modulator, and to a segment of retention gap, which was connected to a split valve. This simple modification enabled gas-flow regulation through the second analytical column. It must be added that J. Phillips, in his first works, recognized that gas flows in the second dimension were too high and, hence, employed a flow splitter prior to modulation: a tee union connecting the two dimensions to a short capillary segment enabled the diversion of ca. 30% of the gas flow (Liu & Phillips, 1991). Analyte quantification is not a problem because flow-splitting between the 2 dimensions can be considered in the same way as split injection (e.g., if a 50:1 split injection and a 50:50 flow-split are applied, the total split is equal to 100:1). Sensitivity loss is due to the reduction of the second dimension gas velocity and not to the flow split which can be counterbalanced by an increase of the injection volume. Nevertheless, the employment of the long conventional + the short rapid column option has always been the most preferred by the average comprehensive GC analyst. Gas-flow regulation is another GC × GC aspect, where there is room for further development.

Fine tuning of the temperature ramp is an issue to be highly considered during method development. GC × GC applications are carried out either in single or dual-oven configurations, the latter being certainly the best, most flexible solution. Temperature increases are generally rather slow (1–3°C/min), enabling the attainment of broader first-dimension peaks. In single-oven GC × GC, an additional advantage is that compounds elute from the primary column at relatively low temperatures, are subjected to more intense interactions with the secondary column stationary phase, and, consequently, occupy a greater portion of the available 2D space. Apart from wrap-around, which will occur if excessively slow temperature rates are applied, losses in sensitivity must also be accounted for. Second-dimension separations are achieved essentially under isothermal conditions: for example, if a 3°C/min ramp is applied with a 6 sec modulation period, then there will be a 0.3°C temperature increase during each secondary column analysis. As a consequence, the second-dimension elution times of successive modulated fractions relative to each analyte undergo a slight decrease.

As aforementioned, the modulation process and high-second-dimension linear velocities generate narrow and rapid analyte bands. Although 2D peak durations vary in relation to the modulator, second-dimension stationary phase-type and thickness, temperature program, and gas flow, they are usually comprised in the 100–500 msec range. Such analyte bands, altogether comparable to those attained in micro-bore column very-fast GC, require detection systems characterized by minute internal volumes, very rapid responses to analyte concentration variations, and high sampling rates (minimum 50 Hz) to avoid extra-column sources of band broadening and incorrect peak re-construction. This last aspect is of considerable importance, because adequate peak quantitation requires at least 10 data points/peak. The most commonly used detection system has been the flame ionization detector (FID), followed by the mass spectrometer. Detailed descriptions of the detectors used in comprehensive GC experiments have been reported in the literature (Adahchour et al., 2006; von Mühlen et al., 2006).

As mentioned above, the sub-optimum exploitation of the 2D space is dependent on two factors; namely, the partial correlation between the dimensions, and the far-from-ideal gas flows. A third cause exists and relates to the analyte plug widths injected onto the second dimension; it is obvious that there is an inversely proportional relationship between bandwidths and peak capacities. The extent of secondary column band broadening is, therefore, dependent on the modulator performance, which varies from system to system. This aspect has been recently considered as a bottleneck in GC × GC, because peak capacities are reduced to only a fraction of the theoretical values [Sandra et al., 2007].

Along with the principal benefits of comprehensive GC, a series of shortcomings have also been emphasized. It is the authors' opinion that there is still much room for improvement and, hopefully, many gaps will be filled in the following years. A final note must be devoted to the fact that many real-world applications can be carried out satisfactorily using a single capillary column. Although the increased costs of cryogenically modulated comprehensive GC per analysis are more than counterbalanced by the analytical performance, the method must only be used whenever truly necessary. It is the authors' impression that many reported comprehensive GC applications might have also been carried out by using conventional GC.


Many GC × GC chromatograms reported in the literature are characterized by literally thousands of peaks. It is unthinkable to rely only upon group-type pattern formation, the use of pure standard compounds, and comprehensive two-dimensional GC linear retention indices (a standardized method for their calculation has not yet been introduced, although a certain amount of research is currently devoted to such a purpose) for positive peak assignment. Although it is true that a GC × GC-FID chromatogram contains a greater amount of information than a GC-FID one, it is unquestionable that MS detection is still necessary.

A wide series of comprehensive GC-MS experiments reported in the last 8 years is listed separately in the reference section: in the 1999–2002 period, only nine articles were reported, whereas in the following 3 years over 60 were published, almost equally distributed. In the first 6 months of the current year, approximately 15 contributions have appeared. Rather than reporting a long table with brief details of every single experiment, an extensive description of what the authors retain as some of the most significant contributions to this field will be provided. As will be seen, characterizing analytical trends and developments, related to the evolution of the technique over the 1999–2006 time range (the year in progress will not be considered), will be linked to specific periods. Many valid experiments cannot be described because of space limitations.

It is worthy of note that the inventor of comprehensive GC affirmed that, in principle, the method was altogether similar to GC-MS: in both techniques, the first analytical system delivers isolated compounds or simplified sub-samples to the secondary analytical system. In the case of GC × GC, the latter is a fast gas chromatograph, in which the sub-samples are separated along an independent secondary elution time axis; in GC-MS, the mass spectrometer subjects the sub-samples to fragmentation, and separates the generated fragments along a secondary m/z axis. The main difference between the two methods consists in the fact that in GC-MS, a single compound can produce several m/z signals, whereas only a single retention time signal is produced in GC × GC (Phillips & Xu, 1995). Several years later, a simulated GC × MS chromatogram was compared to that of a hydrocarbon GC × GC chromatogram (Schoenmakers et al., 2000), whereas very recently, GC × time-of-flight (ToF) MS diesel chromatograms have been generated by using soft ionization methodologies (Wang, Qian, & Green, 2005; Mitschke, Welthagen, & Zimmermann, 2006).

