Development of a spatially controllable chemical vapor ... of a spatially controllable chemical vapor deposition reactor with combinatorial processing capabilities J. O. Choo and R

Development of a spatially controllable chemical vapor deposition reactorwith combinatorial processing capabilities

J. O. Choo and R. A. Adomaitisa!

Department of Chemical Engineering and Institute for Systems Research, University of Maryland,College Park, Maryland 20742

L. Henn-Lecordier, Y. Cai, and G. W. RubloffDepartment of Materials Science and Engineering and Institute for Systems Research,University of Maryland, College Park, Maryland 20742

sReceived 16 November 2004; accepted 12 February 2005; published online 23 May 2005d

Most conventional chemical vapor depositionsCVDd systems do not have the spatial actuation andsensing capabilities necessary to control deposition uniformity or to intentionally inducenonuniform deposition patterns for single-wafer combinatorial CVD experiments. In an effort toaddress these limitations, a novel CVD reactor system has been developed that can explicitly controlthe spatial profile of gas-phase chemical composition across the wafer surface. This paper discussesthe construction of a prototype reactor system featuring a three-zone, segmented showerhead design.Experiments are performed to assess the ability of this reactor system to deposit tungsten films bythe hydrogen reduction process; segment-to-segment process recipes are controlled to depositspatially nonuniform W films. The capabilities of this reactor system for materials discoveryresearch are discussed.© 2005 American Institute of Physics.fDOI: 10.1063/1.1906183g

I. INTRODUCTION

In semiconductor manufacturing, chemical vapor depo-sition sCVDd is one of the essential unit operations used toproduce thin solid films. Conventional CVD reactors consistof showerhead delivering and distributing reactive gasesacross a substrate, and an energy source to heat the substrateso that gas phase precursor chemical species react on thesubstrate surface producing nonvolatile solid films. Becausereactive gas species can be effectively delivered into micro-scale channels and trenches, CVD can produce smooth andconformal films on topographically complex surfaces at highdeposition rates using relatively simple reactor designs.1,2

Most conventional CVD equipment used in industry isdesigned for a specific process and a narrow range of oper-ating conditions targeted to produce a specific product; thesefixed design configurations limit the ability to control filmthickness, composition, and microstructure spatial unifor-mity. Additionally, the few control inputsse.g., pressure, gascomposition, residence time, temperatured and limited real-time metrology capabilities available in most equipment de-signs result in a trade-off relationship between manufactur-ing effectiveness and film quality.

Because of these limitations, significant research efforthas been directed to improving film growth uniformity bysimulation and optimization3–5 and to adding flexibility toCVD reactor operation by modifying reactor and showerheadconfigurations.3,6–9However, the benefits sought in these de-sign studies cannot always be realized; for example, seg-mented showerhead designs created to control gas phase

composition across the wafer surface9 fall short of this goalbecause of the effect convective mass transfer has on distort-ing the relationship between showerhead output and the re-sulting gas chemical composition profile across the wafersurface.10

In an effort to improve reactor flexibility, we introduceda new CVD reactor concept that enables control of filmdeposition characteristics to produce uniform or intentionallynonuniform films across the wafer surface.10 This reactorsystem incorporates several new design features, the first ofwhich is a segmented showerhead that discretizes the spaceabove the wafer surface into individually controlled regionsto maintain a two-dimensional gas concentration pattern overthe wafer. Individually controlled gas feed lines to each seg-ment allow the segment-to-segment management of gascomposition. Second, to minimize interaction between seg-ments in the wafer/showerhead gap region, residual gas ispumped up through each segment to a common exhaust vol-ume. Additionally, a linear motion device attached to thesegment wall assembly can adjust the gap size to control thesegment-to-segment gas concentration gradient across thewafer surface in the gap region. Third, residual gas analysisfor each segment can be performed using the sampling tubeinstalled in each segment.

Numerical simulation studies and a sequence of experi-ments were performed using a prototype reactor to assess thespatial control capabilities of this CVD reactor design.10 Thisfirst prototype reactorsdesignated P1d consisted of a modi-fied commercial Ulvac ERA-1000 CVD cluster tool; in itsoriginal configuration, the CVD tool was equipped with alamp heater and quartz window sieve showerhead. As part ofthe prototype modifications, we replaced the lamp heaterwith a substrate resistance heater and the quartz window

adAuthor to whom correspondence should be addressed; electronic mail:[email protected], Homepage: www.isr.umd.edu/;adomaiti/.

REVIEW OF SCIENTIFIC INSTRUMENTS76, 062217s2005d

0034-6748/2005/76~6!/062217/10/$22.50 © 2005 American Institute of Physics76, 062217-1

Downloaded 24 Jun 2005 to 129.2.19.102. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

http://dx.doi.org/10.1063/1.1906183

shower head with a new hexagonal segmented showerhead.The schematic diagram of the reactor, the diagram of theshowerhead, and the bottom view of the honeycomb show-erhead are shown in Figs. 1 and 2. Preliminary simulationsand experiments with this modified reactor system were usedto quantify segment back diffusion and intersegment diffu-sion contributions to film growth, and to demonstrate inten-tional spatially nonuniform film deposition.

In this paper, the design and construction of a new pro-totype reactor with combinatorial capabilities will be dis-cussed as well as the experimental results demonstrating WCVD by H2 reduction performed with this reactor. In theconclusion section, future research for improving perfor-mance of this reactor system is discussed.

II. COMBINATORIAL CVD

The difficulty of accurately measuring and controllinggas phase composition in the two dimensions across a sub-strate surface has resulted in relatively few studies focusingon combinatorial CVD. Notable exceptions include the com-

binatorial CVD reactor developed by Gladfelter,11–13 a reac-tor design featuring three widely spaced feed tubes throughwhich different single-source precursors were injected overthe wafer to produce compositional spreads in Ti/Hf/Sn ox-ide films. Hydrogenated silicon films of varying microstruc-ture and thickness were produced in the hot wire CVD sys-tem of Wang;14–17 this design employed a fixed mask placednear to the surface of a movable substrate. Typical combina-torial processing with this system consists of a sequence ofdeposition runs using a single substrate, repositioning thesubstrate between runs corresponding to different H2 and/orSiH4 gas feed flow rates.

Because the reactor system described in this paper wasdesigned to have significantly greater sensing and actuationcapabilities relative to conventional CVD reactor designs,the programmable reactor takes its name from its ability torapidly alter reactor geometrical characteristics as well as thereactant gas conditions, all in software. With its fast repro-gramming capabilities, this new CVD reactor design is anideal tool for combinatorial materials development researchbecause of its ability to produce films with properties thatchangessmoothly or sharplyd across the wafer by varying theoperational parameters of the system, which include wafer/segment gap size, gas flow rates, and the gas composition ineach segment.10 For example, in Fig. 3 we demonstrate arapid process prototyping approach to new materials andprocess development. The spatial patterning capabilities ofthe programmable CVD reactor can be used to produce a“library” wafer sFig. 3, middle leftd containing regions of

FIG. 1. sColor onlined Diagram of the original prototype reactor chamber,hexagonal showerhead, heater, and feed and sampling tubes.

FIG. 2. Photograph of the three-segment prototype showerhead with feedand sampling tubes.

FIG. 3. sColor onlined Fast reprogramming for materials and processdiscovery.

062217-2 Choo et al. Rev. Sci. Instrum. 76, 062217 ~2005!


distinct material properties; the reactor system then can bereprogrammed to deposit spatially uniform films of the cho-sen propertysFig. 3, bottom leftd. Carrying out the process-ing concept shown in Fig. 3 requires an accurate simulator toextrapolate the measurements available from the segmentsampling to the wafer surface.

III. THE REACTOR SYSTEM

While preliminary experiments with prototype P1 suc-cessfully demonstrated the feasibility of this new approach tocontrollable CVD, the experiments uncovered limitations inthe prototype system. For example, in the first prototype,only one feed line of each segment was connected to a massflow controller, limiting the range of process recipes thatcould be tested. As a result, this feed gas supply configura-tion did not allow W deposition by H2 reduction and pre-vented a true test of programmability. Also, the first proto-type system could only reach 10−4 torr as a base pressurebecause the original Ulvac system was designed for LPCVD.This relatively high base pressure caused several contamina-tion problems including particle generation, high partialpressures of H2O, CO2, and O2 in gas phase, and unusualdeposition patterns. The reactor system also required a sig-nificant amount of conditioning time to lower the partialpressure of undesirable gas species, such as the previouslymentioned oxides that tend to consume the reactive WF6

during processing. The nonflexible design of the original Ul-vac system also was a serious limitation of the first prototypesystem in terms of lack of ability to add new sensors and gaslines, and to repair chamber components such as segmentedshowerhead assembly and substrate heater. Therefore, to

demonstrate the true performance capability of this spatiallycontrollable CVD system, it was necessary to construct amore flexible prototype CVD system that could be operatedas a cleansUHV capabled systemssee Fig. 4–8d.

The prototype reactor system P2 consists of four majorcomponents:s1d loadlock and reaction chambers,s2d gas dis-tribution box, s3d two sets of pump systems, ands4d a gascomposition sensing system consisting of a mass spectrom-eter and sampling lines. The process equipment layout isshown in Fig. 5.

A. Chambers and pump systems

Two standard stainless-steel 8 in. six-way CF crosses areused for the loadlock and reaction chambers. A detailed me-chanical drawing of the new reactor chamber with the pro-grammable showerhead and heater are shown in Fig. 4. Theloadlock chamber is used to minimize air exposure to thereactor chamber and is pumped with one set of turbo-molecular and mechanical pumps resulting in a loadlockchamber base pressure of 10−8 torr. This UHV condition wasachieved after He leak checking using the mass spectrometeron this system, normally used for residual gas analysissRGAd. The reaction chamber is pumped by a similar ar-rangement resulting in a base pressure lower than 10−8 torr.Because of these low base pressures, the contaminationsources from chamber walls are eliminated. The chamberand pump diagram is shown in Fig. 6, along with the gasdistribution box and mass-spectrometer sampling system.

As shown in Fig. 6, the reaction chamber is isolated bytwo gate valves, minimizing the air exposure when pumps orgauges are repaired. During a CVD process, residual gas ispumped through the top of each segment into the commonexhaust volume and then through valves2d to the four-waycross. This four-way cross allows the gas pumped throughthe turbomolecular pump and valves3d to flow to the me-chanical pump. During chamber cleaning, gas evolving fromchamber walls is pumped through valves1d and the 6 in.four-way cross, then is exhausted to the turbo-molecularpump and backing mechanical pump.

FIG. 4. sColor onlined Detailed mechanical drawing of the new reactorchamber, showerhead, and heater with critical dimensions such as wafer tofeed tube distance, segment length, and heater and baffle positions.

FIG. 5. sColor onlined Overall configuration of the prototype 2 system.

062217-3 CVD reactor for combinatiorial research Rev. Sci. Instrum. 76, 062217 ~2005!


Each chamber has a set of pressure gauges consisting ofan ion gauge measuring chamber pressure from10−8 to10−4 torr and a convectron gauge for the 10−3 torr toatmospheric pressure range. These ion and convectrongauges are easily damaged by corrosive gases such as WF6,therefore, a Baratron diaphragm gauge is used for the pres-sure control sensor in the reaction chamber with throttlevalve s4d.

B. Gas distribution box

To control gas flow rate and composition to each seg-ment, a gas distribution system was developed to providecontrol to nine sets of gas feed lines; three gas lines distrib-ute gas to each segment and each is connected to an Ar, H2,and WF6 gas source. A gas line assembly consists of a massflow controller sMFCd, one manual and two pneumaticvalves, a pressure gauge, and a pressure regulator. A sche-matic diagram of the gas distribution box and gas deliverylines are shown in Fig. 7.

The setpoints of the pneumatic valves and mass flowcontrollers are sent from a Brooks automation system mastercontroller. The one set of pneumatic on/off valves is installedin each gas line near to the reactor, instead of in the gasdistribution box, to minimize exposing gas feed lines to airduring maintenance.

C. Gas composition sensing

A novel approach to gas composition sampling is imple-mented in prototype P2. QMSsquadruple mass spectrom-etryd is one of the most widely used methods forin situreal-time process monitoring. QMS instruments are compactand cost effective, as well as selective and sensitive metrol-ogy tools.18 Also, because the mass spectrometer can directlysense gas composition at specific points within the reactorsystem without disturbance to process performance, it is acandidate sensor for real-time control, such as in processend-point detection. This capability has been demonstratedfor end-point detection and run-to-run control.19,20

The sampling tube of each segment is used to transport asmall amount of gas to a QMSsan Inficon Transpector CIS2mass spectrometerd. From the residual gas analysis of eachsegment, approximate film thickness and composition offilms deposited on the substrate area corresponding to eachsegment can be determined. Coupled with a reactor segmentmodel,10 this capability of individual segment sampling andmonitoring enables fast reprogramming of the recipe acrossthe wafer surface using model-based, real-time control.

To monitor gas composition in each of the three seg-ments pseudo-simultaneously during operation, the singlemass spectrometer must have time-sharing capabilities.Three sampling tubes and on/off valves connect the segments

FIG. 6. sColor onlined The diagram of the prototype 2 systemsgas flow control system, pumps, chambers, gas distribution box, mass-spectrometer, andcontrollersd.



to the mass spectrometer. The three on/off valves are sequen-tially opened and closed by a control signal sent from a Tech-ware Brooks control platform to perform the mass spectrom-eter multiplexing. The schematic diagram of this setup isshown in Fig. 8.

IV. WAFER TEMPERATURE DISTRIBUTION

Wafer temperature is the other key factor affecting filmgrowth uniformity in most CVD processes. To reveal theheat transfer mechanisms primarily responsible for affectingwafer temperature and its uniformity, temperature measure-ment experiments were performed with an instrumented wa-fer over a range of operating conditions and reactor geom-etries. With a thermocouple instrumented wafer,temperatures of 13 specific points were measured. In Fig. 9,the positions of 13 thermocouples and their relationship to

the segment positions are shown. To assess gas compositionand wafer/segment gap size on wafer temperature, the wafertemperatures were measured when the chamber was filledwith typical reactant gasessAr, N2, H2d at 1 torr and at highvacuum for a range of gap sizes. The role of temperaturedistribution in interpreting deposition results will be dis-cussed later in this paper.

V. W DEPOSITION EXPERIMENTS: H2 REDUCTION

The programmable CVD reactor is designed to produceintentional spatial patterns of W films on the wafer surface aswell as uniform W films. In the new prototype reactor P2, H2

reduction is performed to test this concept of spatial control-lability. Intentional nonuniform deposition ability is impor-tant for demonstrating true programmability of depositionand for combinatorial materials discovery research.

FIG. 7. Schematic diagram of the gasdistribution box and its components.

FIG. 8. Schematic diagram of the time-sharing RGA.



A. Experimental conditions

The recipes used for the experiments are shown inTable I. The recipes EXP1 and EXP2 are designed for

segment-to-segment uniform and intentional nonuniform Wfilm deposition, respectively. In each recipe, the ratio of H2

and WF6 in each segment is fixed at the 4:1 ratio used for atypical blanket W deposition because the ratio of the reactantgases affects the growth rate of W film as well as pressure,temperature, and contamination on the wafer surface.21–23Tominimize convective interaction between segments, the totalflow rate to each segment was held constant set at 60 sccmfor both EXP1 and EXP2.

The heater temperature was set to 400 °C and total pres-

TABLE I. Experimental feed gas recipes for uniform and nonuniform Wdeposition. Gap51 mm.

Feed gassegment

EXP1 EXP2

SEG1 SEG2 SEG3 SEG1 SEG2 SEG3

Ar 30 30 30 60 30 0H2 24 24 24 0 24 48

WF6 6 6 6 0 6 12

Total fsccmg 60 60 60 60 60 60

FIG. 9. Wafer thermocouple positions relative to the segment walls.

FIG. 10. Contour map of tungsten film thickness depos-ited by H2 reductionstopd; three-dimensional thicknessprofile of tungsten filmsbottomd for EXP1.



sure of reactor was set to 1 torr, a typical LPCVDslow pres-sure chemical vapor depositiond condition for W deposition.According to the temperature measurement experiments, theactual wafer temperature is approximately 355 °C at the400 °C setpoint of heater temperature.

When W deposition by H2 reduction is performed on aSi wafer without an adhesion or barrier layer, WF6 reactsfirst with Si rather than H2. This Si reduction is self-limitedbecause Si supply from the substrate to the film surface incontact with the reactant gas is limited by the W film grow-ing by Si reduction. This Si reduction is not desirable formicro-electrical device fabrication because of the well-known encroachment and W migration phenomena leadingto device damage.1,2,24 However, in this experiment, seedlayer formation by Si reduction is not avoidable and is nec-essary to grow W film by H2 reduction. Considering the seedlayer formation, deposition time should be sufficiently longto allow neglecting the W deposited by Si reduction. By theQMS RGA of byproduct gas SiF4 from the Si reduction, thedeposition time was chosen to be 10 min, during which Sireduction occurs for the first 10–20 s, producing a several-nanometer-thick W film.

Proper cleaning of the wafer surface is necessary be-cause the W film growth is significantly affected by the sur-face condition of the Si wafer. For example, H2 reductioncannot produce W films on native oxidesSiO2d, a film thatusually exists on the surface of a Si wafer. Also, other or-ganic and nonorganic particles and molecules cause catalyticreactions consuming WF6. To minimize these effects, each Siwafer s100d is cleaned with acetone and rinsed with deion-ized water to remove organic contaminants. Then, the wafer

is dipped into a 10% HF solution for 5 min to remove nativeoxide and metal contamination completely.

B. Uniform deposition experiments

To establish a baseline to which intentional nonunifor-mity control will be compared, recipe EXP1 was fed to thesegments to measure the segment-to-segment uniformity. Us-ing the experimental conditions described in the previoussection, W deposition by H2 reduction was performed. Theresistance of the W film produced in this experiment wasmeasured to generate a two-dimensional thickness contourmap using an automatic high resolution four-point probe thatwas developed for the analysis of films produced by the pro-grammable reactor.

The resistance measurements are converted to film thick-ness based on the assumption that the resistivity of W film isthe same as bulk W resistivity—we note that this assumptionis not strictly valid because W films produced by CVD have1.5 to 2 times greater resistivity than bulk W. However, thismeasurement is still acceptable for comparing depositionperformance of each segment and to evaluate uniform depo-

TABLE II. Average and standard deviation of W film thickness determinedby four-point probe resistance measurements.

SEG1 SEG2 SEG3 Overall

No. of measured points 544 550 550 1644Average thicknesssnmd 154.5 162.6 158.8 158.6Standard deviationsnmd 21.97 17.74 19.64 20.12

FIG. 11. Average thickness of W film produced by the programmable CVDreactor P2 with recipe EXP1.

FIG. 12. Comparison between wafer temperature profile at 5 mm gap,400 °C, and 1 torr stagnant H2 stopd and deposition profilesbottomdalong the line segments defined by thermocouple positions for the EXP1conditions.



sition ability because the conditions of each segment are suf-ficiently similar to produce W films having a similar range ofresistivity.

The map of thickness measured by the four-point probeis shown in Fig. 10. According to the measurements, sharphexagonal W deposition patterns reflecting the segment loca-tions were observed on the wafer surface, demonstratingclearly the ability of the segments to maintain gas concentra-tions that differ from that outside the segments.

To assess the uniformity of W films deposited by theprototype P2 reactor, each segment region is measured indi-vidually. In Table II and Fig. 11, the average thickness andstandard deviation of W films based on the four-point probemeasurements are summarized. According to these measure-ments, the average of W film thickness in each segment isapproximately 159 nm. The standard deviation of the overallmeasurements is approximately 20 nm and this is close tothe individual segment’s standard deviation of W film thick-ness. Uniformity of segment-to-segment thickness can berepresented by the average and standard deviation of thesethickness values, which are 158.65 and 4.06 nm, respec-tively. Based on this analysis, prototype P2 is able to produce

W films, which have 2.6% segment-to-segment and 12.6%intrasegment uniformity. Because the performance of eachsegment is similar, the intrasegment uniformity remains thekey index to be improved.

In Fig. 12, the temperature profilestopd produced using a5 mm gap and the thickness profile of W filmssbottomd areshown. The temperature profiles in Fig. 12 are measured atthe points 1–5 each segment as shown in Fig. 9. The W filmthickness profiles in Fig. 12 are interpolated values along theline segments intersecting the five temperature measurementpoints in each segment region. The temperature profilesshown in this figure are measured at 1 torr H2 without gasflow and conditions otherwise similar to that of the deposi-tion experiment. Comparing the temperature and thicknessprofiles, we observed similar patterns. One can immediatelyinfer that intrasegment uniformity depends on the tempera-ture distribution across the wafer as well as the gas compo-sition at the wafer surface.

