(November 15, 2010) — Delivery of HBr with consistently low water vapor (moisture) levels is critical to prevent delivery system corrosion and device performance issues. Jianlong Yao et al, Matheson, present the effect of cylinder material on delivered moisture concentration in gas phase HBr. Polished AISI Cr-Mo steel, Nickel-lined AISI Cr-Mo steel, and 316L stainless steel cylinders are prepared and tested under similar conditions and show markedly different results.

Anhydrous hydrogen bromide gas is frequently used in the semiconductor industry as a trench-etch for CMOS device manufacturing and must be maintained at a consistently high purity to minimize process variations. Moisture is a particularly detrimental impurity in the hydrogen bromide gas as it leads to corrosion of the delivery system components and generation of particles that can negatively impact device performance. Vapor phase moisture in HBr also inhibits wafer chemistries and ultimately increases the likelihood of defects on the chip. As the features on the wafer decrease in size, an increased emphasis on optimized moisture control in HBr is needed. Requirements for ultra-high purity HBr specify that vapor phase moisture levels must be below 1 ppmv.

Control of moisture in HBr cylinders requires that cylinders are filled with a purified HBr source gas, and that stable constructing materials are used to maintain stored gas purity in the cylinder package. HBr can react with metal oxides on the internal cylinder surfaces to generate H₂O. Therefore, research and testing of various alloys and materials is important to enable selection of appropriate cylinder materials that minimize impurity generation. Past work, done with alloys such as Nickel 200 and Hastelloy C-22, show superior performance compared to 316L stainless steel. However, these materials are costly and therefore are not practical for cylinder package applications.

In this work, nickel-lined AISI 4130 Cr-Mo steel cylinders have been evaluated versus the gas industry standard AISI 4130 Cr-Mo steel cylinders with regard to the concentration of H₂O in the delivered HBr gas. Reasons for the observed differences in moisture concentration are discussed. Further, gas phase cylinder depletion studies are presented to demonstrate how the H₂O concentration can vary as gas is progressively withdrawn from the cylinder, particularly close to and after the phase-break point. The results allow a comparison of the above two packages together with 316L stainless steel and enable selection of the most suitable package for maintaining low moisture concentrations.

Polished Cr-Mo steel cylinders exhibit the highest and most variable H₂O levels due to HBr reaction with iron oxides on the internal surface — generating moisture. Intermediate results are observed with 316L stainless steel cylinders. The lowest and most consistent H₂O levels are obtained with Ni-lined Cr-Mo steel cylinders. This is attributed to the enhanced stability of the Ni-lining, which has a lower surface area and decreased levels of surface oxides.

Figure 1. Schematic diagram of experimental set-up used for gas-phase moisture measurements from cylinders and cylinder depletion studies.

Gas phase moisture comparison of cylinder packages

Moisture comparison experiments were performed using polished AISI Cr-Mo steel cylinders and Ni-lined AISI Cr-Mo steel cylinders fitted with Hastelloy C-22 CGA 634 valves. Cylinders were filled with 70lbs purified HBr via liquid phase transfer from the same purified HBr source. Prior to filling, the cylinders were vacuum-baked, passivated and equilibrated using the same techniques. Cylinders were then connected to the analysis system shown in Figure 1 and analyzed for moisture in the gas phase HBr. The sample lines and manifold components were constructed from Hastelloy C-22 to minimize corrosion and avoid any impurity contribution from the system during sampling. The sample lines were also heated to 65°C, purged with purified N₂ to remove adsorbed water and finally dried down to <100 ppb H₂O in HBr, using a Nanochem MetalX purifier (Matheson Tri-Gas, Longmont, CO). Low level gas phase moisture data were collected with a model MTO-1000 cavity ring-down spectrometer (CRDS) (Tiger Optics, Warrington, PA) using the absorption band at 1392.54nm. The calibration of the instrument was confirmed using a NIST certified H₂O standard. The experimental conditions and procedures were established by a previous study determining the thermodynamic properties of moisture in HBr [1]. Higher moisture levels were detected by FTIR spectroscopy (Magna 550, Thermo-Nicolet, Madison WI) at 800 torr and a flow rate of 1 slpm.

Results of the gas phase moisture studies conducted are presented in Figure 2 and show that the Ni-lined Cr-Mo steel cylinders outperformed the polished Cr-Mo steel cylinders.

Figure 2. Comparison of moisture in vapor phase HBr delivered from Cr-Mo steel and nickel-lined Cr-Mo steel cylinders as measured by CRDS.

The Cr-Mo steel cylinders exhibited the highest and most variable H₂O levels. This is attributed mainly to the reaction of HBr with available reactive metal oxides (particularly iron oxides) on the internal cylinder surfaces that results in water being generated.

