Stop-Flow Lithography and Its Application to
Graphical Encoding
ARCHIVES
by
OF TECHNOLOLGY
7MASSACHUSETTS INSTIT(JTE
Mo Chen
APR 15 2015
B.S. Optics
Fudan University, 2012
LIBRARIES
Submitted to the Department of Mechanical Engineering in Partial Fulfillment of
the Requirements for the Degree of
Master of Science in Mechanical Engineering
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
February 2015
2015 Massachusetts Institute of Technology. All rights reserved.
Signature redacted
Author..........................................
.....................
Department of Mechanical Engineering
December 3, 2014
Signature redacted
Certified by.................................................
Nicholas-
...F
.
Nicholas X. Fang
Associate Professor
Thesis Supervisor
Signature redacted
Accepted by...............................................
David E. Hardt
Chairman, Department Committee on Graduate Students
1
To my family.
2
Stop-Flow Lithography and Its Application to Adaptive
Encapsulation
by
Mo Chen
Submitted to the Department of Mechanical Engineering on December 3, 2014 in Partial
Fulfillment of the Requirements for the Degree of Master of Science in Mechanical
Engineering
ABSTRACT
Colloids of a few to tens of microns have shown great promise in various
applications. For practical purposes, colloidal building blocks which self-assemble into
operational device are sometimes desired. This preprogrammed assembly requires large
quantities of colloidal building blocks with well-defined shape, size and composition,
which cannot be provided with existing techniques.
In this thesis, a new fabrication technique is presented combining Stop-Flow
Lithography (SFL) and a spatial light modulator (SLM). With this technique,
geometrically anisotropic colloid particles are generated at high throughput (~106
particles/h). Fabrication of functional materials such as hydrogel and shape memory
polymer is proven compatible. All candidate materials can be combined to form
chemically anisotropic colloid particles like Janus particles. Further, the feedback
mechanism of our system allows adaptive fabrication according to detected suspensions.
On the one hand, this extends our material selection pool for the building blocks, as
materials incompatible with direct SFL fabrication are incorporated by encapsulation; on
the other hand, this capability applies to single cell encapsulation and graphical encoding.
This powerful tool facilitates fabrication of complex building blocks and potentially
promotes self-assembly and application of colloids.
Another project covered in this thesis is called solid-state superionic stamping
(S4). It is a direct patterning technique for metals, featuring one-step, large fabrication
area, low cost and working in ambient conditions. This technique is complementary to
SFL in the sense that it enlarges material selection pool.
Thesis Supervisor: Nicholas X. Fang
Title: Associate Professor
ACKNOWLEDGEMENT
First and foremost I would like to thank my advisor prof. Nicholas X. Fang, for
offering me the opportunity to work with him. His enthusiasm, patience and experience
in interdisciplinary research guide me through the difficulties during my Master's study.
He created a space where we have the freedom to explore ideas without financial
constraints.
I sincerely thank prof. Howon Lee, former graduate student and post doc in my
group, who not only helped me get started on the project, but also shared with me his
sincere advice towards research and career. I thank Fan Wang, for sharing with me his
experience of ups-and-downs in research and providing encouragement and suggestions.
I would like to express my gratitude to post docs in the group, Dr. Jun Xu for his
interdisciplinary knowledge, Dr. Dafei Jin for his physical pictures, Dr. Qing Hu for help
with fabrication, Dr. Sang Hoon Nam for working with me through the difficult times of
S4 project, Dr. Kevin Ge and Dr. Qiming Wang for helpful discussions of soft materials,
Dr. Nicholas Viard for providing helpful comments on this thesis and Prof. Shengqiang
Cai, Prof. Nicholas Boechler for valuable advice of graduate study. I would like to thank
Yoon Kyung Lee and Anshuman Kumar for their warm welcome when I just settled. I
thank Tian Gan for sharing happiness and hardships through the past two years. I thank
visiting student Rui Xu for helping me with the S4 project. I would also like to thank
other group members together with who we make the group as a whole. Besides, I would
like to thank NSF for supporting the SFL project, and Samsung for supporting the S4
project and also for providing samples.
Finally I would like to thank my parents and my girlfriend, for their unconditional
love and support.
4
Table of Contents
1. Colloids and Their Applications: State-of-the-Art and Remaining
Challenges...............................................................................................................................
1.1 State-of-the-Art of Colloids .............................................................................................
1.1.1 On the cell/tissue level....................................................................................................................14
1.1.2 On the single device level...............................................................................................................15
1.1.3 Self-assem bly ......................................................................................................................................
2.
13
13
16
1.2 Challenges and Opportunities ........................................................................................
18
1.2.1 Projection M icro-Stereo Lithography (PpSL).....................................................................
1.2.2 Stop-Flow Lithography (SFL).......................................................................................................20
1.2.3 Incorporation of PVSL and SFL................................................................................................
1.3 D issertation Organization...............................................................................................
19
20
21
Stop-Flow Lithography with Dynamic Mask........................................................
2.1 Experim ental Setup and Procedure.............................................................................
2.1.1 Optical Setup .......................................................................................................................................
