Coming clean about
03/01/2004
Andy Mackie, Praxair Electronics, Tonawanda, New York
Brian Warrick, Praxair Electronics, Colorado Springs, Colorado
 |
Shrinking feature sizes are making devices increasingly vulnerable to gas-phase contamination. Aggressive gas-impurity milestones set in the International Technology Roadmap for Semiconductors (ITRS) call for a 90% reduction in levels of moisture, oxygen, carbon dioxide, and methane in bulk gases for leading-edge processes between 2004 and 2006. Existing gas supply, analysis, and purification techniques have proven to be reliable at current levels, but the industry faces significant challenges to reach ITRS targets for 2006. Moreover, "off-the-roadmap" purity targets set by individual semiconductor manufacturers for specific markets present additional hurdles, especially for gas analysis. This article examines the tradeoffs of APIMS and GC in addressing these challenges.
To ensure reliable device performance as feature sizes shrink beyond 90nm, process gases will need to be supplied with extremely low levels of contaminants. The 2002 edition of the ITRS [1] already forecasted a 2006 requirement for sub-100ppt levels of four impurities — moisture (H2O), oxygen (O2), carbon dioxide (CO2), and methane (CH4) — in the bulk gases nitrogen (N2), oxygen (O2), argon (Ar), and hydrogen (H2). The ITRS currently calls for sub-1ppb impurity levels of these impurities in the bulk gases.
To achieve ITRS targets for purity in 2006, gas suppliers will need to consider a number of factors when designing a supply system capable of producing, delivering, and analyzing impurities in ultrahigh-purity (UHP) gases. In addition to anticipated roadblocks in the drive toward improved gas purity, some "off-the-roadmap" impurities are difficult to measure at sub-100ppt levels, even using atmospheric-pressure ionization mass spectrometry (APIMS) and high-sensitivity gas chromatography (GC).
Drivers for gas-purity requirements
Existing purification and gas analysis techniques can ensure a reliable gas supply at the current target of sub-1ppb impurity levels. However, the lowering of impurity levels to 100ppt by 2006 poses significant challenges, primarily for gas analysis.
The ITRS itself is somewhat of a "compromise" document, and semiconductor manufacturers' demands for bulk gases at sub-100ppt levels for all impurities are growing well ahead of the Roadmap. The main drivers for these demands include:
Ultraclean technology (UCT). Since the early 1990s, Professor Tadahiro Ohmi [2] has promulgated the ethic of UCT, including a concept of "maximum allowable impurity concentrations" in process gases.
Industry roadmaps. The ITRS is the most notable of the public-domain roadmaps, but semiconductor manufacturers usually have their own versions, tailored to their specific markets.
Ability to analyze and purify. If an impurity cannot be measured or reduced below a certain level, there is little point in requesting it.
Yield improvements. This is the least-understood issue, probably due to several factors:
- high value of the data,
- cost of experiments, and
- low signal-to-noise ratio of changing one variable in a manufacturing process involving 400 or more individual steps.
In addition to the ITRS and "off-the-roadmap" issues, there is a "chicken-and-egg" situation. As one expert in ppb analysis has stated, "APIMS...has been a major factor in driving gas specifications" [3]. Requests for decreasing impurity levels can be shown to correlate well to the availability of instrumentation capable of quantifying those impurity levels.
Cost may be a potential factor in resisting the drive to lower impurity levels. Since gas supply contributes <2% of materials cost/wafer, however, there is little incentive to reduce materials costs by using gas containing higher impurity levels.
Achieving projected ITRS impurity levels
While gaseous products containing impurities at ITRS levels can be produced and analyzed under laboratory conditions, implementation of a production, delivery, and analysis system at sub-100ppt impurity levels presents a challenge for suppliers. Semiconductor fabs routinely employ automated APIMS technologies that can switch between base gases to make frequent measurements of impurities at sub-400ppt levels. However, significant purge time between samples may be required for accurate impurity-level measurement at <1ppb, severely limiting this approach at sub-100ppt levels. Because of its low volatility and affinity for most surfaces, moisture is invariably the worst culprit here.
With a shift to sub-100ppt impurity specifications, ensuring gas purity at the process tool will become even more dependent on a variety of factors, the most important being:
- quality of incoming gas supply;
- sizing and purification capability of gas-purification systems;
- sizing, choice, and installation of gas-filter skids;
- choice, installation, and integration of analytical systems; and
- design and purging of the gas-delivery system.
Most semiconductor fabrication tools are equipped with high-efficiency point-of-use gas purifiers. This implies that a consistent impurity level, and particularly the avoidance of spikes in impurity levels, should be more important than achieving ultralow impurity levels in the bulk gas supply.
Quality of on-site production. Today's UHP N2 and O2 production cycles allow removal of active impurities to sub-ppb levels without product purifiers. Purity is based on efficient pre-purifiers and columns to remove "heavy" (higher boiling point) components from N2 and to remove both "light" (lower boiling point) and "heavy" components from O2 [4].
