Authors: Saba Aslani, Daniel W. Armstrong
Categories: Article, Structure elucidation, NMR detection, MRR detection, FTIR detection, VUV detection
Source: Journal of chromatography. A
Authors: Saba Aslani, Daniel W. Armstrong
Gas chromatography has always been a simple and widely used technique for the separation of volatile compounds and their quantitation. However, the common detectors used with this technique are mostly universal and do not provide any specific qualitative information. There have been some attempts to combine the separation power of GC with the qualitative capabilities of “high-information” spectroscopic techniques including infrared spectroscopy, nuclear magnetic resonance spectroscopy, molecular rotational resonance spectroscopy, and vacuum ultraviolet spectroscopy. Some of these hyphenations have proven to be quite successful while some were less so. The history of such attempts up to the most recent studies in this area are discussed. Most recently, the hyphenation of GC with molecular rotational resonance spectroscopy which provides promising results and is a newly developed technique is reviewed and compared to previous high-information spectroscopic detection approaches. The history, description and features of each method along with their applications and challenges are discussed.
Gas chromatography (GC) is a widely used technique for analyzing volatile and semi-volatile compounds. Detectors that are the most commonly used with GC, like the flame ionization detector (FID), thermal conductivity detector (TCD), nitrogen-phosphorus detector (NPD), electron capture detector (ECD), etc. provide little to no analyte structural information. Therefore, even though such detectors are widely used, standard samples are always needed to identify unknown compounds present in mixtures. When analyzing complex mixtures, the ability of a detector to provide qualitative information gains importance, since the substances may be unknown and reference standards may not be available. Mass spectrometry (MS) hyphenated with GC was a groundbreaking development, given the ability of MS to provide structural information and high sensitivity [1]. However, MS techniques are not adequate in differentiating many isobaric compounds, structural isomers, and enantiomers. Therefore, selective spectroscopic techniques have been investigated as GC detectors to overcome some of the MS detection limitations.
Spectroscopy involves studying the interaction of electromagnetic radiation with matter [2]. Some spectroscopic techniques are very powerful in providing compound structural information, but they usually are used to confirm the structures of single purified analytes rather than mixtures of unknowns. Combining the qualitative power of spectroscopic techniques with the separation power of GC can provide a powerful tool for qualitative and quantitative analysis of unknown compounds in mixtures along with shorter analysis times. Such techniques could be complementary to GC-MS or provide viable substitutes, but these topics are not covered in this review. For those interested in this topic, see representative references [3–10].There have been multiple attempts to hyphenate GC with high information spectroscopic techniques, including infrared spectroscopy, nuclear magnetic resonance spectroscopy, molecular rotational resonance spectroscopy, and vacuum ultraviolet spectroscopy. The regions of electromagnetic radiation that have been studied as possible GC detectors are infrared, ultraviolet, vacuum ultraviolet, microwave, and radio wave regions. However, the amount of structural information varies significantly with each spectroscopic approach as does the practicality of hyphenating each with GC. One of the main challenges of hyphenating these spectroscopic techniques with GC, that immensely affects sensitivity, is developing the right interface that can deliver the GC effluent to the detector efficiently and without diffusional dilution and/or loss of analytes. A heated, low volume transfer line is usually of importance to prevent chromatographic extra-column band broadening. Also, the size of the detection cell is an important factor in VUV detectors and light pipe interface for Fourier transform IR detectors. The reason is that detection volume affects band broadening, residence time, sample dilution/concentration, and therefore sensitivity of the absorbance measurements which also are affected by flow cell pathlength [11, 12]. Another factor is the analysis time. The hyphenated detector needs to be fast enough to acquire data in the short time that analytes are present in the detector, without the need of stopping the flow. Indeed, there often are “trade-offs” between sensitivity, chromatographic efficiency, analysis time (speed), and the amount and type of structural information that can be obtained with each approach. This review will compare, in detail, different spectroscopic detectors hyphenated with gas chromatography [13, 14].
Nuclear magnetic resonance spectroscopy (NMR spectroscopy) is a very powerful analytical method and provides valuable information on the structure of molecules. This technique exploits the magnetic properties of certain atomic nuclei to determine the structure of compounds in which they reside [15]. NMR active nuclei are those possessing a property called ‘spin’, whereby a charged nucleus spins about an axis and generates its own magnetic dipole moment. For a molecule to be NMR-active the nuclear spin of the nucleus (I), must not be zero. For example, ^1^H(I=1/2), ^13^C(I=1/2), ^19^F(I=1/2), ^31^P(I=1/2), ^2^H(I=1), etc. The magnetic field resulting from the environment of atoms around a nucleus in a molecule affects its resonance frequency, thereby providing details pertaining to the electronic structure of a molecule and its individual functional groups [15]. Placing an atom in an external magnetic field induces circulations in the electron cloud surrounding the nucleus which results in electrons generating their own magnetic fields. The induced magnetic field can reduce or increase the strength of the external magnetic field around the nucleus which results in shielding or deshielding of the nucleus affecting its chemical shift. Chemical shifts are measured with reference to the absorptions of protons of a reference compound, which is usually tetramethyl silane (TMS) [16]. NMR is usually performed on purified samples in solution at concentrations of 0.4 to 0.5 mM [17]. The solvent cannot contain atoms with “competing” chemical shifts to the analyte. One possible great advantage of GC-NMR is that there is no solvent. However, the analyte amounts are much lower and residence times are short.
