Arthur Sherman - Chemical Vapor Deposition for Microelectronics (779637), страница 19
Текст из файла (страница 19)
1o These experiments were carried out in aquartz cold-wall system. Best resu Its were fou nd at a deposition temperature of150°C, which is much lower than needed by the chloride process. Typicaldeposition conditions were 250 mTorr pressure, 2 to 6 secm of MoF6 and 100sccm of Si H 4 .
All films were annealed at 1100°C for 30 seconds. They werefound to be very stable and adherent on both silicon and silicon dioxide. Aminimum resistivity of '"'"'100 prl-cm was found for deposition temperaturesbelow 150°C. Compared to the chloride process, the major advantage is the lowtemperature at which these films can be deposited.Finally, the conformality of the MoSi 2 films are good.4.2.3 Tantalum SilicideThe WSi 2 films described earl ier were deposited in a cold-wall reactor. TheMoSi 2 films just described were deposited in a hot-wall reactor. Does it makeany difference whether these films are deposited in one or the other type of reactor? In studying the deposition of TaSi 2 , we have an opportunity to examinethis question, as we have two studies to consider; one was done in a hot tube 11and the other was done in a cold-wall system. 12 ,13 Both study the reaction between TaCl s and SiH 4 at comparable temperatures and pressures.
The desiredtantalum silicide is TaSi 2 , which has a low electrical resistivity and is thermodynamically stable.The hot-wall study l1 introduced TaCl s with an evaporator operating inthe temperature range of 120° to 140°C with a small H 2 flow (about 5 secm)as the carrier gas. A SiH 4 flow of 24 sccm is used at a pressure of 280 mTorr.Deposition rates of 120 Aim in were achieved at temperatures of 615° to 635°C,with uniformity of ± 10%. Actually, a polysilicon layer is deposited first, sothat a polycide structure could be studied.Unfortunately, the tantalum silicide deposited was very metal rich, andclose to Ta sSi 3 . Although attempts to vary the stoichiometry by changing theThermal CVD of Metallic Conductors101process conditions were made, they were unable to report the deposition ofTaSi 2 films.Instead, the approach taken was to anneal the polycide films at temperaotures over 800 e.
In this case, Si diffused up to the Ta s Si 3 film in sufficientquantity to convert it to TaSi 2. An anneal at 1000 e for 15 minutes in argonproduced a resistivity of 48 JJS1-cm in 2500 A thick TaSi 2 layers.Although these results were encouraging, examination of the films by TEMshowed that almost all of the underlying poly was consumed in creating TaSi 2.Also the Si was not extracted uniformly from the poly layer, so that the surfaceof the TaSi 2 was quite rough.The second study was done in a cold-wall reactor12~13 using the same reactants.
The reactor was a single-wafer system, similar to the tube reactor ofFigure 18 in Chapter 2, with the wafer heated by an electrical resistance heaterin the pedestal. In this case, the sublimator was operated at 88°C with a 10sccm flow of H2 . The influence of SiH 4 flow rate on the film stoichiometry andresistivity (after anneal) are shown in Figure 11.Films of TaSi 2 deposited in this process, after anneal, were specular (surface rough ness of onl y 1000 A gra in structure) a nd had good step coverage(thickness on vertical wall equals 65% of thickness on horizontal surface).As long as the films were not Si rich, the resistivity was in the range ofo""75 JJS1-cm. When substrate temperature was varied between 650 e and 750 e,the deposition rates were unchanged. This implies that the reaction is proceeding by a diffusion-controlled mechanisnl. The resistivity of the better filmsafter a 1-hour, 900 e anneal in argon was ""60 JJS1-cm, independent of thedeposition temperature.000200035~Qcm1I10]'SlIo30I.....fillc:500.~u~~:~25.~cu-020010 2.~.60•40102040.
20•60seem8015SiH. flow rateFigure 11: Stoichiometry and resistivity of TaSi 2 films deposited by CVD. 12E102Chemical Vapor Deposition for MicroelectronicsAn interesting effect was observed by varying the pressure between 80and 400 mTorr. At pressures of 180 mTorr and above, the deposition ratejumped from 600 to 2000 A/min. At the same time, resistivity rose as high as4000 ~n-cm. The variation with pressure is shown in Figure 12.
Apparently,the stoichiometry changed dramatically at pressures of 180 mTorr and higher.In fact, there is very little Ta in the film created at 400 mTorr (about 13%),so this is mostly a polysilicon film.35Ia/o030200nmmin150251.,c;tI)~E100c:.~20.~00..tI)~60x1540200 r--...,...---,......-----,..----r----.....----....,...---+1070100150200pressure250.300~ar400Figure 12: Resistivity and stoichiometry of TaSi x CVD films as a function ofpressure.
