Fabrication of hierarchical pillar arrays from thermoplastic and photosensitive SU-8.

Abstract

By exploiting the thermoplastic and photosensitive nature of SU-8 photoresists, different types of hierarchical pillar arrays with variable aspect ratios are fabricated through capillary force lithography (CFL), followed by photopatterning. The thermoplastic nature of SU-8 enables the imprinting of micropillar arrays with variable aspect ratios by CFL using a single poly(dimethylsiloxane) mold, simply by tuning the initial film thickness of SU-8 on a substrate. The pillar array is subsequently photopatterned through a photomask, followed by post-exposure baking above the glass transition temperature (T(g)) of SU-8. The pillars in the exposed region become highly crosslinked and, therefore, neither soluble nor able to reflow above T(g), whereas the pillars in the unexposed regions can reflow and flatten out. Two developing strategies are investigated after UV exposure of the SU-8 pillar arrays including i) solvent development and drying and ii) thermal reflow to create bilevel hierarchical structures with short pillars and single-level, dual-scaled, high-aspect-ratio (up to 7.7) pillars in a microdot array, respectively.

Full text

Page 1

6/2010
www.small-journal.com
D15063
Volume 6 · No. 6 – March 22 2010
Fabrication of Hierarchical Pillar Arrays from Thermoplastic and
Photosensitive SU-8
S. Yang et al.

Page 2

Lithography
Fabrication of Hierarchical Pillar Arrays from
Thermoplastic and Photosensitive SU-8
Ying Zhang, Chia-Tai Lin, and Shu Yang*
By exploiting the thermoplastic and photosensitive nature of SU-8
photoresists, different types of hierarchical pillar arrays with variable aspect
ratiosarefabricatedthroughcapillaryforce lithography(CFL),followedby
photopatterning.ThethermoplasticnatureofSU-8enablestheimprintingof
micropillar arrays with variable aspect ratios by CFL using a single
poly(dimethylsiloxane) mold, simply by tuning the initial film thickness of
SU-8 on a substrate. The pillar array is subsequently photopatterned
through a photomask, followed by post-exposure baking above the glass
transition temperature (Tg) of SU-8. The pillars in the exposed region
become highly crosslinked and, therefore, neither soluble nor able to reflow
above Tg, whereas the pillars in the unexposed regions can reflow and flatten
out. Two developing strategies are investigated after UV exposure of the
SU-8 pillar arrays including i) solvent development and drying and
ii) thermal reflow to create bilevel hierarchical structures with short pillars
and single-level, dual-scaled, high-aspect-ratio (up to 7.7) pillars in a
microdot array, respectively.
1. Introduction
Bioorganisms frequently display highly ordered hierarch-
ical structures over multiple length scales.[1–3]They offer
inspiring marvels to materials scientists as the hierarchical
organizations provide remarkable materials properties as
beautifully exemplified in self-cleaning lotus leaves,[1,4]dry
adhesionofgeckofoothairs,[2]andiridescentstructuralcolorin
butterfly wings.[3]Taking a cuefrom nature, manygroups have
exploited various bottom-up strategies, such as phase-separa-
tion of polymer films in different solvents,[5–7]assembly on
colloidalparticles,[8,9]andlayer-by-layerassemblyofpolyelec-
trolytes[10]
to createcomplex,
However, it remains challenging to fine-tune the geometry
features at different levels with high precision using a bottom-
up process. In contrast, top-down lithographic processes are
multileveled structures.
well-known for high controllability over the structure details,
but are limited to essentially two dimensions.
Among many hierarchical structures, high-aspect-ratio
(HAR) pillar arrays are of particular interest because of their
large surface area, high linearity in mechanical response to
external force, and true surface topography. Researchers have
generated varioushierarchical HAR structures for biomimetic
superhydrophobic surfaces[4,11,12]and tunable dry adhe-
sion,[2,13–15]microneedles for transdermal drug delivery,[16,17]
tissue engineering,[18]force sensing and actuation,[18–20]and
particle-trapping,[21]where surface topography plays impor-
tantroles.Single-scaleHARstructureshavebeenfabricatedby
a number of techniques, including photolithography,[16,19,22]
replica molding,[23]
nanoimprint lithography (NIL),[24,25]
capillary force lithography (CFL),[26]and sequential molding
anddrawingofathinpolymerfilm.[27]Inordertosystematically
study the topography effect to the physical properties, it is
necessary to vary the aspect ratios (ARs) at different levels.
Typically,thisisachievedbyvaryingfilmthicknessorusingaset
of different masters. However, when the feature size is
decreased to the submicrometer range and the AR is above
5, because of their small effective stiffness, HAR structures
tend to collapse when i) demolding from the mold due to the
strong adhesion between pillars and mold, ii) gravity[28]or the
full papers
[?] Prof. S. Yang, Dr. Y. Zhang, C. Lin
Department of Materials Science and Engineering
University of Pennsylvania
3231 Walnut street, Philadelphia, PA 19104 (USA)
E-mail: shuyang@seas.upenn.edu
DOI: 10.1002/smll.200901843
Keywords:
? capillary force lithography
? hierarchical pillar arrays
? SU-8
? thermal reflow
768
? 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
small 2010, 6, No. 6, 768–775

