Fowler Associates Labs

Clean Corona Ionization

Can clean ionization be achieved by using air or nitrogen to minimize particle buildup on emitter points? Here's a look at IBM's research in this area.

Kenneth D. Murray, Vaughn P. Gross- IBM General Technology Div., Essex Junction, VT 05452

Philip C.D. Hobbs- IBM Research Div., Thomas J. Watson Research Center, Yorktown Heights, NY 10598

In an attempt to achieve clean ionization and control particle generation, a standard bipolar DC ionizer was modified so it would supply filtered, dry compressed air or nitrogen to the emitter points via a perforated hose. This technique was designed to restrict the ammonium-nitrate buildup on the emitter points and also restrict the formation of ultra-fine aerosol particles (a product of high-voltage corona discharge) to measurable levels at or near background.

Ionization and Particle Generation

Corona-point air ionizers are widely used in clean-room environments to minimize ESD problems by neutralizing static charges and electrostatic fields (1-4). However, ionizers themselves can generate minute particles (5-8).

Attempts have been made to identify the particles generated by steady-state DC (SSDC) ionization. In 1989, the particles were tentatively associated with ammonium nitrate
(); a material found to precipitate onto the emitter points over time (4,7). Micro Fourier Transform Infrared (MFTIR) and Raman microprobe (RAMP) showed the material to have major spectral features and Raman shifts indicative of , respectively (4). There were not enough aerosol particles collected to determine conclusively, by MFTIR or RAMP, the particles' primary chemical compound. However, energy dispersive x-ray results showed similar elemental associations between the particles captured and the collected from the points (7).

If the particles are , then their formation would require a source of hydrogen, with the most plausible one being atmospheric water vapor. According to test results, particle production can be consistently restricted with dry, compressed air or nitrogen-immersed corona points. This technique can also restrict buildup of the points. Particle generation has also been shown to be a function of humidity.

Test Setup and Procedures

Two SSDC ionizers were modified as shown in Figure 1. A plastic covering, or faceplate, was attached to half the length of each unit. Behind the faceplates, gas bleeder lines were used to flood the inside cavities of each ionizer with reasonably dry, filtered gas. The holes cut in the faceplates, in front of each point, were 1 cm in diameter.

The ionizers were placed 15 cm below the HEPA filters in a class 10, vertical laminar flow (approx. 100 cfm). Four condensation nucleus counters (CNCs) were used to monitor the particle concentrations. These were connected to a multiplexer processor and a computer. One meter of anti-static sample tubing was attached to each CNC. Air samples were taken 8 cm from the negative emitter points, and background counts were taken 3 cm below the HEPAs and above the influence of ionization.

The effectiveness of the modified versus unmodified ionization was evaluated 1 m below the points with a charge plate monitor. The 0.05-micron filtered nitrogen had a water content o f0.1 parts per million (ppm). The 0.05-micron filtered compressed air water content was 1.0 PPM The percent relative humidity (%RH) inside the clean room lab was adjustable. As shown in Figure 1, the setup included particle monitoring under the control, the nitrogen-immersed, the room-air-immersed, and the compressed-air-immersed points. The monitors remained fixed in these positions for the duration of the experiment.

The computer scanned every 10 seconds the average number of particles per cubic foot (diameter >0.01 micron) and recorded this information to the floppy disk every 10 minutes. The particle counts were recorded continuously in both high and low %RH for 30 days. After 30 days the emitter points were examined with a scanning electron microscope.

Particle Concentrations

In Table 1, voltage decay rates are shown in seconds under the modified vs. unmodified ionizations. The gas-treated points took about 3 seconds longer to decay 5000V to zero (averaging 14.4 seconds).

Tables 2 and 3 show typical particulate background concentrations established at the beginning of the experiment. Of particular interest is the average value of the compressed-air concentration over time. The compressed air high-low range in Table 3 is noticeably similar to typical concentrations recorded in Tables 4 and 5. If the control range in Table 3 is subtracted from Tables 4 and 5, the remainder for the "Comp. Air Point" would be at or near zero. For this case, the results may be due to particles from the in-line filter. The same may be true on the nitrogen side. To address this concern, an improved filter system or an extended breaking-in period would be needed.

Relative Humidity

Tables 4 and 5 also reflect a 3X increase (approximately 10%) in range for relative humidity as shown by the "Air/RH Point."

