ULTRAPURE GASES

Certifying gas distribution systems in a fab producing advanced-technology devices

Ciro Napoletano, Texas Instruments Italy

Since the end of the 1980s the cleanliness requirements for ultra-high-purity (UHP) gas and chemical distribution systems in semiconductor manufacturing facilities have become increasingly stringent and the need for system upgrades has recurred at ever shorter intervals. When a 1989 industry forecast for acceptable contamination levels of iron in hydrogen peroxide and hydrogen fluoride is compared with the actual values specified through 1996, the curves of the forecast spec and the actual requirements soon separate, as the iron contamination requirements requested by the semiconductor industry dropped from 10 ppb to 10 ppt in less than 5 years.

Under these conditions, the need to improve on-site contamination control by developing new limits of detection, standards criteria, and laboratory tools quickly became obvious.¹ Additionally, gas distribution systems, having been identified as one of the areas in a fab most susceptible to contamination, became a logical focal point during the construction of ultraclean manufacturing facilities for next-generation devices.² Usually, every technology-level change in semiconductors demands a 10x improvement in the purity specifications for process-critical fluids.³ Clearly, setting and satisfying quality assurance (QA) requirements for gas distribution systems are keys to meeting the challenges of evolving technology. But achieving the required quality levels has been expensive. Indeed, the costs of UHP gas distribution systems have taken a major leap, caused mainly by the use of improved, higher-quality construction materials and components.⁴

Assembling the gas-distribution system at TI's AMOS fab required advances in both specifications and performance of system components. Photo Courtesy of TI.

Balancing the costs of technology changes with expected revenue gains is a major concern of the industry. One example of how technology and costs are related and move together rapidly is the upcoming move to larger-diameter wafers. The change to a larger diameter is being necessitated by the economies of scale possible in the fabrication of large die. The transition will improve the efficiency of silicon material usage through greater product complexity and thereby lower product costs, increasing the competitiveness of finished ICs.⁵ However, as set forth in The National Technology Roadmap for Semiconductors, the next-generation devices will require lower levels of contamination and wafer defects (e.g., metals, oxygen, and defects in bulk).⁶ Thus, only if a QA system for gas distribution system materials is developed will the goal of an increased wafer diameter and larger die be realized.

In the last six years Texas Instruments (TI) has developed two production lines to manufacture advanced IC DRAM devices at its plant in Avezzano, Italy (AMOS). The production lines, which were brought on-line in 1989 and 1995, were built to process 0.8- and 0.5-µm devices, respectively. This article discusses the QA requirements for the lines' gas distribution systems and the gas-line certification program used to satisfy the demands of the different device technologies.

Facilities Issues

Consisting primarily of piping, filters, valves, and purifiers, the gas distribution system represents the largest material-contact area for the gases used in semiconductor manufacturing. Each of these component types have experienced an evolution in both the materials of construction and method of fabrication. The techniques of assembling the components into a completed system have likewise gone through dramatic changes. Table I provides a snapshot of the overall advances in both specifications and performance as devices linewidths have decreased to 0.5 µm.

				Components
Device Structure	Integration Level	N2 Quality Gas (ppb)	H2O	Pipe	Welding	Filter	ESG Cylinders	$/C	Remarks
4.2 µm	16.64 Kb	3.0-4.0	10 ppm	316L, manual	Manual	1 µm	Stainless steel	x	—
1.5 µm	256 Kb	4.0-5.0	1 ppm	316L, manual, electro- polished	Manual	0.5 µm	Stainless steel	1.6x	—
1.2 µm	256 Kb	5	0.1 ppm	316L, auto., electro- polished	Auto.	0.05 µm	Stainless steel + electro- polished	6.6x	CO, CO2, THC
0.8, 0.5 µm	1 Mb; 4 Mb	5.0-6.0	50 ppb	316L, auto., electro- polished	Auto.	0.02 µm	Stainless steel + electro- polished	8x	ppb level, metals
0.5, 0.35 µm	16 Mb; 64 Mb	6.0-7.0	30 ppb	316L, electro- polished, VAR	Auto.	0.01 µm	Stainless steel electro- polished, IIIGEN	10x	Measured in situ, purifiers

Table I: Historical gas distribution system features as device linewidths decreased from 4.2 to 0.5 µm.