22
22
22
2.1.1.1
2.1.1.2
2.1.1.3
2.1.1.4
Dynam ic M ask..............................................................................................................................................23
UV LED ............................................................................................................................................................
UV Light Beam Shaping............................................................................................................................25
Optical M icroscope ....................................................................................................................................
24
26
2.1.1.5 CCD cam era...................................................................................................................................................28
2.1.1.6 Optical Alignm ent.......................................................................................................................................28
2.1.2 M icrofluidic Channel........................................................................................................................29
2.1.2.1 Channel Preparation.................................................................................................................................29
2.1.2.2 Flow control in m icrofluidic channel.................................................................................................30
2.1.3 M aterials................................................................................................................................................31
2.1.4 Control through LabVIEW ........................................................................................................
2.1.4.1 Projecting Patterns....................................................................................................................................32
32
2.1.4.2 LED through pow er supply....................................................................................................................33
3.
2.1.4.3 Relay ................................................................................................................................................................
33
2.1.4.4 DSLR cam era ................................................................................................................................................
33
2.1.4.5 M icroscope sam ple stage and focus.............................................................................................
2.1.5 Scalability ..............................................................................................................................................
2.2 Optical Characterization.................................................................................................
34
35
35
2.2.1 Point Spread Function m easurem ent..................................................................................
35
2.2.2 Grayscale Projection for Cross-Link Density Control....................................................
37
Basic Functions of Digital Stop-Flow Lithography...................
3.1 Experim ent Setup....................................................................................................................
39
39
3.1.1 M icrofluidic channel setup.....................................................................................................
3.1.2 Residue Pressure Adjustm ent................................................................................................
39
39
3.2 High through-put fabrication of complex structures..............................................
39
3.2.1 M odeling of the stop-flow process.............................................................................................39
3.2.1.1 Elasticity of PDM S......................................................................................................................................40
3.2.1.2 Lam inar flow ................................................................................................................................................
3.2.1.3 Coupling of elasticity and fluid flow ..............................................................................................
40
41
3.2.2 Geom etrical anisotropy ..............................................................................................................
3.2.2.1 PEGDA .............................................................................................................................................................
3.2.2.2 HDDA ...............................................................................................................................................................
3.2.2.3 Shape-m em ory polym er..........................................................................................................................45
3.2.2.4 Tem perature-responsive hydrogel ...............................................................................................
5
42
43
44
46
3.2.3 Chem ical anisotropy ........................................................................................................................48
3.2.4 Fabrication speed ..............................................................................................................................49
4.
A daptive Graphical Encoding ................................................................................... 50
4.1 Rotation, off-set and scale adjustm ent ............................................................................. so
4.1.1 Rotation .................................................................................................................................................50
4.1.2 Offset and Scale Adjustm ent .........................................................................................................51
4.2 Im age acquisition and analysis ..........................................................................................51
4.2.1 Im age acquisition and processing ..............................................................................................51
4.2.2 Im age analysis ....................................................................................................................................52
4.2.3 Adaptive projection pattern generation ..................................................................................54
4.3 Graphical Encoding .................................................................................................................54
4.3.1 Identification Num bers ...................................................................................................................54
4.3.2 Barcode ..................................................................................................................................................55
4.3.2.1 1D Barcode .................................................................................................................................................... S5
4.3.2.2 2D Barcode .................................................................................................................................................... 56
4.3.3 Particle/Cell caging ..........................................................................................................................57
4.3.4 Particle sorting ...................................................................................................................................58
5.
Solid-State Superionic Stam ping ............................................................................. 60
5.1 Introduction ..............................................................................................................................60
5.2 Copper .........................................................................................................................................61
5.2.1 M aterials ................................................................................................................................................61
5.2.2 Stam p Form ation ...............................................................................................................................62
5.2.3 Results and discussion ....................................................................................................................63
S.3 A lum inum ..................................................................................................................................66
5.3.1 M aterials ................................................................................................................................................66
5.3.2 Stam p form ation ................................................................................................................................67
5.3.3 Results and discussion ....................................................................................................................67
6.
Su m m ary and O utlook ................................................................................................ 69
6.1 Sum m ary ....................................................................................................................................69
6.2 Outlook .......................................................................................................................................69
6.2.1 Characterization of m icro-gels ....................................................................................................69
6.2.2 Increase of throughput ...................................................................................................................70
6.2.3 Finer resolution ..................................................................................................................................70
6.2.4 Flow in three dim ensions ..............................................................................................................70
6.2.5 Additive m anufacturing ..................................................................................................................70
6.2.6 3D printing inside the m icrofluidic channel .......................................................................... 71
6.2.7 Self-assem bly ......................................................................................................................................71
R eferences .............................................................................................................................. 72
6
List of Figures
Figure 1.1 Histology of frog tarsus segments 6 months after surgery. Dashed
white lines indicate position of the displayed histological sections. (a) and (b) show
sections 6 months after tarsus extirpation without scaffold. The missing tarsus gap is
filled with intact tarsus (white arrow) and muscle and scar tissue (black arrow). (b) is a
magnified image of the tarsus gap. (c) and (d) show section 6 months after tarsus
extirpation with scaffold implantation. The tarsus gap is completely bridge with ossifying
tissue (white arrowhead). (d) shows magnified view of the ossifying cartilage sement in
(c). Scale bars: a, e, 1.0mm; b, d, 500gm. Figure reprinted with permission from ref.
