Chemiluminescence in Analytical Chemistry

to the sample inlet where it is mixed with oxygen (or air) and with hydrogen fuel prior to the burner head, and before entering the chemiluminescence cell. Fuel-rich, hydrogen/oxygen flames are generally preferred, but the combustion mixture must be optimized for specific analytes. As is common in chemiluminescence analysis, the greatest limitation to sensitivity of the FPD is noise associated with the background signal, which arises primarily from other flame emissions. To optimize the limit of detection, the PMT is positioned, and lenses are used to view a region of the flame where the signal-to-background ratio is greatest, while filters are employed to reduce background contributions of flame emissions.

5.2.1Sulfur Detection by FPD

The most common application of the FPD to routine GC is detection of organosulfur compounds in complex matrices. Examples include quantification of sulfur compounds in petrochemical feedstocks [88, 89], measurements of sulfur-con- taining pesticides such as malathion and parathion in environmental samples [90, 91], and detection of sulfur compounds in foods and beverages, especially beer, where they are strong contributors to flavor and fragrance [92]. The FPD also detects inorganic sulfur compounds such as sulfates, sulfites, thiosulfates, and thiocyanates [87, 93]. The trace gases SO2, H2S, OCS, CS2, CH3SH, and (CH3)2S may be detected in the atmosphere, but require a preconcentration technique [94, 95].

Despite occurring in a very complex flame environment, the mechanism for sulfur detection in the FPD is largely understood. Sulfur compounds are fully combusted in the flame, resulting in the formation of sulfur atoms, among other sulfur species. These atoms recombine to form S2 in its B3 u electronically excited state, which relaxes to the X3 g ground state by emission of light in a series of bands with maximum intensities at 284 and 294 nm [96, 97], as shown in Figure 10 for the chemiluminescent reaction between OClO and H2S.

Because the reaction forming S2* is second order in sulfur atoms, the signal is approximately quadratic in concentration of the sulfur analyte. Although the detector may be linearized electronically by taking the square root of the signal, the power dependence on sulfur analyte concentration has been observed to vary from 1.5 to 2.0, depending on flame conditions [85, 87]. Because of this, calibration curves must still be constructed over the entire range for which compounds are to be quantified. The detector can also be linearized chemically at the expense of its dynamic range by adding a large excess of a sulfur compounds such as SF6 to the burner gas [98–100], bringing the kinetics into a pseudo-first-order regime in sulfur. This chemical linearization has the additional benefit of improv-

ing the limit of detection by one to two orders of magnitude [87]. The FPD suffers from quenching by high concentrations of coeluting hydrocarbons. It has been suggested that hydrocarbons may enhance the formation of CS at the expense of S atoms, thereby reducing the sensitivity [101]. Reported detection limits for a chemically linearized FPD have been determined to be 2–10 pg S s 1 [87].

5.2.2Detection of Other Elements by FPD

In addition to sulfur compounds, the FPD has been used for the measurement of organic compounds containing atoms such as phosphorus, nitrogen, boron, arsenic, antimony, and even chlorine [93]. Organophosphorus compounds are of particular interest owing to their common use as pesticides and chemical warfare agents. A modification to the FPD design, discussed below, known as pulsed flame photometric detection (PFPD) has additionally detected hydrocarbons and organic compounds containing tin, germanium, selenium, silicon, iron, and manganese [102]. The mechanism of detection for organophosphorus compounds is through the formation of PO, which subsequently reacts with H atoms in the fuelrich flame to produce HPO*:

emitting light at 526 nm [86]. Nitrogen-containing organics are detected in an analogous manner, generating HNO*, which emits light at 690 nm [93]. Unlike the nonlinear response to sulfur, HPO and HNO give a response proportional to analyte concentration. However, the response is not directly proportional to the number of P or N atoms, as structural and compositional differences between compounds can alter both the chemiluminescence efficiency and the emission spectra of the products [86].

Emissions from BO2* [103], AsO*, and SbO* [103, 104] in an FPD flame have been used to detect organics or highly reduced species containing B, As, and Sb, respectively. These metal atoms react to form the same excited-state metal oxides discussed in their reactions with ozone above. These analytes have limits of detection measured to be approximately 50 ppbv, 10 ppbv, and 20 ppbv, respectively [93].

