Arthur Sherman - Chemical Vapor Deposition for Microelectronics, страница 14

Since sputteredaluminum does not produce a conformal coating, voids and/or thin regions canoccur that will compromise device reliability. Heating the deposited Si0 2 to ahigh enough temperature will cause the glass to soften and flow, thereby rounding off sharp corners. Adding phosphorus and boron to the oxide lowers thetemperature at which such reflow takes place. ,A sketch describing the reflowphenomena is shown in Figure 7.AS - DEPOSITEDAFTER ANNEALINGPLANARIZATIONI~CONTACT WINDOWSFigure 7: Reflow of PSG after heating.sThe top figures refer to the planarization of a deposit over a poly line.On the left is the as-deposited PSG.

On the right, we see how much smootherThermal CVD of Dielectrics and Semiconductors73the profile has become upon heat treatment. The lower figures refer to smoothing the corners of contact windows etched in PSG. On the left, we see thesharp corners produced by the typical anisotropic etch. On the right, thesecorners are substantially smoothed.Experimental results for the reflow of PSG as a function of doping levelfor a 1000°C reflow in a Nz ambient for 30 minutes is shown in Figure 8. Theholes (windows) that had been etched in the l-micron thick PSG had initiallyhad sharp corners and vertical walls.

Clearly, even the first sample with 7.34%doping received an acceptable rounding off.6.<>WT"IPAlFigure 8: PSG reflow for different doping levels. 6 Reprinted by permission ofthe publisher, The Electrochemical Society, Inc.74Chemical Vapor Deposition for MicroelectronicsIncreasing the phosphorus content beyond the 6 to 8% level allows phosphoric acid to form in the presence of moisture.

This acid can attack aluminummetalization, and can be the source of device reliability problems. It can alsocause resist layers to Iift, thereby terminating the processing. Recent research shas shown that adding B20 3 to the PSG by doping with B2H 6 as well as PH 3leads to a borophosphosilicate glass (BPSG) and the desired lowered viscosityat a given tem perature.The BPSG was deposited in a cold-wall CVD reactor operating at atmospheric pressure. Temperatures were 410° to 430°C, and the oxygen/hydrideratio was 20:1.

Evaluating depositions, it was determined that the PH 3 flowhad little if any effect on the B incorporation due to B2H 6 flow. Therefore,the process was implemented by adjusting the 8 2 H 6 flow first, to establish thedesired B doping. Then, the PH 3 could be adjusted to bring the P doping upto the proper level knowing that the 8 doping level would remain uneffected.Reflow experiments established that films prepared with 3.4 wt % Band4.5 wt % P had the same flow behavior, at a given temperature, as a PSG filmwith about 7 wt % P. Therefore, one can either add more phosphorus to furtherlower the flow temperature or maintain it as is and have a more reliable process.3.2.4 Tetraethylorthosilicate (TEOS) SourceIn addition to the SiH 4 or SiCI 2H 2 source gases that are used to depositSi0 2, one can consider TEOS-(C2HsO)4Si for the same purpose.

This materialhas a number of possible advantages. For one, it is a stable inert liquid, so thehazards of dealing withpyrophoricsilane are avoided. Also, it offers the potential of improved step coverage, which we will explore later. It does have thedisadvantage of requiring a higher deposition temperature than the SiH 4 process.As with earl ier work, it would be desirable to dope the films with P and/orB.

This has been tried7 with liquid sources as well as phosphine and diborane.The liquid sources were:Trimethylphosphate TMP-ate (CH30)3POTrimethylphosphite TMP-ite (CH30)3PTrimethylborate TMB(CH30)3BThese have the major advantage of bei ng safe, as contrasted to the very toxicphosphine and diborane.The experimental setup in which TEOS was used for oxide depositions isshown in Figure 9. We see that the reactor is a low-pressure, hot-wall tube, sodeposition uniformity on each wafer and from wafer to wafer will be the primary requirement as far as reactor operating conditions are concerned.Experiments were carried out at 500 mTorr and temperatures from 620°to 680°C. The best results were obtained with phosphine and TMB doping,with added O 2 to increase the deposition rate. The amount of boron and phosphorus incorporated in the film could be varied by changing the TMB flow,as shown in Figure 10.As noted earl ier, a potential advantage of the TEOS process may be improved step coverage. To illustrate this phenomena, BPSG was deposited ina small, deep trench by both the TEOS and the SiH 4 processes.

The resultingSEMs are shown in Figure 11.Thermal CVD of Dielectrics and SemiconductorsWafers in BoatsHeraeus CantileverPressureSensor[f£] =Flow~~~~~~~~~~~~~~~~~~Controller~=~eumaticV~~Figure 9: Schematic of setup for TEOS-BPSG depositions. 7CJPH 3( 15 0/0) / Ar: 150 seem02: 160 seemDeposition Temp.: 665°Cc::CD"'C0en15;:+"0"='::::J:J~Q)c::....~..B ••2:5 6c::Ctl:---........ ........... ..........§-5CJ43..................•ciii••••~-r--~-f...e.2•••c;rt='3"3:;"L....Jp.......-...i.....~.2...75o10-----------1-----&.-7- - .