First Appearances: 1999–2000

One of the first combinations of a mass spectrometer to a comprehensive GC instrument was reported at the end of the 1990s (Frysinger & Gaines, 1999). A marine diesel oil was subjected to a very typical bidimensional volatility versus polarity separation (fuels have been widely subjected to GC × GC analysis). The column set effluent was directed to a Hewlett-Packard 5972-series quadrupole mass spectrometer (qMS), operated in the full scan mode (45–350 amu). It is well-known that qMS instruments are limited by the duty cycle, and by the necessity to scan individual ions from each mass in the scan range; hence, these systems present a limited data acquisition speed. In fact, that MS system generated 2.43 scans/sec, which was far too slow for GC × GC requirements: under normal operational conditions, many analyte bands would have probably not been detected at all. The authors affirmed that a faster acquisition rate could be attained by monitoring single ions, causing, though, a loss in spectral information. They also stated, with hindsight, that the best solution would be the use of a ToF-MS instrument. The analytical problem was resolved by intentionally broadening peaks that entered the ion source, through the application of a very slow temperature program (30–250°C at 0.5°C/min), which led to a 7 hr-plus run time. Peak widths of 1.0 sec were generated and enabled the acquisition of about three MS spectra/peak, enough solely for identification purposes. The generation of such bandwidths greatly reduced the second-dimension peak capacity. Although the experiment was successful, it also fully demonstrated the shortcomings of the quadrupole MS for such a GC methodology. The fact that other GC methodologies were carried out at excessively high gas linear velocities for scanning mass analyzers had been reported previously (Holland et al., 1983).

The suitability of ToF-MS for the detection of rapidly eluting peaks was shown in the later part of the 1990s in high-speed micro-bore column GC experiments (the reduction of column internal diameters is the most common approach towards faster GC analyses): a 40 sec GC-ToF-MS application on a 14-component mixture was carried out, with peaks adequately re-constructed by using a 22 spectra/sec acquisition rate (Van Ysacker et al., 1996); in a later, very-fast GC experiment, a higher MS acquisition rate (50 spectra/sec) was applied for the detection of solvents (Davis, Makarov, & Hughs, 1999). In May, 2000, the use of a reflectron-type ToF-MS (Pegasus II, Leco, St. Joseph, MO, USA) in an ultra-fast GC experiment was reported (van Deursen et al., 2000a): 10 compounds were separated in 500 msec on a micro-bore column segment; the authors applied the maximum acquisition rate possible, 500 spectra/sec (a single acquired spectrum consisted of 10 transients), and a 40–200 amu mass range. Even though bandwidths were extremely narrow, ca. 12 msec, peak re-construction was very good. Moreover, the similarity between experimental and library spectra was also generally satisfactory, and demonstrated the effectiveness of the system under such extreme conditions. Another important characteristic was the system capability to deconvolute unresolved peaks on the basis of mass spectral differences: a 10 msec (or 5 spectra) retention time difference was sufficient for the deconvolution algorithm to recognize two overlapping compounds, hence overcoming insufficient chromatographic separation. Spectral deconvolution is possible because ToF-MS achieves high-speed full-spectrum acquisition, without mass spectral skewing (spectral patterns do not change across the chromatographic peak). It was effectively demonstrated that the optimum acquisition speed must preserve resolution and maximize sensitivity. It is worth noting that the velocity of the experiment described and the peak widths generated both exceed those observed in any rapid separation carried out in the second dimension of a GC × GC system.

In August 2000, van Deursen et al. (2000b), passing from very fast to comprehensive GC, used a Pegasus II ToF-MS for the analysis of kerosene. To avoid excessively high second-dimension linear velocities, a 1 m deactivated column was connected to the 0.7 m (0.1 mm I.D.) secondary column. A TIC GC × GC-ToF-MS kerosene typical fan-type profile is shown in Figure 4: the alkanes are situated in a horizontal band along the first dimension (y-axis, in this case), followed by the mono-naphthenes and di-naphthenes; the mono-aromatics are located in the higher parts of the second-dimension retention space. In this case, there was no need to extend the GC run-time (Frysinger & Gaines, 1999), because the MS system was operated at an acquisition rate of 50 spectra/sec, quite sufficient for the rather broadened 2D peaks (w1/2 values were reported to be ca. 200 msec). The long 2D column was probably the main cause for the wide bands, because gas flows were greatly reduced; the second column He linear velocity was estimated to be 100 cm/sec (rather near to ideal). Although the spectral acquisition rate was not exceedingly high (many current-day applications are carried at a 100 spectra/sec or higher rate), each application generated a total number of 210,000 spectra, that required a powerful PC with a large disk space. The production of large amounts of data is a constant characteristic in any GC × GC-MS application. A good example of the usefulness of comprehensive GC is shown in Figure 5, which shows a single untransformed 2D chromatogram relative to the analysis shown in the previous figure; the modulator-injected band was separated into eight peaks in 6 sec on the secondary column. From the number of data points illustrated in the extracted-ion traces, it can be concluded that the acquisition rate applied was appropriate.

Figure 4.

TIC two-dimensional plot of the analysis of a kerosene sample, using GC × GC-ToF-MS (van Deursen et al., 2000b; reproduced with written permission from Wiley).

Figure 5.

Example of a second-dimension chromatogram from the analysis shown in Figure 4, with the unique ion-traces of each compound shown separately. The first-dimension retention-time was ca. 32 min. For compound identification, see van Deursen et al. (2000b) (reproduced with written permission from Wiley).

The Early Years: 2001–2002

In 2001–2002, further GC × GC-ToF-MS experiments were described (Shellie, Marriott, & Morrison, 2001; Adahchour et al., 2002; Dallüge et al., 2002a,b,c; Shellie & Marriott, 2002). In particular, Dallüge et al. (2002a) reported the analysis of mainstream smoke of a self-rolled cigarette: apart from the application in itself, of interest were also the problems encountered during (partially automated) data processing and the requirements, emphasized by the authors, of specialized GC × GC-ToF-MS software.

The matrix volatiles, initially collected in a sample tube containing three sorbent plugs, were injected onto a primary apolar column through thermal desorption; a polar micro-bore column collected fractions from the first dimension every 6 sec and directed them to a Pegasus II ToF-MS. The latter was operated at a spectrum storage rate of 100 Hz, which was necessary for the very rapid eluting peaks. The use of a thick film in the first dimension caused rather high elution temperatures and restricted the application range to the more volatile compounds, because the second-dimension stationary phase (polyethylene glycol) maximum operating temperature was 250–260°C. In general, polyethylene glycol-type phases are characterized by low upper temperature limits and whenever they are employed in GC × GC experiments, the volatile elution range is always restricted. The extreme complexity of the matrix is shown in Figure 6, which illustrates a baseline-corrected TIC GC × GC-MS chromatogram in the C7–C12 elution range. The entire chromatogram was not shown due to the presence of a series of broadened peaks, which extended from the upper chromatogram side to the lower (see insert). Some conclusions can be drawn after observation of the chromatogram: first, besides the high number of entirely and partially resolved peaks, many unresolved analytes, located mainly in the horizontal non-polar zone of the 2D chromatogram, are evident. In this case, even comprehensive GC appears to fail to meet the enormous peak-capacity requirements for such an application. Second, apart from the upper left-hand corner, the 2D space is rather homogeneously occupied, probably due to rather wide-spread wrap-around suffered by the more polar compounds; excessively long second-dimension retention times are quite acceptable when the major objective is to resolve the highest possible number of solutes.

