There and back again
05/01/1997
There and back again
Anthony Bonora, Asyst Technologies Inc., Fremont, California
Clean manufacturing technology has gone full circle, starting out with localized clean zones, developing into immense, multilevel global clean environments, and returning once more to local clean zones with today`s minienvironments and SMIF technology.
Minienvironments are sealed enclosures installed on wafer-processing tools to isolate wafers, as well as all parts of the tool that can influence process wafer cleanliness (Fig. 1). They create localized clean zones that are isolated from the general fab. When these clean zones are created around wafer cassettes, wafer processing stations, loading ports, and storage locations, the cleanliness of the global cleanroom becomes relatively insignificant. Beyond providing a physical wall to separate the enclosed space from external influences, a minienvironment controls airflow to flush out particles generated within the space as well. Some process tools may benefit from vertical airflow, while others work more cleanly with horizontal airflow. By controlling the source- and exit-points of air in the minienvironment, you can minimize stagnant zones around the wafers.
Standard mechanical interface (SMIF) is a complementary technology that uses standardized, hermetically sealed containers and associated input/output (I/O) products to protect process wafers in transit between minienvironments and during storage. SMIF containers are designed to lock mechanically onto ports integrated with minienvironments to allow loading and unloading of wafers while maintaining the integrity of the environmental isolation (Figs. 2, 3). SMIF and minienvironment technologies are not inextricably linked, but most fabs that employ minienvironments use them together with SMIF.
A global cleanroom to process 200-mm wafers at a rate of 20,000 wafer starts/month typically has a floor area between 3000 and 5000 m2 and a ceiling height of 3-5 m, giving a total volume between 9000 and 25,000 m3 that must be maintained at Class 1 or better. A typical minienvironment, on the other hand, would enclose a volume of around 1 m3 for one process tool. Multiplying by the 300 to 400 different process and metrology tools used in such a fab, the total minienvironment volume is 300-400 m3, a much smaller volume to maintain.
Figure 1. 200-mm wafer fab with integrated SMIF load ports and minienvironments.
Simple fab activities, from opening and closing of facility access doors to operator movements around process tools, generate much of the particulate contamination in state-of-the-art fabs. Minienvironments and SMIF technology control such particle-burst events far more effectively than conventional global cleanroom technology, by isolating the wafer-processing environment from particle-burst sources.
Figure 2. The Asyst-SMIF Indexer directly integrates with atmospheric process tools to provide sub-Class 1 environmental control, wafer position scanning, and Smart-Tag communication.
Separating the actual processing environment from the global environment provides several important benefits. Since roughly 10% of the total clean-air volume must be changed each time it cycles through the system, makeup air filtration requirements are reduced. Relaxed cleanroom protocols make life a lot easier for operators, reducing turnover as well as training costs. Breaking the processing environment into a number of small minienvironments protects wafers in process from contamination through fab mishaps, such as fires, spills, and volatile chemical releases. SMIF standardization also simplifies design of material-handling robots. These factors also make rearranging and upgrading processing equipment much simpler.
New SMIF-based fabs are typically capable of running qualification lots on process tools and debugging process flow while still completing the fab. Many 300-mm manufacturers are expected to use SMIF approaches either partially or completely in their wafer-processing facilities.
Figure 3. Designed with an integral elevating minienvironment for ergonomic loading of vacuum chamber tools, the LPT-2200 features a small footprint combined with a full-function cassette transfer robot.
Historical development
A systematic approach to lean-manufacturing technology has its roots in the precision electromechanical industry for manufacturing product components, such as gyroscope assemblies. However, cleanrooms as we know them did not exist when the semiconductor industry started, more than 40 years ago. At that time, manufacturers used local clean zones, clean benches, and glove boxes. A key development in cleanliness control was the introduction of high-efficiency particulate airfilter (HEPA) technology in the 1960s, which is extrordinarily effective for removing almost all particles. Semiconductor manufacturers of the 1960s applied those techniques of local isolation and filtration to keep diode and transistor processing clean.