Bulk products and purifiers. For many customers, O2, Ar, and H2 products are stored on-site in cryogenic vessels or high-pressure tube trailers. Vapor purity from these products varies greatly, and may contain >10ppm levels of contaminants. Therefore, these products are typically purified before use. The purifier medium reacts with contaminants to produce a product that usually meets or exceeds current (2003 at time of writing) ITRS specifications. Improper sizing of the purifier can have a detrimental effect on gas quality. Additionally, purifier performance is greatly affected by incoming bulk-gas quality. Strict control and monitoring of feed supply may be necessary to guarantee efficient removal.
For externally installed purifiers, the diurnal effect of ambient temperature fluctuations on exposed downstream piping may cause the occasional desorption of ppb moisture levels. Therefore, purifiers and piping should be located within a temperature-controlled facility and be placed close to the delivery system, especially when roadmap purity is required. The gas-supply system will inevitably drydown over the long term, but this effect may prolong initial qualification of the system.
On-site production and purification coupled with APIMS technology can easily meet and verify ITRS impurity-level needs projected for 2006 in N2, according to measurements conducted by a semiconductor manufacturer in Southeast Asia (see Fig. 1).
 Figure 1. a) Asian semiconductor manufacturer's data for three impurities in N2, and b) data for methane in N2. |
With careful design of on-site production and storage systems, bulk product can also be produced and purified to sub-100ppt levels. The concern then becomes preventing product recontamination.
Bulk filtration
The first potential source of contamination is the product filter. Improper choice of filter media can lead to moisture desorption and H2 outgassing at ppb levels. PTFE membrane filter use has continued because of its filtration efficiency and inertness. These filters may rapidly allow the gas supply to meet sub-100ppt purity levels under the correct conditions, especially with sizing to match delivery-system flow rates, hermetically sealed filter housings, and filter pre-baking.
Moisture is loosely physisorbed to PTFE, and low gas flows (under-utilization) can contribute ppb levels of moisture for extended periods of time [5]. Metal filters may improve product purity since the media adsorb less moisture; they can also be pre-baked to higher temperatures than polymers.
Other considerations
Under-utilization of a piping system is common in new facilities. Such a system can cause impurity levels to rise due to moisture desorption from the piping, as explained previously. Unpurged volumes ("dead legs") are another potential source of contamination within a UHP delivery system. To deliver sub-100ppt product to tools, even a perfectly sealed system must have sufficient constant flow to dilute down recontamination effects such as these.
Analytical issues
The choice of industry-proven analytical technologies is limited for gauging impurity at sub-100ppt levels. The generally accepted method of analysis in N2, Ar, and H2 is APIMS. APIMS employs a two-step ionization process to first ionize base gas molecules and then transfer charge to the impurities through collisions.
 Figure 2. APIMS calibration for the 2002 ITRS impurities in N2. |
The four impurities in bulk gases that are referred to in the current ITRS are readily detectable in N2, Ar, and H2 by using positive-ion APIMS. As previously discussed, this analytical technique is typically sufficient to meet the current projections of the ITRS for 2006.
 Figure 3. Modeling of CH4+ peak from CH3+ and H2O+. |
Figure 2 shows calibration plots for H2O+, CO2+, O2+, and CH3+ in N2. The CH4+ ion is not used by Praxair as a methane marker, as it is sensitive to moisture levels. Figure 3 shows modeling of the CH4+ (m/z = 16) peak, compared with the actual result found, where the model CH4+ result (x) is calculated from the simple equation
x = a + (by) + (cz)
where a = 0.0022, b = 0.0088, c = 0.76, and y and z are the concentrations of H2O (from H2O+) and CH4 (from CH3+) in ppb. The reason for the moisture contribution to the (m/z = 16) peak may be protonation of a methyl radical formed in the plasma:
CH3 + H3O+ Æ CH4+ + H2O
Roadblocks with positive-ion APIMS
APIMS sensitivities are typically sufficient to allow sub-100ppt detection limits using both IUPAC and "propagation of errors" analysis. One exception is detection of O2 in H2, which is limited by the low proton affinity of O2.
Also, due to the low ionization potential of O2, direct measurement of impurities in a base gas containing high levels of O2 is not possible with APIMS. However, cluster analysis and H2-addition studies have been conducted, resulting in improved detection limits for H2O, CH4, and CO2 in O2. Unfortunately, these methods are still not sufficient for measuring impurities down to the sub-100ppt level slated for 2006 [6].
 |
The question remains: What about H2, CO, nonmethane hydrocarbons, amines, NOx and other impurities? As an example, the table shows some of the problems and interferences from other potential species that have been observed in N2. It should be noted that the data shown will vary significantly with a variety of factors, including the "tune" used, source pressure and temperature, sample gas flow rate, corona discharge-voltage setting, and the concentration of interfering impurities.
APIMS for H2 in N2. As would be expected, looking for the parent H2+ ion at (m/z = 2) proves fruitless, and every hydrogenated species (H2O, CH4, and so on) must be considered a potential source of protons, although evidence suggests that the contribution from H2O is minimal at low-ppb moisture levels. Using a mixed calibration gas (2ppm of H2, O2, CO, CO2, CH4 in balance helium), the sensitivity of N2H+ (m/z = 29) may be calculated by allowing for a contribution of 0.733% from N14/N15 (also m/z = 29), and subtracting this from the peak height. But the extent of proton contributions from CH4 and any other protonated species present at ppb levels is a major conflicting factor.
