Innovative automation approaches for 450mm factories
07/01/2013
MICHAEL BRAIN, MAY SU, ANTHONY BONORA, and DANIEL BABBS, Brooks Automation, Chelmsford, MA
New approaches can help improve cycle time and achieve higher purity inter-process wafer environmental control.
Automation adds significant value by addressing key issues of cycle time, process variability and contamination. This is true for 300mm manufacturing and will be even more critical to enable the transition to 450mm manufacturing. It is important to look at some of the issues facing the industry right now to understand the impact and benefits that fab automation will enable for the future. Current cycle times are unacceptable at 300mm, hindering the full utilization of assets and dulling market responsiveness. As manufacturers transition to sub-20nm process nodes, the queue time between certain process steps becomes more sensitive to variability and more difficult to support with current AMHS architectures. For example, between etch and post-etch clean the timing must be carefully controlled because exposure to moisture and oxygen immediately begins to destroy critical dimensions of features on the wafer surface.
As the industry progresses to smaller process nodes lithography requirements are driving the need for multiple patterning. This doubles or quadruples the number of process steps per mask layer, further extending factory cycle time. In addition, processing larger substrates will likely require longer process times. An increase is most likely in tools like lithography, implant and some metrology steps where process times are based on the area of the wafer. As a result, the first wafer processed in a 450mm batch will wait longer while the remaining 24 wafers in the lot complete processing. After all 25 wafers are returned to the original front opening unified pod (FOUP) the entire lot must wait for AMHS transport to the next process step. The total queue time is too long, with unacceptable variation lot-to-lot and between the first and last wafer processed.
Shorter cycle time and tighter inter-step timing control are possible through innovative automation approaches. An architecture that enables wafers to be protected from oxygen and moisture between process steps will reduce the dependency on critical process timing and can positively impact yield. In the early 1990's standard mechanical interface (SMIF) systems enabled a tighter level of environmental control by reducing the controlled volume of clean air or nitrogen and isolating clean ???zones' with mini-environments. At 450mm, similar techniques of reduced volume can enable creation of a safe and economical wafer environment to protect wafers from particulate contamination as well as moisture, oxygen, and any other contaminate that is potentially harmful to a given process. As was the case with SMIF, an end-to-end solution will be needed such that the wafers are never exposed to an environment that exceeds the maximum allowable exposure limit of contaminants. The solution will again need to be modular to enable flexibility in fab layout, configuration and equipment choices.
Using smaller lot sizes offers tremendous advantages. In a 25 wafer batch, 24 wafers are waiting while one wafer is in the process chamber. Reducing the lot size shaves valuable minutes from this variability, and significantly reduces the overall cycle time. Many papers have shown that equipment utilization goes up and cycle time goes down as lot sizes are reduced. Unfortunately, conventional AMHS systems are not able to support small lot delivery time without bottlenecks, effectively blocking the benefits. A local tool buffer, such as shown in FIGURE 1, can help relieve AMHS bottlenecks delivering FOUPs thereby enabling smaller lot sizes. Localized tool buffers can increase equipment utilization by removing transport from the critical path. Integrated nitrogen purging capability widens the process window so that the next step can be executed within the allowable critical time period.
 |
| FIGURE 1. 450mm sorter/EFEM with integrated tool buffer and nitrogen purge. |
Delivery time variability caused by AMHS can directly impact process uniformity and yield. An approach which links equipment cells directly can reduce variability and remove the burden of high-priority delivery from the AMHS. The cell approach can use a combination of interconnected tool buffers, interconnected vacuum platforms and interconnected EFEMs to accomplish the required lot arrival time and consistency. Ideal interconnection solutions create a virtual cluster with balanced wafer throughputs while maintaining layout flexibility. A modular approach is needed to maintain full application flexibility.
An exciting building block for the cell approach is interconnected EFEMs that can pass wafers directly from one process chamber to the neighboring EFEM, and then into the successive process chamber. This direct wafer exchange does not require FOUP input/output and AMHS operations, thereby reducing critical transfer times and eliminating variability. In this scenario lots are distributed across multiple tools and must later be merged back to maintain FOUP integrity. As always, software must manage and control these wafer movements without misstep.
Interconnected vacuum platforms will connect individual cluster tools together in a manner similar to a common vacuum handler. Wafers never have to leave the safety and purity of a vacuum environment between steps. A combination of these interconnectivity elements provides a flexible cell approach, an example of which is shown in FIGURE 2.
 |
| FIGURE 2. Interconnected vacuum platform. |
Linear tool architectures offer significant benefits, particularly if multiple equipment types can be integrated. In the atmospheric example shown in FIGURE 3, a traditional EFEM removes the wafers from the FOUP and transfers them inside the equipment. A linear robot transfers the wafers between process chambers arrayed in a linear configuration behind the EFEM. After some process steps are completed, the wafers are returned to the FOUP. If the cell is modeled as one tool with a complex recipe, no software modifications are required at the manufacturing execution system (MES) level.
 |
| FIGURE 3. Linear tool architecture. |
Valuable time is lost in 300mm fabs searching for notches in wafers to assure correct rotational alignment relative to crystallographic orientation and to find the laser-scribed wafer ID. Notches are mechanical modifications to the wafer that can impact the ability of the wafer to withstand shock without breakage. One alternative method could involve laser scribing of an alignment mark at the center of the wafer, either top or bottom side. This would allow handling systems to immediately locate the fiducial mark and take action before handling the wafer, thereby saving time and improving wafer yield.
Beyond the front end, assembly and test will also benefit from more automation in the 450mm wafer generation as the back end adds increasing value. Already, requirements for cleanliness, repeatability and error free handling are driving the need for increased automation. It makes sense to leverage the successes of 300mm front end automation to improve the operations of back end fabs. While the technologies will need to be adapted to the unique needs of back end customers, the lessons learned in 300mm automation are significant and should not be lost.
Automation at 450mm is different from previous generations because advanced processes are very sensitive to the environment around the wafers. This pushes automation equipment into the realm of process control. Space constraints in 450mm factories are requiring equipment manufacturers to be creative in developing solutions that provide more output with less space. Automation providers must therefore also be key collaborators in optimizing equipment footprint. The weight and size of 450mm wafer carriers means that automated movement is essential. The overhead involved in moving and storing 450mm FOUPs will be large enough that running small lots in 450mm FOUPs will challenge fab economics. Flexible cell level automation such as presented in this paper may be needed to address these challenges.
In conclusion, innovative automation approaches will be required to fully realize the promise of increased efficiency and reduced costs in the 450mm generation. Because the semiconductor production process is complex and requires a high level of flexibility, no single approach will serve all fabs or all equipment within the fab. A flexible building block approach must be explored to enable each fab to optimize its facility for its unique production requirements.
MICHAEL BRAIN, is Sr. Director, Fab Solutions; MAY SU is Vice President, Strategic Accounts; ANTHONY BONORA is Vice President, Advanced Technology; and DANIEL BABBS is Sr. Director, Strategic Technology Development.