Auger electron microscopy (AES) studies on polished 4130 Cr-Mo steel examined after 1 year exposure to HBr have shown that the oxide layer in the polished Cr-Mo steel cylinder extends beyond 1000Å into the metal surface [2]. Therefore not only is there a significant amount of oxygen present in the inner cylinder surface for reaction, but the formation of FeBr₃ and H₂O from Fe₃O₄ or Fe₂O₃ and HBr are also both thermodynamically favorable:

Fe₃O₄  + 9HBr  →  3FeBr₃  +  4H₂O  +  1/2H₂      ΔG°_R  = -26.4 kcal
Fe₂O₃  + 6HBr  →  2FeBr₃  +  3H₂O                      ΔG°_R  = -25.7 kcal

In contrast, all 5 Ni-lined cylinders delivered HBr with a moisture concentration that was consistently well below the 1 parts per million, volume (ppmv) specified level. The average vapor phase moisture concentration for all the Ni-lined cylinders was 3.8 times lower than that of the polished Cr-Mo steel cylinders. The AES depth profile of a similarly prepared Ni-lined cylinder to those above showed that the surface oxide penetrated less than 100Å into the nickel surface, which is more than a 10 times reduction compared to that of the Cr-Mo steel surface [2]. Consequently the moisture generation reaction is minimized in Ni-lined cylinders because the Ni lining has decreased levels of surface oxides and is much more protective than porous iron-oxide-containing surfaces.

Another factor to consider is the surface roughness profile of the internal surfaces of cylinders. Table 1 shows average surface roughness based on 9 measurements (made at the top, middle and bottom sections of the cylinder with a portable surface profilometer, Mahr Federal, Providence, RI) and the total cylinder surface area projected from the profilometry measurements for Ni-lined vs polished Cr-Mo steel surfaces. Severe degradation of the cylinder surface roughness profile occurred during long-term exposure in polished Cr-Mo steel cylinders. The average surface roughness increased over 10 times when exposed to HBr for over one year, while Ni-lined Cr-Mo steel cylinders had a roughness 5 times that of polished Cr-Mo steel after similar HBr exposure. Applying a surface area model with the surface roughness data enables a projected surface area of the cylinder surface to be calculated. Surface degradation by HBr effectively adds ~6 times the surface area when compared to the original polished surface, while the Ni-lined cylinders’ surface area only increased by a factor of ~3 over the period of a year. The lower surface area of the Ni-lined Cr-Mo steel cylinders is expected to provide fewer surface interaction sites between the gas and the cylinder material, creating less reaction and impurity generation over time.

Shelf life studies have confirmed this. The data in Table 2 shows the concentration of impurities after storing HBr in a Ni-lined Cr-Mo cylinder for 22 months. The impurity levels, including for H₂O, are all below the specification of the 5N5 purity product and indicate stability with this cylinder package.

Moisture in liquid and vapor phase HBr

Since elevated moisture was detected in several Cr-Mo steel cylinders, studies were undertaken to measure the water concentration in the liquid and vapor phases of two full HBr cylinders. Gas phase measurements were performed using the set-up in Figure 1. To quantitatively analyze H₂O in the liquid phase HBr, the cylinders were inverted and liquid HBr was withdrawn and vaporized within a Hastelloy C-22 vaporizer tube at 180°C. The vaporized liquid phase was introduced into the sampling manifold via a Hastelloy C-22 regulator at ~1slpm and analyzed by FTIR at 70°C and 800 torr.

The results in Table 3 show that the water preferentially distributes itself into the liquid phase. H₂O levels in the liquid phase HBr were approximately 6.5 times higher than in the vapor phase HBr (under the sampling conditions used in the study). Previous work with other compressed liquefied gases has shown that elevated liquid phase H₂O levels can result in variable vapor phase moisture levels as the cylinder contents are consumed. Therefore HBr cylinder depletion studies were undertaken.

Cylinder depletion

The cylinder depletion experiments were performed using polished AISI Cr-Mo steel, Ni-lined AISI Cr-Mo steel and 316L stainless steel cylinders to understand how moisture levels might vary as gas is withdrawn from cylinders until empty. All cylinders fitted with Hastelloy C-22 DISS 634 valves were prepared, passivated and filled with purified HBr under similar conditions. Gas phase HBr was withdrawn from the cylinders at 5 slpm and the weight loss was tracked using a cylinder scale as shown in Figure 1. The moisture level in the HBr stream was analyzed by CRDS or FTIR spectroscopy at 1 slpm. Excess HBr was directed to an acid gas scrubber via an MFC.