APIMS for CO. The CO+ and N2+ species share the same mass/charge ratio (m/z = 28), so a high declustering voltage is set to allow the CO+ ion to be broken up:
CO+ Æ O + C+
However, any other carbon-containing species present may also contribute to the (m/z = 12) peak, making positive identification of the C+ as "CO" a complex task.
TA7000R-N2. As a result of the APIMS limitations for direct measurement of H2 and CO, a trace analytical TA7000R-N2 was tested to quantify these impurities in N2 at sub-ppb levels. The instrument takes samples every 340 sec, and uses GC as the peak separator. As the impurities elute from the chromatography column, they pass over a heated bed of mercuric oxide. The level of mercury vapor formed by reaction with each of the two reducing impurities with this bed is then detected by light absorption.
 Figure 4. H2 breakthrough analyzed by TA7000R-N2 and APIMS. |
During initial tests, the instrument clearly detected a "breakthrough" of H2 from an adsorption bed, well ahead of that seen by the APIMS, which was also monitoring the output from the bed (see Fig. 4), using the N2H+ peak found from the mixed calibration gas.
Comparison of TA7000R-N2 and APIMS. A comparison of the two instruments was performed using a certified calibration cylinder of H2 and CO in N2, singly or doubly diluted with getter-purified "zero gas" N2. The results can be compared by reference to Figs. 5 and 6. Note the extremely low sensitivity of the APIMS (Fig. 5) to H2 (only 0.05, or 5%, of expected) and CO (0.12, or 12%, of expected). This can be compared to the TA7000R's 100% (1.00) sensitivity for H2 and 84% (0.84) sensitivity for CO (Fig. 6). The reason for the reduced sensitivity of the TA7000R-N2 to CO is unknown.
 Figure 5. H2 and CO concentrations from APIMS peaks at m/z = 29 and 12. |
Based on a linear-regression analysis of the TA7000R-N2 measurement, the "zero gas" data gave a mean result of 0.2 ppb H2. Longer-term experiments showed that these results are not normally distributed, and so a mean figure overestimates the H2 level. The distribution clearly demonstrates that the "real" (modal) level is between 0.0 and 0.1ppb (0–100ppt) for both H2 and CO.
 Figure 6. H2 and CO concentrations from TA7000R-N2. |
Other species. Other work on "off-roadmap" impurity species in UHP N2 at Praxair has so far focused on analysis of butane, sulfur dioxide, nitrogen oxides, and ammonia [4], with the APIMS showing the highest sensitivity for the SO2+ ion.
Conclusion
By a judicious choice of ion, positive-ion APIMS may be used to analyze for H2O, O2, CO2, and CH4 impurities in N2, Ar, and H2 into the foreseeable future. Based on the 2002 edition of the ITRS, APIMS technology is sufficient to meet sub-1ppb continuous monitoring requirements, although O2 as a base gas remains a problem. However, careful design and implementation of the incoming gas supply and delivery system will remain a critical factor in ensuring consistent product quality.
Continuous analysis of sub-100ppt impurity levels will also be critical, as application of current APIMS gas-switching technology may prove difficult. This may require the use of dedicated APIMS or the newer ion-mobility spectrometry (IMS) technique for each individual gas.
Reliable quantitative analysis for sub-100ppt levels of many impurities not listed on the ITRS is beyond the capabilities of positive-ion APIMS, and is clearly pushing the limits of other techniques.
Acknowledgments
The authors would like to thank Praxair Electronics team members Dr. Kevin Albaugh, Conrad Sorenson, Dr. Richard Kelly, and Bill Sullivan for their assistance during the initial drafting of this paper.
References
- At time of writing, the current (2002) Edition of the ITRS was available at http://public.itrs.net. Tables 95.a. and 95.b are quoted.
- See, for example, T. Ohmi, "Ultraclean Technology: ULSI Processing's Crucial Factor," Microcontamination, pp. 48–58, October 1988.
- J.J.F. McAndrew, "Humidity Measurement in Gases for Semiconductor Processing," p. 31 in Specialty Gas Analysis, ed. J.D. Hogan, Wiley-VCH, NY, 1997.
- B. Warrick, "Detection of Non-Standard Impurities by APIMS in Semiconductor-Grade Nitrogen," Proc. of Semicon West, 2002.
- J. O'Sullivan, "Use of Fluoropolymer and All-Stainless Steel Filters in Ultraclean Gas Distribution Systems," Pall report STR-PUF 31 (available at http://www.pall.com).
- S.N. Ketkar, et al., "The Use of Proton Transfer Reactions to Detect Low Levels of Impurities in Bulk Oxygen...," Intl. Jour. of Mass Spectrometry, Vol. 206, pp. 7–12, 2001.
For more information, contact Andy Mackie at Praxair Electronics, 175 East Park Dr., P.O. Box 44, Tonawanda, NY 14150; ph 716/879-3151, fax 716/879-2852, e-mail [email protected].