[15], 02011 Mary Ann Liebert, Inc., New Rochelle, NY.............................................
14
Figure 1.2 Applications of shape-memory polymers. (a) shows SMP made of
poly(s-caprolactone) dimethacrylate and butylacrylate at 50 wt% transits from the
temporary spiral shape to the permanent rod shape. The whole recovery process takes 35
seconds at 70 0 C. (b) shows simulation result for the recovery cycle of a PEGDMA/tBA
vascular stent. The stent is injected at deformed shape, and recovers to its permanent
shape at 60 0 C inside the artery for added support. (c) shows biodegradable SMP suture
for wound closure. This photo series from an animal experiment demonstrates the
shrinkage of the suture with temperature. Figure reproduced with permission from: a, ref.
[17], 2002 Wiley; b, ref. [18], 2010 Elsevier; c, ref. [22], C 2002 AAAS............ 15
Figure 1.3 (a) 2D colloidal crystal of polystyrene (PS) beads formed by a wedgeshaped cell. (b)-(e) Examples of well-defined aggregates formed by templating spherical
PS beads on patterned substrate. All cylindrical holes are -2gm in diameter. (b) Dimer
clusters formed from 1.0pm PS beads; (c) trimer clusters formed from 0.9ttm PS beads;
(d) square tetramers formed from 0.8ptm PS beads; (e) pentagon aggregates formed from
0.7grm PS beads. Scale bar: a, 5pm; b-e, 2pm. Figure b-e reproduced with permission
16
from ref. [31], C 2001 A CS..........................................................................................
Figure 1.4 (a) confocal image of the x-y plane (top) and x-z plane (bottom) of the
colloidal microgear. The x-y scan is carried out at z=20gm. (b) and (c) SEM image of the
synthesized colloidal microgear made of densely packed silica microbeads. Scale bar:
a,1O0pm; b, 50ptm; c, 5pm. Figure reprinted with permission from ref. [34], C 2008
17
Wiley .................................................................................................................................
Figure 1.5 (a) Triblock Janus spheres which are hydrophobic on two sides (black)
and charged in the middle (white) sediment in DI water. NaCl is then added to screen the
electrostatic repulsion to allow for self-assembly through short-range hydrophobic
attraction. (b) and (c) show the fluorescence image of colloidal kagome lattice and its
FFT image (inset). Bottom of (c) illustrates the orientation of Janus particles. Scale bar:
4pm. Figure reprinted with permission from ref. [40], C 201 1NPG............................. 18
Figure 1.6 Schematic of projection micro-stereolithography. A 3D model is sliced
and sent to the dynamic mask. The mask patterns UV light and cures one layer of
photocurable resin. The substrate then is lowered to allow for fresh resin. 3D structure is
7
fabricated in this layer-by-layer fashion. Figure reproduced with permission from ref.
[43], 0 2014 A AA S......................................................................................................
19
Figure 1.7 (a) Setup of stop-flow lithography. Photocurable resin flows in a
microfluidic device driven by a pressure source controlled by PC through a valve. The
PC also controls UV exposure. (b) Microscope images showing the stop,
lithography/polymerize, flow process. Figure reprinted with permission from ref. [44],
2007 RSC ..........................................................................................................................
20
Figure 2.1 Experimental setup of digital SFL platform. The setup is based on an
inverted optical microscope. A DSLR camera is used as the CCD sensor. The Liquid
Crystal on Silicon (LCoS) chip from a Canon projector is hack-jacked and serves as the
dynam ic m ask...................................................................................................................
22
Figure 2.2 Optical power versus drive current of the UV LED. Below 20A, it
shows good linear relation, indicating good control over projection intensity through
drive current. Figure reproduced with permission from ref. [46], C 2014 Innovations in
Optics................................................................................................................................2
5
Figure 2.3 Optics of SFL platform. Inside the 2" tube, the UV light is
homogenized by a diffuser, collimated by a series of lens and a pinhole. Then it goes
through a polarizer and hits the polarizing beam splitter. The reflected light then gets
patterned and reflected from the LCoS chip, transmits through the beam splitter.
Unwanted polarization component is further filtered by a polarizer before the patterned
UV light enters the m icroscope....................................................................................
26
Figure 2.4 Schematic of the inverted optical microscope (Olympus IX71). UV
projection enters the microscope from the right side port. The mirror for the right side
port is replaced by an 80:20 beam splitter for simultaneous projection and observation.
Figure reproduced with permission from ref. [47], Olympus America Inc.................. 28
Figure 2.5 The observation and projection planes coincide with each other. It
means the optical path from LCoS chip to the sample is the same as from the sample to
CCD sensor, and they will remain the same when adjusting focus. Therefore, observation
of the synthesized particles in real time without changing focus is possible. Scale bar:
2 00 tm ...............................................................................................................................
