Evaluating the particle containment effectiveness of face masks and head gear

By: Patrick McPherson, Digital Semiconductor;
Daniel Duggan, Dryden Engineering;
and John Manguray, Candescent Technologies

As semiconductor manufacturing has become increasingly automated, efforts to control particle contamination have focused on equipment cleanliness rather than on cleanroom personnel. However, as process geometries are reduced to 0.25 μm and below—sizes where 0.1-μm particles can contaminate a wafer—it is important to reevaluate the contamination contribution of personnel, especially those who come in physical contact with process equipment during cleaning procedures and in proximity to exposed wafers during critical process steps.

Data published 10 years ago indicated that at that time the four main sources of cleanroom contamination and their respective contribution levels were chemicals, 20%; processes, 20%; process equipment, 25%; and personnel, 35%. 1 More recent studies suggest that with the use of standard mechanical interface (SMIF) containers and minienvironments coupled with improvements in cleanroom design, chemical purity, and cleanroom garment materials, the percentage of contaminants from personnel has fallen to the 5—10% range. The contribution from process equipment, however, has increased to between 80 and 90%. Unfortunately, this information regarding a major shift in contamination sources has caused some fabs to become complacent about their cleanroom products and protocols. As a result, personnel wearing particle-shedding products while working near open equipment chambers may be driving the contamination contribution from personnel back up to 35% and even beyond. The problem is not that the technology needed to control these contaminants isn't available; rather, the problem is that the technology isn't being applied because the industry is underestimating the severity of personnel-generated contaminants.

Study Rationale

Digital Semiconductor's Fab 6 in Hudson, MA, which is processing at CMOS 6, or 0.35-μm line geometries, is scheduled to switch to CMOS 7, 8, and 9, or geometries of 0.25, 0.18, and 0.12 μm, respectively. With approximately 400 employees using protective cleanroom garments at the fab, the company decided to study the risk of personnel-generated contamination at the 0.1-μm level. The intent was to evaluate the effectiveness of various face masks and head gear systems for particulate containment at that level.

A similar study conducted by Dryden Engineering in 1984 yielded data that questioned the effectiveness of face masks at the 0.5-μm level. In recent years, some industry experts have theorized that cleanroom face masks will be even less effective at preventing 0.1-μm particle contamination. Therefore, Digital and Dryden agreed to cooperate on the particle-profile study to test that hypothesis. The testing was conducted June 2—4, 1997, at Dryden's Class M1.5 test facility in Fremont, CA.

Besides passive face masks, Digital has used various active head gear systems with hard shields for personal particle containment. These include a half shield that covers the lower half of the face; a half shield and protective glasses; a split full-face shield that allows users to raise or lower the upper portion; and a full-face, one-piece shield. The study included all of these systems as well as four face masks and a beard cover. In addition to its interest in the relative effectiveness of the various types of containment systems, the company was interested in a side issue involving the different head gear configurations. The fab's cleanroom protocol requires that if personnel touch their faces or skin, they must immediately replace their gloves, and that safety glasses be worn whenever the upper portion of the split shield is raised. However, it is recognized that people will frequently touch their faces without being aware they are doing so, and operators have been seen walking around in cleanrooms with raised split shields but without safety glasses. Consequently, there was interest in study data comparing the full-face shield with the half shield and the split shield.

Test Parameters

Ten different types of personal containment systems readily available in the marketplace were tested in the particle-profile study. These systems, which are identified in Table I, are used in cleanroom facilities to prevent the migration of particles generated by cleanroom operators through normal breathing and sometimes by forceful discharge, such as coughing and sneezing. A test setup was devised to measure the nominal particle levels of each system under these conditions.

Table I: The 10 personal containment systems evaluated in the particle-profile study.

All testing was conducted in a Class M1.5 cleanroom to maintain controlled background particle counts at an acceptable level, thereby protecting against interference from outside particle sources that could affect the test results. Airflow velocities at the cleanroom filters ranged from 85 to 90 ft/min. The cleanroom surfaces, test equipment, test setup, and operator cleanroom apparel were cleaned prior to the testing. Gowning and handling procedures typical of airborne particle studies were followed. A cleanroom suit, bouffant, head/face cover, gloves, and shoe covers were worn by the test operator, who remained at a distance from the test subject. The test subject was gowned in a polyester building suit (with no street clothes); Gore-Tex cleanroom coveralls, overboots, and hood (when not wearing a test shield); and vinyl gloves. The gloves were changed between each face mask or shield test cycle.

The laser particle counter used in the study (Model A2100B; Met One, Grants Pass, OR) operates on the light-scattering principle and is capable of sizing particles as small as 0.10 μm. As particles pass through the source light beam in the instrument's detection chamber, the number of particles is determined by the number of discrete light-scattering events, and the size of each particle is determined by the magnitude of the scattering event. The counter samples air at a rate of 1 cu ft/min. A clean isokinetic probe and a particle-sampling hose also were used during testing, as were a cleanroom chair, a holding assembly for the sampling probe, and a metronome. Particle samples for normal breathing were taken for 12 periods of 10 seconds each, resulting in a total air sample of 2.0 cu ft. For coughing tests, three 10-second samples were taken while the test subject coughed every 5 seconds, for a total of two coughs per count cycle.

To create consistent test conditions throughout the study, all testing was conducted with the subject using mouth-only breathing. Mouth breathing was selected because it creates a more severe environment than nasal breathing does, generally involving heavier breathing and the potential for spittle. Using this technique also permitted the 0.5-μm test data to be correlated with data generated during the 1984 study (mentioned above), serving as a reference point for study validation.

Test consistency was further ensured by using a single test subject, a 47-year-old mustached male who was a nonsmoker in good health. Individuals vary not only in face size and contour, but also in their breathing patterns. They may also put on masks differently, regardless of instruction. To eliminate these variables, and to ensure that test data reflected the efficiencies of the containment systems under study and nothing else, the decision was made to use only one test subject. However, even under these conditions, when the tests conducted on the first day of testing were repeated on the second day, some variation was observed between each day's results.

The tests were conducted over a 3-day period. On the first day, breathing and nonbreathing tests were conducted for each of the 10 contaminant systems. For each system, particle counts were taken 14 times at nine positions in front of the face covering (see Figure 1). The 1st and 14th counts were taken with the subject holding his breath; the 2nd through 13th counts represent mouth breaths at 5-second intervals. The respective counts were then averaged to provide nonbreathing and breathing particle data. The nonbreathing test aspect was included because of conjecture that the cleanroom's laminar airflow moving down past a face mask could be dislodging particles. The coughing test was performed in a similar manner, although only three counts were taken at each position, with the test subject coughing at 5-second intervals.

Figure 1: Particle sample probe locations around the subject's face

The same tests were repeated on the second day, the only variable being that the containment systems were tested in reverse order. On the third day, two different types of test were conducted: a marching test and a mask-stretch test. Because some cleanrooms in today's fabs are quite large, reaching a football field or more in length, cleanroom personnel frequently are highly active during the course of their work shift. To simulate these conditions, the marching test required the subject to march in place at a rate of 95 steps per minute, matching a metronome beat. A wide-area particle sampler, encompassing approximately 12 x 8 in., was positioned just below the subject's chin. Particle counts were recorded every minute for 5 minutes and then averaged to arrive at a value for each containment system.

Results from the first two days had indicated that the face mask tests were producing higher particle counts than a control test with the test subject wearing no face protection whatsoever. These high particle counts had to have some source, and the only variable was the addition of the face masks themselves. Were the additional 0.1-μm particles being generated by the mask materials? To try to answer that question, a mask-stretch test was devised. In this procedure, the sides of each mask were connected to two stainless-steel rods that were then flexed at a metronome pace with a particle counter probe mounted 3—4 in. below the mask. The underlying theory was that any particles detected must have come from the face mask material.

Analytical Observations

Figures 2—6 present the data derived from the comparative testing described above in particles per cubic foot. These data are still being analyzed to determine their full significance and to identify what type of additional testing is needed. Nevertheless, there are some test results that deserve immediate consideration.

Figure 2: Results of breathing tests. Each bar represents the average of all readings from both days of testing (12 counts were taken at each of nine locations each day, for a total of 216 samples for each test system).

Figure 3: Results of tests with subject holding in breath. Each bar represents the average of all readings from both days of testing (two counts were taken at each of nine locations each day, for a total of 36 samples for each test system).

Figure 4: Results of coughing tests. Each bar represents the average of all readings from both days of testing (three counts were taken at each of nine locations each day, for a total of 54 samples for each test system).

Figure 5: Results of marching tests. Each bar represents the average of five counts taken with a wide-area sampling system.

Figure 6: Results of mask-stretch tests.

Most importantly, the test data revealed that the control test, during which the subject wore no protective mask, yielded a lower 0.1-μm particle count than the count for any of the passive face masks tested (see Figures 2—5). That is, wearing a face mask seems to increase the count of 0.1-μm particles in the area around the wearer's face. This finding was unexpected and led to follow-up testing. To confirm that the test setup did not bias the control results, a shortened version of the no-mask test was performed using the breathing and coughing protocols described above. A different subject was used, but one with the same general traits: male, mustached, and nonsmoker. During this follow-up testing, the test operator first scanned the entire area above, below, and to the sides and back of the subject's head to determine if there were any high-particle-level locations that had not been included in the initial test protocol, and no such locations were detected. Several counts were then taken at the nine protocol locations, and these readings were at the same general level as those taken during the initial testing, which confirmed that the test protocol was valid for the no-mask testing.

At this point, although the data are clear and repeatable, the explanation for this difference in particle levels between not wearing a mask and wearing any mask is only conjecture; however, it does not contradict the earlier theory that face mask materials that are capable of capturing and containing 0.5-μm particles would be ineffective at the 0.1-μm level. A passive mask will filter the wearer's breath, but the air has to go somewhere, and, while >=0.5-μm particles are captured by the inner facing of the mask, smaller particles may be carried with the air as it escapes through the mask and around mask edges. Another possibility, suggested by the mask-stretch portion of this study is that the mask material itself is a source for 0.1-μm particles when the mask is pulsed by the wearer's breathing (see Figure 6).

The nonbreathing tests also revealed particle count variations among the containment systems, with the masks having the highest counts. The 0.1-μm particles detected during these counts, which were taken immediately before and after the breathing test counts, may have been dislodged from the mask material by the downward laminar flow of clean air in the cleanroom. Perhaps there is some other explanation. In any event, the data are repeatable and the containment system rankings in the nonbreathing tests were almost identical to the rankings in the breathing tests.

In a sense, these face-mask test data should be viewed by semiconductor manufacturers as problematic. Initial analysis points to the conclusion that the masks that are somewhat effective contaminant barriers at 0.5 μm are not only ineffective at 0.1 μm but may be a source of contamination at that level—the masks' facing materials may be the source for the 0.1-μm particles. The good news, however, is that the full-shield and split-shield containment systems produced excellent results in the breathing and coughing tests, offering optimal particle containment at the 0.1-μm level.

Conclusion

Comparative tests of various personal containment systems confirm earlier hypotheses regarding the inefficacy of face masks for particle control at the 0.1-μm level and also suggest directions for further research. If fab management is concerned about contamination at that level, the means used to contain particles from the head and face of personnel should be considered carefully, and the standard solution, face masks, should be questioned. It may be accurate to say that contamination control in the semiconductor industry has evolved to a point where equipment is a far greater contaminant source than personnel, but that will remain true only if state-of-the-art technology is applied to contain personnel-generated particles.

Reference

1. Dixon AM, "Guidelines for Clean Room Management and Discipline," in Handbook of Contamination Control in Microelectronics: Principles, Applications and Technology , Tolliver DL (ed), Park Ridge, NJ, Noyes Publications, chap 4, 1988.

Patrick McPherson is a principal engineer in the CFM group for Digital Semiconductor, Hudson, MA, where he is responsible for protocol and contamination issues in the cleanrooms. He has worked for the company for 10 years. Before joining Digital, he was a process engineer for Motorola with 18 years experience in wafer processing. McPherson has an associate degree in general studies from Mesa Community College (AZ). (McPherson can be reached at 978/568-5236.)

Daniel Duggan is a contamination control specialist with Dryden Engineering Fremont, CA. He has been selling equipment, supplies, and technical services to the worldwide cleanroom market for 17 years. Duggan's areas of expertise include apparel systems, gowning room design, protocol development, and surface particle detection. He holds a BS in entomology from the University of California at Davis. (Duggan can be reached at 800/379-3361.)

John Manguray is a contamination control engineer at Candescent Technologies Corp. in San Jose. He has more than 13 years' experience in the contamination control field, having worked with Lockheed-Martin and Dryden Engineering. His experience includes expertise in the testing and certification of cleanrooms and minienvironments, cleanroom protocol training, and cleanroom materials testing. He has planned, developed, and documented precision cleaning techniques and surface inspection methods for processing of product. He has also conducted several studies on various materials considered for use in cleanroom facilities including particle generation and migration studies, outgassing characteristics, chemical compatibility, ESD properties, and film contaminants on surfaces. He holds a BS in chemistry from the University of California at Santa Cruz.