Anal. Chem. 2005, 77, 1007-1012
VUV Single-Photon Ionization Ion Trap Time-of-Flight Mass Spectrometer for On-Line, Real-Time Monitoring of Chlorinated Organic Compounds in Waste Incineration Flue Gas Shizuma Kuribayashi,*,† Hideo Yamakoshi,† Minoru Danno,† Satoshi Sakai,† Shigenori Tsuruga,‡ Hiroshi Futami,§ and Shigeki Morii‡
Advanced Technology Research Center, Yokohama Research & Development Center, Mitsubishi Heavy Industries, Ltd., 1-8-1 Sachiura, Kanazawa-ku, Yokohama, 236-8515, Japan, and Nagasaki Research & Development Center, Mitsubishi Heavy Industries, Ltd., Yokohama, Japan
In this work, a sensitive and robust vacuum ultra-violet (VUV) single-photon ionization (SPI) ion trap time-of-flight mass spectrometer (VUV-SPI-IT-TOFMS) for on-line, realtime monitoring of chlorinated organic compounds in waste incineration flue gas has been newly developed. The fragment-free SPI technique with 121.6-nm VUV lamp irradiated by a microwave generator and the quadrupole ion trap to accumulate and select analyte ions were combined with a reflectron time-of-flight mass spectrometer to detect chlorinated organic compounds at trace level. This measuring system was tuned up to detect dioxins precursors with the aim at an application to monitoring trace level toxic substances in flue gases from incinerator furnaces. As a result, this technology has made it possible to analyze trichlorobenzene (T3CB), a dioxin precursor, in 18 s with a sensitivity of 80 ng/m3‚N (10 pptv) using the selective accumulation of analyte substances and separation of interfering substances in the ion trap. Moreover, the first field test of the continuous monitoring T3CB in an actual waste incineration flue gas had been done for 7 months. The results show that this system has an exceeding robust performance and is able to maintain the high sensitivity in analyzing T3CB for long months of operation. Polychlorinated dibenzo-p-dioxins/furans (PCDD/F) and dioxin precursors contribute to environmental pollution and cause serious social problems. Waste incinerators as one of major emission sources have to reduce and control PCDD/F and their precursors’ emission under the environmental acceptance criteria. The occurrence of PCDD/F and their precursors is normally due to combustion conditions. Therefore, a stable control system of combustion requires reliable on-line and real-time monitoring. * Corresponding author. E-mail:
[email protected]. † Advanced Technology Research Center. ‡ Yokohama Research & Development Center. § Nagasaki Research & Development Center. 10.1021/ac048761y CCC: $30.25 Published on Web 01/13/2005
© 2005 American Chemical Society
Zimmermann et al.1 have pointed that front-end and back-end measurements in a waste incineration plant are effective to reduce the incomplete combustion product emissions. Front-end measurements deal with an optimization of the combustion condition parameters of temperature, oxygen, and feed supply. Front-end measurement requires fast on-line measurement, because rapid changes of combustion conditions occur even in large incinerators. Monitoring CO, NO, and so on controls the actual combustion conditions in waste incinerations. CO gas concentration usually indicates a good relation to that of PCDD/F and their precursors, but it does not work sufficiently if the concentrations of PCDD/F and their precursors are very low. On the other hand, the concentrations of the precursors indicate a good relation to that of PCDD/F. Moreover, the concentrations of the precursors are generally over 1000 times larger than the density of PCDD/F. Therefore, a real-time measurement of the precursors is expected to be feasible and effective. Back-end measurements deal with any kind of flue gas cleaning procedures, such as dust precipitators, wet scrubbers, or fabric filters. Generally, catalytic conversion (SCR) or adsorptive trapping is applied to reduce PCDD/F and their precursors as back-end technology. Back-end measurement of PCDD/F in waste incineration flue gas is very important for public security. However, realtime measurement of PCDD/F is not easy, because the concentrations of PCDD/F are much less than the concentrations of the precursors. Therefore, the standard method to analyze PCDD/F is necessary to take a long time to concentrate PCDD/F, separate into various isomers, and assess the respective toxicities of each compound. When public security in the future will require a continuous and on-line measurement of the PCDD/F, a rapid (convenient), sensitive, and robust detection method will be necessary. These requirements encouraged this development of a sensitive vacuum ultraviolet (VUV) single photon ionization (SPI) timeof-flight mass spectrometer (TOFMS) with an ion trap for online, real-time monitoring of chlorinated organic compounds. (1) Zimmermann, R.; Heger, H. J.; Blumenstock, M.; Dorfner, R.; Schramn, K. W.; Boesl, U.; Kettrup, A. Rapid Commun. Mass Spectrom. 1999, 13, 307314.