29
Figure 2.6 Pressure control system of the digital SFL platform. Compressed air is
connected to a 3-way pressure regulator. Flow is on when the regulator is turned on and
off when the regulator is connected to the atmosphere. The 4 pressure relief valves are
connected to the regulator and could output lower pressure individually. Each valve is
connected to a inlet of a microfluidic device, giving independent control of the inlet
pressure and correspondingly stream width inside the channel.................................... 31
Figure 2.7 Automatic control of the SFL platform through LabVIEW. (a) An
alternative way of sending images to the LCoS chip. The LCoS is recognized as a second
monitor by windows. The image is first displayed in a window and moved to the second
monitor. (b) Control of the power supply through its LabVIEW driver. The voltage is
8
fixed while electric current regulated. The range is 0-20A. (c) Control of the RS232 relay
through a typical LabVIEW VISA application, given the address, baud rate and other
specifications. (d) Control of the Canon DSLR camera through a third-party LabVIEW
driver. Shown here is the real-time display of viewfinder.............................................
34
Figure 2.8 Experimentally measured pixel size versus objectives with different
magnification. Blue line corresponds to pixel size of the long side and red line of the
short side. The LCoS pixel is thus rectangular rather than square, with an estimated size
of 11.2pm xI0.5
. ...................................................................................................
35
Figure 2.9 Characterization of the point spread function. (a) and (b) show the
projection and observation of one random noise pattern, respectively. Random noise is
used to excite all frequency components to avoid mathematical infinities during the
calculation of MTF. Five such measurements are averaged to obtain the MTF for each
objective. The OTF is obtained by taking the amplitude of MTF. (c) The inverse Fourier
Transform of OTF gives PSF. Since the LCoS is of rectangular shape, the PSFs along the
long (bottom part) and short (top part, lifted by 0.45) side are slightly different. The short
side has lower side lobes. (d) A representative comparison of the PSFs along the long
side (blue dashed line) and short side (red dash dot line). The minimums occur at the
same position, while the short side has lower side lobes, meaning less cross talk between
adjacent pixels and thus higher resolution ...................................................................
37
Figure 2.10 Grayscale projection for cross-link density control. (a)
Experimentally measured relationship between input grayscale and output light intensity
(normalized by the intensity at 255 grayscale projection) by CCD. The curve shows good
linearity. (b) Discoids synthesized by grayscale 255, 230, 205, 180, 155, 130 inside a
microfluidic channel. The contrast with surrounding material shows difference in
refractive index, an indicator of cross-link density. Inset shows the corresponding
38
lithographic pattern. Scale bar: 50pm ..........................................................................
Figure 3.1 (a) Microfluidic channel setup with a micro pipette loaded with
photocurable resin at the inlet. The rubber tube delivers pressure and seals the pipette
through elasticity. The outlet is connected to a centrifuge tube to balance the residue
pressure. (b) shows the simplified 2D model of the deformed microfluidic channel, where
the height is averaged across width at all cross sections. Inset shows the bulged crosssectional view of the channel when the external pressure is turned on. b Figure reprinted
40
with permission from ref. [44], 0 2007 RSC. .............................................................
Figure 3.2 Structural formulas for (a) PEG and (b) PEGDA. (c) Resolution test
pattern, with pitch size of 10, 5, 3, 2 ,1 pixels, respectively. Inset: photo mask; (d) SEM
image of identification numbers. Scale bar: 50pm........................................................
43
Figure 3.3 (a) Structural formula of HDDA. (b)Resolution test pattern synthesized
with the same photo mask as shown in Fig3.1(c). (c) Micro-gear and (d) Identification
numbers. (e) Grayscale projection with the same photo mask as used in Fig 2.10(b)
synthesized at a more robust range. The grayscale follows 255, 230, 205, 180, 155, 130.
44
Scale bar: 50p . ...............................................................................................................
9
Figure 3.4 (a) Three-micromechanism model of springs and dashpots that
qualitatively explains the thermo-mechanically-coupled large-deformation behavior of
amorphous polymers in a temperature range spanning their glass transition temperature.
(b) and (c) show the structural formulas of BMA and PEGDMA, respectively. In (d),
DMA results for different constitution of BMA and PEGDMA with 5% photo initiator
are given. The data were measured by cooling the test sample from 120'C to 00 C at a rate
of 2*C/min with 1Hz loading frequency. (e) and (f) show high resolution particles of
SMP synthesized inside the microfluidic channel (slightly overcured for visualization
purposes). Figure a reprinted with permission from ref. [60], C 2010 PERGAMON.
Scale bar: 50 m ................................................................................................................
46
Figure 3.5 (a) and (b) show structural formulas of NIPAAm and MBAm,
respectively. (c) and (d) compare PNIPAAm discoids synthesized by our SFL platform
before and after swelling in ethanol. (e) gives the temperature-sensitive swelling behavior
of PNIPAAm discoids in water. All PNIPAAm hydrogel here is polymerized from a mix
of 100 mL ethanol, 70g NIPAAm, 1 g MBAm and 5g photoinitiator........................... 47
Figure 3.6 (a)-(c) conceptually show tuning chemical composition by controlling
inlet pressure individually. When (a) Ph=P2=P3=P4, (b) P2=P4>P1=P3 (c)
P4>PI=P2=P3, the hypothesized particles in the white dashed box have different
chemical composition, though the projection patterns remain the same. (d)-(h)
experimentally validates the concept. Inlet 1 is transparent HDDA and inlet 2 red
PEGDA 700. The same star pattern is projected with varying pressure ratio between
inlets, rendering star patterns of various compositions. White dashed line indicates the
interface between HDDA and PEGDA. Scale bar: 100pm. .........................................