Figure 6.

GC × GC–TOF-MS contour plot (total ion chromatogram, TIC) of cigarette smoke showing the first-dimension range of 500–2,600 sec (Dallüge et al., 2002a; reproduced with written permission from Elsevier).

As mentioned before, many early-eluting second-dimension apolar analytes (essentially alkanes and alkenes) were characterized by severe overlapping in an enlarged band located along the x-axis. The deconvolution algorithm of the software derived good quality mass spectra for many of these compounds, as shown in Figure 7A–D: an expansion of the contour plot illustrated in the previous figure, with a vertical line to indicate a single rapid 2D chromatogram, is shown in A; the single unconverted 2D chromatogram illustrated in B shows an overlapping cluster of mainly hydrocarbons; in this 2 sec second-dimension elution range, 18 peaks were unraveled by the software and are indicated by the horizontal lines. Of these 18 peaks, 9 were positively assigned by considering a minimum value for the MS match factors. The good quality of the mass spectra attained is illustrated in C and D, which report the deconvoluted spectrum of 2-methyl-1,5-hexadiene and its best library match, respectively.

Figure 7.

A: Detail of the GC × GC contour plot of Figure 6. The vertical line at 583 sec indicates the second-dimension chromatogram, which is shown separately in (B). In (B), the horizontal lines indicate the positions where peaks were found by the deconvolution algorithm of the GC-TOF-MS software. Provisional identifications are summarized on the right-hand side of the figure. C: The deconvoluted mass spectra of the peak at 0.24 sec. D: The corresponding library spectrum (Dallüge et al., 2002a; reproduced with written permission from Elsevier).

The enormity of the amount of information generated in comprehensive GC-MS applications and the consequential data-processing difficulties were immediately evident in this experiment. With regards to the GC × GC-MS operational parameters, these were set by using the conventional GC-MS software. The latter, as is obvious, did not foresee the employment of a different I.D., twin-column combination and, hence, could not calculate the gas flows through each column. At the end of the application, contour plots were generated by using two external programs. Further data processing was carried out by using the conventional GC-MS software: initially, all peaks in the unconverted GC × GC chromatogram, with an arbitrary S/N value higher than 30, were found; mass spectra, for all peaks, were calculated by means of deconvolution. This process was carried out on vicinal 500 sec chromatogram portions, due to the software peak table restrictions (only 9,999 peaks could be listed!), and was followed by MS library searching. These entirely automated three steps, peak finding-deconvolution-library matching, took 7 hr for the “raw” chromatogram relative to the contour plot shown in Figure 6. The final large table contained 30,000 peaks, with each component defined by a compound name, CAS number, similarity [this factor expresses the similarity (0–999 range) between the experimental and library spectra, considering all masses], reverse [this factor expresses the similarity (0–999 range) between the experimental and library spectra, considering only the library masses], and probability [this factor expresses the uniqueness (0–9,999 range) of a spectrum, compared with all other library spectra]. Obviously, because peak modulation produces a series of consecutive pulses that contain the same analyte, the number of different component names was much less, namely 7,500. The peak table was exported to a spreadsheet program for further processing that required manual interaction. To filter out mass spectra with low match factors, the authors chose minimum values of 800 and 900, for similarity and reverse, respectively; a total of ca. 1,500 peaks (520 different component names) met these requirements. It was also affirmed that a number of peaks (8,000–9,000) with lower but acceptable match factors could have been identified after manual inspection or by using first-dimension linear retention indices (I): with the additional support of literature-derived I values, only 152 compounds, from a group of 2,500 volatiles with good and acceptable match factors, could be identified (reference I were found only for 238 compounds). In a further GC × GC-ToF-MS investigation carried out by the same group (Dallüge et al., 2002c) similar software-related problems were encountered. Moreover, various aspects relative to method optimization (column characteristics, modulation period and temperature, gas velocities, temperature program) and validation (precision, sensitivity, linearity) were discussed. Analyte detectability was improved considerably (by a factor of 5–7) by using the highest multichannel plate voltage (2,000 V). However, as affirmed by the authors, such an approach reduces the detector life-span. Limits of detection (LOD) were compared between GC × GC and single column GC experiments for a series of pure standard pesticides: comprehensive GC-MS LODs, which were in the 5–23 pg range, were 2- to 5-times higher than those observed in GC-MS. The authors stated that the limited sensitivity increase was due to the fact that peaks were modulated 2–3 times, rather than once. In these, and other works it became immediately clear that powerful PCs and fully integrated software were necessary for instrumental control, to locate and identify peaks in automatically generated contour plots and, in general, to process the enormous amount of data generated by GC × GC-ToF-MS experiments.

A noteworthy GC × enantio-GC-qMS experiment was reported in 2002 (Shellie & Marriott, 2002). The mass spectrometric step was not emphasized, because the qMS system (HP5973, Agilent Technologies, Burwood, Australia) generated an 8.33 Hz acquisition rate in the SIM mode. Of higher interest was the method optimization; namely, the proper selection of column dimensions to exploit the low-pressure outlet conditions: a micro-bore first-dimension apolar column (10 m × 0.1 mm I.D. × 0.1 µm) and a second-dimension wider-bore chiral column (1 m × 0.25 mm I.D. × 0.25 µm) were employed. The reason for such a choice was related to the fact that, in GC, higher optimum linear velocity values can be attained by increasing the diffusion coefficient of the solute in the mobile phase. Apart from a lower mobile-phase molecular weight, this increase can be obtained through a reduction of the pressure across the column. The latter option is achieved by using an MS system, thus creating vacuum outlet conditions. By using short, mega-bore columns, sub-atmospheric conditions extend across a great part of the column length: the ideal gas velocity nears the 100 cm/sec mark, by using a 0.53 mm I.D. capillary. On the other hand, if a micro-bore column is used, then the effects of low pressure are exploited only in a short end segment. Generally, a restrictor is connected to the head of the mega-bore column to prevent an excessively low pressure at the sample inlet and to reduce the flow rate (which can exceed the MS pumping capacity). Additional advantages are represented by the high sample capacity and by greater peak widths; therefore, very high mass-spectral acquisition rates are not required. By using the aforementioned column combination, the required conditions were met: the primary micro-bore column was exploited for solute separation and as a restrictor; the secondary wider-bore column was suited for reduced-pressure rapid chiral separations. A GC × enantio-GC-qMS application was carried out on a bergamot essential oil and, although wrap-around was rather evident, some very nice second-dimension enantiomer separations were shown.