Analytical Chemistry, Vol. 77, No. 4, February 15, 2005 1007
The SPI process has advantages compared with other processes such as electron impact (EI), chemical ionization (CI), and resonance-enhanced multiphoton ionization.2 EI causes massive fragmentation of organic compounds during the ionization process. SPI is a threshold ionization process, so it can avoid fragmentation in the ionization process. CI is very sensitive to matrix effects (namely, the chemical composition of the sample gas can influence the ionization efficiency of a specific target compound). MPI is suited to detection of aromatic compounds, but it has a tendency to decrease the sensitivity to aromatics with increasing degree of chlorination. SPI is not sensitive to the increasing degree of chlorination. The single photon ionization process using VUV (VUV-SPI) is direct and so soft that it can be adjusted to be nearly fragmentation free. Thus, VUV light is an excellent ionization source, so VUV-SPI-TOFMS has been applied to diagnose several chemical systems. There are two types of VUV sources (laser and lamp) in the applications. Generally, laser-based photoionization techniques have the drawback that it is necessary to use expensive, bulky, and sophisticated laser systems. Therefore, a microwave-excited VUV lamp system is selected as a practical VUV light source, because it was sufficient to use a commercial (low in price) and robust microwave generator, and it was not necessary to use sophisticated VUV laser optics except for a VUV window. Most VUV-SPI-TOFMS experiments2-4 have reported a detection limit of over 1 parts-per-billion volume (ppbv). However, the higher sensitivity is necessary to detect dioxin precursors in real time. In this system, an ion trap-TOFMS was used instead of the usual TOFMS in order to increase amounts of ionized molecules per one measurement and improve the selectivity of the target molecule. The selective ion-trapping technique is expected to improve the sensitivity of the actual flue gas spectrometry because there are many chemical compounds in the flue gas that often prevent the concentration of the target molecule ions or cause background noises. There are many dioxin precursors whose ionization potentials are low enough to be ionized at 10.2 eV. This study has focused on trichlorobenzene (T3CB), because its concentration is relatively high and correlates closely with the amount of dioxin toxicity equivalency quantity (TEQ), which is calculated by conversion of PCDD/F concentrations into a toxicologically equivalent concentration of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD). EXPERIMENTAL SECTION Figure 1 shows the schematic diagram of the experimental setup. This analysis system consists of a VUV lamp, a quadrupole ion trap (IT, Jordan Co.), and a reflectron TOFMS (Jordan Co.). The sampling gas is introduced into the ion trap and ionized by the VUV light, and the ions are accumulated in the quadrupole ion trap. The accumulated ions are selected into the target ions preliminarily and separated from interfering substances that have the same molecular weight. This process is to increase the sensitivity and extend the practical applicability of the system (2) Mu ¨ hlberger, F.; Wieser, J.; Ulrich, A.; Zimmermann, R. Anal. Chem. 2002, 74, 3790-3801. (3) Mu ¨ hlberger, F.; Zimmermann, R.; Kettrup, A. Anal. Chem. 2001, 73, 35903604. (4) Tonokura, K.; Nakamura, T.; Koshi, M. Anal. Sci. 2003, 19, 1109-1113.
1008 Analytical Chemistry, Vol. 77, No. 4, February 15, 2005
Figure 1. Schematic drawing of the VUV-SPI-IT-TOFMS. It consists of VUV light-emitting part, ionization and ion-trapping part, and a reflectron TOFMS. The first part is composed of a VUV lamp exited by a microwave generator and a VUV window (lens). The second part is composed of an ionization chamber, a gas inlet column, and a quadrupole ion trap.
because actual gases contain various interfering substances. The TOF signal is detected by a three-stage microchannel plate (MCP, Jordan Co). The output of signal from the MCP was digitized by a 2-GHz multiple-event time digitizer (FAST ComTec). In this system, a 121.6-nm VUV lamp is selected as a SPI light. Because the ionizing potential of Lyman R light (wavelength, 121.6 nm; photon energy, 10.2 eV), which is irradiated from hydrogen plasma, is slightly higher in photon energy than 8.5-10 eV of PCDD/F5 and their precursors’ ionization energy,6-13 it can be used to perform ionization efficiently without the occurrence of any dissociation. The lamp consists of a commercial microwave generator and a microwave cavity where a gas tube is located in the center that is filled with a H2/He mixture gas at 5 Torr. H2 gas is ∼1% in density. In this condition, the intensity of the VUV light is estimated to be on the order of 1014 photons cm-2 s-1.14 The VUV light is introduced through a MgF2 window (lens) into the ion trap in the vacuum chamber. Although a VUV lamp is weak in energy per unit area and unit time compared with a laser, the lamp is continuously irradiated by a 300-W microwave generator in order to increase the amount of ion generation. Therefore, a sufficient number of ions for achieving high sensitivity can be accumulated in the ion trap in a limited time. The gas to be analyzed entered into the interior of the ion trap through a deactivated fused-silica capillary column. The sampling tubing and the capillary column were kept at a temperature of ∼200 °C, to minimize the adhesion of any of the dioxins and their (5) Koester, C. J.; Hites, R. A. Chemosphere 1988, 17, 2355-2362. (6) Fujisawa, S.; Ohno, K.; Masuda, S.; Harada, Y. J. Am. Chem. Soc. 1986, 108, 6505. (7) Fujisawa, S.; Ohishi, I.; Masuda, S.; Ohno, K.; Harada, Y. J. Phys. Chem. 1991, 95, 4250. (8) Maier, J. P.; Thommen, F. J. Chem. Phys. 1982, 77, 4427. (9) Lipert, R. J.; Colson, S. D. J. Chem. Phys. 1990, 92, 3240. (10) Cockett, M. C. R.; Takahashi, M.; Okuyama, K.; Kimura, K. Chem. Phys. Lett. 1991, 187, 250. (11) Marier, J. P.; Turner, D. W. J. Chem. Soc., Faraday Trans. 2 1973, 69, 521. (12) Zimmermann, R.; Boesl, U; Lenoir, D.; Kettup, A.; Grebher, T. L.; Neusser, H. J. Int. J. Mass Spectrom. Ion Processes 1995, 145, 97. (13) Williams, B. A.; Cool, T. A. J. Chem. Phys. 1990, 93, 1521. (14) Davis, D.; Braun, W. Appl. Opt. 1968, 7, 2071.
Figure 2. Timing chart of the quadrupole ion trap for T3CB measurement. About 100-300 kHz notched SWIFT pulses (black condensed waveforms in the figure) are applied to one end cap electrode. After 0.28-s SWIFT mode, a VUV light is put out. Consequently, an auxiliary ac voltage is applied to the end cap electrode. After this tickle mode (0.01 s), a short cooling mode begins (0.01 s). A ring electrode is applied with 1-MHz rf voltage during one cycle (0.28 s).
