Roland A. - PVD for microelectronics (779636), страница 60
Текст из файла (страница 60)
9.29b).Reducing the reflectivity at the metal-film interface also reduces the sensitivity of the lithographic process to variations in resist thickness, whichimproves across-wafer uniformity of line width. While sputtered films ofamorphous Si have been similarly used to control lithographic variationsover reflective topography, TIN is preferred due its more robust adhesionto photoresist and the fact that it can be left in place after metal patterningas part of the Al slab interconnect.
The TiN ARC layer also protects theunderlying Al layer from corrosion by the chemically basic photoresist developer, and for this reason a pinhole-free ARC layer is required.9.6 Titanium-Tungsten (Ti-W) Alloys9.6.1 METALLURGICALISSUES FOR PVDTiW is actually a pseudo alloy of Ti in W, with the Ti added to improvethe adhesion of the W to oxide and its oxidation resistance.
While TiW iscommonly sputtered from targets with 10% Ti by weight (equivalent to atarget of Ti,,.,W,,,, atomic composition), the deposited film is depleted inR. POWELL AND S. M. ROSSNAGEL32411111-90 ~i.............................i~!~6094~~.2010:.Wavelengthn m ': 436................ ~ ............... '................ ~................ '!. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .,- .................................................................................~............................... ...................020TIN406080T h i c k n e s s (nm)(a)FIG. 9.29 Reflectivity of an AI-Si alloy with a PVD TiN overlayer as a function of (a) TiN layerthickness for a wavelength of 436 nm, and (b) wavelength for a TiN film thickness of 35 nm (afterref.
9.34). (Reprinted with permission from M. Rocke and M. Schneegans, J. Vac. Sci. & Tech. B6(4):113-115 (1988). Copyright 1988 American Institute of Physics.)Ti such that its Ti content is only ~ 5-7% by weight. Since the films are-~ 95% tungsten by weight, one should really refer to them as "tungstentie" as opposed to the conventional "tie-tungsten." In any event, the resistivity of such TiW films ( ~ 50-80/zf~-cm) is comparable to that of PVDTiN, and the films are similarly applied in IC processing, e.g., as a diffusion barrier between A1 and Si.
Unlike TiN, however, TiW can be sputtered easily in a nonreactive magnetron process but is generally not considered as good a barrier as TiN, so its use in advanced devices isbecoming less common. The step coverage of TiW films over topographycan be quite good, relative to, say, AI-Cu, because the large mass of W(184 amu) minimizes any loss of directionality due to gas-phase scatteringwith Ar (40 AMU).PVD MATERIALS AND PROCESSES90......................~..........!; ..............................i~ .........~. ..............................i TiN T h i c k n e s s325i ..........i .........:.35.nm~ ..........9.o,o iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii i i i i i i i,iiiiiiiiiiiiiiiiiiiii iiiiiiiii! iiiiiiiiiiiiiiiiiiiiiiiiiiii2O'~.......................! .........!..................................................
i....................:..........i................... :0 - ' ; .......... ; ......... ~": ...... i ......... i ......... ~.......... ; ...... ': ...........i.......... i ......... ~"......... ;.................. :~4005OOWavelengthFIG. 9.29600(nm)(b)9.6.2 PVD OF T[xW~_xAs mentioned above, PVD TiW films are Ti-deficient with respect to thetarget composition, and this can affect both their electrical resistivity andbarrier properties. In addition, TiW film composition in topography is affected by differences in the angular distribution of incident Ti and W fluxat the wafer [9.36, 9.37].
While the angular emission of Ti and W from analloy target is similar, the lower-mass Ti has a higher average scatteringangle than W with the Ar sputter gas. Therefore, a wider incident flux distribution is expected for Ti than for W at the wafer, although this differencewill be more pronounced at greater source-to-substrate spacing due to theadditional gas-phase scattering. This is illustrated in Fig.
9.30, where angular distributions are calculated for sputtering at 7 mTorr at source-tosubstrate spacing (sss) of sss = 2 cm (Fig. 9.30a) and 7 cm (Fig. 9.30b)from a commercial 5-cm-diameter magnetron. The two well-defined peakscorrespond to the annular erosion track designed into the sputter source,and the integrated area under each curve is proportional to the total flux ofR. POWELL AND S.
M. ROSSNAGEL326o . 1 0-,, . ,,-i-,-#",-.t..--.-st'ii0.04,,,!.I0.02o,;t0"%o.0o . s o,o0--- - - - . -.3,.~-~t0~o~;.o,o.o,o.0,9ii!9.I.0 . 4 0e9.i9i~.0_t0 . 3 0i0 . 2 00it0 . 1 o0.00-gO.O-60.0..30.00.030.060.090.0FIG. 9.30 Calculated Ti and W incident flux distributions for magnetron sputtering of a 5-cm-diameter TiW target at 7 mTorr and a source-to-substrate spacing of 2 cm (upper) and 7 cm (lower)[9.371 9PVD MATERIALS AND PROCESSES327Ti or W reaching the wafer surface. It is clear from Fig.
9.30 that the stoichiometry of a blanket TiW film deposited at sss = 7 cm will be heavilydepleted in Ti compared to the film deposited at sss = 2 cm. In addition, ithas been shown that preferential resputtering of the Ti by energetic Ar(40 AMU) that is backscattered from the W target atoms can have a dominant effect on further depleting the Ti content of the deposited TiW film[9.38].With regard to Fig.
9.30b, we see that the W flux is much more directional than the Ti flux. Hence, we would expect the W component to penetrate deeper into a via and exhibit relatively better bottom coveragethan the Ti. On the other hand, the Ti component should exhibit relatively better sidewall coverage due to a higher fraction of Ti atoms arriving at oblique incidence. The net result is that sputter deposition ofTIW over topography is expected to lead to T i r W I p ron the field, an increased concentration of Ti on via sidewalls (Ti,W,-Vwith y > x). anda decreased concentration of Ti on the via bottom (Ti,,W, with y < x).This is seen i n the simulated film composition profile of Fig.
9.3 I, whichshow W-enriched TiW at the bottom and Ti-enriched TiW on the sidewails - an effect supported by experimental data of Liu et a/. 19.361.- V9.7 Refractory Metal SilicidesThe four basic applications for refractory metal silicides in advanced 1Cprocessing are as ( 1 ) contacts to Si, (2) polycide gate electrodes in whicha layer of silicide is deposited on top of polysilicon to produce a lowerresistance gate stack, ( 3 ) short-length local interconnects, and (4) selfaligned silicides (so-called salicides) in which the source, drain, andpolysilicon gate regions of an MOS transistor are selectively shuntedwith a low-resistance silicide by a process of metal deposition and a twostep annealing that leaves the oxide regions free of both unreacted metaland silicide. PVD is widely used for all of these applications to depositeither the metal component of the silicide (e.g., PVD Ti from a Ti targetfor a Ti-salicide process) or the silicide itself (e.g., PVD MoSi, from acompound Mo-Si target for a polycide bilayer).
While many metals formsilicides with a combination of desirable properties (e.g., metallic conduction with resistivity much lower than heavily doped polysilicon, stable contact formation to Si, high-temperature stability, and self-passivation in oxidizing ambients), the principal silicides exploited in VLSIdevices have been M S i where M = Mo, W, Ta, and Ti. Interconnectroadmaps suggest that Co will join this list for ULSI devices due to the328R. POWELL AND S.
M. ROSSNAGELFIG. 9.31Simulated Ti W~_~film composition in a 1:1 aspect ratio trench showing variation in stoichiometry with location [9.36]. The legend bar indicates relative Ti concentration in weight percent.attractive scaling properties of CoSi 2. Reviews of the physics and chemistry of silicides used in IC processing are provided in refs. 9.39-9.42,and in this section we simply point out several PVD-related aspects ofsilicide use.9.7.1 MSIx WHEREM = TA, MO, OR WAlthough these refractory silicides can be cosputtered using separate metaland Si targets, they are more often sputtered from a single, composite target with ultrahigh density (_> 95% of bulk density). Such targets are produced by mixing and pressing together fine particles of refractory metaland Si and then vacuum-sintering at high temperature and pressure to givea bulk density close to the theoretical maximum for the metallurgical alloy.Since target microvoids in a low-density target can trap gases and contamination that are subsequently released during sputtering, high-density targets have allowed silicide films with improved purity and reduced particulates.
Also, since a greater in-plane atom density is exposed to Ar §PVD MATERIALS AND PROCESSES329bombardment, a higher deposition rate is achieved at a given magnetronpower.Another aspect of composite silicide target sputtering is that the targetcomposition must be controlled to give the desired MSi x film on the substrate. For IC applications, a film with [M]:[Si] ratio close to the disilicidecomposition (TaSi2, MoSi 2, and WSi2) is favored. However, a film with anover-stoichiometric amount of Si is often used to optimize stress and resistivity and to provide an in-film source of Si for a surface-passivating oxide.Due to the different sputter yields of the alloy constituents in the target andthe difference in gas-phase scattering of the higher-mass metal atoms(181ya, 96M0, 184W) and the 288i atoms, the as-deposited films often have aslightly different stoichiometry than the t a r g e t - typically metal enriched.If an RF bias is applied to the substrate during PVD, the deposited film canbe further depleted in Si that is preferentially resputtered relative to thehigh-mass metal.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