Hyphenation of NMR with GC provides a way to analyze mixtures of organic compounds more effectively since synthetic organic compounds and/or natural compounds often are complex. There have been multiple attempts to combine separation techniques with ^1^H-NMR to take advantage of its power in structure elucidation [18–22]. In contrast to liquids, gases have rarely been investigated by NMR [23–26]. The reason is the experimental difficulties in handling gasses and the low S/N obtained for NMR signals for gaseous samples at atmospheric pressure. However, with the advent of higher magnetic fields and solenoidal microprobes, these limitations could be overcome [11, 12]. The first attempts of combining GC with NMR spectroscopy obtained the NMR spectra of trapped GC fractions after condensing them [25–30]. In these cases the column flow had to be stopped prior to measuring the NMR spectra of the fraction containing each analyte [25–30]. Typically, samples were condensed and then transferred to another vessel for the NMR measurements [26–30]. These methods required sufficient time to collect enough sample for analysis. Using direct trapping procedures requires less sample for obtaining NMR spectra of samples (100–400 μg instead of milligrams) [25]. The most recent trapping method was performed by Nojima et al. using a preparative GC system to purify nanogram amounts of geranyl acetate for NMR analysis [31]. However, this was an off-line hyphenation and did not provide acquiring the spectra in real time of chromatographic peak elution [31]. In 1981 Buddrus, et al. used a system with a very simple design to combine GC and NMR spectroscopy without interrupting the flow [32] (see Fig. 1.A). This system did not have a heated transfer line and therefore was limited to samples with boiling points of ≤ 60 °C. The compounds analyzed with this instrumental setup were diethyl ether, 2,2-dimethylbutane, and cyclopentene [32]. A packed column GC stationary phase was used. The NMR data were obtained with a pulsed Fourier transform NMR spectrometer (JEOL FX-100). The magnetic field strength of the instrument was much lower than those of instruments used today, which contributed to the low sensitivity of this setup [32].
The free induction decay (FID) was stored in a computer of 8 K words (one word = 32 bits); only the first eighth of the words of the FID spectrum was taken for Fourier transformation, the content of the remaining words being of minor importance and substituted by zero (‘zero-filling’ technique) [32]. The pulse repetition time (the total time between the start of acquisition of the first FID and the start of acquisition of the second FID) was 1s. When only the first eighth of the words of the FID spectrum are used for FID storage, the acquisition time can be reduced to 0.5 s. Considering the pulse repetition time as 1 s, as many as 130 or even 160 accumulations could be performed. More repetitions result in higher S/N which is favorable for increasing the sensitivity [32]. In this design, the line widths are 2 Hz and it is partially since the NMR tube is not spinning and partially due to the flow rate of 0.57 cm/s (corresponding to a flow rate of 5.2 ml/min in a tube with 4.4 mm i.d.). With flow rates more than 1 cm/s the NMR lines become broader due to shorter residence times [33]. Even though this represented a successful coupling of a GC with an NMR spectrometer, it was not clear if this set-up provided online detection of the compounds eluting from the GC or the flow had to be stopped and analysis of each sample had to be performed separately by NMR, before the flow was turned back on. The second attempt to hyphenate of GC with NMR was performed by Herzog and Buddrus in 1984 (see Fig. 1.B) [34]. In this design, a mixture of α-pinene, β-carene, γ-terpinene, and fenchone (10 μL of each) was injected on a packed GC column (4 mm i.d.), coated with Carbowax 4000 and heated from 150 to 200 °C. The flow of carrier gas (helium) was reduced from 40 ml min^−1^ (van Deemter optimum) to 10 ml min^−1^ to reduce NMR line broadening [33]. The improvement in this instrument included a heated transfer system, thereby allowing compounds of higher molecular weights (up to 150 Da) to be analyzed. However, the sensitivity of this instrument was lower due to the higher temperature of the NMR probe (450 K) which resulted in a 2.1-fold decrease in S/N compared to room temperature measurements. Continuous recording of a “^1^H NMR- chromatogram” was still not possible as the chromatogram obtained in this study was recorded using a thermal conductivity detector [34]. In recent years with improvements in NMR sensitivity due to higher magnetic fields and utilization of solenoidal microprobes, new combinations of GC-NMR tend to have improved sensitivities. There are two geometrical types of continuous-flow NMR probes available, which were described in detail in several papers [35–37]. One was a saddle-type coil probe, which is commonly used in conventional NMR probes [35]. Another approach used a solenoidal-type probe, which was constructed by directly wrapping the radiofrequency (r.f.) transmitter coil around a capillary column, immersing the coil into a susceptibility-matching fluid for improved field homogeneity, and placing it transverse to the magnetic field [36, 37]. Using the aforementioned improvements, the first online GC-NMR spectra could be recorded by hyphenating capillary GC to micro coil NMR [11]. A custom-built solenoidal NMR microprobe was used with an active volume of 2 μL. The NMR detection of several compounds at 400 MHz was accomplished, first in a mixture and then with full coupling to capillary GC to identify the compounds separately. The injected amounts of each analyte were 3.3 μL. A 50 m SE-54 coated capillary GC was used. High purity helium was used at a flow rate of 0.72 mL min^−1^, with splitless injection and a column temperature of 60 °C [11]. It can be seen that the flow rate is lower than that normally used for GC analysis today. The flow rate was a compromise between obtaining reasonable chromatographic resolution and having a sufficient S/N for NMR detection. Even through the flow rate was not at the van-Deemter optimum, higher flow rates could result in to too short residence times in the detection cell [11]. This limited residence time of the nuclei in the flow cell reduced the spin- lattice (T1) and the spin-spin (T2) relaxation times and produced very broad NMR linewidths, resulting in lower signal intensities [11]. Also, the column length was somewhat longer than usual, but necessary for achieving a separation for the mixture of diethyl ether, tetrahydrofuran, acetone, and dichloromethane. Fig. 2 is a simplified schematic illustrating this hyphenated system. A splitless injection of 10 μL of diethyl ether, dichloromethane, and tetrahydrofuran (3.3 μL each) was needed to obtain the pseudo 2D plot depicted in Fig. 3. In this system the gas flow did not stop, and FIDs were accumulated as the solution was going through the flow cell. For more experimental details see Ref. [11]. The same GC-NMR instrument was used with a shorter transfer line (2m instead of 3m) to analyze stereoisomers of dimethylcyclohexane, hexene, and pentene which are usually difficult to separate. Contour plots are shown in Fig. 4. It can be observed that the combination of highly selective GC stationary phases and NMR detection, with the help of a spectral database, can achieve identification of relatively simple geometrical isomers in a complex mixture. Also stopped-flow measurements of very low sample amounts, which are often essential for analysis of compounds at very low concentrations, can be performed. The sample amount for a successful online detection is ∼100–300 μg at 400 MHz, but instrumental improvements as well as stronger and shielded magnets are anticipated to decrease this amount.
