 |
 |
 |
 |
|
Comments? Suggestions? Send us your feedback.
MicroMagazine.com |
|
Process Tool Support Increasing PVD tool uptime and particle control with twin-wire- arc spray coatingsReed W. Rosenberg, Pentagon Technologies Tests on PVD chambers show that an aluminum twin-wire-arc spray coating effectively reduces the flaking of refractory metals from chamber shields and helps control particle generation. Escalating demands for product,together with highly conservative capital expenditures in recent years, have placed the semiconductor industry in a phase of constrained capacity. This situation makes it imperative for fabs to improve overall wafer die yields. To help achieve this goal, semiconductor manufacturers have focused on ways to improve tool uptime and utilization to drive down the cost of ownership. Equally important, in the face of an inexorable trend toward smaller geometries, is the need to reduce and control process-generated particles that can cause product defects. The industry has focused on particles generated by physical vapor deposition (PVD) refractory metal processes involving titanium nitride (TiN), titanium tungsten, cobalt silicide, tungsten silicide, tantalum, and tantalum nitride. These metal compounds are used for barrier/liner processes because they form high-stress films when deposited. However, high stress, along with plasma and thermal cycling, tends to cause refractory metals to flake off chamber shielding and generate particles. Historically, the industry has combated flaking and particle generation by performing shield-change preventive maintenance and increasing the level of pasting. But as device geometries have shrunk and particle size and limit specifications have tightened, the frequency of shield-change preventive maintenance has also increased, resulting in a higher cost of ownership. To increase the shield life of PVD systems and help control particles, a shield-coating process known as twin-wire-arc spray (TWAS) was refined for PVD systems by Thermal Coating, a start-up firm that has become a subsidiary of Pentagon Technologies (Fremont, CA). This article discusses the application of the TWAS method to semiconductor process operations and presents the results of experiments that were conducted in order to advance the use of the coating beyond the R&D stage and into production applications. The Cause of Particle Generation Several mechanisms are responsible for particle generation inside PVD systems. The most common is simply the failure of the highly stressed deposited material to adhere to the shielding substrate. The resulting peeling and flaking that occur cause the film to break up into small particles that are then transported throughout the chamber. When particle detection techniques identify critical contamination levels, a preventive maintenance procedure is called for to correct the situation. In part, the problem stems from a mismatch in the coefficients of thermal expansion, crystalline lattice, and adhesion strength between the materials in use. Commonly deposited barrier materials have stress levels that can exceed 1010 dyn/cm2. These films create enough force in a thin layer to overcome the strength of adhesion to some shield substrates. As the plasma is cycled on and off, alternate periods of heating and cooling are created. Even a small mismatch in the thermal expansion coefficients can produce enough force to delaminate the deposited film from the shielding. Thus, thermal cycling of the shields can come from the plasma, the heater table, or the heat of condensation from the deposition material coming from the target. The Quest for Effective Shielding Materials Traditionally, attempts to eliminate flaking have involved increasing the surface roughness of the shielding and using shielding materials that more closely match the expansion and contraction rates of the deposited films. Although these efforts have helped limit flaking, none have proved completely satisfactory. For example, stainless steel is the most common shielding material because it is inexpensive, durable, strong, and chemically resistant. Parts made of stainless steel have a much longer life cycle than those made of other shielding materials. Moreover, stainless steel can be easily cleaned and is easy to fabricate. Nevertheless, stainless-steel shielding material has a fairly short life cycle, as indicated in Figure 1, which presents coupon stud pull data from a series of PVD tantalum film adhesion tests. When the shielding material's adhesion strength falls to between 1 and 2 kpsi, as is the case with stainless steel, the tantalum film undergoes catastrophic failure and delaminates from the substrate surface. Although these tests show relative differences in film adhesion and expected shield life, actual shield life and performance depend on the type of tool, the process, the frequency of preventive maintenance, and parameters that affect time, temperature, and cleanliness.