The HBr cylinder depletion studies performed with the three cylinder packages reveal how moisture, once generated (as discussed above), will partition into the liquid phase and increase the gas-phase concentration towards and after phase break (Figure 3). Data measured from the Cr-Mo steel cylinder showed that the HBr from this cylinder exhibited the highest gas-phase moisture increase from 0.8 ppmv at the start to 1.5 ppmv after two thirds of the HBr had been withdrawn, and 19 ppmv after phase break, when fully depleted. The stainless steel cylinder showed more consistent H₂O levels. However, moisture levels still rose from 0.4 ppmv when full to >1 ppmv after three quarters of the HBr had been removed, reaching 3.7 ppmv after phase break. This indicates that 316L stainless steel is partially reactive in the presence of HBr and confirms results of other researchers, who show that under certain conditions surface corrosion does take place [3,4]. The Ni-lined cylinder was the most stable package. The moisture concentration delivered in the HBr from this cylinder remained consistently low at below 0.3 ppmv as the entire contents of the cylinder was withdrawn, including after the phase-break point.

Figure 3. Moisture concentration in gas phase HBr delivered from Cr-Mo steel, stainless steel and Ni-lined Cr-Mo steel cylinders as the cylinder contents are depleted.

Conclusion

The cylinder materials used to store reactive gases such as HBr can have a major impact on delivered gas purity. Cylinder material comparison and depletion studies establish that the use of Ni alloy adds significant benefits to control contaminants, such as moisture, that originate from the cylinder package. Minimizing moisture is important for maintaining consistent gas purity. Cr-Mo steel cylinders even when well prepared and passivated will generate moisture that partitions into the liquid phase and then gradually increases the HBr gas phase moisture concentration as gas is progressively withdrawn from the cylinder. This is undesirable as it can lead to corrosion of delivery systems and cause process inconsistencies. 316L stainless steel is not an optimal material either for storing HBr, as shown by the depletion data. We conclude that Ni-lined Cr-Mo steel cylinders are the most suitable cylinder package for maintaining and consistently delivering low moisture concentrations in HBr delivered to process tools. The Ni lining is stable and protective in the presence of HBr, and has a very thin oxide layer that minimizes moisture generation.

References
1. J. Yao, E. Olsen and M. Raynor, Thermodynamic properties of trace water in high purity hydrogen bromide, ECS Transactions Vol. 6 Advanced Gate Stack, Source/Drain and Channel Engineering (1) 2007 p 309.
2. A.J. Seymour, C.L Wyse, J.L. Yao, P. Jha, E.W. Olsen, M.W. Raynor and R. Torres, Cylinder materials of construction for ultra-high purity HBr in advanced semiconductor etch processes, ECS Transactions Phoenix AZ, Vol. 13, Plasma Processing (17) 2008.
3. G.M. Smudde, Jr. W.I. Bailey, B.S. Felker, M.A. George and J.G. Langan, Materials selection for HBr Service, Corrosion Science, 37 (12) 1995 p. 1931-1946.
4. S.M. Fine, R. M. Rynders and J.R. Stets, The role of moisture in the corrosion of HBr gas distribution systems, J. Electrochem. Soc. 142 (4) 1995, 1286-1293.

Jianlong Yao received his B.S and M.S. degrees in Chemistry from The Univerity of Science and Technology of China and his Ph.D. degree in Chemistry from Colorado State University, Fort Collins CO and is Senior Scientist at Matheson, Advanced Technology Center, 1861 Lefthand Circle, Longmont CO 80501, (303) 6782067, [email protected]

Eric W. Olsen received his B.S. degree in Chemistry from University of Minnesota, Duluth and is Quality Manager at Matheson.

Adam Seymour received his B.S. degree in Biochemistry from University of  Colorado, Boulder    and is Senior Scientist at Matheson.

Robert Torres received his B.S. degree in Chemistry from University of Wyoming and Ph.D. degree in Chemistry from University of Colorado, Boulder and is Director of R&D, Gas Technology at Matheson.

Gianni Leonarduzzi received his B.S. in Chemistry from the University of Padova, Italy and Ph.D. in Chemistry from University of California at Santa Cruz and is Product Marketing Manager at Matheson.

Mark W. Raynor received his B.S. (Hons) degree from University of Natal, South Africa and Ph.D. in Chemistry from University of Leeds, England and is Director of R&D, Gas Development and Analytical Technology at Matheson.

Subscribe to Solid State Technology/Advanced Packaging. Follow Solid State Technology on Twitter.com via editors Pete Singer, twitter.com/PetesTweetsPW and Debra Vogler, twitter.com/dvogler_PV_semi. Or join our Facebook group