High Flow Delivery Systems for Bulk Specialty Gases

Belgin Yficelen, Joseph Vininski, and Robert TorresMatheson Tri-gas

Advanced Technology CenterLongmont, Colorado

Biography

Dr. Belgin Yiicelen is a Senior ResearchEngineer at the Matheson Tri-Gas AdvancedTechnology Center. She received her Ph.D. inChemical Engineering from the Colorado Schoolof Mines. Prior to joining Matheson Tri-Gas, herindustrial background included petroleumrefining, fluid-bed combustion systems and phaseequilibria. At Matheson Tri-Gas, Dr. Yiicelen'sresearch involves developing heat and masstransfer models for corrosive gases, designinghigh purity gas distribution systems, anddeveloping analytical techniques for the detectionof trace gas impurities. She has been a memberof AIChE since 1997.

Abstract:Two independent high-flow bulk deliverysystems were designed and built. These systemswere employed to investigate the moisture andthe temperature changes and to test the efficiencyof the purification methods for hydrogen chlorideand ammonia as a function of flow rate.

The moisture concentration in gas phase forhydrogen cWoride and ammonia was measuredby an FI1R spectrometer. The efficiency ofMatheson Tri-gas' Nanochem@ series purifierswas tested at a wide range of flow rates. Heatwas introduced to the Bulk Specialty Gas Unit(BSGU) and the distribution system during thepurification experiments to ensure saturatedvapor phase in the purifiers. The results fromtesting bulk specialty gas purifiers for ammoniaand hydrogen cWoride under severe moisturechallenges and various flow rates are described.

The temperature changes were monitoredthroughout the delivery manifold and the BSGUto accurately characterize evaporative coolingand Joule-Thomson effects. A comprehensivemodel based on heat transfer concepts wasdeveloped to estimate the temperature profilewithin a BSGU. The temperatures predicted bythe theoretical model were compared with theexperimental data to assess the reliability of theheat transfer model. The results of the

temperatute measurements and the modelpredictions are reported for hydrogen chlorideand ammonia for flow rates up to 900 and 500Urnin, respectively.

Introduction:Bulk Specialty Gas Systems (BSGS) are nowreplacing the traditional cylinder delivery systemsdue to their lower cost, safety, product quality,consistency and reduced cylinder changes [I ].BSGS are also capable of high flow rates (up to1500 slpm) which will be required by thesemiconductor industry moving towards largertabs along with 300 mm wafers. Information onthe temperature and moisture variations at highflow regimes is critical to the design andoperation of bulk distribution systems in order toensure the supply of high purity gas to theprocess tools.

Hydrogen chloride is employed for the plasmaand thermal etching of silicon and galliumarsenide wafer surfaces prior to epitaxial growth.A small amount of water may cause a reductionin the etching selectivity. Moisture in a corrosivegas such as hydrogen chloride also acceleratescorrosion leading to metallic particulatecontamination within the gas distribution system[2], [3]. These particles can be carried to the

Workshop on Gas Distribution SystemsSEMICON West 2000

@ SEMI 2000

G-I

process tools and have a direct impact on thewafer yield. Ammonia is used for the depositionof silicon nitride layers in integrated circuits andfor MOCVD growth of gallium nitride films inLED's. Moisture can adversely affect theperformance both processes.

Source gas and point of use purification (POU) ofspecialty gases minimize the effect of particulatecontamination caused by corrosion [2] and reducethe chemical contamination to negligible levels.Therefore purification of gases is necessary toprotect down stream components from corrosionand deterioration and to improve the processquality. Matheson Tri-gas' Nanochem@ seriespurifiers provide effective gas purification inmulti-tool and single-source applications.Utilization of these purifiers for BSGS presentsmany challenges due to the problems associatedwith delivery of condensed gases at high flowrates. The effects of severe temperature drops,liquefaction of the gas, and large linear velocitiesmust be considered to ensure optimal purifierperformance.

Data:Experimental ProceduresTwo independent high flow gas distributionsystems with 3/8" and Ill" 316L stainless steeltubing were used for the experiments. Theschematic of the gas distribution system forhydrogen chloride is shown in Figure 1 and forammonia in Figure 2. Temperature changes weremonitored throughout the delivery manifold andthe BSGU using type K thermocouples interfacedto a computer. For the purification experimentsthe BSGU was heated with 5000 W and 5760 Wheat blankets for hydrogen chloride and ammoniaBSG units, respectively.