A major advance was the use of laminar airflow to ensure delivery of clean air to and around the product. Ordinary turbulent flow randomly mixes the air within a clean zone and can lead to sporadic contamination in the presence of particle sources. In laminar flow, air velocity and direction are well controlled and the possibility of contaminants backwashing onto work in process is minimized.
Clean-facility managers initially used this approach to build horizontal laminar-flow clean benches, which was a breakthrough technology of the 1960s. Human operators worked in front of these benches facing a bank of filters. Air moved through the filters toward the operator, so particles generated by the operator or by the process moved toward the operator and away from the material in process.
The next step was to move from horizontal laminar flow (which required a lot of packaging depth) to building vertical laminar-flow clean benches. Filters and air blowers were mounted in a hood up above the workspace while the operator remained in front of the bench. The main advantages of vertical laminar-flow clean benches were packaging efficiency and reduced airflow obstruction, due to process equipment mounted on the bench. These clean benches would sit in ordinary laboratory/production environments, although some facilities started filtering the air going into those environments to provide a cleaner external zone.
As fab facilities got larger, it was logical and initially cost-effective to start fitting the airflow and filtration system into the building structure. Facilities planners specified blower/filter systems mounted on the wall and blowing toward the work area, or hung from the ceiling and blowing downward. The idea was to dispense with the bench/hood structures by making them part of the facility.
By 1970, transportation of materials between work areas became a significant issue. Filtration units mounted in the ceiling over work areas protected material at a particular process tool, but when operators moved it out of that environment to the next process tool, it got dirty. So, facilities planners added filters over aisleways and hallways, covering more and more of the fab facility with HEPA filters. Federal standards were developed in cooperation with industry to allow classification of cleanroom performance based on particle counts and particle size distribution. The standard currently in use is Federal Standard 209E.
The Class 10,000 clean rooms of the 1970s required 10-15% coverage. That is, 10-15% of the ceiling in the fab was covered with
HEPA filters. By the 1980s, processes for making 2 to 3-?m linewidths required Class 100 or Class 10, and 30-50% coverage. One-micron processes required Class 10 or better and 100% coverage - the only areas not fully covered were in the tool maintenance bays.
Bunny suits
Cleanroom fashions progressed from basic smocks and gloves in the early 1970s to pants-suits and gloves, then to additional footwear. The full-blown bunny suit came into general use in the late 1970s. The operator would step into the bunny suit, which zipped up the front. There were minimal seams and no waist-pullout. The suit had elastic around wrists and an integrated or separate hood with beanie underneath.
Some facilities went to double-suits by the early 1980s. The double suit has the usual outer bunny suit along with an inner layer of full polyester underwear. Operators would have to shed their street clothes before entering the fab, and in some cases take a shower and use a lotion to prevent skin flaking. Then they would don the undersuit, then the oversuit. Some facilities even went to three layers of protective garments.
In the mid-1980s, Dryden Engineering, of Fremont, CA, introduced the idea of a breathing recirculator with a bubble helmet, such as the ones used at Intel, Santa Clara, CA, today. The operator wears a plastic "space helmet" and an exhaust blower pumps his or her breath through a HEPA filter pack strapped to the waist, to prevent exhaled particles from contaminating the fab work space.
Cleanroom engineering
Today`s state-of-the-art Class 1 global or open cleanrooms typically have close to 100% filter coverage over the operator aisles, and tool loading and processing areas. Design factors include the number of operators working in each aisle, operator activities, operator particulate contamination, quality of their cleanroom garments, use of respirators, etc.
It is hard to find employees who want to work under these restrictive conditions. For example, in many cleanrooms, smokers are not allowed. Some facilities forbid consumption of soft drinks for several hours before entering the cleanroom. In some facilities, employees are not allowed to wear cosmetics for a certain number of hours before entering.
