 |
 |
 |
 |
|
Comments? Suggestions? Send us your feedback.
MicroMagazine.com |
SURFACE CLEANLINESSComparing cleanroom wipers with a dry abrasion resistance testOsmond Atterbury, Himansu R. Bhattacharjee, Douglas W. Cooper, and Steven J. Paley, Texwipe The continuing effort to improve the cleanliness of semiconductor fabs and other contamination-sensitive manufacturing environments includes selecting non-particle-shedding fabrics for use in cleanroom wipers, garments, and swabs. Traditionally, wiper fabrics have been evaluated for releasable particles by agitation in air (Helmke drum and flex tests) or in water (shaking, swirling, or ultrasonicating, with or without the addition of solvents or surfactants). Although such tests may include some fabric/fabric and fabric/vessel contact, particle release by abrasion is not their major concern. Other fabric tests include those for sorption capacity and rate, permeation, electrical resistance, and the amount of various types of chemicals— nonvolatile residue, specific organic compounds, and specific ions, for example—leachable from the wipers by various solvents. Tests designed to measure particles released by abrasive interaction between wipers and a surface, however, have rarely been reported. Exceptions to this are the stress-strain studies by Mattina and Paley, and Oathout and Mattina, which showed that increases in the amount of particulate material released in aqueous solutions correlated with increases in the stress (force per area) applied to the fabrics tested.1, 2 Cleanroom wipers are used on a variety of surfaces, including glass, plastics, metals, and ceramics. Some of these materials and the contaminants on them are relatively abrasive. The wiper itself may need to be made of a somewhat abrasive fabric to successfully remove contaminants. The abrasive nature of both wipers and the surfaces to be wiped is particularly important in applications where the internal areas of process equipment are being cleaned after sputtering, etching, chemical vapor deposition, or chemical-mechanical polishing. The goal of the study reported in this article, therefore, was to develop a dry test that would allow comparison of wipers on the basis of particle release caused by abrasion. The general literature on tribology (lubrication, friction, and wear) indicates that the volume of material produced by abrasion is proportional to the normal force per unit area and the distance traveled, with the proportionality constant itself dependent on the surface materials, environment (including temperature and humidity), and a host of other factors.3 Surfaces change as abrasion continues, as well, so we sought to study a degree of mild abrasion likely to be comparable to that which occurs during typical equipment-wiping applications. Developing a useful test involved compromises between idealized conditions that were not representative of conditions of use and more realistic, nonideal conditions that are hard to describe, control, and reproduce adequately. Thus, much preliminary experimentation went into development of the wiper test method described below, which might also be useful for evaluating fabrics used for garments and swabs. Equipment, Procedures, and Materials Tested An important step in the development of the cleanroom wiper abrasion test was the fabrication of an automatic wiping machine, which was done in-house. Figure 1 is a schematic drawing of this apparatus. Significant features of the equipment include a counting device to monitor the number of wiping strokes performed and a closed-loop line to channel the abraded debris to an optical particle counter. The particle counter's 60-second recording cycles are synchronized with the strokes of the wiping machine. Standard-grade 316-stainless-steel wire mesh screen pieces encased in a 10 x 10-cm stainless-steel housing are used as the wiping surface. During testing, a wiper is placed in a flat, mainly stainless-steel holder having specified dimensions, pressed against the wire mesh with specified force, and caused to move back and forth against the mesh at a specified speed and stroke length while air is drawn through the mesh. Particles released by friction are entrained in the airstream and drawn to the optical particle counter, where they are sized and counted. (This method is similar to that reported by Miller et al. for measuring particles released from a pin-on-disk abrasion test.4)
Figure 1: Schematic drawing of the abrasion test apparatus. Specifically, the wiper holder used in the developmental tests was 4.9 cm long and 3.8 cm wide and the mesh pieces were 5.1 x 7.7 cm. The data reported below were obtained with a 60/in. screen, though other meshes were also explored. Abrasion is determined by weight per area, among other variables; the weights applied to the wiper holder totaled 1377 g, giving a weight per area of 74 g/cm2 or about 0.07 atm pressure. The stroke length was 3.6 cm, and the wiping rate was 60 back-and-forth cycles per minute (120 strokes per minute), giving an average velocity of 7.2 cm/sec. The wipers were tested using two wiping motions, designated the machine direction and the cross direction. Figure 2 shows a knitted-wiper section and indicates how these directions, which produced very different results, correlated with the fabric weave.
