INSTITUTE OF PHYSICS PUBLISHING
JOURNAL OF MICROMECHANICS AND MICROENGINEERING
J. Micromech. Microeng. 13 (2003) 380–382
PII: S0960-1317(03)55287-1
Building thick photoresist structures
from the bottom up
Mark C Peterman1, Philip Huie2, D M Bloom3
and Harvey A Fishman2
1
Department of Applied Physics, E L Ginzton Laboratory, Stanford University,
Stanford, CA 94305-4090, USA
2
Department of Ophthalmology, Stanford University School of Medicine,
Stanford, CA 94305-5308, USA
3
Department of Electrical Engineering, E L Ginzton Laboratory, Stanford University,
Stanford, CA 94305-4085, USA
E-mail: hfishman@stanford.edu
Received 28 October 2002, in final form 23 January 2003
Published 28 February 2003
Online at stacks.iop.org/JMM/13/380
Abstract
Micromachining has pushed the limits of photolithography in the vertical
direction. We demonstrate a new technique for producing structures in thick
photoresists, with excellent contact alignment and high aspect ratios using a
standard UV source. In this technique, we exposed the resist from the
bottom by defining the mask as part of the substrate. By doing so, we
successfully created high aspect ratio and multiple-layer structures.
Additionally, the structures have a slight inward taper, a very useful
feature for molding. This technique is a new way of thinking about
photolithography, allowing novel structures and simplified processing.
1. Introduction
While much of photolithography drives to produce smaller
lateral dimensions, microelectromechanical systems (MEMS)
push large vertical dimensions and aspect ratios. One of
the more popular thick photoresists in MEMS, SU-8, has
yielded aspect ratios of greater than 15:1 with conventional UV
photolithography [1–3]. Here, we demonstrate a technique to
expose this photoresist from the bottom, providing excellent
contact alignment, high aspect ratio structures with small
features, and a new tool for producing three-dimensional
structures.
There are multiple challenges in photolithography,
especially with thick photoresists. First, underexposure is a
common problem. With thick photoresists, getting enough
energy to the base of the resist can be difficult or timeconsuming. If there is not enough energy to drive the
crosslinking reaction at the base, the photoresist will remain
unexposed and subsequent development will lift the entire
structure from the substrate. Second, aligning a mask to an
underlying feature can be challenging with thick photoresists.
During alignment, the mask is brought into close proximity
with the substrate, allowing features on both the mask and
substrate to be co-focal. If the photoresist is thick, it may not
0960-1317/03/030380+03$30.00
be possible to focus on the mask and substrate simultaneously.
Additionally, an uneven photoresist surface provides uneven
contact during exposure. Sections of the substrate not in
contact with the mask will not resolve small features. Uneven
surfaces are especially a problem with thick films (>100 µm),
where the photoresist may flow during baking steps. When the
photoresist is thick, a small percentage variation in the surface
becomes very significant.
One solution to these problems is to expose the photoresist
through the substrate. By spinning the thick photoresist
directly on the mask, perfect contact alignment is achieved. By
exposing from the base of the structure, the structure will not
lift off during development due to underexposure. Since using
a chrome mask directly as the substrate would be an expensive
R
glass wafers upon which
option, we chose Schott Borofloat
amorphous silicon was deposited. The glass passes the light
necessary to crosslink the photoresist, while the silicon acts as
the mask. This technique will work with any opaque material
that can be deposited and patterned.
The objective of our project was not only to produce
tall structures, but also to produce a structure useful for
molding. For a mold to be useful, the molded material must
be removable, preferably without damaging the master. This
typically requires at least vertical sidewalls, if not an inward
© 2003 IOP Publishing Ltd Printed in the UK
380
Building thick photoresist structures from the bottom up
Figure 1. (a) Process schematic for generating pillars. (i) Start with
1500 Å of amorphous silicon on a glass wafer. (ii) Pattern and etch
the amorphous silicon. (iii) Spin on a thick layer of SU-8
photoresist. (iv) Develop SU-8. (b) Scanning electron micrograph
of pillars. The pillars are 9 µm across at the top and 220 µm tall.
(c) Higher magnification micrograph of the pillar base.
taper from the base to the top. Exposure through photoresist
often causes a taper to occur, narrowing away from the source,
due to the loss of energy as the photoresist absorbs the light.
By exposing from the base, the taper narrows to the top, a
desirable feature for molding.
2. Fabrication
A schematic of the fabrication is shown in figure 1. Four-inch
R
wafers, 700 µm thick, were obtained with
Schott Borofloat
1500 Å of amorphous silicon already in place. The amorphous
silicon was patterned lithographically using 1 µm AZ3612
photoresist. Reactive ion etching (SF6/F-115) transferred the
pattern to the amorphous silicon, after which the AZ3612 was
stripped in acetone. SU-8 2100 (MicroChem Corp.) was
spun on the wafer (100–300 µm) and baked according to the
manufacturer’s specifications. The wafer was inserted into an
aligner upside down, with the SU-8 toward the vacuum chuck
(figure 1(a)). The exposure time was chosen to yield a dosage
150% higher than the recommended dosage, to ensure that the
structures were as tall as possible. The exposures were pulsed,
typically 10 s on followed by 5 s off, to allow the resist to
relax between doses. Development was performed in either
PGMEA (SU-8 developer) or in ethyl lactate (SU-8 thinner)
as supplied by MicroChem Corporation.