Even though restrictions and controls on people have become much more stringent, it is still very difficult to achieve the desired level of cleanliness and, even more important, it is nearly impossible to prevent particle burst events. Within any 24-hr period, a global cleanroom may maintain Class 1 or better for several hours and then a 10-minute burst event may take the local region to Class 10 or Class 100 before settling back down to Class 1. That localized 10-minute burst event can be critical to process quality, negating the hours of steady, ultraclean conditions.
Figure 4. 200-mm SMIF pod and port plate assembly (typical integration for nonvacuum tools).
Despite all these preventive measures, the performance of a global clean facility cannot match that of even a relatively simple minienvironment. In studies comparing the best-managed, state-of-the-art Class 1 cleanrooms to minienvironment-based fabs, the minienvironment-based fabs invariably exhibit much steadier, more predictable conditions.
Minienvironment and SMIF development
In the late 1970s and early 1980s, researchers began to explore alternatives to the global cleanroom. IBM worked on pioneering a concept called the quick-turn-around-time (QTAT) manufacturing facility. Its material-handling system consisted of localized clean tunnels through which process wafers would move while never being exposed to the general room environment. These clean tunnels had HEPA filtration systems, and wafers moved through them from one process station to another on air tracks. The idea was to isolate the material and processing stations completely from the global environment while providing full automation of material transfer.
QTAT was not ultimately successful, partly because floating wafers on the air track resulted in particulate generation by accidental contact with mechanical features. Additionally, the "hard automation" approach lacked flexibility.
Other efforts to utilize isolation principles were undertaken at both Texas Instruments and Hewlett-Packard in the early 1980s. HP, in Palo Alto, CA, developed and patented a sealable, mechanically openable box with a standard interface and latching mechanism. Used in conjunction with a standardized port-plate, it prevented intrusion of particles from the external environment to the clean zone around the process tool during a load-lock transfer. These boxes, used for transport and storage of wafers moving between process tools, were the first of what came to be known as SMIF containers. These containers would latch onto or plug into a port at a given process-tool location. Operators would transfer wafer cassettes through this load-lock into what is now called a minienvironment around that process tool.
The door of the container mates with the mechanical port in the minienvironment, which has the same dimensions as the perimeter (Fig. 4). The outside surface of the box door seals to the outside of the minienvironment port door to create a "particle sandwich." Particles on these surfaces become trapped between them. The mated doors are transferred as a unit into a clean zone, so the clean zone is never exposed to a dirty surface. It only sees surfaces that have been protected inside the container. These fundamental principles of isolated material transfer are currently being applied to systems that utilize either bottom- or front-opening containers.
Dr. Mihir Parikh, the manager of the automation lab that developed the first generation of SMIF products, later went on to found Asyst Technologies, Milpitas, CA. Under Parikh`s guidance, HP Labs developed the early working prototypes, which demonstrated the feasibility and effectiveness of isolation and captured the interest of people in the industry. HP did not, however, feel that cleanroom technology was one of their core businesses and ultimately granted licenses to parties wishing to commercialize minienvironment technology using their initial patents.
A sealable, mechanically openable container works well for transport and storage, but it was also quickly realized that the minienvironment attached to the tool needs constantly and actively cleaned air. A significant amount of engineering goes into successful minienvironment design; air distribution, tool maintenance access, and vibration control are among the many considerations involved. The first full fab with SMIF and minienvironments was VTC in Minnesota, circa 1986. It is still operating today as part of Cypress Semiconductor.
Minienvironment vendors have started to incorporate direct original equipment manufacturer (OEM) integration as SMIF becomes a more standard product. The tool clean zone volume continues to shrink, so that the zone created on a given process tool is now often only 0.1 m3 in volume.
Standards
In 1984, a SEMI standard was drafted for 100-150-mm SMIF ports that became the foundation for defining all the measurements and the opening means for a standard machine-openable container. This standard was issued in 1987, followed by a standard for 200-mm in 1990, and a 300-mm SMIF standard in 1997. There is already a 400-mm standard under discussion. Today, approximately 25% of all new 200-mm fabs are being built with SMIF, and the availability of effective standards has undoubtedly been a decisive factor in the growing acceptance of SMIF.