Figure 2: Section of wiper fabric in relation to test machine's wiping motions. The machine direction parallels the fabric's knit rows, while the cross direction is at right angles to them. Testing was carried out in a clean hood, having nominally Class 100 air. The optical particle counter used was a Model CI-500 (Climet Instruments, Redlands, CA), drawing a sample at 472 cm3/sec (1 cu ft/min). The air velocity at the mesh screen was 23 cm/sec—the volume flow rate of 472 cm3/sec divided by the difference between the screen area (39.3 cm2) and the wiper holder area (18.6 cm2). The air was drawn through 100 cm of tubing with an inner diam of 0.95 cm. Each counting cycle lasted 1 minute (120 wiping strokes), and the resulting data represent the number of particles having an optical-equivalent diameter of > 0.5 µm. Before testing, each wiper was visually divided into quadrants. After background (or control) readings were taken, the first quadrant of the unfolded wiper was placed in the holder, and back-and-forth wiping in one direction (machine or cross) was performed for 1 minute with the counter recording particle data. Another minute of wiping (and counting) was done on the wiper's second quadrant, this time in the other direction. The third and fourth quadrants were tested similarly. Eight to 10 wipers each of the 10 knitted polyester cleanroom wiper types listed in Table I were tested, along with a conventional nonwoven composite blend wiper. The latter released many more particles than did the knitted polyester wipers and its results were not included in the subsequent statistical analyses.
Table I: The 10 types of cleanroom wipers tested. Results For all samples of the 10 polyester wiper types tested, the particle counts when wiping was performed in the machine direction were typically a factor of two or three lower than the counts during wiping in the cross direction. This difference is quite significant statistically and practically, suggesting that wiping in the direction of the fabric's knit rows could reduce wiper abrasion. If the wiping direction made no difference, the probability of getting these unanimous results is (2)(0.5)10 = 0.002. To compare the wiper types, however, the total counts (machine direction counts plus cross direction counts) for each test wiper were used. Table II shows the total counts for the 10 polyester wiper types by wiper and wiper type. Also included for each wiper type are the mean, standard deviation of the counts, standard error of the mean count (the standard deviation divided by the square root of the number tested), number tested, and relative standard deviation (RSD), which is the standard deviation divided by the mean, also known as the coefficient of variation. The relative standard deviations include variations from wiper to wiper and random experimental variation and range from 20 to 54%.
Table II: Total counts by wiper and wiper type, and the mean, standard deviation, standard error, number tested, and relative standard deviation for each wiper type. Table III compares the various wiper types in terms of probability levels, which are based on the results of a statistical comparison of their respective means, the Student's t-test. The t-statistic for two sample groups is the difference in their means divided by the square root of the sum of the squared standard errors. A computer program uses the Student's t-test results to determine the probability of getting means as different as those observed for the two groups when sampling from identical populations. The smaller that probability, the greater the likelihood that the actual sampled populations are different.
Table III: Probability levels based on results of the Student's t-test, a statistical comparison of the means. Boldface results indicate a statistically significant difference. The boldface values in the table are those comparisons where there was <5% (0.05) probability of getting such different values from identical populations, which is frequently the criterion used to say that two sets of results are statistically significantly different. Of the 45 nontrivial comparisons performed, 32 showed statistically significant differences. These results also indicated that the wipers fell into three groups: wipers WT01, WT02, and WT03 released significantly fewer particles due to abrasion than wipers WT04, WT05, WT06, and WT07, which, in turn, released significantly fewer particles due to abrasion than wipers WT08, WT09, and WT10. Testing more of the wipers would be expected to increase the t values by a multiplier of about the square root of N' divided by 10, where N' is the total number of tests run per wiper type; for example, testing 10 more wipers each for W01 and W03 would yield a multiplier of 1.41 and the t-test would be expected to go from near t = 2.0 to near t = 2.8, which would be statistically significant (p = 0.02). Figure 3 compares the wipers in another way, by plotting particle count data (the mean and an error bar representing plus-and-minus one standard error) against the wiper type.
Figure 3: Plot of particle counts versus wiper type, based on data in Table II. Statistically significant difference in this case translates roughly into those comparisons where the error bar for one wiper type does not overlap with the error bar of another. If the standard deviation were used, rather than the standard error, the error bars would be about 3.1 times as large (the square root of 10) and would indicate how the individual wiper counts vary, rather than the uncertainty in the means of those counts. Not only are the differences in mean particle counts of practical importance in comparing various wiper fabrics, the differences in the variation of the counts may also yield useful information. For example, wipers WT01 and WT02 had very similar means but WT02 had twice the standard deviation (four times the variance) of WT01. The usual test for differences in standard deviations is the F-test of the ratios of the variances. Whether or not a ratio of four is statistically significant depends on how many measurements have been taken. Using the computer program mentioned above, the F-test command was used to compare the variances from the abrasion tests. The results shown in Table IV indicate the probabilities of getting the F-values calculated for the wipers from populations that are the same. Again, a probability of ¾5% was used to indicate statistical significance; those values are given in boldface type. To return to the example, wiper WT02's variability was statistically significantly greater than that of wiper WT01, although it was not statistically significantly greater when compared with the other wipers. The axiom "the less variability, the better" is a standard rule in quality control.
