Microsystem Technologies 11 (2005) 282–291
Springer-Verlag 2005
DOI 10.1007/s00542-004-0432-1
Deep photo-lithography characterization of SU-8 resist layers
E. F. Reznikova, J. Mohr, H. Hein
282
Abstract SU-8 layers of different thickness that were
formed by spinning and soft baking or a casting process
with a solvent content of not more than 4% were used for
experimental investigation of deep UV lithography using
modified radiation of a mercury lamp. The specific
absorbance of the SU-8 layers have been measured in
dependence on the wavelength to calculate the power of
the absorbed radiation as a function of depth in the resist
layer. The resist layers which were formed on a photo-
mask have been exposed with a variation of exposure dose
to study the residual thickness of the resist. There are
threshold exposures for the formation of insoluble SU-8
resist layer and for the disappearance of shrinkage of the
resist layer, which depends on the size of the irradiated
area, on microstructure topology and on the resist thick-
ness for fixed parameters of pre- and post-exposure bak-
ing. It has been shown that exact filtration of a low-
intensity band of exposure radiation at 334 nm allows to
reduce strong diffraction distortion in the upper layer of
the resist. Microstructures with an aspect ratio of 25 in a
SU-8 resist layer of 1 mm thickness have been obtained
using a 100 lm thick SU-8 resist layer as a filter for the UV
radiation and with optimal dose of exposure.
1 thick SU-8 resist layer with illustrations of some important
Introduction
Negative SU-8 resist was widely investigated for LIGA
(German acronym for Lithografie, Galvanik, Abformung)
applications [1–8]. As the multi-component resist which is
based on epoxy resin including photo-acid generator
(PAG) molecules is sensitive to near-UV light also high
aspect ratio microstructures would be formed by low cost
photo-lithography [1]. Anyway, the practical application
of deep photo-lithography for such a resist system
encounters the problems of control of geometry and
physical-chemical properties of polymerized resist struc-
tures [5, 6].
The producer of SU-8 resist [6] and optimization
studies on the exposure parameters [7] recommend to use
350–500 nm UV light for improving lateral resolution and
side-wall profile of the resist structure. For this purpose an
Received: 8 August 2003 / Accepted: 6 November 2003
E. F. Reznikova (&), J. Mohr, H. Hein
Forschungszentrum Karlsruhe,
Institut für Microstrukturtechnik,
Postfach 3640, 76021 Karlsruhe, Germany
e-mail: reznikova@imt.fzk.de
exposure set with an optically modified radiation of an UV
lamp is used. The UV exposure process is characterized by
the dose of the incident radiation [4, 7, 8]. Whereas, in
concordance with absorption, reflection and diffraction
laws, the distribution of the dose of absorbed radiation in
the resist layer and the characteristic exposure–develop-
ment behavior in dependence on the absorbed dose
determine the profile of the microstructures for fixed
characteristics of pre- and post-exposure baking of the
resist layer. The control of absorbed radiation allows to
purposefully change the cross-linking extent of SU-8
epoxy pre-polymer molecules and, correspondingly,
physical-chemical properties of the polymerized SU-8
resist layer. The calculation of absorbed radiation is pos-
sible if the exact experimental spectral functions of the
resist specific absorbance and of the power of the exposure
radiation are known. Unfortunately, the numerous effects
and results of different authors which have been obtained
for SU-8 resist by trial-and-error methods can not be
compared with each another quantitatively because nei-
ther spectral intensity of incident radiation nor spectral
sensitivity of used intensity meters have been mentioned
in the literature.
In this paper the parameters of absorbed radiation in a
diffraction phenomena are considered for the quantitative
characterization of photolithography using a modified
radiation of a mercury lamp.
2
Experimental
SU-8 resist with different solid contents produced by
MicroChem Corporation were used to form the resist
layers.
For studying the specific absorbance of SU-8 layer in
dependence on wavelength SU-8(5) layers of different
thickness were coated on 4¢¢ photo-mask quartz glass. A
Perkin Elmer UV/VIS Spectrometer Lambda 2 with an
absorbance limit of 3 and a resolution of 1 nm in a region
between 200 nm and 1000 nm was used to measure the
absorbance.
SU-8 layers up to a thickness of 200 lm were spin
coated onto the substrate by a Hamatech spin coater in a
two stage spinning process. For pre- and post-exposure
baking we used the usual process regimes described in Ref.
