ANALYTICAL TECHNOLOGIES

Using emission spectroscopy to perform impurity analyses in UHP gases

Mark L. Malczewski, Hollis C. Demmin, James D. Borkman, William A. Kilinskas, and Joseph Wegrzyn, Praxair

The last decade of the 19th century saw the discovery of all the noble gases.¹ The ability to liquefy air in large quantities made it possible to concentrate measurable quantities of these rare gases, but of equal importance was the use of a new technique: emission spectroscopy. When excited in an electric discharge, the gases dissipated some of the energy as electromagnetic radiation at wavelengths that were specific to each gas. The spectral lines were intense enough to be easily observable by the simple spectroscopes of the time and even by the human eye. The appearance of unfamiliar spectral lines in a gas sample indicated the presence of new elements, which greatly facilitated the discovery of previously unknown gases. Early on, the sensitivity of the technique for certain combinations of gases was noted, leading to the use of emission spectroscopy for quantitative analysis as well as qualitative identification.^2,3

Equipped with sophisticated electronics and detectors, today's spectroscopic analyzers can perform continuous, multicomponent analyses with limits of detection (LODs) in the single-digit parts-per-billion range for a number of impurities. As a result, these instruments have applications in the control of the ultra-high-purity (UHP) process gases used in the semiconductor industry. In addition to monitoring the purity of delivered gases, they can be used for point-of-use purity monitoring and for the certification of newly installed gas distribution systems. After discussing the operating principles of spectroscopic systems, this article presents some examples of these applications.

Spectroscopic Analyzer Operation

The heart of a spectroscopic analyzer is a pair of analytical tubes that contain the plasma undergoing excitation. In contrast to conventional low-pressure gas-discharge tubes, those using the silent electric discharge (SED) excitation technique operate at atmospheric pressure with high-voltage ac power supplied by two external electrodes. SED, which has been used for decades for monitoring nitrogen in argon,⁴ has several advantages. Operating at atmospheric pressure greatly simplifies sampling system requirements. Because the electrodes are external, they neither contribute to the spectral lines emitted by the plasma nor contaminate the UHP gas being analyzed. The technique is "soft" in the sense that it uses a low-energy discharge in comparison with that of an inductively coupled plasma source. Relatively few spectral lines are generated, yielding simple spectra for analysis and reducing the problem of spectral interference between impurities. Also, there is no interconversion between chemical species such as carbon monoxide and carbon dioxide, which allows simultaneous measurement of various impurities. However, the technique has a serious disadvantage in that several impurities of interest, such as hydrogen, are not excited by SED and therefore cannot be analyzed directly. Thus, while ideal for a number of measurements, the technique is limited to impurities that can be excited in the SED.

An example of a portion of the spectral output of an argon SED is shown in Figure 1. Note that there is a continuum of output from the sample gas in addition to the sharp emission lines from various impurities. This emission background becomes increasingly important at the sensitive 0—100-ppb range and must be compensated for in the zero offset of the detector electronics.

Figure 1: Multiple impurity emission lines excited in an argon silent electric discharge.

Two detection and signal-processing techniques are shown schematically in Figure 2. In both types of system, photomultipliers (PMTs) convert the light output from various impurities to a current. Narrow-bandpass optical filters isolate the strongest emission line from each impurity, making respective PMTs specific for that impurity. An analyzer can accommodate as many as four PMTs, one at each end of the two analytical tubes. The detectors and electronics for each impurity are independent of one another, which provides added reliability.

Figure 2: Schematics comparing the (a) conventional emission spectroscopy frequency-modulation technique with (b) electro-optical modulation.

Signal Modulation. To improve the signal-to-noise ratio, the PMT signals are modulated. The detector electronics selectively amplify signals at the modulation frequency, minimizing unwanted noise and drift. In conventional emission spectroscopy, a chopper wheel is used to interrupt the light from the discharge to the PMT, as shown in Figure 2a.⁵ This modulation technique works well when the light output of the discharge source is constant over time, such as in the case of an incandescent solid. However, an ac-driven SED does not provide a constant output; rather, it turns on and off at twice the frequency of the excitation voltage. This cycling impresses an inherent modulation frequency on the output of the light source, which is in addition to the modulation frequency from the chopper. The resulting PMT output is a combination of the two modulation frequencies, which limits the chopper's usefulness.

