Evaluating plasma-etch resistance of high-performance plastics
04/01/2001
Richard W. Campbell, DSM Engineering Plastic Products, Reading, Pennsylvania
overview
Plasma etch is used throughout wafer processing, including initial oxide etch, various pre-cleaning steps, etching in various reactive chemistries, and removal of specific layers at several steps in the process. Most of these etch processes are designed to remove organic materials from wafers, and there is an attendant erosion of any other components in the plasma chamber. The longevity of these components is an important issue, whether they are disposable items, such as clamp rings, or more permanent components, such as screws, seals, and supports. In addition, the erosion of these components releases constituent trace metal impurities into the chamber environment. Thus, the rate of material attack and removal by a plasma and the content of trace impurities are critical factors in the selection of components for use in a plasma chamber environment.
Ceramic or plastic?
For components such as clamp rings, polyimide plastics and ceramics are the materials of choice. Ceramics are less costly and longer-lasting than polyimides, but they cost more to machine and are more brittle. While the erosion rates for ceramics are lower, there is concern about the release of sodium, aluminum, zirconium, and other inorganic elements from which ceramics are made.
 |
Unfortunately, comprehensive comparative data have not been available in the open literature for the various plastics, including polyimides, which are potential candidates for making wafer clamp rings.
Now, erosion rate data, as well as data about the by-products of erosion, are available for:
- polyamide-imide (PAI),
- polybenzimidazole (PBI),
- polyetheretherketone (PEEK),
- polyimides (PI), and
- polyphenylene sulfide (PPS).
These data pertain to specific grades of these heat-resistant polymers and are based on samples taken from the rod or plate typically used to machine wafer-clamp rings. The seven specific high-performance plastics tested are: Celazole PBI, Torlon 4203 PAI (with and without being post-cured after machining), Torlon 4503 PAI, Ketron PEEK 1000, Techtron PPS, Duratron XP PI, and Vespel SP-1 PI.
Test conditions
In this work, we wanted to ensure that the test conditions mimicked actual wafer processing etch conditions. To define the processing parameters and gas chemistries that would be most representative of those used in the field, we held very helpful discussions with several wafer-processing engineers representing a cross-section of the industry. Since wafer processors use somewhat different plasma conditions, and to protect proprietary information, a series of four plasma gas mixtures was chosen as being representative of those in use. These may or may not be exactly what a given semiconductor manufacturer uses.
We placed half-disk samples of the various plastics (~25mm in diameter and 3mm thick) in a Tek-Vac Model MPS-3000-LL. This is a multichamber, plasma-enhanced, chemical-vapor deposition system with a parallel plate capacitance plasma source. We used the following operating conditions:
- 100 mtorr working pressure,
- 200W RF power,
- 55°C maximum surface temperatures,
- 20mm plasma "showerhead" elevation,
- 15-20sccm total flow rate, and
- 100-hr total exposure time.
In addition, we used four gas combinations to represent functional etch cycles:
- 100% O2 pre-clean etch,
- 95% CF4-5% O2 silicon etch,
- 50% CHF3-25% HBr-12.5% O2-12.5% Cl2 polysilicon etch, and
- 75% Cl2-25% HBr main etch.
 |
Figure 1. a) Weight loss and b) thickness loss from O2 plasma etch.
Figure 2. a) Weight loss and b) thickness loss from 25% HBr-75% Cl2 plasma etch.
We arranged samples on the etch system's wafer chuck and exposed them to the plasma. Although there were cooling water channels in the chuck, we could not use this cooling because of the low thermal conductivity of the plastic samples. Instead, we monitored the surface temperatures of the samples with sensors and turned the plasma system on and off manually using the 50°C mark on the sensor as our indicator; we shut the system off until the surface cooled sufficiently to continue. At intervals of 25 hrs cumulative exposure, we wiped the samples with methanol and weighed them. They were also rotated around the sample holder at that time, to ensure randomness of exposure.
Typical exposure times in wafer fabs are on the order of 3-6 min/cycle. Therefore, the 100 hrs chosen for the total exposure represents the accumulation of 1000 to 2000 cycles. We believe that these conditions provided a fair determination of each material's performance.
We measured both thickness and weight loss as an indication of material erosion. None of the materials chosen for the tests were particularly hygroscopic, nor were they exposed to high humidity conditions, so there was good agreement between thickness and weight changes. Surface roughness was also determined, but did not prove to be quantitatively useful in comparing samples.
Test results
We recorded our data as weight and thickness changes from each plasma etch chemistry, for the various plastics tested (Figs. 1-4), except for one. Data for the two Torlon 4203 PAI samples (as machined and after post-machining re-cure) are not included in the data because the surface of the test samples became chalky during plasma etching, making the materials unsuitable for wafer fabrication applications. This surface residue was analyzed and determined to be titanium dioxide, an inorganic additive whitening agent (3%) in Torlon 4203 PAI; during plasma etching, the organic resin was etched away, leaving the inorganic pigment as a surface residue and potential contamination source. Torlon 4503 PAI is the same base resin as Torlon 4203 PAI without titanium dioxide.