The Turning-Point Year: 2003

The post-1999–2002 years saw a rather considerable increase in the employment of GC × GC-MS, in academic and industrial areas. A fundamental reason relates to the introduction of a twin-oven comprehensive GC-ToF-MS instrument (Pegasus 4D, Leco) with a cryogenic two-stage quad-jet modulator (Dimandja, 2003a); the performance of the latter is very good, although it does require a considerable amount of liquid N2 per analysis. The system was equipped with fully integrated software (Chroma-TOF) for instrument control and entirely automated data processing. Different visualization options (2D, 3D, reconstructed 1D chromatogram) could be employed during the GC run-time. Moreover, as with standard GC-MS programs, specific ions or a selection of different ions could be monitored during the analysis. GC × GC-MS quantitative profiling was possible at the end of the application.

The introduction of rapid-scanning quadrupole mass spectrometers also acted as a stimulant for the expansion of the technique. In comparison to ToF systems, qMS detectors cost much less and, hence, quite a lot of research was devoted to the evaluation of the feasibility of such systems for GC × GC analysis. In the GC × GC history only quadrupole and ToF-MS systems have been used, except for a single ion trap (IT) MS experiment carried out in a portable multidimensional GC system (Wahl et al., 2003). The latter possessed the capability to carry out either heart-cutting or comprehensive GC analyses. The second-dimension bandwidths were necessarily wide (a value of 6.5 sec was reported) because the MS acquisition speed was very low (1 spectra/1.5 sec); the modulation period corresponded to 120 sec. Since then, the employment of ITMS instrumentation has not been reported.

In 2003, the use of qMS systems, operated under relatively fast-scanning conditions, was reported in a series of GC × GC experiments (Shellie & Marriott, 2003a; Shellie, Marriott, & Huie, 2003b; Kallio et al., 2003). The main advantage of such systems over ToF-MS instrumentation appeared to be the much lower costs. In particular, Shellie, Marriott, and Huie (2003b) applied GC × GC-qMS to the analysis of ginseng volatiles. The use of a rapid scanning system (5973 mass selective detector, Agilent) enabled the application of common GC × GC operating conditions. Mass spectra were acquired at a 20 scan/sec rate in a reduced range of 41–228.5 m/z to enable the detection of molecular ions up into the oxygenated sesquiterpene region. The authors reported that reliable quantitative data could not be derived with only three to four data points per peak, which sufficed, on the other hand, for identification. Although not observed in the experiment, insufficient peak sampling can also cause inconsistent second-dimension retention times for pulsed peaks that belong to the same compound. Mass spectral skewing was also investigated, with only slight spectral variations observed across a single peak. The main restrictions of the method applied, highlighted by the authors, were related to the impossibility to derive quantitative data, and to the non-identification of higher molecular weight compounds due to the limited mass range.

Ambient particles, as demonstrated by many epidemiological studies, have a highly negative impact on human health. These effects are caused by physical and chemical properties of the inhaled particles. The potential of GC-ToF-MS (widely used for these studies) and GC × GC-ToF-MS for the characterization of semi-volatile organic compounds (SVOCs) in German airborne particulate matter, with a diameter of up to 2.5 µm, has been compared (Welthagen, Schnelle-Kreis, & Zimmermann, 2003). The GC-MS chromatogram (the separation was volatility-based) was characterized by a typical kerosene-style hump of unresolved carbonaceous matter (UCM) that contained a great amount of SVOCs. The authors estimated that only 15% of the latter were quantified, and affirmed that probably many toxic compounds were contained in the UCM band. GC × GC-ToF-MS was exploited to increase knowledge on the constituents and chemical classes present in aerosol samples. An additional objective was the reduction of the amount of data attained to a suitable dimension for epidemiological studies. The employment of a second polar dimension was successful in distributing SVOCs along the y-axis on the basis of increasing polarity: the Chroma-TOF software found ca. 15,000 peaks. Chromatographic peak data (first- and second-dimension elution times, peak areas, MS peak list) were exported, and were subjected to analysis with external programs (e.g., EXCEL 2000 software, Microsoft, Inc., USA). The 2D chromatograms were re-constructed as bubble plots, where each chromatographic peak was visualized as a circle, the position and diameter of which were related to retention data and areas, respectively. The authors classified the huge number of resolved peaks and set selection rules based on the exploitation of analyte 2D retention data (group-type patterns) and well-known MS fragmentation information. Seven groups of components were defined in relation to a specific y-axis retention time range [e.g., alkanes were in the 1–1.5 sec range, and polar benzenes (with or without alkyl groups) presented 2 sec-plus elution times] and to mass fragmentation criteria (e.g., alkanes present an m/z 57 or 71 base peak, with the second largest peak at m/z 71 or 57, whereas polar benzenes present peaks at m/z 77 above 25% relative intensity). Each group was defined by a color in the bubble plot to facilitate data interpretation. The approach was of interest, although the employment of additional software was again necessary. The fact that GC × GC-ToF-MS methods were limited by the lack of a universal data analysis tool was highlighted by the authors in the concluding comments.

Gradual Expansion: 2004–2006

In 2004, Ryan et al. used a rapid scanning qMS (MS-QP2010, Shimadzu, Milan, Italy) and a ToF-MS (Pegasus III) system in LMCS GC × GC experiments on the analysis of coffee volatiles; the latter, extracted by means of solid-phase microextraction (SPME), were separated on a polar–apolar column set, which provided a more satisfactory separation than the orthogonal set-up. The high complexity of an Arabica sample was revealed by the thousands of peaks scattered across the 2D plot, as can be seen in the qMS result illustrated in Figure 8 (not published in the article). Of interest was the formation of the pyrazine-group pattern, as shown in Figure 9 (not published in the article): pyrazines with the same degree of carbon substitution (e.g., ethylpyrazine and dimethylpyrazines) were located in distinct horizontal bands. The volatility-based secondary separations enabled the isolation of pyrazines from other sample compounds (the degree of co-elution in the first dimension was considerable). Characteristic and high-quality pyrazine mass spectra were generated, with group-specific pyrazines characterized by very similar MS fragmentation patterns. Reliable peak assignment was carried out with the support of information derived from contour-plot pyrazine locations and one-dimensional LRIs. Such an amount of valuable analytical data could not have been derived from single-column GC-MS.