[6]. To form a layer thickness less than 4 lm the SU-8 (5)
resist was diluted by c-butyrolactone with fixed weight
relation. Layers thicker than 200 lm were formed by
casting the resist on the photo-mask quartz glass placed on
a Hamatech hot-plate heated in stages: first at 60–65 C multiplied by a refractive index n ¼ 1:67 that was deter-
and then at 95 C. To determine the solvent content of mined in Ref. [5] for the unexposed layer. The measure-
SU-8 layer weight measurements (Mettler Toledo balance) ments of thin layers was made by an interference
of the initial resist layer mr and the dried resist layer mL microscope ERGOLUX with an accuracy of about 2%.
after soft baking or casting process were performed with A scanning electron microscope (JEOL JSM 6600) was
an accuracy of 5 · 10-5 g. The solvent content v is used to observe microstructures with high aspect ratio and
calculated in accordance with equation (1): to measure their sizes.
mr
v ¼ 100
v ð1Þ 3
mL S Results and discussion
where vS is the solid content of SU-8 resist. Usually, the
time of resist drying was chosen to achieve a solvent 3.1 283
content not more than 4%. UV exposure parameters of SU-8 resist layer
For studying deep photolithography behavior resist The recorded spectra of SU-8 layers in the UV and visible
layers of 1 mm thickness were formed on the photo-mask region correspond to those reported in [6, 7]. The specific
surface by a casting process and afterwards exposed absorbance l(k) was determined for SU-8(5) resist layers
through this mask. The quartz glass photo-mask had 128 of different thickness L from 0.23 lm to 350 lm in
areas. Each one of them included chromium patterns in accordance with the ratio of absorbance A(k, L):
form of circles and rectangular slits of different sizes from Aðk; LÞ
2 to 400 lm and of positive and negative tone. The dis- lðkÞ ¼ ð2Þ
tance between circles was 100 lm and that between slits L
was equal to the size of the slit. For the UV exposure a Karl The data of absorbance less than the measurement error
Suss MA-6 exposure set with OSRAM mercury short ARC (0.06) were cut off. The result is presented in Fig. 1. The
lamp having an electrical power of 350 W was used as a values of SU-8 layer specific absorbance for the wave-
source of UV radiation. The stability and distribution of lengths of the intensive bands of a mercury lamp are given
the total incident radiation power was controlled by Karl in Table 1. They have different orders of magnitudes.
Suss UV Intensity meter of model 1000 with an America The function l(k) shows a maximum of absorption at
P405 detector having an accuracy of 0.1 mW/cm2. Another 231 nm, 279 nm and 285 nm. It is known that triar-
nonreflective and lightproof mask with 1 cm2 opening was ylsulfonium hexafluoroantimonate salt is used as PAG in
disposed on the quartz glass surface. It was displaced in the SU-8 system [9, 10]. This salt has absorption bands at
the plane of the sample in order to vary the exposure dose 231.5 nm, 268.5 nm and 276 nm in methanol solution
for the different areas. The exposure with different doses
through other photo-masks with various chromium pat-
terns were used to study the microstructure profile for
different geometries. In this case SU-8 resist layers were
formed on silicon wafers of 4¢¢ or 6¢¢ diameters by the
technique mentioned above and exposed with different
doses. The post-exposure bake (PEB) of the samples was
performed at 95 C. The PEB duration was 30 min for
SU-8 layers of a thickness between 100 lm and 1 mm and
between 20 to 10 min in dependence on layer thickness for
layers thinner than 100 lm.
The development of the exposed resist layers was car-
ried out in propylene glycol monomethyl ether acetate
(PGMEA) with slight agitation in two steps. The duration
of the first step was equal to a time necessary for total
dissolving of the unexposed area of the SU-8 resist layer.
At the second step, the sample was placed in a bath with
fresh PGMEA for the same time. After development, the
sample was rinsed in isopropyl alcohol with slight agita-
tion during several minutes and air-dried in an oven at
40 C. It should be noted that for all development-rinsing-
drying operations the samples were placed on a special
holder with the resist layer downwards. Fig. 1. Specific absorbance of SU-8(5) layer in dependence on
A long scan profiler Tencor P-2 was used to measure the wavelength
micropattern thickness up to 250 lm with an accuracy of
25 Å. The measurements for a lager thickness of micro- Table 1. Specific absorbance of SU-8 layer for the intensive
patterns were performed by the optical microscope ER- bands of a mercury lamp
GOLUX with the help of a Heidenhain MT25 micrometer
with an accuracy of 0.5 lm. In the case of a SU-8 layer k, nm 313 334 365 405 436
without patterns the measured optical length was l(k), lm)1 1.19Æ10)1 5.85Æ10)2 2.71Æ10)3 1.41Æ10)4 6.90Æ10)5
 [11]. The shifts of the maximum values can be explained
by different intermolecular interactions of PAG molecules
with other ones in the SU-8 solid layer and in the solution.
