Wafer cleaning is an essential step in the manufacturing of integrated circuits. The two main contaminants encountered in the fabrication process are particles and organic residues. These contaminants arise in photoresist stripping operations and must be removed from the surface. Vacuum plasmas are routinely used for photoresist stripping, descum of resist residues, and general wafer cleaning. After the plasma process, the wafer is subjected to a megasonic or ultrasonic clean with deionized water to remove particles. This step is necessary because vacuum plasmas generate particles inside the chamber by ion bombardment. In other words, if you use a vacuum plasma to strip resist, descum, or otherwise remove contamination from a surface, you are going to need a second step to remove particles generated in that process.
Now there is a new way to clean silicon wafers with a plasma that does not require a second particle removal step. The Atomflo™ atmospheric pressure (AP) plasma removes organic contamination in seconds without adding particles. Shown in Table 1 is a comparison of cleaning 200mm Si wafers with a vacuum plasma and the Atomflo™ AP plasma. Before cleaning, the native oxide on the silicon is covered with a layer of adsorbed organic compounds that passivates the surface, yielding a surface free energy of 64 mN/m and a water contact angle of 38o. After 2 minutes in the vacuum plasma, the wafer is cleaned and activated with a surface free energy of 77 mN/m and a water contact angle below 7o. Identical results are achieved by scanning over the wafer with the Surfx 100-mm-wide plasma source at 200 mm/s. The plasma was operated at 500 W with 0.8% oxygen in argon. In this case, the total process time is 2.5 seconds. The AP plasma is 48 times faster than the vacuum plasma!
Table 1. Comparison of vacuum and Atomflo™ plasma for cleaning 200mm Si wafers.
After cleaning the wafers, a laser scanner was used to determine the distribution of particles on the surface between 0.2 and 1.0 micron in diameter. These results are presented in Table 2. Exposing the wafer to the environment in the class 100 cleanroom led to the accumulation of 17 particles on the surface, with 8 of these in the 0.8 to 1.0 micron range. Vacuum plasma treatment caused approximately 2,800 particles to be deposited on the wafer, with over 1,100 of these in the 0.8 to 1.0 micron range. A secondary water cleaning is definitely required to remove these. By contrast, Surfx’s AP plasma only deposited 8 more particles than the control. Most likely the few additional particles came from the scanning motion of the robot over the wafer. In conclusion, by switching from vacuum plasma to Surfx plasma treatment, you can eliminate a wet cleaning step from your IC fabrication process.
Table 2. Measurement of the particle size distribution on 200mm Si wafers before and after plasma treatment.
The reason the vacuum plasma deposits particles on the wafer and the Surfx argon plasma does not has to do with the physics involved. Radio frequency powered, capacitive discharge plasmas build up a sheath next to the wall in which there is a large voltage drop that repels the free electrons, see Figure 1. This is necessary to maintain the plasma, because the electrons move so much faster than the positively charged ions, and would instantly drain out without the potential barrier. If the ions do not encounter any molecules in the sheath, then they will be accelerated by the voltage potential and hit the wall with enough energy to sputter away material. This ion bombardment is the source of particles in the vacuum plasma. For operation at 0.2 Torr, the sheath thickness, S, is approximately 250 microns, while the mean free path between collisions, λ, is 60 microns. Only a few collisions occur as the ion travels across the sheath, leaving it with sufficient energy to sputter material off the walls. By contrast, at atmospheric pressure, 760 Torr, S is on the order of 20 microns, and λ is 0.06 microns, so that there are many hundreds of collisions before the ion reaches the wall. Consequently, there is no sputtering with particle generation in the Atomflo™ atmospheric pressure plasma.
Figure 1. Schematic of the potential, ϕ, as a function of the distance, 2d, between the electrodes. The sheath is the region near the wall where the potential rapidly decreases to ϕW (Chen and Chang, “Lecture Notes on Principles of Plasma Processing,” (Kluwer/Plenum, N.Y., 2003)).
Dr. Robert F. Hicks
CEO & President
Surfx Technologies LLC