Intelligent second-generation MES solutions for 300mm fabs
06/01/2000
overview
To meet growing requirements for agile fab management, tomorrow's factory control solutions must extend well beyond traditional manufacturing execution systems. They must be proactive, powerful, adaptable, extensible, and automated, with a tighter integration of equipment and information. Today's software-based factory control and automation solutions, particularly for emerging 300mm fabs, are changing to reduce factory ramp and cycle times, and to improve overall equipment effectiveness and yield.
Sukhi Nagesh, Nick Ward, Consilium Inc., Mountain View, California
A key goal for second-generation manufacturing execution system (MES) tools — we will refer to them here as MES II — is to acquire more data throughout a fab and make it much more usable to management. This includes tracking the precise location and processing status of every wafer (work in progress — WIP) in the fab at all times, both product and nonproduct, and recording a complete history of each piece of processing equipment.
 Figure 1. MES is the hub of all software in the fab. |
By contrast, "legacy" MES systems have been largely limited to tracking lot locations. For example, even though a legacy MES knows that a lot is at an etch step, it does not know that this step involves other procedures such as splitting the lot, processing a test wafer, then processing the remaining wafers in the lot if the test wafer succeeds. Lacking wafer-level processing and tracking information, legacy MES systems cannot deal with such extra processing steps within the framework of their limited architecture.
Another goal of MES II is to provide a mechanism to automate as much of manufacturing as possible, from wafer movements to process environments. Closed-loop process monitoring and automatic defect review are emerging as ways to self-correct process conditions for enhanced yield and increased equipment uptime. The MES II system should be able to automate 60-70% of manufacturing, including all process tools in a facility, and provide a pathway for future automation of the rest.
MES evolution
 Figure 2. ROI for a 300mm fab |
Simply stated, MES is a software system that controls the computer and paper systems that in turn manage the manufacturing process on the factory floor (Fig. 1). Today's semiconductor factory management and automation solutions are essentially derivatives and evolutions of a generation of software spanning the early 1980s to the present. They consist of a bewildering variety of components and custom modules from multiple vendors, including chipmakers' internal groups, usually tied together using code written by internal IT, CAM, or automation departments or system integrators. Their architectures are complex and costly to manage, and they are also quite difficult to change in response to new requirements.
These MES installations are largely unplanned; years of additional functions are piled on top of the original systems. Adding new releases to any one of the many components is a lengthy and expensive process, as it must be tested with all the other pieces. Eventually, it becomes extremely difficult — or impossible — to achieve the system's core function: to get a single, controllable definition of the manufacturing operation.
Many fabs today use a commercial MES, which tracks WIP. Module handling dispatching or real-time scheduling often comes from a different vendor. Similarly, another vendor's external recipe manager is responsible for storing and controlling tool recipes. Each specific lot of wafers holds and looks up equipment process recipes based on product or operation context.
The majority of MES systems in fabs today were originally designed for WIP management. Equipment control is a secondary and more recent consideration. It is difficult to link processing activities in the tool with nonprocessing tasks like setup and maintenance. It is even harder to implement predictive maintenance operations effectively or adjust tool recipe parameters while material is being processed. This MES model is not well suited to full automation, the level to which the industry is now switching.
Transition to 300mm fabs
The industry's transition to 300mm wafers is expected to bring cost reductions of 20-30%/unit area of silicon and 30-40% lower cost/chip. The projected $2 billion fab cost means, however, that even small inefficiencies in revenue generation will be costly. Stewart McIntosh, COO at Philips, notes that a $1-billion fab, depreciated over five years, has to cover depreciation costs at the rate of $4 million/week; a $2.4 billion fab, $9.6 million/week [1].
Companies operating in extremely cost-sensitive markets, such as DRAM manufacturers and foundries, are already under intense cost and efficiency pressures.
For those companies that have chosen to compete primarily on manufacturing expertise, the 300mm transition involves many risks. The increase in capital costs from mature 200mm manufacturing technology to relatively unproven 300mm technology, and time-to-market pressure for generating revenue, means that they will have to move quickly along an effective manufacturing curve by optimizing return on investment (ROI) (Fig. 2).
Through their ability to manage a vast array of different manufacturing variables, including much greater interaction with the details of equipment operation, chipmakers envision an important new role for MES II systems in maximizing ROI in 300mm fabs, as well as new 200mm fabs.
MES II systems have the potential to move the industry beyond the operator-driven paradigm of the last 20 years, to an event-driven paradigm characterized by learning cycles (i.e., any learning that takes place in the fab, from managing maintenance cycles to optimizing a process result in a single piece of equipment).
