Bonding Tools, Growing With, Not On You!
MEMS manufacturing has numerous features that distinguish it from CMOS fabrication. The obvious one is lower product volumes that lead to other different needs such as flexibility, process standardisation, substrate size and special wafer handling technologies. Therefore the best choice for wafer bonding tools in production MEMS environments should provide a direct path from R&D to manufacturing while remaining versatile enough to handle a wide variety of products adapted to market changes.
First and foremost, it is important to ensure that the technology transfer from R&D to production is unfettered by tooling or mechanical changes between equipment. This should include such items as construction components of the chamber, heaters, cooling systems, wafer chucks etc. Most commercial bonding systems today will satisfy this requirement although some smaller R&D tools can lead to process redevelopment. A quality commercial system is a modular system such as the ABC200 by SUSS MicroTec. This platform uses process modules that are replicas of the standalone systems used for the R&D phase.
When standalone systems are identical to the production module, the transition from development to income will become seamless. A production system typically includes cleaning and/or coating, aligner and bonding modules. The deck layout or platform will hold a given number of process modules, typically up to six for each tool. The layout of the cluster platform is process dependent and herein lies the essence of true flexibility and growth. Let's consider a typical startup production scenario that leads into a full blown foundry operation as an example. Let's also assume that the initial product is an anodic bonded accelerometer.
Stage one of the startup production would only require systems for alignment and bonding. But with rapid growth, the business needs automation. The solutions chosen already at this stage have impacted the future growth and capital expenditures of the budding company. By choosing a module system, the two initial chambers will replicate the R&D systems. For example, in transitioning from a SUSS SB6e standalone bonder to an ABC200 cluster bonder module, the same hardware, wafer chuck design and recipe software is used. The technology transfer engineering staff will simply load the recipes onto the new tool and continue the process. So the initial "production" tool is a cluster platform with two modules and a robot.
Stage two of the production will include a production ramp, and once the utilisation limit of the slowest module is reached, additional equipment will be needed. The beauty of the cluster system is that you still have four remaining slots to populate with process modules. At this time you may also desire to increase throughput to avoid running additional shifts. Some of the options might be to include cooling plates, which can be ordered and installed as the bond chambers are added. Note that aligners are usually the fastest module, taking between 2-3 minutes per aligned bond, so the aligner becomes the CEO of the cluster platform dishing out product to the other chambers as needed.
The addition of extra bond chamber modules is done in the field and generally takes only a few days. Once installed the new bond chamber can be used for the same anodic bond process or for new bond processes including eutectic or metal diffusion bonding. The flexible software of a quality commercial cluster system will allow each chamber to be dedicated to specific bond types to prevent cross-contamination or to be used interchangeably with all other modules.
In the final stage the cluster platform will become fully populated. Assuming that the cluster is still functioning as a production support tool for product, it is reasonable to assume that the equipment is still used for more than one type of bond. The platform should easily transition between bond and substrate types. The software should allow for absolute full module flexibility.
1. As you populate the clusters it should be possible to place any type of module in the tool. For example you may need to add two coaters, one aligner and one bonder or one coater, one aligner and two bonders. If it is not possible to populate the frame as needed in the remaining physical locations then the throughput and productivity will be negatively impacted.
2. The software should allow modules to be skipped or processes to be skipped in the recipe. This might occur for example if the original anodic bond is replaced or partially replaced with a direct bond process in which the bonds will occur in the aligner directly after the cleaning or surface activation step. Thus the materials handling and scheduling software should not assume that the process ends only at a specific type of module. This flexibility is also invaluable should a module need maintenance. In a lock-and-dock system, the module is simply removed from the cluster tool to a maintenance area and production continues.
3. Materials handling systems must be designed to allow for rapid change over between wafer size or type. Auto sizing cassette interfaces are a great advantage during expansion when the fab launch might coincide with a protracted or overlapping substrate diameter change over. These cassette interfaces accept any size of cassette and automatically detect and grip the cassette onto the materials handling interface unit (MHU).
Now let's assume the products go to true high volume production and are relocated off-shore. Or perhaps, the products are going to be produced by a third-party foundry. During the production transfer between companies and/or continents, the permanent layout of the platform will become evident. The challenges will be many, but the reproducibility of tool performance, delivery schedule and qualification times should not be one of them. The tool supplier must have field proven production know-how, with the service and support infrastructure necessary to satisfy 24/7 operations. Things to look out for will include detailed installation and acceptance test protocol (ATP) schedules and procedures, service and support contracts, and access to applications engineers when advice is needed.
In summary, it is fair to say that the argument over when MEMS commercialisation begins may never be resolved. If we continue to believe that commercialisation of a product ends its revision or enhancements, then the ever evolving MEMS products will never be ready for commercialisation. If, however, we recognise our reliance on the millions of accelerometers that provide safety in our daily lives, the RF resonators that guide us to our final destinations or at least allow us to call for directions, we should see that indeed there is a huge production volume that requires flexibility. Cluster platform manufacturing allows for intelligent capital equipment utilisation and growth while retaining the diversity needed to adapt to the changing world that is MEMS production.