Piping material with a high level of purity already was available when construction of the AMOS Phase 1 plant began. It was later possible to use the same electropolished material during Phase 2 with only incremental changes in the required level of surface smoothness. Stainless steels produced by more-advanced technologies, such as vacuum arc remelt (VAR), were also considered, but these materials did not demonstrate the level of advantage necessary to justify their selection as the piping material. On the other hand, for certain specialty-gas applications selection of VAR material is justifiable, especially for valves and regulators.

A wide selection of available filters offered an opportunity to exploit newer technologies at points in the gas system where the advantage of doing so was significant. The filters installed for the Phase 2 line were assembled in a cleanroom and have a guaranteed 9-log retention efficiency for particles as small as 0.003 µm. The media itself is a high-pressure-differential material without hydrocarbon-outgassing potential. The valves and regulators were specified with VAR material, as mentioned above. Further, the valves for bulk gases were designed to be free of dead space. Integral purge valves were specified for all gases to facilitate qualification of assembled pipelines and reduce on-site welding requirements.

The on-site assembly of the piping system represented the greatest challenge to realizing UHP gas delivery, mainly because of the number of construction actions that had to be accomplished and the number of people who were needed to execute them. Fully developed procedures that are well supported by documentation were the key to successful system installation. For the Phase 2 project, for example, procedures were adopted that governed welding schedule development, welder certification, coupon witnessing, incoming material inspection, material storage, purging, and qualification. In recognition of its criticality, system qualification was executed by a third-party contractor dedicated to testing high-purity gas systems.

Gas Purity Specifications

It seems intuitive to assert that process gas purity must improve with each advance in device technology. However, experience suggests the bulk-gas containers and distribution lines are more likely to be sources of contamination than are the gases themselves. Approximately 95% of all contamination problems involving gases are operation problems and not directly related to gas quality. Examples include corroded valves, cylinder leaks, and mismatched fittings. To ensure they will not contaminate the gas, cylinders are electropolished, cleaned, and passivated internally, then filled with UHP nitrogen for certification based on the collection of particle data. They may be reinspected before reuse to review the passivation condition, but since there are no set maintenance procedures or shelf-life limits, in some cases contaminants can be released from a cylinder, especially if it has contained a corrosive or oxidant gas.

	Process Gases
Contaminant	CF4	SF6	N2O	PH3/N2	BCl3	Ar
H2	<0.05	<1	<50	—	1.2	6.2
Ar + O2	NA	<0.5	<50	—	6.8	—
N2	1.4	<0.5	157	—	>5000	<0.1
CH4	0.03	<0.1	<20	—	—	0.05
CO	0.05	<0.5	<50	—	—	0.04
CO2	0.17	<1	—	—	—	0.11
H2O	0.05	<1	—	—	—	0.18
O2	0.15	NA	—	—	—	0.05
SF6	NA	NA	—	—	—	—
CF4	—	<1	<100	—	—	—
PH3/N2	—	—	—	7940	—	—

Table II: Specialty gas analysis report. All units are parts per million. (NA = not applicable.)

Purity-level investigations of several gases as received from suppliers were carried out at the AMOS facility in cooperation with a contract laboratory and it was found that most of the gases did not differ significantly from the suppliers' certification data, confirming gas quality and the reliability of vendor claims for contaminants of interest (see Table II). When process problems have indicated gas contamination as a root cause, our experience has confirmed that the critical issues related to gases are concentrations, moisture, and oxygen control; hydrocarbons and metals can be involved in the case of cylinder contamination. The hypothesis that low levels of gas impurities contribute to process defects cannot be dismissed but has not been proven. We have seen air in a cylinder of sulfur hexafluoride (SF^₆), and found that hydrogen bromide (HBr) with a higher quality specification for hydrogen and hydrogen chloride (HCl) was needed to meet process specifications for polygate profile etching.⁷ Such etching is among the process steps at Avezzano that are more critical than others and require a purer gas. To be specific, polygate profile etching requires an HBr product of ULSI quality, while at metal oxide contact an HBr of VLSI grade is adequate (Table III compares the specifications of both products). In addition, there are many cases where contamination from the tools used for a process cancel the benefits derived from the use of UHP gases. For example, plasma etching with a resist mask leaves such reaction by-products as water vapor, carbon oxides, and organics.⁸