Figure 8.

TIC GC × GC-qMS result relative to SPME-extracted Arabica coffee bean volatiles.

Figure 9.

Pyrazine zone relative to the contour plot illustrated in Figure 8. Peak identification: (1) pyrazine; (2) 2-methylpyrazine; (3) 2,5-dimethylpyrazine; (4) 2,6-dimethylpyrazine; (5) 2-ethylpyrazine; (6) 2,3-dimethylpyrazine; (7) 2-ethyl-6-methylpyrazine; (8) 2-ethyl-5-methylpyrazine; (9) 2,3,5-trimethylpyrazine; (10) 2-ethyl-3-methylpyrazine; (11) 2,6-diethylpyrazine; (12) 2-ethyl-3,5-dimethylpyrazine; (13) 2,3-diethylpyrazine; (14) 2-ethyl-3,6-dimethylpyrazine.

The rapid-scanning qMS instrument enabled the application of a normal mass range (40–400 amu) at a scanning rate of 20 spectra/sec, which was sufficient for reliable peak identification but not for correct peak re-construction. The ToF mass spectrometer, on the other hand, was operated at an acquisition rate of 100 Hz over a 41–415 amu mass range. TIC chromatograms were automatically processed with the Chroma-TOF software; the maximum number of processed peaks was restricted to 1,000 (S/N > 100), a choice related to the extensive time (8 hr) required for data processing (the generation of large data files was stressed by the authors). Peak tables contained first- and second-dimension retention times, CAS number, similarity, probability, relative percent area, and unique mass (defined by the peak find algorithm and used for purity calculations).

In the same year, the first example of quantitation in a comprehensive GC application, by using a qMS system (5972, Agilent), was reported (Debonneville & Chaintreau, 2004): known analytes (fragrance allergens) were analyzed by monitoring selected ions (SIM) in pre-defined retention-time windows. A detection frequency of 30.7 Hz was reported and affirmed to be sufficient for peak quantitation: comprehensive GC-MS data were in good agreement with those data derived from a single column GC-MS experiment. Although such an approach can only be applied to known target analytes, the sampling rate still appears to be insufficient for the proper construction of the narrower analyte bands (50–100 msec).

In gas chromatography, matching chromatogram profiles between GC-FID and GC-MS applications is often sought for and is a rather well-known procedure. As demonstrated by Shellie et al. (2004a), in comprehensive GC the issue is much more challenging, because the column set-up consists in the combination of two capillaries with differing internal diameters. Consequently, although FID and MS experiments might be carried out at the same nominal average linear velocity, different retention times will be noted in both dimensions; this factor is due to the different pressure profiles in each column, under atmospheric and vacuum conditions. It was shown that, by employing an additional gas supply, which was directed to a T-union between the secondary column outlet and the MS interface, it was possible to attain identical GC × GC-FID and GC × GC-MS chromatograms, relative to mixtures of pure standard compounds. A further advantage of the described approach was that column changing was a much easier task.

Over the years, persistent organic pollutants (POPs) have been widely subjected to conventional GC analysis, generally with selective detection and/or MS detection. Among these toxic analytes, the separation of the 209 polychlorinated biphenyl (PCB) congeners has proven to be a very challenging task. PCBs can be classified in 10 homolog groups on the basis of their chlorine content and, due to their similar chemical-physical properties, the analytical difficulties encountered are not surprising. An experiment focused on the separation of the highest possible number of PCBs, through comprehensive two-dimensional GC-ToF-MS, has been described (Focant, Sjödin, & Patterson, 2004a): four thermally stable column sets were tested, with the best set consisting of a slightly polar first dimension and a moderately polar second dimension. Obviously, the use of thermally stable columns was related to the low vapor pressure (and, therefore, high elution temperature) of many PCBs. A separation of a PCB standard solution is illustrated in Figure 10 with the analytes distributed in a diagonal band across the 2D plane. The great amount of unoccupied 2D space was essentially dependent on the similar chemical-physical properties of the PCBs. The good deal of pattern formation present in the chromatogram was of great help in terms of peak assignment: homolog groups are located in distinct zones, even though vicinal families tend to overlap. Furthermore, depending on the number of chlorine substituents, up to five sub-series were separated within the same group. In fact, the first-dimension phase carborane group enabled the separation of PCBs on the basis of the degree of ortho-substitution, from non-ortho-CBs to tetra-ortho-PCBs. The method developed allowed the chromatographic separation of 188 congeners, an additional four required deconvolution and the remaining 17 analytes were distributed in 8 complete co-elutions. All the dioxin-like WHO PCBs, as well as the EU marker PCBs, were well-resolved.

Figure 10.

GC × GC-ToF-MS chromatogram of the 209 PCB congeners using an HT-8/BPX-50 column set (Focant, Sjödin, & Patterson, 2004a; reproduced with written permission from Elsevier).

Comprehensive GC, in combination with isotope-dilution ToF-MS, has also been applied for the simultaneous analysis of 59 target POPs in human serum and milk (Focant et al., 2004b): 38 PCBs, 11 organochlorine pesticides (OCP), and 10 brominated flame retardants [polybrominated diphenyl ethers (PBDEs)]. The authors (from Centers for Disease Control and Prevention, USA) affirmed that no previously reported method (usually GC-IDHRMS) allowed the simultaneous determination of these compounds in human fluids. The GC × GC column combination consisted of a completely apolar first dimension and of a slightly polar second-dimension. The separation of the 59 target POP standards is shown in Figure 11A: the rather closely eluting analytes are located in a diagonal band, which would appear to show a lack of orthogonality of the column set-up (a substantial part of the contour plot is unoccupied). In truth, the (sufficient) second-dimension analyte scattering is due to the similar chemical properties of the POPs. The polar column was also responsible for the isolation of an enlarged band of matrix-related interferences, located along the x-axis (Fig. 11B,C). These results highlight a further favorable aspect of GC × GC, viz. the reduced requirements of tedious sample clean-up processes (obviously, this is also dependent on the number of interfering analytes). The instrumental detection limits, which were determined by considering the lowest quantity of a solute that generated a S/N ratio > 3, ranged between 0.5 and 10 pg/µL. Method detection limits, determined by spiking bovine serum, were between 1 and 15 pg/µL. The newly developed approach was tested against a validated GC-IDHRMS procedure, which consisted of three separate applications. Precision was nearly as good as in the single-column technique, and the POP levels measured were very similar.

