Roland A. - PVD for microelectronics (779636), страница 17
Текст из файла (страница 17)
The result is that the plasmaspreads farther around the chamber and substrate biases are increased.There is also the possibility of the formation of secondary discharges intight, confined areas of the chamber. This is due to a hollow-cathode trapping effect [3.17] in which secondary electrons are geometrically trappedin places such as pump ports, tubes, gas lines, and so on. Since virtually allproduction semiconductor PVD tools are operated DC, this is not an issue.However, RF tools will be needed for the complex dielectrics (k = veryhigh, or k < 4), and this extraneous discharge may become a problem.The shields in a magnetron tool are almost always grounded by design:They are made of metal and screwed or clamped to the grounded chamberwalls.
As such, they function as anodes in the plasma circuit and may actually carry a reasonable current to ground from the plasma. If a shieldwere oxidized at its contact point, this situation could change as the shieldbecame negatively charged and attained the floating potential. At thispoint, the shield is like a large capacitor: It contains a stored negativecharge that may be released rapidly in the form of an arc, which may result in material emission and deposition onto a sample. This same effectcould occur for a heavily deposited shield that develops cracks or breaksin its thick sputter-deposited coating. Upon air exposure, these cracks mayPLASMA SYSTEMS83allow moisture or air penetration under the film, which may result in flaking or eventual arcing.The arcs that occur on the cathode surface (see Chapter 11 for a discussion of target issues) are called unipolar arcs, in that they appear to haveonly one point of contact.
In reality, since an arc implies current flow, thearc is between the cathode surface and the plasma, hence the term unipolar. As will be described in Chapter 11, arcs may result in droplet emissionfrom the cathode as well as physical damage to the cathode. Present-daymagnetron power supplies (1998) are solid state, switching supplies, whichare less susceptible to arcing than older style, RC-filtered supplies.However, unless specific precautions are taken w such as arc-supressingc i r c u i t s - miniarcs of a few amperes may occur during sputtering and itwill be very hard to detect these arcs. This is an ongoing field of interest,and one that is only recently been recognized as a contributor to defect formation during deposition.3.11 Plasma Diagnostics and OpticalEmission in MagnetronsCommercial magnetron sputtering tools are generally not equipped withany sort of plasma, film, or optical diagnostics.
For laboratory-scale tools,several authors have described the use of Langmuir probes, which aresmall wires inserted into the plasma. By biasing this wire tip in the presence of the plasma, it is possible to determine the electron temperature,electron density, and plasma potential [3.17-3.20]. This type of experimentis greatly complicated by the large fluxes of sputtered metal atoms thatcoat the insulators holding the Langmuir probe and make determination ofthe functional area of the tip difficult.Laboratory-scale tools are also occasionally outfitted with depositionrate monitors that can be operated during deposition. These rate monitorsare typically small quartz disks driven at 5 Mhz or so, whose resonant frequency changes in a known way depending on the mass of the depositedfilm. Unfortunately, the geometry of cluster tools (Chapter 5) is such thatthere is no room for this sort of film deposition diagnostic.One laboratory-scale technique that could be applied to magnetron sputter deposition is optical emission spectroscopy (OES).
This is a passivediagnostic, in that it only measures light emitted from the discharge.However, the amount of light emission turns out to be a rather complexfunction of both the species being observed and the operating conditions ofthe tool. For example, if a wavelength of light is observed that is due to84R. POWELL AND S. M. ROSSNAGELneutral Ar, it would be expected that the intensity of that light would increase as the discharge power is increased. (The excitation probability foran atom depends on the plasma density, which should scale linearly withthe discharge power.) And this is indeed what is observed.