NMR spectroscopy is widely used in organic chemistry to confirm the structure and purity of synthesized compounds. NMR spectroscopy can identify structural patterns, isomers, and isotopes, however identifying enantiomers is challenging. Also, since NMR is inherently insensitive, the residence time of analytes has to be high and therefore the GC column flow rates are usually well below van Deemter minima which results in a loss of separation power [11]. Further, real time GC chromatograms have not yet been shown to be feasible with hyphenated NMR. Moreover, NMR instruments take considerable space, are expensive and need a high level of expertise to operate and maintain [38]. That is why the probability of a commercial GC-NMR is very low and there have not been as many attempts to do this type of hyphenation.
Infrared (IR) spectroscopy is the measurement of the intensity of absorption of infrared light by a sample versus wavelength [39]. For molecules to absorb IR energy, the vibrational motions must change the dipole moment of the molecule [40]. IR spectra show unique vibrational frequencies that correspond to structural characteristics and functional groups of molecules. The combined spectroscopic features of all intrinsic functional groups can provide a unique molecular fingerprint to characterize and sometimes confirm the identity of a sample [41]. Traditional dispersive IR instruments cannot be used as GC detectors since acquiring a single IR spectrum with these can take several minutes [42]. However, FTIR spectrometers have three major advantages in their performance over dispersive infrared spectrometers. These are : the multiplex or Fellgett’s advantage, the throughput or Jacquinot’s advantage and the wavelength accuracy or Connes’ advantage [14]. These advantages are the reasons for the switch to modern Fourier transform IR (FTIR) techniques over the last few decades, and they make FTIR spectrometers a more suitable choice as a GC detector.
The fundamentals of FTIR detectors that are coupled to GC, are the same as those of single-beam FTIR spectrometers. There is a two-beam interferometer that produces a modulated IR signal, aka the interferogram, and it is based on the Michelson interferometer. The interferogram shows the change in intensity of the beam versus time and can be measured as a function of the light path difference by an IR detector [43, 44]. Usually, rapid-scanning interferometers operate by signal-averaging successive scans. In order for the signal averaging process to be precise, each interferogram must be sampled at exactly the same retardation delay for every scan. Therefore, the movement of the mirror that controls the optical path should be controlled by use of the interference pattern of the monochromatic light of a He-Ne laser. This laser acts as a “built-in” way to calibrate the wavenumbers with a very high precision (~ 0.01 cm^−1^). This also is known as the Connes’ advantage [43, 44]. Available commercial GC-FTIR instruments have detection systems that are analogous to conventional FTIR spectrometers (Fig. 5.A). After the IR beam exits the interferometer, it is focused on the sample and then reflected by a movable mirror to the port of connection to the GC effluent [14, 44].
After collecting the interferograms, the next step is to transform the set of collected interferograms into a reconstructed IR chromatogram. There are two widely used methods to construct chromatograms directly from the interferometric data [44]. The first method relies on the fast Fourier transform (FFT) algorithm. In this method, a reference spectrum is stored with the same spectroscopic resolution before the GC run. Then the spectrum is plotted in real time versus the reference spectrum. In order to produce the spectral window chromatograms, the IR absorbance is calculated in discrete spectral regions. These spectral regions generally correspond to characteristic absorptions of selected functional groups [44–46]. Modern computers allow the production of “windows” chromatograms that refer to multiple separate spectral regions. After averaging the signals in the spectra, they can be computed in real time to obtain the interferograms [45]. There is another group of chromatograms called the functional group chromatograms (FGC). These chromatograms identify the parts in GC eluents that have a specific functional group and can be used to identify certain groups of compounds. This is somewhat analogous to single-ion monitoring (SIM) in GC-MS [45]. Accurate determination of discrete wavenumbers of each analyte makes the optimization of the S/N of the FGCs easier. However, in the case of unknown samples for which the absorption spectra are unidentified, it is not possible to know the wavenumbers that are specific to each analyte [45]. The second method for chromatogram reconstruction is the Gram-Schmidt (GS) or vector orthogonalization process which was developed by De Haseth and Isenhour [47]. This method directly reconstructs gas chromatograms from single scan GC/IR interferograms [47]. This process describes the inherent instrumental and collection characteristics by collecting the background interferograms (only the carrier gas) as a basis set. In the orthogonalization process which occurs after collecting the basis set, all background information is removed from the collected sample interferograms and differences between the background and any infrared active species that have been eluted is measured which eventually results in a gas chromatogram [47]. The reconstructed gas chromatograms, being derived directly from the single scan interferometric data, precisely indicate which interferograms best represents sample spectra [47]. Due to advantages of the GS method over FFT on selectivity, S/N ratio and data acquisition and handling speed, the GS method remains the standard method to reconstruct total IR chromatograms [44].
The function of the infrared radiation detector is to convert the light intensity received into an electrical signal. The two most commonly used detectors in FTIR instruments are the deuterated triglycine sulfate (DTGS) detectors and the mercury cadmium telluride (MCT) detectors [48]. A change in the intensity of IR radiation striking a DTGS detector will cause a proportional change in temperature. The change in temperature will then lead to a change in the dielectric constant of the detector and its capacitance [14]. The MCT detector is a semiconductor. When the chemical components present in MCT absorb IR radiation, electrons move from the valence band to the conduction band and generate an electrical current proportional to the IR intensity. MCT detectors are more sensitive and faster than DTGS detectors and hence provide spectra with higher S/N [14].