Figure 11.

Comprehensive two-dimensional gas chromatography results for a contour plot of total ion chromatogram of (A) the 100 pg/µL native compound multianalytes calibration solution, and (B) a human serum sample. The shaded surface plot and the reconstructed one-dimensional trace in (C) are based on specific extracted ion current for the same human serum sample as in (B). The 2D scale was shifted by 1.5 sec (Focant et al., 2004b; reproduced with written permission from the American Chemical Society).

Chemometric techniques have been widely used as an essential mathematical tool to obtain information from the wealth of data generated from GC × GC applications. In particular, GC × GC–MS applications are capable of generating trilinear data; viz., columns 1 and 2 retention times and the mass spectrum. Chemometric calibration methods such as trilinear decomposition (TLD) and parallel factor analysis (PARAFAC) were developed to find relationships in trilinear data. It has been shown that, by using these techniques, pure chromatographic profiles and mass spectra can both be attained from partially resolved compounds in GC × GC-ToF-MS applications (Sinha et al., 2004a). Specifically, mathematical deconvolution was effectively achieved on overlapping compounds in an environmental sample, by using PARAFAC initiated by TDL. It was demonstrated that this combination provided improved deconvolution in comparison with TDL alone, and required only a single data set for qualitative analysis.

In the 2004–2005 period, there was a significant interest towards the analysis of metabolites by using GC × GC-ToF-MS (Sinha et al., 2004b,c; Hope et al., 2005b; Shellie et al., 2005; Welthagen et al., 2005). The analytical method appeared to be particularly suited, because the determination of volatile metabolites in a specific biological situation is a large task. Welthagen et al. (2005) applied GC and GC × GC, both combined with ToF-MS, to the analysis of spleen extracts of obese (NZO strain) and lean (C57BL/6 strain) mice. A 10-fold higher sample amount was injected onto the single GC column to compensate for the lower sensitivity. GC-ToF enabled the detection of 538 peaks, whereas >1,220 were reported for the comprehensive GC experiment. The difference between the number of detected peaks would have been considerably more if equal sample amounts were injected. Despite the number of detected compounds, the authors reported that only 10% were positively identified through library matching and the use of LRIs (a 500-compound MS library was used). A further aspect, worthy of note, was the “so-called” analytical purity, which refers to the combination of the chromatographic and mass spectrometric capacities to isolate compounds. A compound is considered as “pure” if either one or both of these separative dimensions fully resolves it from the rest of the sample constituents. Purity approaches zero in the ideal case and (theoretically) infinitum in the problematic situations. In the GC × GC-ToF-MS experiment, the number of peaks with acceptable purity (<1 values were considered as sufficient) was increased by a factor of seven. The effectiveness of the three-dimensional method for differential metabolomic biomarker determination can be observed in Figure 12: four obese mice were compared to five lean controls. The first point that is immediately evident is the apparently good stability of the separation pattern. Moreover, metabolite variability between the samples becomes observable if 2D peak intensities are considered; for example, two sugar alcohols (indicated by yellow circles) are present in lower amounts in the obese mice spleen tissues. The Student's t-test was used to evaluate differences between the two sample-types: although the number of replicates were too limited to achieve actual statistical significance levels, the results indicated that the content of a series of compounds might indeed be different between the two groups.

Figure 12.

A direct comparison between the analyzed (lean) C57BL/6 female mouse spleen samples (left panel, samples 1–5) versus the (obese) NZO female mouse samples (right panel, samples 6–9). The encircled compounds were used for an example of statistical evaluation of biomarker efficiency (Welthagen et al., 2005; reproduced with written permission from Springer).

The contrast between ToF and quadrupole instrumentation continued in 2005. Adahchour et al. (2005a) described the principles, practicability, and potential of rapid-scanning qMS instrumentation in comprehensive GC. The performance of an ultimate generation qMS system (Shimadzu QP2010), which neared GC × GC requirements, was studied for qualitative and quantitative purposes. The detector was characterized by a scan speed of up to 10,000 amu/sec and could even reach the ultimate (GC × GC) goal of 50 spectra/sec, at a restricted mass range (95 amu). In the experiment, it was emphasized that the MS sampling rate should be sufficiently high to correctly reconstruct the chromatogram and to avoid any peak-skewing. First, the influence of the acquisition rate on peak-area precision was determined through standard solution analyses, carried out at 20, 25, 33, and 50 Hz. To generate the last two acquisition rates, the MS scan ranges were restricted to 195 and 100 amu, respectively. The authors found that a 33 Hz sampling rate produced at least 7 data points (only above-baseline points were counted) for peaks with a 200 msec or higher base width. This number of experimental data points was found to be the minimum necessary for reliable quantitation. If a wider mass range was required, then a 20 or 25 Hz sampling rate needed to be selected, but peak reconstruction was much less reliable. In the case of <200 msec peaks, a rate of 50 Hz was necessary, and limited even more the mass range. Second, peak skewing was evaluated by plotting the ratios of abundant ions in mass spectra relative to a series of compounds at acquisition rates of 33 and 20 Hz. It was found that considerable ion-abundance variation occurred at the lower rate, whereas ratios were essentially constant at the higher frequency. From the study, it appeared that a 30 Hz rate could be fast enough for GC × GC-MS applications, if the operational conditions were properly tuned. If a normal mass range is required, then the only option is to reduce the rate down to 25 or 20 Hz, inevitably reducing the mass spectral quality. Although the quadrupole instrument tested did provide a more than satisfactory performance, the final impression was that of the superiority of ToF systems for qualitative/quantitative GC × GC analysis.

