EXTREME SILICON

Investigating ionizer use in Class 1 and Class 100 cleanrooms

Michael G. Harrison, formerly with MEMC Electronic Materials

A series of investigations were initiated to evaluate the costs and benefits of using ionization systems in Class 1 and Class 100 cleanrooms and to determine what impact ionization use has on airborne particle attraction to silicon wafers.

As submicron particles moving through the HEPA and ULPA filters in cleanroom and minienvironment air supplies rub against the filter media, they generate a triboelectric surface charge. These charged particles are then attracted to surfaces with the opposite charge, which can quickly cause process equipment, wafer carriers, and wafers to accumulate a high surface charge. If the affected surfaces are nonconductive or not grounded, or both, these electrostatic forces will build up and the surfaces' attractiveness to airborne particles will increase.

Ion emitters are typically used to neutralize triboelectrically charged particles and cause them to cascade harmlessly with the airflow over wafers and tools without being attracted to these critical surfaces. Ion emitters can also dissipate any charges that do build up on process equipment, wafer carriers, or wafers. This article describes a series of investigations that were initiated to evaluate the costs and benefits of using ionization systems in Class 1 and Class 100 cleanrooms and to determine what impact ionization use has on airborne particle attraction to silicon wafers. The investigations were aimed specifically at evaluating preventive maintenance requirements, effects on device yield, and potential cost savings for new construction.

The manufacturing cleanrooms studied contain a variety of emitters in either the ceiling grids or minienvironment hoods. The ion emitter manufacturer recommends that the emitter tips be cleaned four times yearly and balanced twice yearly. However, no data are available from the manufacturer or in the literature on discharge-time variation or changes in emitter voltage swings as a function of time. Therefore, in the first set of tests, voltage and discharge time data were collected just prior to and just after regularly scheduled emitter cleaning and balancing and over a period of months thereafter to monitor emitter performance as a function of time. Additional tests were performed to evaluate particle attraction to witness wafers when the ion emitters were purposely set out of balance and when the surface charge on the witness wafers was in excess of ±1000 V. Finally, the cleanrooms were audited with a noncontact field meter to determine electrostatic charge levels on conductive and nonconductive surfaces. These results were used as the basis for corrective action plans that included installing additional ion emitters, grounding process tools, replacing nonconductive insulating materials with conductive materials, and, in some cases, doing nothing at all.

Experimental Procedures

Collecting Voltage and Discharge-Time Data. In the initial tests, a Model 210 charge-plate monitor (Ion Systems; Berkeley, CA) was used to collect preliminary data for 10 ceiling ion emitters located in a Class 100 gowning area and 10 ceiling ion emitters in a Class 1 cleanroom. Discharge-time measurements were taken after applying a positive or negative 1000-V charge to the monitor. The first measurements of voltage swings and discharge-time variation occurred just prior to the manufacturer's cleaning and balancing service. Six months had passed since the emitters in the Class 1 area were last serviced, and eight months had elapsed since servicing in the Class 100 area. To determine what benefits can be achieved from performing this preventive maintenance, a second set of measurements was taken after the ion emitter tips had been cleaned with isopropyl alcohol swabs to remove contamination buildup. A third set of measurements was then performed after the ion emitters had been balanced by adjusting the positive and negative knobs. These readings represented the baseline data for future testing. To investigate ion emitter performance as a function of time, voltage swing and discharge-time measurements for the same 20 emitters were taken monthly for the first four months after they had been cleaned and balanced.

Emitter Unbalance Tests. These three series of tests focused on determining what changes in particle attraction to witness wafers would occur if ion emitters became severely out of balance as a result of equipment malfunction or personnel neglect. During testing, an ion emitter was set to release all positive (+400-V) or all negative (—400-V) ions, or was turned off. Both 10- and 100-minute tests were performed to determine if the length of exposure to these out-of-balance conditions had any influence on particle attraction. Approximately 200—250 >0.1-µm particles, 150—200 >0.2-µm particles, 100—150 >0.3-µm particles, 50—75 >0.5-µm particles, 20—40 >1.0-µm particles, and 1—10 >5.0-µm particles were generated directly over the witness wafer by rubbing cleanroom gloves and cleanroom garments in the airstream. An airborne particle counter was positioned directly beside the wafer to ensure that the contamination level was consistent from test to test. For the 10-minute test, particles were generated the entire time. For the 100-minute test, particles were generated during the first and last 10 minutes (0—10 and 90—100 minutes). Particle adders on the witness wafers were then counted using a standard electronic inspection tool.