5.2.3Improvements to FPD Design

Two improvements to the FPD design have been made in recent years. First, the addition of an oxygen-rich burner upstream of the FPD oxidizes hydrocarbons to CO and CO2 and thereby eliminates the hydrocarbon interference within the analytical flame [105, 106]. This increases the sensitivity of the detector to approximately 20 pg S s 1 for a nonlinearized FPD [102]. In a second improvement, the PFPD was developed to significantly reduce the background and increase the sensitivity to all detectable atoms [107]. This detector is now commercially

available from Varian Instruments (San Fernando, CA). In this design, the burner is constructed to generate a noncontinuous flame, which is reignited at a frequency of about 1–10 Hz. This periodic interruption allows for the acquisition of the time-dependent emissions from the various excited-state species present in the detector. As each of these has differing fluorescence lifetimes on the order of milliseconds, they can be differentiated in the time domain between flame pulses. Partially because the flame emission background is absent from the signal and other coeluting interferences may be eliminated, this technique has lowered the detection limits for sulfur, phosphorus, and nitrogen to 0.2, 0.01, and 2 pg s 1, respectively. Chemically linearizing the sulfur detection resulted in a detection limit of 30 fg S s 1 [108]. Additionally, carbon can be detected at levels near 60 pg s 1 with this technique based on the emissions of both C2* and CH* species [102].

5.3Fluorine-Induced Chemiluminescence Detection of S, Se, Te, and P Compounds

In the FCLD [65], F2 is reacted with the effluent of a GC in an evacuated cell at a pressure of 1 torr. Chemiluminescence is monitored using a red-sensitive PMT in conjunction with a band-pass filter that isolates wavelengths in the range 660–740 nm, as in the generic chemiluminescence detector shown in Figure 1. A commercial version of this detector was manufactured as the Model 300 SCD by Sievers Research (now Ionics-Sievers, Boulder, CO). It was later replaced by the Model 350 SCD, which is based on SO O3 chemiluminescence, described below. In the Model 300 instrument, fluorine was generated online by means of a high-frequency electrical discharge of SF6. The instrument was found to be linear over at least three orders of magnitude and displayed a limit of detection of 1 pg S s 1 for a wide range of mercaptans, sulfides, disulfides, and trisulfides. Cyclic sulfur compounds such as thiophene were detected with about one order of magnitude less sensitivity. The detector has been successfully interfaced to high-performance liquid chromatography (HPLC) [109] and supercritical fluid chromatography (SFC) [110], in addition to GC.

The gases H2S, CS2, OCS, and SO2 do not provide significant responses in the FCLD [65], apparently because their ionization potentials are too large to form long-lived charge-transfer complexes, and because they do not contain β hydrogens that can be eliminated to form the CCS bond, as in Reaction (33). The inability to detect these compounds is a disadvantage for some analytical applications (but can be an advantage for the highly selective detection of atmospheric DMS, discussed below). The detector also responds analogously with high sensitivity to organo-selenium and -tellurium [67] compounds and to phosphines, alkyl phosphines, and phosphinate esters [68]. As previously mentioned, the FCLD exhibits selectivities of 107 against alkanes, but compounds with

weak CEH bonds also provide weak responses. This lack of adequate selectivity was the ‘‘Achilles heel’’ for this detector from a commercial standpoint, since one of the largest markets for sulfur-selective detectors is for the analysis of petrochemical feedstocks for sulfur compounds that can foul catalysts. These feedstocks often contain relatively high concentrations of olefins.

The FCLD has been particularly useful as a highly sensitive means of measuring volatile sulfur, selenium, and tellurium compounds in studies of bacterial methylation [67, 111–114]. Because of its high sensitivity to phosphinate esters [68], the FCLD could potentially serve as a monitor for nerve gases.

5.4 Sulfur Chemiluminescence Detector

The sulfur chemiluminescence detector, invented by Benner and Stedman [115, 116], combusts the GC effluent in a fuel-rich H2/O2 flame. The combustion products are reacted with ozone at low pressure ( 1–10 torr), and the chemiluminescence is detected by a PMT in combination with an optical filter to select for SO2* [Reaction (4)] chemiluminescence, as in the generic chemiluminescence detector shown in Figure 1. Commercial versions of this detector, manufactured by Ionics-Sievers and Antek Instruments (Houston, TX), combust the GC efflu-

ent with a H2/O2 mixture in a ceramic or quartz furnace resistively heated to 800–1000°C, although the original SCD manufactured by Sievers Instruments

used a ceramic capillary to sample the exhaust gas from the flame of an FID [117]. The SCD is linear over four or more orders of magnitude and exhibits a nearly equimolar response to all sulfur compounds [115, 116]. Because potential interferents such as olefins are combusted to form products (CO2 and H2O) that do not react with ozone, the detector is highly selective. An example of the selective detection of trace sulfur compounds in a hydrocarbon fuel is given in Figure 11. The limit of detection of the SCD has been reported to be as low as 25 fg S s 1 [118]. The SCD has advantages over the FCLD for sulfur-compound detection in that it responds to all sulfur compounds and exhibits much higher selectivity. In comparison to the FPD, the SCD is more sensitive and exhibits a linear response.

Although there has been some dispute among the manufacturers about the mechanism of detection, it is clear from the scientific evidence that the signal is derived from the reaction of SO with O3 [Reaction (4) discussed above]. Martin and Glinski [119] have identified SO2* as the emitting species. That the SO2* is derived from the reaction of the SO intermediate with O3 has been confirmed by flow-tube kinetics studies in which the flame or furnace effluents were reacted with O3 and NO2 [23, 120]. The rate constants for reaction of the chemiluminescent intermediate with O3 and with NO2 agree with the literature values for those rate constants and with the rate constants measured in the same apparatus using SO generated in a microwave discharge of SO2/He. One proviso, however, found

Figure 11 SCD chromatogram of catalytic cracked (FCC) gasoline, demonstrating the selectivity for sulfur compounds. (A) SCD response; (B) FID response. (Reprinted from American Laboratory 23(3):117, 1991. Copyright 1991 by International Scientific Communications, Inc.)

by Burrow and Birks [120], is that with the higher sulfur concentrations used in the kinetics experiments, the SO was derived from some other chemical species, X, which itself reacts rapidly with O3 to form SO:

Based on kinetics considerations, it was possible to rule out a large number of HxSyOz species, including S atoms, which are known to react rapidly with ozone. The most likely candidate for X is S3 formed in the association of S atoms with S2, a reaction that could occur within the transfer line linking the furnace with the chemiluminescence chamber [120]. Watson and Birks (unpublished results,

1998) demonstrated that sulfur vapor, which contains S3, does chemiluminesce with ozone.

5.5Thermal Energy Analysis and Nitrogen-Selective Detectors

The NO O3 chemiluminescent reaction [Reactions (1–3)] is utilized in two commercially available GC detectors, the TEA detector, manufactured by Thermal Electric Corporation (Saddle Brook, NJ), and two nitrogen-selective detectors, manufactured by Thermal Electric Corporation and Antek Instruments, respectively. The TEA detector provides a highly sensitive and selective means of analyzing samples for N-nitrosamines, many of which are known carcinogens. These compounds can be found in such diverse matrices as foods, cosmetics, tobacco products, and environmental samples of soil and water. The TEA detector can also be used to quantify nitroaromatics. This class of compounds includes many explosives and various reactive intermediates used in the chemical industry [121]. Several nitroaromatics are known carcinogens, and are found as environmental contaminants. They have been repeatedly identified in organic aerosol particles, formed from the reaction of polycyclic aromatic hydrocarbons with atmospheric nitric acid at the particle surface [122–124]. The TEA detector is extremely selective, which aids analyses in complex matrices, but also severely limits the number of potential applications for the detector [125–127].

Both N-nitroso and nitro groups are thermally labile and can be pyrolyzed to liberate nitric oxide. Because of the low thermal stability and low volatility of the compounds usually detected with this technique, GC cannot always be used for their separations. An analyte simply may not survive the temperature gradients required to elute it [121]. To circumvent this problem, the TEA detector has also been interfaced to HPLC with some success [128, 129]. The HPLC column effluent is discharged through a restricting orifice and nebulized into the heated detection cell, where the mobile phase is evaporated and the analysis proceeds as with GC. The GC/TEA design possesses the advantage of simplicity, consisting only of a heated tube or catalyst bed interfaced to an evacuated cell supplied with a flow of ozone and viewed through an optical filter by a PMT as in Figure 1.