Efforts to select other effective shielding materials have had mixed results. The use of shielding made of titanium or molybdenum (Mo) has resulted in increased system operating costs. Titanium shielding is difficult to clean and reuse. Because it is not very chemically inert, it is particularly susceptible to etching by chemical solutions such as KOH, HNO3, HF/HNO3, and HNO3/HCl, which are required to remove most refractory materials during the cleaning process. Molybdenum shielding is expensive and difficult to fabricate. Although molybdenum's coefficient of thermal expansion is closest to that of many refractory metals, it does not adhere ideally to titanium, titanium nitride, or titanium tungsten. Because they are expensive to manufacture, pure molybdenum shields have not been tested to verify their expected life cycle. Stainless-steel shielding works well with PVD aluminum, because aluminum adheres to stainless steel well even when the shielding is electropolished. Because of that, attempts have been made to sputter-coat stainless-steel shielding in an aluminum deposition chamber and then place that shielding into a refractory-metal deposition chamber. Although such quick feasibility tests have not shown great promise, they have indicated that aluminum films deposited over a stainless-steel substrate seem to increase shield life in PVD TiN processes. Consequently, more tests were developed to investigate the use of aluminum shielding in TiN technology and the viability of the TWAS method in semiconductor manufacturing processes. Of the various shielding materials that have been tested, aluminum has been the most successful. It is easy to fabricate, and refractory materials adhere to it well for several reasons:
Despite its strengths, aluminum shielding, like titanium shielding, is difficult to clean and reuse. It, too, is susceptible to etching by the chemical solutions used to remove refractory materials during the cleaning process. In short, there seems to be no single shielding material that is easy to clean, chemically resistant, reusable, easily fabricated, highly adhesive to refractory metals, sufficiently texturable via abrasive blasting, and strong enough to resist bending. Various materials meet some combination of these criteria, but none provides them all. Perhaps a hybrid of two materials could provide a possible solution to the problem of flaking and particle generation in PVD chambers, but in the absence of such a hybrid, the semiconductor industry was in need of a new approach. Enter Twin-Wire-Arc Spraying For years, aerospace maintenance facilities and machine shops have used twin-wire-arc spraying (TWAS) to deposit highly textured films onto almost any metal. The TWAS process is simple. Two rolls of metal wire, each with an opposing electric charge, are fed together through a motorized gun. During this procedure, the wires nearly touch each other, creating an electrical arc. Meanwhile, a highly compressed gas is fed through an orifice located behind the point of electrical contact. As the arc is created, a gas nozzle atomizes the molten metal and propels it onto a shielding surface. A coating of the metal measuring several thousandths of an inch thick is sprayed onto the shielding, and multiple layers are used to obtain the desired film thickness. The system is fully adjustable for deposition rate, surface roughness, bond strength, porosity, and density. TWAS offers several benefits. Unlike other coating technologies for semiconductors, including flame spray and plasma spray, TWAS does not produce high levels of contaminants such as water vapor, hydrocarbons, or carbon black. When they appear during the TWAS procedure, these contaminants are trapped in the film as it is deposited and then outgassed once the PVD system is under vacuum. As the shields are heated and plasma is applied, even more outgassing occurs. Films produced by TWAS also produce high bond strengths and do not require preheating or bond coats. Contamination is kept to a minimum because >99.4% pure aluminum wire is used, and the compressed gases are filtered to <0.1 µm and gettered to <50 ppm of moisture. TWAS coating is fairly porous and gas permeable, which aids in outgassing during vacuum and process usage. Less-porous coatings with lower gas permeability trap gases during deposition, causing slow outgassing at a rate of 0.6 to 1.5 µg/cm2 after 4 hours under vacuum. Repeated parts washing after spraying only increases this outgassing problem. Although TWAS coatings increase the shielding surface area significantly, they do not increase system pumpdown times or affect ultimate base vacuum if they are cleaned after spraying, baked, and vacuum packaged. TWAS can be used to deposit highly textured films onto almost any metal surface. Arc-sprayed aluminum films are highly adhesive to stainless steel and have bond strengths up to 7500 psi. Great adhesive strength and surface roughness produce shielding that can hold substantial quantities of refractory materials, especially titanium nitride, tantalum, and titanium tungsten. In some PVD systems, TWAS-coated shields can hold these materials for as long as one full target life. The deposited film does not delaminate even when the system is vented. TWAS coating addresses five critical concerns: increased surface roughness, deposited film stress, the mismatch in the coefficients of thermal expansion, compatibility with refractory materials, and bond strength.
TWAS Testing and Results After it was determined that the TWAS method was beneficial to particle control in PVD chambers, a program was initiated to take TWAS shield coatings from R&D into commercial use. The first process that was investigated was PVD tantalum on aluminum shields. Whereas the interval between preventive maintenance cycles had been 300 kWh before the initial TWAS application, that interval doubled to 600 kWh after a TWAS coating was administered. That result closely matched the coupon stud experimental data for aluminum and molybdenum TWAS coating shown in Figure 1. In a test performed in an Endura 5500 from Applied Materials (Santa Clara, CA), the results of which are shown in Figure 2, titanium nitride particles that had flaked off in chamber 2 before TWAS coating were discovered to average 1.7–2.3 µm in size. Post-TWAS data indicated that the average particle size dropped dramatically to <0.5 µm in size. When researchers examined the chamber interiors that had uncoated stainless-steel shields, they found that 36 to 48% of the contamination within those chambers was from TiN flakes. After TWAS coating, TiN contamination measured 9 to 12%. These data confirm that the use of the TWAS method decreases the time required to perform chamber-wipe cleans by 50% or more.
It is extremely difficult to determine the yield impact of a single process change on a semiconductor device or process tool, but some test methodologies can predict yield impact with some accuracy. Figure 3 displays the results of one such method, an electrical test with millions of test cells in a comblike structure. This test was performed before and after TWAS coating of the stainless-steel shields in two identical chambers with very similar histories. Figure 3a shows a pre-TWAS mean level of 0.063 defects/cm2, or 20 >0.2-µm particles per 8-in. wafer, while Figure 3b shows a post-TWAS mean level of 0.037 defects/cm2, or 12 >0.2-µm particles per wafer. The pre-TWAS maximum defect density level was 1.42, compared to a much lower post-TWAS level of 0.28. The relatively out-of-control standard deviation of 0.115 before TWAS coating was reduced to 0.042 after TWAS coating, a threefold improvement.
If coated shields are not properly cleaned, the TWAS film can delaminate early or not meet particle-performance requirements. Such delamination can have several causes, including too-heavy substrate oxidation, heavy oxidation of the aluminum arc spray coating, the presence of contaminants at the substrate surface, too-low Ra roughness prior to arc spray coating, and the existence of too many ingrained particles before coating. Table I presents electrical test device yield data demonstrating the need to use the most advanced cleaning processes to clean PVD shields, especially for device generations below 0.18 µm. Two different cleaning technologies were used to clean the shields in PVD chambers processing 0.22- and 0.13-µm devices. While cleaning process A performed well for 0.22-µm devices, it failed for 0.13-µm devices. In contrast, the implementation of cleaning process B, which can remove ingrained particles from the shields, enabled high yields for both 0.22- and 0.13-µm devices. The difference between the two processes is that process B includes special steps to reduce and remove residual, loosely held, and embedded grit particles in the shielding.
Conclusion The data presented in this article demonstrate that aluminum TWAS coatings can increase the life span of PVD chamber shields and help control particle generation. It is possible to achieve a two- to threefold increase in effective shield life if the proper shielding material and coating are used. Increased shield life and better particle control can reduce the inventory of spare kits that a fab must have on hand, lengthen preventive maintenance cycles, shorten preventive maintenance events, improve overall equipment efficiency, and reduce the cost of ownership. The benefit each customer will derive from implementing the TWAS method depends greatly on process module conditions, process recipes, process module utilization, module cycle times between wafers, and the control of processes and protocols for performing shield-change preventive maintenance. While the TWAS coating helps to mitigate the effects of these variables, its ultimate performance level is still affected by thermal and plasma fluctuations, gas phase impurities, target performance, and the existence of tool areas where shields or targets have been grit blasted and not coated with TWAS-deposited films. Reed W. Rosenberg is the director of engineering at Pentagon Technologies (Fremont, CA). Before joining the company, he ran the applications laboratory at Novellus Systems and headed up laboratory efforts at BOC Specialty Gases. Rosenberg has more than 17 years of experience in contamination control and yield enhancement, is the author of more than 20 papers, and has two patent applications pending. He received a BS in chemistry from the University of Arizona (Tucson) and an MS in chemistry from San Diego State University. (Rosenberg can be reached at 510/723-2121 or rrosenberg@pen-tech.com.)
 |
Buyers Guide | Media Kit
All rights reserved.