The spectral intensity of the output radiation I0(k) of
the MA-6 exposure set was measured in Karl Suss Com-
pany at the photo-mask plane. It is represented as a sum of
Lorentzian peaks with an accuracy of 0.1% in Fig. 2 for
SU-8 resist layer thickness L ¼ 0. This spectra is typical
[12], when the lamp radiation is filtered by an exposure set
optical system. It consists of strong intensity bands at
365 nm, 405 nm, 436 nm and a weak band at 334 nm. The
284
value of total intensity of MA-6 output radiation in the
range of 300–470 nm is 24.36 mW/cm2 and it is reduced
during the lamp operation time. Therefore, the current
measurement of the lamp intensity by a calibrated inten-
sity meter can be determined as IðkÞ ¼ p  I0 ðkÞ where p is
an intensity factor. It should be noted p is considered as an
independent parameter on k during recommended oper-
ation time of the lamp. In our experiments the values of p
were in the range of 0.99 ‚ 0.78.
The spectral intensities of the radiation that transmits
through SU-8 resist layers with different thickness I(L,k)
are described by an exponential law and are shown in
Fig. 2 also. The radiation of the band at 334 nm and
shorter wavelengths are already absorbed in thin resist
layers. In thicker layers only an absorption of the radiation
of band at 365 nm can be detected. The two last bands
(405 nm, 436 nm) are not absorbed practically. Thus, SU-8
layer can be used as a filter to cut off the radiation of short
wavelengths.
The dose of radiation D(z, t) which is absorbed by a thin
layer of SU-8 molecules on the depth z for exposure time t
is determined by equations 3 to 6. The thin layer can be
defined as a layer of SU-8 molecules with thickness DL
bedded on the depth z of SU-8 resist layer and satisfying
the condition of l(k)DL <<1.
Dðz; tÞ ¼ t  p  ðW0 ðzÞ þ Ws ðzÞÞ ð3Þ
where W0(z) and Ws(z) are SU-8 layer specific powers of
absorbed radiation, the first of which is directed to the
Fig. 2. Spectral intensities of MA-6 exposure set output radiation sure set with 350 W Hg lamp in dependence on the depth of the
transmitted through SU-8 resist layers with different thickness
substrate and the second is reflected from a substrate
surface; p is the above described intensity factor.
Z kb Y
W0 ðZÞ ¼ I0 ðkÞ  TFðkÞ  lðkÞ  e
lðkÞz  ð1
Ri ðkÞÞdk
ka I¼1
ð4Þ
where k is the wavelength of radiation; ka and kb are the
integration limits; I0(k) is the intensity function of the
output radiation of the exposure set; TF(k) is the trans-
mittance of the used filter (without a filter TF(k) ¼ 1); l(k)
is the function of the specific absorbance of resist layer;
Ri(k) is the reflectivity of a surfaces of a filter and/or a
photo-mask and the resist layer.
Z kb
Ws ðZÞ ¼ I0 ðkÞ  TFðkÞ  lðkÞ  e
lðkÞð2L
zÞ 
ka ð5Þ
Y
ð1
Ri ðkÞÞ  Rs ðkÞdk
I¼1
where Rs(k) is the reflectivity of a substrate surface. Rs(k)
and Ri(k) are determined for normal incident radiation
by the Eq. (6) as the reflectivity of the boundary of two
media:
ðn1 ðkÞ
n2 ðkÞÞ2 þ ðK1 ðkÞ
K2 ðkÞÞ2
R12 ¼ ð6Þ
ðn1 ðkÞ þ n2 ðkÞÞ2 þ ðK1 ðkÞ þ K2 ðkÞÞ2
where nðkÞ is the reflective index; KðkÞ ¼ klðkÞ
4p is the
extinction coefficient. Data n(k) and k(k) from ref. [13]
were used for the materials of the substrate and the photo-
mask to calculate the exposure parameters of SU-8 layer.