Mondello et al. (2005) carried out a GC × GC-qMS experiment (a Shimadzu QP2010 was used) on a highly complex perfume sample; the MS instrument was operated at 20 Hz with a 40–400 amu mass range. Pure standard compounds, MS library matching, and one-dimensional linear retention index data (LRIs were used as a filter during library searches, with the elimination of matches outside a predefined LRI window) were used for reliable peak identification. The authors used an apolar 30 m × 0.25 mm I.D. primary column and a polar 1 m × 0.25 mm I.D. secondary capillary. It was found that the influence of the 1 m polar column, in terms of LRI variation with respect to reference values, was negligible for the apolar analytes; on the contrary, the more polar components, subject to more intense interactions, were characterized by a greater LRI variability. A total of 866 peaks were detected in the GC × GC experiment, but only 186 were detected in an unmodulated experiment; that difference revealed the extent of analyte co-elution. The shortcomings of GC-MS in such an application was fully demonstrated: a peak, unreliably identified as estragole and with a 72% spectral similarity, was reported as an example; in the comprehensive GC-MS experiment, the same analyte band was fully resolved into eight compounds, four of which were reliably identified (a 98% similarity was attained for estragole). As previously reported (Shellie & Marriott, 2002), the use of a wider-bore second dimension, under low-pressure outlet conditions, had beneficial effects on analyte separation. Furthermore, band broadening is increased and, hence, excessively high sampling rates are not required. In fact, the authors reported the attainment of at least three or four data points, sufficient for identification purposes, for the narrowest peaks. Again, however, it was the chromatographic operational conditions that needed to be tuned to meet the qMS requirements, and not the contrary.

In commonly used software, compound peak location information is only available after processing a complete data set, with the generation of a list of compound peaks and cross-matching them to MS library spectra. As discussed above, this automated procedure can take many hours, depending on the sample complexity. This route is not efficient, especially when only a low number of target analytes are of interest. To avoid such time expenditure, an algorithm defined as “DotMap” was developed (Hope et al., 2005a) to enable the rapid determination of peak locations in GC × GC-ToF-MS data. The algorithm uses a spectrum of interest (from a commercial or laboratory-constructed library) to define a specific compound in an entire GC × GC-ToF-MS data set. After searching for the target spectrum, a contour plot is generated to report the location of the most similar spectra. The best match is extracted and a conventional MS library search is performed to evaluate the effectiveness of the DotMap analysis. Whenever a DotMap-located compound overlaps with interferent signals, then PARAFRAC can be subsequently employed for the deconvolution of chromatographic/MS profiles and, hence, for quantitative analysis.

Polychlorinated n-alkane (PCA) mixtures are characterized by high complexity, because single constituents present different degrees of chlorination and chain lengths (short, medium, or long). Due to the extremely high world consumption, the presence of these toxic compounds in the environment and foods is unavoidable. Conventional GC-MS, widely employed in PCA analysis, generates typical “humped” chromatograms, which indicate the occurrence of extensive overlapping. Moreover, electron ionization is not advisable, because the considerable fragmentation, hinders identification. Consequently, the electron-capture negative ionization (ECNI) mode is preferred.

Comprehensive GC has been exploited for the separation and introduction of as many pure PCAs as possible to a recently developed ToF-MS system operated in the ECNI mode (ThermoElectron, Austin, TX, USA), (Korytár et al., 2005a) [the use of the same system, for the analysis of polybrominated diphenyl ethers, and of GC × GC-qMS, using the same ionization mode, were reported in the same year (Korytár et al., 2005b,c)]. The 100% methylpolysiloxane column used in the first dimension and the thermally stable 65% phenyl-methylpolysiloxane stationary phase employed in the second-dimension, generated highly structured chromatograms. With regards to ECNI-ToF-MS detection, methane was used as reagent gas, the mass range was from 50 to 700 Da, and the acquisition rate was 40 Hz. The chromatographic and ECNI mass spectral behavior of PC decanes with a Cl content of 55% (a standard mixture) is shown in Figure 13; the GC × GC-ToF-MS m/z 70–73 extracted ion chromatogram (EIC), which corresponds to the non-specific [Cl2]. and [HCl2] ions, is illustrated in Figure 13A. The group-type patterns are visible, with PC decanes characterized by the same number of Cl atoms (4–7) aligned along distinct bands; within each group, polarity increases when the substituents are distributed over the entire chain length (e.g., 2,5,6,9-C10Cl4 is more polar than 1,1,1,3-C10Cl4). This behavior was confirmed through the visualization of EICs for m/z 243–245 (C10Cl4 cluster), 276–283 (C10Cl5 cluster), 311–319 (C10Cl6 cluster), and 345–355 (C10Cl7 cluster), which correspond to the [M[BOND]Cl]. and [M[BOND]HCl] ion clusters. However, for penta-, hexa-, and hepta-Cl PCAs more than a single group was present on the space plane. For example, in Figure 13B, apart from the evident C10Cl6 band, additional small groups are located below and above the band. The upper compounds belong to the C10Cl5 class and are characterized by the mass spectral presence of the [M]. ion; the lower components belong to the C10Cl7 class, characterized by the [M[BOND]2HCl]. ion (mass spectra inserts are shown in Fig. 13B). To verify the location of specific congeners in the different bands, eight individual PC decanes were added to the standard mixture (their positions are indicated by black circles). Although many PC decanes followed the expected 2D chromatographic behavior, a series of congeners appeared to “slip” a chemical-class band: for example, 1,1,1,3,9,10-C10Cl6 is situated at the rear of the C10Cl7 pattern and, on the basis of its location, could be misidentified. The authors affirmed, though, that for technical mixtures the content of these “outliers” can be considered as negligible. PC decanes with a Cl content of 65% were also subjected to study (Fig. 13C); the C10Cl6–C10Cl9 clusters are now evident and widen as the degree of chlorination increases. This phenomenon is related to the higher number of congeners and diastereoisomers. Finally, three sub-groups of strongly retained second-dimension compounds, displayed in the insert in Figure 13C, remained unidentified.

Figure 13.

GC × GC–ECNI-TOF-MS extracted-ion chromatograms of polychlorinated decanes, (A) 55% (w/w) m/z 70–73, (B) 55% (w/w) m/z 311–319, (C) 65% (w/w) m/z 70–73. Inserts of (B) show its 3D presentation and averaged mass spectra of selected peaks. Inserts of (C) show its zoom-out visualization and mass spectrum of selected peak exploited (Korytár et al., 2005a; reproduced with written permission from Elsevier).

