Anal. Chem. 1998, 70, 1802-1804
Use of a Clustering Reaction To Detect Low Levels of Moisture in Bulk Oxygen Using an Atmospheric Pressure Ionization Mass Spectrometer A. D. Scott, Jr., E. J. Hunter, and S. N. Ketkar*
Air Products and Chemicals, Inc., 7201 Hamilton Boulevard, Allentown, Pennsylvania 18195
Atmospheric pressure ionization mass spectrometry (APIMS) is being routinely used to quantify trace impurities in bulk gases used in the manufacture of semiconductor devices. APIMS has been successfully applied for the quantification of ppt levels of O2, H2O, CO2, and CH4 in Ar, N2, and He. However, it has not been successfully used to quantify trace impurities in bulk O2 due to the low ionization potential of O2. APIMS relies on chargetransfer reaction between the ions of the bulk gas molecules and impurity molecules. Since all the relevant impurity molecules have ionization potentials higher than that of O2, APIMS has not been used to analyze for impurities in O2. We report here the detection of subppb levels of H2O in O2 by making use of the clustering reaction between O2+ and H2O. The declustering region in an APIMS, which is normally used to break apart unwanted and interfering clusters, has to be carefully adjusted to keep intact the weakly bound cluster O2+‚H2O. Our results indicate a statistical detection limit of less than 300 ppt for the detection of H2O in O2. Atmospheric pressure ionization mass spectrometry (APIMS) is being routinely used to quantify sub-ppb level impurities in inert gases being used in semiconductor processing. This technique is capable of achieving detection limits in the ppt range for O2, H2O, CH4, and CO2 in H2, N2, Ar, and He.1 In the ionization source of an APIMS, a corona discharge operating at atmospheric pressure is used to create the primary ions of the gas being sampled. At such a high pressure, the mean free path is small and ion/molecule reactions play a key role in ionizing the trace impurity atoms/molecules present in the sample gas. For impurities that can undergo fast charge-transfer reactions (rate constants of the order of 10-9 cm3/s) with the sample gas ions, APIMS can be used very effectively. This includes O2, H2O, CH4, and CO2 in bulk N2, Ar, and He. If the sample gas is H2, proton-transfer reactions play a key role in the secondary ionization mechanism. In this case, the primary ion is H3+ and impurity molecules that can undergo fast proton-transfer reactions can be detected as the protonated molecule. Impurities that can be detected include N2, H2O, CO2, and CO in H2. Cluster ions produced in the atmospheric pressure reaction region usually interfere with the detection of impurities and also lower the analytical sensitivity of the (1) Siefering, K.; Berger, H.; Whitlock, W. J. Vac. Sci. Technol. A 1994, 11, 4.
1802 Analytical Chemistry, Vol. 70, No. 9, May 1, 1998
technique. These cluster ions are usually dissociated in the declustering region.2 APIMS has not been successfully used to analyze for trace impurities in bulk O2 because O2 has a fairly low ionization potential and consequently O2+ cannot undergo charge-transfer reactions with impurities of interest to the semiconductor industry. APIMS spectra for O2 have been reported earlier.3,4 Siegel and Fite3 reported an APIMS spectrum of moderately dry air dominated by O2+ and cluster ions including O2+‚H2O, whereas Kambara and Kanomata4 reported an APIMS spectrum of 99.99% pure O2 (which was purified using a molecular sieve trap and a quartz combustion tube). The reported spectra were dominated by O2+ and clusters, including O2+‚H2O. Due to the large amount of moisture in the O2 sampled by Kambara and Kanomata, a rather large H3O+ peak was also present. They were able to dissociate the weakly bound cluster, O2+‚H2O, by changing the declustering voltage. They used H3O+ as an indicator of moisture in the highppb and ppm level. Advances in purification technology coupled with the use of ultrahigh-purity distribution systems have made it possible to deliver O2 to semiconductor fabrication facilities with impurity specification below 10 ppb. Dedicated moisture analyzers and GC techniques are used to certify compliance with the specification at the 10 ppb level. With the shrinkage of design rules in the semiconductor fabrication industry, the specifications for O2 are going to be lowered to the 1 ppb level and below. We report here the use of an APIMS, operating with a weak declustering field, to quantify moisture in oxygen at the sub-ppb level. EXPERIMENTAL SECTION A VG APIMS coupled to an Air Products dilution system was used for the experiments. The APIMS uses a single quadrupole mass spectrometer coupled to a corona discharge source operating at atmospheric pressure. The ions produced in the corona discharge, together with the sample gas being analyzed, flow through a 100-µm aperture into a low-pressure region maintained at a pressure of about 1 Torr.2 A potential gradient can be applied in this region to help facilitate the breakup of clusters which are generated in the atmospheric pressure corona. The dilution system uses dynamic dilution and is capable of generating dilution (2) Ketkar, S. N.; Dulak, J. G.; Fite, W. L.; Buchner, J. D.; Dheandhanoo, S. Anal. Chem. 1989, 61, 260. (3) Siegel, M. W.; Fite, W. L. J. Phys. Chem. 1976, 80, 2871. (4) Kambara, H.; Kanomata, I. Anal. Chem. 1977, 49, 270. S0003-2700(97)01273-0 CCC: $15.00
© 1998 American Chemical Society Published on Web 03/19/1998
Figure 1. APIMS spectrum of purified oxygen.
