Roland A. - PVD for microelectronics (779636), страница 46
Текст из файла (страница 46)
1995).7.21. D. Butler, "Options for multilevel metallization," Solid State Tech., s7-sl0 (Mar. 1996).7.22. P. Singer, "Filling contacts and vias: A progress report," Semicond. hTt., 89-94 (Feb. 1996).7.23. M. Saran, "Grain-separation over vias filled by high-pressure aluminum extrusion," TechnicalProc. of the Electrochem. Soc. Syrup., vol. 97-I, abstract no. 380, pp. 48 !-482 (1997).Chapter 8 Ionized Magnetron Sputter Deposition: I-PVDPhysical sputtering from a magnetron cathode is dominated in most casesby the emission of single, neutral atoms.
Cluster emission is rare and if apositive (sputtered) ion were formed at the cathode surface, it could not beemitted due to the strong, opposing electric field present in the sheath.Negative ions, which are formed during the sputtering of some compoundssuch as titenates, zirconates, and other oxides [8.1 ] are not a factor in mostareas of inert gas sputtering of metal targets. The sputtered, neutral atomshave a wide variety of angular trajectories, and directional deposition isonly possible by means of filtering, such as using collimators or long target-to-sample distances. This is due, of course, to the inability to controllably direct the trajectory of a neutral atom.On the other hand, a metal ion, if it were formed away from the cathodesheath, it could be controlled by the presence of an electric field. The metalion's trajectory would be "straightened out" in a suitably strong electricfield located at the edge of the plasma, perhaps at the location of the sample.
This concept is the basis of ionized PVD, also known as I-PVD, IMP(ionized metal plasma), or IMD (ionized magnetron deposition). The intrinsic advantage of I-PVD is that a large fraction of the sputtered metalatoms can be ionized and controllably directed to the sample surface atnormal incidence. This compares to the relatively small fraction of sputtered atoms (typically a few percent) that pass through a directional filtersuch as a collimator.In I-PVD, metal atoms are sputtered from a conventional magnetronsource using an inert gas and a conventional magnetron power supply.
Asecond plasma, nominally different from the magnetron source plasma, isproduced in the region between the sputtering source and the sample usingthe same background gas. Some fraction of the sputtered metal atoms areionized as they transit this second plasma. Finally, just above the samplesurface, the metal ions are accelerated to the sample by the difference between the plasma potential (usually slightly positive) and the sample potential, which can be controlled externally to be either a zero or a negativevoltage.8.1 Experimental SystemsThe deposition of metal ions to form films has been reported many timesand is not a new phenomenon.
This is particularly true of ECR (electroncyclotron resonance) plasmas, which can be quite dense and which if not241R. POWELL AND S. M. ROSSNAGEL242tightly confined result in reasonable amounts of wall sputtering and subsequent metal ionization by the plasma. The practical application of thistechnology was demonstrated by Kidd et al. in the mid- 1980s using a highfield ECR plasma in which the sputtering plasma doubled as the ionizationplasma [8.2].
While this technique was not manufacturable, it led to workby Holber et al. at IBM Research [8.3] and by Barnes, Forster and Keller,at IBM East Fishkill [8.4]. The Holber group used evaporation from ametal boat of Cu or Al rather than sputtering into an ECR plasma chamber(Fig. 8.1) [8.2].Sputtering in the axial geometry of Kidd's system results in the formation of high-energy secondary electrons that can directly impact the sample. In the Holber work, once the ECR plasma was started in an inert gassuch as Ar, the metal evaporation source was turned on and the Ar gassupply turned off. This resulted in essentially a pure metal plasma, whichwas then used to deposit metal films onto negatively biased wafers justout of the field of view of the evaporation source. This work showed filling of various semiconductor features, although the overall tool designbased on evaporation and ECR was not easily converted to a manufacturable system.-.
