Next Generation 300mm Wafer Handling
With upwards of 4 billion invested to bring a 300mm fab on line, IC manufacturers are eager to ensure speedy returns on investment. Their suppliers, the equipment companies use technology as their primary point of differentiation. Secondary technologies, such as wafer handling both inside and outside the tool, are increasingly outsourced to specialists to ensure focus and efficiency. As the semiconductor industry looks to bring its 300mm capability to full throttle, these trends are driving the area of factory automation to be more efficient.
The argument for moving to 300mm is simple and straightforwardincrease the wafer size to make more die per wafer pass (PWP). But the reality of 300mm poses another daunting challenge co-ordinating the movement of wafers that are now contained in front opening unified pods (FOUPs) to the vast range of tools and inventory management resources. At 200mm, fabs took on the responsibility of linking the material handling systems (either human or automated) to all of the 200-plus tools in the fab. At 300mm the game has changed. Fabs no longer want to be responsible for the interface between the automated material handling system (AMHS) and tools. The equipment front-end module (EFEM) - the newly defined sub-system that integrates the AMHS to the tool - is now part of the toolmakers responsibility and is now seen as their challenge. This includes not only the EFEM, but also all of the associated issues with communication, FOUP handling, wafer handling, environment, identification, and, in some cases, process interaction.
To date, SEMI has played an instrumental role in facilitating this shift. It has defined standards so that both the AMHS suppliers and the tool suppliers have consistent reference specifications for achieving interoperability and integration. By using these standards for the FOUP, the FOUP drop-off and communications, along with best practice guidelines for wafer handling, toolmakers and automation providers were able to develop first generation (1G) EFEMs.
These standards have provided the uniformity essential for fully automated transfer and movement. At the same time, they have allowed for a high level of specification flexibility in the 1G EFEM design. This has resulted in little component or overall EFEM design commonality. In fact, there may be over 50 different EFEM designs in todays fab. Instead of becoming a uniform, almost commoditised system, EFEMs and wafer sorters have a diverse array of design and component methodologies. Both design (or lack of it) and component commonality are having a significant impact on the operational efficiency of 300mm fabs - primarily in areas such as set-up, serviceability and extendibility. At 200mm this was not seen as a major issue, but at 300mm, with billions of dollars invested, it has become a major issue that equipment manufacturers must address.
The 1G EFEM design can be characterised as a box structure consisting of a welded frame with separate sheet metal skins. Added to the box are Box Opener/Loader to Tool Standard (BOLTS) loadports, a Selective Compliant Articulated Robot Arm (SCARA) unit with controller, a fan filter unit (FFU) and a power supply. Other components such as pre-aligners, optical character recognition (OCR), ionisation units, closed loop differential pressure controllers, carrier ID or light curtains are occasionally added based on system requirements and fab preferences. All of these components are off-the-shelf” and thus the control and co-ordination of these components have traditionally been through low-level direct communication from the tool controller.
Extending the time needed to set up a tool directly extends time-to-money for a fab. With the traditional EFEM design, set-up has always been an issue that directly affects the IC makers. Due to the box construction and BOLTS-style loadports, once assembly is complete a significant amount of time is needed to fully calibrate and teach the 1G EFEM. Many times, specialised calibration equipment and training are needed. In addition, this assembly, calibration and teaching process occurs at the place of manufacture of the EFEM, then at the place of integration into the tool and finally during the final installation at the fab. This repetitive, non-value-added loop recurs several times because the 1G EFEM is typically too large to ship as a single unit. The loadports must be removed and shipped separately.
Costs and complexities also exist in the choice of robot. Typically, the robot is a SCARA or complex linkage device. SCARA robots require involved set-up procedures due to the compound angles found in angular motion robots. These robots are also limited in their extendibility to future needs.
Serviceability is another significant challenge for the 1G EFEM. Since the components used are off-the-shelf, packaging is often ad hoc. Controllers and power supplies are either placed in hard-to-reach areas around the robot or high above the FFU. But what makes servicing even more difficult is the stand-alone nature of the components. Many of these components do not have the bandwidth to monitor and facilitate service intelligence - this needs to come, therefore, from more highly specialised field technicians.
Extendibility is an area that is poorly addressed by the 1G EFEM. Many fabs now require EFEMs to support three loadports to ensure operational continuity of material supply into a tool. To add a third loadport, the EFEM design needs an entirely new frame structure. The initial SCARA robot designs typically do not have an extra degree of freedom in the wrist to allow reach into distant loadports. So the robot choice must change to a fixed SCARA robot with yaw axis capability at the wrist or a track mounted arrangement. Regardless of either design, reintegrating a new robot is necessary.
Finally, lack of extendibility also degrades the cleanliness of the mini-environment. Adding a pre-aligner, OCR or ionisation usually means placing irregular structures in the middle of the clean airflow region. New wafer positioning requirements may mandate a track- or SCARA-robot with more yaw capability to position the end-effector into a FOUP or tool station. Beyond the development costs to implement the first order changes, operational changes must also be made. Additional parts need to be carried; training for other designs has to be undertaken; and the technical support infrastructure stretched to cover the new solution.
Asyst Technologies has developed 1G EFEM designs for many toolmakers throughout the industry. Intimately aware of the complexities and challenges, Asyst began designing the next generation EFEM two years ago. During this project, Asyst recognised that the problems seen with the EFEM structure were exactly the same as with the wafer sorter. After all, a wafer sorter is mechanically identical to an EFEM, the only difference being in software control.
While EFEMs have been governed by SEMI standards, there is nothing standard about 1G EFEMs in a fab. Going forward, IC manufacturers have stated that the first priority should be the adoption of a unified product design. The complexity of todays designs is at the root of many problems. With a unified design these problems would be easier to identify and address. This was the basis of the project begun at Asyst.
Simplicity may be achieved by building from the wafer-transferring system outwards, and using best-of-breed designs instead of individual components. Factor in extendibility and scalability, serviceability and manufacturability, as well as an optimised mini-environment, and you would be effectively addressing both production and cost of ownership (CoO) concerns.
Asysts solution is the Spartan 2G EFEM that takes the key aspects of wafer handling and ties them together into a major assembly that ensures easy set-up and operational efficiency. The core” of Spartan consists of a central datum structure (CDS), or backbone, the Wafer Engine and the Loadport 300i. By integrating these three sub-systems together, in both hardware and software, Spartan is smaller, lighter and better performing than a 1G EFEM. Depending on the toolmakers application - implant, CVD, RTP, strip, inspection, metrology, etc. - the mini-environment is tailored to meet reach, handling or option needs. Process toolmakers require front access to their tool and longer reach due to a slit valve into the process chamber. Metrology or wafer sorter solutions aim for smaller, more compact mini-environments. Both of these requirements can be met with the Spartan architecture.
The CDS of the Spartan architecture provides the basis for all wafer-handling componentsloadports, wafer engine and pre-aligner. It consists of computer numerical control (CNC) machined ribs” placed at 505mm intervals (the standard SEMI spacing for loadports) and longitudinal sheet metal channels that create a light and stiff space frame. The Loadport 300i and the Wafer Engine are both mounted and aligned to the CDS in the factory. This alignment is maintained throughout shipment and set-up, thus eliminating the need for time spent re-assembling and re-aligning the Spartan in the field. The CDS is scaleable for two, three or four loadport configurations. By adding additional ribs and longer sheet metal channels, Spartan scales to meet the need for various numbers of loadports or packaging requirements.
The XYZs of motion
In the middle of Spartan is the Wafer Engine robot, given this name because of its high-degree of specificity for wafer logistics/transferring. The Wafer Engine is a track-mounted, Cartesian (x, y, z) motion robot. The track is scaleable to the width required by the toolmakers application. The same Wafer Engine is used regardless of whether there are two, three, or four loadport widths.
Being based on Cartesian (x, y, z) motion, the Wafer Engine is easier to set-up and troubleshoot since each motion direction is coupled to one assembly. This helps eliminate the inherently complex compound axes trajectories found in SCARA robots. The Wafer Engine is equipped with dual end-effectors that enable fast wafer swap times and the movement of two wafers simultaneously for high-throughput applications such as ion implanters.
The design of the Wafer Engine has also led to some significant packaging and service benefits for the overall system. In traditional SCARA robots, the z-range is based on a telescoping design. To achieve full z-range, the robot needs to be twice as long as its z-reach capability. With Spartans z-mast above its base only, the Wafer Engine does not extend to the floor of the EFEM. The benefits are multifold - the system has unencumbered packaging space below the Wafer Engine, the Wafer Engine is 40% lighter than a SCARA robot, the mini-environment volume is reduced by 40% and rigidity and motion precision are enhanced.
In order to gain operational efficiency for both toolmakers and fabs alike, the Spartan architecture has been developed based on standardised assemblies and major components across different versions of Spartan. For example, the Loadport 300i has many of the same sub-assemblies and components found in Asysts IsoPort. By using the same FOUP advance mechanism and door mechanism (minus considerations for packaging the door drive) supply chain benefits are incurred. This has also lead to a greater understanding of certain AMHS and FOUP interoperability issues, all of which have been resolved in the new design.
Due to the compact Wafer Engine design and BOLTS-free mounting of the Loadport 300i, the volume of the clean wafer transfer region is reduced by up to 60% relative to 1G EFEM designs. Each component within the clean region has been evaluated for its impact on airflow and particle generation. For example, air is always drawn inward across any exposed openings to prevent particle migration into the mini-environment. With computational fluid dynamics (CFD) modelling and laboratory testing, the laminar flow approach has been optimised to eliminate particle entrapment. This, along with a closed loop differential pressure controller, ensures an ISO Class 1 environment at all times, leading to excellent PWP performance.
Fig.1: Spartan integrated sorterfor300mm wafer handling
Fig.2: Interior view of Spartan featuring the Wafer Engine