Roland A. - PVD for microelectronics (779636), страница 30
Текст из файла (страница 30)
28N1+2' IH+ , and 2H 2).also a reasonable compromise given the mass of common sputtered metalatoms such as 27A1 and 48Ti.In addition, Ar is inert and will not react chemically with the wafer, thematerial being sputtered, or the residual gas atoms. It is true that Ar can beincorporated into the growing PVD film and thereby affect film densityand growth morphology. However, at typical incorporation levels ( < 0.5%due to very low sticking coefficient of Ar) these effects are generally notsevere, and the levels are reduced further at elevated wafer temperature. Inany event, the situation would be far more complicated if the Ar werechemically reactive. Although other inert gases can be used (e.g., 4He,84Kr, 131Xe), Ar is readily available even in ultrahigh purity grades such as"6 nines" grade - 6N = 99.9999% and is relatively inexpensive (e.g., Krand Xe are ~ 20 and 75 times more expensive than Ar, respectively).
Also,He is difficult to pump with cryopumps and turbopumps and could interfere with vacuum leak checking with He leak detectors.The important vacuum concept with regard to Ar gas purity is that of relative arrival rate. Namely, to ensure high-purity sputtered films, the arrivalrate of sputtered atoms should be much greater than the arrival rate ofresidual gas atoms/molecules. From the kinetic theory of gases, the surfacearrival rate R of gas-phase molecules at temperature T (K) and pressure P154R.POWELL AND S. M.
ROSSNAGELis related to the average molecular velocity <v>the expressionR (molecules/cm2-sec) = n<v>-=43.51and the gas density n byXlO2'---P(MT)' I 2where P is in Torr and M is the mass of the molecule in amu. F o r the caseof Ar at room temperature, this reduces to R = 3.2 X 10" P. Therefore atatoms/cm2typical PVD pressure of 3 rnTorr, the Ar arrival rate is =sec.
Given the extremely low sticking coefficient of Ar for typical PVDfilms and process temperatures (< 0.001), the effective Ar incorporationrate at the film surface can then be below 1 0 1 ~ t o m s / c m ' - s e c ,which isabout 1 monolayer per sec (= 1 A/sec). Since film deposition rates for A1are = 1 pmlmin (160 A/sec), Ar incorporation of 0.5% is understandable.While A r incorporation is generally not a problem. small levels of contamination in the Ar can have a significant impact on film quality. Assuming a sticking coefficient near unity (not a bad assumption for residualspecies such as 0 and metals such as Ti and Al), = 1-2 A/sec of contamination will "deposit" at a partial pressure of 10-'Torr - which representsonly about 0.03% of the total 3-mTorr Ar pressure.
Viewed in another way.incorporation of contamination at rt rate of 2 A/sec during A1 depositionat 167 A/sec (1 pm/min) leads to a 1 % level of contamination in the film.Figure 5.33 shows the effect of residual gas on film resistivity and reflectivity of a sputtered 1.2-pm A1-I7cSi film 15.33. 5.341 and indicates theneed to keep reactive gas partial pressures very low (c 10 Ton-j for critical applications. The use of ultrahigh-purity Ar gas ( 6 N grade or betteris recommended) with short Ar delivery lines to reduce outgasing frominterior line surfaces is also necessary.
In addition. point-of-use purifiersare available for low-flow-rate applications such as PVD that can costeffectively purify 5N grade Ar to 7 N grade; remove H,O. 0,. CO. CO,.H,. and N, to below I 0 ppb; and thereby cost-effectively purify 5 N gradeA; up to 7IL' at the point of entry to the PVD chamber without the tenfoldcost premium usually associated with a 7 N vs 5N purity cylinder of Ar.Even if the Ar gas is 100% pure. outgassing from the walls of the chamber and/or from the wafer itself can introduce contamination into the sputtering ambient.
For a given pumping speed on a vacuum module, the ultimate base pressure is determined by the outgassing rate of the chamber(see Eq. (5.1)), and this in turn is strongly influenced by such things aschoice of materials and seals. surface finish, and bakeout procedures.Figure 5.34 plots process pressure versus level of contamination introduced from outgassing. Assuming that base pressure is lo-' Torr.
we see'155SPUTTERING TOOLSAII't=EyE::- 4 . 5 .......................................................................................................................................................ovi_~m-5.o.................................... i .......................................................................i ~ ...........................................i.,m0,o"::-5.5. . .
. . m ~ ....................... ';'9 ..... 9 ........ m . ~ . . m - m . - ~.j.......................... ;. ........................ ,. .................9~ii::,~- 6 . 0 ........ i ................. i ................................... ~,. . . . . . . . . . . . ~ . . . . . . . . . .
~~10-1010-910-810-710-610-5Partial Pressure (Torr)(a)10080,A60v...o..0~i40|w=,,,.2010-7,,10-6....10"510-4Residual Gas Pressure (Torr)(b)(a) Residual level of O~ versus AISi resistivity [5.33]; (b) residual level of H,, H,O, andN 2 versus AISi reflectivity [5.34]. (Figure 5.33b reprinted with permission from P. S. McLeod and L.D. Hartsough, J. Vac. Sci. & Tech.
