Roland A. - PVD for microelectronics (779636), страница 29
Текст из файла (страница 29)
However,the molecular regime is the one relevant to PVD processing. In this case,the efficiency of heat transfer is independent of the exact gap spacing,and the heat conductivity is proportional to the gas pressure.R. POWELLAND S.
M. ROSSNAGEL148Figure 5.30 presents calculated heat transfer parameters for a variety ofgases. Even though lighter gases such as H 2 and He conduct heat betterthan Ar, they are not typically used for the practical reasons stated earlier.In any case, as long as the gap is small compared to the gas mean free path,the heat transfer coefficient h at pressure P is given by h = aAP, where Ais the free molecular conductivity and the accommodation coefficient ameasures the extent to which molecules reflecting from a wall will adjusttheir kinetic energy to the temperature of the wall.
For completely roughened surfaces, a ~ 1; for highly polished surfaces, a << 1. Taking ce = 0.5and using the data for Ar in Fig. 5.30, we see that at 4 Torr, a value of h =aAP = 0.5 x 9.3 mW/cm2-K-Torr x 4 Torr ~ 20 mW/cm2-K is obt a i n e d - provided that the gap spacing is less than the mean free path,which, from Fig. 5.30, is h = (52/xm-Torr)/(4 Torr) = 13/xm. Maintaininga gap this small is challenging since the pressure behind the wafer tends tobow its center outward.The theory of mechanics allows one to calculate the central deflection,y, of an edge-clamped Si wafer of radius r and thickness t subjected to aBSA gas pressure P:3 pgr4)[l][ (m - 1)(5m + 1)]Y= ~~-~-3}(m 2)(5.4)where the material-dependent quantities in brackets are the elastic modulus, E, and m is the inverse of the Poisson ratio v, i.e., m = l/v.
Wheny = 0, the wafer is assumed to be directly in contact with the planar waferchuck. For Si, E = 1.5 X 10 7 psi ~ 7.8 x 10 8 Torr and m = 3.3 (v = 0.3).FIG. 5.30 Gas cooling coefficients for a variety of gases encountered in PVD processing, with Arhighlighted because of its use in backside-gas-assisted heat transfer. (Adapted from ref. 5.31 with kindpermission from Elsevier Science-NL, Sara Burgerhartstraat 25, 1055 KV Amsterdam, TheNetherlands.SPUTTERING TOOLSFor a 200-mm Si wafer (rduces to=149100 mm; t = 0.675 mm), Eq.
(5.4) then re-y (mm)=0.29P (Torr)(5.5)We see that an increase of 1 Torr of BSA widens the gap by 0.29 mm .=10 mil. Hence, as one increases backside gas pressure to improve heattransfer efficiency, the added displacement due to wafer bowing can resultin z >> A at the center. This results in nonuniform heat transfer. In addition, even though the thermal conductivity of the backside gas is greater atthe higher pressure, the added gap spacing translates into an increasedthermal transfer path length and can actually increase the thermal resistance between the wafer and chuck. A widely used method to deal with thissituation is to clamp the wafer over a domed platen whose shape is machined in advance to anticipate the bowed wafer shape taken during BSAoperation.Although the backside-gas-coupled chuck is widely used to controlwafer temperature during PVD, other methods can be used.
For example,radiant heating of wafers by a tungsten-halogen lamp array located outsideof the process chamber has been reported by Clarke [5.32] as a way of improving wafer thermal response time and reducing contamination fromchuck outgassing and mechanical clamps. In this case, radiation directlycouples into the backside of the Si wafer through a quartz window and nobackside gas is used during PVD - an approach similar to the methodused in lamp-based rapid thermal process (RTP) systems for annealing andfor rapid thermal CVD.5.3.5 PVD MODULEFigure 5.31 shows schematically the basic elements of a generic PVDmodule. Sputtering is provided by a planar DC magnetron with a large-diameter circular target (e.g., = 12-14-inch diameter for an 8-inch wafer)and relatively slow mechanical rotation (= 40 rpm) of a permanent magnet array behind the target is used to improve andlor to tailor the erosionprofile of the target.
This type of sputter source is not the only one used inPVD for microelectronics (e.g., an S-GunTMsputter source with separatelybiasable dual cathodes is used by Sputtered Films); however, the planarmagnetron is quite common, and the technology of this source was discussed in more detail in Chapter 4. Source cooling is provided by flowingwater over the magnets within the source housing - the magnets beingsuitably potted in a water-resistant epoxy to prevent corrosion. The entire150R.
