MATERIALS: Low-defect target metallurgy development for sub-0.18?m Al-based interconnects
12/01/1999
Murali Abburi, Vikram Pavate, Murali Narasimhan, Sesh Ramaswami, Applied Materials, Santa Clara, California
Janine Kardokus, Jane Buehler, Lip Yap, Johnson Matthey Electronics, Spokane, Washington
Arcing between the sputtering target and plasma during physical vapor deposition (PVD) of metal films is one of the primary causes of defect generation in interconnect metallization processes. With rapidly shrinking device geometries and stringent defect requirements for sub-0.25µm technology nodes, there is significant interest within the semiconductor manufacturing community in reducing arcing-related defects during metal deposition.
Of particular interest to manufacturers is the generation of defects during aluminum (Al) alloy deposition that result in catastrophic electrical shorts and opens. Such defects can vary from 0.5µm to greater than 100µm. Current commercially available microprocessors consist of five or six Al layers (0.5-2µm in thickness) where the cumulative effects of these defects can result in substantial yield loss during device manufacturing.
Figure 1. shows a typical arcing-related defect generated in a semiconductor device during Al film deposition. Due to the size and shape of this defect, the metal line has been destroyed. The unusual molten characteristic of this defect is commonly referred to as a "splat" or a "blob." Depending on the die layout on the wafer, this particular arcing event could have potentially destroyed a significant number of microprocessor or DRAM chips.
Arcing phenomenon
An arc is a general term used for any low-impedance condition that is equivalent to a high-power-density short circuit. Arcing in PVD sputtering systems can be broadly classified into two categories: 1) unipolar or cathodic, caused by arcing between the plasma and sputtering target; and 2) bipolar, caused by arcing between the sputtering target and shields or the substrate and shields. The majority of arcing events (up to 99%) are caused by unipolar arcing in which the plasma cloud discharges to a point on the target surface. This point is charged up because its surface conductivity differs from adjacent areas.
 |
Figure 1. Representative focused ion beam cross section of a splat (molten droplet) in the aluminum metal line of a defective semiconductor device.
One possible source of unipolar arcing is dielectric inclusions in the sputtering target that are exposed to the plasma. As previously reported [1], an electrical field applied to thin nonconducting films on a cathode surface generates a strong electron emission current. Another study [2] showed that electric field enhancement, due to dielectric surface charging, also increases the generation of secondary electron avalanches over the dielectric surface. If these are forced into motion, the resulting regenerative ionization causes formation of a conductive plasma channel, leading to dielectric breakdown due to surface flashover. This would cause localized melting on the target surface, resulting in a splat defect with a globular, molten appearance. Aluminum, with a lower melting temperature than titanium, tantalum, or copper, has a higher propensity toward this localized phenomena. Any of the above factors may initiate an arc, which can be sustained for a considerable time, leading to continuous creation of splats.
Target metallurgy and arcing correlation analysis
 |
Figure 2. In-film defect performance for Al sputtering as a function of Al target metallurgical properties.
The Al alloy sputtering targets investigated in this study were Al-0.5% Cu, with metallic purity greater than 5N5, i.e., the metallic impurity content was less than 5ppm. The Al-film deposition took place in the PVD chamber on an Endura PVD system. The operating conditions (pressure = 2mtorr; DC magnetron power = 10.6kW; deposition thickness = 1µm), which remained unchanged throughout the study, are representative of deposition conditions used for DC magnetron sputtering of Al films in semiconductor manufacturing. To quantify defect densities, we inspected as-deposited films using a Tencor 6200 or 6400. Additional analysis using scanning electron, confocal laser, and optical microscopy helped characterize defect morphologies. We also used energy dispersive spectroscopy (EDS) to identify the composition of the defects. Figure 2 shows the results from the testing of Al alloy targets, with contrasting metallurgies, specially manufactured for this study.
 |
Figure 3. Representative optical and Ultrapointe micrographs of splats.
Figure 3 shows the comparison of splat morphology from two targets with different metallurgies. Splats from Target A suggested that an isolated arc on the target surface, typical of secondary electron heating phenomena, was the cause for splat formation. By contrast, splats from Target B appeared to have been the result of an explosive arcing mechanism, typically associated with dielectric breakdown or due to entrapped gases. EDS analysis of these defects showed the composition to be pure Al. Correlation with the destructive metallurgical analysis of these targets suggests that Target B had significantly higher alumina inclusion content as well as higher gaseous oxygen, hydrogen, and carbon content when compared to Target A. While Target A appeared to have entrapped alumina inclusions, the level was lower than that found in Target B. Random splats from both targets appeared to have alumina particles entrapped in the center, indicative of the effect of alumina inclusions. The use of filtration during the casting of Al ingots provided optimum target metallurgy that was characterized by low inclusion content and gaseous impurities.
Improvements in metallurgy
Looking at the refining and casting processes for Al, dissolved gases in the melt come out of the solution during the solidification process and form voids in the solid Al. Foreign material contaminants or disturbance of the melt surface in the presence of reactive gases can result in inclusions. Disturbance of the melt surface in the presence of gases that can react with Al, particularly at high temperatures, can form nonmetallic inclusions such as aluminum oxides. Therefore, casting process design and controls are critical in producing good metal quality with minimal inclusions and voids.
Continuous casting allows access to the melt for high-efficiency degassing and inclusion removal. Done properly, the process protects metal from contamination sources and exposure to reactive gases and makes it possible to monitor the gas content in the melt and make adjustments, prior to solidification, to reduce gas levels substantially.
Ultrasonic C-scans are useful in detecting the presence of material defects. It is difficult to differentiate between a void and an inclusion with ultrasonic testing. Destructive analytical techniques, such as chemical dissolution using hydrochloric and nitric acid, can detect alumina inclusions in the sputtering target.
 |
Figure 4. Ultrasonic C-scan images showing differences in porosity and inclusion content between a) conventional and b) improved targets.
Improvements made to the casting process for the Al alloy targets significantly reduce inclusions and hydrogen incorporation. Figure 4 shows ultrasonic C-scans of the conventional and improved target metallurgies achieved by optimizing the continuous casting process.
Experimental verification of improved target metallurgy
While the above sputtering target characterization sampled targets for short durations, extended "through life" testing is a more effective method for characterizing performance because it detects random variations of undesirable metallurgical imperfections in the target. Representative targets with conventional and improved metallurgies, respectively, were tested through life (>4500µm, equivalent life of 4500 wafers of 1µm deposition each) to characterize splat performance. A comparison of in-film defect performance for conventional and improved metallurgy targets is shown in Fig. 5. The improved metallurgy targets result in a significant decrease in the in-film defect density. Results showed that targets with improved metallurgy significantly reduced film defect density.
 |
Figure 5. In-film defect performance for a target with improved target metallurgy.
We conducted systematic investigations to identify the cause of splat formation. Results revealed a strong correlation between the metallurgy of the Al alloy sputtering target and unipolar arcing. Our studies showed that undesirable metallurgical characteristics in the Al alloy target, such as dielectric inclusions, hydrogen content, porosity, grain size, and surface finish are the cause of unipolar arcing during sputtering and subsequent splat formation.
Acknowledgments
The authors thank Keith Hanson, Glen Mori, Hai Duong Pham, Cathy Cai, and Sunny Chiang (Applied Materials) for their advice and suggestions; and Fusen Chen, Jaim Nulman, Ashok Sinha, Tom St. Dennis, and Zheng Xu (Applied Materials) for their support of this project.
References
- Malter, Phys. Rev., 50, 48, 1936.
- Anderson and J.P. Brainard, J. Appl. Phys., 51(3) 1414, 1980.
For more information, contact Janine Kardokus at Johnson Matthey Electronics, 15128 East Euclid Ave., Spokane, WA 99216-1801; ph 509/252-8750, e-mail [email protected].