To overcome this problem, a recently developed system known as the UltraSpec2000 (Praxair, Tonawanda, NY) uses the inherent modulation of the gaseous discharge by precise-ly controlling the frequency of the high-voltage excitation signal; thus the desired modulation frequency is generated without the use of the mechanical chopper. Represented schematically in Figure 2b, this patented method is referred to as electro-optical modulation.⁶ The analyzer's variable-frequency oscillator can drive the high-voltage transformer at any desired input voltage and frequency. This frequency is then doubled and used by the detector electronics to selectively amplify the PMT signal. If a specific application warrants the use of a different excitation frequency, the detector electronics will automatically return to the new frequency.

Compared with modulation using a chopper, electro-optical modulation provides a greater reduction in noise and greater light throughput, resulting in higher signal levels. A direct comparison between the two techniques can be seen in Figure 3, which shows the results when the same UltraSpec analyzer, using the same gas standards, was operated with and without a conventional mechanical chopper. The high signal gains achievable with PMTs coupled with the noise rejection of frequency-selective amplification make possible single-digit parts-per-billion LODs for several key impurities. The analyzer's calibration curves for the 0—1000- and the 0—100-ppb ranges for nitrogen, moisture, methane, and carbon dioxide in argon are presented in Figures 4 and 5, respectively. The range for each impurity can be set independently, according to the composition of the sample gas.

Figure 3: Comparison of signal-to-noise ratios for conventional and electro-optical frequency-modulation techniques.

Figure 4: Calibration curves for various impurities in argon using a 0—1000-ppb full-scale range: (a) nitrogen and moisture, and (b) methane and carbon dioxide. In all cases the flow rate was 1.380 std cm³/min.

Figure 5: Calibration curves for various impurities in argon using a 0—100-ppb full-scale range: (a) nitrogen and moisture, and (b) methane and carbon dioxide. In all cases the flow rate was 1.380 std cm³/min.

Other Instrument Features. In addition to its use of electro-optical modulation, the UltraSpec2000 offers such benefits as rapid response and microprocessor control. The analyzer has a low-dead-volume heated sampling system that helps speed up response times, particularly for moisture. For example, its response time—which is fast—is compared with that of an atmospheric pressure ionization mass spectrometer (APIMS) for moisture in argon in Figure 6.

Figure 6: Comparison of response times of the UltraSpec2000 and an APIMS system for moisture in argon.

A high-quality sampling system and a sensitive analytical technique are critical for an accurate analysis. Once these attributes are in place, a microprocessor control feature offers an easy-to-use operator interface. Two important capabilities provided by the microprocessor are data communication to host computer systems and validation of the analytical data.

Data Communication. A recent trend in analytical instrumentation has been to provide serial outputs. Such serial communication has several advantages over analog outputs, including the ability to easily transmit a large number of process variables to a host system; the ability to perform monitoring, troubleshooting, and control via a remote modem; the elimination of the analog-to-digital and digital-to-analog conversion errors that can occur when passing data between an analyzer and its host system; and a reduction in installation costs resulting from the use of a single communications link rather than a discrete pair of wires for each process variable.

The analyzer has two serial ports that are mirror images of one another; either can be used for monitoring or control. One port is RS-232C compatible and the other can be configured by the user as either RS-232C or RS-485. The analyzer can output data at preset user-defined intervals or on demand from the host system, facilitating data transfer to the user's facility management system. In addition, a limited amount of data are shown on the instrument's rear-panel analog and discrete digital readouts. Impurity analyses, alarms, range information, and a series of status outputs are available.

Data Validation. The accuracy of impurity concentration data is a critical concern if the real-time analyses are used to decide whether to begin a processing step on a boatload of wafers or if they are part of a closed-loop process control scheme. Manufacturers cannot afford to change process parameters based on faulty analytical data. To ensure that data are valid the analyzer microprocessor should perform self-diagnostics as part of its data-scanning routine.

A portion of the analyzer's self-diagnostic flowchart is shown in Figure 7. The diagnostic routine runs as part of the scanning routine so that each new set of impurity analyses is validated. Status inputs provide the microprocessor with the information necessary to answer the questions in the flowchart's various decision boxes. Because providing a complete set of status inputs is the key to successful diagnostics, an analyzer design must include whatever sensors are necessary to provide the required data. Each required input must either be measured directly or inferred unambiguously from other status inputs that are available. To avoid a misdiagnosis that is caused by a faulty sensor, part of the diagnostic routine is to verify proper sensor operation before a particular status input is used. This involves either checking for an out-of-range sensor output or cross-checking each sensor's output against that from another sensor. Examples of these methods can be seen at points A and B, respectively, in Figure 7.

Figure 7: Portion of flowchart outlining the Ultra Spec2000's self-diagnostic routine.

Potential Applications

To demonstrate two potential applications of the UltraSpec2000, we conducted various experiments. In the first, the analyzer was used to evaluate the outlet-purity performance of a rare gas POU purifier challenged by a flow rate exceeding its rated capacity. Situations may arise in the fab where the installation of additional tools overtaxes POU purification systems. The experiment demonstrated how the analyzer can help users clearly understand how flows higher than specification levels will affect impurity concentrations at the purifier outlet.

The purifier studied was a used, heated alloy-based system. Data were collected using the analyzer calibrated for nitrogen and moisture measurement in dewar-supplied argon. Use of a bypass maintained constant flow and pressure at the analyzer under all test conditions. The inlet challenge concentration of nitrogen was 250 ppb. Moisture was removed from the supply stream via a molecular-sieve dryer to simulate process argon that had been purified using a catalytic system. This purification technology would not remove trace amounts of nitrogen or methane from the bulk supply. Figure 8 shows nitrogen and moisture concentrations at the outlet as a function of total flow across the purifier bed. While moisture levels remained fairly constant, for each step-function increase in flow there was a corresponding increase in nitrogen concentration. These results clearly indicate there was a loss of purifier effectiveness when flow-rate specifications were exceeded.

Figure 8: Nitrogen and moisture concentrations measured by the analyzer at a POU purifier outlet as a function of flow rate.

The second analyzer application studied was simulating the certification of gas distribution system tubing. Because argon is widely used as a shielding and purge gas during piping installation work, the ability of the analyzer to measure impurities in argon fits well with purity measurement requirements. However, use of the analyzer in this application is only as successful as the sampling method used to determine system cleanup and system certification on weld gas prior to introducing process gases. Thus, two different sampling approaches were evaluated in our study.

In the first of these experiments the analyzer was used to monitor the cleanup of a newly installed 10-ft length of 1/4-in. electropolished-stainless-steel tubing which was sampling from a high-purity argon system that had been under argon purge for several days. The gaseous flow rate through the entire sample tubing length was approximately 3 std cu ft/hr. The results, shown as dotted lines in Figure 9, indicate the times that can be expected to thoroughly purge a severely air-contaminated line of the selected impurities under these conditions. It is interesting to note the staged response of moisture, methane, carbon dioxide, and nitrogen. This variation was expected given the relative atmospheric concentrations of these constituents in the air, except for moisture, which responds to drydown based on chemisorption rather than physicosorption properties.^7,8

Figure 9: Comparison of cleanup times for a poorly purged and a well-purged sample point, measured by the UltraSpec2000.

A second set of test data, depicted by the solid lines in Figure 9, was collected for a sampling system that had been prepared more rigorously. In this case, the sample line had been purged and capped prior to its connection to the piping system. The outlet vent of the instrument was also capped during instrument transfer to avoid back-diffusion contamination of the analyzer cell. In this instance, cleanup and drydown were achieved considerably sooner than with the previous sampling approach, which demonstrates the importance of proper sample handling to ensure rapid equilibration. This second experiment also provides a more representative example of the analyzer's response time under expected conditions in a semiconductor fab, suggesting that the use of this instrument for line certification after weldment could shorten the time required for analysis compared with conventional certification methods. Even with the poorly purged sample line in the first test, the analyzer recovered to parts-per-billion levels within a reasonable time. However, Figure 9 clearly demonstrates that proper purging can greatly shorten the time required for certification.

Conclusion

Emission spectroscopy can be used to perform multiple-impurity analyses of semiconductor gases based on the spectral lines that are specific to each impurity. The analyzer described in detail above, which uses electro-optical modulation to achieve parts-per-billion LODs, can be used to monitor a number of impurities of concern in UHP argon and helium, and efforts are under way to extend its capabilities to additional impurities as well as to the analysis of other process gases. Potential applications for the system include purifier monitoring and gas distribution system post-installation certification.

Acknowledgments

The authors wish to acknowledge the support of Turner Design Associates, which provided design and fabrication expertise during the development of this analyzer.

References

1. Davis HM, The Chemical Elements, Science Service, pp 142—150, 1959.

2. Travers MW, The Experimental Study of Gases, London, Macmillan, 1901.

3. Argon, Helium and the Rare Gases, vol 2, Cook GA (ed), Interscience Publishers, pp 506—516, 1961.

4. Fay H, et al., "Emission Spectrometric Method and Analyzer for Traces of Nitrogen in Argon," Analytical Chemistry, 34(9):1254, 1962.

5. Fay H, Mohr PH, and Cook GA, "Emission Spectrometer," U.S. Pat. 3,032,654, 1962.

6. Malczewski M, Demmin HC, Brown DE, and Wiltse DR, "Gas Emission Spectrometer and Method," U.S. Pat. 5,412,467, 1995.

7. Brimblecombe P, Air Composition and Chemistry, Cambridge, UK, Cambridge University Press, p 2, 1986.

8. Kubus JM, and Leggett GH, "Piping System Drydown Predictions with Field Verification," in Proceedings of the 1993 Microcontamination Conference, Santa Monica, CA, Canon Communications, pp 212—221, 1993.

Mark L. Malczewski, PhD, is a development associate at Praxair, Tonawanda, NY. His research interests include particulate measurement in gases other than air or nitrogen and design of automated, continuous monitoring systems for particulate and gaseous impurities in UHP bulk gases. In addition, he contributed to the design of the UltraSpec2000 as well as to research on extension of the method to measure impurities in other base gases. Malczewski earned both his BS and PhD in chemistry from the State University of New York at Buffalo. He is a member of the American Chemical Society and ASTM. (Malczewski can be reached at 716/879-2525.)

Hollis C. Demmin is also a development associate at Praxair, where he has been responsible for the design and development of Praxair's automated, continuous monitoring systems for gaseous and particulate measurements in UHP bulk gases. He is a member of the National Council of Standards Laboratories and hold an associate's degree in electrical technology from Erie Community College.

James D. Borkman is an analytical systems manager for Praxair's turnkey electronics business. His responsibilities include product management of the company's advanced analytical systems as well as supporting project management of analytical installations for new semiconductor fabs. A member of various SEMI standards committees, he holds a BA in chemistry from Buffalo State College and an MBA from Canisius College. (Borkman can be reached at 716/879-2494.)

William A. Kilinskas is a senior development associate in process and systems R&D at Praxair. His research interests include software design and programming of analytical instrumentation and continuous analytical instrument data collection and analysis systems. He holds a BS in physics from the University of Dayton and an MS in physics from Ohio University.

Joseph Wegrzyn is a technologist in Praxair's R&D group, where he is responsible for final assembly and testing of analytical instrumentation. He recently participated in the mechanical design and development of the UltraSpec2000 for measurement of trace impurities in UHP gases by emission spectroscopy. Wegrzyn's studies have also focused on mechanical and electrical technology.

MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit