Tailoring sputtered Cr films on large wafers
07/01/1999
Valery Felmetsger, Sputtered Films Inc., Santa Barbara, California
As a sputtered film, chromium may exhibit a broad range of stress and electrical resistivity, depending on process conditions. Once obtained, a narrow range of stress and resistivity values enables successful use of these films in a wide variety of critical applications from advanced photolithography mask fabrication to under-bump metallization for high-density flip chips. Two new approaches for sputter deposition of uniform low-stress chromium involve post-deposition film treatment by RF plasma and a multistep process that alternates deposition and surface plasma treatment.
During sputter deposition, nonuniform distributions of charged particles and temperature across the wafer will create nonuniform film growth conditions. Resulting film properties, such as bulk resistance and stress, will likewise vary. This can limit many sputter deposition applications where it is necessary to create films with highly uniform physical properties.
Consider the need for uniform low stress. Stress control is very important for films used in advanced mask manufacture, high-density under-bump metallurgy, metallization of semiconductor devices used in high-power electronics (i.e., fast switching diodes, insulated gate bipolar transistors [IGBTs], etc.), and in any other applications requiring thinned wafers. High stress can cause mask image distortion, increased wafer warpage, and may even cause interface failure by buckling (compressive stress) or microcracking and peeling (tensile stress).
In another example, thin-multifilm membranes applied for x-ray lithography mask fabrication using silicon wafer substrates are especially sensitive to film stress and stress uniformity because the stress gradient across the pattern membrane is correlated to image placement errors [1].
We set out to characterize and improve intrinsic stress and electrical resistance uniformity during chromium sputter deposition. Our films were deposited on the Sputtered Films Endeavor AT cluster physical vapor deposition (PVD) system equipped with RF plasma pre-clean and Series IV S-Gun magnetron sputter sources. The S-Gun is a cylindrical concentric ring cathode sputter source with a central biasable anode. The two independently controlled cathodes, which can be energized by DC or AC (bipolar) power, enable uniform film deposition on up to 200mm dia. wafers. RF power can be applied to the wafer to create a negative DC bias.
Typically, 100-250? Cr films deposited with this equipment onto 200mm dia. silicon wafers will have a thickness uniformity better than ?4.5% (3s), a stress uniformity ?5.0 ? 108 dynes/cm2, and an average stress level <1.0 ? 108 dynes/cm2.
Both DC and bipolar deposition methods will produce high-quality, low-stress Cr films, but fine tuning of film properties requires development of special approaches.
Cr resistance vs. real thickness
Film thickness is measured directly using a surface profiler. The accuracy of stylus profilometry is typically ?50?, satisfactory for measuring film thickness greater than ~1000?. Thickness is also calculated from sheet resistance measurements, assuming that resistance is inversely proportional to thickness. This assumption requires uniform bulk resistance.
Our investigation, however, has shown that, at least for Cr films, the assumption of uniform bulk resistance is not always correct. Depending on deposition conditions, a Cr film can have areas with essentially different bulk resistance across the wafer surface.
With a series of graphs we show calculated bulk resistance distributions measured along a 150mm wafer radius using thickness measured with a profiler and sheet resistance data.
* From bipolar deposition, the top plot in Fig. 1 shows data for a 3180-3400? thick Cr film deposited at ambient temperature without wafer bias. Its physical thickness uniformity - (max. - min.)/(max. + min.) - equals 3.3%. However, thickness uniformity calculated from only sheet resistance measurements of film thickness is 11%. This difference is a result of the film`s bulk resistance change from 19.7W-cm in the center of the wafer to 24.8-25.5W-cm on the wafer edge. In comparison, Cr films deposited on 150mm wafers with applied RF bias and with applied heat (Fig. 1, plots b and c) reveal less bulk resistance variation - 15.5-17-cm and 16.5-18-cm, respectively - and much better agreement between profiler and sheet- resistance-determined thickness measurement.
* We also observed bulk resistance variation for DC sputtered films. For example, the data in Fig. 2 show a Cr film deposited at ambient temperature with no RF bias. Bulk resistance of this film ranged from 59?W-cm in the center of the wafer to 85?W-cm at the edge. The data in Fig. 2 show that DC sputtering deposition without wafer heat or RF bias can produce very low density films, if we assume that the areas of higher bulk resistance are less dense than the areas of low bulk resistance. When we applied wafer heat before or during deposition, we achieved Cr films with more uniform bulk resistance and a dramatic reduction in bulk resistance (15.3-15.7?W-cm).
Wafer heat is very important for processing large wafers. The PVD modules we used are equipped with radiant infrared heaters. This feature enabled us to apply and decouple highly uniform heat input, allowing the creation of heat profile recipes for 200mm wafers.
Our conclusion from these tests is that a 350-370?C pre-heat for ~30 sec applied in the process module immediately before sputtering allows creation of a film with more uniform physical properties. Film thickness distribution data measured with ellipsometry confirmed an essential improvement of optical property uniformities for 100? Cr films sputtered on 200mm Si wafers (see Table 1).
Stress uniformity
Our work showed also that magnetron sputtered Cr films have a significant intrinsic tensile stress, as high as 1.0-2.5 ? 1010 dynes/cm2. Applying an RF bias power in the range of 100-200W during film deposition decreased tensile stress, and in some cases drove it to compressive stress. Films deposited without RF substrate bias or with low RF substrate bias typically show very good tensile stress uniformity, particularly on 150mm or smaller wafers. The application of elevated RF bias power results in even lower tensile stress or compressive stress, but these Cr films have less stress uniformity.
Stress uniformity is more difficult to attain on larger wafers. For example, Cr films on 200mm wafers have significantly varying areas of stress on their surface; a film with an average stress of
108 dynes/cm2 can have different areas where the stress varies from -2.0 ? 109 to 2.0 ? 109 dynes/cm2. In these cases, a typical stress map consists of a quasi-circular field with compressive stress and another field, surrounding the first, but nonconcentric with it, that has more tensile stress.
A stress map clearly illustrates the effects of nonuniform ion bombardment and temperature on a growing film. We assume that areas of compressive stress are those that receive more bombardment compared to areas of tensile stress. In most cases we observed that the area of relatively uniform compressive stress has a diameter of approximately 5-6 in. We therefore focused our effort on making the RF bias-induced ion bombardment as uniform as possible.
We have found that dividing the deposition process into a few steps allows us to eliminate the effects of RF plasma localization. We call this "multistep processing." We tested this with a series of experiments completed in the same Cr process module, employing just RF discharge without a primary discharge at the target. Arguably, the RF discharge is strongly affected by the existence and intensity of the primary cathode discharge at the targets. Also, the RF bias in a PVD module is not optimized for etching. These experiments were pursued nonetheless to try to gain insight into RF bias localization. We looked at etch profiles on 200mm wafers covered with SiO2 vs. process parameters such as gas flow and RF power for continuous and multistep processes. The results in Table 2 show that the uniformity of this "quasi-etch" process was better when we used low gas flow and high power, and we achieved better uniformity when using the multistep etch regimes (i.e., 10-step etching, in this case).
We then tested whether or not these etching results could be directly extended to multistep sputtering. Our experimental observations indicate that during sputter deposition with RF bias there are two simultaneous discharges occurring in the process volume: the magnetron discharge localized on the targets and RF plasma discharge localized near the wafer. Through a series of experiments, we found that a multistep process, including a post-deposition RF plasma treatment in the same or special separate process chamber, enabled us to achieve a low-stress Cr film even on 200mm wafers. Specifically, we found we could change a film with tensile stress to zero or to a film with compressive stress by applying comparatively low 300-500W of RF power for 10-30 sec. For example, 100-400?-thick Cr films were deposited on oxide-coated silicon wafers without applying RF bias; they were transformed from a deep tensile to a deep compressive state via plasma treatment.
It is interesting that an unusually strong effect of this plasma treatment on stress was observed even for medium-thick films up to 2000?, deposited on the Cr under-layer completed in the same Cr process module, employing just RF discharge operation (without DC or bipolar discharge initiation). This implies that each thin layer in the Cr film during multistep deposition or the whole film after deposition can receive an additional RF plasma treatment between steps to tailor stress control precisely.
Table 3 presents some data illustrating significant influence of the post-deposition plasma treatment on film stress. Initial experiments have shown this multistep sputtering with interstep plasma treatment approach can improve film stress uniformity, especially with multiple plasma switching. A combination of the multistep method with wafer pre-heat is a highly efficient process for producing very uniform film properties across a 200mm wafer.
Conclusion
The physical properties of Cr sputtered films can be quite different across a wafer because of nonuniform deposition conditions. Thus, sheet resistance measurements sometimes do not reflect real film thickness uniformity because the film has areas with different bulk resistance. Sheet resistance data can be assumed correct for a film`s uniformity evaluation only when film deposition includes elevated temperatures on a substrate that is nonreactive with the material being deposited. RF bias-induced ion bombardment on the growing film surface can vary. This directly results in stress nonuniformity in thin Cr films.
Apart from the traditional methods of stress control by means of RF substrate bias applied during film deposition, we have developed two new approaches for low-stress film formation: post-deposition film treatment by RF plasma and multistep sputtering with inter-step plasma treatment. Both of these approaches achieve better stress uniformity than conventional methods. n
Acknowledgments
S-Gun and Endeavor are trademarked by Sputtered Films Inc. The author thanks Andrew Clarke for his helpful discussions.
Reference
1. W.J. Dauksher et al., "Method for Fabricating a Low-Stress X-ray Mask Using Annealable Amorphous Refractory Compounds," J. Vac. Sci. Technol., B13, p. 3103, 1995.
Valery Felmetsger received his PhD in material science and engineering from the Institute of High-Current Electronics, Tomsk, Russia. His work includes serving as head of the laboratory of vacuum and ion-plasma technologies, Metal-Ceramics Instrumentation, Ryazan`, Russia. Felmetsger is a process development engineer at Sputtered Films Inc., 320 Nopal Street, Santa Barbara, CA 93103; ph 805/963-9651, fax 805/963-2959, e-mail [email protected].