In the matrix isolation (MI) technique the compounds are trapped and frozen using low temperatures and the IR measurements are performed off-line along with extensive signal averaging which increases the S/N significantly by acquiring the data for a longer time and increasing the interaction of samples with the modulated IR beam [49]. Condensation of molecules takes place under vacuum and at very low temperatures (cryogenic temperatures, < 20 K) on a gold-plated disk which is an IR-reflecting surface. The carrier gas is usually 98% helium with 2% argon. The helium in the carrier gas is not condensable at the operating temperature. Therefore, only the argon and sample molecules are condensed on the surface of the cryogenic disk [49]. Under these conditions, the rotational motion of molecules is absent, and the individual molecules do not interact with each other. As a result, IR absorption bands sharpen significantly (frequently <0.5 FWHM cm^−1^ for small molecules) and because there is no molecular interaction or rotation present, this results in almost pure IR vibrational spectra [49]. This feature makes it feasible to distinguish between closely related isomers which is not possible with the light pipe interface (see Section 3.3.3) or even traditional condensed phase spectra [49, 50]. The early work on combining GC to MI-FTIR suffered from insufficient focusing of the GC eluate on the cryogenic surface. Also, it did not show any significant gain in sensitivity since the collecting area on the cryogenic surface was large (~3mm^2^ ) resulting in lower absorbance of IR radiation [49, 51–53]. An optimized version that claimed sensitivity (in terms of detection limit) in the 100 pg range resulted in the commercialized version of GC-MI-FTIR, which was known as Cryolect^™^. This interface was developed by Cryolect^™^ Scientific Corp. and was provided by Mattson Instruments, but is no longer available due to its complexity and maintenance issues [44, 49, 54].
Fig. 5.B shows an enlarged view of the sample collection surface. The surface of the cryodisk is mirrored and maintained at temperature of 11 K and in a vacuum. The low temperature is provided by a closed-cycle helium refrigerator, which does not require liquid helium. The GC effluent is directed onto the surface of the cryodisk. The analytes coming from the GC column are retained in sequence in the track of frozen argon on the surface of the rotating cryodisk. The diameter of each collected compound is approximately 0.3 mm leading to high level of IR spectral absorbance per nanogram of sample compared to the bigger 3 mm diameter of sample matrix in the earlier version of the GC-MI-IR introduced in 1971 [49, 51]. After trapping of the samples, IR beams will be reflected from the surface multiple times and interact with each sample when it is under the focus of the IR beam. Multiple scan signal-averaging could be performed to improve the S/N. In order to clean the disk surface after analysis, the surface temperature is increased to evaporate the deposited compounds. The sensitivity of this technique is 10 to 100 times higher than the light pipe GC-FTIR (see Section 3.3.3) and very close to the current GC-MS sensitivity (in low nanogram range) [50]. The biggest disadvantage of the GC-MI-FTIR technique was its complexity and the precision required to build the interface which resulted in a higher cost. Also, considerable experience and knowledge was needed to maintain and operate this instrument [54]. It should be mentioned that even though the high resolution spectra obtained could provide much structural information and excellent sensitivity, these condensed phase spectra are different from gas phase spectra and no commercial compound library was/is available for this method [44, 49].
Fig. 5.C is a schematic of the DD interface. The sample slide consists of a ZnSe window which is transparent to IR beams. The ZnSe window is held under a vacuum of 10^−5^ Torr, maintained by a turbomolecular pump. The cooling temperature (97 K) here is somewhat higher than that of the MI interface. It is maintained by placing flexible thermal transfer lines between a moving block connected to the window and a liquid nitrogen reservoir [55]. The space between the orifice connected to the transfer line and the focus of IR beam on the surface of the window is 50 μm [56]. The width of the frozen spots of analytes on the moving window is around 100 μm. This provides an increase in absorbance per unit weight (~10 fold) compared to the MI technique in which the width of the sample spot is 300 μm wide [55]. By continuous movement of the sample slide each spot is placed in the focus of the IR beam and interacts with it. Since the interaction takes place after a short time delay, this device is not a real-time detector. Next, the IR beam will transmit through the ZnSe window onto an MCT detector. Also, it is possible to control the movement of the disk in order to increase the interaction time and have more signal averaging processes to increase the S/N and bring the sensitivity of this instrument (in terms of detection limit) to low sub-nanogram levels or even a few tens of picograms [56]. The spectra that are collected using this method can be searched against the compound libraries obtained by KBr disks [55, 56]. The commercial GC-DD-FTIR is known as DiscovIR-GC^®^. This instrument provides unequivocal differentiation between isomers and other molecules with similar or identical MS patterns. Identification of these compounds can be difficult or impossible with mass spectrometry alone [57].
Fig. 5.D shows a typical arrangement for a GC-FTIR system equipped with a LP interface. The effluent from the GC reaches a heated gas cell after passing through a heated transfer line. After the IR beam is modulated, it is focused on the entrance of the LP to interact with the analyte and the resulting beam will be reflected onto the MCT detector. The effluent can be redirected to a conventional GC detector i.e. flame ionization detector (FID) using a heated transfer line after exiting the LP, as IR is a nondestructive detector [44]. The LP interface consist of a glass tube that is gold coated which will result in multiple reflections of the IR beam and more interaction with the sample before reaching the detector. This enhances the optical performance of the cell [44]. The higher detector temperature was shown to negatively affect the signal intensity of the GC-FTIR [58]. Since the volume of the flow cell and transfer lines will result in band broadening, optimizing the size and geometry of the LP is important. [14, 59]. The “practical” limit of detection/identification with the LP interface is around tens of nanograms of sample if the compounds absorb the IR radiation, but amounts are higher if compounds are poor IR absorbers [60]. There have been no significant improvements in the optics of this instrument since the development of LP interfaces were adapted to narrow-bore fused-silica WCOT GC columns [45, 50]. The commercial GC-FTIR that utilizes the LP interface is the NICOLET^™^ FTIR spectrometer. This LP interface offers high-performance and an easy-to-use approach to GC-IR separation and identification. The LP consists of gold-coated glass flow cell with gold-coated alloy end caps; LP seals are made of a high-temperature polymer capable of operating continuously at 325°C. Also, the low dead volume design of the LP (15 cm length, 1.0 mm ID) produces the highest absorbance with minimal peak broadening [61]. Capillary columns can be inserted directly into the LP. This eliminates component degradation and peak broadening in the transfer line. Also, this instrument is able to collect, co-add up to 7 scans at 8 cm^−1^ spectral resolution, process, display, and store to hard disk; IR spectra and IR chromatograms in real-time, i.e., in less than 1 second [61].
