As demand for device slips below 0.18m m, are relying more machinable help get silicon wafers through the fab line quickly with less consumable costs and fewer defects per wafer. OEMs and fabs want to reduce the cost per wafer so they are looking for longer life and/or low-cost components.
WHY MACHINABLE PLASTICS?
Chemical-mechanical planarization (CMP) is a highly accurate polishing process that flattens and smooths silicon wafer surfaces. The process creates a more uniform platform on which to build ultrafine, multilayer integrated circuitry. The more widespread use of CMP is posing challenges for semiconductor manufacturing consumables.
Use of CMP is growing as wafer-planarity requirements become more stringent for both within-die (WID) and within-wafer (WIW) processing. One method for boosting WID and WIW repeatability and yield is to use one retaining ring material for the entire range of CMP processes. A single ring per wafer also lowers overall consumable costs as one material can handle rings for the entire fab line. A single-ring approach, however, means the ring will see a wider array of chemistries including a broad pH range. It will also get exposed to a number of highly aggressive polishing processes, much more so than currently is the case for multiring setups.
Metal retaining rings corrode when exposed to acidic chemistries and are more easily abraded. Metal-particle contamination causes short circuiting and reduces the number of viable ICs per wafer. Ceramics, on the other hand, historically have been materials of choice for retaining rings. Ceramic is less costly and provides a longer-lasting consumable, but is more brittle and costs more to machine than its plastic equivalent. Ceramics also generate abrasive particles that scratch wafers.
Advanced machinable plastics are increasingly becoming the material of choice for consumables used in building 300-mm silicon wafers. That's because they can be tailored to withstand the host of chemistries associated with the oxide, copper, and tungsten CMP processes. They also remain dimensionally stable over a wide temperature range and provide enough stiffness for handling larger wafers. These qualities help promote superior WID and WIF wafer planarity results during polishing.
Advanced plastics can also be formulated so they are inherently free of ionic impurities and outgas very little. This is a boon to designers looking to thwart the trace contamination from sodium, aluminum, iron, copper, lithium, and other inorganic elements that commonly leaches out of conventional ceramic retaining rings.
|COMPARISON OF OTHER CRITICAL VARIABLES|
|Critical properties||PPS||PEEK||CMP LL5||CMP XL20|
Heat-deflection temperature (264 psi), °F
1 to 14
1 to 14
2 to 9
2 to 10
Coefficient of friction
Coefficient of linear expansion, in./in./°F
|2.8 x 10-5||2.6 x 10-5||4.5 x 10-5||1.7 x 10-5|
Flexural modulus, kpsi
Hardness (M scale)
Stability of machined dimensions (flatness, parallelism)
Relative cost ($/wafer Normalized to Techtron PPS)
p = Poor, m = Medium, m+ = between medium and good; g = Good
MATERIALS FOR CMP CONSUMABLES
Key drivers for designers looking to replace metal and ceramic CMP retaining rings with advanced plastics are lower ring cost per wafer, processing ease, longer consumable service life, and a good balance of high chemical, thermal, and abrasion resistance.
Originally acetals then Torlon PPS had been predominantly used. Recently, a few more contenders have been added to the stable of highperformance plastics suitable for CMP wafer-retaining rings. According to Richard Campbell, manager of product development, Quadrant Engineering Plastic Products, Reading, Pa., one of the key performance needs for next-generation CMP materials is greater stiffness. Stiffer materials are a must for handling delicate 300-mm wafers. They also ensure better planarity for both WID and WIW polishing and ease wafer loading on the fab line. These advanced plastics also need better chemical and abrasion resistance to ensure longer consumable life. In addition, they must be highly pure to eliminate contaminants.
To test new candidate materials, Campbell devised an in-house procedure to simulate the CMP process. Test results using the procedure effectively screen and rank materials for use as CMP consumables.
The basis of this test, says Campbell, is a Buehler polisher/grinder assembly. "The apparatus uses a Vector Power Head and Beta model single wheel grinder/polisher turntable that polishes four 1.25-in.-diameter samples simultaneously. One sample, Techtron PPS (polyphenylenesulfide), serves as a control to which the results of the other three samples are normalized."