precursors on the walls. The adhesion causes a lot of loss and time lags since they adhere easily to any walls they come in contact with. In addition to above, the ionization chamber was also kept at a temperature of ∼200 °C. The quadrupole-type electrodes were adopted for use as the ion trap. The VUV light from the source lamp was irradiated onto the analyte molecules that were led into the interior of the ion trap. The higher the pressure in the trap, the greater was the amount of ionized molecules present. However, excessively high pressure results in gas breakdown. As a result of optimization, an appropriate internal pressure of the ionization chamber was determined to be lower than 10-4 Torr. The internal pressure of the ion trap is estimated to be on the order of 10-3 Torr because the total conductance of six holes on the ion trap is very small compared with pumping speeds of the turbomolecular pump attached to the ionization chamber. Since the VUV-SPI can ionize all molecules whose ionization energies are lower than the VUV photon energy, selectivity of this method in the ionization is low. Therefore, if there are a lot of interfering molecules that have the same mass numbers as that of the target molecules, the measurement accuracy deteriorates. It occurs often in actual flue gas analysis. This problem is solved as follows with the description of T3CB. Figure 2 shows a timing chart for each process. When ions are accumulated in the quatrupole ion trap under continuous ionization using a VUV lamp and a selective electric signal is applied to the ion trap, only ions with the same mass number as that of T3CB are accumulated. Other ions with different molecular weightrs are simultaneously eliminated (selective accumulation).15 This technique is called a notched stored waveform inverse Fourier transform (SWIFT).16 The selecting signal has an inverse Fourier transform waveform of a continuous frequency spectrum in which only the secular frequency of T3CB to be accumulated is not included. The secular frequency is given by the mass number of T3CB and the trap (15) He, L.; Wu, J.-T.; Parus, S.; Lubman, D. M. Rapid Commun. Mass Spectrom. 1997, 11, 1739-1748. (16) Soni, M. H.; Cooks, R. G. Anal. Chem. 1994, 66, 2488-2496.
Figure 3. Procedure of selective accumulation and fragmentation for a T3CB standard gas. (A) shows a typical mass spectrum of T3CB standard gas. (B) shows a typical mass spectrum of selected T3CB by the SWIFT technique. (C) shows a typical fragmentation pattern of T3CB by the tickle technique.
frequency (1 MHz, for instance) applied to the rf ring electrode. Figure 3A shows the mass spectrum of a typical gas sample when the selective accumulation is not performed, while Figure 3B shows the mass spectrum when a selective accumulation is performed. The latter shows that ions with m/z 180, 182, and 184, which are the isotopic ions of T3CB, can be selectively accumulated. Then, T3CB and its interfering molecules that have the same molecular weight as T3CB, which are selectively accumulated together with the T3CB, are separated. In this procedure, an auxiliary ac voltage (tickle) is applied to the end caps of the ion trap to activate the target ion to induce collision with the ambient buffer molecules and the consequent dissociation into fragments. As shown in Figure 3C, T3CB is dissociated into molecules with m/z 145 and 147, which have lost one chlorine atom, respectively, and the molecules with m/z 109, which have lost two chlorine atoms and one hydrogen atom. Interfering molecules can then be separated from the T3CB molecules because they have different fragmentation patterns. Furthermore, the sensitivity of the system was improved by lighting the VUV lamp only during accumulation. This was done in order to prevent ions produced during the fragmentation from becoming noise. Thus, it becomes possible to realize a highly sensitive analysis by ordinary reflectron TOFMS. Analytical Chemistry, Vol. 77, No. 4, February 15, 2005
1009
The selecting signal applied to carry out the selective accumulation was generated at an arbitrary waveform generating board installed in a computer, amplified by an amplifier, and applied to the ion trap through a filter that protects the flow back of pulse voltage applied for the ion extraction. Since one cycle of the selecting signal is ∼3 ms, the signal was repeatedly applied during an accumulation time of 0.1 s to several seconds. The signal for the tickle was given from the same system described above. As shown in Figure 2, a timing sequence of the ion trap includes an enrichment and isolation (SWIFT and cooling) time that is typically 0.28 s and a fragmentation (tickle, typically100 kHz) period that is typically 0.01 s. The total time is typically 0.3 s. Ions dissociated into fragments were led to the TOF and mass analyzed. Then, the concentration of the analyte molecule was calculated from the total number of fragments originating from the analyte molecule. A series of these procedures was repeated multiple times during an 