However, thesame observation for light from Ar ions is more complex. Since the number of Ar ions is dependent on the discharge power, as is the probability oflight emission, the light intensity from the Ar ions will scale roughly as thesquare of the discharge power.The same argument can be extended to the metal species from the cathode. Since the rate of metal sputtering scales with the discharge power, thelight intensity will also scale with the square of the discharge power. Formetal ions, since the number of metal atoms scales with the power, theprobability of ionization scales with the power, as does the emission probability, the light intensity from the metal ion species scales as the cube ofthe discharge power (Fig. 3.21).
This is all explained in more detail in[3.21]. The end result, though, is that unless the emission wavelengthsfrom the discharge are well identified, measurements of their intensity todiagnose or monitor the discharge will be very hard to decipher.Needless to say, production PVD tools are rarely, if ever, outfitted withoptical emission detectors. In fact, most tools have virtually no windowsFIG. 3.21 Optical emission spectroscopy (OES) of emission from Ar neutrals, Ti neutrals, and Tiions as a function of discharge current during magnetron sputtering of Ti in Ar [3.21 ].PLASMA SYSTEMS85and it is quite hard to see the plasma.
This is because the tooling is generally quite tight to capture all of the deposition for subsequent tool cleaning:If a window to see the plasma were there, it would eventually become deposited with metal, and eventually that metal would flake off and form particulates. However, since OES is widely used for diagnosis of reactive etching technology, its use in PVD tools cannot be precluded.
Any such use,though, should be carefully tailored to observe only the desired species.A very common diagnostic now being outfitted on most production PVDtools is a mass spectrometer or residual gas analyzer (RGA). The role ofthis tool is obvious: to observe and quantify the composition of the background gas inside the chamber both during sputtering and at the nominalbase pressure. Since most sputtering runs occur at pressures of a fewmTorr, it has often been necessary to differentially pump the RGA, whichtends to not function well at pressures above 0.5 mTorr.
Differential pumping requires a small turbo pump and backing pump and brings up the possibility of oil back-contamination to the chamber. It also provides only amoderately accurate look at the internal gas composition of the chamber,as the plasma region is generally sampled through a small aperture andwith some distance to the RGA (typically tens of cm). However, the RGAcan be extrememly useful for leak detection as well as for observation ofresidual water vapor and impurities in the gas lines. Since newer-generation RGAs have become avaiable that are compact and may not even require differential pumping, they have become nearly standard equipmenton the manufacturing production-scale tools. Simply detecting a problemwith the RGA that would compromise a single wafer run could easily justify the added expense of the RGA ($10k), as wafers late in their production cycle can easily be worth $5-1 Ok per wafer.References3.1.
H. R. Kaufman and R. S. Robinson, Operation of Broad Beam Ion Sources, CommonwealthScientific, Alexandria, VA, 1987.3.2. D. B. Meveded, P. Mahadevan, and J. K. Layton, "Potential and kinetic electron ejection fromMo by argon ions and neutral atoms," Phys. Rev. 129:2086 (1963).3.3. C. D. Child, "Discharge from hot CaO," Phys. Rev. 2:492-511 (1911).3.4.
B. Chapman, Glow Discharge Processes, Wiley, New York, 1980.3.5. D. Bohm, "Minimum Ionic Kinetic Energy for a Stable Sheath," in The Characteristics ofElectrical Discharges in Magnetic Fields, pp. 77-86, A. Guthrie and R. K. Wakerling, Eds.,McGraw-Hill, New York, 1949.3.6. E Chen, Introduction to Plasma Physics, Plenum Press, New York, 1974.3. 7. J. S. Logan, "RF Diode Sputter Deposition and Etching," in Handbook of Plasma ProcessingTechnology, pp.
140-150, S. M. Rossnagel, J. J. Cuomo, and W. D. Westwood, Eds., NoyesPublications, Park Ridge, NJ, 1990.86R. POWELL AND S. M. ROSSNAGEL3.8. Applied Materials, Santa Clara, CA.3.9. S. M. Rossnagel, "Induced drift currents in circular planar magnetrons," J. Vac. Sci. & Tech.AS: 88-91 (1987).3.10. P. Clarke, U.S.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