Given the availability of commercial GC-FTIR instruments, it is not surprising that there have been several reports on applications that have utilized the LP or DD interfaces. The LP interface is the most widely used due to its simplicity and lower price. There have been a wide range of applications in drug analysis [62–67], biological studies and pheromone analysis [68–71], essential oils, fragrances, and fatty acids [72–77]. Investigations on the interconversion of oximes [78, 79], Li-ion and Na-ion batteries [80, 81], and the analysis of hazardous materials [82] have been reported. Although the direct deposition approach is more complex and needs a higher level of expertise, it has been widely used due to its superior resolution and sensitivity compared to the LP interface.
Fig. 6 shows a comparison of the FTIR spectra of 3-hydroxyhexanal obtained via a LP detector and a DD interface. The OH bond stretching band appears as a very broad band between 3500 and 3000 cm^−1^ (Fig. 6.b) since in the solid phase spectra obtained with the DD interface intermolecular forces such as hydrogen bonding are present, as they are with KBr disks or in neat films. Also, note that the amount of sample injected using DD was 20 times less than that for the LP. The DD interface was used in studies including drug analysis, and forensic toxicology [83–88], biological studies and pheromone analysis [89–92], volatile organic compounds, essential oils, and fatty acid analysis [93–95].
Fig. 7 shows the advantage of using FTIR molecular simulations in industrial synthesis over MS or NMR when library data or injection of authentic standards are not available [65]. The GC is needed for separation of components and the FTIR detector provides spectra of the pure compounds. The IR spectra of the isomers have sufficient differences because of the different symmetries and geometries of E and Z isomers (i.e. C2v and C2h point groups, respectively). The comparison of match scores of simulated data with the experimental data for each isomer shows that the major component is the Z isomer, with the minor one being the E isomer. However, MS cannot differentiate between Z and E isomers of 2,3-bis (thiosulfanyl)-but-2-enedinitrile. Also, even though NMR chemical shifts are different for the two isomers, because of the difference in their abundance in the analyzed sample (99.85:0.15 relative abundance), it cannot be assured that the minor component is detectable by NMR, considering the limited sensitivity of this method [65].
FTIR can be a very good option to combine with GC as it provides good sensitivity, fast analysis, and some degrees of structural information. Moreover, both vapor phase and solid-state compound libraries are available which make unknown identification possible. Also, it should be taken into consideration that solid state and vapor phase spectra are different, and the libraries are specific to the techniques used. A problem arises when the unknown compounds are not present in the libraries. In this case the information obtained from FTIR spectroscopy is limited to the presence of specific functional groups. Also, the structural information obtainable from IR is limited in terms of spatial arrangement, i.e., enantiomers cannot be differentiated [44]. All said, the availability of the Nicolet series and DiscovIR-GC instruments plays an important role in accessibility and success of this hyphenated GC method.
In the past, high intensity sources were needed to measure VUV absorption spectra to overcome significant background absorption, but these sources were restricted to synchrotron facilities. The reason for high background absorption is that all molecules absorb strongly in the VUV range (115–185 nm). Virtually, all transitions are excited with photons of this energy, especially σ → σ* that cannot be monitored in ultraviolet-visible spectroscopy and short wavelength π → π* transitions. There were some early efforts at combining of UV and VUV absorption detection with GC [96–98]. The far UV detector used by Middleditch, et al. could only probe a single wavelength absorption at 122 nm (10.2 eV) [96]. The issue is that this will not provide any qualitative information even though it is feasible for detection of some chemical species like alkanes. Lagesson et al. reported a broadband UV absorption detector capable of quantitative and qualitative analysis [98]. They were able to record VUV spectra from GC peaks in the 168–330 nm range. They also were able to detect molecules containing heteroatoms and conjugated systems based on characteristic gas phase absorption spectra. The limitation of this setup was a lack of universal detection and limited transmittance of light in the 168–180 nm range [98]. Broadband vacuum ultraviolet (VUV) absorption spectroscopy has gained attention in the last few years due to the recent availability of commercial instrumentation, i.e., the VGA-100 shown in Fig. 8 [13]. The new commercial version of this instrument (VGA-101) has improved spectral libraries and data acquisition features [99]. After eluting from the gas chromatograph, analytes enter a heated transfer line (~ 300 °C) consisting of a length of uncoated deactivated capillary. A make-up flow, the same as the carrier gas, is introduced at the end of transfer line in order to control the residence time of the sample zone in the flow cell (80 μL volume, 10 cm path length). Incorporation of specially coated reflective optics and a back-thinned charged coupled device (CCD) light path monitor, enables the instrument to collect high quality VUV (and some UV) absorption data between 115−240 nm [13]. A dark noise reading is taken for background subtraction at the beginning of each run. It was shown that optimizing the residence time of the analyte in the flow cell provides some advantages in terms of signal averaging. Absorption data are sent to the data station for processing. Data acquisition rates for the GC-VU can be set as high as 100 Hz which means that the measurements are fast enough to be compatible as a GC detector. After residence in the flow cell, the analyte zone is swept out through the exit vent. [13]. Spectra taken in the gas phase do not suffer from broadening which is commonly the issue observed for UV absorption spectra in solution. Also, the VUV spectra for closely related species are somewhat different. Data acquisition and analysis software, which are part of the GC- VUV package, have the capabilities to compare and deconvolve closely related spectra [13]. It is difficult to differentiate some compounds, such as structural isomers based on electron ionization mass spectra. However, at short wavelengths and in the gas phase, these isomeric compounds are distinguishable (see Fig. 9). Moreover, VUV and UV spectra can be calculated theoretically, using a software like Gaussian09, which implements semiempirical approaches and density functional theory calculations to sophisticated post-Hartree−Fock methods [100–102]. It is true that the spectrum for each compound has characteristic features, but the challenge is that the calculated spectra do not match the experimental spectra. This problem arises since the VUV spectrometer uses a broadband source to excite electronic transitions, which also will excite the underlying vibronic modes and this results in deviations from theoretical calculations, which only consider the electronic transitions. To address this problem, correction factors had to be developed, to match the calculated spectra with the experimental ones. However, more studies are needed to optimize these correction factors since their accuracy decreases when the molecular weight of the analytes increases [13, 103, 104].
Quantitative analysis by VUV spectroscopy follows Beer-Lambert Law principles. The VUV detector provides a level of performance on par with or exceeding those of more established detectors especially for quantitation of water which has a limit of detection of 246 ng in the 130−175 nm range [13].
A new time interval deconvolution (TID) algorithm has been created to provide a more advanced deconvolution of chromatograms of complex mixtures in an automated fashion [105]. This method works by dividing the GC-VUV chromatographic data into time interval sections and analyzing each section using a general linear least squares regression procedure to get responses corresponding to each analyte (or groups of analytes depending on what setting is used). Availability of assigned retention indexes and relative response factors for each compound/class of interest are necessary in order to use the TID reference spectral libraries [106]. Walsh et al. used the automated approach to TID for the first time to provide bulk characterization and speciation of finished gasoline samples (Fig. 10) [105]. The advantage of performing TID is faster separations and simplified analysis of data. Also, the deconvolution can be used to differentiate several coeluting compounds in one chromatogram [107].
The ability to deconvolute co-eluting peaks is an important ability of this detector. Absorption is additive; therefore, in the VUV detector overlapping peaks give a spectrum that corresponds to the sum of absorbances of the compounds present [108]. In order to determine individual contribution of each compound, the classical least square algorithm (CLS) can be used as shown below if the absolute molar absorptivity values across wavelength ranges are known for each analyte. Also, the reference spectra of coeluting compounds should be sufficiently distinct from each other [107, 108]: (1)A(λ)=b∑j=1Nεj(λ)c
λ: wavelength
ε is the molar absorptivity coefficient, which is directly related to the absorption cross-section by the Avogadro number
By using the individual reference absorbance spectra (Aref) to determine the relative cross-section of each analyte, Eqn.1 can be simplified to Eqn.2 [107]: (2)A(λ)=b∑j=1Nfj×Ajref(λ)=(f1A1ref+f2A2ref+⋯+fnAnref)
fj: the fitting coefficients (representing the amount of analyte j contributing to the total absorbance value relative to the amount of analyte contributing to the reference absorbance)
Ajref(λ): reference spectra for the co- eluting components at each increment in wavelength.
In order to deconvolute each absorbance spectrum, each absorbance scan is fit to give a distinct set of optimized fj values. By applying Eqn.2 to a chromatogram, variation of the individual fj values with time produces n new deconvoluted chromatograms each corresponding to a certain analyte [107]. An example of deconvolution is given in Fig. 11. The co-eluting peak consisting of p,p’-DDE and dieldrin pesticides was deconvoluted into two individual peaks [108]. Also, it should be mentioned that the additive property of the absorbance might fail when concentrations of the overlapping species is high enough that they would interact significantly [107, 109]. There are more powerful deconvolution techniques that can be considered good alternatives for CLS to be used in VUV deconvolution software, including multivariate curve resolution alternating least square (MCR-ALS) analysis, and parallel factor (PARAFAC) analysis [110–112].
With all of the above-mentioned features GC-VUV can be used in a wide range of applications including oil and gas analysis [9, 13, 105, 113–122], food, flavors and fragrances analysis [6, 7, 10, 123–125], biological studies and drug analysis [8, 126–137], and environmental analysis [108, 138–143]. With improvements in spectral libraries and with fast analysis times, the VUV detector also has been coupled with multi-dimensional GC systems [144–149]. This combination can be advantageous since GC×GC systems provide a higher peak capacity as well as improvements in minimum detectable amounts compared to conventional one-dimensional GC systems [150]. Also, GC×GC systems introduce improved selectivity when deconvolution of coeluting isomers is not achievable due to reference library limitations. However, high acquision frequency (20–100 HZ) is required for detectors to process the narrow peaks (50–600ms) generated in fast second-dimension separations. This requirement is attained in the VGA-101 detector with acquisition frequency of up to ∼90 Hz. In the VGA-100 detector, due to large volume of the flow cell (∼80 μL), adjusting the make-up gas pressure in necessary for tuning the residence time of analytes in order to detect narrow peaks. Increasing the make-up gas is not needed in the VGA-101 due to lower cell volume (40 μL) [106, 145]. Also, VUV detection can be complementary to mass spectrometry in the areas that MS has limitations, including distinguishing between compounds of similar retention characteristics and similar fragmentation patterns such as closely related structural isomers and labile pesticides [151–153]. Fig. 12 shows an example of how coupling VUV and MS provides more confidence in identification [6].
Since almost all compounds absorb in the VUV region, this detector can be considered a universal detector. Also, there is no broadening presence as with solution-based UV detectors, since spectra are collected in the gas phase which results in more structural information. Even with the similar spectra of isomers, the lower wavelength region has different spectral features special to each isomeric compound. With the ability of deconvolution of coeluting peaks, calculation of spectra, along with the growing libraries, the GC-VUV combination can provide a powerful tool in mixture analysis and can be used to shorten analysis times (due to deconvolution abilities). However, the needed correction factors are not always accurate since they only account for electronic transitions and not for underlying vibronic transitions. More work is needed in order to make this aspect of the technique practical [100]. Also, VUV has higher sensitivity for the analysis of most compounds, which is beneficial for quantitative purposes [13]. However, it should be considered that compound libraries are limited and utilizing them requires some degree of knowledge as to reference library spectra, retention indexes and relative response factors. Therefore, totally unknown analytes cannot be identified. Another challenge in GC-VUV is regulating the pressure of the make-up gas in order to optimize the residence time of the analyte in the flow-cell which affects band broadening [13].
