Issue



Industry Insights: The urgent need for integrated metrology


08/01/1999








John Heaton
Click here to enlarge image

Currently, the processing industry can at best achieve about 65 percent overall equipment effectiveness (OEE). The financial reality is that this is not good enough. Figures for equipment availability, performance efficiency, and rate of quality, which make up the OEE calculation, must all be raised.

This challenge sets the stage for an inevitable and logical migration toward integrated metrology and inspection systems. In the past, most metrology and inspection tools were used on unpatterned monitor wafers. This technique proved to be an inadequate predictor of process performance, since monitor wafers were not exposed to many of the processes applied to actual product wafers. For example, to check for added particles, wafers were cycled through a film deposition system without actually depositing a film. The next-generation tools added the ability to measure on actual product wafers, thus allowing the engineer to verify real process performance. But there was a penalty for this capability. The systems were quite slow relative to the actual process, so only a small sampling of wafers could be measured.

In both of the previous methods, metrology was viewed as a non-value-added monitoring step that had no effect on device performance. Today's continuing reductions in feature size and increasing demands on device specifications are squeezing the process window and pushing the limits of tool capabilities. This fact, combined with the widespread adoption of inherently unstable processes such as CMP, is forcing a new monitoring method in which metrology and inspection are put right into process tools. At the same time that the process window is shrinking, skyrocketing fab costs are mandating that yield, equipment utilization, and overall ROI be maximized.

We can see an analogy in modern automobiles, which squeeze out every ounce of performance through the use of computers and sensors. These technologies have decreased braking distance and increased horsepower while improving fuel efficiency. By applying similar strategies and technologies to wafer-processing equipment, IC fabs would also improve performance. Integrating metrology and inspection into the tools would improve fab efficiency by reducing process examination-verification time. Rapid insight into process excursions would also increase yields by allowing finer process adjustments or faster halts in the event of a catastrophic problem.

Let's look at some examples. An integrated particle counter could be used to identify definitively when it's time to PM a process chamber, thus reducing scheduled downtime. Integrated CD monitoring, or film characterization systems, could speed process feedback, thus reducing process equipment standby time. Of course, all of the above systems could be used to immediately detect process excursions that adversely affect yield or quality.

Data from these integrated measurement systems could be used in a closed-loop feedback mode, to allow process tools, or parameters, to be adjusted on a wafer-to-wafer or run-to-run basis, maintaining the tight control required. In fact, this approach has already been taken for CMP, with excellent results. CMP, however, is not the only process where such systems will find application. Several other processes are inevitable candidates. In lithography, integrated CD/overlay measurements could ensure proper exposure and alignment, as well as monitor and control resist thickness. In etch, integrated CD measurements could provide a combination of tight process control and end-point verification with the same metrology tool. In film deposition, thickness, composition, optical constants, and dopant concentration could be monitored and controlled on line.

Beyond process improvement gains, the next most obvious benefit of integrated metrology is a reduction in the use of monitor wafers. It has been estimated that test wafers make up as much as 40 percent of the silicon used by a given fab. Thus, the cost savings from monitor wafer reduction could potentially amount to millions of dollars. At the very least, process tool utilization, OEE, and overall process throughput will be increased.

If the throughput of the integrated metrology tool is sufficiently high, it will allow for the measurement of each product wafer without slowing the process. Deviations will be observed and corrective action taken by the process engineer or by some closed-loop control algorithm in the process tool. Feedback from stand-alone metrology has been shown to take from 30 minutes to as much as two hours to reach the appropriate engineer. This is time enough to create some very expensive scrap. In many cases, saving only one lot of product wafers from being scrapped could be enough to pay for the metrology system.

Estimates indicate 50 percent of all process defects are caused by human material handling. This highlights another benefit derived from integrating the metrology and process tools together. A process step that, in most cases, requires an operator to transfer wafers from one tool to another, thus putting the wafers at risk, could be eliminated. This could have a direct impact on final device yield.

Of course, there are barriers to widespread implementation of integrated metrology. The mindset of the process engineer must shift toward allowing metrology systems to monitor and regulate a process closely. It seems likely that shrinking process windows will provide much of the needed "influence" for this shift to take place. The seamless integration between process and metrology tools that customers require must also be achieved. Only close partnerships that work toward a common goal of delivering a one-stop solution will ultimately succeed. Integration schemes have been tried in the past and have failed because users are confused and don't know who to call, leading to further downtime. Also, since tools haven't been designed to work together, innovative integration schemes and compact designs need to be devised. This has been accomplished by Nanometrics and others, and will evolve and improve as leading-edge companies adopt these new technologies.

Quality must be emphasized because users will not accept an integrated metrology or inspection tool that repeatedly brings a process down. The instruments must be extremely robust with an MTBF significantly longer than the process tools they're integrated into. Generally speaking, process tool manufacturing companies tend to be much larger than metrology equipment suppliers, and have more extensive worldwide service and support organizations. Therefore, they must be willing to accept the responsibility for at least frontline service on the entire integrated system, with backup from the metrology supplier. Again, this necessitates very close cooperation between the two vendors. Proper training and documentation must be provided, along with a detailed field reporting and escalation system.

When these barriers are overcome, the door will be open to a leaner, more efficient manufacturing process. We believe the pioneers who first adopt the new integration solutions will enjoy real cost and technology advantages over their competitors, which can be hard to do in an era of megafabs and advanced foundries. n

John Heaton is president and CEO of Nanometrics, 310 De Guigne Drive, Sunnyvale, CA 94086-3908; ph 408.746.1600, fax 408.720.0196.