AI4(1): 263-265 (1977). Copyright 1977 American Institute ofPhysics.)FIG. 5.33R. POWELL AND S. M. ROSSNAGEL156J,10-o -. . . . . . . . . . . . . .10-2c._o10 -4"~.......................i P r e s s u r eilppmEc.co0q,,,o>..I10-6 10-1o lppb 10_14 .10-16 -.......-''............ iii.l0.1iii .................. ~ . . . . . . . . . . . . .w J illW1WorkingIW9I[IW~10PressureiIi 10-12 T ~w a I waa IiiiiI I100(mTorr)Calculated level of contamination introduced due to outgassing versus process pressure,for several base pressures. Contamination-free PVD manufacturing favors UHV base pressure.FIG.
5.34that contamination levels < 1 ppm can be obtained with process pressure> 2 mTorr w assuming the Ar process gas is 100% pure. Since the trendin conventional PVD is toward lower process pressure ( < 0.5 mTorr) to reduce gas-phase scattering, this will increase the need for UHV-type vacuum practices. One exception to this is ionized PVD, which is carried outat much higher than conventional (15-20 mTorr) process pressure (seeChapter 8). In this case, it is necessary to throttle the chamber pump, whichdegrades the base pressure.Wafer HoldingThere are basically only two ways of holding a wafer during PVD: mechanical and electrical. As mentioned earlier, some PVD processes can bedone clampless; however, this rules out the possibility of backside gas,which requires several Torr for effective heat transfer, and would blow thewafer off the platen (the weight of a 200-mm wafer corresponds to a pressure of ~ 0.08 Torr).
Clampless processing also makes knowledge andcontrol of deposition temperature more difficult. Holding at the wafer edgewith a spring-loaded mechanical clamp ring or using only the weight of theSPUTTERING TOOLS157clamp (a gravity clamp) are commonly practiced, although provision forthe physical flat or notch at the wafer edge must be considered. Since thePVD film also deposits on the clamp ring, the ring can be a source of particles as the film builds up and flakes off. In addition, the ring could "stick"to the wafer during some elevated-temperature processes (e.g., a Si ringcould stick to a CVD W film by forming WSi2), requiring careful designand process-specific choice of materials.
Finally, the edge exclusion of theclamp ring (typically ~ 6 mm) prevents coating of valuable "Si real estate." For a 200-mm wafer this means that ~ 12% of the wafer area is covered by the clamp and cannot be occupied by a die. For these reasons, thetrend is away from frontside, mechanical holding and toward backside,electrical holding of wafers using electrostatic chucks (ESCs), sometimescalled e-chucks.The physics and technology of ESCs are dealt with in a number of excellent reviews (e.g., see refs. 5.35 and 5.36). ESCs operate via the principle of coulombic charge attraction, in which charges on the chuck electrode attract real or image charges in the Si wafer being c l a m p e d - whichis analogous to the attractive force between the plates of a parallel-platecapacitor, only in this case one of the plates is the wafer.A simple e-chuck is illustrated in Figure 5.35 and compared with amechanical edge clamp.
For simplicity, the lift pins that would raise andlower the wafer off of the chuck and onto the end effector of a robotic armare not shown. An upper insulator of thickness hdie! and relative dielectricconstant k coats the metal electrode of the chuck and is separated fromthe wafer by an effective vacuum gap h gap . A vacuum gap is inevitableover some fraction of the wafer surface when using a nondeformabledielectric (e.g., a material such as alumina and not an elastomer) due to themicro-roughness of the wafer backside and possible wafer bow or warp.Also, gaps are intentionally created at the dielectric surface by the groovesor channels that are used as a distributed path for backside gas.The expression for the electrostatic pressure PESC holding the waferdown (force per unit area of clamping) can then be shown to be12%V 2PF~sc+ hgapwhere the dielectric constant of free space e0 = 8.85 • 10 -1~ farad/meter.Assuming perfect contact (hgap = 0), Eq.
(5.7) reduces to158R. POWELL AND S. M. ROSSNAGELM=hmnlcal Clamp RingGer Pressurn ( = 5 Torr)Gas Prwaura ( = 5 Torr)Eleetrostatlc P ~ 3 u r =e $12 QFIG. .5..35ing PVD.v2f(hdkllk+ hgap)Illustration o f ~nrchanicalchuck and electrostatic chuck (ESC) for wafer holding dur-where Edlc,is the electric field across the dielectric in voltslmeter and P,,,is the pressure in pascals ( 1 Pa = 7.5 mTorr).
Therefore, the maximumpossible holding pressure is determined by the electric field strength of thedielectric at breakdown.A popular ceramic insulator for e-chucks is alumina (A1,0, with k =lo), whose breakdown strength depends on the method of preparation(e.g., bulk, plasma sprayed, anodized) but is typically in the range of10-1 5 Vlpm. Using these values in Eq. (5.8) gives a maximum pressure of= 350-500 Torr.
To provide a margin of safety, one would operate at avoltage several times lower than the breakdown point of the dielectric, andSPUTTERING TOOLS159the ESC pressure would be much lower than the calculated maximum(PEsc goes as V2). In addition, as noted earlier, perfect contact with h gap = 0is never possible over the entire wafer area given a wafer's backside roughness and the presence of gas conduction grooves machined into the chucksurface.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