. . . 1 .- - 15TMB- Flow [Arbitrary Units]9Figure 10: Influence of TMB flow on dopant content of TEOS-BPSG films. 776Chemical Vapor Deposition for MicroelectronicsFigure 11: Step coverage for (a) TEOS-BPSG deposition at 640°C, and (b)SiH 4 process at 400°C. 7The explanation offered for this phenomena is that in the TEOS process,the molecule decomposes slowly after adsorbing on the surface, so there ismore time for it to diffuse along the surface before breaking up and leavingSi0 2 behind. In the silane process, it is argued that the SiH 4 is so reactive thatit decomposes as soon as it hits the surface, and therefore does not have timeto diffuse down to the bottom of the trench before leaving Si0 2 behind.

Also,there may be some temperature effect with molecules in the higher temperature process, (TEOS) being more mobile on the surface.The films produced by these authors 7 exhibited good chemical stabilitywith doping levels of 5 wt % Band 5 wt % P. Since these doping levels arehigher than typically achieved in the SiH 4 process, reflow temperatures required can be lowered somewhat.3.2.5 Diacetoxyditertiarybutoxysilane (DADBS) SourceThe TEOS process just described has potential for use as an interlayerdielectric over metal layers that can withstand the high deposition temperature(i.e., doped polysilicon). If we want to cover aluminum, this process cannotbe used. An alternative is to use a process where diacetoxyditertiarybutoxysilane (DADBS)-(AcOhSi (OtBU)2 will decompose at temperatures between450 and 600°C.

8Again, a low-pressure, hot-wall CVD reactor was used for the depositions.Pressures ranged from 300 to 900 mTorr. Phosphorus doping was carried outwith trimethylphosphite. For these experiments, some depositions were carried out with oxygen addition. In this case, deposition rates were lowered.Combining trimethylphosphite with oxygen, on the other hand, increaseddeposition rates.0Thermal CVD of Dielectrics and Semiconductors77In comparison to the TEOS process, it is claimed that the step coverageis perfect. For example, taking the top surface adjacent to a trench as unitythickness, a trench with aspect ratio of 2: 1 when coated with a 700°C TEOSprocess produces an 85% sidewall thickness and a 70% bottom thickness. It0is claimed that a 450° to 500 e DADBS film is 100% on both sidewall andbottom.3.3 SILICON NITRIDEAgain, Si 3N 4 can be deposited by atmospheric-pressure CVD, but theeconomics are more favorable to the LPCVD approach.

Films from SiH 4 andNH 3 can yield stoichiometric Si 3N 4. More uniform and higher deposition ratefilms are created, however, from SiH 2 CI 2 and NH 3 in a reactor such as theone shown in Figure 4. 3 Under conditions such as 750°C, 500 mTorr, andNH 3/SiH 2 CI 2 == 10:1 flow ratios, excellent thickness uniformities are obtained.Within a wafer, uniformities are ± 3%, and from wafer to wafer they are ± 1.5%.Deposition rates are 37 A/min.A very detailed analysis of the chemical nature of thin (50 to 500 A)LPCVD Si 3N4 films has been carried out.

9 These authors show that Si 3 N4,by the above LPCVD process when deposited on Si, exhibits a 15 to 20% ASi0 2 layer between the Si and Si 3N4. It is felt that this layer originates fromthe native oxide on the Si, even after careful cleaning. Also, they detected0.4% chlorine in the film originating from the SiH 2 CI 2 , although the Si 3N 4was indeed stoichiometric.There did not appear to be any gross evidence of hydrogen in the films,but none of the analytical techniques used by these authors to evaluate thefilm composition were sensitive to this impurity.3.4 POLYSILICONUp to this point, all of the films we have considered (Si0 2 , Si 3N4 ) weredeposited under conditions such that they were amorphous.

The only defectsof interest were particles from the gas phase that might be incorporated intothe growing film, or pinholes. Low-pressure CVD has reduced the incidenceof particles, and thicker films can minimize the presence of pinholes.When we consider silicon films, on the other hand, the nature of the soliddeposit is crucial to the behavior of the film. Depending on deposition conditions, we can deposit amorphous, polycrystalline, or single crystal films.As was noted in Chapter 1, the morphology of polycrystalline films can becomplex. In the present section, we will review some aspects of polysilicon(poly) thin films deposited by CVD. The final section of this chapter will bedevoted to epitaxial silicon thin films.3.4.1 Deposition BehaviorFirst, let us consider some of the factors that arise when we form CVDpoly, and then when we try to dope it in-situ.

Although CVD poly can be de-78Chemical Vapor Deposition for Microelectronicsposited in a cold-wall atmospheric-pressure reactor, economic considerationsare such that today it is only deposited in a low-pressure hot-tube furnace.The basic chemistry used is the pyrolysis of Si H 4 , usually with a diluent gassuch as N 2 , He, H2 , etc. Other chemistries have been tried, but none has yieldedas uniform depositions at the high deposition rates as the Si H4 process.