Most weight and thickness losses from the different etch chemistries are linear, with two significant exceptions: Weight losses for the different materials during the main etch flattened out after about 50 hours.
Celazole PBI performed consistently better than other materials tested. It showed the least weight and thickness loss in the pre-clean etch, as well as the silicon etch. Further, it showed the second least weight and thickness loss in the polysilicon glass etch.
 |
Torlon 4503 PAI was better than most of the materials tested. It was the best in weight and thickness loss in three of the four chemistries: the main etch, the polysilicon glass etch, and the silicon etch. In the pre-clean etch, it was the second-best performing material.
Techtron PPS was readily attacked by nearly all plasma chemistries. Its best performance was in the main etch, where it was second best in thickness loss, even though it was the worst material in weight loss in the same system, suggesting loss of material from the interior, not just surface erosion.
The polyimides, which are widely used for making wafer clamp rings, did not perform as well as expected. Both PI materials included in the tests performed about equally in all four etch systems. Their best resistance to weight and thickness loss was in the main etch, where they were second to Torlon 4503 PAI.
 |
Figure 3. a) Weight loss and b) thickness loss from 50% CHF3-25% HBr-12.5%O2-12.5% Cl2 plasma etch.
Figure 4. a) Weight loss and b) thickness loss from 95% CF4-5% O2 plasma etch.
Ketron PEEK 1000, while not outstanding in any of the etch systems with regard to weight and thickness loss, performed quite well in all systems and may, therefore, be attractive from a cost-performance perspective. Its performance is equal to that of the PI currently in use, but at a much lower material cost.
What about impurities?
Other factors must be taken into account when choosing a material for a retaining ring or other plasma chamber components. One of the primary factors is the purity of the material. Since all polymeric materials are etched to varying degrees, it must be assumed that any trace metals in them will be released into the chamber environment during use. Therefore, it is critical to ensure that the material is as contamination-free as possible.
Two types of purity data are often reported: leaching or extraction and total digestion.
Leaching data are useful in environments where the component is exposed to liquids, such as tanks, valves and tubing, but since only the soluble compounds near the surfaces are extracted, this may underestimate the impurities pertinent to the etch chamber situation.
 |
Analysis by total digestion is accomplished by completely dissolving or ashing a sample and analyzing the total composition for trace elements using mass spectroscopy inductive-coupled plasma (MS/ICP).
The following ionic impurity data are for specific trade name plastic products and were determined by the total-digestion method. All data were generated from stock shapes used to machine semiconductor manufacturing equipment components.
Outgassing
As shown in Table 2, several high performance plastics have very low outgassing (tested in accordance with ASTM E595). Ideally, total material loss (TML), after 24 hrs at 125°C and a vacuum of 10-5mm Hg, is <<% and collected volatile condensable material (CVCM) should be 0.00% to ensure that no unwanted deposition occurs in the chamber.
Other properties
When choosing a plastic material for use with wafer processing applications, additional important factors include the spectrum of mechanical and thermal properties, availability in efficient sizes, and machinability. Extensive materials properties can be determined from manufacturers' published data sheets and fabrication experience. Typical data for the materials tested for plasma etch resistance are given in Table 3.
Conclusion
Polyimides are widely used for making wafer clamp rings in plasma etch systems, and were assumed, going into the tests reported here, to have the best performance. However, test data do not necessarily support this assumption. Several other materials were equivalent to, or better than, the incumbent polyimide in their resistance to the attack of the plasma and should be considered based on performance, cost, and availability. Ionic impurity and low outgassing are also critical to the choice; for example, Celazole PBI is an excellent candidate, but Torlon PAI and Ketron PEEK should be considered for their attractive cost/performance profiles.
Acknowledgments
The author thanks Tek-Vac Industries Inc., S. Brentwood, NY, for in-chamber plasma etch exposures.
Celazole is a registered trademark of Celanese Acetate Inc. Duratron polyimide, Ketron PEEK, and Techtron PPS are registered trademarks of DSM Engineering Plastic Products Inc. Torlon polyamide is a registered trademark of BP Amoco Performance Products Inc. Vespel polyimide is a registered trademark of DuPont.
Richard W. Campbell received his PhD in polymer science and engineering from the University of Massachusetts. He is manager of product development at DSM Engineering Plastic Products, 2120 Fairmont Ave., Reading, PA 19612; ph 610/320-6600, fax 800/366-0301, e-mail [email protected].
Selecting a material for making components, such as wafer clamp-rings, which are exposed to plasma etch processing requires information on their erosion rates and the metal impurities that they can release. Here, for the first time, are comparative data for plastics generally considered for such components.