18-s period, and the integrated values were adopted as measured values. A MgF2 lens through which the VUV light is irradiated is used as a window that separates the VUV lamp from the ionization chamber. This lens with a focal length of 33 mm focuses VUV light on the center of the ion trap. It was found in laboratory tests that a transparency of the material deteriorated easily, in a typical case, by 50% after 7 days. The reasons for the deterioration have been identified, and the following countermeasures have been taken to improve the performance and longevity of the window. The observation of the surface of the deteriorated MgF2 lens indicated that there were two reasons. One was carbon adhesion on the surface facing the ion trap, and the other was discoloration of the MgF2 surface layers facing the VUV lamp. The effect of carbon adhesion is estimated as follows. When trace amounts of hydrocarbons contained in residual molecules in the ionization chamber came to the surface of the window, carbon adhesion occurred due to a photochemical reaction of hydrocarbons and VUV light. On the other hand, it was also found that carbon adhesion was not observed if water vapor was present inside the chamber, for instance, when actual waste incineration flue gases were measured. It is estimated that adhered carbons reacted with radicals of water vapor excited by VUV light to produce volatile carbon compounds. This presumption is based on the fact that VUV light at 10.2 eV is sufficiently higher than the threshold energy at 6.7 eV for OH radical production and is not enough high to produce H2O+ ion (ionization potential, 12.6 eV). This phenomenon is used as a cleaning procedure in which oxygen or water vapor is injected and supplied by design to the surface of the window during normal operation or maintenance service. It was observed that the discoloration of the MgF2 surface layers was caused by the change from MgF2 to MgF2-xOy. This change is estimated to be caused by the microwave-induced plasma or VUV light. On the other hand, it was also found that production rate of discoloration would be decreased by cooling the window. Consequently, the window and its vicinity are to be cooled during operation as a means of minimizing the formation of MgF2-xOy. A color center formation in the region of 260 nm in a near-surface layer of MgF2 under VUV irradiation has been 1010 Analytical Chemistry, Vol. 77, No. 4, February 15, 2005
Figure 4. Ion signal intensities versus concentrations of T3CB standard gases. The detection limit for a signal-to-noise ratio of 2 is 80 ng/m3‚N ()10 pptv), because the noise level is estimated to be 4.2, which is the ion signal at the zero point.
reported,17,18 and it was actually observed in this case. It was not estimated that the color center formation mainly caused the deterioration. Continuous VUV irradiation causes color center formation not only in a near-surface layer but also in the bulk of the crystal.18 Thus, saturation of the deterioration was not expected. However, it was found that the discoloration showed a tendency to saturate in laboratory tests after the elimination of carbon adhesion. Thus, it was estimated that the formation of MgF2-xOy was the main reason for the deterioration after the elimination of carbon adhesion. In addition to the countermeasures described above, a design in which eight windows can be used in turn without exposing the vacuum system to the air has been developed in order to prevent time loss to obtain high vacuum after the air exposure. This makes it possible for the system to run continuously maintaining a high level of sensitivity over six months. RESULTS AND DISCUSSION Laboratory Results. First, the system was calibrated. Because no commercial standard gas whose concentration is less than 1 ppbv was available, the sample gas for calibration was made from a standard gas (80 µL/m3‚N T3CB, Takachiho Chem. Indust. Co.) and a pure nitrogen gas (99.999%, Air Liquid Japan Ltd.). They were separately measured by two mass flowmeters to calculate the concentration of the sample gas for calibration by proportion. All gas line components (tubes, valves, mass flowmeters) were heated to 200 °C to minimize memory effects. Figure 4 shows the calibration line. The detection limit is estimated by the following. The count number of ions at the zero point of the X-axis was 4.2, which corresponds to the noise component that is suspected to come from the pure nitrogen gas, because the gas may contain impurities at less than 1 ppmv. The coefficient of the calibration curve was 0.052 (count number of ions)/(ng/m3‚N) with respect to the least-squares approximate straight line. Hence, the lower detection limit was determined to be 80 ng/m3‚N (10 pptv). At a concentration of 80 ng/m3‚N, the count number was found to be 8.4, which was sufficiently larger (17) Preston, R. C.; Brookes, C.; Clutterbuck, F. W. J. J. Phys. E. Sci. Instrum. 1980, 13, 1206-1213. (18) Zhukova, E. V.; Zolotarev, V. M.; Shishatskaya, L. P. Opt. Spectrosc. 1996, 81, 723-727.
Figure 5. Fragmentation pattern of m/z 180-184 on-line monitoring of T3CB in an actual exhaust gas. In addition to the fragments in the T3CB standard gas (indicated by solid line arrow marks), new fragments (indicated by dotted line arrow marks) are found.
than the noise (S/N ) 2). The longer the trapping time, the larger is the count number. However, excessively longer trapping time leads the accumulation of ions to a point of saturation at high concentration. In this case, the trapping time was settled to be 0.3 s. No saturation was found as shown in Figure 4. Integration time was 18 s. The photon fluence of VUV light has been reported as on the order of 1014 photons cm-2 s-1. Thus, it is estimated that the photon fluence of VUV light becomes ∼5 × 1013 photons cm-2 s-1 at the entrance of the ion trap because of the MgF2 lens with a transparency of 0.5 and focusing effect from 7- to 3-mm diameter. Sample gas is directly introduced into the interior of the ion trap at a maximum flow rate of 1.5 × 10-2 Torr L s-1. An ion trap has six small holes through which injected gases flow out. Thus, the pressure in an ion trap is higher than in the ionization chamber by 1 or 2 orders of magnitude. In case a buffer gas is nitrogen and a flow rate of sample gas is 1.5 × 10-2 Torr L s-1, it is estimated roughly that the total conductance of the six holes is 2 L s-1, and the pressure in the ion trap is 7.5 × 10-3 Torr. We assumed tentatively that the cylinder with a radius of 1.5 mm and a length of 14 mm (the gap length between two end gap electrodes) is defined as the photolysis volume in the ion trap. It is also assumed that the photon fluence in the photolysis volume is constant and homogeneous and the density of a gas is homogeneous. The amount of 10 pptv T3CB corresponds to 2.5 × 102 molecules in the photolysis volume. Therefore, the yield rate of T3CB ionization is estimated to be 0.13 ions s-1 by typical photoionization cross section of chlorobenzenes at 10.2 eV. This value is in good agreement with the experimental result described previously, because the detected 4.2 ions (net value) at 10 pptv correspond to 0.23 ions s-1. However, this discussion leaves efficiencies of trapping generated ions, efficiencies of TOFMS, and influence of the gas flow in the ion trap out of consideration. Especially the influence of gas flow is important because it causes the increase of the amount of molecules in the photolysis volume per unit time. On the other hand, the increase of generated ions may not cause the increase of trapping ions because a dynamic range of an ion trap is so narrow. Although we do not have sufficient experimental data for a quantitative discussion, further research would be necessary to improve the detection limit in future. First Field Application. The monitoring system was applied in a measurement of an actual exhaust gas at an outlet of a waste
Figure 6. Difference between the fragmentation pattern of T3CB sample gas with a pure N2 buffer gas (A) and with H2O-added N2 buffer gases (B). New fragments are found in (B) that are estimated to be caused by a reaction of H2O and T3CB.
