SI CHINA     SI JAPAN
Login  |  Register          Free Newsletter Subscription
Subscribe
Email
Print
Reprint
Learn RSS

Vacuum Pump Change Improves PECVD Process Uptime

Stephen Munley, National Semiconductor Corp., Greenock, Scotland
Ian McNaught, Dan Mrotek and C.Y. Lin, Leybold Semiconductor Vacuum Solutions, Export, Pa. -- Semiconductor International, 10/1/2001

At a Glance
To solve vacuum pump downtime issues associated with harsh PECVD processes, a new screw pump was installed. Downtime was reduced and the conditions of the vacuum pump exhaust were improved.

The Novellus Concept-One tool is used to deposit nitride and doped oxide layers. A requisite process step also includes PFC plasma treatment to clean the process chambers. Figure 1 describes the deposition reactions of silicon nitride (Si3N4) and P2O5-doped SiO2 (PSG) by plasma-enhanced chemical vapor deposition (PECVD). The deposition of P2O5 and SiO2 for PSG occurs simultaneously.

One concern is the formation of excess particles such as SiO2, Si3 N4 and P2O5 that will be either transported to pumps or deposited in the process chambers. Depending on pumping mechanisms, the particles can easily accumulate inside pumps and create more PM (preventive maintenance) and unexpected downtime. PFC plasma cleaning is needed on a regular basis to remove the particles deposited in the chamber. The cleaning process of SiO2, Si3N4 and P2O5 using PFC plasma is summarized in Figure 2.

Particulate generation

The major contributor to downtime of vacuum drypumps is the large amount of particulate that must be pumped through the system. The Table lists typical amounts of solid particles generated.

Particulate Generation From Processes
Process Solid byproducts (grams/min)
SiO2 P2O5 Si3N4 Total
PSG oxide 1 0.4 0.6 1.0
PSG oxide 2 0.4 1.2 1.6
Nitride 2.5 2.5

This means that the vacuum pump and associated piping system must transport 960-2400 g/day of powdery particulate with an assumption of ~16 hr/day of run time. The N2 purging system at the exhaust line helps efficient transportation of process particles. The pumping mechanism and system design are the other critical factors for preventing generation and efficient removal of particles.

Another major concern for pumps is the generation of ammonium complexes such as ammonium fluorosilicate as a byproduct from the cleaning process. In the cleaning step, the C2F6 gas combined with plasma releases various radicals as described in Figure 3. The generated radicals will react with solid particulates of SiO2, Si3N4, and P2O5 to form gas species described above.

1. How silicon nitride and P2O5 -doped silicon oxide films are deposited by PECVD.

Excess NH3 and H2 entrapped in the Si3N4 and SiO2 powders also react with plasma-generated species. Once the complex ammonia fluorosilicates start to deposit in the exhaust of the pump, it is difficult to remove them by nitrogen flush purge, and exhaust passages will become restricted. Small amounts of other ammonium salts can also be formed through process and cleaning steps. Process powders can then be more easily entrapped in the fluorosilicates, and rapid clogging of pump exhaust will occur. One way to prevent or minimize these byproducts is to heat the exhaust line.

2. PFC plasma cleaning is needed on a regular basis for removing the particles deposited in the chamber

If an excess of N2O is used in the oxide deposition step, H2O is another possible byproduct that can make the powder-transport mechanism complicated in the pump and associated pipes.

When water is adsorbed on the powder, the powder can become a slurry-like or sticky material and be collected in the pump passages. Water can hydrolyze the SiO2 to form a gelatin-type material that could also clog the pump. Phosphoric acid (H3 PO4) and its conjugated base forms (H2 PO4 ¯ and
HPO4 2-) can be created when P2O5 is exposed to water. These phosphoric acid derivatives are corrosive and sticky and can be deposited along the gas passages. Further ammonium salts of these derivatives also can be formed. Again, depending on the temperature of the exhaust lines, these salts can be minimized. Other chemical reactions forming solid materials based on available chemicals are also possible.

3. During the cleaning step, the C2F6 gas combined with plasma releases radicals that react with solid particulates of SiO2, Si3N4 and P2O5 to form various gas species.

System modifications

The formation of these particulates is more complicated than the possibilities described above. Some of the controlling parameters are related to the reaction possibility and thermal stability of the particles along the gas passage (from the pump to the exhaust). The reaction possibility will be a function of length, duration and temperature of gas passages. Further details of each parameter may not be necessary here.

4. The exhaust line above was almost totally clogged after only two weeks of operation. On the revamped system with the new screw pump and particle trap, the exhaust line remained clear even after six weeks of use.

Based on the study of the compounds and the processes done by National Semiconductor and the pump manufacturer, temperature can be relatively easily controlled at the given conditions to reduce the exhaust pressure increase. Higher temperature means increasing vapor pressure of the materials even though it generally increases reaction rates. However, the formation of solid ammonium-silicate salts can be minimized by a proper control of exhaust temperature. The effect of temperature control on particle reduction was significant, as expected. The careful management of temperature in the pump mechanism and associated piping also keeps water and other ammonium salts in gas phase. It may keep process powders dry, help effective transportation of powders out of the pump, and will eventually reduce the risk of the exhaust line being significantly clogged.

Pump selection

There are several drypump mechanisms on the market today for harsh semiconductor processes. They include multistage roots; roots/hook & claw; hook & claw; and screw. National studied these mechanisms and found several key factors in determining their particle handling abilities:

  • Gas (particle) path length and complexity.
  • Temperature excursions during compression.
  • Gas compression method.

The ideal pump mechanism should have a short gas path length, smooth temperature transition from inlet to exhaust, and minimum gas direction changes from vacuum to atmosphere. Screw pumps were chosen because they have gas paths that can be 5× shorter than other types. Screw mechanisms have been used successfully for decades as particle movement mechanisms in other industries. Because of their design, screw pumps avoid big temperature and gas velocity changes that can cause particles to quickly become trapped in a complex pumping mechanism, and cause the pump to seize due to particle entrapment. For this reason, the screw pump was deemed the likeliest candidate for success.

Addition of a particle trap

After careful examination of the entire system, it was determined that downtime was occurring because of two main issues: 1) premature pump failure due to clogging of the pumping mechanism, and 2) clogging of the exhaust line with process and/or ammonium-silicate particles.

Every two weeks, the piping between the vacuum pump and the Centrotherm abatement device had to be disassembled, cleaned and reassembled. Based on the evaluation of pump mechanisms by National, it was expected that the new pump would solve the pump clogging issue, and that the amount of powder being transported to the exhaust would increase due to the new pump's capability to transport more powder. In addition, the pump manufacturer felt that a modification must be made to the exhaust system to ensure increased uptime.

The new process drypump with an exhaust particle trap was installed. This was designed to handle the expected increase of powder transport due to the more efficient pumping mechanism, and to reduce the frequency of exhaust cleans. It was anticipated that, although the particle trap would need maintenance, this would be simpler and quicker than a full exhaust clean.

The new system with the screw pump of split-flow pumping mechanism was optimized and evaluated against the original system. After three weeks of trouble-free operation, a scheduled exhaust clean was due. Instead of carrying out the usual strip-down — clean and reassembly that could take 12-18 hr and involve the handling of possibly toxic compounds — a simple inspection was carried out. The exhaust line was opened at the particle trap and the abatement system, and found to be clean. This inspection was repeated every three weeks.

Twelve weeks after installation, the drypump monitoring system indicated an increase of back pressure at the exhaust. The exhaust line was opened for inspection, and restrictions due to particulate formation were observed. Using the system with the DuraDry pump, the exhaust PM was extended to 12 weeks from every two weeks. Previously, the exhaust plumbing required full strip-down and rebuild every two weeks. In Figure 4, the relatively clean exhaust line of the revamped system after six weeks (bottom) is compared to that of another system after only two weeks of operation (top).

When the project began, it was expected that this particulate trap would require weekly maintenance. However, because the trap maintenance was simple and quick to perform, it was not expected to increase downtime significantly. In conclusion, the DuraDry pump performed as expected with zero downtime and effectively moved the high powder load into the trap. The powder trap also proved very effective in condensing and collecting particulate, thus minimizing facility exhaust maintenance. Pump downtime was eliminated and the trap exchange interval significantly exceeded expectations by having a lifetime of more than six months. For additional information, see http://homepage.ntlworld.com/stephen.munley.


Author Information
Stephen Munley is staff equipment engineer in deposition at National Semiconductor. He has worked as an equipment engineer specializing in epi, CVD and diffusion processes since 1984, and has been employed by the company's Greenock, Scotland, manufacturing site since 1979. He has an HNC in electrical and electronic engineering.
Ian McNaught is Leybold's account manager for Europe North, managing the team supporting the semiconductor end-user business. He joined the company in March 1999 as sales engineer, and has been employed in equipment support roles in the semiconductor industry since 1989. He has an HND in electrical and electronic engineering.
Dan Mrotek is Leybold's worldwide forevacuum applications marketing manager. He has worked at Leybold for >20 years in various engineering and marketing roles, focusing primarily on application-specific solutions for semiconductor customers. He has a B.S.M.E.
C.Y. Lin is senior application specialist at Leybold. He previously was employed at Westinghouse R&D in Pittsburgh for more than 30 years as a principal research scientist in all phases of vacuum deposition research. He has a Ph.D. in physical organic inorganic chemistry.


Email
Print
Reprint
Learn RSS

Talkback

We would love your feedback!

Post a comment

» VIEW ALL TALKBACK THREADS

Related Content

Related Content

 

By This Author

There are no other articles written by this author.

SPONSORED LINKS



 
Advertisement
SPONSORED LINKS

More Content

  • Blogs
  • Podcasts
  • Videos

Blogs

Podcasts

Videos

Advertisements





NEWSLETTERS
Plug in and get the latest SI news, trends and industry updates delivered free, directly to your inbox!

SI NewsBreak and Special Reports (Weekdays)
Wafer Processing Report (Monthly)
Lithography Report (Monthly)
Metrology Report (Monthly)
Clean Processing Report (Monthly)
Packaging Report (Twice Monthly)
©2008 Reed Business Information, a division of Reed Elsevier Inc. All rights reserved.
Use of this Web site is subject to its Terms of Use | Privacy Policy
Please visit these other Reed Business sites