W0(z) is presented in Fig. 3 for all wavelengths of the
output radiation of MA-6 exposure set and for the
assumption that n(k) ¼ 1.67 for k ¼ 300 ‚ 470 nm. The
specific power of absorbed radiation for the used resist
system and the exposure set is characterized by strong
change almost by an order of magnitude from z ¼ 0 to
Fig. 3. Specific power of absorbed radiation for a molecular layer
of SU-8 resist layer and for the output radiation of MA-6 expo-
molecular layer bedding
z ¼ 100 lm and by a weak reduction for z > 100 lm. This
feature of W0(z) is explained by strong and fully absorp-
tion of energy for the source radiation band at 334 nm by
the top layer of the SU-8 resist; whereas the underlying
layer of the SU-8 resist is irradiated by wavelengths at
365 nm only.

3.2
UV exposure and development behavior of thick SU-8 resist
layer
The exposure parameters of SU-8 resist layers of 1000 ±
60 lm thickness and a normalized residual thickness of
irradiated areas are indicated in Table 2. The areas of the
sample were irradiated for different time t, whereas the
conditions of PEB and development were identical for each
exposure area. The post-exposure baking was done at
95 C during 30 min. Each irradiated area has identical
size of more than 5·5 mm2, therefore the diffraction
phenomena are negligible. The dose of absorbed radiation
was calculated by equations 3 to 6 for z ¼ 0, z ¼ L and
z ¼ L(t) taking into account the internal reflection of resist
surface at z ¼ L, where L(t) is the residual thickness of the
SU-8 resist layer after exposure during time t, after PEB
and after development processes. The dependence of
normalized value L(t)/L on D(z, t) which has been
achieved experimentally and which is presented in Fig. 4
has the typical character for a negative resist.
It should be noted that approximately 30 min is needed
to dissolve an unexposed SU-8 layer of 1 mm thickness in
PGMEA. After this time the value L(t) is independent on
development time. This means the residual exposed SU-8
resist layer is insoluble in PGMEA and isopropyl alcohol
and does not undergo a swelling in the solution at normal
conditions. Such exposure and development behavior of
the resist is totally different from the sensitivity curves of
typical positive or negative resists which are described in
[14] and the traces of which can be affected by develop-
ment conditions. The sensitivity curve of the SU-8 resist
layer is determined by the function D(z, t) only for fixed
pre- and post-exposure baking conditions. The observed
exposure and development behavior of the resist layer can
be explained by an existence of soluble and insoluble
Table 2. Dose of absorbed radiation of SU-8 resist layers with
initial thickness L from 940 lm to 1040 lm and the corre-
sponding normalized thickness of the insoluble layers after de-
velopment. Each SU-8 resist layer was exposed through quartz
glass on which it was formed
285
Fig. 4. Normalized value of residual thickness L(t) of SU-8 resist
layer with 1 mm initial thickness in dependence on the dose of
absorbed radiation
phases of SU-8 molecules in the irradiated volume after
PEB. Besides of that, the fact of the dependence of the
resist residual thickness L(t) on the dose of absorbed
radiation at the level of the residual thickness D(z ¼ L(t),t)
is a display of a shrinkage phenomena of the insoluble
phase of the resist layer. The role of the resist layer
shrinkage as the determinant factor for the sensitivity
curve is demonstrated in the next section for an uniform
distribution of the power of the absorbed radiation in the
SU-8 layer depth.
The residual thickness is equal to the initial one L(t) =
L » 1000 lm for an exposure D(z ¼ L) > 60 J/cm3. This
value can be considered as a threshold for the SU-8 layers
when shrinkage effect of insoluble polymerized resist
molecules does not change the initial resist layer thickness
of 1 mm. It is reproducible for a given resist thickness and
the chosen pre- and post-exposure baking conditions
when diffraction phenomena are not taken into account.
On the other hand the value of approximately 1 J/cm3,
which is a cross of the extrapolated dose dependence curve
of normalized thickness and the dose axis can be consid-
ered as a threshold for arising of an insoluble phase of a
polymerized molecule in the negative resist. Thus, the
exposure – PEB– development behavior of the SU-8 resist
layer can be characterized by these two threshold values of
absorbed dose.

Time of Dose of absorbed radiation Normalized 3.3

exposure D(z,t), J/cm3 thickness Shrinkage phenomena
t, sec L(t)/L The shrinkage phenomena of an insoluble phase in a
Z=0 Z=L Z = L(t) partly exposed area of a SU-8 resist layer after develop-
10 14.55 0.359 1.63 0.15 ment is possible to observe for an uniform distribution of
20 29.10 0.708 2.40 0.26 dose in the resist layer depth at a variation of the dose
50 72.75 1.735 4.07 0.43 values within one wafer. Using a 100 lm thick SU-8 layer
100 145.5 3.40 6.20 0.58 deposited on a photo-mask quartz glass as a filter of the
200 291.0 6.68 9.38 0.73 exposure radiation the dose variation from top to bottom
500 727.5 16.25 18.5 0.88 for thin SU-8 resist layers can be reduced essentially. For a
1000 1455 32.3 33.3 0.97 12 lm thick SU-8 layer on a Si wafer, the power of ab-
2000 2910 63.8 63.8 1
sorbed radiation for a factor p ¼ 1 falls from 0.254 ‚ 0.266

286
W/cm3 to 0.247 ‚ 0.258 W/cm3 for the top to the bottom
taking into account a thickness distribution for the SU-8
filter. So, the dose variation is only 3% in case of a stan-
dard Si wafer. It is about 5% for a 12 lm thick SU-8 layer
deposited on a Si wafer with thermal oxide.
The used scheme for a dose variation during an expo-
sure of the layer is shown in Fig. 5a. The wafer with the
SU-8 layer is exposed through the lightproof mask with
round holes. The central part of each hole of the mask was
covered by a lightproof square after a certain time of the
exposure of the wafer. This results in square areas with
variable doses located inside areas with maximal dose
which was chosen more than 60 J/cm3 identical for each
square area with a variable dose. The procedure was
necessary because the partly exposed areas showing strong
shrinkage can lose adhesion to substrate without being
stabilized by the fully exposed ring which is stable after
Fig. 6a–f. Profiles of 10 min developed SU-8 structures which
were exposed in accordance with the scheme in Fig. 5a. Maximal opment, f 29 J/cm3
dose was 60 J/cm3. Variable doses were the following: a 1.5 J/cm3,
Fig. 5a,b. a Scheme of expo-
sure of SU-8 layer with use of
lightproof mask with round
holes and additional lightproof
squares which are put conse-
quently in order to dislocate
area with variable dose inside
area with maximal dose. b
Example of optical microscope
image of the area after devel-
opment of the wafer with
12 lm thick SU-8 structure
which is indicated at the top of
the scheme
development. An example of an obtained SU-8 structure
viewed by an optical microscope is presented in Fig. 5b.
The cracking central square of this structure was exposed
with a dose of 12 J/cm3 and the dose of the ring was 60 J/
cm3.
The profile of each structure on the wafer was measured
by means of the long scan profiler TENCOR P-2 (Fig. 6).
Pre- and post-exposure baking conditions and develop-
ment condition were identical for each structure with a
variable dose on the wafer. The pre-exposure baking of the
used layers was done at 95 C for 5 min that results in a
solvent content of about 1%. The normalized value of the
residual thickness L(t) of the SU-8 resist layer with 12 lm
initial thickness in dependence on the absorbed dose for
different time of development and for different duration of
PEB are presented in Fig. 7. It should be noted that 20 s of
development are sufficient for a transition of 12 lm thick
b 2.4 J/cm3, c 5.2 J/cm3, d 9.0 J/cm3, e 15 J/cm3 at 15 h devel-

Fig. 7a,b. Normalized value of residual thickness L(t) of SU-8
resist layer with 12 lm initial thickness in dependence on the
dose of absorbed radiation for different time of development a at
15 min PEB and for different duration of PEB b at 10 min
development
unexposed resist layer into PGMEA solution. After this
time a viscous liquid remains within the central part of the
structures which were exposed with a dose near to 1 J/cm3.
It looks like a silt and wastes the substrate surface if to
blow off the wafer, and its thickness can not be measured.
This viscous liquid is dissolved totally after 4–5 min even
without any agitation of the development solution. The
threshold dose for an appearance of a completely insoluble
phase of a polymerized layer is about 2 J/cm3 for 15 min
PEB at 95 C and 2.7 J/cm3 for 3 min PEB (Fig. 7b).
An increase of the development time up to 1 h, 2 h, 5 h,
15 h and 120 h did not cause any changes of the absolute
values of the residual thickness of the SU-8 layers exposed
with doses below  5 J/cm3 and above  30 J/cm3
(Fig. 7a). Some fluctuations of the residual thickness are
observed within 5 ‚ 20 J/cm3 and can be explained by
different motion of polymerized molecules of the soluble
phase through the volume of the insoluble phase into the
developer. The density of the insoluble phase is larger at
the resist layer top than at the layer bottom for the used
287
Fig. 8. SEM micrographs with viewing angle of 45of micro-
structures in SU-8 layer of 1000 lm initial thickness. The layer
was exposed through the Cr photo-mask on which it was formed.
Circle openings of the photo-mask had sizes from 2 lm to
200 lm. The time of exposure was t ¼ 200 s
exposure geometry when the SU-8 layer was deposited on
a wafer. The more dense top of the layer can delay the
going out of the soluble phase, a destruction of the layer is
possible with an increase of the normalized residual
thickness above 1 (Fig. 6e). Values of L(t)/L >1 were not
observed if the SU-8 layer was exposed through the quartz
glass on which it was deposited. In this case the distri-
bution of power of absorbed radiation was the same but
the largest dose was absorbed by the layer near the surface
of the glass substrate.
From these results it can be concluded that for a dose
less than 30 J/cm3 uniformly distributed in the resist layer
(W(z ¼ 0) / W(z ¼ L ¼ 12 lm) < 1.05) the exposed area
of the initial resist layer after post exposure bake consists
of an insoluble and a soluble phase homogeneously dis-
tributed. The most reasonable explanation of the curve
trace is that a volumetric development process occurs with
a simultaneous shrinkage of the insoluble phase of the
polymerized resist layer.
The cracks shown in Fig. 5b are characteristic for SU-8
layers with a normalized residual thickness from  0.96
to  0.99. Their width and density are reduced with a an
increase of dose. The minimal dose when the cracks dis-
appear can be considered as a threshold for the formation
of a completely insoluble SU-8 layer with minimal
shrinkage. It is  30 J/cm3 for 15 min PEB at 95 C
and  50 J/cm3 for a PEB of 3 min only.
3.4
Illustrations of diffraction phenomena
The analysis of the function of the specific power of the
absorbed radiation in a 3-dimensional space of a SU-8
resist layer in accordance with different sizes and forms of
openings and dark patterns of the photo-mask and for an
optically modified mercury lamp radiation will be pre-
sented in our next work. In this study we illustrate the
typical diffraction phenomena of deep photolithography
that are the most important for the top (the resist surface)
288

Fig. 9A,B. SEM micrographs of structures in

SU-8 layer of  1 mm thickness for dark A
and light B cusped angle square patterns on
the Cr photo-mask. The substrate is silicon,
the exposure time is 2500 s. Angles of view
are A 0 and 45 and B 45
and the bottom (substrate surface) parts of SU-8 resist The diffraction distortion in the top layer can be re-
structures. duced essentially if the filter efficiency parameter EF of the
The diffraction influence on the residual thickness of filtrations is less than 0.1. Without a filter EF ¼ 4.65. If a
microstructures of different sizes is shown in Fig. 8. A 100 lm thick SU-8 resist layer on a quartz glass is used as
1 mm thick SU-8 layer was exposed through the photo- a filter the EF parameter will be equal to 0.07. It should be
mask quartz glass on which it was formed by the above- noted that the function l(k) of a SU-8 resist layer is
mentioned process sequence. The incident radiation was changed very slowly in dependence on exposure time.
identical for all circle openings on the Cr photo-mask but Therefore the SU-8 layer can be used several times as an
the residual thickness (height) of corresponding resist effective filter to cut wavelengths shorter than 350 nm of
pillars reduced with reducing diameter of the photo-mask the mercury lamp radiation. The power of absorbed
openings. These insoluble pillars manifest themselves in radiation W(z) for an exposure using a 100 lm SU-8 filter
289
the fact that the distributions of dose of absorbed radiation are shown in Table 3 for the surface layer and the bottom
in the depth of the resist layer distinguish for different layer for example for 1 mm thick SU-8 layer on different
circle openings of the photo-mask. It should be also real- substrates. To set the values the integrand in equations
ized that the observed round-off of the pillar end and the (4, 5) are multiplied by TF(k) ¼ exp(-l(k)Æ100). The values
pillar profile as a whole can be a result of the shrinkage for an exposure without a filter are indicated in Table 3 for
effect of some part of the insoluble irradiated volume of comparison also.
the resist. For an usual geometry of exposure, when the In addition, the roughness of a substrate surface can
resist layer is formed on a substrate, these effects become result in scattering of the incident radiation and in a re-
evident for the bottom of the resist layer. Thus, small size duced power of the absorbed radiation in the bottom of
elements can be absent on a substrate after development or the resist layer (z ¼ L).
they can have bending form and a narrowing to the sub- As for our practice, the use of W(z) calculations and
strate direction. taking into account the measured data of threshold doses
The method of exposing the resist layer through the allow to obtain reproducible results of structure quality for
photo-mask on which it is formed allows together with any SU-8 layer thickness and any typical substrate mate-
the variation of the exposure time to choose experi- rial. The microstructures in 1 mm thick SU-8 resist layer,
mentally the minimal dose of absorbed radiation on the which were obtained using a 100 lm thick SU-8 layer as a
bottom of the resist layer for given thickness and filter for the MA-6 output radiation, are presented in
topology of microstructures and for fixed PEB conditions. Figs. 10 and 11. The doses of absorbed radiation in the
The aim of the method is to determine the exposure time bottom of the resist layers with these structures and that
and, consequently, the dose of absorbed radiation for the structures in Fig. 9 were chosen in the range of
D(z ¼ L) for which the narrowing of the microstructures 90 ‚ 100 J/cm3 in order to obtain maximal aspect ratio
disappears at z ¼ L practically. The absence of shrinkage parameter and microstructure side-walls close to vertical.
effect allows for an usual exposure geometry to obtain In comparison with structures in Fig. 9 these structures
microstructures on a substrate surface with high aspect have a significant shortening of the diffraction distortion
ratio without bending phenomena for microstructures of for cusped angles of photo-mask patterns at the top of
smallest size and without any cracks for large areas of the 1 mm thick SU-8 layer. Diffraction distortion at the bot-
resist layer. tom of the layer can still be seen. Nevertheless, the use of
Nevertheless, this condition results in a formation of a the filtration of MA-6 output radiation allows to obtain the
tun-like profile of microstructures with high aspect ratio in
thick resist layer for the top part of the microstructures.
Besides of this effect the strong diffraction distortion of Table 3. Power of absorbed radiation in the top and in the bot-
cusped angles of photo-mask patterns are characteristic tom of SU-8 resist layer with 1 mm thickness for different sub-
strate materials and for MA-6 exposure set with optically
for the resist top layer (Fig. 9). These diffraction problems modified radiation of mercury lamp of 24.36 mW/cm2 incident
in thick SU-8 resist layer are caused by a large value of intensity I0
power of absorbed radiation for z ¼ 0 ‚ 100 lm (Fig. 3). It
can be reduced by filtration of used modified radiation of Substrate W(z) = W0(z) + Ws(z), W/cm3
mercury lamp.
without a filter with 100 lm thick
SU-8 filter
3.5
Filtration of modified radiation of mercury lamp Z = 0 Z = L = 1000 lm Z = 0 Z = L = 1000 lm
The efficiency of the filtration can be defined by the ratio
thermal SiO 1.454 0.0319 0.187 0.0276
of powers of radiation for the ranges 300 ‚ 350 nm and quartz 1.455 0.0324 0.1872 0.0281
350 ‚ 390 nm that are transmitted through the filter and air 1.455 0.0334 0.1875 0.0287
absorbed by surface SU-8 layer at z ¼ 0: Au 1.458 0.0392 0.1903 0.0339
R 350 Ti 1.459 0.0413 0.1911 0.0358
I0 ðkÞ  TFðkÞ  lðkÞdk Ni 1.459 0.0417 0.1916 0.0362
EF ¼ R300390 ð7Þ Si 1.458 0.0424 0.1906 0.0367
350 I0 ðkÞ  TFðkÞ  lðkÞdk Cu 1.46 0.0428 0.1922 0.0372
where TF(k) is the function of transmittance of the filter Cr 1.463 0.0486 0.1946 0.0421
Al 1.467 0.0592 0.199 0.0512
material for k ¼ 300 ‚ 390 nm
290

Fig. 10. SEM micrographs of

square pillars and through
holes in 1 mm thick SU-8 layer
obtained with the use of
100 lm thick SU-8 filter

microstructures with aspect ratio up to 25 in case of single The SU-8 layer of any given thickness is characterized
