Halogenated gases are employed in semiconductor manufacturing for a wide number of processes, including plasma etching of polysilicon, diffusion furnace cleaning and oxidation processes. The selection of components for use in a corrosive gas delivery system is influenced by safety considerations and the corrosion resistance of the materials of construction. The choice of components for use in corrosive gas service will determine the extent of corrosion and dictate the quality of the gas delivered to the process tool. The selection of a suitable POU filter for corrosive gas service plays a critical role in the delivery of UHP gas, as the filter is typically the last component in contact with the process gas prior to entering the tool.

Stainless steels have a number of corrosion modes including general and localized corrosion, stress cracking, pitting and crevice attack. The corrosion of a gas delivery system can result in the introduction of corrosion particulates and volatile corrosion by-products into the gas stream, which can ultimately affect device performance.

The careful selection of the 316L stainless steel utilized in the fabrication of filter assemblies is critical as materials of similar composition can perform very differently. Wang and coworkers ⁽¹⁾ noted variability in the corrosion behavior of steel alloys with bulk compositions that are virtually identical when exposed to moist HCl. In addition, Takahashi et al. ⁽²⁾ noted that bromide penetrated 100 Å into the surface of electropolished 316L stainless steel and less than 10 Å into the surface of Cr₂O₃ passivated stainless steel when exposed to moist HBr (1,000 ppm). The latter result illustrates the corrosion resistance imparted to electropolished surfaces by a chromium oxide layer. The alloy composition and surface preparation will dictate the corrosion resistance of the component.

Fine and coworkers ⁽³⁾ investigated the effect of moisture content on the extent of HBr corrosion for 316L electropolished stainless steel. The scanning electron microscopy (SEM) and x-ray emmision spectroscopy (XES) analysis of the exposed sample coupons indicated no effect upon exposure to HBr containing less than 0.5 ppm of moisture. A moisture content of 10 ppm resulted in bromide incorporation and the onset of corrosion. The formation of corrosion pits was noted upon increasing the moisture level in the HBr to 100 ppm. A dense bromide scale was noted at a moisture concentration of 1,000 ppm.

The latter work indicates that the degree of corrosion is strongly dependent on the moisture content in the corrosive gas. The moisture concentration in commercially available high purity HBr is typically < 3 ppm. However, it is important to note that the moisture concentration is dependent on the quantity of the corrosive gas remaining in the cylinder. In the case of HCl the moisture concentration increases significantly at quantities below 100 grams ⁽⁴⁾ - see Figure 1.

Figure 1

Moisture concentration in HCl as a function of the remaining cylinder weight.

In addition, the concentration of moisture present in the distribution system may differ significantly from the moisture concentration of the gas in the cylinder. The introduction of moisture into a corrosive gas delivery system downstream of the cylinder is typically attributed to a leak in the system, incomplete drydown of the system or outgassing of moisture from components in the gas system. The change-out of the high pressure cylinder and maintenance and repair of the delivery system must be accomplished without the introduction of atmospheric contaminants into the gas system. An additional concern is the change in moisture content due to Joule-Thomson cooling. When HBr passes through a restricted orifice, Bhadha and Greene ⁽⁵⁾ reported that the cooling phenomenon increases the corrosion of a regulator at moisture levels as low as 5 ppm.

A number of maintenance practices are recommended to eliminate the introduction of moisture into the corrosive gas distribution system

In order to meet the requirements of the gas delivery system, the POU filter must provide the means to remove any particulate contamination added to the gas stream during transportation to the process tool. In addition, the POU filter must not outgas any chemical contaminant into the gas stream. The tendency of organic materials to outgas over a prolonged period of time has restricted the use of polymeric filters in a number of applications. All-stainless steel filters were developed to allow POU filtration without any measurable outgassing. The Ultramet-L^TM POU filter is ideally suited for corrosive gas service due to the absence of any measurable moisture outgassing and the transparency of the all-metal filter to moisture slugs.

The performance and change-out schedule for gas filters will be influenced by the process conditions (flow rate, temperature, particulate contamination and moisture levels) experienced in service. A critical factor influencing the change-out schedule of a gas filter is compatibility of the materials of construction with the process gas.

While it is possible to estimate a preventive maintenance change-out schedule for gas filters based on laboratory scale corrosion testing, the testing cannot accurately duplicate actual service conditions. In order to determine an acceptable PM change-out schedule for POU gas filters, a number of Ultramet-L Gaskleen® 4400 Series Assembly (GLFF4400VMM4) filters were evaluated after more than two years service in corrosive environments. The returned filter assemblies were employed in BCl₃, HBr and Cl₂ with moisture levels in the 5 -10 ppm range. The service conditions experienced by the filters included cylinder change-out and replacements of components upstream of the POU filter.

The returned filters were evaluated in terms of differential pressure, retention efficiency, surface finish and viewed for evidence of corrosion. The following report outlines the results of the testing, and based on the findings a PM change-out schedule for POU gas filters is recommended.

The Ultramet-L Gaskleen 4400 Series Assembly is constructed of an electropolished stainless steel housing and an all 316L stainless steel medium. The housing of the assembly has an internal surface finish of ≤7 µin R_a and a chromium enriched surface layer ⁽⁷⁾. The 316L stainless steel media pack consists of layers of sintered fibers of a specified diameter. The layers of sintered 316L stainless steel fibers are supported between a sintered woven 316L stainless steel mesh.

The results of the evaluation of the returned filter assemblies which were employed in BCl₃, HBr and Cl₂ are as follows.

Aerosol Cleanliness and Retention Studies

The GLFF4400VMM4 filters used in corrosive service for more than two years were tested for aerosol cleanliness, and retention. A test stand conforming to SEMATECH standard # 93021511A-STD, "Test Method for Determination of Particle Contribution by Filters in Gas Distribution Systems", was used to assess the particulate cleanliness of the filters. The test stand was modified to accept a stream of unfiltered - 40°F dew point air containing > 1E4 particles per cubic foot of 3 nm and greater in size as a challenge to measure retention characteristics. The challenge aerosol of unfiltered air was preferred, as we did not want to contaminate the internal surface with NaCl (a typical challenge aerosol), due to subsequent evaluation for corrosion. A flow of 50 slpm was used to measure cleanliness and retention performance, utilizing a TSI CNC model 3025 sensitive to 0.003 micron for particle measurement.

The filters were in service for more than 2 years, at a moisture level of 10 ppm for the HBr and Cl2 and a level of 5 ppm moisture for the BCl3 gas. The cleanliness and retention data are summarized in Table 1:

Table 1

Gas Service	Cleanliness (cts/ft3)	Upstream Challenge (cts/ft3)	Downstream Challenge (cts/ft3)
HBr	4	8 E 4	4
Cl2	44	8 E 4	4
BCl3	32	9 E 4	2

The filters performed well, with expected retention. The initial higher counts in the cleanliness results were likely due to surface contamination, which was cleaned off by the time the challenge test was performed. Based on the aerosol testing, the filters, at this stage, will continue to perform well in these semiconductor grade gases.

Differential Pressure Measurements

The differential pressure results for the filters (all were equivalent) were 6 psid at 25 slpm, and 10 psid at 50 slpm, with an inlet pressure of 30 psig indicating a very slight increase in pressure drop due to contaminant loading. This is consistent with typical point-of-use gas particle loading, and demonstrates no particulate generation causing increased filter differential pressures.

Microscopic Inspection

The returned filter assemblies were sectioned and subjected to microscopic analysis to determine the extent of corrosion of the 316L stainless steel. The effects of corrosion were analyzed in terms of surface roughening, change in fiber diameter, pitting and localized corrosion at the weld interface between the assembly housing and the media pack.

The internal surface finish of the upstream and downstream of the returned filter assemblies, including inlet and outlet bore, were measured to be < 7 µin R_a with the exception of the inlet of the filter assembly employed in HBr. The measured surface finish of the returned assemblies is within the maximum internal surface finish specification of the 4400 Ultramet-L filter assembly. The fine surface finish exhibited by the returned assemblies suggests that the internal surface of the filter assemblies were not subjected to corrosion. The apparent absence of any significant corrosion is supported by photomicrographs of the internal surface of the filter assembly housing. The photomicrograph of the upstream internal surface of the housing removed from BCl₃ revealed a smooth bright surface with no indication of pitting - see Figure 2.

Figure 2

Internal surface of the 4400 Ultramet-L filter assembly removed from BCl3 gas service.

The inlet bore of the assembly housing removed from HBr service revealed a bromide scale. The spotted appearance of the bromide scale suggested the onset of corrosion pits - Figure 4. The appearance of the corrosion pits was unexpected as Fine and coworkers ⁽³⁾ reported that onset of corrosion pits occurred for 316L electropolished stainless steel at a moisture concentration no lower than 100 ppm for HBr. The apparent onset of corrosion observed for the filter assembly removed from HBr service was further investigated by analyzing the weld region between the filter housing and medium pack. The weld region of a component is typically considered to be most susceptible to corrosion. The weld region revealed no evidence of corrosion (see Figure 3). The low particulate counts reported during the cleanliness testing of the filter assembly are in agreement with the corrosion free weld region observed. The degradation in surface finish noted on the inlet bore of the filter assembly can therefore be attributed to the bromide scale.

Figure 3

The weld region between the media pack and the assembly housing of the 4400 Ultramet-L assembly removed from HBr service.

The inlet bore of the assembly housing removed from Cl₂ service revealed no scaling or corrosion - see Figure 5. The lack of corrosion or scaling is not unexpected as chlorine is not considered as corrosive a gas as HBr or HCl.

Figure 4	Figure 5
Inlet bore of the 4400 Ultramet-L filter assembly removed from HBr service.	Inlet bore of the 4400 Ultramet-L filter assembly removed from Cl2 service.

The internal surface area of the assembly housing is less than 1% of the total BET surface area of the filter assembly. The higher surface area of the medium pack suggests that corrosion is most likely to occur at the medium pack. The upstream and downstream sintered woven 316L stainless steel mesh support layer of the filter assemblies removed from Cl₂ and BCl₃ service revealed no scaling or corrosion - see Figure 6. Conversely, the sintered woven 316L stainless steel mesh support layer of the filter assembly employed in HBr service revealed a bromide scale - see Figure 7.

Figure 6	Figure 7
Downstream of medium pack from 4400 Ultramet-L filter assembly removed from Cl2 service.	Downstream of medium pack from 4400 Ultramet-L filter assembly removed from HBr service.

The diameter of the sintered 316L stainless steel fibers of the media pack from the returned filter assemblies was determined with the aid of a scanning electron microscope and compared to a filter assembly supplied from manufacturing stock. The diameter of the fibers from the returned filter assemblies are identical to that of the filter assembly removed from manufacturing stock - see Figures 8 - 10. The SEM photomicrographs revealed some contamination but no degradation of the sintered fibers. The retention efficiency observed for the filter assemblies during the aerosol challenge is further evidence that no degradation of the media pack occurred during extended exposure to the corrosive gases.

Figure 8	Figure 9
SEM of 316L Stainless Steel medium from a 4400 Ultramet-L filter assembly.	SEM of 316L Stainless Steel medium from a 4400 Ultramet-L filter assembly removed from BCl3 service.

Figure 10

SEM of 316L Stainless Steel medium from a 4400 Ultramet-L filter assembly removed from HBr service.

The Ultramet-L filter assemblies employed in corrosive gas service displayed a high surface finish, minimal corrosion, no degradation of the media pack and the expected retention. The results indicate that the gas filter assemblies evaluated remain suitable for use in semiconductor grade gases after more than 2 years service in BCl₃, HBr and Cl₂. Through our experience, with this study no degradation was noticed at a period of more than 2 years. We suggest that in a effort to maintain good house keeping practices that the filters be changed out every two years. This however is highly dependent upon the specific conditions of each specific system.

1. "Using Atomic Force Microscopy to Evaluate Alloys For Corrosive Gas Service", S. Chesters and H.W. Wang, Microcontamination, June 1994.
2. S. Takahashi, S. Miyoshi, T. Kojima, T. Koyama and T. Ohmi, in Microcontamination '93 Conference Proceedings, p. 596, Canon Communications, Santa Monica, CA (1993).
3. "The Role of Moisture in the Corrosion of HBr Gas Distribution Systems", S.M. Fine, R.M. Rynders and J.R. Stets, J. Electrochem. Soc., Vol. 142, No. 4, April 1995.
4. "Reducing the Effects of Moisture in Semiconductor Gas Systems", E. Flaherty et al., Solid State Technology, July 1987, p.69.
5. "Joule-Thomson Expansion and Corrosion in HCl Systems", P. M. Bhadha and E.R. Greene, Solid State Technology, p. S3, July 1992.
6. "The Effects of Corrosive Gases on Metal Surfaces", P.M. Clarke. R.A. Hogle and S.M. Lord, in Microcontamination '93 Conference Proceedings, p. 433, Canon Communications, Santa Monica, CA (1993).
7. "Selection of 316L Stainless Steel for High Purity Semiconductor Gas Filter Assemblies", W. Murphy and B. Gotlinsky, STR PUF 21.