ELSEVIER Microelectronic Engineering 46 (1999) 287-290
Process Optimisation of DUV Photoresists for Electron Beam Lithography
S. Thoms and D. Macintyre
Nanoelectronics Research Centre
Department of Electronics and Electrical Engineering, The University of Glasgow, Glasgow G12 8QQ, UK
MI~OEI.ECrRONIC
ENGINEERING
This paper describes work carried out to optimise two DUV photoresists for high resolution electron beam
lithography using experimental design techniques. Shipley UV5 XMpositive tone resist and UVN2 negative tone
resist have been studied. Screening experiments were initially carried out to identify significant factors
influencing processing. For UV5 a Box-Behnken design was carried out and used to produce response surface
plots. In-air delay effects were significant for UVN2. Optimised procedures were used to produce 70 nm lines
and spaces in 300 nm thick layers of UV5 and discrete lines of 60 nm in 300 nm thick films of UVN2.
1. INTRODUCTION
In recent years chemically amplified DUV
resists have gained increasing importance in electron
beam lithography largely due to their much improved
sensitivity and dry etch resistance relative to PMMA
[1, 2]. DUV resists have less obvious merits in that
they are now widely used in the semiconductor
industry with the result that they are readily sourced
and meet present day expectations on health and
safety. Many factors influence the electron beam
performance of chemically amplified resists for
instance softbake (SB) time and temperature; post
exposure bake (PEB) time and temperature;
developer time, temperature and concentration; and
various exposure parameters such as dose, beam
voltage and spotsize. Experimental design techniques
help to reduce the number of runs needed to optimise
processes which can be influenced by more than one
factor. This paper describes the use of these
techniques to optimise Shipley UV5 TM and UVN2
chemically amplified resists for high resolution
electron beam lithography applications.
2. EXPERIMENTAL METHOD
Statistical experimental design techniques are
well established and Montgomery [3] gives a good
account of the methodology available.
Initially for each resist a number of factors were
shortlisted for study and screening tests were carried
out to determine which were the most significant.
Fractional factorial designs using five or six factors
were used here. Further work for each resist varied
depending on the initial results.
Silicon substrates were used for all the
experiments and were treated with HMDS prior to
the application of resist. A vacuum hot plate with
temperature control to within I°C was used for all
0167-93t7/99/$ - see front matter
PII: S0167-9317(99)00083-0
bake steps. The exposure system used was a Leica
Microsystems EBPG5-HR100. All pattern
development was carried out using Shipley CD26
developer and immediately followed the PEB. Resist
patterns were inspected using a Hitachi S-900 SEM.
Two types of test pattern were used to evaluate
the resists. The first was an exposure dose wedge
with 20 p.m wide lines which was used to make
sensitivity measurements. The second pattern type
comprised sets of 2 mm long lines which were
suitable for cleaving. Each set of lines was written
with a range of doses. The widths and spaces used
with each resist were different. The measured
responses used to evaluate the resists were: the
process window (defined as the number of dose
values for which the pattern was resolved and fully
cleared); the resist profile; the LDS (defined as the
slope obtained by plotting linewidth against log
dose); the minimum fully resolved linewidths; and
linewidths at a fixed dose.
3. EXPERIMENTS ON UV5
The initial screening test used the factors shown
in Table 1. The ranges were as shown and are
centred around Shipley's recommended processing
conditions. Four centre runs were included to look
for curvature and to provide a measure of process
variance. The test pattern contained 50 nm lines with
200 nm spacings written at doses ranging from 16 to
400 ~C cm-2.
The process window was measured for each
substrate and the contrast of each factor on the
process window was calculated. These are shown in
tabular form in Table 1 while Figure 1 presents the
data graphically as an analysis of means (ANOM)
test [4]. In this the magnitude of the contrasts are
shown together with calculated 0.01 and 0.05
© 1999 Elsevier Science B.V. All rights reserved.