R
e
Lithography
t
i
c
l
e
I
n
s
p
e
c
t
i
o
n
Leveraging Scatterometry to Enhance STI Etch Process Development Matthew Sendelbach and Andres Munoz, IBM Microelectronics Pedro Herrera, KLA-Tencor Corporation
Scatterometer profiles on line/space arrays are often compared to XSEM images to determine how well they match. However, these comparisons lack precision because they are typically qualitative. This problem can be circumvented by comparing multiple measurements of critical dimension (CD), sidewall angle, and height/depth from XSEM images to scatterometry measurements previously collected from the same locations. A series of etched shallow trench isolation (STI) wafers subjected to a range of etch processes was used in this study. The end result indicates that increased sampling of scatterometry can be used to improve etch process development.
Introduction
Etch process development requires more than just CD measurements. Full profile information showing CD, sidewall angle (SWA), depth, and film thickness is needed. Cross-section SEM (XSEM) is usually used to provide this information because of the detailed images they provide. But XSEM metrology is slow and only provides the profile at a very localized spot. Each wafer is typically sampled using XSEM in only a couple of places, often center and edge. Depending on the XSEM lab workload and the priority of the work, results can take anywhere from several hours for urgent work to several days for normal priority work. Process decisions are then made using this very limited information. Furthermore, XSEM metrology is costly and destructive, so wafers cannot be reused for further process development downstream. Scatterometry has the potential to be an alternative to XSEM metrology because it can also provide full profile information. Although it cannot provide as many subtle details as XSEM metrology, it collects information over a much larger area—tens
of micron in diameter per measurement. This is preferable because it provides greater sampling. In addition, scatterometry can measure at a rate of only a few seconds per measurement. This enables the measurement of every field on the wafer in only a few minutes. Lastly, scatterometry measurements are relatively inexpensive and non-destructive, so the wafers can be reused. But how accurate are scatterometry’s measurements of CD, SWA, and depth? They need to be at least as accurate as XSEM measurements. And how should scatterometry measurements be compared to XSEM measurements in order to determine this? Scatterometry can provide measurements using either library-based or regression-based models. But is one of these more accurate than the other? Or do they provide the same result? This article addresses these questions using a specific case study. Wafer samples and their measurement
Wafer Samples
STI refers both to the series of processes used to form electrically isolating structures in the wafer substrate and to the isolating structures themselves. The nine 300-mm wafers used in this study were part of development work in which various etch processes were evaluated. After etch, the wafers were processed until the Spring 2006
www.kla-tencor.com/magazine
1