Arthur Sherman - Chemical Vapor Deposition for Microelectronics (779637), страница 7
Текст из файла (страница 7)
Allof which says that in spite of our best efforts, cold walled reactors may havetheir cold walls an undesirable source of particulates which may end up on thehot substrate. The occurence of such particulates can be minimized by frequentcleaning of the chamber walls to remove deposits.1.5.1.1 Tube Reactor, Parallel Flow: A sketch of this reactor configuration is shown in Figure 17. Here the reactant gas or gas mixture flows axiallydown a tube (circular or rectangular cross section) over a heated susceptor32Chemical Vapor Deposition for Microelectronicsaligned parallel to the flow.
The wafer to be coated is placed on this susceptor.The tube is most often fused quartz. Techniques for heating the susceptor andnot the tube walls will be discussed later. Operation can be at atmospheric orat low pressures depending on the process being developed.
Low pressure operation will, of course, require vacuum pumping capability.TUBE WALLLUNREACTED-----I~GASES•a:c::c:::::::::PARTIALLYREACTEDEXHAUST...':::::::TUBE WALLFigure 17: Tube reactor, parallel flow.1.5.1.2 Tube Reactor, Normal Flow: An alternative arrangement for sucha reactor would be to align it vertically, and place the wafer on a pedestal normalto the flow direction.
Such a system is shown in Figure 18.REACTANT GASES1 1 1TUBE~1WAFERPEDESTAL1PARTIALLYREACTEDGASESFigure 18: Tube reactor, normal flow.Fundamentals of Thermal CVD33The difference in these two configurations is in the flow pattern developedover the wafer. In Figure 17 the flow will develop as a classical boundary layerwhich will be thinner at the front and thicker towards the trailing edge. Towhatever extent the deposition is diffusion controlled, it will be influenced byboundary layer thickness, so deposition rates may be higher near the front ofthe wafer than in the rear.
For Figure 18 the flow pattern will be quite different and whatever nonuniformities occur will at least be axisymmetric.1.5.1.3 Heating Systems: Over the years, three methods of heating thesusceptor but not the tube walls have been used. For one, the susceptor canbe made of an electrically conducting material (i.e., graphite) and used as aresistor with an electrical power supply. Joule heating will then heat it readily.Alternately, one could heat the conducting susceptor inductively with aconducting coil placed around the tube, as shown in Figure 19. The coil wouldbe activated by an AC power supply.8ABBeBBB~sssssssssssss,f}f}f}f}f}f}f}!nFigure 19: Inductive heating of tube reactors.Finally, heating can be done by irradiating the susceptor with high intensitylamps that will transmit readily through the fused quartz.Direct heating of a resistive substrate requires high temperature electricalconnections within the reactor proper. For this reason, inductive heating seemsto be preferred.
Optical heating has one advantage over either of the first twoschemes. When the susceptor is heated electrically, the wafer sits on a heatedsurface and its bottom surface can be somewhat hotter than its top surface.For certain high temperature processes, this small difference can be enough tomake the wafer warp, which can cause disturbances in the flow or in extremecases damage (i.e., slip) to the wafer. With optical heating, one can heat thewafer from above while it is inductively heated from below, or from both aboveand below, so that the wafer remains in a uniform temperature environment.1.5.2 Cold Wall Systems-Multiple WafersObviously, it would be desirable to process more than one wafer at a timein a CVD reactor, since wafer throughput (i.e., wafers/hour) could be an important factor in determining how commercial a process will be.
Multiple waferreactors can all trace their roots to some variant of the two simple systemsjust described.34Chemical Vapor Deposition for Microelectronics1.5.2.1 Tube Reactor: When initial attempts were made to extend the susceptor of Figure 17 in the flow direction, it was discovered that deposits on thelast wafers were thinner than deposits on the initial ones. As discussed earlier,the explanation proposedl4 was that deposition was diffusion controlled sothat the thicker downstream boundary layer limited the diffusion of reactantspecies to the downstream wafer. An obvious solution is to tilt the susceptorup towards the rear (see Figure 8). This would have the effect of thinning thedownstream boundary layer which should correct this problem.
Experimentall4data have verified this effect as shown in Figure 20.0·50·40·30·2--....0·1·s~ 0·5E~0·4(.!)0·3.eeVo=34cm/slPoI',~0tcm2,..0·2-c-wto=639 dyne0·1IJI, ,ItIItII0·50·40-30-20-1 '--.L-~L--I"""--l.~--'-"--""--"'---'----'---'--"'---t-o~~~......~Fbsition x along the susceptor (em)Figure 20: Deposition along length of tilted susceptor. 14 Reprinted by permission of the publisher, The Electrochemical Society, Inc.1.5.2.2 Bell Jar Reactor, Barrel Susceptor: If we wanted to place evenmore wafers in a reactor, one can go to a bell jar configuration and extend thetilted susceptor tube concept just described. Consider a reactor within a belljar arranged as shown in Figure 21.Fundamentals of Thermal CVD35BELL JAR-;\Figure 21: Bell jar reactor, barrel susceptor, axial flow.In this configuration, reactant gases enter at the top of the bell jar and flowbetween the bell jar and susceptor before exiting at the bottom.