Purity Specification	VLSI HBr	ULSI HBr
N2	<40	<3
O2	<40	<1
CO	NS	<5
CO2	<20	<5
H2O	<3	<3
H2	NS	<500
HCl	<1500	<500
Cu	<5	<0.01
Fe	<5	<0.5
Mg	<5	<0.01
Ni	<5	<0.1
Cr	<5	<0.1
Co	<5	<0.05
THC	NS	<5

Table III: Comparison of the specifications for VLSI and ULSI HBr gases used for metal oxide contact and polygate profile etching, respectively. All units are parts per million. (NS = not specified.)

System Certification

As stated above, the distribution system materials used for the Phase 1 and Phase 2 projects were similar; both the tubing and the components were passivated, electropolished 316L stainless steel meeting the Semaspec requirements. The tube materials provided by all the suppliers had the same specifications for <0.005-ppm sulfur, a surface chromium/ iron ratio of 5.1:1, and a profile analysis of 1.1:1 with a chromium oxide thickness of 75 Å.

The certification process was performed in two steps: the main pipelines were tested first, followed by the piping from the submain valves to the tools. The system was filled with UHP nitrogen for all tests. Certification data for the Phase 1 project are reported in Table IV for gas-related tools. The tools are identified as CVD (chemical vapor deposition), ETC (etchers), IMP (implanters), and FUR (furnaces), along with a MISTI number that identifies the specific tool and its position in the wafer fab. Multiple tools are connected to each gas pipeline.

Gas	Tool	Oxygen (Spec: <30 ppb)	Moisture (Spec: <50 ppb)
HCl	FUR051-061 071-091-	10	29
HCl	ETC201-2-3-4	8	24
SiH4	CVD401-2	23	25
SiH4	CVD501-2	23	29
N2O	FUR111	1	28
N2O	FUR121	14	26
N2O	CVD 401-2	1	45
CHF3	ETC110-1-2 401-2-501-2-3-611-	7	35
CF4	ETC301-2-3-4- 101-2-3-4-5-6- 7-8-9-901-2	15	32
CF4	CVD501-2	22	30
Cl2	ETC601-2-611	20	27
Cl2	ETC801-2	20	32
SiF4	ETC701-2	5	20
NF3	CVD201-2-401-2	15	28
0.8% PH3	CVD401	20	36
Ar	IMP011-2-3- 001-2-3-4-5- ETC401-2	21	30

Table IV: Certification data from the Phase 1 project for single-gas pipelines and equipment tools. The oxygen and moisture data refer to measurements in UHP nitrogen.

As outlined in Table V, the test series included a pressure test, a pressure decay test, helium leak testing, and particle counts. Inboard leaks and moisture were the key considerations from a quality standpoint because of the potential for external contamination of the lines due to leaks and the potential for internal corrosion due to moisture. Usually, because of the high quality of the gas and newness of the filters, it was easy to achieve the particle-level specification in both Phase 1 and Phase 2. An example of Phase 2 particle data is provided in Figure 1.

Figure 1: Particle data from Phase 2 certification testing.