C. Nonuniform deposition experiment

To test the concept of programmability of the prototypeP2, intentional segment-to-segment nonuniformity producing

FIG. 13. Contour map of tungsten film thickness depos-ited by H2 reductionstopd; three-dimensional thicknessprofile of tungsten filmsbottomd for EXP2.



deposition experiments were performed using the recipeEXP2 in Table I; the resulting film is shown in Fig. 13. Asshown in this figure, W film grown in segment 1 is thincompared to the other two segment regions but is thickerthan that found outside of the segment region on the wafersurface. The reactive gases, H2 and WF6, make their wayinto segment 1 by back diffusion from the common exhaustvolume and by diffusion through the wafer/segment gap be-tween this and segments 2 and 3. The thickness differencebetween the W film in segment 1 and the outside region canbe used as evidence to conclude that the contribution of theprecursor gas species by back diffusion from the commonexhaust volume is significant.

In Fig. 14, we illustrate the gradientsdenoted by arrowsdof film thickness in segment region 1. The gradient is a par-tial indicator of the role diffusion has in transporting reactantgases across the wafer surface. From the directions of thegradient, it can be concluded that more reactive gases insegment 1 diffuse from segment 3 than segment 2 becausesegment 3 has a higher concentration of reactive gases, H2

and WF6, compared to segment 2. It is expected that byvarying the wafer/segment gap size and concentration of re-active gases in segments 2 and 3, the diffusion pattern, filmgrowth rate, and thickness gradient can be controlled.

D. W film properties

The electrical and physical properties of W films depos-ited by CVD methods usually depend on the film thickness,growth rate, and species of gas used in the process.21,25 Par-ticularly, the resistivity of W film deposited by CVD usuallyis higher than that of bulk W because of impuritiesssuch asthe remaining F atoms from the WF6 precursord and smallgrain size. The faster grown W film has larger grains andfewer F atoms incorporated in the films. Commonly, thinnerW films have smaller grain size leading to the general obser-vation that faster grown thick W films have lower resistivitythan slow grown thin W films.1,21,25Under EXP2 experimen-tal conditions, the W films have significantly different thick-nesses, growth rates, and processing recipes in each segment;therefore, the W film thickness measured by a four-pointprobe must take into account the spatially varying resistivityto assess segment-to-segment and intrasegment uniformity.

To compare W film thickness measured by the four-pointprobe with true thickness measurements, a SEMsscanningelectron microscoped was used to take cross-sectional pic-tures of the W film and Si substrate, sampled from the centerof each segment region. The W film thickness measured bySEM and the four-point probe resistance measurements aresummarized in Table III. The true W film resistivity of eachsegment is calculated from a comparison of the four-pointprobe and SEM measurements and is shown in the thirdcolumn of Table III. According to the calculated results, Wfilms in segment 2 and 3 regions exhibit a lower resistivityrelative to that of segment 1. This corroborates with the pre-viously discussed general observations on film resistivity, inthat the faster growing W films in segments 2 and 3 havelarger grains, and, as shown in Fig. 15, the grain size of Wfilm in segment 1 is much smaller compared to the two othersegments.

By controlling gas phase composition in this depositionprocess, the potential of this system for performing combi-natorial CVD investigations was demonstrated by the pro-duction of thin films having different electrical and physicalproperties on a single wafer in a single experiment. In thisexperiment, where an inert gas was fed to one of the threeshowerhead segments, a gradient of film thickness inside thissegment region resulted from precursor diffusion from adja-cent segments. While the thickness gradient was limited tothe one segment region in this experiment because of thesmall gap size used, experiments are being planned wherelarger gap sizes will be used to diffuse this gradient acrossmultiple segment regions. It is from combinatorial experi-

TABLE III. Thickness and resistivity measured by four-point probe andscanning electron microscopy.

SegmentFour-pointprobesnmd SEM snmd Resistivity smV cmd

SEG1 56.2 270 26.88SEG2 146.1 420 16.10SEG3 180.7 580 17.97

Bulk W ¯ ¯ 5.6

FIG. 14. Cross section of W films by SEM: segment 1stopd, segment 2smiddled, and segment 3sbottomd.



ments such as these that a desirable film quantity will beidentified; the process conditions at this spatial point on thewafer will be determined from the reactor simulator, and thereactor system then will be reprogrammed to deposit thisdesired film uniformly over the wafer surface.

ACKNOWLEDGMENTS

The authors acknowledge the support of the NationalScience Foundation through Grant No. CTS-0085632 forseed support and Grant Nos. CTS-0219200 and DMR-0231291 for construction of the prototype reactor described

in this paper, and the National Institute of Standards andTechnology for fabricating several showerhead components.The assistance of Jing Chen in several aspects of simulationwork related to this project also is acknowledged.

1C. Y. Chang and S. M. Sze,ULSI TechnologysMcGraw-Hill, New York1996d, Chap. 8.

2L. Xia, P. W. Lee, M. Chang, I. Latchford, P. K. Narwankar, and R.Urdahl, inHandbook of Semiconductor Manufacturing Technology, editedby Y. Nishi and R. DoeringsMarcel Dekker, New York, 2000d, Chap. 11.

3C. A. Wang, S. H. Gorves, S. C. Palmateer, D. W. Weyburne, and R. A.Brown, J. Cryst. Growth77, 136 s1986d.

4H. K. Moffat and K. F. Jensen, J. Electrochem. Soc.135, 459 s1988d.5C. R. Kleijn, Th. H. van der Meer, and C. J. Hoogendoorn, J. Electrochem.Soc. 136, 3423s1989d.

6P. N. Gadgil, J. Cryst. Growth134, 302 s1993d.7A. Kobayashi, A. Sekiguchi, K. Ikeda, O. Okada, N. Hosokawa, Y.Tsuchiya, and K. Ueno, inAdvanced Metalization and Interconnect Sys-tems for ULSI Applications, edited by R. Havemann, J. Schmitz, H. Ko-miyama, and K. TsubouchisMater. Res. Soc., Pittsburgh, PA, 1997d, pp.177–183.

8B. N. Kim and H. H. Lee, J. Electrochem. Soc.144, 1765s1997d.9M. M. Moslehi, C. J. Davis, and R. T. Matthews, United States Patent5,453,124s1995d.

10J. O. Choo, R. A. Adomaitis, G. W. Rubloff, L. Henn-Lecordier, and Y.Liu, AIChE J. 51, 572 s2005d.

11R. C. Smith, N. Hoilien, J. Roberts, S. A. Campbell, and W. L. Gladfelter,Appl. Chem. Mater.14, 474 s2002d.

12R. C. Smith, N. Hoilien, J. Chien, S. A. Campbell, J. T. Roberts, and W. L.Gladfelter, Chem. Mater.15, 292 s2003d.

13B. Xia, R. C. Smith, T. L. Moersch, and W. L. Gladfelter, Appl. Surf. Sci.223, 14 s2004d.

14Q. Wang, G. Yue, J. Li, and D. Han, Solid State Commun.113, 175s2000d.

15Q. Wang, J. Perkins, H. M. Branz, J. Duncan, and D. Ginley, Appl. Surf.Sci. 189, 271 s2002d.

16Q. Wang, Thin Solid Films430, 78 s2003d.17Q. Wang, F. Z. Liu, and D. X. Han, Macromol. Rapid Commun.25, 326

s2004d.18R. K. Waits, J. Vac. Sci. Technol. A17, 1469s1999d.19T. Gougousi, R. Sreenivasan, Y. Xu, L. Henn-Lecordier, J. N. Kidder, Jr.,

G. W. Rubloff, and E. Zafiriou, Characterization and Metrology for ULSITechnology: 2000 International Conference, Gaithersburg, MD, 26–29June 2000, AIP Conference ProceedingssAIP, Melville, NY, 2001d, Vol.550, pp. 249–253.

20Y. Xu, T. Gougousi, L. Henn-Lecordier, Y. Liu, S. Cho, and G. W.Rubloff, J. Vac. Sci. Technol. B20, 2351s2002d.

21S. Sivaram, M. L. A. Dass, C. S. Wei, B. Tracy, and R. Shukla, J. Vac. Sci.Technol. A 11, 87 s1993d.

22C. R. Kleijn, C. J. Hoogendoorn, A. Hasper, J. Holleman, and J. Middel-hoek, J. Electrochem. Soc.138, 509 s1991d.

23A. Hasper, J. Holleman, J. Middelhoek, C. R. Kleijn, and C. J. Hoogen-doorn, J. Electrochem. Soc.138, 1728s1991d.

24E. K. Broadbent and W. T. Stacy, Solid State Technol. December 51s1985d.

25C. M. McConica and K. Krishnamani, J. Electrochem. Soc.133, 2542s1986d.

FIG. 15. Surface of W films by SEM: segment 1stopd, segment 2smiddled,and segment 3sbottomd.



Documents

Development of a spatially controllable chemical vapor ... of a spatially controllable chemical vapor deposition reactor with combinatorial processing capabilities J. O. Choo and R