Excessive Voltage Tests. These series of tests focused on evaluating particle attraction to witness wafers when the surface charge on the wafers was in excess of ±1000 V. Various levels of surface charge were obtained by holding the wafers at different distances from the emitter tip. At a distance of ~1 in. from the tip, ±6000 V was achieved; at a distance of ~5 in., ±3000 V was achieved; and at a distance of ~20 in., ±1500 V was achieved. The ionization to the room was turned off as soon as the desired surface charge was obtained to prevent the charge from being dissipated. Consistent particle levels were generated and particle adders were measured as in the emitter unbalance tests.

Field Meter Audit. A Hand-E-Stat electrostatic field meter (Simco Static Control and Cleanroom Products; Hatfield, PA) was used to provide noncontact surface voltage readings from conductive and nonconductive equipment and materials in the cleanrooms. Among the items tested were walls, floors, and shelves; stationary and moving equipment and tools; and plastic cassettes and packages.

Results and Discussion

Effects of Cleaning and Balancing Emitters. The preliminary data in Figure 1 show that the emitters in the Class 100 gowning area were experiencing large voltage swings just prior to preventive maintenance. The average positive voltage was 379 compared with the average negative voltage of 88, for an average change increment (or delta) of 290. In addition, the discharge time averaged 16 seconds for a positively charged surface and 9 seconds for a negatively charged surface, for an average delta of 7, as shown in Figure 2. These data indicate that neglecting ion emitters for more than six months in a Class 100 area may cause significant performance problems. The manufacturer recommends cleaning every three months and balancing every six months, but servicing in the Class 100 area had not been performed for eight months. Before maintenance was conducted, the emitter tips were coated with visible, white, silicon-based, organic and inorganic contaminants, which undoubtedly contributed to their performance variability. In a gowning area, the effect of out-of-balance emitters on actual wafer yield or on the quantity of particles being brought into an adjacent cleanroom on operators' uniforms was not known at the time of the study.

Figure 1: Voltage swings of ion emitters before and after preventive maintenance.

Figure 2: Discharge-time variation of ion emitters before and after preventive maintenance.

In contrast to emitters in the Class 100 area, those in the Class 1 area did not exhibit severe voltage swings or discharge-time variation before cleaning and balancing. The average positive voltage was 148 and the average negative voltage was 159, while the discharge time averaged 10 seconds for a positively charged surface and 10 seconds for a negatively charged surface. In addition, these emitters appeared clean, with little or no white contaminant on the tips. These results suggest that neglecting preventive maintenance in a Class 1 cleanroom does not detrimentally affect emitter performance.

The data shown in Figures 1 and 2 also indicate that cleaning the emitter tips in the Class 100 area did little to change the voltage swings or discharge-time variation. After cleaning, the average positive voltage was 374 and the average negative voltage was 112, for an average delta of 262. The discharge time averaged 14 seconds for a positively charged surface and 10 seconds for a negatively charged surface, for an average delta of 4. Cleaning the emitters in the Class 1 area did not improve the voltage or discharge-time deltas, but these already were very good. Following the cleaning procedure, the average positive voltage was 143 and the average negative voltage was 162; the discharge time averaged 9 seconds for a positively charged surface and 10 seconds for a negatively charged surface.

Balancing, however, greatly reduced voltage and discharge-time variations for the emitters in the Class 100 area. After the emitters were balanced, the average positive voltage of 193 more closely matched the average negative voltage of 210. The discharge times also were very similar, measuring 12 seconds for a positively charged surface and 11 seconds for a negatively charged surface. Balancing the emitters in the Class 1 area did little to enhance their already excellent performance. The average positive voltage was 158, compared with an average negative voltage of 158; the discharge time averaged 9 seconds for a positively charged surface and 10 seconds for a negatively charged surface.