POWELL AND S. M. ROSSNAGELBackside GasUHV Vacuum ChamberI,7.1Basic elements o f a generic PVD rnodulesource housing containing the source-target assembly can be pivoted intothe face-up position to allow for target changes or chamber maintenance.Due to the weight of such assemblies (> 100 Ib is not unusual), the mechanical leverage of a source manipulator is often used to assist the operator. Figure 5.32a illustrates how a PVD source on a vented module can berotated from the face-down to the face-up position for target change ormaintenance. A new and eroded target on two such adjacent modules areshown in Fig.
5.32b with an eroded target (right) and newly replaced one(left).To confine the plasma discharge to the vicinity of the target face and toprevent unwanted sputtering of other parts of the target assembly, a metalground shield is contoured around these surfaces at a distance less than thatof the cathode dark space. Metals shields are also used to prevent sputteredmaterial from directly coating the chamber walls, although the coatedshields themselves must be periodically removed for cleaning. Processgases such as Ar and N, are introduced into the chamber by single-pointSPUTTERING TOOLSPVD source accessible fortarget change or maintenanceFIG. 5.32(a)injection from a simple tube or by multiple-point injection - e.g., from aperforated tube or "spider" manifold.For reactive sputtering of films such as TIN, a uniform distribution ofthe reactive gas is desired over the wafer surface, and equipment supplierscarefully engineer the location of both gas injection and pumping ports(often guided by computer modeling of gas flow around the sputter shieldsand other chamber fixtures).
A mechanical shutter is also sometimes provided between the source and wafer to allow a Ti target that was nitridedduring PVD TIN deposition to be cleaned before a subsequent PVD Tistep. The wafer is located relatively close to the target in most systems(= 5 cm) and is positioned on a substrate holder that usually is configuredfor active wafer heating via a heater block with backside-gas-assisted heattransfer.Mechanical clamping of the wafer to the wafer holder is generally accomplished through mechanical pressure at the wafer edge using a continuous clamp ring or finger clamps. More recent designs employ electrostatic152R. POWELL AND S.
M. ROSSNAGELFIG. 5.32 Example of a manipulator used to assist the operator in removal of PVD source assembly (Quantum T M PVD source shown courtesy of Varian Associates).chucks (ESCs), which permit uniform holding over virtually the entirewafer area and avoid such things as edge exclusion and particles associatedwith mechanical clamp rings. However, it should be noted that many PVDsteps can successfully be carried out clampless m only gravity holds thewafer down. Some chuck assemblies also allow for z-axis motion to varysource-to-substrate spacing up or down by a few cm. This z-axis motion canbe used as a "tweak" to optimize film uniformity during target life or to usethe same chamber to perform processes such as PVD Ti and PVD TiN thatin general require different source-to-substrate spacing.As we have seen, RF bias can be applied to the wafer to enhance and/orcontrol ion bombardment during precleaning.
Wafer bias is also used toSPUTTERING TOOLS153improve the directionality of PVD, as in the case of RF-ionized PVD (seeChapter 8), and to improve film properties during film growth ~ e.g., toincrease film density by removing loosely bound gas atoms or to improvestep coverage by resputtering metal atoms and increasing their surface mobility.
As a result, some wafer chucks for PVD also provide an RF bias capability. We note that designing a high-temperature electrostatic waferchuck for PVD ( > 500~ wafer temperature) that is also capable of RFbias and z-motion is a nontrivial engineering exercise.Argon Gas for PVDArgon gas makes up less than 1% of the composition of room air in a PVDmodule before pump-down, but can make up 20-100% of the process gas(e.g., Ar/N 2 or Ar) during PVD. Being the primary sputtering medium, Aris deserving of some discussion.
There are several reasons why Ar is usedso widely in PVD. Physical sputtering involves a direct momentum transfer, which depends on the relative masses of incident ion (m) and targetatom (M) through the reduced m a s s / x = 2(mM)/(m + M). 4~is relatively heavy (m = 39.948 amu) so that its sputter yield will in general bemuch larger than that of a typical residual gas that is ionized in the sameDC magnetron plasma (160+ ' 320++ Mass 40 is"-'2 . InN+. . .
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
