B. Koek et al., PATTERNING ISSUES OF 256MB DYNAMIC RANDOM-ACCESS MEMORY X-RAY MASKS, Journal of vacuum science & technology. B, Microelectronics and nanometer structures processing, measurement and phenomena, 11(6), 1993, pp. 2876-2880
Optical UV lithography has probably the best chance to become the prod
uction technique for 256MB complexity devices [M. Rothschild, Proceedi
ngs of the 37th International Symposium on Electron, Ion, and Photon B
eams, Orlando, FL, 26-29 May, 1992 (unpublished). N. Nomura et al., Mi
crocircuit Engineering Conference, Erlangen, Germany, Date 1992 (unpub
lished) 1. X-ray lithography (XRL) is still a promising alternative, b
ut this technology is in an earlier state of development [S. Yoshida,
MicroProcess Conference, Kawasaki, Japan, Date 1992 (unpublished)]. Ma
ny research centers are putting effort into this to accelerate the dev
elopment. Maskmaking is one of the critical issues in XRL. With increa
sing pattern complexity, not only the accuracy of the lithographic too
l, but the whole path from computer-aided design conversion to lithogr
aphy tool optimization to exposure becomes important. 256MB complexity
pattern data have been converted by using both a flattening and a hie
rarchical convertor. The total conversion time from both convertors is
similar. By maintaining the hierarchy during the fracturing, the hier
archical convertor has more flexibility for pattern manipulation. The
Leica Cambridge EBPG5-HR was optimized and the exposure time of a 256M
B complexity mask was less than 7 h. Four masks were exposed and fully
analyzed. After correcting for substrate drift in the first two expos
ures, the overlay between the second pair of masks was 41 nm (3sigma).
Pattern placement in all masks was better than 25 nm (3sigma).