48
Figure 4.1 Process flow of the adaptive projection. (a) Original digital image of
the microfluidic channel with suspension of PS beads. There are a lot of stains on the
substrate. The stains together with the channel walls provide strong contrast with the
surroundings and are major distractions for the identification of beads. (b) Digital image
after background subtraction. The grayscale is adjusted for visualization purposes.
Almost all stains and channel walls are removed. It is superior to (a) in terms of
identification. (c) Remaining distractions are removed by subtracting the image with an
optimized dynamic coefficient. The result is transformed into black-and-white image.
Inset compares the output with (top left green box) binary image without background
subtraction or optimization of dynamic coefficient, and (top right red box) binary image
with background subtraction but no optimization of coefficient. The signal-to-noise ratio
is greatly enhanced by these two steps. (d) A decision is made based on the area,
perimeter etc. of the connected regions to distinguish beads (circled out by red) from
noise. (e) An adaptive projection pattern is generated (inset) and encapsulates the
detected beads. Scale bar: 1 00p ..................................................................................
53
Figure 4.2 Encoding ID numbers to micro beads. (a) Digital image of suspended
beads in a mix of PEGDA700, PEG and ethanol. (b) The same beads encoded with ID
numbers. These IDs are attached to them through further physical or chemical processes
and they are easy to readout for non-specialists. Scale bar: 50pm. .............................. 55
10
Figure 4.3 1 D barcode encoded micro beads. 1 Oiim beads are attached with
simplified UPC barcodes for tracking. The barcodes are of SLRE pattern. Inset: digital
image of the beads right before adaptive encoding. Scale bar: 50 m. ......................... 57
Figure 4.4 Conceptual demonstration of QR code ODA. (a) QR code ODAs
attached to micro beads (in position of drugs). Black regions of QR code is adjusted to
gray so the structure crosslinks as a whole. Otherwise isolated parts will float away. The
error correction level is L, which means 7% of the codeword could be restored when the
QR code is corrupted. (b) and (c) show machine readout of the two QR codes in (a). The
three numbers indicate tracking number, size of the micro bead and production date.
57
Scale bar: 50 m . ...............................................................................................................
Figure 4.5 Caging targets inside a microfluidic channel. Ring patterns are created
to confine targets inside while allowing activities. Dashed white circle is representative of
the shape and size of cage generated in this case. Scale bar: 100 m. ........................... 58
Figure 4.6 Sorting mechanism based on geometric restrictions. A 3 inlet 2 outlet
microfluidic device is used. Three streams flow from right to left and the middle stream is
focus to the connection point of outlets. (a) Upward and downward angles synthesized in
the middle stream of the microfluidic channel; (b) and (c) Due to geometric restrictions,
the angles direct themselves into different outlets. Scale bar 100p........................... 59
Figure 4.7 Categorizing and encapsulating particles with sorting patterns. (a)
Digital image of intact 10pm and 20pim beads. (b) The same beads encapsulated with
sorting patterns. The angles direct themselves towards designated outlets in microfluidic
59
channel. Inset: projection pattern. Scale bar: 1 00m....................................................
Figure 5.1 Schematic of S4 process. (a) The initial state. The stamp and metal to
be patterned are connected through electrical bias, with the stamp as cathode and metal as
anode. (b) When the stamp is in physical contact with metal surface, electrochemical
reaction happens at the contact area and the anode metal is dissolves progressively. (c) S4
gives complementary pattern to the pattern on the stamp. Figure reprinted with
permission from ref. [89], 0 2007 ACS. ......................................................................
61
Figure 5.2 Solution for preparing the superionic polymer stamp. (a) A mix of 3.3g
PAN, 8g DMSO and lOg DMF. (b) 2.37g CuTf2 dissolves in lOg DMF. (c) A mix of (a)
61
and (b), which is finally cast into S4 stamp.................................................................
Figure 5.3 S4 stamp for copper. (a) 6 inch polymer stamp with 400nm L/S
pattern. The color comes from light diffraction from the grating pattern. (b) Digital image
of the L/S stamp taken under optical microscope. Uniform pattern over a large area is
transferred onto the stamp. (c) Polymer stamp with pyramid array pattern transferred onto
a glass slide. The stamp shows no crease when transferred properly. (d) Pyramid pattern
array observed under optical microscope. All pyramids from the hard master mold are
successfully transferred to the stamp (1 inchx 1 inch). The square bottom of pyramid can
62
be seen from the figure. ................................................................................................
11
Figure 5.4 (a) S4 setup. A manual stage is installed on a motorized translational
stage. Screwed to the manual stage is a load cell connected with a tilting stage and the
stamp holder. (b) Close-up view of the experimental setup. The bias is applied through a
copper conductive tape. ................................................................................................