Certification Test	Phase 1	Accuracy	Phase 2	Accuracy
N2 pressure test	10 kg/cm2 no loss per 30 min	1%	10 kg/cm2 no loss per 30 min	1%
N2 pressure decay	7 kg/cm2 no loss per 24 hr	0.25 kg/cm2	7 kg/cm2 no loss per 24 hr	0.25 kg/cm2
He inboard leak	10 x 10-9 atm cm3/sec	2 x 10-1 atm cm3/sec1	10 x 10-9 atm cm3/sec	2 x 10-11 atm cm3/sec
N2 blowdown	flow = 30 m/sec	—	flow = 30 m/sec	—
Particles with N2	Avg. = <1 particle/cu ft; max = <5 particle/cu ft;	>0.1 µm	Avg. = <1 particle/cu ft; max = <5 particle/cu ft;	>0.1 µm
Oxygen	50 ppb	—	NS	—
Total hydrocarbons	50 ppb	—	NS	—
Moisture (H2O)	50 ppb	5 ppb	30 ppb	2 ppb

Table V: Gas distribution system certification requirements. (NS=not specified).

After the particle tests, oxygen, total hydrocarbons, and moisture tests were performed. During the Phase 1 certification oxygen and hydrocarbons were consistently within specifications, but in several cases moisture was found in the pipeline; it was then necessary to flush with nitrogen until the piping met the moisture specification. Although moisture and oxygen can be thought of as related from a chemical standpoint, no correlation was found according to the certification tests performed in Phase 1. The reason is related to the length of time gas was exposed to the analyzers. Figures 2 and 3 show the results of the Phase 1 moisture tests for the lines for bulk process gases (nitrogen, oxygen, helium, and hydrogen) and specialty gases, respectively. For the bulk-gas lines, moisture was always within a range of 20—40 ppb.

Figure 2: Moisture data from Phase 1 certification testing of bulk-gas lines: (a) process nitrogen, (b) oxygen, (c) hydrogen, and (d) helium. In all cases the moisture specification was <50 ppb.

Figure 3: Moisture data from Phase 1 certification testing of specialty gas lines. The moisture specification was <50 ppb.

Because of the experiences in Phase 1, in Phase 2 several tests that contributed only moderately to the quality certification of the gas system were eliminated, including oxygen and total hydrocarbons measurements. Instead, moisture was considered the key parameter and the specification was tightened to 30 ppb. As seen in Figure 4, moisture data from Phase 2 certification in UHP nitrogen was within a range of 15—30 ppb. The tighter limit did not result in dramatically different results than those of Phase 1 because the analyzer was disconnected immediately after the specification was achieved.

Figure 4: Moisture data from Phase 2 certification testing of process nitrogen line. The moisture specification was <30 ppb.

The effect of the certification goals reported herein on device performance was evaluated in correlation studies of yield results from sister wafer fabs producing identical product. The results of these broad correlations were positive; results were also very good from the aspect of process test releases. For example, oxidation processes such as DZ annealing and tank drive did not deviate from the worldwide TI baselines in terms of oxide performance and defect formation. Comparing actual production data with process baselines confirmed that the gas system designs work even better than expectations, so the lessons learned during these projects should be useful in setting future goals.

Future Trends

As device technology continues to evolve there will be a continuing demand for new wafer fabs with increasingly stringent contamination requirements. The possible next-generation requirements for bulk- and specialty gas distribution systems are outlined in Table VI.⁹ Several questions arise: How can the semiconductor industry measure the parameters set forth in the table and define reliable limits? Once that is done, how can those parameters that respond to many variables, including gas containers and shipping effects, be controlled to the point of use?

Parameter	Bulk-Gas Specification	Specialty Gas Specification
Pressure decay	<1% (24 hr­full system) <1% (2 hr­segment)	<1% (2 hr)
Helium leak check	3 x 10-9 atm cm3/sec	3 x 10-9 atm cm3/sec
Particles (>0.1 µm)	0.5/cu ft	5/cu ft
Moisture	5 ppb	50 ppb
Oxygen	5 ppb	—
Total hydrocarbons	10 ppb	—

Predicted future certification requirements for bulk-gas and specialty gas distribution systems.