Page 3

adhesive force between pillars or pillars and the substrate
become dominant over the elastic energy of pillars,[23,29]and
iii)dryingfromasolventduetocapillaryforce.[30]Furthermore,
introducing hierarchy into HAR structures often requires
multiple steps and complicated procedures,[31–34]which could
furthercompromisethepatternfidelity.Forexample,amixand
match of NIL and UV lithography processes has been
suggested. However, the solvent development step to resolve
the patterns during photolithography process often causes
patterncollapseofHARstructures,[35]whichlimitsthehighest
achievable AR of pillar arrays.
Significant advancement has been made by Suh et al. in
hierarchicalHARstructurefabricationwithCFL.Bycontrolling
capillarity and surface adhesive forces during the demolding
process,theycreatedHARpillarstructureswithvariousARs(up
to >20).[27]Using a two-step capillary molding process with
molds of different sizes and UV curing, they demonstrated
the fabricationofdual-scale
Nevertheless, the fabrication processes rely on careful tailoring
oftheadhesiveforceatthemold–substrateinterfaceanddynamic
tuningofthepolymerviscosityandcrosslinkingdensity.Here,by
combiningCFLandconventionalphotolithography,wereporta
simple,yetrobust,approachtofabricatehierarchicalmicropillar
arrays with variable ARs using a single poly(dimethylsiloxane)
(PDMS)moldbyexploitingthethermoplasticandphotosensitive
nature of a commercially available photoresist, SU-8.
As an ultrathick photoresist, SU-8 has been widely used in
photolithography and X-ray lithography to fabricate high-
density HAR microstructures.[38]The crosslinked SU-8 film
possesses high thermal and mechanical strength (glass transi-
tion temperature, Tg>2008C, and Young’s modulus of up to
5GPa), which are necessary for stable HAR microstructures.
On the other hand, the unexposed SU-8 film has a low Tg
(?508C) and relatively low viscosity above Tg, thus, making it
well-suitedfortheCFLprocess.Here,HARSU-8micropillars
(AR up to 8) were rapidly imprinted within 2min by CFL. By
simplytuningtheinitialSU-8filmthicknessforCFL,wecreated
a series of polymer pillar arrays with different ARs using a
singleHARPDMSmembraneasthemold.Moreover,because
SU-8 is both thermoplastic and photocurable, we could
photopatternthepillararraysthroughaphotomask.Bymixing
CFL and conventional photolithography, which involved the
solvent-development step, we demonstrated bilevel, dual-
scale, hierarchical structures with relatively short pillar arrays
without pattern collapse. However, as the pillar AR increases,
solvent swelling during the development step and capillary
forces during drying of the developer could cause pattern
collapse of HAR pillars. Alternatively, instead of developing
the photopatterned SU-8 pillar arrays in a solvent, we heated
the film above the Tgof SU-8, thus reflowing and flattening the
imprinted pillars in the unexposed regions while maintaining
the highly crosslinked pillars in the exposed regions. With this
novel thermal reflow approach, we developed single-level,
dual-scale HAR structures.
polymerstructures.[33,36,37]
2. Results and Discussion
In a typical CFL process, a patterned elastomeric mold is
placed onto a thin layer of liquid either in melt or in solution
state.Thecapillaryforcedrivesthespontaneousfillingofliquid
into the voids of the patterned mold, replicating the mold
morphology when the liquid becomes solidified upon cooling,
UV crosslinking, or solvent removal. Depending on the
available volume of the liquid film relative to the void volume
of the mold, two scenarios can occur, as shown in Figure 1.[39]
When the liquid volume is greater than the mold void volume,
liquid can completely fill the void and replicate the mold with
identical size and geometry, however, leaving an excess
residue layer. This process is often referred as over-filling.
When the liquid volume is less than the mold void volume,
incomplete filling (or under-filling) occurs, where the residue
layerisminimized.CFLhasbeenwidelyusedtocreatepatterns
for pattern transfer to functional substrates.[40,41]However,
therehavebeencomparativelyfewstudiesontheapplicationof
CFL to create HAR micro- and nanostructures.[27]
To faithfully fabricate HAR structures by CFL, we first
needtoaddressafewissues,includingairbubblesthatcouldbe
trappedinsidethechannelsofthemold,patterncollapseduring
mold release, and possible swelling of the mold by the
imprinting liquids (e.g., monomers and solvent), leading to
deformationoftheHARstructures.HARstructureshavebeen
fabricated by NIL, however, with limited success.[24,42,43]In
NIL, air bubblesareeasilytrappedinthemicro-/nanochannels
of the mold, which are difficult to remove when using a rigid Si
or quartz mold.[25,44]This problem is further aggravated when
theARofthechannelsisincreased.Becauseofthelargesurface
area of HAR structures, which significantly increases the
adhesion between the mold and the replica, release of HAR
structures from a rigid mold could be problematic. Defects
Hierarchical SU-8 Pillar Arrays
Figure 1. SchematicillustrationoftwoscenariosofCFL:overfillingversus
underfilling.
small 2010, 6, No. 6, 768–775
? 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.small-journal.com
769