Figure 4. Change in the APIMS spectra resulting from the addition of 5.1 ppb of moisture to purified oxygen.
Figure 5. Change in the APIMS response at m/z ) 50 for changing concentrations of added moisture.
Figure 2. Effect of declustering voltage on the intensity of cluster ion peaks.
via the following reactions:6
O2 + e- f O2+ + 2eO2+ + O2 + O2 f O4+ + O2 O2+ + H2O + O2 f O2+‚H2O + O2 O2+ + O3 + O2 f O5+ + O2 O3 is formed by the neutral reaction O + O2 + O2 f O3 + O2
Figure 3. APIMS spectrum of purified oxygen with 5.1 ppb of added moisture.
ratios as high as 10000:1.5 The oxygen used for the experiments was obtained from a liquid source and further purified using a catalytic bed purifier followed by a molecular sieve bed.
RESULTS AND DISCUSSION Figure 1 shows a mass spectrum of the purified oxygen with a declustering voltage of 20 V. The base peak is at m/z ) 32 with smaller peaks at m/z ) 50, 64, and 80. These peaks occur (5) Ketkar, S. N.; Scott, A. D., Jr.; Martinez de Pinillos, J. V. J. Electrochem. Soc. 1994, 141, 184.
The cluster peaks at m/z ) 50 (O2+‚H2O), m/z ) 64 (O4+), and m/z ) 80 (O5+) are weakly bound and can be easily broken up in the declustering region. Figure 2 shows the effect of the declustering voltage on the intensity of these peaks. As the declustering voltage is increased, the intensity of these cluster peaks decreases. From the rate of decrease it is clear that the water cluster O2+‚H2O is somewhat tightly bound compared to the oxygen clusters O4+ and O5+. On reducing the declustering voltage below 20 V the entire spectrum vanished, presumably due to scattering losses in the declustering region and the loss of optimal tuning on the lenses used to carry the ions to the entrance of the quadrupole mass spectrometer. On addition of moisture to the oxygen stream being sampled by the APIMS, the water cluster peak increases in intensity. Figure 3 shows the APIMS spectrum obtained after the addition (6) Albritton, D. L. At. Data Nucl. Data Tables 1978, 22, 1.
Analytical Chemistry, Vol. 70, No. 9, May 1, 1998
1803
Figure 6. Calibration curve for m/z ) 50 as a function of added moisture. Figure 7. Results of the monitoring of H2O in O2 for a period of 33 hours.
of 5.1 ppb of moisture to purified oxygen, and Figure 4 shows the difference spectrum between this and the spectrum of purified oxygen. On the addition of moisture, the peak at m/z ) 50 increases, and as shown in Figure 4, this increase is almost exactly compensated by a corresponding decrease in the peak at m/z ) 32 (O2+). This indicates that our identification of m/z ) 50 as a water cluster peak, O2+‚H2O, is indeed correct. Figure 5 shows the response of the peak at m/z ) 50 to different levels of added moisture. As can be seen in the figure, the peak at m/z ) 50 responds to changing moisture concentration and the APIMS can easily detect changes of 1 ppb in the level of moisture in oxygen. Figure 6 shows the linear calibration curve for moisture in oxygen at m/z ) 50. The statistical LOD based on this calibration is 0.27 ppb and the background moisture level in the purified oxygen is 3.8 ppb. Figure 6 shows the results on the monitoring of a stream of purified O2 for a period of approximately 33 h. The slow variations in the H2O concentration are due to the variation of ambient temperature in the laboratory, which will change the outgassing of moisture from the rather long sampling lines. The average moisture concentration during this time period was 2.6 ppb with a standard deviation of 0.07 ppb. The maximum and minimum values were 2.8 and 2.4 ppb, respectively.
1804 Analytical Chemistry, Vol. 70, No. 9, May 1, 1998
At higher concentrations of moisture, a peak at m/z ) 19 (H3O+) starts to appear. This is the ion that was used by Kambara and Kanomata as an indicator of moisture in the high-ppb to ppm range.4 The peak arises from the following reaction
O2+‚H2O + H2O f H3O+ + OH + O2 The protonated molecule should exhibit a second-order response to increasing moisture concentration, as shown in ref 4. CONCLUSION This work has shown that, despite the low ionization potential of O2, an APIMS can be used to quantify sub-ppb levels of moisture in oxygen. It is accomplished by monitoring the water cluster ion O2+‚H2O. This cluster ion is fairly weakly bound and care has to be exercised so as not to dissociate this cluster in the declustering region of the instrument. Received for review November 19, 1997. January 22, 1998. AC971273O
Accepted
