Advanced semiconductor circuits, such as DRAMs, are based on very comp
lex fabrication processes. Because of the cost and complexity involved
, it is rapidly becoming impossible to adopt a ''trial-and-error'' app
roach in the development stage of a new process. Fortunately, the adva
nces in computer power spurred by the new semiconductor devices have m
ade it possible to compute the response of complex systems in a reason
able time on workstations. Thus, the study of a virtual representation
of the process (that is, a model) can represent a solution to the hig
h cost of process development-of course, after verification of the mod
el accuracy through controlled experiments. A correct physical interpr
etation of the process under study is necessary in order to implement
a model that is both accurate and extendible. This is particularly tru
e for new approaches, such as those involved in X-ray lithography. We
have studied the process of image formation in X-ray lithography and h
ave implemented several models to predict the intensity distribution a
t the wafer plane. The models can be applied to the definition of an o
ptimal exposure system that will provide the maximum exposure latitude
, and to the study of new types of X-ray masks.