Page 4

appear readily even when the mold surface
is coated with a low-surface-energy surfac-
tant.[42]PDMS,onthe otherhand,hasbeen
widelyusedinsoftlithographyandCFL.Itis
a highly gas-permeable elastomer. Studies
have shown that N2and O2can be easily
dissipated through the PDMS mold[45]
without building up a negative pressure
thatcouldpreventcapillaryfilling.[46,47]For
the pillar arrays fabricated in our experi-
ments, the capillary filling process was
completedinlessthan2minandnoobvious
air trapping was observed. In addition,
PDMS possesses very low surface energy
(20mN m?1) and is highly elastic, thus
allowing for effortless release of the fabri-
cated HAR SU-8 pillars from the PDMS
mold.
Because PDMS can be easily distorted
under pressure or by swelling of the
imprinting liquid, it is important that the
liquid does not interact with PDMS during
CFL. Previously, by partial polymerization
of the monomers as molding precursors, we
successfully fabricated HAR hydrogel pil-
lars through replica molding via a PDMS
mold without swelling deformation.[48]
Here, SU-8 is a bisphenol-A novolac resin
derivativewithanaveragemolecularweight
of 1397g mol?1. Upon heating to 958C, a
typical soft baking temperature for SU-8
above its Tg(508C) and melting tempera-
ture, Tm(808C), the film begins to flow and
its viscosity is significantly lowered to ?5Pa s, thus suitable for
CFLtopatternvariousstructures.Atthesametime,becauseof
its relatively larger size versus monomers, diffusion of
the melted SU-8 into the PDMS mold is minimized. Once
imprinted, the SU-8 film is cooled down to room temperature
andbecomesglassyagain,allowingforeasyseparationbetween
the imprinted pillars and the soft PDMS mold.
In Figure 2, we show SU-8 micropillar arrays with different
ARsfabricatedbyCFLusingasinglePDMSmold.TheARwas
determined by the available volume for CFL, that is, the initial
SU-8filmthicknesscastonasubstrate.FromanSU-8filmwith
an initial thickness of ?3.7mm, micropillars of diameter 1mm
and height ?7.7mm (AR¼7.7) were obtained (Figure 2a),
close to the original Si master height (8mm). The small
discrepancy in height may be attributed to the thermal
shrinkageofSU-8uponcoolingandtheresiduelayerremained
unfilled beneath the pillars. The residual layer is rather thick,
?1.7mm, which is typical in the case of overfilling during CFL
(Figure 1). In the case of underfilling, where incomplete filling
occurs in the channels due to the limited polymer reservoir
surrounding them, it is possible to vary pillar height and
minimize the residual layer. Figure 2b–h shows a series of
micropillararraysimprintedfromthinSU-8filmswithdifferent
initialfilmthickness(0.25to1.55mm),resultinginpillarheights
rangingfrom0.74to5.85mm(ARof0.74to5.85),respectively,
whilethelateralfeaturesizeremainedas1mm.Meanwhile,the
residuelayeronsubstratewasdecreasedto?117nmforpillars
with AR?1.15 (Figure 2i). We note that the crosslinked SU-8
structureshavebeenwidelyusedasmastersinsoftlithography.
Using the imprinted micropillars with different ARs as new
mastersforPDMSmolds,wefurtherfabricatedasetofnewSU-
8 micropillarsby CFL for
structures.
Themaximumfillingratioofmold,f,canbeestimatedusing
a simple geometric analysis. If the molding material is an
incompressible liquid and the volume conservation rule holds,
f can be calculated from the initial film volume under the mold
(Vfilm)assumingthereisnoresiduelayer.Forthesquarelattice
of micropillar arrays in our experiments, 1mm in width and
2mminpitch,thefractionofvoidprojectedareaonfilmsurface
is wvoid¼0.25. Thus, the theoretical pillar AR can be
calculated by
fabricationof dual-scale
AR ¼H
D¼
T
’void
¼ 4T
(1)
where T is the initial film thickness in micrometers and H and
D are the pillar height and width (or diameter), respectively.
We compared the theoretical and experimental results of the
pillar AR as a function of the initial film thickness. As seen in
Figure 3, the experiments show a linear relationship between
full papers
Y. Zhang et al.
Figure 2. SEM images of micropillar arrays with various ARs, which were obtained from SU-8
filmsofdifferentthickness,T.a,b)T¼3.70mm(AR¼7.7).b)Anenlargedviewof(a),showinga
residuelayerwiththicknessof?1.7mm.c)T¼1.55mm(AR¼5.85).d)T¼1.21mm(AR¼4.5).
e) T¼0.93mm (AR¼3.15). f) T¼0.79mm (AR¼2.17). g) T¼0.38mm (AR¼1.15).
h)T¼0.25mm(AR¼0.74).i)Aclose-upviewofg)atpillar–substrateinterface,showingathin
residue layer with thickness of ?117nm.
770
www.small-journal.com
? 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
small 2010, 6, No. 6, 768–775

Page 5

the pillar AR and the initial film thickness as
AR ¼ 4ðT ? 0:15Þ
(2)
From this linear relationship, we can control the mold void
filling and obtain micropillars with variable ARs. The
calculated ARs are slightly but systematically larger than
the experimental ones. This can be explained by the thin
residue layer that always exists in the imprinting films. By
extrapolating the experimental AR to T¼0, we obtained an
average residue layer thickness of 150nm, close to the
observed minimal value seen in Figure 2i, ?117nm.
A recent study by Lee et al.[49]on low-pressure imprint
lithography suggests that both capillary force and viscous flow
contribute to the initial imprinting stage (in their case, the first
10min with polystyrene (PS) Mn¼100000g mol?1at 1608C),
while capillary forces become dominant as time progresses.
However, we were not able to distinguish between these two
mechanisms in our system because the imprinting time of the
SU-8 photoresist was much shorter in comparison to that
required for a typical IL or CFL, where a very high-molecular-
weight thermoplastic polymer is used. For example, for a
poly(styrene-b-butadiene-b-styrene) (SBS) block copolymer
(molecular weight of 3?105, 30% styrene), the viscosity is
106Pa s at 1008C.[39]In contrast, the viscosity of SU-8 is
measured as ?5Pa s at the imprinting temperature (958C),
which is six orders of magnitude smaller.
According to a capillary filling theory that excludes the
effect of lateral viscous flow,[50]the filling time is
t ¼
2hsh2
Rgcosu
(3)
where hsis the viscosity of the filling liquid, h is the channel
height, R is the hydraulic radius of the channel, g is the liquid
surface tension, and u is the liquid contact angle at the liquid–
PDMS mold interface. For SU-8 to fill a microchannel 8mm
long and 1mm wide, the filling time is estimated less than 0.1s,
which is much shorter than the typical filling time for CFL, for
example ?20min for imprinting PS (Mn¼100000g mol?1)[49]
at 1608C with a pattern depth up to 130nm, and 30min for the
SBS block copolymers to completely fill the void in the mold
withastepheightof600nmandawidthof400nmat1008C.[39]
To ensure the complete filling in our experiments, we left the
PDMS mold on the SU-8 film at 958C for ?2min.
We also investigated the possible viscous flow effect on
liquid filling by perform imprinting experiments at a
decreased pressure. Considering the elastic nature of the
PDMS mold, the imprinting pressure was normally kept very
low(e.g.,5?10?2bar)toavoiddeformationofthemold.When
the imprinting pressure was lowered to 5?10?3bar while the
imprintingtimeremainedunchanged,weobservedadecreased
pattern quality. As seen in Figure 4, completely and partially
imprinted pillar patterns coexisted throughout the film, and
fingeringcouldbeclearlyseenatthesampleedge,similartothe
reports by other groups.[47,49]Based on this observation, we
believeourresultssupporttheimprintingmechanismproposed
Hierarchical SU-8 Pillar Arrays
Figure 3. Comparison between experimental and theoretical ARs
obtained from capillary imprinting SU-8 films with different initial film
thickness.
Figure 4. SEM images of a non-uniformly imprinted SU-8 film when an
insufficient pressure was applied to PDMS mold for conformal contact.
a)Alarge-scaleviewoftheimprintedfilmwithbothcompletelyfilledand
incompletely filled regions. b) A close-up view of a HAR pillar array with
nearly complete filling.
small 2010, 6, No. 6, 768–775
? 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.small-journal.com
771