Table IV: Probability levels based on results of the F-test, a statistical comparison of the variances. Boldface results indicate a statistically significant difference. Discussion A caveat should be given with regard to the use of abrasion testing. Cleanroom wipers may serve as clean work surfaces as well as being used to clean process equipment and other abrasive surfaces. They can be used wet and dry, with a range of pressures, on a variety of materials. This potentially wide range of usage conditions makes it hard to specify test conditions that are relevant to all users. Ideality and practicality must be balanced. The test described in this article enables users to compare wiper types (and perhaps wiper lots) with regard to the particles released by dry abrasion under the specified set of conditions. The more closely the test conditions resemble real-world usage conditions, the more helpful the rank ordering of wiper types and ratios of particles generated during testing will be. As mentioned above, the testing revealed that wiping in the cross direction, perpendicular to a fabric's knit rows, generated many more particles than did wiping in the machine direction. Possible causes include differences in the local contact geometry and in the relative velocity distribution and more global differences in the stretching of the knit during the motion. Knit fabrics produce rows that run like rails in the machine direction but run at right angles to the cross direction like ridges on a file. It would be valuable to know whether these ridges make wipers more effective at removing contaminants when used in the cross direction rather than in the machine direction. If there is no difference in cleaning efficiency, particle generation by abrasion could be reduced by wiping only in the machine direction. Other factors affecting a fabric's susceptibility to abrasion are likely to include the stress modulus of the basic (bulk) polymer and the treatment of the filaments and the geometry of the yarns and the knit. Further experimentation in these areas should lead to further reductions in abrasion-related particle shedding. Conclusion When used to compare 10 different wiper types, the abrasion resistance test described here was able to demonstrate a substantial number of statistically significant differences between the wipers with respect to the release of particles. The 10 wiper types fell into three groups, based on having means that were statistically different from each other. Statistically significant differences in variability were also demonstrated. Because of the specific test conditions utilized, the test should be particularly useful in comparing wipers to determine their suitability for cleaning abrasive (rough) surfaces such as the interiors of chambers used for etching, sputtering, and other deposition processes. References 1. Mattina CF, and Paley SJ, "Assessing Wiping Materials for Their Potential to Contribute Particles to Clean Environments: Constructing the Stress-Strain Curves," Journal of the Institute of Environmental Sciences, 34(5):21—28, 1991. 2. Oathout JM, and Mattina CF, "A Comparison of Commercial Cleanroom Wiping Materials for Properties Related to Functionality and Cleanliness," Journal of the Institute of Environmental Sciences, 38(1):41—51, 1995. 3. See, for example, Rabinowicz E, Friction and Wear of Materials, 2nd ed, New York, John Wiley, 1995. 4. Miller RJ, Cooper DW, Nagaraj HS, et al., "Mechanisms of Contaminant Particle Production, Migration and Adhesion," Journal of Vacuum Science and Technology A, 6(3): 2097—2102, 1988. Osmond Atterbury is an associate research chemist at Texwipe (Upper Saddle River, NJ). He is involved in developing measuring techniques for cleanroom materials and applying them to continuous improvement of those materials. He also carried out the measurements reported in this article. He has a BS in chemistry and zoology from the University of the West Indies (Kingston, Jamaica) and has authored numerous technical reports and standardized test methods. Himansu R. Bhattacharjee, PhD, is laboratory director and senior scientist for Texwipe, where he is responsible for technical research and new product development. Previously, he was a senior research chemist at AlliedSignal's Morristown, NJ facility. Bhattacharjee received his PhD in physical chemistry from Wayne State University, and is the author or coauthor of 25 peer-reviewed publications, including papers on particle counting by SEM. He holds 16 U.S. patents in such varied areas as laser printing plates, photoactivatable time and temperature indicators, and the modification of polyamides and polyesters. He is a member of the American Chemical Society and a senior member of the Institute of Environmental Sciences and Technology, where he serves on the working groups for wiper testing and swab testing. (Bhattacharjee can be reached at 201/327-9100, ext. 270.) Douglas W. Cooper, PhD, is the director of contamination control at Texwipe, where he is involved in R&D for advanced cleaning materials. After stints with GCA and as a member of Harvard's faculty, he worked at IBM's T. J. Watson Research Center where he conducted contamination-related research for 10 years. He received his AB in physics from Cornell, an MS in physics from Pennsylvania State University, and his PhD in engineering and applied physics from Harvard University. Cooper has published more than 100 technical articles and several book chapters, holds several patents, and has served on various editorial and research advisory boards. Active in the Institute of Environmental Sciences and Technology, he received the Willis Whitfield and Maurice Simpson awards from the institute and was elected a fellow in 1995. (Cooper can be reached at 201/327-9100, ext. 397.) Steven J. Paley is president of Texwipe and oversees all of the company's technical research, engineering, and new process development. Before joining Texwipe he held positions at IBM, where he worked on System 38 computer development, and at AT&T Bell Laboratories, where he was involved in the development of specialized flat-screen computer terminals for medical information systems. Paley has coauthored several papers on contamination metrology for consumable products and holds five U.S. patents with several others pending. He is a member of the Institute of Electrical and Electronics Engineers and a senior member of the Institute of Environmental Sciences and Technology, where he serves on the working group for wiper testing. He received his MS in mechanical engineering product design from Stanford University, and a BS in electrical engineering and a BA in English literature from Tufts University.
MicroHome |
Search | Current Issue | MicroArchives Questions/comments about MICRO Magazine? E-mail us at cheynman@gmail.com. © 2007 Tom Cheyney |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
