Integrated Metrology: A Faustian Bargain?
Alexander E. Braun, Senior Editor -- Semiconductor International, 12/1/2001
Integrated metrology is like sex: Everybody feels a need for it, everyone considers himself an expert in the field, and nobody wants to pay for it.
Users rightly view integrated metrology with analysis capabilities as APC enablers, but nurture vague visions of a plug-and-play world where tool hardware sockets accept any kind of integrated metrology, although standards are lacking. Hardware interfaces — how to bolt in integrated metrology — have not been defined. Most integrated metrology today results from OEM work with users and is customized, or even designed, to meet requirements. Now users want these input ports to be capable of queuing up material going in or coming out of the tool, and to talk to everything else. Integrated metrology expectations need to be set — it is not meant to replace off-line metrology.
Tools today are designed to do the same thing to every wafer. If the flow is interrupted to do sampling or metrology, potential throughput hits are introduced into the sequencing. Flexibility requires conservative decision-making. Given CD and film thickness integrated metrology capability in a tool, for example, an engineer might set up wafer mapping recipes without considering the impact to wafer scheduling.
Excessive flexibility can create problems in something as simple as wafer scheduling in the tool. Everything works in sequence — a wafer arrives here, then goes there. Users want the capability to bypass wafers in a particular lot, run another lot on the tool and measure two points on the third, fifth and 10th wafer. This would require that modules be bypassed and queued, and most tools are not designed to do this. Complete flexibility will not come all at once. It must be staged with some extra functionality and evolve as OEMs' capabilities do. The ultimate goal is to have the whole software system — a combination of APC and sampling and scheduling — to be self-interactive, enabling the user to adjust throughput, crank it up and reduce sampling.
For true APC feedforward system feedback, software standards for sensor and integrated metrology integration — protocols, data formats, etc. — are needed, and standards task forces are soon expected to report to SEMI. However, hardware is not part of this effort.
As long as the hardware complies with the software, everything should work. The complication is that the software doing the analysis and metrology is not always equivalent from hardware to hardware or from company to company — even with the same hardware. Spectroscopic ellipsometry (SE) is an example. All SE modules generate an SE signal, but the software that analyzes the signal to provide CD, film thickness and profile may not be equivalent.
Users must agree on a standard governing download control and data, and on factory-level integration. Consensus is also needed regarding which data are important and on protocol and format. Users must also decide where integrated metrology goes in the process flow. It could go on a furnace or an etcher and do the measurement before the wafer goes in. It should not go on everything. Calibration is another crucial area — instead of only four stand-alone systems, a user may be looking at 30 or more different calibrations. Agreement on a business model and the value proposition are also necessary: Do you use integrated metrology on all etchers or only two? Who pays for the R&D?
Integrated metrology should be 100% plug-and-play, completely transparent. However, for at least the next five years, this is an unrealistic expectation. To attempt it now is a devil's bargain. A dedicated roadmap is needed.
For additional information on inspection, measurement and test, go to www.semiconductor.net/imt.