Mass spectrometry can be exploited as a separation method, resolving analyte molecular ions and fragments on the basis of their masses, and also as a spectrometric technique, where the fragmentation pattern acts as a molecule-specific fingerprint. However, MS experiments are generally carried out by using EI, in which the fingerprint aspect overshadows the other. As emphasized by J. Phillips, GC × GC and GC-MS are characterized by several common points. The similarity increases between the two methodologies if soft ionization techniques are employed; the lack of fragmentation and the formation of the sole molecular ions highlights the second-dimension separation capabilities of mass spectrometry. In 2006, conventional gas chromatography combined with soft photoionization (SPI) ToF-MS was employed to generate GC × MS fuel chromatograms, which were completely similar to their GC × GC-FID counterparts (Mitschke, Welthagen, & Zimmermann, 2006). Although the method is not quite within the marked boundaries of the present review, a brief description will be provided: a 30 m × 0.25 mm ID moderately polar column was used to introduce analyte bands to a laboratory-constructed laser-ionization ToF-MS. Selective resonance-enhanced multiphoton ionization was employed for aromatic compounds, whereas single-photon ionization was used for universal soft ionization. The generated GC-MS data were visualized as a 2D chromatogram, with elution times plotted against m/z values. The GC × SPI ToF-MS result for a diesel sample is shown in Figure 14: the first-dimension separation is polarity-based, and the second-dimension analysis resembles a volatility-based one. As can be seen in the figure, the chromatogram was modified to eliminate the temperature program influence on the GC separation to create two completely orthogonal dimensions. Consequently, because the boiling point component of the GC process was eliminated, the y-axis represents a 100% polarity-based separation. For example, the (isopolar) n-alkanes fall along a straight line parallel to the x-axis and are distinguished on a mass basis. The adjustment was achieved by doing a 10-point polynomial fit on the retention times of the n-alkanes and then shifting the retention times so that the n-alkanes aligned with one another.

Figure 14.

Comprehensive gas chromatography coupled to soft ionization mass spectrometry (GC × MS) representation, using SPI-mass spectrometry. A: With the GC × MS chromatogram before transformation and (B) the chromatogram after doing the “Kovats Index”-like transformation of the retention time axes (Mitschke, Welthagen, & Zimmermann, 2006; reproduced with written permission from the American Chemical Society).

As seen, in the 2005–2006 period, a certain number of GC × GC-MS (or related) contributions that reported the use of low- or no-fragmentation ionization methods, had come onto the scene. Briefly, and finally, in a further experiment, carried out by Kochman et al. (2006), flow-modulated GC × GC was combined with supersonic molecular beam (SMB) qMS. The scope of the investigation was: (1) to tune the GC separation to meet the qMS scanning-speed requirements; (2) to increase the reliability of peak assignment through the generation of molecular ions (common EI procedures often do not produce any molecular ions). Flow modulation had been introduced years earlier, as an alternative to cryogenic modulation (LaClair et al., 2004). The slightly modified version used by Kochman et al., generated peak widths in the 200–300 msec range, and, hence, were compatible with the MS instrument. MS ionization used SMB EI (defined as “cold EI”) to generate intense molecular ion peaks.


The wealth of undiscovered information relative to the composition of samples in all scientific fields is enormous. The awareness of this fact and that the complexity of real-world matrices often exceeds the analytical capabilities of conventional GC methods has increased recently. A further wide-spread opinion is that the development and employment of more powerful techniques to enable a deeper insight into the composition of natural and synthetic matrices is not only desirable but also a necessity. Comprehensive two-dimensional gas chromatography in combination with mass spectrometry fulfills these requirements: unprecedented selectivity (three separation dimensions, related to volatility, polarity, and mass), high sensitivity (through band compression), enhanced separation power, and increased speed (comparable to ultra-fast GC experiments, if the number of peaks resolved per unit of time is considered).

A comprehensive perspective on the evolution of GC × GC-MS and on its analytical potential in different research fields has been provided here. The method, however, is of relatively recent introduction, and is still far from being well-established. A series of reasons behind its limited diffusion, related both to the chromatographic and mass spectrometric steps, has been reported and discussed in the text. In particular, the main hindrances probably derive from a natural scepticism towards new methodologies and the high instrumental costs. It may be anticipated that, if further improvement occurs, wherever necessary, GC × GC-MS will undergo a gradual and constant expansion in the following years.

Biographical Information

Prof. Dr. Luigi Mondello is Full Professor of Analytical Chemistry at the Dipartimento Farmaco-chimico of the University of Messina, Italy. Prof. Mondello teaches the course of Analytical Chemistry at the School of Pharmacy and School of Medicine of the University of Messina, and the course of Food Chemistry at the “Campus Biomedico” in Rome. He is the author of 180 scientific papers, 13 book chapters, 5 reviews, co-editor of a book on Multidimensional Chromatography (John Wiley & Sons), and has been chairman and invited lecturer in national and international meetings. His research interests include single column chromatography techniques (HRGC, HPLC, HRGC-MS, HPLC-MS, OPLC) and the development of coupled techniques such as LC-GC-MS, GC-GC, GC × GC, LC × LC, LC × GC and their applications in the study of natural complex matrices.

Biographical Information

Dr. Peter Quinto Tranchida possesses a degree in Pharmaceutical Chemistry and received a Ph.D. in Food Chemistry and Safety in 2006. At present, he occupies a position as Assistant Professor in the Food Analysis Division at the School of Pharmacy of the University of Messina. His main interests regard the development and application of classical multidimensional and comprehensive chromatographic techniques. He is author of 35 scientific papers and has been lecturer in national and international congresses.

Biographical Information

Prof. Dr. Paola Dugo is an Associate Professor of Food Chemistry at the School of Chemistry of the University of Messina since 2000. Her research interests include the study of the composition of citrus essential oils and the study of components with a possible biological activity contained in natural matrices. She is actively engaged in the development of comprehensive multidimensional GC and LC methods with mass spectrometric detection. She is the author of approximately 130 scientific papers and has been lecturer in national and international congresses and symposiums.

Biographical Information

Prof. Dr. Giovanni Dugo is Full Professor of Food Chemistry at the School of Chemistry of the University of Messina since 1986. He is author of over 260 papers and co-editor of a book (Citrus, Francis and Taylor) covering all aspects related to Citrus science. His main research interests are directed towards the development and application of innovative chromatographic methods (both hyphenated and non) to the analysis of complex natural matrices. He was one of the founders of the Food Chemistry Group of the Italian Chemical Society and chaired this group for a long time. He has been in the Scientific Committee and Chairman in several national and international meetings.