ty for

300mm Wafer FabricationROBERT TORRES, JOSEPH VININSKI, BELGIN YUCELEN & VIRGINIA HOULDING, Matheson Tri-Gas Advanced Technology Center,

Longmont, CO, USA

DAVID TREADWELL, CP Industries, Inc., McKeesport, PA, USA

JOHN W. fELBAUM, Digital Wave Corporation, Englewood, CO, USA

Gas

costs of maintenance of multiple systems and of labourfor cylinder changeouts continued to rise.

The transition from conventional gas delivery to bulk

speciality gas systems involves many teclmical considerations.Bulk systems must operate at much higher flow rates andsource pressures, and under different thermal conditions.

Condensable, corrosive speciality gases offer the largestchallenges to the gas providers. The nomJal operating condi-tions used in bulk speciality gas delivery require specialised

components such as regulators, valves, mass flowcontrollers and filters that operate at high pressure andhigh flow rates. Additionally, source gas purifiers will needto be altered to meet the new operating parameters ofbulk systems. Temperature effects must also be consid-ered seriously. Owing to the much larger thermal massof a ton unit, the insulating properties, Joule- Thomson

cooling, and the l~rge evaporative cooling that willoccur under high flow conditions, bulk systems must be

carefully designed to ensure that the gas is delivered underthe required conditions. Another deleterious factor forbulk systems is liquefaction of condensable speciality gasessuch as HCI, ~, DCS, TCS and ~. This condition is muchmore likely to occur in bulk systems thari in conventional

cylinders owing to the flow and pressure characteristicsof the bulk systems.

INTRODUCTIONSemiconductor manufacturing is trending towardslarger conventional fabs along with start-up of 300mm fabs. This cun-ent trend has placed an enormous strainon conventional gas cabinet delivery. Bulk deliveryoffers many advantages over conventional cylinders, withvery few disadvantages.

During the 1980s, speciality gas-handling systemsevolved from simple regulators with pigtail connectionsto automated, high-purity manifolds linked together onlocal area networks. This increase in complexity was drivenby semiconductor process requirements, manufac-turing demands and more stringent government safetyregulations. The request by semiconductor manufacturersfor >99.999% process gas purity supplied to multipleprocess tools with no flow interruptions has been a signif-icant challenge. Since many of the requested process gasesare corrosive, toxic or flammable, the need to usespecific materials and components continues to drive gascabinet prices upward. Additionally, conventionalspeciality gas delivery typically used one gas cabinet foreach speciality gas cylinder.

By the early 1990s, a new scheme using distributionboxes had been adopted by many wafer manufacturers.While this helped reduce capital expenditures, the

SEMICONDUCTOR FABTECH -14th EDITION 131

Figure 1

Ra surface roughness measure-

ments of nickel-Iined ton unit

The major cost saving associated with bulk systemsis the reduction in the unit cost of the consumable gasproduct. Bulk systems give the gas producer the abilityto offer the customer lower unit costs owing to the reduc-tion in cylinder preparation, analysis and rental.

Start-up costs are reduced for bulk-system installa-tions since a customer can realise the cost savingsassociated with installing one single bulk systemcompared with ten comparable gas cabinet systems. Other,ancillary advantages of a bulk distribution systeminclude simplified cylinder logistics, easier manage-ment of cylinder inventory, smaller footprint in the faband decreased transportation costs.

The safety benefits of bulk speciality gas systems area major factor in switching from conventional cylinderusage. Fewer source changeouts, fewer componentconnections, remote ~ placement, fewer monitor pointsand less material handling illustrate the safety advantagesof the bulk system. It was recently reported that 35 of 39silane safety incidents investigated were a result ofhuman intervention [1]. Additionally, cylinder changeoutswere responsible for 23% of incidents reported [2].

Product purity and product consistency are majoradvantages of bulk systems. Consider HCl as anexample. One tube trailer is equivalent to 350 cylinders,and one ton unit is equivalent to ten cylinders. This illus-trates that a wafer process can run uninterrupted fromthe same source gas 10-350 times as long as withconventional cylinder delivery. This point is critical ifone considers the root cause of contamination withingas distribution systems. It has been reported thatcontaminants at the wafer tool are due to the cylinderand the gas distribution system [3]. Gas purity levelsin the cylinder are tested at the production site andsometimes at the fab prior to connecting the cylinderto the gas distribution system. However, this onlyensures that the gas coming out of the cylinder meetsthe designated purity specifications of the cylinder.Unfortunately, many contaminants can be added to theprocess gas prior to reaching the tool or point of use.It has been reported that the gas distribution system isresponsible for 68% of the contamination in a CMOSprocess [4]. Impurity molecules can be introducedinto the gas distribution system during cylinder changing,inadequate purge procedures or exposure to contam-inated equipment. It should be noted that although the

Matheson Tri-Gas and other gas suppliers haveworked with component vendors to obtain the neces-sary equipment to deliver speciality gases in bulkquantities. Matheson is also currently involved in an exten-sive research programme to experimentally characterisethe physical and chemical properties of many of thespeciality gases when delivered at high flow rates andhigh pressures. A sophisticated model is under devel-opment to predict the concentrations of impuritieswithin the gas as a function of flow rate, pressure andtemperature, and as the gas is consumed.

Although we have outlined some of the difficulties ofusing a bulk speciality gas distribution system, the advan-tages are quite significant. The primary advantages are in:

.total cost of ownership

.safety

.product purity

.product consistency.

The cost of ownership of a bulk system offerssignificant reductions in expenditures on new capital,

operations, procurement, start-up, qualification,training and consumable gas. The primary savings incapital expenditures with bulk delivery systems comefrom the reduction of the piping and componentsnecessary to deliver the speciality gas. A bulk systemwill typically use two pressure reduction manifolds inan automatic switchover configuration serving multiplepoints of use, whereas traditional gas cabinets typicallyserve one point of use each. As an example, a toncontainer bulk system can provide the same gasdelivery capacity as ten standard gas cabinets atroughly one-fifth of the capital cost. Such a system alsoreduces the number of spare parts required to maintainthe bulk system to one-tenth of that required tomaintain a similar gas cabinet system.

Operational cost savings are achieved with a bulk systemthrough reductions in routine maintenance, labour, partsand materials, and reduced time to change cylinders. Ifa ton container is used, the cylinder changeouts arereduced to one-tenth of those for a gas cabinet system usingstandard gas cylinders. An additional operational cost savingcan be realised from the requalification of fewer processlines after changing gas sources.

132 SEMICONDUCTOR FABTECH -14th EDITION

Figure 2

Totol particle concentrotions {0.

Jlm! for a conventional carbon

steel ton unit compared with a

nickel-Iined ton unit

gas from the cylinder is changed and analysed frequently,the gas distribution components are sometimes exposedto the atmosphere, and gradually become contamina-tion sources for the wafer tool. This demonstratesthat fewer cylinder changeouts using a bulk system willimprove gas quality, minimise contamination sourcesdue to components and lengthen the lifetime of the gasdistribution system.

Tube trailers offer the largest advantages overconventional cylinders for the reasons stated above.However, tube trailers are a large capital investment forcustomers that are merely considering switching to bulksystems. Ton units offer the advantages of bulk deliverywithout the large initial expenditure. Ton units also offerthe customer the ability to evaluate the advantages anddisadvantages of bulk delivery on a small scale, beforefully committing to the idea. Additionally, tube trailersare prepared differently from a ton unit or conventionalcylinder, and require special passivation and processingtechniques in order to meet ultra-high purity specifi-cations. In fact, conventional speciality gas cylinders havealways had an advantage over bulk containers owingto the exotic, proprietary coatings, linings and passivationsdeveloped for cylinders.

Recently, CP Industries (CPI) collaborated withMatheson Tri-Gas to develop a bulk package for corro-sives that exceeds the performance of any conventionalcylinder package on the market. The CPI MIRRO-CLAD@ container is the next leap forward in deliveryof ultra-high-purity speciality gases. These nickel-platedcontainers are made from high-quality seamless steel pipe,with the ends hot-forged to form the cylinder. Theinterior is then electroplated with nickel- 200 and thecylinder is heat treated. Following surface conditioningand end machining, the nickel interior is polished to anaverage reading of <20 microinches Ra. The endclosures are manufactured from Hastelloy@ C-22@,which means that the entire gas-wetted surface of thevessel is corrosion resistant. These end closure glandsfeature a radial primary seal and a redundant face seal,both made of VITON@ or other suitable material. Forliquefied gases, dip tubes, which are GTA W welded ontothe gland, are also made from Hastelloy C-22, and areavailable on one or both outlets of the cylinder. The glandis secured by a closure nut which screws onto theexternal threads of the cylinder and is locked in place

with set screws. The external threads allow the packageto be sealed without disturbing the integrity of thenickel coating. Additionally, a conventional valve can beinserted into the Hastelloy gland before inserting the glandinto the bulk container. This ensures that none of themetallic particles generated during the valving proce-dure are trapped within the bulk container. This bulkcontainer is available in sizes up to 24 inches OD by 17feet in length. The container is typically designed forpressures up to 2400 psig.

Although the package can be used for nearly allspeciality gases, it has been demonstrated to be quite effec-tive with corrosive gases. HCI was selected for thepurpose of evaluating the package since it has beenproven to be highly corrosive to conventional carbon steelbulk containers, and is widely used for processing of wafersin the semiconductor industry. It is believed that HCI iswidely employed for plasma and thermal etching ofsilicon and gallium arsenide wafer surfaces prior toepitaxial growth. HCI is also used as an additive inpolysiliconetching and serves to getter unwanted impuri-ties during the thermal oxidation of silicon. Moisture inthe presence of HCI is a corrosion catalyst to purificationequipment, gas panels and components in semicon-ductor process equipment. Typically the gas distributionsystem has been constructed of 316L stainless steel or morecorrosion-resistant materials such as Hastelloy. Untilrecently, the carbon steel bulk container was always theweak link in the corrosion resistance chain. The corro-sion that occurred in the carbon steel bulk containerproduced impurities that were deposited into the gas distri-bution system in the form of volatile iron chloride andparticulate iron species. This iron contamination canremain as a residue on the surface of the piping and beincorporated into wafers in subsequent process runs. Inalmost all applications, iron contamination carried to thewafer will ioesult in degradation of the semiconductor device.The nickel-Iined bulk container eliminates the possi-bility of iron contamination since none of the wetted partsare made from material susceptible to corrosion.

This package has been extensively investigated, andresults are presented below from measurements ofsurface roughness, particle contamination, moistureconcentration, inboard leak rates, outboard leak ratesand metal contamination. Metals data from shelf life studieswas obtained after one year, and is also presented.

SEMICONDUCTOR FABTECH -14th EDITION 133

DATA

Surface roughness measurementsSurface roughness has been consid=l a critical parameterin semiconductor wafer p=ssing. A smooth surface elimi-nates or minimises abrupt changes in surface morphologythat may entrap and/or generate particulates. It has alsobeen discovered that in some cases the actual surface areacan be reduced by 90% [5]. Furthennore, it has been demon-strated that the coefficient of friction can be reduced byup to 75% [5]. These factors will have a direct impact onparticle concentration and on purge times of the container.Currently, the industry standard for surface roughness isconsidered to be 5--20 microinches Ra. Surface profilom-etry measurements were conducted on a nickel-Iinedbulk container. Typically, an average of five independent'Ra' measurements on the sidewalls of both ends was obtainedusing a cut-off length of 0.090 inches. This correspondsto three times the normal or recommended cut-off length.This was done to obtain additional data to adequately repre-sent the actual average roughness of the surface. The averageof each set of five readings from such a measurement mustbe less than 20 microinches for the unit to be accepted forshipment. Results from the ton unit that was used in theseexperiments are illustrated in Figure 1.

the sample outlet stream, insIde the H~P-101, providedthe necessary 0.1 cfm flow rate. The mass flow controllercontrolled the flow to :t2%, independent of the pressureover the range 40-150 psig.

The optical system used a 5 mW, 633 nm helium-neonlaser, an external mirrored quartz crystal oscillator anda photodiode array to count the particles in a gas streamin the path of the laser. This information was then sentto the particle counter readout, where it could bedisplayed as total counts, size differential counts orparticles/ft3. The particle counter had the ability to countparticles in eight different size ranges simultaneously. Theavailable particle size ranges ranged from 0.1 to 5 J1m.

The ton unit (nickel lined or carbon steel) was filledwith fIltered (0.01 J1ill) cryogenic N2 to 100 psig at a rateof 50 litres/min. Mter the ton unit was filled, thecylinder valve was closed. N2 was used to purge the sampleline of any residual particles. A baseline reading wascollected on the sampling manifold for 30 minutes, witha typical particle count of zero counts/ft3. After a stablebaseline was achieved, the purge gas was turned off, theton unit valve opened and the particle counter started.The particle counter readout was set to record particlesin one-minute intervals at a flow rate of 0.1 ft3/min. Thesample time for each test was 30 minutes. Figure 2 illus-trates the particle-counting results for the conventionalcarbon steel ton unit and the nickel-Iined ton unit.

The results show that the initial particle concentrations,as received from the factory, are quite high for bothcontainers. The particle concentration then increases aftera vacuum bake, owing to a much drier internal surface andto thermal desorption of particles. A proprietary Mathesoncleaning procedure was then applied to the nickel-linedcontainer only. The results showed that after more than 4fp of gas had been sampled, the particle concentration forthe nickel-lined ton unit was -SI particle per cubic foot ofgas for all particles of size 0.1 J1ill and larger. This exper-iment was repeated the next day with identical results.

Particle concentrations

Particulates are a large source of wafer contamination insemiconductor processing. Particulates are any bits of materialpresent on the wafer surface that have discrete, defmable

boundaries. As feature sizes shrink, the sizes of particulatesthat can cause defects also decrease. Particulate sources

include silicon dust, quartz dust, atmospheric dust,

particles originating from cleanroom personnel, processing

equipment, photoresist, bacteria, the gas distributionsystem and the gas container. The data presented hereindescribes work conducted to measure the particlesemitted from a conventional carbon steel ton unit and anickel-lined ton unit. The data was collected using aParticle Measuring Systems Inc. HPGP-1011aser particlecounter. The laser particle counter monitored particle concen-trations at pressures of 40 to 150 psig. The constructionof the wetted gas lines used electropolished 316L stain-less steel with VCR@ fittings. A mass flow controller in

Moisture concentration in inert gasAnalysis of moisture in nitrogen was conducted using anOlin hlStnnnents CO1p. model 71oocR ana1yser. The analyseruses the electrolytic current generated by the electrolysis

ofH2O in the sample gas stream. The current is directly

Figure 3

Analysis of moisture in nitrogen

from nickel-Iined ton unit

134 SEMICONDUCTOR FABTECH -14thEDITION

1

0.9

EO.8

!0.7

(3 0.6~.= 0.5

~ 0.4~-

.~ 0.3o~ 0.2

0.1

°

Figure 4

Analysis of moisture in hydrogen

chloride from nickel-Iined ton unit

The analysis was o direct

measurement of unpurified HCI

gas from the bulk container

0 10 20 30 40 50

Time (min>

proportional to the moisture concentration in the samplevia Faraday's la\\( Therefore, calibration can be easily achievedat the factory or in the field. The nickel-lined ton unit wasfilled to 1 DO psig with dry nitrogen and allowed to equili-brate completely. Moisture analysis of the unit at 100 psigrevealed a concentration less than 5DO ppb. Figure 3 showsthe results of the analysis over 6 hours.

Moisture concentration in HCI gasAnalysis of moisture in hydrogen chloride was conductedusing a Fourier transform infrared (FfIR) spectrometer.A Nicolet Magna 760 FfIR bench was equipped withadditional internal purge lines and an external purge boxto ensure low, constant background moisture levels inthe beam path. Purified nitrogen (> 1 ppb moisture) wasused as the purge gas. The lines and valves were heatedwith heating tapes to avoid adsorption and desorptionof moisture. A long-path-Iength cell was used for gassampling. The pressure in the FfIR cell was monitoredwith an MKS Baratron pressure gauge (0-1000 torr).Low noise levels and high sensitivities were obtained witha liquid-nitrogen-cooled MCTA detector. Sampling wasperformed at a resolution of 4 cm-1 with Happ-Genzel

apodisation.

Figure 5

Inboard leak rates of externally

threaded connection after

exposure to high-pressure nitrogen

LEAK RATE MEASUREMENTS

Inboard leak rateSince the externally threaded connection was new to themarket, a comprehensive leak check of the connection wasnecessary. The externally threaded connection flfSt under-went an inboard leak check. A Vaccum Instruments

Corporation model MS-170 helium leak detector was usedfor determination of the inboard leak rate of the externally

threaded connection. The Vacuum Instnunents CorporationMS-170 is a basic mass spectrometer tuned to 4 atomicmass units (helium). Leak rates of 1 cc/sec to 6 x 10-11 cc/seccan be detected with automatic or manual range selection.The response time is typically less than 1.5 seconds in the10-1° cc/sec range. The normal operating pressure is 1 x1 Q--6 torr, which is obtained by means of a mechanical forepump, a roughing pump and a diffusion pump. The leakdetector was calibrated by means of a NIST traceable heliumleak source (NIST test number Tl18 255729-95, NISTleak identification number NBSLCl18). Figure 5 displaysinboard leak rates obtained for the extemally ~ed connec-tion directly after exposure to high-pressure gas.

Outboard leak rate in heliumA Gow-Mac Instruments mode121-250 gas leak detectorwas used for determination of the outboard leak rate of

DATA ANALYSISA classicalleast-squa:res (CLS) method, implementedin Nicolet's OMNIC OuantPad software, was used fordata analysis. This CLS method mathematically combinescalibration spectra for moisture and HCl to match thesample spectra and calculates the actual concentra-tions corresponding to the individual calibration spectra.The method can analyse several components simulta-neously, and the residual spectra yield informationabout additional impurities. The algorithm requiresone to choose spectral regions where the variouscompounds have the least interference with one another.Several spectral windows between 3682 and 3988cm-1 were used for this work. The detection limits of themethod were calculated using linear regression analysis.This method takes variations of the calibration systemas well as changes in the bench moisture backgroundand the reproducibility of blanks into account. Thelimit of detection (LOD) for the experiments reportedhere was -0.075 ppm in a dynamic range between 0 and20 ppm. Figure 4 shows the moisture level in unpuri-fied HCl from the bulk container.

SEMICONDUCTOR FABTECH -14th EDITION 135

the TLD-l analyser nor the CM4 tape showed any signof silane leakage. This indicates that if a leak wereoccurring at the externally threaded connection, theemission level was less than 1 ppm silane in thesurrounding stagnant ambient air.

ANALYSIS IF METALIIN HCIIron contamination is a major metallic impurity in HCI whenthis gas is delivered in conventional carbon steel packages.Prior to the nickel-lined bulk container, iron contamina-tion from bulk HCI systems was unavoidable. Before theadvent of this container, the iron specification for HCI gaswas responsible for the majority of the total metal impuri-ties allowed from the cylinder. The SEMI StandardC3.35-95lists the iron specification for 99.997% pure HCIas 1000 ppb. The data reported below demonstrates thatiron is no longer a consideration when one is evaluatingthe total metals emitted from the gas package.

The metals analysis was conducted by first samplingthe HCI gas using a technique known as the hydrolysismethod. This method consists of bringing the gas outof the container through a corrosion-resistant metalmanifold and then collecting the gas in a Teflon@ vesselcontaining ultra-high-purity deionised water. Theanhydrous HCI from the cylinder or bulk container isdissolved in the water to produce aqueous hydrochloricacid. The aqueous acid is then analysed using inductivelycoupled plasma mass spectroscopy (ICP-MS). Typicaldetection limits using this analytical technique formetallic impurities are in the single-digit ppb range.

Mter HCI had been introduced into the nickel-linedcontainer, the HCI was allowed to remain in the cylinderfor 60 days prior to collecting the initial metals sample.The second metals sample was obtained more thanone year after the initial sample.

Figure 6 displays the primary metals of interestobserved in the initial metals sample. Figure 7 displaysthe primary metals of interest observed in the sampleobtained one year after the initial sample.

The results demonstrate that the iron concentrationis very low and is no longer a major contributor to thetotal metals emitted from the gas package. On thebasis of the duration of the HCI exposure, it appears thatthe package is not susceptible to corrosion, unlikeconventional carbon steel ton units.

Figure 6 (top)

Anolysis of metals in HCI from the

nickel-Iined ton unit after 60 days

exposure to HCI The dark bars

refer to actual values; all others

represent the LOD of the instrument

LODs are reported since the

concentration observed was

below the LOD

Figure 7 (above)

Analysis of metals in HCI from the

nickel-Iined ton unit after > 1 year

exposure to HCI The dark bars

refer to actual values; al{others

represent the LOD of the

instrument

the unit. The Gow-Mac 21-250 is a thennal-conductivitydevice with a detection limit of 1 x 10-5 cc/sec. The exter-nally threaded connection was tested at an internalpressure of 2380 psig helium. Mter probing for leaksfor more than 30 minutes, the leak detector was unableto detect a leak above the detection limit of the instru-ment. Additionally, the externally threaded connectionwas subjected to a temperature range of 32°F to IIo°Fwhile under pressure, and no leaks were found with theoutboard leak detector.

Outboard leak rate in silaneOwing to the insensitive response of the Gow-Mac leakdetector, the container was filled with a 3% silane,balance nitrogen mixture up to a pressure of 2300 psig.This was done so that a hydride gas analyser, which hasa much lower detection limit, could be used. The exper-iment used a 1W-1 hydride gas analyser with CM4 hydridetape. The principle of operation of this instrument relieson the hydride gas reacting with the CM4 tape so as tochange the colour of the tape from white to grey. The opticalcomponents of the instrument are factory calibratedsuch that the colour change is correlated to the concen-tration of the hydride gas. The detection limit of the analyseris 1 ppm for silane. Additionally, it is possible to visuallyobserve the colour change of the tape at concentra-tions at or above 1 ppm.

The externally threaded connection was pressurisedto 2300 psig for 20 minutes and wrapped with CM4hydride tape, while the connection was probed with theTLD-1 gas analyser. Mter20 minutes of testing, neither

CINCLISIINSThe experiments conducted to measure the perfor-mance of the nickel-Iined ton unit clearly demonstratethe improvement in the gas package over conventionalgas delivery containers. Surface roughness measurementson the nickel surface exhibited an average Ra value of6.6 microinches (0.17 pm). In contrast, conventional bulkcontainers are unpolished, with a typical surface rough-ness of > lOO microinches (2.5 pm).

Particle-counting results showed that the initialparticle concentration, as received from the factory,was quite high for both containers. The particle concen-tration increased after a vacuum bake, owing to a muchdrier internal surface and thermal desorption of parti-cles. A proprietary cleaning procedure was then appliedto the nickel-lined container. The results showed that aftermore than 4 cubic feet of gas had been sampled, the particleconcentration for this unit was ~1 particle per cubic footof gas for all particles of size 0.1 pm and larger.

Analysis of moisture in nitrogen was conducted.The nickel-lined ton unit was filled to 100 psig .withdry nitrogen and allowed to equilibrate completely.

136 SEMICONDUCTOR FABTECH -14thEDITION

ABOUT THE AUTHORS

Robert Torres is a principal research

scientist at Matheson Tri-Gas, working

in the areas of microcontamination,

corrosion and purification. He man-

ages a research group focusing on

investigations involving live gas testing

of components in gas distribution sys-

tems. This also includes development and testing of new materi-

als and packages to be used in corrosive environments.

Additionally, Dr Torres manages a research group dedicated to the

characterisation and development of new purifiers for inert,

hydride and corrosive gases. A member of the American Chemical

Society, Electrochemical Society and National Association of

Corrosion Engineers, Dr Torres holds degrees in chemistry from

the University of Wyoming (BS) and the University of Colorado

(PhD), and held a postdoctoral position at the National

Renewable Energy Laboratory. General research interests include

electrochemistry, sudace analysis and intedacial phenomena.

Moisture analysis revealed a concentration less than500 ppb. Analysis of moisture in HCI was conductedat flow rates from 2 SLPM to 900 SLPM. Resultsshowed that the unpurified moisture level neverincreased above 600 ppb.

The leak rate experiments consisted of inboard andoutboard measurements. Inboard leak rates weremeasured to be <5 x 10-1° cc/sec after exposure to 2200psig nitrogen. Outboard leak rates were measured inhelium and in a silane mixture. A helium outboard testrevealed no leak at a detection limit of 1 x 10-5 cc/sec.A silane outboard leak check was conducted becauseof the increased sensitivity of the hydride gas analyserrelative to the helium leak check device. The externallythreaded connection was pressurised to 2300 psig for20 minutes and wrapped with CM4 hydride tape,while the connection was probed with a TLD-1 gasanalyser. After 20 minutes of testing, neither theanalyser nor the tape showed any sign of silane leakage.This indicates that if a leak were occurring at theexternally threaded connection, the emission level wasless than 1 ppm.

A metals analysis was conducted by sampling the HCIgas using a hydrolysis method. The HCI was allowedto remain in the cylinder for 60 days prior to collectingthe initial metals sample. The results from this analysisrevealed an iron concentration of less than the detec-tion limit of the instrument «13 ppb). The second metalssample was obtained more than one year after theinitial sample. The results from this analysis revealedan iron concentration of 18 ppb. The results demon-strate that the iron concentration is very low and no longera major contributor to the total metals emitted from thegas package. It appears that the package is notsuscep-tible to corrosion.

John W. Felbaum is Chief Operating Officer at Digital Wave

Corporated. He has been involved with cylinder technology for 20

years and has also been chairman of the Compressed Gas

Association's Cylinder Specification Committee. He is currently

representing ANSI/CGA as the United States Technical Advisory

Group Leader for ISO 9808, Part 111- Carbon Steel Cylinders, for

ISO/TC58-WG7 -Hydrogen Compatibility Testing and for the

ISO/TC58/SC3-11120 Trailer Tube Specification. Mr Felbaum

holds a BS degree in metallurgical and materials engineering from

the University of Pittsburgh. He is a member of the American

Society for Testing and Materials. the American Society of

Mechanical Engineers. the Society of Automotive Engineers, the

American Welding Society, the American Society for Metals and

the Society for the Advancement of Material and Process

Engineering.

IF YOU HAVE ANY ENQUIRIES REGAROING THE

CONTENT OF THIS ARTICLE, PLEASE CONTACT:

Brenda P. Kennedy

Marketing Communications Manager

Matheson Tri-Gas

625 Wood Creek Drive

San Jose

CA 95112

USA

REFERENCES

[ 1] SEMATECH, "Silane safety improvement project S71 ",Technology Transfer Ooclnnent No. 94062405A-ENG,1994.

[2] SEMATECH, "Comparative analysis of a silanecylinder delivery system and a bulk silane installa-tion", Technology Transfer Document No.95092976A-ENG, October 31, 1995.

[3] R. Torres, D. Fraenkel, J. Vininski, E. Hennig, T.Watanabe and \I: Houlding, "High pressure POU purifi-cation of corrosive gases; effect on gas distributioncomponents", Semicon West Conference Proceedings,Workshop on Gas Distribution Systems, July 1998,San Francisco, pp. 11-118.

[4] S.D. Cheung and G.L. Mooney, "Designing, installing,a gas distribution system in a sub 0.5 ~m facility",Micro, October 1995, pp. 59-67.

[5] Anewalt, "Surface studies of semiconductor processpipes and tube: 316L stainless steel",Microcontamination, Apri11985.

ACKNDWLEDGEMENTIJohn W. Felbaum would like to acknowledge David W.Treadwell and Joe Tassone for being co-inventors of theMIRRO-CLAD package.

Robert Torres would like to acknowledge Dr GregJohnson of Matheson Tri-Gas for conducting the ICP-MS analysis, and Ehrich Diede of Diede PrecisionWelding, Inc. for welding and assembly of parts neededto conduct the particle analysis.

Tel: +1 (408) 971-6500 ext: 270

Fax: +1 (408) 275.8643

E-mail: bkennedy@matheson-trigas.com

Web site: www.matheson-trigas.com

137SEMICONDUCTOR FABTECH -14 :DITION