RICHARD W. CAMPBELL
Quadrant Engineering Plastic Products
Reading, Pa.
KING T. WILSON
Tek-Vac Industries Inc.
Brentwood, N.Y.
|
Silicon wafers aren't the only things that lose material in the various etching steps of IC manufacturing. The clamp rings which hold the wafers also erode during the process.
Plastic clamp rings are disposable items. Still, it would help reduce manufacturing costs if the rings had a longer life. And rings that erode more slowly during plasma etching would release fewer trace metal impurities into the chamber environment.
With these factors in mind, a few tests have shed light on the rate of material attack and removal by plasma and the content of trace impurities. Tests took place on plastics typically machined into wafer clamp rings, as well as on a few used less frequently.
The assumption going into these evaluations was that polyimides would perform best since they are widely used for making wafer clamp rings. But test data do not support this idea. Results show that several other materials resist attack from plasma etch either as well as or better than polyimide. They merit consideration based on performance, cost, and availability. The importance of ionic impurity and low outgassing also make materials that include Ketron XP PEEK (from Quadrant Engineering Plastic Products), Torlon (BP Amoco Performance Products Inc.), and Celazole PBI (Celanese Acetate Inc.) excellent candidates.
Ceramics are also materials of choice for clamp rings. They are less costly and last longer than high-performance plastics, but are more brittle and expensive to machine. Also, ceramic rings can release traces of sodium, aluminum, zirconium, and other inorganic elements.
Etching environments
It is useful to consider the environmental factors that can arise during wafer etching. All wafer processors use plasma conditions that are each somewhat different and proprietary. But discussions with wafer-processing engineers reveal that there are four plasma-gas mixtures representative of those in commercial use. They are 100% O2 (typical of preclean etch), 95% CF4/5% O2 (silicon etch), 50% CHF3/25% HBr/12.5% O2/12.5% Cl2 (polysilicon glass etch), and 75% Cl2/25% HBr (main etch).
The test samples seeing these conditions were 25-mm-diameter half disks that were 3-mm thick. They sat on the wafer holder in a Tek-Vac Model MPS-3000-LL. This is a multichamber, plasma-enhanced, chemical-vapor deposition unit. A robot insertion and extraction system handled the samples, and the plasma source was in a parallel-plate capacitance configuration. Important settings during the tests included working pressure (100 mtorr), RF power (200 W), approximate total flow rates (15 to 20 sccm), surface temperatures (55?C max), showerhead elevation (20 mm), and total exposure time (100 hr).
Despite the existence of channels for cooling water running through the platform, surface temperature of the sample disks could exceed control settings because of the plastic's low thermal conductivity. So technicians checked the actual surface temperatures with sensors, then shut the system off and let things cool when readings hit 55?C. After every 25 hr of cumulative testing, the samples were wiped with methanol and weighed. They also rotated to a different position in the sample holder to ensure consistent exposure.
Fabs use exposure times that are typically on the order of 3 to 6 min/cycle. Thus 100 hr of total exposure represents the accumulation of 1,000 to 2,000 cycles. It also seems to portray real-world conditions well enough for ranking purposes.
Both thickness and weight loss were measured as an indication of material erosion. No materials tested were particularly hygroscopic, nor did they see high humidity, so there was good agreement between changes in thickness and weight. Test personnel also checked surface roughness, but the results weren't useful in quantitatively comparing samples.
In general, all the plastics lost weight and thickness at a more or less constant rate in the different etch systems.
Celazole PBI performed consistently better than other samples. It showed the least weight and thickness loss in the preclean etch (100% O2), as well as in the silicon etch system (95% CF4/5% O2). Further, it showed the second least weight and thickness loss in the polysilicon glass etch system (50% CHF3/25% HBr/12.5% O2/12.5% Cl2).
Torlon 4503 PAI was better than most plastics tested. It lost the least weight and thickness in all but the preclean etch (100% O2). There, it came in second.
However, samples of Torlon 4203 PAI were eliminated from the test. It turned out their surface became chalky during plasma etching, making the material unsuitable for clean-room use. Analysis of the surface residue revealed it to be titanium dioxide, an inorganic additive used as a whitening agent at a level of 3% in Torlon 4203. The organic resin was etched away leaving the inorganic pigment behind as a surface residue and potential source of contamination. (Torlon 4503 is formed by compression molding the same base resin as 4203, but without the titanium dioxide.)
Polyimides, which are widely used for making wafer clamp rings, did not perform as well as might be expected in that they did no better than other materials. Both PIs in the test did about the same in all four etch systems. They lost the least weight and thickness in the main etch test where they were second to Torlon 4503 PAI.
Techtron PPS (from Quadrant Engineering Plastic Products Inc.) was readily attacked by nearly all plasma chemistries. Though its resistance to erosion wasn't outstanding, Ketron PEEK performed quite well in all systems. Thus it may be attractive from the standpoint of cost/performance.
Impurities and outgassing
Purity and outgassing qualities are also important when deciding on retaining ring materials. Any trace metals get released into the chamber environment during plasma etching. To quantify this behavior, samples of the various plastics were first completely dissolved or ashed. The total composition then underwent analysis for trace elements by inductively coupled plasma/mass spectrometry.
Finally, tests reveal that several highperformance plastics outgas very little when tested to ASTM E595. No collected volatile condensable material (CVCM) was found after 24 hr at 125?C and a vacuum of 10-5-mm Hg. Water vapor recovery (WVR) reflects the hygroscopic nature of the material; a large number indicates that water makes up a significant amount of the total material loss value.