During polishing the surface of the rotating pad is flushed with slurry solution while the head containing the four test specimens rotates in a corotational direction. The 10-in.-diameter pad rotates at 100 rpm (3,200 ipm) while the head rotates at 60 rpm (300 ipm). Each sample is held against the polyurethane pad with 4 psi of pressure.
Three types of slurries are commonly used in the industry, says Campbell. These are known as oxide, tungsten, and copper. Wafers may see all three slurries as they travel through the various CMP steps.
Oxide uses a high pH caustic solution such as Semi-Sperse SS-12 from Cabot Microelectronics Corp., Aurora, Ill., says Campbell. "Cabot's tungsten slurry, Semi-Sperse W-2000, was chosen because it's a typical highly acidic (i.e., low-pH) industry grade. The same is true of the copper slurry selected for the test. The ICue 5003 slurry also from Cabot is, however, neutral and not as aggressive an abrasive as the other two slurries tested."
After a 24-hr run in the oxide or tungsten slurry, samples are weighed and measured to determine the amount of material loss through wear. Likewise, a 48-hr run was needed for the less-abrasive copper slurry due to its reduced rate of wear.
"To analyze the data the results from the three samples were normalized to the Techtron PPS control that is assigned a value of one. The lower the resulting normalized number, the less material that is abraded away and the better the polymer resists wear," says Campbell.
The comparable wear rates of previously used advanced engineering polymers and many new candidates were tested. According to Campbell, they included the industry-standard CMP polymer,-Techtron PPS, as well as unreinforced and carbon-fiber (CF) reinforced Ketron PEEK, Delrin, Celazole PBI, Duratron XP polyimide (PI), and polycarbonate (PC). The recently developed polymers tested were Semitron CMP LL5 polyester and Semitron CMP XL20 polyamide-imide. All the samples were evaluated under simulated oxide, tungsten, and copper CMP processing.
Copper slurries are typically mildly abrasive. During the CMP simulation test, most materials did not wear. Large potential for error was built into the data as a result.
Oxide CMP testing used the highly abrasive Semi-Sperse SS-12 slurry. According to Campbell, the wear-enhanced Semitron CMP XL20 polymer exhibits extremely good wear performance in each chemistry slurry, out pacing even the Techtron PPS.
The modified Semitron CMP LL5 grade, likewise, performed well offering the best wear performance of its POM, PBT, PET equivalents, says Campbell. "However, despite better wear simulation than Techtron PPS, tests have shown that in structurally unsupported designs typical POM, PBT, and PET materials don't perform on par based on performance and consistency and, therefore, are generally used in a composite-ring design."
Although the CF-reinforced grades have better wear performance than the Tectron PPS standard, says Campbell, these materials are prone to "slough off" or shed CF particles. These particles can result in microscratching the silicon-wafer surface. Conductive carbon particles deposited in the structure of a silicon wafer can also have disastrous effects on its functionality. The wear-enhanced PEEK and PPS grades, on the other hand, offer little or no advantage over their virgin grade counterparts.
LCPs and PC materials have poor wear performance, as did the modified PTFE which showed reduced wear results due to softness in its X-Y plane. And finally, PI and PBI materials also have excellent wear results, but are often regarded by fabs as being prohibitively expensive, says Campbell.
Tungsten CMP slurries such as the Semi-Sperse W-2000 are less abrasive than typical oxide slurries, but sport a much more acidic composition (pH: 1 to 3). "Of the test materials, PC was an extremely poor wear performer," says Campbell. "Its wear rate was 5.8 times that of the Techtron PPS standard." The Tungsten's strong acidic chemistry also aggressively attacked POM and polyester materials, says Campbell.
Semitron CMP LL5 is a new, lowercost alternative offering good wear resistance, says Campbell, and is especially suitable for CMP fabs that use lighter pressure and less-aggressive slurries.
The Semitron CMP XL20 material, on the other hand, outperformed most materials, says Campbell.
The Semitron CMP XL20 is the "magic bullet," says Campbell. "It sports the best combination of wear life and mechanical strength of any plastic tested in the CMP simulation study."