 Figure 3. Of the major categories contributing to OEE, the goal is to increase equipment-adding value while decreasing all other nonvalue-adding categories. |
For example, in a 300mm fab the operator will place a wafer-carrying FOUP on a tool and walk away. From this point, the MES II system should be able to identify which FOUP was placed on the tool and what the tool should do with it. In this scenario, a single event (the FOUP loading into a system) leads to the processing of the wafers in the FOUP. If an incorrect FOUP were placed on the tool, the MES II system should take the necessary business action to notify an external entity (operator, status board, tool alarm, etc.) that an exception occurred.
Automatic data collection also must be an integral part of the MES II solution. Automatic data collection in legacy MES systems involves expensive systems integration among software products from multiple vendors, making these systems hard to maintain and replicate. The MES II system's ability to store and retrieve context-based data in a 300mm fab will be especially important in the process ramp-up and process improvement cycles, and should lead to an increase in overall equipment effectiveness (OEE).
Improving OEE
Much of the improvement and optimization of chip manufacturing refers to the need to improve OEE. This measures the value-added productivity of equipment, expressed as a percentage of the time equipment is actually making product compared to a theoretical maximum. Averaged across the industry, today's OEE is believed to be in the range of 30-50% (Fig. 3). This means that for approximately two-thirds of the time chipmakers are getting no value from a machine from a production viewpoint. Even when individual equipment performance and reliability is high, OEE is usually reduced by:
- reducing equipment downtime through advanced data collection and feeding equipment data into algorithms that can better predict when the equipment will need maintenance, allowing for optimized maintenance scheduling,
- reducing the number of nonrevenue wafers run on the equipment by providing tighter, more predictable control of the process through feed-forward and feedback controls, and automated recipe tuning, and
- providing higher levels of automation and real-time equipment state changes.
- automated handling of nonproduct wafers (test and dummy wafers) that prevents delays, links test wafer process data to product lot decisions, and ensures that non-product wafers are not overused,
- individual wafer tracking,
- real-time information capture and detailed processing history, and
- feedback and feed-forward control in advanced control systems.
- S. McIntosh, "Conquering semiconductor's economic challenges by increasing capital efficiency," a keynote address at the 1997 IEEE International Symposium on Semiconductor Manufacturing Conference.
- T.E. Byrd, A.M. Maggi, "Challenges to plug-and-play CIM," Future Fab International, Issue 3, Vol. 1, pp. 79-81.
- M. Weiss, "300mm tool automation and its impact on fab design and OEE," IEEE/Semi 7th Annual Advanced Semiconductor Manufacturing Conference, 1996, Cambridge, MA, pp. 156-161.
MES II systems address those issues by:
Managing higher levels of automation
With the pending adoption of 300mm wafers, higher levels of fab automation are driving changes in the manufacturing model and in the required scope of MES systems. The most obvious physical aspect involves automated material handling. Beyond that, a new vista of automated processing is coming into view, with higher degrees of intelligent integrated and in situ process control guiding and manipulating process conditions.
 Figure 4. Tracking of CVD temperature differential index during wafer process. |
Material handling and WIP. Prior to 200mm, there was little use of automated material-handling systems (AMHS) in semiconductor fabs, often due to the perception of high cost and lack of flexibility. Human labor was thought to be relatively low in cost, while high in flexibility. With the transition to 200mm wafers, chipmakers recognized that savings due to accurate and timely delivery of product material between workstations made AMHS attractive for interbay, and in some cases intrabay, use.
Because 300mm carriers are large and heavy, ergonomics mandates the use of automated handling for all wafer containers, with zero manual handling [2]. The elimination of manual handling requires delivery of all material to the equipment load port through fully automatic carriers, or person-guided vehicles (PGVs) as an alternative to fully automated intrabay transport [3].
To manage all the material in motion, MES II systems must feature a user interface that completely reveals all existing operations and functions and services to all potential clients, both human and automated. The systems must maintain models of all key components, such as the material-handling system and the process tools, so that instructions to these systems can be tied to WIP and equipment management function. Furthermore, MES II should be able to make intelligent setup decisions, such as delivering WIP and tooling to equipment before processing is definitively committed. Failure to recognize such needs will result in tools waiting on WIP or needing operators to request deliveries manually in advance.
MES II process plans and WIP management models require other new capabilities:
Intelligent components *** In a MES II system, many components need to be modeled to control their functioning. Each of the modeled components possesses state and behavior sub-models. For example, a CVD system could show its state (e.g. "receiving wafers into a load port, Chambers 1 and 3 processing wafers") as well as behavior (e.g., "Process Recipe X being run in Chambers 1 and 3 with process conditions such and such").
The system's model in the software must have an "intelligence" that contributes toward a fab's business processes. Business processes include each process' steps and rules, the fab's equipment scheduling and maintenance rules, collection and analyses of tool operating data, system performance metrics, and many others.
Given the rapidly changing business environments, each MES II component's behavior and state need to be easily accessible to engineers and manufacturing personnel so they can be modified in response to changing requirements. For example, if a cluster tool on the factory floor is re-configured with an additional chamber, the equipment-management component of the MES II system should be easily accessible to the equipment engineer for re-configuration.
Process control and maintenance. One major cause of equipment inefficiency in a fab involves equipment downtime (scheduled and unscheduled), maintenance and repair procedures, and inadequate communication about equipment status. MES II systems can enable integration of the maintenance schedule into the production schedule.
 Figure 5. A typical defect review process. |
Advanced software tools may allow users to judge a tool's continued operation or need for maintenance based on data analysis capabilities embedded in the equipment that can judge its "self health" and issue alerts that describe maintenance requirements. Instead of relying so much on test wafers to detect many equipment drift problems, MES II has the potential to use more fully all the data streams available in the equipment itself and correlate them to process results. Such "smart" maintenance dispatching creates a maintenance and repair schedule that reflects the actual state of the equipment instead of a fixed maintenance and repair schedule drawn up once or twice a shift. Figure 4 shows an example of how a MES system could use sensor data from within a CVD system to trigger an event-based maintenance procedure that can take place before total hardware failure and before process performance reaches out-of-spec levels.
Comprehensive in-line defect monitoring. Equipment and instrument makers now offer defect-detection and analysis technology that enables users to inspect virtually every wafer as it moves through the production line. Once a wafer is inspected, its defects can be automatically classified. The wafer can then be passed to a SEM defect review tool that identifies significant excursions and performs an imaging and material analysis of individual defects to determine their origin.
Integrating defect-review data into MES can relate deviation or failure to event-related reference points such as a specific machine, lot, wafer origin or supplier, etc. This activity has tremendous potential to increase OEE because of its impact on tool availability and repair schedules. Figure 5 shows a typical defect review process flow where the MES II system takes defect information from defect monitoring and statistical process control systems, and when a significant defect or distribution of defects is detected, takes appropriate action. This action could include tool shutdown and placing all lots that were processed on the tool on hold for further analysis.
Conclusion
MES systems have evolved from computerized replacements for paper to control systems that keep a manufacturing body running. With 300mm fabs, we reach the point that the automobile industry arrived at some years ago: one cannot revert to manual systems when automation fails. This has, in part, driven the development of next-generation MES technology — MES II. The complexity of the manufacturing operation requires a MES II system directly to aid in optimizing and ensuring quality production.
This type of next-generation MES integrates the factory management and automation system components with equipment that has greater intelligence — with embedded process control elements that share data with a factory-wide system, and predictive maintenance (i.e., "self health"). This intelligence must be designed to contribute to the fab's business objectives. Having intelligence distributed widely throughout the fab allows a switch to event-based workflow. No longer is the action of an operator the basis for the control system's structure and rationale. This is the meaning of automation. In addition, fab management will have, for the first time, extensive access to data of high granularity. The MES II will use a system-wide data model, shared by all components, allowing management to add or replace components without disrupting the other components.
References
Sukhi Nagesh received his master's degree in mechanical engineering, specializing in control systems. He has worked in the semiconductor MES and automation industry for the last four years. Nagesh is product marketing manager for Consilium Inc./Applied Materials Inc., 485 Clyde Ave., Mountain View, CA; ph 408/584-6225, e-mail [email protected].
Nicholas Ward holds a post-graduate diploma in computing, specializing in manufacturing. He has worked in the semiconductor manufacturing and MES-automation industries for the last 15 years. He is a product management director at Consilium Inc.
Yesterday's first-rate factory control not good enough *** At ARC, we use the term EPM to refer to enterprise production management software whose primary function is to support the plant in managing production. E-business and other market forces are driving the adoption of a new generation of EPM systems that are fundamentally different than earlier manufacturing execution systems (MES). ARC forecasts that global shipments of these systems across all industries will exceed $2.6 billion by 2003. Plant performance that was considered first-rate yesterday will no longer be good enough. Manufacturers need to keep pace with the accelerating speed of business today while dealing with increasing complexity. The new generation of production management systems, including many of those for semiconductor manufacturing, provides tools to do this, while improving flexibility and operational precision.
Gregory C. Gorbach, senior analyst, ARC Advisory Group Inc. (specializing in EPM and e-business solutions)