by Robert P. Donovan
The editors at CleanRooms received so many letters in response to Robert P. Donovan's October 1999 Electronics column on page 8 entitled, “The minienvironment mystery,” that this month we will publish the letters and author responses in this space.
A possible solution
I have always enjoyed reading and learning from articles written by Mr. Robert P. Donovan. The most recent article titled “The minienvironment mystery” also greatly attracted my attention. It was very informative and contained a concise description about some advantages of minienvironments. Especially, the “mystery” took me into a deep thinking. Mr. Donovan wrote “the current consensus that the major source of particulate contamination in wafer production is no longer the ambient cleanroom environment or personnel within the cleanroom but rather the production process and the process materials themselves.” Mr. Donovan then mentioned about the “mystery,” “it is not clear to me how a minienvironment reduces these equipment and process-related resources of particle contamination. On a surface, a minienvironment only reduces a contamination sources that is no longer thought to be dominant.”
After carefully analyzing the entire problem, I came to one possible solution to the mystery. Indeed, due to good practice of a strict cleanroom protocol, as pointed out by Mr. Donovan, the major source of particle contamination is the production process and the process materials themselves. Before I discuss about the proposed solution to the “mystery,” I first would like to emphasize that, I think, for a well designed cleanroom with a good cleanroom protocol, the performance of a traditional cleanroom should be as good as a well designed minienvironment. Therefore, the problems mentioned by Mr. Donovan may only exist if a traditional cleanroom was poorly designed or the cleanroom protocol was not followed and caused the flow pattern in the cleanroom to be highly turbulent. Under these situations, some of the particle contaminants generated by the tool may not be washed away immediately through the cleanroom floor. Instead, some of the contaminants are spread out into the big cleanroom environment by the disturbances of air, or big eddies, created by either the poor cleanroom design or improper cleanroom activities. Regional turbulent flows also possess a mechanism to pass particle contaminants to each other, just as they are “practicing baseball,” inside the cleanroom environment. It will depend on the level of disturbances of air that how long and how high concentration of the particle contaminants will “hang in” the big cleanroom environment. Some contaminates attracted in the cleanroom environment end up landing on some wafers and others are eventually washed away through the floor.
Although minienvironments have some distinct advantages, there are many important reasons that still make a well designed traditional cleanroom highly preferred for a big number of semiconductor manufacturers. These factors include liability, cost, speed, and…, etc. Mr. Donald Wadkins had very good descriptions about the reasons in an article, “Minienvironments vs. ballrooms; challenging ULPA filters and more,” published in the May 1997 issue of CleanRooms.
I think this may be only one of the solutions to the mystery and would appreciate to hear comments.
George Lei, PhD, president
Fluid Dynamics Solutions Inc.
Author's reply: Dr. Lei states that a cleanroom of ballroom design allows equipment-generated particles that are emitted into the environment immediately external to the equipment to persist longer in that vicinity than a minienvironment. A minienvironment rapidly sweeps away all particles from the space surrounding the equipment. A ballroom cleanroom does not flush this critical area nearly as effectively or as rapidly. And it is these particles, right next to the equipment and present in high concentrations in a ballroom but largely absent in a minienvironment, that produce the reported differences in wafer particle counts that are so favorable to the minienvironment; that is, the improved cleanliness of wafers processed in a minienvironment is at least partially attributable to the reduction/elimination of the impact of the particles emitted externally by the processing equipment.
Lei's explanation is plausible. It postulates that a minienvironment not only protects process wafers from the general cleanroom environment but more particularly from the ambient air immediately adjacent to the process equipment and whose quality is dominated by the process and process equipment emissions themselves. It questions my implicit assumption that equipment and process related particles are confined to the interior of the process equipment, remote from the external surroundings achieved by a minienvironment.