. . . . 4.. . . .. -. .. I ::. B-Field1 Coils . , ,'4IMicrowaveWaveguide andWindowCu or AlEvaporationSourceHigh-VacuumValve andPumpFIG. 8.1Experiment of Holber rt ul. (8.3)1IIONIZED MAGNETRON SPUTTER DEPOSITION: I-PVD243The Keller et al. group used a 4-cm metal sputtering source and arrangeda second plasma antenna/electrode in the region between the cathode andthe sample (Fig. 8.2) [8.4]. The plasma antenna was configured to be amultiple-turn metal coil, which was operated at 13.56 MHz and tuned toform an inductively coupled RF plasma. The deposition chamber was confined with a multipole magnetic bucket to enhance the degree of metal ionization as well as to increase the uniformity of the plasma and reduce edgelosses. This work showed significant promise toward a manufacturingscale system because each of the components was known to be scalable._r--](-'32 %,,-34":" 383g4oi ~.37! ~'~"2424J1s~Doool3m2022FIG.
8.2I-PVD system from Barnes, Forster, and Keller patent [8.4].244R. POWELL AND S. M. ROSSNAGELLater work by Rossnagel and Hopwood examined the transition of thisgeneral scheme to manufacturing-scale proportions [8.5, 8.6]. This isshown schematically in Fig. 8.3. The magnetron was scaled first to 20 cmdiameter and then 30 cm, the latter case using a rotating magnet design. Awide variety of RF coils were examined, with the eventual use of coils of1-3 turns of diameter about 20% larger than the wafer, located about 1/4of the cathode radius from both the cathode and the sample. The RF coilsare operated at 13.56 MHz primarily due to the wide availability of powersupplies at that frequency. Some commercial versions of this technologyhave moved to lower f r e q u e n c y - as low as 1.9 M H z - because this isbelieved to result in a slightly higher level of ionization [8.7].A variety of RF coil designs have been explored that will have an effecton either the uniformity or the deposition processes.
The RF antenna in anI-PVD system is immersed in the plasma, unlike inductively coupled RFcoils used for etching applications. In that case, the coil is situated behinda quartz window outside the vacuum system. Since I-PVD tools are usedin most cases for the deposition of metal films, any window used to separate the coil from the plasma would quickly become coated with metal andwould no longer function as a window. One proposal is to shield the RFFIG.
8.3Manufacturing-scale I-PVD tool design.IONlZED MAGNETRON SPUTTER DEPOSITION: I-PVD245antenna/coil in the vacuum behind some sort of metal shielding. The RFfields could then penetrate into the plasma region through small holes,slits, or gaps in the shielding.
Unfortunately, because of the high densityof the plasmas involved (10'?/cm3), the Debye length of the plasma isshort enough (100 micron range) that the physicaI gaps would be quite difficult to build.Since the RF coil is located within the plasma chamber, two significantprocesses occur: First, the coil develops a negative floating potential dueto the higher intrinsic mobility of the plasma electrons compared to theions. Second, the coil may receive substantial amounts of metal depositionfrom the magnetron source.
The latter issue is of extreme importance tosemiconductor applications in that it may lead to thick film buildup, subsequent flaking, and particle formation.There are three approaches to this issue of film deposition on the RFcoil. The first is to simply remove the coil from time to time for cleaningor replacement.
This adds a regular maintenance task and the expense ofdowntime and base pressure recovery. which would be nominally similarto the expense of changing collimators.A second approach is to tune the inductively coupled discharge to reducethe rate of film buildup on the coil. The RF coil antenna is coupled to thedischarge both inductively as well as capacitively, the latter due to thephysical presence of the metal coil antenna in the plasma. By altering thetuning o f the coil. the ratio of the inductive coupling to the capacitive coupling can be altered.
At high levels o f inductive unupling. the negative biason the coil is low and there is little sputtering of the coil surface. At highlevels o f capacitive coupling. the sputtering rate of the coil is high. By altering the ratio of the coupling, the level of sputtering can be adjusted tobe just slightly lower than the deposition rate of metal from the magnetron.This results in a very slow buildup of metal on the coil surface and longlifetimes for the coil in the system.The third approach overcomes the concern of deposition on the coil byalways operating the coil in a net-erosion mode.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