Moisture experiments were run with BSGUcontainers of hydrogen chloride and ammoniathat were prepared with known amounts ofmoisture. A Nicolet Magna FfIR bench wasequipped with additional internal purge lines andan external purge box to assure low and constantbackground moisture levels in the beam path.

Purified nitrogen « 1 ppb moisture) at 20 LIminwas used as purge gas. Lines and valves wereheated with heating tapes to avoid adsorption-desorption of moisture. A 10m path lengthnickel coated cell with gold coated mirrors andIR quartz windows was used for gas sampling.The pressure in the FfIR cell was monitored witha MKS Baratron pressure gauge (0-1000 torr).Low noise levels and high sensitivities wereobtained with a liquid nitrogen cooled MCTAdetector. Sampling was performed at a resolutionof 4 cm-l and Happ-Genzel apodization. A 1/8"316L sample line was connected to the high flowmanifold down stream of the purifier. This lineflowed a continuous sample (at 2 Umin) to theFfIR. The 118"sample line was heated and heldat a constant elevated temperature. The set upalso allowed the purifier to be by-passed. Allflows were adjusted with metal seated meteringvalves and measured with rotameters.

N2

HighPressurePurifier

FTlRHeatedlines

Figure 1. Gas distribution system for hydrogenchloride

N2

NHaTonner

Heated /1Une

Figure 2. Gas distribution system for ammonia

Workshop on Gas Distribution SystemsSEMICON West 2000

@ SEMI 2000

G-2

Moisture Concentration MeasurementsHydrogen ChlorideThe moisture concentration in the gas phase wasmeasuredfor a wide range of flow rates from I to900 Umin. The concentration was stablethroughout the experiments with a range from0.24 to 1.22ppm and was not a strong function offlow rate as seen in Figure 3. The averagemoistureconcentration was 0.63 ppm.

1.4

1.2II>~..E~ 8: 1.!!~0 c:E 0 0.811>:;:'" ..! E0.6Q.II>'" .... :50.4Clo

0.2

. .. .0

0 200 400 600 800 1000

Row Rate (Umin)

Figure 3. Change in the moisture concentrationin gas phase hydrogen chloride as a function of

flow rate.'!!oislure challenge: 10 ppm :JAmmonia

The moisture concentration in gas phaseammonia was measured at flow rates up to 500Umin. The water concentration increases withflow rate, reaches a maximum, then declines asseen in Figure 4. This behavior was observed intwo independent experiments and is believed tobe a result of the combined effects ofthermodynamics, fluid dynamics and heattransfer. It is difficult to model the waterconcentrationas a function of flow rate but sometheoretical explanations are under considerationto describe this interesting phenomenon.

4

3.5"'~:; K 3~.5~:5 2.5:~ 2...::~ ~ 1.5",g(;8 1

0.5

..I

iI

II.Ex~.11I ..Ex.2!,I

I

I

...

. .. ...

00 200 400

Row Rate (Umin)

600

Figure 4. Change in the moisture concentrationin gas phase ammonia as a function of flow rate.Moisture challenge: 40 ppm (liquid phase)

Purification ExperimentsHydrogen ChlorideAn 8L Nanochem@ bulk hydrogen chloridepurifier was tested at cylinder pressure (627.7psia at 70 OF) for source gas purificationapplications. Heat was added both to the BSGUand the distribution manifold to ensure thathydrogen chloride was being introduced to thepurifier as gas phase. With moisture challengesup to 10 ppm. the purifier performance was testedat three different flow rates and the results areshown in Figure 5.

12purifier

: :5O!i-:: :Umin:. : : : +

i J:50 i. 400. ,:lJmin : + Umir>: :~

I I

! I

..+- bypass--

10...~:; K 81.;.5~S 6"'1;:.:: 4s::.cQ.CII

<II ~ 2"'0Go

700 250Umin UmiD. .

2U. . : :

~i I

: 100

,-0-2

0, .

100 200

Time (min)

300 400

Figure 5. Moisture concentration in gas phasehydrogen chloride with high-pressurepurification

Workshop on Gas Distribution SystemsSEMICON West 2000

@ SEMI 2000

G-3

It .wasassumed that there are three modes of heattransfer; heat loss due to evaporation of liquidammonia, heat gain by convection from air andheat conduction within the liquid. The

temperature distribution was detennined from the Figure 7. Model predictions and experimentalsolution of the one-dimensional, unsteady-state data measured on the BSGU surface below thecylindrical heat conduction equation subject to a liquid level line for hydrogen chloride

Workshop on Gas Distribution SystemsSEMICON West 2000

@ SEMI 2000

Ammonia

The efficiency of a Nanochem@ bulk ammoniapurifier was tested at flow rates of 2, 50, 100,250, 500 and 700 Umin and a moisture challengeof 40 ppm (liquid phase). Heat was applied to theBSGU and the distribution manifold during theseexperiments. The results of the purification testsshowed that the moisture concentration wasreduced to below the FfIR detection limit of 100ppb at all flow rates. Figure 6 shows the moistureconcentration in the gas phase at each flow rateusing a purifier together with the results at 400Umin bypassing the purifier.

12purifier

: :500: :: :Urnin:. : :

700 250Urnin Urnin

j +: ~j50 :+ 400:lJmin :. lJmin: : ,.

I I

: j

. +- bypass-

10CI>~:; [ 8"'0.I/)~

"0 c 6:!Eo

CI>~::~ 4.r:.CQ.~I/) C 2'" 0CJo 2~ i I i

: 100

j Urnin

0

-20

. .100 200

Time (min)

300 400

Figure 6. Moisture concentration in gas phaseammonia withpurification

Heat Transfer ModelA model was developed to determine thetemperature distribution in a tonner based on heattransfer concepts. Information on thetemperature profile inside the tonner is essentialfor the understanding of the boiling mechanismand to estimate the pressure inside the BSGU inorder to be able to maintain a constant flow.

set of appropriate boundary and initial conditions.It was assumed that the system was cylindricallysymmetric and that the volume of the liquid wasthe only variable. The finite difference methodwas employed to approximate the partialdifferential equations of heat conduction by a setof algebraic equations in both space and timedomains at a number of nodal points. For theradial temperature profile 5 nodes were used anda total of 54 ordinary differential equations weresimultaneously solved for 54 unknowns.

The validity of the model developed wasestablished by comparing the predictions with theexperimental data for both ammonia andhydrogen chloride. The data were taken using athermocouple placed on the BSGU wall belowthe liquid level. No heat was supplied for theseexperiments. This model can also be used topredict the efficiency of heat transfer of variousmethods by simply measuring the temperature ofthe BSGU surface.

Hydrogen ChlorideThe comparison of the model predictions with theexperimental data for hydrogen chloride is givenin Figure 7 for 100, 500 and 900 Umin. Theseresults show that the model can predict thesurface temperature of the BSGU for hydrogenchloride with an average deviation of ::to.3°Cbetween the calculated and the measured values.

30 Experimental - Model

25

e2O~.a 15'"(;;~10..I-

100 Umin

900 Umin5

0

0 10 20 30

Tme (min)

40 50

G-4

The predicted temperature profile at 5 nodalpoints for 900 Umin is illustrated in Figure 8.The temperature at the center of the lonner goesdown to temperatures below -110°C (hydrogenchloride freezing point) after 15 minutes. Theseresults may suggest the formation of hydrogenchloride ice. However. it should be noted that athigh flow rates. the vigorous boiling mechanismmay increase the mixing process and decrease thetemperature gradients. As a result. thetemperature distribution may be more uniformand the possibility of having hydrogen chlorideice inside the tonner will be reduced. But. thepressure calculated from the predicted averagetemperature correlates very well with theexperimental pressure at all flow rates. This alsoconfirms the validity of the model.

so

54

00-~~ -so::sa;8:100

1freezin9J)()int .....

i. ofHCI@1atm1--150

3

2Tonne,

1

-200

0 10 15

Time (min)

20 255

Figure 8. Temperature profile for hydrogenchloride BSGU at 900 Umin

AmmoniaFigure 9 shows the model predictions togetherwith the experimental data measured at theBSGU surface for ammonia at 100. 250 and 500Umin. The model predictions were generally invery good agreement with the data and theaverage deviations between the calculated and themeasured values were :tOAoC throughout thecourse of the experiments.

40 Experimental- Model

30

E2OQ)

:Je10Q)a.

E 0{!

-10 500 Um!n

-200 40 60

Tame(min)80 10020

Figure 9. Model predictions and experimentaldata measured on the BSGU surface below theliquid level line for ammonia

The predicted temperature profile at 5 nodalpoints for 500 Umin is illustrated in Figure 10.The calculations were made until there was no

more pressure inside the tonner to provide flow.

so

~ 00

~-50 .5 ~_aml,llQOla0;-100~ point@1atm~~1SOto1--200

GTonne,

-250

0 40 60

Time (min)

60 10020

Figure 10. Temperature profile for ammoniaBSGU at 500 Umin

These predictions show that running only for 90minutes may cool the center temperature to below-200°c. These results may suggest the formationof ammonia ice (freezing point of ammonia = -77.7°C) assuming the mixing process does noteffect the temperature gradients. However, thefact that the measured and the calculated

pressures agree for all flow rates indicates thatthe effect of mixing is negligible.

Workshop on Gas Distribution SystemsSEMICON West 2000

@ SEMI 2000

G-5

Conclusions:

Water is differently distributed between the liquidand the vapor phase for each gas. Both hydrogenchloride and ammonia are gases in which water isvery soluble, yet they have dramatically differentflow rate dependence of moisture. The hydrogenchloride moisture level was not a strong functionof flow rate. The performance of Matheson Tri-gas' Nanochem@series purifier was consistent forall flow rates tested at both low and highpressures. At high moisture the average moistureobtained downstream of the purifier was reducedsignificantly for low and high-pressurepurification.

The ammonia moisture concentration increasedwith flow rate reached a maximum then declinedat higher flow rates. Further studies are underdevelopment in order to explain this behaviorwith the combined effects of thermodynamics,fluid dynamics and heat transfer. A cylinder isbeing built to observe the boiling mechanism andmodel the boiling regimes of ammonia as afunction of flow rate. Matheson Tri-gas'Nanochem@series purifier reduced the moistureto a level that would not be detrimental tostainless steel [4]. The purification of the gaslowered the moisture level to below the FTIRdetection limit at an flow rates tested, with highmoisture challenges.

An integrated approach combining experimentaland modeling techniques was presented tocharacterize the thermal and physical propertiesof a bulk delivery system. The heat transfermodel showed good agreement with theexperimental data for both hydrogen chloride andammonia at all flow rates. The mixing effectsdue to temperature gradients and evaporativeboiling were not taken in to consideration but thismodel allows us to predict the average pressureinside the BSGU and the efficiency of heattransfer for various methods.

References:

L R. Torres, J. Vininski, and E. Hennig, «UltraHigh Purity Bulk Specialty Gas Package for

300 mm Wafer Fabrication", Semicon WestConference Proceedings, Workshop on GasDistribution Systems, July 1999, SanFrancisco, California, 11-110.

2. R. Torres, D. Fraenkel, J. Vininski, E.Hennig, T. Watanabe and V. Houlding, "HighPressure POD Purification of CorrosiveGases; Effect on Gas DistributionComponents," Semicon West ConferenceProceedings, Workshop on Gas DistributionSystems, July 1998, San Francisco,California, 11-117.

3. R. Torres, V. Houlding, S. Lau and T. Bogart,«Evaluation of the Performance of theDiaphragm Valves in UHP Chlorine Gas",1998. Semicon China 98 Technical

, Symposium.

4. S. M. Fine, R. M. Rynders, J. R. Stets, "TheRole of Moisture in the Corrosion of HBr GasDistribution Systems", J. Electrochem. Soc.,Vol. 142, No.4, April, 1995, 165-175.

Aclrnowledgements:We would like to thank to Scott Thompson andSean Williamson from the La Porte branch ofMatheson Tri-gas and to Joe Giagnacova and hisgroup from the Montgomeryville branch ofMatheson Tri-gas for their assistance in buildingthe distribution systems used in this study.

We also would like to acknowledge Ehrich Diedeof Diede Precision Welding, Inc. for welding andassembly of the distribution manifolds.

Workshop on Gas Distribution SystemsSEMICON West 2000

@ SEMI 2000

G-6