I think Lei's analysis is the most convincing solution to the “minienvironment mystery” that I've heard. It does leave some questions unanswered such as the mechanisms by which these neighboring, external particles end up on the process wafers. Perhaps during wafer loading/unloading or maybe the particle concentration inside a tool depends in some way on the particle concentration immediately adjacent to its outer walls via in leakage, diffusion or other pathway. Regardless I feel enlightened by Dr. Lei's analysis and I thank him for sharing his thoughts. RD
The difference is wafer handling
I too have considered the minienvironment paradox (CleanRooms, Oct., 1999) and I have a theory that may help explain the difference between ballroom and minienvironment yields. In both cases the process-induced defect levels should be the same, since the same processing tools are used. The consensus is that the ambient is not a major contributor, either for ballroom or minienvironment. I think that the major difference is in wafer handling. Using the robotic SMIF unloader may generate significantly less contamination than the actions of manually opening a wafer carrier, removing the cassette(s), reorienting them to match the wafer handling mechanism, and loading the cassettes. Given that a typical 250-nm process has a least 300-400 process steps, that means that the load/unload process may occur up to 800 times. Over the long-term, it would only take a few fractions of particles per wafer event difference to generate the yield delta that others have observed. In fact, the difference is so small that a typical short-term particle per wafer pass (PWP) measurement might not be sensitive enough to show the difference.
What do you think?
Daren L. Dance, vice president of technology
Wright Williams & Kelly
Author's reply: Daren Dance's explanation is somewhat similar to that of George Lei in that it postulates that a third region must be considered; that is, particle deposition on the product wafers depends on more than just the process interior (region #1) which is assumed to be identical for both the ballroom and the minienvironment; and the external environment remote from the process equipment (region #2), which is assumed to be cleaner in a minienvironment than in a ballroom. The third region is the transition region (region #3) between these first two regions and through which the wafers pass during loading and unloading.
Both Dance and Lei postulate that region #3 is cleaner for a minienvironment than a ballroomDance because the robotic wafer loading/unloading mechanism associated with a minienvironment typically generates fewer particles than the manual methods often used in ballroom environments and Lei because the cleaner region #2 air of the minienvironment also improves the air quality of the region #3 air immediately adjacent to the process equipment through which wafers pass during loading and unloading.
Dance's explanation has the advantage of being readily verifiable by someone with sufficient time and patiencesimply set up the two wafer loading and unloading mechanisms side by side and repeatedly run the same number of wafers through each mechanism until enough particles are deposited to justify a statistically meaningful conclusion. These measurements of wafer handling differences can be conducted in an isolated, non-minienvironment. Verifying Lei's hypothesis requires that a minienvironment be part of one test configuration so that the ambiguities associated with identifying the origin of the particle contributions remain. My guess is that both explanations identify part of the answer to the “minienvironment mystery” raised in my October column. RD
Only the future can tell
Regarding the article, “The minienvironment mystery,” (CleanRooms, October 1999, page 8) by Robert Donovan, I wish to make a few comments:
First, minienvironments were invented to work in the manual handling and operator environment. Thus, most success stories relate to this. In the semiconductor industry, even today, this is the norm. Stocker and storage systems have automated transport and handling, but operations in the bay are mostly manual. Exceptions to this have been a multitude of factories in Japan, where wafer transport and handling is totally automated. Of course, these factories use open carriers and no SMIF. They experience no problems with contamination in the open. Indeed, they believe that with the open approach they would be able to live with the open environment to below the 0.1-micron feature sizes.
Of course, the question remains why use SMIF for 300mm wafer size manufacturing since it is not intended to use manual handling methods at all. SMIF promotion thus became a cost issue. However, evidence for cost advantages is, at best, anecdotal. On the contrary, while large cleanroom costs are fast declining with new technology for air handling, additional robotics, SMIF pods, and minienvironment pass-through handling is growing expensive because of many unresolved issues. We may have brought a curse on the industry. Only the future can tell. One thing is sure. It is not a particle contamination issue!
George W. Horn, applications manager
Middlesex General Industries
Author's reply: Cost is always an important issue. Unfortunately, cost information in the open literature regarding minienvironments vs. ballroom designs is much scarcer than particle data and even yield data. The growing number of sites adopting minienvironments, however, suggests that at present the cost issue is not deterring the minienvironment trend.RD
Robert P. Donovan is a process engineer assigned to the Sandia National Laboratories as a contract employee by L & M Technologies Inc., Albuquerque, NM. His Sandia project work is developing technology for recycling spent rinse waters from semiconductor wet benches.