In microwave molecular spectroscopy, rotational spectra are obtained from the transitions between the rotational energy levels associated with a given vibrational state of a particular electronic state. Molecules having a nonzero permanent electric dipole moment, is a fundamental requirement of detection by microwave spectroscopy [154]. When the electric dipole of molecules interacts with electric vector of the microwave radiation, rotational transitions are induced. These transitions fall in the microwave region of the electromagnetic spectrum (from 1000 MHz (λ=30 cm) to 1 million MHz (λ=0.3 mm). In terms of wavelengths, the region can be divided into the centimeter-wave (λ=1–30 cm), millimeter-wave (λ=1–10 mm), and submillimeter-wave (λ<1 mm) regions and frequencies are expressed in megahertz (MHz) or gigahertz (GHz) units (30 and 300 GHz correspond to wavelength of 1 cm and 1 mm, respectively) [154]. The rotational energy levels of the molecules are quantized and depend on the principal moments of inertia (I). Therefore, they are very sensitive to the molecular structure, isotopic composition, and numerous other molecular properties. That is why these spectra provide a considerable amount of structural information [154]. It is possible to calculate the transition energies accurately using a Hamiltonian if angular momentum and moments of inertia are known for each molecule [155, 156]. Also, the response factor can be accurately calculated for a given molecule, therefore, compounds can be quantified even without reference standards [157–160]. Fig. 13 shows the coordinate system for 2-chloropyridine, which is an asymmetric rotor (Ia≠Ib≠Ic). Rotating this molecule about the c-axis requires the spatial movement of the relatively heavy Cl (green) and N (blue) atoms, which requires more energy compared to rotation about the a-axis [157].
Typically, the accuracy of the transition frequencies measured in rotational spectroscopy is better than one part per million. Also, the spectroscopic resolution is very high since the spectra are measured for isolated gas environments and the lifetime of excited rotational states are longer which results in narrower bands [161, 162]. Moreover, transition lines are extremely narrow providing unique fingerprints even for isotopomers, which are not easily distinguishable by any other spectroscopic method. Fig. 14 shows the unambiguous identification of acetonitrile isotopomers using GC-MRR spectroscopy. A compound can be identified using MRR spectroscopy, by using its molecular structure to calculate the spectra or by using a standard. When using the calculated values, the instrument will only look for the calculated transitions in the mixture. Therefore, even isobaric compounds and compounds with similar exact masses can be easily distinguished unlike MS techniques [161]. The mass resolution performance of some techniques like Fourier transform ion cyclotron resonance mass spectrometry (FTICR MS) can reach ~ 10^6^. However, this resolution is only possible at high a vacuum (10^−9^ Torr or lower) [163]. Also, this technique requires a high-field superconducting magnet which causes safety concerns regarding high magnetic fields as well as high cost of operation and maintenance due to continuous consumption of liquid helium [164]. Other powerful mass analyzers with lower resolution power (10^4^ or less) can distinguish between isotopologues, but they cannot differentiate isotopomers that only differ in position of isotopes and have the same isotopic content and therefore, exact masses. Also, since isotopomers have similar fragmentation patterns and cannot be separated by chromatography, more complication is introduced in MS analysis [157].
The first MRR detector that was coupled with GC utilized chirped-pulse Fourier transform microwave (CP-FTMW) spectroscopy which provided a sensitive method for broadband rotational spectroscopy in the 75–110 GHz frequency region corresponding to the millimeter wavelength of electromagnetic radiation. The term chirped pulse means that frequency of the excitation pulse increases or decreases linearly with time which enables the simultaneous excitation of broad bandwidth molecular spectra [161, 165, 166]. A schematic of the first GC-hyphenated MRR instrument is shown in Fig. 15. In order to record the rotational spectra, a short excitation pulse (in the system used here, typically 250 ns) is transmitted into a low-pressure sample cell (pressure~ 10^−2^ torr). The excited molecules then emit coherent radiation at their characteristic rotational frequencies through free induction decay (FID) for several microseconds, which are recorded with the horn antennae in the time domain and Fourier transformed to the frequency domain [161]. CP-FTMW has its origins in several well-known techniques including Fourier transform NMR spectroscopy (FTNMR), FTMW spectroscopy, and rotational coherence spectroscopy (RCS) which are described elsewhere [167–170]. The latest chirped-pulse FT molecular rotational resonance spectrometers incorporate powerful microwave and millimeter-wave sources, broadband active multiplier chains (AMCs), arbitrary waveform generators (AWGs), and high-speed digital electronics for both pulse generating and broadband detection [161, 171–174].
One of the applications of the broadband GC-MRR was in separation and identification of isotopologues and isotopomers since traditional separations with GC stationary phases have limitations [161, 175]. This is shown in Fig. 16, wherein two types of chromatograms are produced. Fig. 16.a shows the total molecule chromatogram (TMC) which is produced by summation of all of observed transitions. Each point in the TMC corresponds to a broadband, high-resolution spectrum. Fig 16.b shows the extracted molecule chromatograms (EMC) of all of the isotopically distinct compounds present in the mixture. This indicates the selectivity of this technique, because even compounds with identical exact masses that are not distinguishable by other GC detectors (e.g., H3C– (C=O) – ^13^CH3 and H3C- (^13^C=O)–CH3 or H3^13^C–C≡N and H3C–^13^C≡N or H3^13^CH2– OH and H3C–^13^CH2–OH) can be easily identified since each of them have different rotational transitions due to the distinct moments of inertia corresponding to each compound. It is noteworthy that this kind of analysis can be performed quickly, and quantitation of analytes is possible even in complex mixtures with ng amounts of sample. However, this is not the case in NMR spectroscopy which is inherently insensitive and has less reliability in analysis of complex mixtures [161]. Another application was the simple determination of natural isotopic abundances of chlorine and bromine atoms in the mixture of bromoethane and five other heterocyclic compounds, since integration of GC-MRR peak areas provides quantitative information for all analytes (see Fig. 17) for more details see ref. [161].
The first-generation GC-MRR had limitations in terms of molecular weight (only compounds <120 Da were analyzable) and suffered from low sensitivity. The reason arises from the broadband (75−110 GHz) excitation source. Also, the absence of a heated transfer line and room temperature analysis (up to 80 °C) contributed to the low molecular weight cut-off and low sensitivity, respectively. However, the second design of the GC-MRR, known as targeted-mode GC- MRR, addresses the prior limitations by utilizing a pulsed-jet mechanism that enables spectral cooling of molecules. In addition, the working range of the instrument was in the 9–18 GHz frequency range which enables the analysis of larger molecules. These improvements lowered the limits of detection of this instrument relative to the first generation GC-MRR instrument and also to what has been shown in the entire field of MRR spectroscopy [157, 168]. As shown in Fig. 18, subjecting the GC effluent to a pulsed-jet supersonic expansion valve connected to a high-vacuum chamber (∼10^−5^ Torr) and an adiabatic expansion of gases, results in gas-phase molecules at very low rotational temperatures [176]. This happens very fast (<1 ms) which prevents post-column band broadening effects observed with large flow cells. Using Ne, He, or Ar as make-up gasses (with a 1mm pinhole nozzle) enables achieving typical rotational temperatures of 1−2 K [157, 176, 177]. Fig. 19 shows the effect of spectroscopic temperature on intensity of spectral lines of 1-(2,3-dihydro-1,4-benzodioxin-6-yl) propan-1-one (192.21 Da). The room-temperature spectrum intensity for this molecule is ~ eight orders of magnitude lower compared to the 1 and 10 K spectra. Also, it can be seen from Fig. 19 that lowering the spectroscopic temperatures results in fewer transition lines that have higher intensities. In addition, the more intense lines tend to be at lower frequencies. These low temperatures can routinely be achieved using pulsed-jet device, and buffer-gas cooling techniques [176–179]. Lowering the temperature down to ~2 K will cool almost all vibrational states for most of the molecules in addition to reducing the number of populated rotational energy levels [180]. Measuring the room-temperature transition intensity for 1-(2,3-dihydro-1,4-benzodioxin-6-yl)propan-1-one will result in higher transition intensity (~two orders of magnitude) when measured at 75−110 GHz ( broadband instrument) compared to 6−18 GHz ( targeted instrument), but still the increase in sensitivity achieved through cooling is much more pronounced. This effect is enhanced with higher molecular weight analytes [157]. The next improvement in design of the second-generation GC-MRR was incorporation of a Fabry-Perot cavity to further increase the detection sensitivity [181]. The advantage of the cavity is that the microwave pulse can make several thousand reflection rounds (cycles) between two Al mirrors and have more interactions with the molecules (Fig. 18) [157, 176]. Table 1 lists the analytes studied by Wahab et al. using the second-generation GC-MRR. These analytes have a wide range in molecular weights and total dipole moments. It can be seen that due to the higher sensitivity of the broadband instrument, limits of detections are close to those of a GC thermal conductivity detector (and sometimes lower than that depending on the make-up gas) can be achieved [157].
After achieving higher selectivities and sensitivities using the targeted mode GC-MRR, another experiment was performed to show the power of the technique in structural identification and quantitation of isomers and isotopologues of a group of dibromo butanes and bromonitrobenzenes. The results of this experiment are shown in Fig. 20, which demonstrates how GC-MRR hyphenation can be to even differentiate between different isotopes of a bromine and quantify them as well [157].
Several studies have shown chiral analysis using MRR spectroscopy [155, 156, 182, 183]. A study by Reilly et al. showed the ability of MRR spectroscopy in measuring the enantiomeric excess of pantolactone by complexation with small “tag” molecules such as propylene oxide, and trifluoro propylene oxide in the gas phase [182]. These complexes are generated in the pulsed jet expansion and are formed through non-covalent interactions between the “tag” and the analyte molecules. Configuration of the “tag” molecule can be the same or opposite of the analyte which results in homochiral or heterochiral complexes, respectively. The homochiral and heterochiral complexes are diastereomers and therefore distinguishable by MRR since they have different rotational spectra. The transition frequencies can be measured for each complex to be used for identification and quantitation. The difference in the intensities of the spectra of the complexes corresponds to the enantiomeric excess [182]. It is also possible to determine absolute configuration of compounds by calculating the transition frequencies of the possible complexes that might form in the gas phase (homochiral or heterochiral) and compare them with the frequencies observed for analytes with unknown configurations [183, 184]. Combining this ability with the separation power of GC, the absolute configuration of enantiomers can be determined when pure standards are not available.
Possibility of calculation of the rotational transition lines for compounds and the fact that rotational lines are extremely narrow, provides a unique fingerprint for each compound, to the extent that compounds can be unambiguously identified and quantified even if standard compounds are not available. Even compounds containing different isotopes of bromine and chlorine can be identified and quantified in MRR spectroscopy. This technique provides the highest degree of structural information compared to the other spectroscopic approaches reviewed herein. Further it can provide greater structural information than GC-MS. As shown in the example of acetonitrile isotopomers in Fig. 14, the narrow line widths of MRR spectra (FWHM= 0.7) provide exceedingly high resolution which can be used to distinguish compounds that have the same exact masses and similar structures. Moreover, there have been studies performed on identification of enantiomers using MRR spectroscopy showing a promising future for absolute determination of enantiomers with GC-MRR without a need for reference standards [182, 183]. Sensitivity of GC-MRR is close to TCD detector (or more depending on the compound and the make-up gas). However, this is a newly developed technique that still needs improvements in sensitivity and instrument design in order to be commercialized.
Combination of high information spectroscopic detectors with gas chromatography provides useful means to identify and quantify unknown compounds in a mixture. Among the discussed detectors, the degree of structural information provided is the most for MRR and NMR detectors and decreases in going to FTIR and VUV detection. It should be mentioned that even though libraries have been developed for FTIR and VUV techniques, these techniques do not inherently provide detailed information on the structural arrangement of a compound. Although NMR spectroscopy provides a high degree of information, its lack of sensitivity and poor hyphenation with GC does not make it a good candidate for a commercial GC-hyphenated instrument. Current sensitivity of GC-MRR is as good as LP-FTIR and nanogram amounts of analytes is enough for unambiguous determination of structure as well as quantitation, however this technique needs to be improved in terms of sensitivity in broadband mode. Sensitivity of MI-FTIR is 100 times more than the LP-FTIR and DD-FTIR is ~10 times more sensitive than MI-FTIR. Currently, VUV provides the highest sensitivity of the reviewed detectors, but provides the least structural information. It should be noted that the range of sensitivity can vary depending on the functional groups, dipole moments and other relevant properties of the compound being analyzed with each technique. All in all, each spectroscopic technique has its strengths and limitations, but they can all work in an orthogonal manner to provide the highest amount of structural information.