Figure 5. Benefits of an Inert Ambient at tool loadlocks are achieved by purging SMIF-Pods with nitrogen in conjunction with ULPA filtered recirculation of nitrogen in an automated cassette buffer station.
Beyond particulate control
The contamination challenge that is awaiting us at linewidths below 0.25 ?m extends beyond sub-?m particulate control. It involves airborne molecular contamination such as water vapor, acid vapors, hydrocarbons, and gases that can dope or otherwise modify surfaces. These molecular contaminants can be present in the air brought into the fab or liberated during semiconductor processing, or evolve from basic materials of construction.
The atmosphere in a typical fab is an alphabet soup of different chemical species and is even influenced by the external airborne environment around the fab. Fabs near low-lying coastal areas may have sodium chloride contamination, while fabs near freeways or airports have to deal with hydrocarbon contamination. Despite strenuous efforts to remove these contaminants, they persist in the facilities.
Within the fab, many process steps are extremely sensitive to trace levels of certain molecules, so process engineers must try to control them. Controlling molecular contamination on a global basis in a fab is impractical, whereas it is possible and is being done with minienvironments.
The most common airborne molecular contaminant is water vapor. Water vapor can be damaging to metallized wafer surfaces and vacuum processing tool performance. It remains as a dominant gas load for hours during pumpdowns of vacuum-processing chambers and can influence the nucleation of particles and the regrowth of native oxide on silicon in an uncontrolled manner.
These problems can be controlled with a minienvironment by introducing a relatively inert gas, such as dry nitrogen, argon, or even dry air, in selected locations (Fig. 5). A special port allows a purge of the SMIF container with pure nitrogen, taking the residual water vapor and oxygen levels down to ppm levels prior to opening the container into the loading area of the tool. The loading or processing area of the tool can operate with its own localized circulation of dry air or nitrogen, keeping water vapor contamination down to the ppm level. This greatly reduces water attachment to chamber walls, hot quartzware, wafer surfaces, etc. In some cases, use of carbon filtration or other chemical getters at the entry point of a minienvironment filtration system can add protection by selectively removing certain chemical species.
Right now, a relatively small percentage of tool requirements go in this direction, but they are becoming more frequent, and I think that, for some of the emerging technologies, especially in metal deposition, controlling the environment may enhance yield and advance process technology.
Future developments
Projecting these trends from the past into the future, we can predict probable development trends for minienvironments and SMIF technology:
1. Molecular contamination control applications will increase, and may become a primary enabling technology beyond 0.18 ?m.
2. SMIF-based, 200-mm fabs will proliferate, with SMIF retrofits of the existing 200-mm fab installed base augmenting new design wins. The worldwide acceptance of isolation-based facilities for 300-mm provides further evidence that minienvironment technology is a superior manufacturing approach.
3. Fab-flexibility aspects of SMIF will gain recognition. In a traditional fab, replacing or adding process tools is a major contamination event that can force shutdown of part of the fab, or require zoning off the affected section while you do the work. With SMIF, new process tools can be brought online while conducting business as usual on adjacent equipment. This enables dynamic modification of the factory toolset to improve overall fab effectiveness and economic lifetime.
4. Automation at the intra-bay level will be accelerated by using automation-friendly characteristics intrinsic to SMIF.
ANTHONY BONORA received both his bachelor`s and master`s degrees in mechanical engineering from the University of California at Berkeley. He joined Asyst in 1984. Prior to that, he was employed for nine years at Siltec Corporation in Menlo Park, California, where he held the positions of vice president of R&D and general manager of the Cybeq equipment division. Bonora has more than 20 years of extensive engineering experience and holds 19 patents in the areas of semiconductor processing and equipment. Asyst Technologies Inc., 48761 Kato Rd., Fremont, CA 94538; ph 510/661-5000, fax 510/661-5160.