CD-SEM: Precision or Accuracy?

Alexander E. Braun, Associate Editor -- Semiconductor International, 2/1/1999

With the increased use of advanced lithography
  enhancement techniques such as OPC or PSM,
  accuracy becomes more important.

Precision and accuracy are critical in metrology. To ensure consistent processing, precision is the primary requirement for a process control metrology tool. For years, the industry has avoided the problem of accuracy in CD-SEM metrology, relying instead on precision to provide process control solutions. Now it can no longer afford to do so.

In the days of 0.5 µm technology, errors in estimates of accuracy were acceptable. You would print something, look at it under the CD-SEM, break the wafer and measure from the cross-section to calibrate the accuracy of your top-down CD-SEM linewidth measurement. Due to uncertainties in the location of the cross-section and the magnification calibration of the cross-section SEM, the accepted error range was ~25 nm, which for 0.5 µm work was not significant. However, 0.18 µm is another story; 25 nm would be over half the allowed lithography budget!

With no better method for CD-SEM linewidth calibration, inaccuracies, like line percentages, are getting larger. Accuracy is more important now, because with the increased use of advanced lithography enhancement techniques such as optical proximity correction (OPC) or phase-shift masks (PSM), accurate determination of process and reticle bias is more critical. There are more things to measure, on reticles and wafers, to verify the appropriate correction, and the importance of an accurate measurement of each artifact is increased.

This problem can no longer be ignored.Equipment and user communities must bite the bullet and invest in the creation of standards. So far, equipment makers have used pitch standards to calibrate magnification. But features are not necessarily completely straight sidewalls, and, when working with a CD-SEM, some material interactions are not easily quantified. Thus, it is extremely difficult to estimate true linewidths; pitch is no longer the answer; linewidth standards are.

SEMATECH is working on this, but too slowly mostly due to lack of resources. It needs someone to produce wafers for circulation to suppliers, and this is low in everyone's priorities. It requires extra people, time and equipment: in short, money. A linewidth standard, based on a material commonly used in processing, would be a good beginning. Like the standard meter maintained by NIST, it would allow other standards to be derived from it. SEMATECH funds this aspect of NIST's efforts, and the industry funds SEMATECH. We need funding support to drive the creation of a productized version of the NIST-developed standard.

An all-out effort could bear fruit in two years or less. This is not going to be an instance where the neighborhood's 500 lb gorilla develops a de facto standard that everyone uses. For the most part, individual fabs work off their own 'gold wafer,' which rarely is applicable to another operation or even another process layer.

A lack of standards will not lead to doomsday but will waste an opportunity to shorten development times considerably.

There is a gap in the device-measuring process. Finished devices are needed to close it, so a design is done, placed on a mask, printed, and a circuit with the particular test device is produced. This can take months.

These standards will not be come by cheaply, but with three shrinks a year, if we do not invest now, we will certainly regret it later.