Figure 7. Relative variation of the VUV light intensity through the MgF2 window. Deterioration of windows caused by VUV exposure tends to saturation.
incinerator furnace. Since the monitoring system was installed upstream of the aftertreatment unit, such as a dust collector, the combustion conditions inside the furnace could be monitored directly. After eliminating solid materials such as ashes, the exhaust gas was quickly cooled to ∼200 °C and led to the system. The fragments’ spectrum obtained by this procedure is shown in Figure 5. In addition to the same fragments as those originating from the standard gas, which are indicated by solid line arrows, new fragments indicated by dotted line arrows (m/z 127-130, 99, 101, 87, 89) were found. The actual exhaust gas is composed mainly of N2, CO2, and H2O, whose percentage was usually 1020%. Because the actual exhaust gas was injected directly into the ion trap without a buffer gas, the fragmentation pattern of T3CB was supposed to change. Since N2 and CO2 gases are stable, it is estimated that these fragments are produced by any reaction of H2O gas and T3CB. This assumption was confirmed by an experiment in which H2O was added to the standard gas. Figure 6 shows the difference Analytical Chemistry, Vol. 77, No. 4, February 15, 2005
1011
and all parts were overhauled, and any defect except windows was not found. This result proves that the system is robust and suitable to continuous monitoring. Figure 8 shows the results of measurements of T3CB in various conditions by this system and a standard measuring method (a GC/MS after over 4-h sampling and condensation by an absorbent) for comparison. The results by this system were in good agreement with the other. Therefore, this real-time measuring system is proven to have a high reliability in field applications compared with a standard measuring system. Moreover, this system is of great advantage because this technique needs only 18 s.
Figure 8. Comparison between the concentrations of T3CB measured by the VUV-SPI-IT-TOFMS and by a GC/MS. The data measured by both methods indicate good consistency.
between the fragment spectrum with a pure N2 buffer gas and with H2O-added N2 buffer gases. The same fragments as the new fragments in Figure 5 were observed in Figure 6. According to the figure, it was found that the sum of new fragment peaks and residual T3CB fragment peaks in Figure 6 equaled the sum of T3CB fragment peaks in Figure 5. Hereafter, the concentration of T3CB is calculated taking into consideration these fragments together. The fragments corresponding to other mass numbers are supposed to originate from interfering molecules except T3CB. Before a long months’ operation monitoring T3CB in an actual waste incineration flue gas, the greatest concern was endurance of an MgF2 window. However, it was a needless worry. Although this system had eight windows that could be used in turn without exposing the vacuum system to the air, in order to prevent time loss to obtain high vacuum after the air exposure, only three pieces of windows were exchanged due to periodic inspection during 7 months of operation. Typical characteristic change of transparency of the window is shown in Figure 7. The result indicated that deterioration of the window came to saturation by ∼75% transparency. All three pieces of windows showed the same saturation characteristics. If the system in the initial tuning condition is preset at 30% higher sensitivity, it can maintain the high level of sensitivity in the final condition. After 7 months of operation, the system
1012
Analytical Chemistry, Vol. 77, No. 4, February 15, 2005
CONCLUSION The VUV-SPI-IT-TOFMS system has been developed in order to monitor the emission of trace-level chlorinated organic compounds from an incinerator furnace. As a result, the sensitivity of 80 ng/m3‚N (10pptv) in 18 s for T3CB was achieved. In addition, this mass spectrometry system was applied to actual incineration plants by adding a process to eliminate interfering molecules and a means for extending the longevity of the irradiating windows. Because the actual exhaust gas with 10-20% H2O was used as a buffer gas in an ion trap instead of a pure N2 gas, it caused the change of fragment pattern of T3CB. However, it was found that the sum of new fragment peaks and residual T3CB fragment peaks equaled the sum of T3CB fragment peaks in case of a standard gas. Hereafter, the concentration of T3CB is calculated taking into consideration these fragments together. The results of measurements of T3CB under various conditions of actual combustion by this system were in good agreement with the results by a standard measuring method. The VUV-SPI-IT-TOFMS has proved to be robust and maintain a high sensitivity for on-line, real-time monitoring of T3CB in the first field test during 7 months of operation. ACKNOWLEDGMENT The authors gratefully acknowledge Professor Mitsuo Koshi (Tokyo University) and Associate Professor Kenichi Tonokura (Tokyo University) for useful advice and encouragement. Received for review August 20, 2004. Accepted November 12, 2004. AC048761Y