openings of a Cr photo-mask. by threshold doses of absorbed radiation for the appear-
ance of an insoluble phase of the layer, that is of the order
4 of 1 J/cm3, and for minimal shrinkage effect of the insol-
Conclusion uble phase. The determinant role of shrinkage phenomena
In this work the influence of parameters of absorbed in the formation of sensitivity curve of SU-8 resist layer
radiation in dependence on wavelength on the structure have been demonstrated for the uniform distribution of
quality are described for deep photo-lithography of SU-8 power of absorbed radiation in the layer depth. The
negative resist. They are more informative in case of minimal dose when the SU-8 layer is developed with the
deep photo-lithography characterization of SU-8 negative normalized residual thickness close to 1 and without
resist layer than parameters of incident radiation be- cracks can be considered as a threshold for the formation
cause they allow to determine the universal dose of an insoluble SU-8 layer with minimal shrinkage. This
thresholds for exposure-PEB-development behavior of threshold dose ranges from 30 J/cm3 to 60 J/cm3 depend-
the resist layer. ing on the distribution of the dose of the absorbed radi-

Fig. 11. SEM micrographs of microstructures with high aspect
ration in 1 mm thick SU-8 layer obtained with the use of 100 lm
thick SU-8 filter
ation in the depth of the resist layer (or the resist layer
thickness) as well as on the processing conditions of the
SU-8 layer.
The increase of exposure time due to diffraction at the
bottom of the resist layer can be determined experimen-
tally by a method of exposure with a variation of exposure
time of the resist layer through the photo-mask on which
the resist layer is formed.
The absorbed radiation dose depends on the resist layer
depth. The value falls down approximately by an order of
magnitude in the range z ¼ 0 ‚ 100 lm and an additional
order of magnitude in the range z ¼ 100 ‚ 1000 lm. This
feature results to strong diffraction distortions at the layer
top for exposure of thick SU-8 layer. Therefore the output
radiation of the exposure set must be filtered; 100 lm
thick SU-8 layer can be used as a filter for exact filtration
of the radiation.
Taking into account parameters of absorbed radiation it
is possible to control the quality of SU-8 resist structures
in case of even a 1 mm thick SU-8 layer. Structures with an
aspect ratio up to 25 and minimal diffraction distortion
were obtained.
References
1. Lee K; LaBianca N; Rishton S; Zohlgharnain S (1995) Mi- 291
cromachining applications for a high resolution ultra-thick
photoresist. J Vac Scien Technol B13: 3012–3016
2. Lorenz L; Despont M; Fahrni M; LaBianca N; Vettiger P;
Renaud P (1997) SU-8: a low-cost negative resist for MEMS. J
Micromech Microeng 7: 121–124
3. Lorenz L; Despont M; Fahrni M; Brugger J; Renaud P; Vet-
tiger P (1998) High aspect ratio ultrathick, negative-tone near-
UV photoresist and its applications for MEMS. Sens & Act
A64: 33–39
4. Lorenz L; Despont M; Vettiger P; Renaud P (1998) Fabrica-
tion of photoplastic high-aspect ratio microparts and mi-
cromolds using SU-8 UV resist. Microsyst Technol 4: 143–146
5. http://aveclafaux.freeservers.com/SU-8.html
6. http://www.microchem.com/products/su_eight_faq.htm
7. Ling Z; Lian K; Jian L (2000) Improved patterning quality of
SU-8 microstructures by optimizing the exposure parameters.
In: Proc SPIE 3999: pp. 1019–1027
8. Flack W; Fan W; White S (1998) Optimization and charac-
terization of ultrathick photoresist films.In: Proc SPIE 3333:
1288–1303
9. Maretial Safety Data Sheet (1997) Microlithography Chemical
Corp
10. Smith GH; Olofson PM (1980) Complex salt photoinitiator.
Patent USA 4231951
11. Crivello JV; Lam JHW (1979) Photoinitiated Cationic Poly-
merization with Triarylsulfonium Salts. J of Polymer Sci:
Polymer Chem Ed 17: 977–999
12. http://www.oriel.com/homepage/down/pdf/02003.pdf
13. Handbook of Optical Constants of Solids. Parts I – III. Edited
by Palik AD, Academic Press, Inc. (1985)
14. Hess DW (1989) Microelectronics processing: chemical engi-
neering aspects. Washington, DC: American chemical Soc