The typical data averaged for Tables 4 and 5 are plotted in Figures 2 and 3. During the month-long data collection period, there was a distinct behavior of the particle presence (See Figs. 2 and 3). The particle presence under the room-air-immersed point appeared in bursts with the lengths of time between low and high concentrations varying, regardless of %RH. The particle concentration under the room-air-immersed point consistently increased as predicted when the %RH increased (See Figs. 2 and 3). Also shown, when moisture from the immediate vicinity of the corona discharge only was eliminated, the humidity-dependent particle production is restricted. This supports the theory that the particles are associated with .

Particle Buildup

After 30 days of operation, the precipitate could easily be seen on the room-air-immersed emitter points. Scanning electron micrographs (SEMs) of the points were taken at 100 power (Photos 1-9). The SEMs show that all of the positive points had buildup to some degree. For the positive points the buildup was dendritic. The room-air-immersed negative points had deposits which were nodular in appearance. In comparison, the overall buildup was much less pronounced on the dry gas-treated points.

Faceplate Collars

A further reduction of the buildup of may be achieved by the attachment of collars
(3 mm in length) around the faceplate openings cut in front of each point (Figure 4). These collars or sleeves would prevent room air (with its moisture content) from being entrained by turbulence to the vicinity of the corona. The differences in the dendrite growth, for example, on the nitrogen versus compressed air points in additions to the different particle counts found under the two gasses are largely due to this room air entrainment.

As indicated in Table 6, the use of faceplate collars eliminated the ion/electron induced chemical reactions which lead to the aerosol formation and underscored the significant influence that plays in the creation of these ultrafine particles.

The mechanisms of aerosol particle formation under corona discharge treatment of combustion flue gasses and the important role that plays on the formation was recently discussed in detail by Chang (9).

Material Loss

Finally, the positive room-air-immersed point shows the typical tungsten material loss, which is consistent with our earlier findings (7). The gas-treated positive points show no such material loss. The loss of positive electrode material appears to be prevented by isolation of the corona points using the method described here.

Conclusions

The research and testing on a modified SSDC ionizer described here shows that:
* The production of aerosols by high-voltage corona discharge can be reduced or eliminated;
* The buildup on the corona points can be avoided or significantly restricted;
* The erosion of the positive electrodes can be prevented;
* The potential problems of metallic and organic contamination to clean room environments can be prevented;
* The ammonium nitrate buildup on the corona points and the aerosol particle presence are highly humidity dependent;
* The data presented demonstrates for the first time the effectiveness of a new method for truly clean corona ionization.

Acknowledgments

The authors appreciate the efforts by these individuals of IBM Essex Junction: S.J. Pierce (MFTIR, RAMP) and R.B. Dunn (SEMs).

References

1. Dillenbeck, K., "Selection of Air Ionization Within the Clean Room," Proceedings of the 32 Annual Technical Meeting of the IES, 1986.
2. Turner, T., "Static in a Wafer Fabrication Facility; Causes and Solutions," Semiconductor International, 1983.
3. Huffman, T.R.; Nicholas, G.; and Bossard, P.R., "Room Ionization: Can it Significantly Reduce Particle Contamination?", Ninth Annual EOS/ESD Symposium, 1987.
4. Murray, K.D.; Ainsworth, G.F.; and Gross, V.P., "Hood Ionization in Semiconductor Wafer Processing: An Evaluation," 10th Annual EOS/ESD Symposium, 1988.
5. Liu, B.Y.H.; Pui, D.Y.H.; Kinstley, W.O.; and Fisher, W.G., "Aerosol Charging and Neutralization and Electrostatic Discharge in Clean Rooms," J. Envir. Sci., March/ April 1987.
6. Donovan, R.P.; Clayton, A.C.; and Ensor, D.S., "The Dependence of Particle Deposition Velocity on Surface Potential," Proceedings of the Inst. Envir. Sci., 1987.
7. Murray, K.D.; and Gross, V.P., "Ozone and Small Particle Production by Steady State DC Hood Ionization: An Evaluation," 11th Annual EOS/ESD Symposium, 1989.
8. Suzuki, M.; Matsuhashi, H.; and Izumoto, Y., "Effectiveness of Air Ionization Systems in Clean Rooms," 34th Annual Technical Meeting of the IES, 1988.
9. Chang, J.S., "The Role of and on the Formation of Aerosol Particles and De-NOx Under the Corona Discharge Treatment of Combustion Flue Gases," J. Aerosol Sci., Vol. 20, No. 8, pp..1087-1090,1986.