The susceptoris multifaceted, the number of facets being determined by wafer size and belljar diameter. It is tilted to enable wafers to sit in pockets and be held in placeby gravity. If we look at one sector of this reactor (delineated by the dottedlines) we see that we have simply replicated a tube reactor with a tilted susceptor several times around the bell jar periphery. The narrowing of the flowpassage towards the exit again serves to thin the exiting boundary layer. Theentire susceptor can be rotated to minimize any tangential nonuniformities.With this reactor design, the susceptor can be heated inductively by wrapping a coil around the bell jar, or radiantly from outside the bell jar.1.5.2.3 Bell Jar Reactor, Barrel Susceptor, Radial Flow: In much thesame way that we extended the tube reactor with parallel flow to a bell jargeometry, we can do the same with the tube reactor with normal flow. Consider Figure 22. Since the flow is entering radially from the outside, one wayto heat the susceptor is with high powered lamps within the central cavity.Again, the susceptor could be rotated to improve uniformity.WAFERS~oo--=INLET FLOWII--\ (------"Figure 22: Bell jar reactor, barrel susceptor, radial flow.36Chemical Vapor Deposition for Microelectronics1.5.2.4 Pancake Reactor: The tube reactor with normal flow leads toanother type of bell jar reactor referred to as the pancake reactor.
It is shownin Figure 23. Here the susceptor is a disc placed horizontally and heated byinduction by coils placed below it. The reactive gas flow could be introducedfrom above, but the favored approach is to introduce it from below at thecenter of the susceptor disc. Gas exhaust is at the periphery between the discand the bell jar.BELL JAR\(!~\!WAFERSINDUCTION COILSFigure 23: Pancake reactor.1.5.3 Cold Wall Systems-Continuous BeltThese systems are similar to the tube reactor with normal flow exceptthat wafers are passed horizontally under the reactant gas flow while beingheated from below.
One version uses a premixed gas stream and flow distributor long enough to deposit on many wafers simultaneously. A sketch of sucha system is shown in Figure 24. Wafers are heated from below by resistanceheaters radiating to the wafer carriers. Wafers enter on the left and leave onthe right after passing through nitrogen purge curtains which protect the operators from any toxic gases used as reactants.GAS DISTRIBUTORWAFERINPUT+ + + + , t + + + J.....'--.'......i-'-+ i t~-.,._~~c=::::Jc=::::J~6 -_ _-~--t.~MOVING BELTGAS EXHAUSTWAFERSFigure 24: Continuous CVD reactor-premixed gas flow.WAFEROUTPUTFundamentals of Thermal CVD37The primary advantage of these continuous systems is the high waferthroughput.
Also they are run at atmospheric pressure so the expense of avacuum pumping system can be avoided. The disadvantage is that any depositson the cooled distributor plate end up as particles on the wafers.The second approach is to use separate reactor streams that mix veryclose to the moving wafers. In Figure 25 we show a typical arrangement.EXHAUST\INPUTGASESfEXHAUSTWAFERS\HEATERMOVING BELTFigure 25: Continuous CVD reactor-separate gas flows.1.5.4 Hot Wall SystemsOne of the problems with cold wall systems is the difficulty in maintaining a very uniform temperature on the wafers.
Such concern can be eliminatedif the entire reactor chamber is placed within a furnace maintained at a very uniform temperature. An ideal candidate for such a furnace is the standard diffusion tube furnace already in wide use for integrated circuit fabrication. Ifin addition, wafers could be loaded vertically as in a diffusion ·furnace, thereactor throughput could be substantial.The problem of assuring uniform depositions on many wafers closely spacedin a long uniform tube was solved when operation of the reactor at low pressure was considered. 22 Normally, in an atmospheric pressure cold wall CVDsystem, the reactant gas is heavily diluted in N2 in order to reduce gas phasenucleation. At the pressures used for low pressure CVD (0.5-1.0 Torr), this isless of a problem so less diluent is needed.
The net effect then is that deposition rates only fall by a factor of five. However, as many as 100 wafers can beprocessed in such a reactor at one time (see Figure 26), and this more thancompensates for the lower deposition rate. In addition, due to the low pressure,diffusion occurs at high rates and the deposition tends to be controlled by thesurface temperature. Given the very uniform temperatures available in a diffusion furnace, the deposition uniformity tends to be excellent in such a system.38Chemical Vapor Deposition for MicroelectronicsHEATERWAFERS1lH_EA_T_ER___QUARTZ TUBEFigure 26: Low pressure hot wall CVD reactor.REFERENCES1.
Glasstone, S., Textbook of Physical Chemistry, Princeton, NJ: D. VanNostrand (1946).2. Zeleznik, F.J., and Gordon, S., Calculation of complex chemical equilibria.Industrial and Chemical Engineering 60: 27 (1968).3. Ban, V.S., and Gilbert, S.L., Chemical processes in vapor deposition ofsilicon.JECS 122: 1382 (1975).4. "JANAF Thermochemical Tables." Dow Chemical Corp. (1965).5. Kern, W. and Ban, V.S., in Thin Film Processes (J.L. Vossen and W. Kern,eds.) p.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
