by John F. O'Hanlon
How clean are your cleanroom walls? How clean should they be?
Is it a mark of quality for walls to be maintained in pristine condition, or should the walls function as “sorption pumps,” silently, passively collecting particles? How frequently should your walls be cleaned?
With what techniques are they best cleaned?
Isopropyl alcohol used in cleaning solutions is becoming an identifiable source of volatile organic carbon. Can a more environmentally favorable chemistry be substituted? What is the best “solution?”
These are some of the questions being asked about cleanroom maintenance. Today, there are ISO recommended practices, which specify procedures for cleaning cleanroom surfaces. These standards, as adapted by local facility and maintenance personnel, become the de-facto internal standards by which a company maintains its facility.
The goal of facility management is simple and clear: maintain the cleanroom in such a condition that it does not impact manufacturing. Do it for the minimum cost, and without the release of unwanted vapors. Standards are based on the best existing practice at the time of their writing. They are modified periodically by studies suggesting alternative procedures, which could lead to reduced labor costs and environmental impact.
Limited studies of particle shedding from cleaning personnel have demonstrated that particle release is related to speed of arm movement. If one wipes at twice the normal speed, the particle release rate doubles, and approximately the same total number of particles is released, but in half the time. These studies also showed no correlation between particle release rates and the direction of motion—horizontal or vertical. They did show significant differences between the kinds of motion—for example, motion of the torso caused more particles to be released through garment openings than did arm motion. It was also observed that maintenance personnel were not always gowned in the same quality garments worn by operators, and this should be a concern in facilities where cleaning personnel work near manufacturing personnel.
Studies on wall cleaning yielded understandable conclusions. The number of particles deposited on a wall increased with the cleaning time interval. The efficiency of cleaning decreased as the cleaning time interval increased, implying that particles became more firmly bonded to the surface with time.
More than one study has shown that re-deposition increases with wipe length. Consequently, one should wipe from the clean side to the dirty side of a surface, and limit the wiping stroke. Most importantly, the quantity of particle deposition on walls was a strong function of nearby tool and personnel activities. Armed with this information, all areas of a facility are routinely cleaned on the same schedule. None of these conclusions would be challenged by anyone who has ever cleaned windows or swept floors. My mother long ago learned when a window was clean by looking through it in the sunlight; she tried different materials and solutions until she perfected her technique. She knew the kitchen floor and the glass near the doorknob needed cleaning more frequently than parlor floor and window, but she cleaned them all every Saturday. She also knew that, if she procrastinated, cleaning was not easy. A janitor knows when to lift a push broom and shake it, and why a broom is wrapped with a cloth. A janitor also learns early on to sweep from the clean side of the room toward the muddy doorway. But didn't we just say this?
The real differences between cleaning home and chip factory are the detection instruments we use, and the particle levels we tolerate. Our eyes are no match for a surface laser counter, but they serve admirably well in our home. A cleanroom wall maintained with Windex and a newspaper would not pass muster, but the level we accept is a subject of discussion. One philosophy demands that cleanroom surfaces be cleaned frequently and maintained as clean as possible. Another school considers cleanroom walls to be impaction surfaces whose role is to collect and remove material from the gas phase; if the particles are on the wall, they're not on the wafer, it is said. This issue cannot be decided without a good model—a model that elucidates the independent relationship between particle levels on surfaces and product yield loss—a model that we do not have. There are those who claim that cleaning improvements can demonstrably increase yield. It is this writer's view that such a simplistic relationship (change in yield to change in wall particle count, with all other factors held constant) cannot be established for such a complex process as high-end semiconductor manufacturing. How clean are your cleanroom walls? How clean should they be? What is the best “solution?”
John O'Hanlon is professor of electrical and computer engineering at the University of Arizona, and co-director of the National Science Foundation Center for Microcontamination Control at Arizona and Rensselaer Polytechnic Institute.