Figure 3: Ion emitter positive voltage as a function of time in a Class 100 area.

Figure 4: Ion emitter negative voltage as a function of time in a Class 100 area.

Figure 5: Ion emitter positive-voltage discharge time as a function of time in a Class 100 area.

Ion Emitter Performance as a Function of Time. The measurements taken during the four-month period following preventive maintenance revealed that the voltage swings and discharge-time variations generally did not increase or decrease as a function of time, although some of the emitters exhibited severe variability. Figures 3 and 4 show that both the positive and negative voltages measured on a charge-plate monitor in the Class 100 area remained fairly constant for eight out of ten emitters. The other emitters required continued monitoring. The representative results for five emitters depicted in Figures 5 and 6 show that there was much variation in discharge time from month to month in the Class 100 area, but there were no data trends for any of the emitters. Air currents, adjacent equipment, or people could have been interfering with the emissions and causing this variability. It was recommended that a cleaning regimen of three to four times per year and a balancing regimen of two times per year be maintained.

Figure 6: Ion emitter negative-voltage discharge time as a function of time in a Class 100 area.

Figure 7: Ion emitter positive voltage as a function of time in a Class 1 area

Figures 7 and 8 show that both the positive and negative voltages measured on the charge-plate monitor in the Class 1 area remained constant over time, while Figures 9 and 10 indicate that there was much variation in discharge time from month to month in this area, since times for some of the emitters appeared to increase over the test period. Again, air currents, adjacent equipment, or people could have been the cause of this variability. Monitoring was scheduled to continue, but the data from this four-month period suggest that preventive maintenance on ionizers in Class 1 areas can be performed less frequently than recommended by the manufacturer. Cleaning twice a year and balancing annually would probably be sufficient.

Figure 8: Ion emitter negative voltage as a function of time in a Class 1 area.

Figure 9: Ion emitter positive-voltage discharge time as a function of time in a Class 1 area.

Figure 10: Ion emitter negative-voltage discharge time as a function of time in a Class 1 area.

Effect of Unbalanced Emitters on Particle Attraction. Figures 11 and 12 illustrate particle attraction to silicon witness wafers as a function of time when the ionizer is set to emit all positive and all negative ions, respectively. Figure 13 shows particle attraction as a function of time when the ion emitter is turned off. An analysis of variance performed on these results indicated that none of the test variables (positive, negative, off, and time) was significant.

Figure 11: Particle attraction to silicon wafers as a function of time with a positive (200- to 400-V) charge on wafers.

Figure 12: Particle attraction to silicon wafers as a function of time with a negative (–200- to –400-V) charge on wafers.

Figure 13: Particle attraction to silicon wafers as a function of time with ion emitter off.

Figure 14: Particle attraction to silicon wafers as a function of excessive surface voltage.

Discussions with experts in the field revealed that results such as these should be expected, because with a surface charge of <500 V, electrostatic attraction forces are insufficient to pull particles from the airstream to a wafer surface. Based on these findings, conservative preventive maintenance specifications should require that an ion emitter be rebalanced whenever the difference between its positive and negative readings exceeds 250 V.

Particle Attraction to Wafers with Excessive Voltage. Figure 14 shows particle attraction to silicon witness wafers as a function of surface charge when the charge was in excess of ±1000 V. (Actual charges in this test were ±1500, ±3000, and ±6000 V.) The figure indicates that as the surface charge increases, the potential for attracting particles from the airstream increases significantly. These data suggest that an audit of the facility's cleanroom equipment, materials, and processes was needed to ensure that there were no high-surface-charge materials or process equipment surfaces that could attract particles.

