Smart sensors: The next evolutionary step in metrology
03/01/1997
Smart sensors: The next evolutionary step in metrology
Over the past 20 years, there has been a steady evolution in testing during IC manufacturing. Initially, testing was performed after an IC was produced, but with increases in process complexity, more testing was required after individual process steps. During this same period, sensor functions evolved from simple measurement of the equipment state to complex measurement of the process state. Now the paths of testing and sensing have converged.
"Smart Sensors" are a new development that will revolutionize IC manufacturing. What makes these sensors "smart" is advanced software that capitalizes on recent developments in sensor technology, extracting process-state and wafer-state information from the raw signal measured by the hardware. In advanced IC manufacturing, it is no longer enough to infer process information by measuring tool parameters such as pressure and flow rate. Measuring process characteristics, such as rate and uniformity, is imperative.
In many cases, even process information is not enough; direct measurement of the wafer state, such as film thickness and film composition, is required during processing. Since the industry sells chips from wafers, it is the final state of the wafers, not the tool, that must be controlled. Smart sensors provide the process and wafer state information necessary to keep the final product in spec.
These sensors create significant new capabilities that improve the cost and quality of IC manufacturing. With smart in-situ sensors, ICs can be tested as they are being manufactured, thereby catching errors earlier, increasing throughput, and decreasing manufacturing costs.
The payback from in-situ metrology sensors is enormous. For example, a smart sensor to measure rate and rate-uniformity could be added to a polysilicon gate etching process. Examination of Overall Equipment Effectiveness (OEE) by both Texas Instruments` Microelectronics Manufacturing Science and Technology (MMST) program and SEMATECH revealed that equipment is making product only 35% of the time. Scheduled and unscheduled calibrations, setup, maintenance, and test take approximately 35% of the time, with test wafers accounting for 8% of the total.
Since an added smart sensor performs key measurements on every product wafer, the number of test wafers can be decreased by at least 25%. This increases throughput of product wafers, saves the cost of test wafers and ex-situ testing, and decreases scrap by detecting faults as they occur. Increasing OEE by 2%, through reduction in the number of test wafers, translates to $25 million in additional annual revenue for a 3000 wafer/week fab (assuming reasonable yields and $10,000 revenue/wafer), and decreases testing costs by $500,000/year (assuming 200-mm wafers at $150 each).
In addition, higher profits will be achieved by measuring every wafer processed. During initial fab ramp-up, when misprocessed lots frequently occur, detecting faults after one or two misprocessed wafers rather than after the next test wafer will save hundreds of thousands of dollars. After the fab is fully operational and the process is stable, measuring every wafer gives a wealth of information for statistical process control (SPC). SPC has increased yields in both the US and Japan, resulting in millions in increased revenue.
Barriers to smart sensors in both existing and future fabs include acceptance, access, and communication. Sensor acceptance is fundamental to the success and growth of the entire semiconductor industry. Since acceptance must be earned, sensor manufacturers must make it easy, tool manufacturers must make it possible, and chipmakers must keep an open mind.
Sensor manufacturers must make sure that their smart sensors can be easily used in R&D and production, since it is in R&D that IC manufacturers will initially use new smart sensor technology. At the same time, IC manufacturers must be willing to try what the sensor community develops in order to reap the benefits down the road. In the future, economic issues will become more dominant and sensor-based manufacturing promises to reduce costs.
Sensor access is mandatory and requires a partnership between equipment suppliers and sensor manufacturers. For sensors to work, equipment manufacturers must design-in access to the wafer and the tool during processing. At the same time, sensor manufacturers must keep tool requirements in mind when designing sensors. Optical sensors are the most promising because they do not affect the manufacturing environment, but both mechanical and electrical access to the process chamber are required for many applications.
Communication between the sensors and the tools is essential for implementation in modern production facilities. Sensors need to know the processing recipe from the tool, while the tools need to get endpoint, go/no-go, and fault information determined by sensors. In many cases, a sensor must also communicate to the fab control computers to facilitate fab-wide SPC. Well-implemented communications will free equipment operators from supervising both tools and sensors.
Growing interest in smart sensors and process control is demonstrated by the establishment of the Control Systems Standardization Working Group and the variety of SEMATECH meetings concerning advanced sensors. Several companies -Low Entropy Systems, Fourth State Technology, High Yield Technology, Luxtron Corp., and Verity Instruments, for example - are dedicated to providing smart sensors. SEMATECH`s Equipment Productivity Improvement Team is a good model for the deployment of smart sensors. Under SEMATECH`s leadership, smart sensors are being integrated with existing equipment to improve standard manufacturing processes and to further understanding of the effect that sensors have on OEE.
Without taking risks and adopting new technologies, our industry will be unable to deliver the increasingly low cost, high quality, and high functionality that are its hallmarks. With smart sensors we can continue to deliver the dramatic results that are the cornerstone of the industry`s reputation and success.
Igor Tepermeister is chief technical officer at Low Entropy Systems in Boston, MA; ph 617/787-2100, e-mail [email protected]