The name thermal energy analysis refers to the amount of heat energy required to break a bond in an analyte to produce the detected NO. The detector can be made more or less selective, depending on the pyrolysis conditions used. Although the simplest design uses only a heated tube as the pyrolysis chamber, these detectors have more often made use of catalysts to detect a wider variety of compounds at lower temperatures [121]. While many catalysts have been investigated, WO3 and W20O58 have been shown to be especially effective [125]. Other catalysts investigated included platinum and nickel alloy oxides to achieve

similar results [130, 131]. For a simple pyrolysis tube made of quartz or ceramic,

the NENO bond in N-nitroso compounds cleaves at temperatures in the range 200–300°C with few interferences from other nitrogen-containing compounds

(with the exception of some organic nitrites). At much higher pyrolytic zone temperatures, up to approximately 600°C, the signal for nitrosamines begins to decrease and those for nitroaromatics, nitroalkanes, and nitroamines become significant. For each compound class, detection limits on the order of picograms of analyte have been determined [121].

Analogously, nitrogen-selective detectors have been developed to detect all nitrogen-containing organics [132]. This is accomplished by adding oxygen

to the column effluent upstream of the pyrolysis tube. At catalyst temperatures near 800–1000°C, any nitrogen present in an analyte is converted to NO, while

all carbon and hydrogen is oxidized to CO2 and water, respectively. The NO is then detected in the same manner as the TEA detector described above. Limits of detection have been demonstrated to be in the picogram range for organic analytes containing at least one nitrogen [121]. The response to nitrogen in compounds containing differing numbers of nitrogen atoms is not always equimolar. For example, the ratio of responses to N-nitrosodimethylamine and pyridine is 2:1, not 3:1 as expected from their nitrogen molar ratios [121].

5.6 Redox Chemiluminescence Detector

The redox chemiluminescence detector (RCD), developed by Nyarady et al. [133], and marketed for a short time by Sievers Research, selectively detects compounds capable of reducing NO2 to NO at a heated gold surface. The NO is detected by chemiluminescence with ozone [Reactions (1–3)], as in the TEA detector and the nitrogen-selective detectors described above. The RCD detects most compounds containing oxygen, nitrogen, sulfur, or other reactive functional groups, and the selectivity can be tuned by varying the NO2 concentration or the temperature of the gold catalyst bed. Atmospheric gases that respond include CO, H2O2, SO2, H2S, OCS, CS2, and H2. Figure 12 compares FID and RCD chromatograms of jet fuel containing 10 ppmv of the antioxidant BHT (2, 6-di-tert- butyl-4-methylphenol), illustrating the high degree of selectivity that can be achieved with this detector. The RCD also has been adapted for detection in SFC and HPLC [134, 135]. The principal difficulty encountered with the RCD is the tendency for the gold catalyst to become poisoned; it is often necessary to recondition the catalyst by flowing oxygen through the bed at high temperature. Perhaps the main reason that the RCD did not become widely accepted is that it responds to a wide range of compounds rather than having an element-specific or functional-group-specific response. As a result, little chemical information is provided for unknown chromatographic peaks. The detector could be useful for

Figure 12 FID and RCD chromatograms of JP-4 jet fuel containing 10 ppm of BHT, demonstrating the high degree of selectivity against hydrocarbons. (Reprinted with permission from Ref. 153. Copyright 1985 American Chemical Society.)

some applications, however, provided the irreproducibility associated with the catalyst is solved.

5.7 Active Nitrogen Detectors

There has been considerable interest in the application of active nitrogen as a universal GC detector [56–58, 136]. Melzer and Sutton et al. named this detection technique metastable transfer emission spectroscopy (MTES) after the metastable N2* component of active nitrogen. An advantage of the use of active nitrogen, as discussed above, is the ability to distinguish between paraffins and olefins by selectively detecting olefins with the atomic component of active nitrogen [Reaction (20)], present in low-pressure discharges, and then doping the column effluent with HCl [Reactions (21,22)] to detect both simultaneously. The system can also be applied to the detection of organometallics in GC separations with high specificity [58]. More recently, Rice et al. applied an active nitrogen detection scheme at ambient pressure, the atmospheric pressure active nitrogen (APAN) detector, for nonselective detection of both hydrocarbons and organometallics [57]. Although clearly useful, detection by active nitrogen does not appear to

have sufficient advantages to supplant the FID as a universal detector, and GC is unlikely to compete with atomic absorption or atomic emission as a practical method for metal analysis. The main advantage of active nitrogen detection is flexibility; this one detector, when properly configured, can perform a variety of analyses that would otherwise require a number of individual detectors.

6. APPLICATIONS TO ATMOSPHERIC RESEARCH

Chemiluminescence is the method of choice for measurements of a few select atmospheric species, especially the oxides of nitrogen. Where chemiluminescence is sufficiently selective that it can be used without prior separation, the high inherent sensitivity of the technique can be exploited. Although chemiluminescence requires no light source, the necessity of a vacuum pump to create low pressures and gas cylinders to provide reagent gases often makes the instruments large and heavy. The NOx and NOy instruments described below, for example, weigh up to several hundred pounds. An additional difficulty associated with the application of gas-phase chemiluminescence to atmospheric measurements is that the detector response changes with reactor pressure, while the atmospheric pressure changes with altitude. Nevertheless, there are several atmospheric measurements that are best made by chemiluminescence.

6.1 ‘‘NOx Box’’ Detector for NO, NO2, and NOy

The most widely used chemiluminescent reaction in atmospheric field studies is the reaction of NO with O3 [Reactions (1–3)]. The so-called ‘‘NOx box’’ measures both NO and the sum of NO and NO2, which is defined as NOx [1, 2, 4]. The concentration of NO2 is obtained by difference. First, NO is measured by mixing air with ozone and measuring the chemiluminescence signal with a redsensitive PMT and optical filter to discriminate against interferences from alkenes and reduced sulfur compounds. The air is then diverted through a photochemical reactor prior to entering the reaction chamber where a fraction, on the order of 50%, of the NO2 is photolyzed to form NO. The sensitivity to NO and efficiency of photolysis of NO2 are periodically measured by means of standards. Limits of detection for integration times of the order of 1 s are in the range 1–10 pptv [4, 137, 138].

The measurement of the total reactive oxides of nitrogen, NOy ( NO NO2 NO3 2 N2O5 HNO2 HNO3 HNO4 CIONO2 organic nitrogen compounds particulate nitrate), has been of great interest in the atmosphere for several years. It has been found that all these compounds may be quantitatively converted to NO by either passing air through a heated molybdenum tube (which has a catalytically active oxide surface) or mixing CO or H2 with the air stream

in a heated gold tube in a manner similar to that of the RCD described above [137–139]. The NO produced may then be quantified using a NOx box.

New laboratory data on conversion efficiencies for a wide range of gasphase odd-nitrogen species and a review of the status of NOy measurements were published recently [140]. Also, measurements of NOy have been subject to recent scrutiny [140, 141]. During the field experiment known as PEM-West (fall of 1991), for example, two NOy instruments operating aboard a DC-8 aircraft showed significant disagreement. Also, the measured NOy was greater than the sum of its separately measured components, often by a factor of two or more. This discrepancy, which has been observed several times before, has come to be known as the ‘‘missing NOy’’ [141] and is of great interest to atmospheric chemists. It is especially important to be able to account for all of the reactive nitrogen species in the atmosphere. Although the chemiluminescence measurements of NO and NO2 are now well established, measurements of NOy are still problematic. Especially poorly established is the efficiency of conversion of particulate nitrate to NO in the catalytic converters.

6.2 Fast Ozone Detector

Ozone is usually measured in the atmosphere by UV absorbance at 254 nm with a response time of the order of several seconds. Faster measurements are desired for aircraft measurements owing to the high speed of the aircraft, and measurements of ozone flux using the eddy covariance technique require a method having a response time of up to 10 Hz [142]. This can be accomplished by using the chemiluminescent reaction of O3 with a large excess of either NO, as in the detection of NO by O3 [Reactions (1–3)], or ethylene [Reactions (5–8)] [36, 143– 146]. A fast-response detector based on the heterogeneous, gas-solid reaction of ozone with a fluorescent dye such as rhodamine B may also be used and has the advantage of being extremely lightweight [147–149]. The latter method requires frequent calibration, however, which can be accomplished with the slower UV absorbance instrument. Miniaturized UV ozone sondes weighing less than 1 kg have been developed recently and may be coupled with fast-response ozone detectors [150].

6.3 Isoprene Detector

Isoprene, the most abundant hydrocarbon emitted to the atmosphere by plants, can also be measured using ozone chemiluminescence. As discussed above, alkenes react with ozone to produce formaldehyde in its 1A2 electronic state, in addition to several other chemiluminescent products. In a fast isoprene detector manufactured by Hills Scientific (Boulder, CO), the chemiluminescence is detected using a blue-sensitive PMT to maximize the sensitivity for isoprene detec-