Developing advanced humidity standards to measure trace water vapor in specialty gases
Joseph T. Hodges and Gregory E. Scace,
National Institute of Standards and Technology (NIST)
Surface-adsorbed water and water vapor in gas streams can be important contaminants in the ultraclean gas-handling environments required for semiconductor manufacturing. In applications in which high-purity conditions must be maintained, water is difficult to monitor and control for two reasons: It is ubiquitous in the atmosphere and on surfaces that have been exposed to the atmosphere, and it has physical and chemical properties that make it difficult to eliminate. Small quantities of water vapor (below 1 nmol mol–1) can adversely affect the performance and yield of silicon-based semiconductors and compound semiconductors used in photonic devices. Processes affected adversely by water include sample preparation, in which high-purity purge gases are used to sweep atmospheric constituents or residual compounds from prior processes, and reactive flows, in which high-purity gases are used for chemical incorporation in film growth.
Typically, point-of-use purification systems are used in conjunction with high-purity gas supplies to ensure that water is not introduced into the process stream. However, in-line methods for monitoring water vapor concentration are required to ensure that low water vapor levels are maintained downstream of purified source gases. Hence, field-deployable humidity-generation devices and humidity analyzers must be available for specialty gas suppliers and end-users. Furthermore, these systems must have the requisite sensitivity and stability to provide meaningful results, and they must have response capability that can be calibrated based on the primary methods of humidity generation.
The National Institute of Standards and Technology (NIST; Gaithersburg, MD) has developed both primary and secondary humidity-generation standards to address the needs of the semiconductor industry. The former have a thermodynamic basis, while the latter are customer-portable artifacts whose calibration is based directly on the NIST primary standard humidity generator.
Primary Standard Humidity Generator
The low-frost-point humidity generator (LFPG) is a NIST primary standard humidity generator that produces trace quantities of water vapor in inert carrier-gas streams. It is based on the idea that at thermodynamic equilibrium, the concentration of water vapor in a carrier gas that is exposed to an ice surface depends only on total system pressure (P) and saturation temperature (T). At the same time, that concentration is independent of gas flow rate. Specifically, the partial pressure of water vapor is equal to the product of the vapor pressure of ice, ew(T), and a thermodynamic quantity called the enhancement factor, f(T,P), that is close to unity and accounts for nonideal gas and nonideal mixing effects. This gives
where xw is the mole fraction of water vapor in the carrier gas. Based on thermodynamic arguments, combined with measurements of the latent heat of water vaporization and the specific volume of saturated water vapor, ew(T) has been determined over a wide range of temperatures.1
The realization of water vapor mole fractions below 100 nmol mol–1 requires operation at temperatures well below the freezing point of water. For example, according to the equation above, xw = 14 nmol mol–1 and 96 nmol mol–1 for T = –100°C and –90°C, respectively. This corresponds to a change in xw of approximately 20% per degree centigrade, implying that 0.2% precision in the output water vapor concentration would require saturator temperature control at the 0.001°C level. The LFPG was designed to perform to these specifications, and a detailed description and uncertainty analysis for the system have been presented previously.2,3
The LFPG consists of an isothermal saturator operated at pressures ranging from 100 to 300 kPa. It contains a 4-m-long ice-coated channel through which the inert carrier gas (usually nitrogen) flows. Volumetric flow rates up to 4 std L/min can be accommodated. Saturator temperature can be adjusted to less than –100°C using a fluid-based refrigeration system. The temperature is measured using standard-grade platinum resistance thermometers and is actively controlled using a set of thermoelectric heating/cooling elements with a repeatability of 0.001°C. Combined uncertainty in the saturator temperature is less than 0.01°C. Likewise, the absolute gas pressure in the saturator is controlled to within 50 Pa, so that its contribution to the overall uncertainty budget is negligible. By operating the saturator at elevated pressure, the LFPG can produce water vapor mole fractions as low as 3 nmol mol–1 with a standard relative uncertainty of 0.4%.3
Finally, for the flow rates of interest, high-precision measurements indicate that the mole fraction produced by the LFPG is essentially independent of flow rate, which is consistent with the assumption that there is thermodynamic equilibrium in the saturator.
