 |
Figure 1: Sampling system |
Mark W. Raynor, Hans H. Funke,
and Virginia H. Houlding
Matheson Tri-Gas Inc.
1861 Lefthand Circle, Longmont, CO 80501
Terry L. Ramus, and Scott J. Hein
Diablo Analytical, Inc.
1110 Burnett Ave, Suite C, Concord, CA 94524
Ralph W. Kirk
Advanced Technology Development
3305 Bay Court, Belmont, CA 94002
Real time impurity monitoring in corrosive gas environments is of considerable interest in the semiconductor community to enhance productivity and reduce costs by minimizing detrimental contamination. There are few commercially available instruments with sufficient sensitivity that can withstand the harsh environments associated with etching and deposition. Optical absorption methods such as FTIR or cavity ringdown spectroscopy [1,2] are some of the most promising approaches, in particular because optical methods are based on first principles and theoretically don’t require calibration. Mass spectrometry (MS), a widely used technique in more traditional residual gas analyzer (RGA) systems, however, offers important benefits such as fast response time, a wide range of species that can be monitored simultaneously, and relatively low cost of ownership [3]. One drawback of RGA systems is their inability to produce reliable quantitative information.
In this work, an HP5973 mass spectrometer was adapted for corrosive gas use with a custom build dynamic flow-inlet system and custom software for quantitative real time monitoring of impurities. The results of sensitivity, response time, and stability testing using moisture and other atmospheric analytes of interest in various matrix gases are presented.
|
Figure 1: Sampling system |
Sample gas from a standard cylinder or a permeation tube was diluted with purified matrix gas using a mass flow controller based dilution system. The sample was introduced at atmospheric pressure through a 50 mm orifice into a vacuum line system that was pressure controlled with an MKS throttle valve and a rotary vane mechanical pump to ~10-1 torr. A fraction of the sample entered through a 1500 mm orifice and an air actuated valve into the electron impact (EI) ionization source of the MS chamber for analysis (see Figure 1). Precisely controlled sample inlet pressure was a critical requirement to obtain quantitative information, since instrument response changes with inlet conditions.
Electron multiplier detectors from two different suppliers (K&M Electronics and SGE Instruments) were evaluated, but otherwise the HP5973 hardware was unaltered. The MS was manually tuned in the m/z <50 range as the HP auto-tune algorithm did not optimize sensitivity in this range. Typical EI conditions (70 eV electron energy, 230 °C source temperature) were employed. For moisture in HCl, low electron energies (~20 eV) and repeller voltages (<10 V) were used to optimize mass resolution and response. The MS was operated in single ion mode.
 |
Figure 2: O2 response in helium matrix (K&M detector) |
Detection limits of <100 ppb and good stability were obtained for O2, N2, CO, CO2, H2O, and CH4 impurities in helium matrix. Figure 2 shows an example for O2 response in helium matrix. Response times were in the range of few seconds. Heavier matrix gases such as Ar, N2 or HCl showed significantly decreased sensitivity, even if only present in small quantities. Figure 3 shows the change in response to N2 and O2 upon addition of small percentages of Ar to the He matrix gas.
 |
Figure 3: O2 and N2 response in He matrix blended with Ar (K&M detector) |
Measurements in Ar and N2 matrix required replacement of the factory supplied K&M electron multiplier because degradation or contamination [4] caused baseline drifts and loss of sensitivity in the range of 2 to 3 % per hour. A possible cause for the degradation is insufficient mass filter selectivity that allows large numbers of matrix ions to hit the detector during dead times while scanning from one mass to the next. In He, the matrix ion is removed from the mass range of interest (e.g. 18, 32, 44) and interference is minimized. Another possibility is surface contamination due to the high background levels of the matrix gases in the vacuum chamber during operation. Several attempts to stabilize the detector response, such as introducing a constant He purge through a separate input port, use of different ion source materials, and changing the source magnet for higher sensitivity in the low mass range decreased the rate of degradation but did not prevent it.
 |
Figure 4: H2O in N2 matrix (SGE detector) |
Improvements by one order of magnitude were obtained by switching to an SGE detector that uses discrete dynodes instead of the continuous dynode design of the K&M detector. Drifts decreased to <0.2 % per hour and could be handled by daily calibration. Tests with ppm levels of moisture in nitrogen matrix also demonstrated improved sensitivity compared to the K&M detector (Figure 4). The instrument was sensitive to sub-ppm changes and calibration curves were linear.
 |
Figure 5: SGE multiplier degradation in HCl matrix |
With the SGE detector, results similar to inert gas matrices were obtained in HCl matrix and moisture impurities. Moisture was measured at m/z=18+19 since a significant fraction of the moisture was protonated (H3O+) in the HCl environment. Doubly charged matrix ions (H35Cl2+ and H37Cl2+) caused elevated background levels at m/z=18 and 19, which limited the ultimate sensitivity. The SGE electron multiplier response initially decreased by ~1% per hour but eventually stabilized after ~24 hours of continuous operation (Figure 5). No explanation for the observed ‘conditioning effect’ is currently available. After the baseline and response stabilized, several calibration curves were measured. Figure 6 illustrates the response to moisture challenges between 2.3 and 11.4 ppm in HCl. Detection limits of ~0.5 ppm were estimated based on 3s (standard deviation) and calibration curves were linear. The sampling system, ion source, and quadrupole chamber did not show any visible signs of corrosion during several weeks of study in HCl matrix.
 |
Figure 6: Moisture in HCl matrix (m/z=18+19) |
 |
Figure 7: Iron chloride species detected in HCl matrix |
Of particular interest in semiconductor applications is the detection of volatile metal contaminants in corrosive matrices. These contaminants may be produced by reactions of the corrosive gas in the distribution system and are usually analyzed off-line after a complicated sample collection step. The possibility of on-line metal detection in corrosive matrices using the MS analyzer was investigated by collection of spectral data in the 100 m/z mass range. FeClx species were confirmed while analyzing HCl. Figure 7 shows a spectral scan that indicates several peak clusters that can be assigned to iron chlorides. TiN, Au, and Ni coated EI sources were installed to rule out formation of Fe species from the ionization source. Results were inconclusive since FeClx peaks were still present using the Au and Ni source but TiClx was found with the TiN source indicating at least partial attack by the matrix gas. However, the real-time observation of volatile Fe species is an encouraging result that merits further study.
Quantitative real-time monitoring of impurities at the sub-ppm level in inert and corrosive gas matrices was accomplished with a prototype MS analyzer. Optimization of tune parameters and use of a discrete dynode electron multiplier were necessary to minimize baseline drift and maximize sensitivity.
[1] S. Salim and A. Gupta, Proceedings of CleanRooms ’96 West, 22-32 (1996)
[2] H. Jones-Bey, Laser Focus World, March 2000
[3] T. Schneider, K.J. Tayor, D.A. Rothenbury, M. Chavis, T. Hoff, and C.H. Huffmann, Micro, 1, 35-39, 1999
[4] J. Gray, American Laboratory, 32(15), 20-21, 2000