12Finally, we can comment on the influence of the reactor type on the filmsthat can be deposited. Evidently, the hot-wall reactor tends to deposit veryTa-rich films. Although it may be possible to alter the stoichiometry in thistype of reactor, the choices are limited. One must operate under conditionswhere uniform depositions are achieved both on each wafer and from wafer towafer, because this is a batch system.
In the cold-wall reactor, it was possibleto obtain the proper stoichiometry at high deposition rates. Since the higherdeposition rates perm it development of a si ngle-wafer reactor, there are morechoices in the process conditions to be used.It is probable that a fundamental difference exists in processes operatingin the two types of reactors considered here. In the hot-wall system, the reactant gases have ample time to react before reaching the wafers, so gas phasechemistry probably plays a role.
In the cold-wall system, this is probably minimized.Thermal CVD of Metallic Conductors1034.2.4 Titanium SilicideThe lowest resistivity silicide film of the four we are considering is theTiSi 2 film, so such films have always been of interest. A recent stud y 14 hasshown that these films can also be deposited by low-pressure CVD.
For these experiments, a cold-wall reactor similar to the parallel-flow tube reactor sketchedin Figure 17 of Chapter 1 was used. The wafer was heated by heating the susceptor from below by optical radiation.Depositions were carried out with Si H4 and TiCI 4 reactants. The TiCI 4 ,which is a liquid at room temperature, is evaporated in a sublimator at 28°C.The structure grown was the polycide structure, as before. Films were depositedat 650 to 750°C and pressures from 50 to 460 mTorr at several flow rates(TiCI 4 /SiH 4 ).
Stoichiometric films that were slightly Si rich (TiSi 2 ) wereachieved with as-deposited resistivity of 22 J,lS1-cm reported. Also, surface roughness was small (about 50 A).In summary then, good quality TiSi 2 films were produced with low asdeposited resistivities. The only concern, as far as using such films is concerned,is the fact that TiSi 2 etches readily during wet HF etch procedures. Such etchprocedures are an integral part of many of the integrated circuit process steps,and one must be concerned about the integrity of the TiSi 2 films.
If the ICmanufacturer is willing to use all dry etch procedures (plasma etching), thisconcern can be alleviated.04.3 TUNGSTENAs mentioned earlier, there is a considerable need for a conformal metalliccoating with resistivity close to that of aluminum, but with a higher meltingpoint. Of the ones we have been considering, the two lowest resistivity candidates are molybdenum and tungsten. Tungsten has received the most attentionsince the H2 reduction of WF 6 process has been under development for a varietyof appl ications since 1967.
15 The extent of current interest can be seen in arecent publication. 16There are two aspects of tungsten CVD for integrated circuits that havetaken on commercial importance. One is the blanket deposition and subsequentpatterning, so it can be used as a conductor to replace high-resistivity dopedpoly.
The second area of interest is the "selective" CVD of tungsten, wheredeposition occurs on silicon but not on silicon dioxide. Here one can selectivelyfill via holes to either provide a thin barrier metal or to deposit a thicker layer tohelp planarize the circuit. Both applications involve only one processing step,and are attractive for this reason.We will review recent work in the blanket tungsten process first.4.3.1 Blanket TungstenTungsten can be deposited by CVD by a number of different processes.Several that have received considerable study are:(2)WF6 + 3H 2~W(s) + 6HF104Chemical Vapor Deposition for Microelectronics(3)2WF6 + 3Si(s)(4)WCI 6 + 3H 2(5)2WF6 + 3SiH 4~~2W(s) + 3SiF4W(s) + 6HCI~2W(s) + 3SiF4 + 6H 2The first is the hydrogen reduction process which can proceed on any surfaceraised to a suitable telTlperature.
The second is the silicon reduction processwhere silicon reduces WF 6 . The third process is similar to the first, but substitutes chlorine for fluorine. The final process is related to the WSi 2 depositionstudied earlier. It has been shown!? that depending on the deposition conditions, one can deposit either W, WSi 2 or Ws Si 3 from these two reactants.The second reaction, Equation (3), is the basis for the selective tungstenprocess we will discuss later. It also plays some role in the blanket process.The first study of CVD tungsten for application to integrated circuits wasdone by Shaw and Am ick, 18 working with the hexafluoride. They carried outtheir depositions in an atmospheric-pressure horizontal cold-wall tube reactor(see Figure 17, Chapter 1), where the susceptor that held the wafers was inductively heated.The major problem, then and now, in attempting to deposit blanket tungsten is the adhesion to silicon dioxide.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