Field Meter Audit. Triboelectric charging from airborne particles was expected to be the largest source of charging in the cleanrooms being audited, because both the structural components and the equipment are grounded and most areas have emitters in the ceiling grids. Noncontact field meter measurements of the charge levels on walls, floors, shelves, and tools in the cleanrooms with ceiling emitters revealed that the surfaces of all stationary metal items showed no charge buildup (±200 V). The measurements of moving items such as robots, motors, conveyors, and carts also revealed no charge buildup. However, although no electrostatic voltage buildup was measured on plastic shielding, Teflon and polypropylene cassettes, and packages when they were undisturbed, as soon as they were moved, they exhibited a 5000- to 18,000-V charge.

The field meter measurements of the walls, floors, and shelves in the cleanrooms without ceiling emitters also revealed no charge buildup, which can be attributed to the fact that these structural components are grounded. Nonconductive plastic cassettes and packages in these rooms held surface charges in excess of 15,000 V. However, these surface voltages dissipated rapidly (in <60 seconds) as soon as these items were moved under the ionization emitters inside minienvironments. These data demonstrate that minienvironments with ionization emitters can be as effective in controlling static charge as classic ballroom cleanrooms with ceiling emitters.

A final set of field meter measurements of equipment and tools was taken after the ion emitters in a cleanroom had been turned off completely for 30 minutes. In this instance, there was no charge buildup on personnel, walls, floors, or robots, but nonconductive cassettes and plastic shielding held a charge, which caused triboelectric charges to build up when these items were moved. Based on these findings, it is recommended that operations in the affected process area be suspended whenever an ionization system fails.

Conclusion

Testing has revealed that cleaning and balancing ion emitters in a Class 1 environment after six months of operation did not provide much performance benefit, whereas cleaning and balancing the emitters in a Class 100 area after eight months did improve voltage swings and discharge-time variation. These differing results are attributable to the fact that the cleaner Class 1 environment slows down the buildup of organic and inorganic contaminants on the emitter tips, which ultimately causes the emitters to operate inefficiently and go out of balance.

Follow-up monitoring indicated that ion voltage and discharge times remained constant during the four-month period after the emitters were cleaned and balanced. Based on these results, it appears that cleaning ion emitter tips in a Class 1 area every six months will effectively maintain their performance level. Manufacturer-recommended maintenance and spot checking is indicated for emitters in Class 100 areas until more data can be collected. Monitoring results also indicates that balancing ion emitter tips every six to twelve months will effectively maintain their performance levels in both Class 1 and Class 100 areas.

Evaluations of particle attraction to witness wafers showed that a surface charged in excess of ±1500 V is necessary to pull particles from the airstream. Surface charges below ±500 V did not cause particles to deposit on wafer surfaces even when ionizers were unbalanced. Therefore, a conservative guideline would be to rebalance an ion emitter when the difference between the negative and positive charge-plate monitor readings is ±250 V.

Finally, a cleanroom audit using a noncontact field meter indicated that nonconductive cassettes, packages, and plastic shielding were at high risk of surface charge buildup. However, placing these materials under properly operating ionization emitters in a minienvironment dissipated their surface charge. Conductive and grounded materials were not a source of surface charges.

Semiconductor industry cost constraints require that such systems as ion emitters be analyzed when upgrading or building new fabs so that possible non-value-added expenses can be avoided. The typical cost of a traditional ballroom-type cleanroom with ion emitters in the ceiling grid is greater than the cost of installing point-of-use ionizers in minienvironments. This cost difference mandates that a thorough design analysis be completed before construction or reconstruction begins.

Michael G. Harrison was the cleanroom engineer at MEMC Electronic Materials (St. Peters, MO) from 1996 to 1998. He was responsible for standardizing all the plant's cleanroom protocols and for approving consumables and future design requirements. He was also responsible for the removal of particles and surface metals from wafers by means of scrubber cleaning technology. Harrison has worked as a material scientist in R&D for General Electric and as a process engineer for DuPont and MEMC. He holds five patents for work in advanced composites and has authored two publications. He received his BS and MS in ceramic engineering from Clemson University (Clemson, SC) in 1983 and 1985, respectively. (Harrison can be reached at 314/236-4316 or mharrison@meridianmt.com.)

MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit