Pilot study explores best methods for cleaning cleanroom surfaces
Research indicates cleaning motion has equal impact on particle concentrations on cleanroom walls
By John F. O`Hanlon, Alamgir Arif, Robert G. Cates, Larry DeShane & John A. Rice
This study records particle concentrations on the walls of a functioning Class 1 manufacturing cleanroom before and after cleaning by a commercial cleanroom contractor. Data was taken with two procedures: a vertical wipe, as described in the IES protocol, and a horizontal wipe. The results
indicated that the cleaning motions produced walls of virtually equal cleanliness, whereas the condition of the walls prior to cleaning was greatly influenced by the manufacturing activities taking place adjacent to the wall.
Introduction
The goal of these studies is to develop the best method for cleaning a cleanroom surface. In order to determine the best methods, several questions need to be answered. What is the condition of a surface prior to its being cleaned? How clean is the surface after it is cleaned? How much do people shed in the process of cleaning? What is the best measurement method to validate the condition of a surface after cleaning? What is the most cost efficient way to clean a wall or other surface in a cleanroom? The answers to these questions would provide the necessary data to answer the ultimate question: Can we improve current protocol?
The strategy is to perform pilot studies to explore the feasibility of this goal and eventually to establish working groups through the Institute of Environmental Sciences/International Organization for Standardization (IES/ISO) to validate the feasibility studies. If appropriate, modifications to the recommended practices would be suggested.
In Part I of this study1, particle shedding by cleaning personnel was studied as a function of the arm and body motions made by personnel in the simulations of cleaning floors, walls and ceilings, e.g., the test subjects held the appropriate tools in hand, but did not touch the walls. In this manner, the component of particles shed by the movements of the personnel were measured. The results of that study indicated the shedding was proportional to the extent and rate of the motions made by the operators. For example, large cross-body torso rotating motions generated more particle shedding than vertical wiping motions made with the torso held relatively still.
In Part II of this study an air-based surface counter was used to study the condition of the walls prior to and after cleaning. The third and future part of this study will develop a liquid-based technique for quantifying surface extractable material. The liquid extraction procedure will allow analysis of the collected material and perhaps, answer questions relating to possible wafer contamination. This article will review the data obtained in the air-based surface counting study of walls in a Class-1 manufacturing facility.
Experimental design
Four wall surfaces were selected in a Class-1 production facility. Each surface was 12 feet high and 24 feet long. All four surfaces were characterized before cleaning. Two surfaces were cleaned with part of the standard IES recommended practice2, whereas the remaining two surfaces were cleaned with a variation on the recommended practice. All four surfaces were characterized after cleaning. The wall surfaces were metal which was coated with baked epoxy.
A 1:14 IPA:UPW mixture was used to clean the walls. Each wall section was cleaned in sub-sections 12 feet high and 4 feet wide, beginning at the top and working downward, using overlapping strokes and wipes, that were folded before each pass. The cleaning was performed by the commercial cleaning personnel normally working in that facility. Two wall sections were cleaned with vertical motions and two walls were cleaned with horizontal motions. The vertical motions used were those described in the IES recommended practice. In addition, the experiments with horizontal and vertical motions were done on different days and on different walls, to avoid any systematic error. A vacuum wand was not used, as the integrated vacuum system was not yet operational in this manufacturing facility. The summary of the experimental design is shown in Table 1.
This manufacturing facility is cleaned on a seven-day rotating schedule, so that at any one time some walls are being cleaned. The four walls cleaned in this study were cleaned between four and seven days prior to the day of taking data on each wall.
Each surface, whose area was 288 square feet, was characterized by measurements at 96 locations with the Dryden3 Q-3 surface particle monitor. For purposes of cleaning and measurement, the 12- by 24-feet area was divided into sections 12 feet high and 4 feet wide. The Dryden counter used a 2-inch-diameter head; the operators attempted to move the counter 6 inches consistently in the 3-second measurement time. Sixteen measurements were taken on each of the 12- by 4-feet sections. Four along the upper elevation (6 inches from the top) at 1-foot intervals, followed by four at the 8-foot elevation, four at the 4-foot elevation and four measurements approximately 6 inches from the floor. After completing one 12-feet high by 4-feet wide section, the equipment was moved to the next section. The surface particle counter operator wore a Dryden face helmet and was positioned in such a way that the unmeasured portions of the panel were not exposed to operator or equipment contamination. In this manner, operator contamination of the unmeasured sections was minimized. This procedure was repeated six times in measuring the 12- by 24-feet wall section. The quality of the room air was checked at each end of the 24-in. panel before and after the measurements were taken with a PMS4 Lasair airborne particle counter which read 0 particles at 0.3 µm diameter.
Results and analysis
The results of this experiment are presented in six figures. Figures 1-4 present the particle counts measured on each of the four walls prior to cleaning. Figure 5 gives the averaged results after cleaning walls 1 and 2, whereas Fig. 6 reports the averaged results after cleaning walls 3 and 4. It was observed that wall sections 1 and 3 were rather dirty near the floor, whereas section 3 showed high contamination near the ceiling. Sections 1, 3 and 4 showed high contamination in the center of the panels. One particular location on wall section 1 showed over 100 particles per sq. in. After cleaning, this disappeared. Two days later, when the same spot was measured, it showed 25 counts. Over that spot, an opening between the sidewall and the HEPA filter was observed. One could infer that a venturi was formed causing particles to be pulled into the facility, and in the process they were deposited on the wall by impaction. Analysis of the data closest to the ceiling revealed several other locations where particle counts were high, 10 to 25 particles per sq. in., and each location occurred under a gap between the wall and the HEPA filter.
The average particle concentration on the four walls prior to cleaning was 3.48 particles per sq. in. After cleaning it was observed that cleaning with the horizontal and vertical wipes was essentially identical. The horizontal stroke showed a slightly cleaner wall, however, the variance in the data was large enough to prevent any further conclusions. This data is presented in Table 2.
Conclusions
Analysis of the as-cleaned walls four to seven days later revealed that local contamination along the top edge of the wall was caused by leaks between the wall and the HEPA filters. It is not possible to determine whether these remained from the original construction or separated at a later time. It does illustrate that the surface becomes an impaction counter and is therefore much more sensitive than an airborne counter held under the same leak site. The surface impaction counter integrates the deposit and the sensitivity of measurement.
The second conclusion which may be drawn is that there is little difference between the horizontal and vertical cleaning motions used in this study. The horizontal wiping motion showed some improvement, 0.37 vs. 0.49 particles/sq. in., however, the large standard deviation precludes any definitive statement. The data does suggest two further conclusions: the procedure in the IES recommended practice is not conclusively better than the modified procedure chosen here, and there may be other procedures which are superior. There is room for improvement in specifying the process. This is an area where those interested could work together with the IES/ISO groups in further refining the recommended practices.
Lastly, it is concluded that there is a great deal of improvement possible in defining the time intervals between cleaning cleanroom surfaces. Most facilities clean on a time cycle; typically one week for routine cleaning, less frequently for major cleaning and more frequently, perhaps daily, for surfaces adjacent to wafer load-unload locks. Analysis of the four walls showed that wall 2 was extremely clean compared to the other three walls. This related to the equipment installed on the opposite wall as well as to personnel traffic through the area. What this data implies is simply that one needs to take periodic measurements through the entire facility to best define the time interval between cleaning operations. Some walls might best be cleaned every two to three days; whereas others might go two weeks between cleanings. This analysis could result in a cost savings to the facility. As a result of this work, the authors have been asked many times `how often should walls be cleaned?` That question cannot be answered quantitatively from this work. It can be answered by similar studies within the individual facility. Commercial office cleaners are asked `how often should carpets be cleaned?` The answer — when they become dirty, and the time interval will vary accordingly. The difference between maintaining the office and the cleanroom is then only a matter of the contamination level acceptable, how it is removed and how it is measured.
The cause and effect, if any, between wall contamination and wafer contamination has been questioned. This study cannot address that issue. However, analysis of the contamination removed by a liquid-based sampling technique might be able to determine such a relationship.
Acknowledgments. The authors would like to thank Stephen Rodgers, microcontamination manager, Intel Fab 12, Chandler, AZ; Rey Santillano, engineering technician, Intel Fab 12, for permission to take the data; and Dick Dryden of Dryden Engineering for discussions relating to use of the Q3 counter. This work was sponsored by ServiceMaster Company, Project #431,520.
REFERENCES
1. John F. O`Hanlon, Sean M. Collins, Robert G. Cates, Larry DeShane and John A. Rice, “Cleanroom Cleaning Studies,” Cleanrooms `96 East Conference Proceedings, April 22-25, 1996, Boston, MA. pp. 531-560.
2. IES-RP-CC018.2, “Cleanroom Housekeeping — Operating and Monitoring Procedures,” Institute of Environmental Sciences, 940 East Northwest Highway, Mount Prospect, IL 60056
3. Dryden Engineering Company, Inc., 210 Hammond Avenue, Fremont, CA 94539-7465.
4. Particle Measuring Systems, Inc., 5475 Airport Road, Boulder, CO 80301.
John F. O`Hanlon is a professor of electrical and computer engineering at the University of Arizona and director of the University of Arizona NSF Industry/University Center for Microcontamination control. He is also a principal investigator in the UA SEMATECH Center for Contamination/Defect Assessment and Control.
Alamgir M. Arif is researching effective protocol for the cleaning procedure of cleanrooms as part of his graduate coursework for the master of science in electrical engineering degree he is pursuing at the University of Arizona (Tucson) under Prof. John O`Hanlon. His work also involves the analysis of particles to detect probable sources of contamination.
Robert G. Cates is facility manager for ServiceMaster Co.`s Business and Industry Group at Intel in Chandler, AZ, where his responsibilities include training managers in cleanroom procedures, writing cleanroom cleaning protocols and working with microcontamination managers. He has also serves as a training consultant for cleanroom services on new fab plants.
Larry DeShane is ServiceMaster Co.`s director of technical support. He is the co-author of ServiceMaster manuals – “Contamination control and cleanroom technology manual;” “Paint systems manual;” and “Food processing manual.” DeShane is a member of the IES and the society of manufacturing engineers.
John A. Rice is the West Division Director of operations for ServiceMaster. He has Arizona, Connecticut and California contractor`s licenses, and he as received various certifications, including the ASU Microcontamination and control cleanroom certification.
 |
The figures above show particle concentrations measured at four elevations on each of four wall sections. On wall section 1, the measurements were taken four days after routine cleaning by commercial cleaning personnel; on wall 2 they were taken six days after routine cleaning; on wall section 3, four days after routine cleaning; and on wall section 4, seven days after routine cleaning.
 |
 |
Particle counts on the cleanroom wall were measured with the Dryden Q3 surface particle counter.
 |
 |
The figures above show particle concentrations measured at four elevations on wall sections 1+2 and 3+4, just after cleaning by commercial cleaning personnel. Walls 1+2 were cleaned using the routine vertical wiping motion, while walls 3+4 were cleaned using a modified horizontal wiping motion.