One tentative answer is to focus more on system contamination rather than on contamination of the gases themselves. For example, bulk-gas lines will need to maintain continuous flows in order to guarantee that specifications will be maintained, and analyses will need to be performed in situ rather than in the laboratory. In addition, better protection will be needed at the points of use. For instance, ensuring that moisture levels in nitrogen will be <5 ppb in the tool chamber after the gas has passed through 100—200 m of pipeline will require installation of a purifier at the tool inlet. Specialty gases have even tighter requirements in this respect because some are corrosive and contain significant amounts of moisture (ppm levels). The use of on-site manufacturing plants to produce such chemicals as hydrogen peroxide, hydrogen chloride, and hydrogen fluoride from an industrial gas phase and deionized water could also help guarantee that stringent purity specifications are achieved by eliminating drum and shipping effects.

Future gas distribution systems will probably require lower levels of moisture and other key contaminants, but it may not be necessary to tighten requirements for all contaminants. If a new system is guaranteed leakproof and includes adequate in-line purifiers and filters, has been certified via a process such as that described above, and is accompanied by on-line process monitoring, current specifications may be sufficient. This hypothesis is based on the similar process performance results obtained for the two AMOS production lines built to meet the requirements of different DRAM technologies. There were four main factors that contributed to the achievement of these results: (1) design, certification, and installation procedures were kept within very tight specifications; (2) the systems deliver high-quality gases to the points of use without additional contamination; (3) system-commissioning procedures were followed by subcontractors, and in Phase 2 the QC subcontractor was an independent company that controlled the installation subcontractors; and (4) AMOS data were correlated with that of its sister companies abroad. It is not impossible that current systems are already in the proper condition for the next generation of device technology; they meet the requirements of The National Technology Roadmap for Semiconductors except for moisture in bulk gas, which is 1 ppb.⁶

In conclusion, as device technology evolves, gas purity will remain an important concern in the semiconductor industry but the QA emphasis will shift from the purity of the gases themselves to distribution system certification.

Acknowledgment

The section on facility issues was written by Ray Niemiec, facility manager of the Avezzano plant, who also provided helpful advice on other sections of this article.

References

1. Caputo C, Della Vedova S, and Napoletano C, "Monitoring of a Deionized Water Production Plant," TI Journal, 13(6):51—55, 1996.

2. Hauser PM, "Ultraclean Manufacturing: If Not Now, When?," Microcontamination, 11(7): 24—28, 1993.

3. Cheung SD, Mooney GL, and Jensen DL, "Designing, Installing a Gas Distribution System in a Sub-0.5-µm Facility," MICRO, 13(9):59—67, 1995.

4. Klastermejer J, "Design of UHP Gas Distribution System," presented to the UHP Gas SAES Getters Workshop, Nice, France, October 1994.

5. Grimes M, "Economics and Materials Science of Large Diameter Silicon Wafers," TI Journal, September, pp 5—12, 1995.

6. The National Technology Roadmap for Semiconductors, San Jose, CA, Semiconductor Industry Association, 1994.

7. Napoletano C, Miccoli G, and Russo F, "Effect of HCl HBr Chemistry on Polygate Profile Etch," TI Journal, May, pp 12—18, 1995.

8. Illuzzi F, and Gualandris F, "Silicon Sub-Half-Micron Technology: The Liquid and Gas Purity Impact," presented to the UHP Gas SAES Getters Workshop, Geneva, March 1996.

9. Collier P, "0.35-µm Certification Requirements for Bulk Gases," TI internal information, 1995.

Ciro Napoletano is a member of Texas Instruments' group technical staff. He is responsible for new materials introductions and new technologies activities at TI Avezzano, Italy, where he has worked since 1989. He has also been involved in QA activity for the company, including quality systems and chemical laboratory supervision as well as supplies management and sourcing of new chemical, gas, and wafer materials. Napoletano graduated with a degree in chemistry from Rome University—Laurea (1985) with a thesis in organic semiconductor synthesis and applications. He then joined Fiat's R&D chemistry center. He has coauthored more than 15 papers in the fields of chemistry, DI water, polymers, and gases and serves on several Italian chemistry committees. (Napoletano can be reached at 39 863 42 3351, or via E-mail at ciro@msg.ti.com.)

MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit