 |
Figure 1. Calibration curve. Moisture as measured by the diode laser sensor vs. moisture supplied by the low frost point generator, showing good agreement throughout the range studied. |
Chris Hovde and Joel A. Silver
Southwest Sciences, Inc.
Santa Fe, NM USA
www.swsciences.com
Diode laser wavelength modulation absorption spectroscopy is a practical method for measuring nanomole per mole (ppb) levels of water vapor. As verified in experiments conducted at NIST, a prototype instrument shows excellent linearity and sensitivity, with noise of less than 0.2 nmol/mol. Another instrument developed in cooperation with Delta F Corp. can resolve 0.5 nmol/mol moisture challenges.
The measurement of trace concentrations of water vapor is important in many industrial applications, especially semiconductor fabrication. A variety of methods exists for measuring water vapor at concentrations approaching one nmole/mole (one ppb). However, problems with calibration stability and cross- sensitivity to other gases have led researchers to consider new approaches. Optical absorption spectroscopy has excellent calibration properties (it has linear response over several orders of magnitude and can be calibrated from first principles) and very low cross-sensitivity to other gases. Optical techniques can measure moisture in inert gases such as nitrogen or argon, in reactive gases such as hydrogen or oxygen, and in corrosive gases such as hydrogen chloride and ammonia.
Optical methods for measuring moisture include Fourier-transform spectrometry, intracavity laser absorption spectroscopy, and cavity ring-down spectroscopy. However, we have chosen to focus on wavelength modulation spectroscopy (WMS) as a practical method to achieve sufficient sensitivity.[1] WMS was invented in the 1970s, but has enjoyed renewed interest with the emergence of diode lasers engineered for fiber optic communications. These lasers operate at room temperature and require less than a Watt of electric power. They provide a few mW of light at a wavelength in the near-infrared part of the spectrum. The wavelength of the light can be tuned by changing either the laser temperature or the injection current.
WMS achieves high sensitivity using these laser sources by rapidly modulating the laser wavelength to avoid 1/f noise. Very small changes in laser power can be measured by detecting the photocurrent at twice the modulation frequency. This ac current is zero unless a wavelength-dependent absorption feature is present. When WMS is combined with a multiple pass absorption cell, ultra-sensitive detection of water vapor is possible, even when detecting H2O using weaker overtone transitions with near-infrared InGaAsP diode lasers near 1392.5 nm. This approach has the desired combination of sensitivity and accuracy while enjoying the benefits of rugged communications-type diode lasers. Other researchers have employed similar techniques.[2,3]
To minimize the effects of noise and background artifacts, the instrument records the spectrum of a single rotational line within a vibrational overtone band of water vapor and fits this line using a patent pending flexible linear least squares algorithm. To measure the absorption line profile, the laser drive current, and hence its wavelength, is ramped over the absorption line and simultaneously modulated at frequency f = 50 kHz with a sine wave. The absorption signal as a function of ramp position is quantified by lock-in detection at twice the modulation frequency, 2f. To suppress étalons, the laser/lens assembly is mechanically vibrated[4] and a jitter modulation[5] is used. Least squares fitting of the spectral line shape determines the concentration of water vapor and any drift of the laser wavelength.
Careful design and material selection is required to avoid contaminating the sample with water that outgasses from the spectrometer. Our sensor design is based on a sealed laser head connected to a Herriott multipass cell constructed from ultra-high vacuum stainless steel flanges sealed with copper gaskets. The sample pressure is maintained near 13 kPa (about one eighth of an atmosphere), balancing signal strength while keeping the line shape sufficiently narrow so that the full shape can be recorded within the wavelength scan range of the laser. Higher sample pressures can be used; at atmospheric pressure the sensitivity is reduced by a factor of about five.
|
Figure 1. Calibration curve. Moisture as measured by the diode laser sensor vs. moisture supplied by the low frost point generator, showing good agreement throughout the range studied. |
The diode laser moisture sensor was used in a series of calibration measurements against the new Low Frost-Point Humidity Generator (LFPG) at NIST.[6] The LFPG saturates an inert carrier gas stream with water vapor by flowing the gas over a plane surface of ice at fixed temperature and pressure. Its output can be predicted from thermodynamic principles and is nearly independent of gas flow rate. The calibration of the WMS instrument was tested by stepping the humidity generator through a series of set points ranging from 5 µmol/mol to 2.5 mmol/mol of water vapor in air (Fig. 1). Good agreement was obtained throughout this range, although outgassing of a few nmol/mol within the laser sensor limited the accuracy at the lowest moisture levels.[7]
Noise in the prototype instrument was characterized by examining the variance in measured moisture over a period of about half a day while keeping the input moisture constant. The spread in measured moisture values is the same at 15 nmol/mol and 100 nmol/mol and is well described by a bell curve with 0.3 nmol/mol standard deviation. The Allan variance provides a useful measure of the noise as a function of averaging time and shows that the best resolution can be achieved with an averaging time of about 20 min. Adopting a three sigma criterion for statistical significance (99.7 % confidence level), the minimum detectable change in mole fraction of water vapor is ~0.2 nmol/mol.
Southwest Sciences has worked with Delta F Corp. to develop a prototype trace moisture instrument suitable for manufacture. The instrument has a faster time response (updating once per second) and a smaller footprint (rack-mountable) than the system described above, without sacrificing sensitivity. Experiments at Delta F showed three instruments could resolve 0.5 nmol/mol. Other features include improved ruggedness and ease of use.
Diode laser absolute moisture detection is sensitive, precise and accurate. Using a combination of methods to suppress étalons and achieve high signal/noise, a detection limit of ~0.2 nmol/mol can be achieved. The frost- point temperature equivalent to this detection limit is -117º C. The linearity of the instrument is excellent, exhibiting an error of only 1% at water concentrations in the mmol/mol range (frost points near -65º C), and this range can be improved by using the full exponential form of Beer’s Law. The most significant limitation to the accuracy appears to be from outgassing within the sample region.
We are grateful to James Whetstone, Joseph Hodges and Greg Scace at NIST for their assistance and co-operation in permitting us to use the Low Frost Point Generator, Andy Wright and Clayton Wood at Delta F for the results of the prototype, and David Moschella for providing an Allan variance routine written for LabView. This work was funded by the U.S. Department of Commerce under Contract No. 50-DKNB-5-00189.
The authors are with Southwest Sciences, Inc. Dr. Hovde is at the Ohio Operations office at 6837 Main Street, Cincinnati OH 45244, dchovde@ swsciences.com. Dr. Silver is at 1570 Pacheco Street, Suite E-11, Santa Fe, NM 87505, jsilver@swsciences.com.
1. Hovde, C., Extended Abstracts of the NIST/AVS Workshop, Gaithersburg, MD, p. 47-48 (May 23-25, 1994).
2. Wu, S.-Q.; Morishita, J.-I.,; Masusaki, H.; Kimishima, T., Anal. Chem. 70, 3315-3321 (1998) .
3. J. M. Girard and P. Mauvais, Proc. ISSM’96, 5th Int. Symp. Semicond. Manuf., Ultra Clean Society, Tokyo, Japan 325-328, (1996).
4. Silver, J. A.; Stanton, A. C., Appl. Opt. 27, 1914 (1988).
5. Cassidy, D. T.; Reid, J., Appl. Phys. B 29, 279-285 (1982) .
6. Scace, G. E.; Huang, P. H.; Hodges, J. T.; Olson, D. A.; Whetstone, J. R. 1997 NCSL Workshop and Symposium, Atlanta, GA (July 27-31, 1997) .
7. Hovde, D. C.; Hodges, J. T.; Scace, G. E.; Silver, J. A., Appl. Opt. (in press).