The LFPG output water vapor mole fraction can be controlled with a precision well below 1 nmol mol–1, although long equilibration times are required. To illustrate this performance, stepwise changes of 0.2 nmol mol–1 in the LFPG output water vapor mole fraction were monitored using a calibrated quartz-crystal microbalance (QCM). The baseline-corrected results (with drift of <0.5 pmol mol–1 h–1) are expressed by the symbols in Figure 1. Exponential fits to the data, expressed by the solid line, reveal time constants for decay and growth that are both approximately equal to 10 hours. These results indicate a precision of <50 pmol mol–1 in the water vapor mole fraction for equilibration times in excess of 24 hours, a result that is consistent with the temperature stability of the LFPG saturator.
The LFPG serves as the standard source of trace water vapor for a variety of commercial and custom humidity analyzers. Calibration of reference-grade hygrometers, which are used as transfer standards and for prototype testing, are common applications. Analyzer types include chilled-mirror hygrometers, QCMs, atmospheric-pressure ionization mass spectrometers (APIMS), and oxide and electrolytic humidity sensors. Also, relatively new instruments based on absorption spectroscopy of water vapor, such as Fourier-transform infrared spectrometers and diode-laser-based absorption spectrometers, have been characterized.
Customers often wish to have their analyzers calibrated directly and are interested in their steady-state response to water vapor mole fraction. However, more-detailed information can be extracted from special tests that yield information about analyzer repeatability, sensitivity and linearity, hysteresis, response time, residual-flow-rate dependence, zero-drift, and residual water vapor background. Such studies provide customers with quantitative data and provide a common link to the invariant thermodynamic properties of water.
Portable Humidity-Generation Transfer Standards
Field-deployable humidity-generation standards complement humidity analyzers by enabling on-site calibration and the testing of analyzer response. To that end, permeation tube humidity generators (PTGs) are commonly used for the precise production of calibration gas-mixture streams that contain trace quantities of water vapor. PTGs are simple devices that are based on the controlled dilution of water vapor by a purified carrier-gas stream. They consist of a liquid-water-filled permeation tube (PT) from which water vapor diffuses, a temperature-regulated chamber that houses the PT, a purified carrier-gas supply, and one or more mass-flow controllers for dilution of the water vapor stream. Since flow rates can be changed quickly and repeatably, PTGs exhibit fast response times and high stability. They are also small, transportable, and relatively inexpensive, often costing less than many of the analyzers they calibrate.
PTGs have gained broad acceptance within the semiconductor industry as reliable devices for producing reference-gas streams that contain water vapor mole fractions as low as 1 nmol mol–1. Specialty gas suppliers use PTGs for calibrating a wide range of trace-humidity analyzers, including APIMS and other point-of-use analyzers that monitor water contamination in silicon wafer processing facilities. Analyzer manufacturers rely on PTGs for their in-house metrology programs and even install PTGs inside humidity analyzer instruments as in situ reference-gas streams with a fixed water vapor mole fraction.
Although PTG systems provide a precise and repeatable way to generate water vapor mole fractions at nmol mol–1 levels, they are subject to large systematic uncertainty. In a study comparing the performance of commercial PTG systems (in the range 10 to 100 nmol mol–1), PTG supplier calibration errors of 10 to 30% were measured.4 While such errors have raised doubts about PT permeation rates, it is important to note that PT permeation rates must be measured because they cannot be predicted reliably from first principles. Usually they are measured gravimetrically—a tedious approach with significant uncertainty of
≈10%, since low permeation rates (typically in the range 30 to 300 ng min–1) are considered. Furthermore, permeation rates must be determined at the temperature(s) at which the PT is operated, complicating the implementation of gravimetric-based permeation-rate measurements.