For structures with multiple layers, the processing
progressed similarly (figure 2(a)). After the mask was defined
in the amorphous silicon and the photoresist stripped, a layer of
SU-8 2050 (25–50 µm) was spun on the wafer. This layer was
exposed from the top, just as in standard topside lithography,
with alignment to the features etched in the silicon. After
development, a second layer of thick photoresist (100–
300 µm) was spun onto the wafer, and exposed from the
backside of the wafer.
Figure 2. (a) Process schematic showing how multi-layer structures
are generated. (b) Scanning electron micrograph of single-layer and
two-layer SU-8 pillars. Note the smoothness of the sidewalls on the
alignment cross. (c) Higher magnification micrograph at the base of
the two-layer pillars.
3. Results and discussion
The result of this processing is shown in figure 1(b). The
structures are 10 µm across at the base, 9 µm at the top,
and 220 µm tall. A higher magnification image at the base
of the pillars is shown in figure 1(c). These dimensions
correspond to an aspect ratio of 22, but the process can be
pushed further. Deep reactive ion etching can produce similar
results in silicon, but is limited to the removal of a lesser
area of silicon. Our process can remove any percentage of
the surface desired. Additionally, the sidewalls of the SU-8
structures are smooth; there is no scalloping like that seen with
the Bosch process. This is an advantage for molding, where
the scalloped edge would tend to hold the molded material in
place. Other techniques using x-ray or proton sources have
shown excellent aspect ratios (100:1 in >150 µm of SU-8),
but they require advanced sources and equipment [4–6]. If
the amorphous silicon were wet etched (e.g. KOH), then our
technique requires nothing other than a fume hood and UV
source.
The limits to structure size are two-fold. First, limitations
in plasma etching and lithography on the amorphous silicon
produce a lower limit on the lateral dimensions. The
lateral dimensions are defined using standard photoresist and
lithography, so this limitation is no different than any other
lithographic technique. The use of electron beam lithography
and reactive ion etching could yield features smaller than
100 nm. Second, the vertical resolution limit is due to the
diffraction of light passing through the small mask features and
the absorption of light passing through the photoresist. After
381
M C Peterman et al
a certain distance, the light cannot provide enough energy to
crosslink the film, for any exposure time. Because of the
perfect contact between photoresist and mask, our technique
eliminates loss of resolution due to the mask–photoresist
gap.
Beyond the advantages for photolithography in thick
films, this technique also provides the ability to create novel,
three-dimensional structures. Because the mask is part of the
substrate, the mask is an integral part of the device. Some
groups have deposited and patterned metal layers to create an
embedded mask [7]. While our technique is advantageous in
that it does not require metal deposition, a combination of these
two techniques could be very useful. In either case, if wafers
are to be bonded together, exposure can occur after assembly.
Moreover, multi-layer structures can be generated, as shown in
figure 2(b), with a high magnification image in figure 2(c). If
all exposures were done from the top, the alignment between
the thick top layer and the underlying layers would have
been very difficult. In this case, alignment occurred during
processing of the underlying thin layer. Problems with lack of
co-focality were eliminated.
Multi-layer structures such as those shown in figure 2(b)
with a pillar on top of a pedestal have an additional advantage
for molding. The SU-8/SU-8 interface is stronger than the
SU-8/substrate interface. Although exposure from the top
leads to a taper in the wrong direction, the larger contact area
between the pedestal and substrate increases the strength of
the contact, effectively increasing the strength of the narrow
pillar.
4. Summary
The concept of building a photoresist structure from
the substrate up introduces a new way of thinking about
382
photolithography.
While the methodology requires an
additional processing step to etch the amorphous silicon mask,
it provides a tool to make novel structures, with nothing
other than a UV source and etchants. Additionally, the
technique simplifies some of the more difficult processing
steps encountered with thick photoresists: (1) alignment
problems due to focus, surface unevenness, and sticking
and (2) photoresist liftoff because of underexposure. This
approach to photolithography is a powerful new tool for
simplifying the current techniques and building new structures.
References
[1] Lorenz H, Despont M, Fahrni N, Brugger J, Vettiger P and
Renaud P 1998 High-aspect-ratio, ultrathick, negative-tone
near-UV photoresist and its applications for MEMS Sens.
Actuators A Phys. 64 33–9
[2] Pan C T, Yang H, Shen S C, Chou M C and Chou H P 2002 A
low-temperature wafer bonding technique using patternable
materials J. Micromech. Microeng. 12 611–5
[3] Conradie E H and Moore D F 2002 SU-8 thick photoresist
processing as a functional material for MEMS applications J.
Micromech. Microeng. 12 368–74
[4] Bogdanov A L and Peredkov S S 2000 Use of SU-8 photoresist
for very high aspect ratio x-ray lithography Microelectron.
Eng. 53 493–6
[5] Osipowicz T, van Kan J A, Sum T C, Sanchez J L and Watt F
2000 Use of proton microbeams for the production of
microcomponents Nucl. Instrum. Methods Phys. Res. B 161
83–9
[6] Tay F E H, van Kan J A, Watt F and Choong W O 2001 A novel
micro-machining method for the fabrication of thick-film
SU-8 embedded micro-channels J. Micromech. Microeng. 11
27–32
[7] Alderman B E J, Mann C M, Steenson D P and
Chamberlain J M 2001 Microfabrication of channels using
an embedded mask in negative resist J. Micromech.
Microeng. 11 703–5