Figure 1: Quartz-crystal microbalance response over time. The symbols represent the baseline-corrected results, while the solid line represents exponential fits to the data, revealing time constants for decay and growth that are both approximately equal to 10 hours.
While recognizing PTs’ value as portable devices for producing reference-gas streams with low water vapor mole fractions, NIST investigators, prompted by the instrument’s limitations, developed a precise method for calibrating customer PTs directly using the LFPG as a primary standard reference. This technique exploits the extraordinarily high precision with which LFPG output water vapor concentration can be controlled. Figure 2 shows that the NIST method includes an LFPG, a purified-nitrogen carrier-gas stream, a temperature-stabilized oven that contains the customer-supplied PT, a laminar flow element (LFE) for flow metering, high-purity mass-flow controllers (MFCs), and a commercial water vapor analyzer. The standard relative uncertainty of the LFE is 0.25%, making flow measurement uncertainty negligible. The system is designed to minimize dead legs and virtual sources of water vapor that might otherwise create large background signals.
Figure 2: Schematic diagram of the NIST method for calibrating customer PTs directly using the LFPG as a primary standard reference.
Figure 2 also illustrates the flow arrangement, in which the respective flow streams of the flow-dilution system and the LFPG are alternately sampled by the analyzer. The LFPG output is adjusted to produce an analyzer response that is nearly indistinguishable from the analyzer response to the PTG system. With this comparative method, sensitivity to the nonlinearity, absolute response, and zero drift of the analyzer is eliminated. The ability to null the respective signals is usually limited by the signal-to-noise ratio of the analyzer response. Using a commercial QCM system as the comparative analyzer yields a resolution approximately equal to 50 pmol mol–1 in water vapor mole fraction.
Five sets of measurements for a single permeation tube were carried out over a nine-month period to quantify the repeatability of the calibration technique. Each set corresponds to six measurements that cover the water vapor mole fraction range of 10 to 100 nmol mol–1 at a permeation tube temperature equal to 50°C. The results of these measurements are summarized in Figures 3a and 3b.
Figure 3: (a) Permeation rate derived by fitting the data to a simple linear model in which the LFPG output and the reciprocal of the PTG system dilution-gas flow rate were independent variables, and (b) the fit residuals for each data set.
Figure 3a shows the permeation rate that was derived by fitting the data to a simple linear model in which the LFPG output and the reciprocal of the PTG system dilution-gas flow rate were independent variables, while Figure 3b shows the fit residuals for each data set. The permeation rate inferred from these measurements is 36.5 ng min–1 with a relative standard deviation of 0.5%, and the standard deviation of the fit residuals shown in Figure 3b is <0.35 nmol mol–1. Notably, these results represent a nearly tenfold improvement over gravimetric-based determinations of the PT permeation rate.
Portable water-vapor-generation standards can be used to characterize spectroscopic analyzers, which measure the absorption coefficient of the water vapor/carrier-gas sample. These instruments have the great potential of providing accurate measurements of water vapor concentration with high sensitivity.5 Compared with analyzers that rely on the physical and/or chemical interactions of water with other materials, spectroscopic methods offer responses that can generally be modeled with high confidence. When an appropriate physical model is used, the response can be characterized by measuring a known sample concentration and by inferring the strength of an individual absorption transition or band of transitions.
Based on this method, NIST researchers have developed a high-resolution variant of cavity ring-down spectroscopy (CRDS) called frequency-stabilized cavity ring-down spectroscopy (FS-CRDS). This method measures individual water vapor transition line shapes and line strengths.6 The FS-CRDS technique was used to measure trace water vapor concentration in high-purity phosphine and to make high-resolution line-shape measurements of water vapor.7,8 Like the vapor pressure of ice on which the LFPG thermodynamic standard is based, line strengths are a property of the water molecule. In practice, these measurements are directly linked to the LFPG using LFPG-calibrated PTGs as sources of water vapor for CRDS sample cells. These data are generally applicable to other spectroscopic methods that probe the same spectral features of water vapor.
Water Vapor Background
Sinks and virtual sources of water in flow manifolds become increasingly problematic and limiting as source-gas purity improves. Indeed, it is impossible to suppress these background water effects in gas manifolds completely. However, adherence to certain standard practices can reduce their influence on the sample water vapor concentration significantly. These standards include the use of low-roughness all-metal tube walls and seals, welded bends and tees, the complete avoidance of elastomer components, and the use of well-designed flow networks with purging capabilities and minimal stagnant volumes or dead legs. Measurement practitioners must be aware that the performance of water vapor generators and analyzers cannot be completely decoupled from the manifolds that connect system components. Quick-and-dirty measurements of trace humidity are a recipe for failure. Invariably, they will yield ambiguous results.
As technical demands in the semiconductor manufacturing industry evolve, measuring the purity of electronic specialty gases will become increasingly challenging, especially when contamination by water vapor is involved. Standardized test and calibration methods such as those presented here can help ensure measurement quality and enable the development of improved measurement techniques.
The work described in this article was supported by the NIST Office of Microelectronics Programs.
1. A Wexler, “Vapor Pressure Formulation for Ice,” Journal of Research of the National Institute of Standards and Technology 81A, no. 1 (1977): 5–20.
2. GE Scace et al., “The New NIST Low Frost-Point Humidity Generator,” in Proceedings of the 1997 Workshop and Symposium of the National Conference of Standards Laboratories (Boulder, CO: NCSL International, 1997), 657–673.
3. GE Scace and JT Hodges, “Uncertainty of the NIST Low Frost-Point Humidity Generator,” in Proceedings of the International Symposium on Temperature and Thermal Measurements in Industry and Science (TEMPMEKO 2001) (Berlin: VDE Verlag, 2002), 597–602.
4. PH Huang and R Kacker, “Repeatability and Reproducibility Standard Deviations in the Measurement of Trace Moisture Generated Using Permeation Tubes,” Journal of Research of the National Institute of Standards and Technology 108, no. 3 (2003): 235–240.
5. HH Funke et al., “Techniques for the Measurement of Trace Moisture in High-Purity Electronic Specialty Gases,” Review of Scientific Instruments 74, no. 9 (2003): 3909–3933.
6. JT Hodges et al., “Frequency-Stabilized Single-Mode Cavity Ring-Down Apparatus for High-Resolution Absorption Spectroscopy,” Review of Scientific Instruments 75, no. 4 (2004): 849–863.
7. SY Lehman, KA Bertness, and JT Hodges, “Detection of Trace Water in Phosphine with Cavity Ring-Down Spectroscopy,” Journal of Crystal Growth 250, nos. 1–2 (2003): 262–268.
8. D Lisak, JT Hodges, and R Ciurylo, “Comparison of Semi-Classical Line Shape Models to Rovibrational H2O Spectra Measured by Frequency-Stabilized Cavity Ring-Down Spectroscopy,” Physical Review A., 73, (2006): 012507.
Joseph T. Hodges, PhD, is a research engineer at the National Institute of Standards and Technology (NIST) in Gaithersburg, MD. Formerly project leader of the humidity standards program at NIST, he has more than 10 years of experience in humidity standards and quantitative gas standards based on absorption spectroscopy. He received a BS from Purdue University in West Lafayette, IN, and a PhD from the University of Wisconsin–Madison in mechanical engineering. (Hodges can be reached at 301/975-2605 or firstname.lastname@example.org.)
Gregory E. Scace is a mechanical engineer at NIST. He has 13 years of experience in generating and measuring trace water vapor/gas mixtures containing water vapor mole fractions as low as 400 ppt. He received a BS in mechanical engineering from the University of Maryland in Baltimore. (Scace can be reached at 301/975-2626 or email@example.com.)