63
Figure 5.5 SEM images of (a) Ipm L/S pattern and (b) pyramid pattern. (a) The
L/S pattern is successfully transferred over most area. There are small fraction of area
where missing/connected lines indicated over/under etching. The line edge roughness is
~3Onm. The ratio between peak and valley change from 1:1 of the hard master mold to
-9:7 of etched Cu. (b) Top view of the pyramid array pattern under SEM. The square
bottom of pyramid is clearly seen. Top inset: Pyramid arrays etched on Cu sample. The
spacing between adjacent pyramid arrays indicate successful transfer of pattern over large
area. The patterned region is -0.6 inch 2 . Bottom inset: close-up view of the 3D pyramid
structure. The square bottom is precisely transferred with a side length of -2tm.
However, the aspect ratio is significantly smaller than hard master mold due to
deformation of the stamp. Scale bars: a, lOjpm; b, 20prm, bottom inset: 2m.............. 64
Figure 5.6 During the imprinting process, the thin film stamp is a soft interface
sandwiched by two hard substrate, causing wrinkling. (a) Optical microscope and (b)
SEM images of a 1 pm L/S sample showing wrinkling pattern. Scale bar: 20pm...... 65
Figure 5.7 Selective deposition by reverse use of the S4 stamp. (a) Schematic of
selective deposition. A layer of metal to be deposited is attached to the ionic stamp as
anode. The substrate is in contact with the patterned surface of stamp and connected to
the cathode. When an electrical bias applies, the anode metal is dissolved and the ions
transport through the stamp and deposit onto the contact regions of substrate. When the
deposition is finished, the stamp is removed from the substrate and the selectively
deposited metals are left on the substrate, forming the same pattern as that on the stamp
(compared to the reverse pattern in the S4 case). (b) and (c) show preliminary result by
selectively depositing Cu (b) 1 gm L/S and (c) pyramid array pattern onto a Cu substrate.
In (b), long range ordering is observed and in (c), the array is clearly seen................. 66
Figure 5.8 (a) Schematic of the selective anodization. A Nafion film exchanges
proton and promotes growth of AAO at contact areas. (b) SEM image of AAOs formed
by this electrochemical process. Scale bar: 500nm. ....................................................
67
Figure 5.9 SEM images of (a) 1 00nm and (b) 65nm L/S Nafion stamp formed by
hot embossing under -650 psi at 150'C. Scale bars: a, 2pm; b, 2pm. ......................... 67
Figure 5.10 SEM images of selectively anodized aluminum with (a) 150nm and
(b) 1 00nm L/S pattern. Inset shows close-up views of the nanostructures. As can be told
from the inset of (a), over-anodization occurs The inset of (b) shows appropriate
anodization condition, where AAO nanopores form the desired structure. Scale bars: a,
5[tm, inset: 500nm ; b, 2ptm, inset: 500nm . ...................................................................
68
12
1. Colloids and Their Applications: State-of-the-Art and Remaining
Challenges
This thesis introduces a Stop-Flow Lithography (SFL) fabrication platform with a
dynamic mask. The fabricated particles are from a few to tens of micrometers, belonging
to colloids, which have shown great promise in tissue engineering and biomedical
applications. However, implantation of the whole device is inconvenient and sometimes
inaccessible. Small building blocks of the device, though, could be delivered through a
minimally invasive surgery and self-assemble into operational device inside the body.
This preprogrammed assembly requires large quantities of colloidal building blocks with
well-defined shape, size and composition. Current fabrication methods, however, either
have low throughput, limits in particle shapes, or high cost.
In this thesis, a new fabrication technique is presented combining Stop-Flow
Lithography (SFL) and a spatial light modulator (SLM). With this technique,
geometrically anisotropic colloid particles are generated at high throughput (~ 106
particles/h). Fabrication of functional materials such as hydrogel and shape memory
polymer is proven compatible. All candidate materials can be combined to form
chemically anisotropic colloid particles like Janus particles. Further, the feedback
mechanism of our system allows adaptive fabrication according to detected suspensions.
On the one hand, this extends our material selection pool for the building blocks, as
materials incompatible with direct SFL fabrication are incorporated by encapsulation; on
the other hand, this capability applies to single cell encapsulation or graphical encoding.
This powerful tool facilitates fabrication of complex building blocks and potentially
promotes self-assembly and application of colloids.
Chapter 5 presents another fabrication technique called solid-state superionic
stamping (S4). It is a direct patterning technique for metals, featuring one-step, large
fabrication area, low cost and working in ambient conditions. This technique is
complementary to SFL in the sense that it enlarges material selection pool.
1.1 State-of-the-Art of Colloids
Colloids are usually referred to as micro or nanometer sized particles dispersed
throughout another phase [1]. This range of size is of particular interest because
Brownian motion, which is the random motion caused by collisions from surrounding
molecules of the dispersed phase [2], is important at this scale. Colloidal dynamics is
generally observable under optical microscopes, making colloids a perfect candidate for
fundamental research. Besides, colloids can be considered as giant molecules in a loose
manner, and find increasing applications in foods, drinks, inks, paints, coatings,
cosmetics, photographic films, rheological fluids and magnetic recording media etc. [3].
Although Graham [4] and Ostwald [5] pioneered the field more than 150 years ago,
preparation of colloid particles with well-defined size, shape and composition at highthroughput remains a challenge. A relatively new but powerful approach to generating
colloids is made possible with the emergence of microfluidics. By co-flowing immiscible
phases and introducing instability, emulsion drops are generated inside microfluidic
channels. These drops are then photo-polymerized to form stable hydrogel or polymeric
particles. This approach is widely used, as it manufactures colloid particles in large
13
quantities not only in spherical shape, but also in rods and discoids with appropriate
microfluidic channel design [6]. These colloids can be modified with cell-adhesive
ligands for manipulating cultured cells [7]. Moreover, taking advantage of laminar flow,
researchers are able to generate Janus particles, named after the two-face Roman god
Janus, which has two or more distinct surface properties [8]. A brief literature review on
these topics is covered below.
NoG
scaffold
Scaffold + GF
Figure 1.1 Histology of frog tarsus segments 6 months after surgery. Dashed white lines indicate position
of the displayed histological sections. (a) and (b) show sections 6 months after tarsus extirpation without
scaffold. The missing tarsus gap is filled with intact tarsus (white arrow) and muscle and scar tissue (black
arrow). (b) is a magnified image of the tarsus gap. (c) and (d) show section 6 months after tarsus
extirpation with scaffold implantation. The tarsus gap is completely bridge with ossifying tissue (white
arrowhead). (d) shows magnified view of the ossifying cartilage sement in (c). Scale bars: a, c, 1.0mm; b,
d, 500pm. Figure reprinted with permission from ref. [15], c2011 Mary Ann Liebert, Inc., New Rochelle,
NY.
1.1.1 On the cell/tissue level
Colloids have a characteristic size in the micrometer region, comparable to cells.
They can be modified with biological functionalities, like ligands to interact and even
control cell behavior. One representative demonstration of such kind is the addition of
amino acids like arginine-glycine-aspartic acid (RGD) peptide sequence to a copolymer
system, which enhances both adhesion and spreading of mammalian cells [9]. This kind
of materials potentially acts as a scaffold for tissue regeneration or repair. Dissociated
cells are encapsulated by the polymer scaffold and delivered to the appropriate site. The
polymer scaffold defines the space for tissue growth and confines the total size and shape
for the engineered tissue. This approach has shown success in engineering skin, bone and
cartilage tissues and recently in growing bone tissues [10]. It enables in vivo tissue
engineering from a small number of implanted cells, in contrast to in vitro tissue growth
and transplantation. An improvement is grafting poly(ethylene glycol) (PEG) to an
artificial protein with functionalities such as cell adhesion, heparin binding and
degradation since PEG reduces undesired protein adsorption and colloid aggregation.
Halstenberg et al. show that protein-graft-poly(ethylene glycol) diacrylate (PEGDA)
featuring cell survival, spreading, migration and proliferation serves as an excellent
candidate for tissue repair [11]. PEGDA is also utilized to generate cell encapsulations
[12], cell carriers [13] and protein detectors [14]. Other materials like 1,6-hexanediol
diacrylate (HDDA) are also investigated for biomedical purposes (Fig 1.1) [15].
14
1.1.2 On the single device level
Polymers capable of switching configurations in response to environmental stimuli
such as temperature, hydration, pH, specific molecules, magnetic or electric field are
referred to as "smart" materials [16]. One example is hydrogel that swells or shrinks
when hydrated or dehydrated. Another example involves shape-memory polymer (SMP)
that recovers from a temporary shape to its permanent shape upon heating. A
demonstration of shape memory effect is shown in Fig 1.2(a)[17]. It promises application
in implanting large devices through a small incision, for example, laparoscopes in a
minimally invasive surgery. A potential candidate is a stent that expands and supports a
blocked artery. Conceptual demonstrations are done both theoretically (see Fig
1.2(b))[18] and experimentally [19,20,21]. Polymer-coated stents have also been
approved for controlled drug delivery in both Europe and the US [16]. Another promising
application of SMP is to accomplish programmed complex mechanical deformation in
vivo, such as suturing. A biocompatible multi-block copolymer is developed by Lendlein
et al. and demonstrates the ability to close an incision with programmed force. Their
SMP has mechanical stresses resembling those of soft tissues (~1 MPa). After tying a
loose knot in a conventional way, the SMP suture tightens and presses the wound lips
together under proper pressure to avoid formation of necrosis or hernias (Fig 1.2(c)).
Moreover, the suture is biodegradable through hydrolyzation [22]. We expect more
biomedical devices that accomplish subtle and complex tasks to be developed.
(a)
(b)
0#
Undeftnd dpp at60 C
Deformed
de g a60 C
wdetoaryae. This phto sriaeat5trss fromanaiaxeiete demosrates sthea shrae o the strmaenth
temperature. Figure reproduced with permission from: a, ref. [17],
Elsevier; c, ref. [22],
2002 A AAS.
15
2002 Wiley; b, ref. [18],
2010
1.1.3 Self-assembly
In tissue engineering, scaffolds are introduced for in vivo tissue regeneration/repair
to avoid immune responses. Besides, it minimizes the incision size needed compared to
transplantation. Smart materials like SMP further reduce the size of devices like a
scaffold to be implanted. However, more significant reduction is desired for minimally
invasive surgery. In other cases, it is difficult or even impossible to implant a device
directly. It is desired that small colloidal building blocks self-assemble into large
functional devices in vivo, in a similar way as DNA strands bond together to finally form
lives. In fact, traditional materials are selected by their properties; the next generation of
materials is supposed to be designed from building blocks according to applications and
programmed for self-assembly [7]. The most straightforward assembly is closely packed
spheres. A flow cell [23,24,25] or wedge-shaped cell [26,27,28] is used to crystalize 3D
or 2D granular crystals consisting of colloidal spheres as shown in Fig 1.3(a). The
colloidal beads are initially monodispersed in an aqueous solution. As the solvent
evaporates, these beads pack and crystalize. Embedding polystyrene spheres in PVA
matrix and stretching at elevated temperature results in ellipsoidal beads due to
viscoelasticity. Closely packed ellipsoids are achieved in this way [29]. Peanut and rodshaped colloids of iron oxides are synthesized, offering options as building blocks for
self-assembly [29,30].
During the crystallization, physical confinement such as cylindrical hole array
patterned on the substrate renders polygonal or polyhedral clusters consisting of colloidal
beads as shown in Fig 1.3(b)-(e). The structure of resulting cluster is mainly determined
by the ratio between the dimensions of the holes and the radius of the colloidal sphere
[3 1]. This provides even complex building blocks for medical and biological applications
as well as photonics.
(b)
(c)
Figure 1.3 (a) 2D colloidal crystal of polystyrene (PS) beads formed by a wedge-shaped cell. (b)-(e)
Examples of well-defined aggregates formed by templating spherical PS beads on patterned substrate. All
cylindrical holes are -2pm in diameter. (b) Dimer clusters formed from I .0pn PS beads; (c) trimer clusters
formed from 0.9pmi PS beads; (d) square tetramers formed from 0.8pm PS beads; (e) pentagon aggregates
formed from 0.7pn PS beads. Scale bar: a, 5prm; b-e, 2pm. Figure b-e reproduced with permission from
ref. [31], C 2001 ACS.
16
Another powerful approach for assembly granular crystals is through flowlithography. In microfluidic channels, aqueous latex particle suspension turns into waterin-oil emulsion caused by droplet break-off. As the droplet moves along the channel,
water keeps diffusing into the oil phase and the droplet shrinks to a spherical colloidal
aggregate. This aggregate serves as an efficient scatterer, and could be used for diffusers.
Thus, it is named as "photonic balls" [32]. Further improvement is achieved by
combining lithographic approach with flow-lithography. After careful study of the
jamming phase diagram, a very dense suspension of silica microspheres in a photocurable
solution can flow inside a microfluidic channel without jamming [33]. Granular crystals
are shaped by shining patterned UV light with a physical photomask. To increase the
structural integrity, a sintering process follows to create dense glassy silica structures
(See Fig 1.4) and a partial sintering renders porous structures [34].
Figure 1.4 (a) confocal image of the x-y plane (top) and x-z plane (bottom) of the colloidal microgear. The
x-y scan is carried out at z=20ptm. (b) and (c) SEM image of the synthesized colloidal microgear made of
densely packed silica microbeads. Scale bar: a,I00ptrm; b, 50ptm; c, 5pgm. Figure reprinted with permission
from ref. [34], C 2008 Wiley.
More recently, trajectory engineering by manipulating particle shape is
demonstrated both theoretically [35] and experimentally [36]. A method for synthesizing
anisotropic colloids suitable for trajectory engineering also shows up [37]. Though this
does not achieve assembly yet, it could potentially serve as a means to orient nonspherical particles. As Younan Xia once prQposed for self-assemble complex shaped
building blocks, positional ordering and orientational ordering would do the job [29].
It's not only shape, but also composition that play a role in organizing particles into
assembly [38]. Flow-Lithography allows researchers to fabricate amphiphilic Janus
building blocks in large quantities. In emulsions, these Janus particles move to the wateroil interface and assemble to minimize their total surface energy [39]. In another case,
fluorescent latex particles are coated with gold and hydrophobically treated on their
poles, resulting in "triblock Janus". The particles are thus electrostatically repulsive in the
middle and hydrophobically attractive on the poles to each other. After further selfassembly conditions are applied, these particles form a quasi-two-dimensional colloidal
kagome lattice (See Fig 1.5) [40]. Assembly can also be initiated by magnetophoresis
when there exists a magnetic susceptibility contrast between suspended building blocks
17
and fluid. By embedding nickel grid pattern just below PDMS channel, strong spatially
varying magnetic field immerses the channel. The difference between suspensions and
fluid of their magnetic susceptibilities molds a template for assembly, just as optical traps
do. This approach works not only for single but multiple component assembly [41].
(a)
RsptLsIn
~~ 00~H20
Sedkment
Seff-mmemb
~O
AtrctinOUnU
(b)W
Figure 1.5 (a) Triblock Janus spheres which are hydrophobic on two sides (black) and charged in the
middle (white) sediment in DI water. NaCl is then added to screen the electrostatic repulsion to allow for
self-assembly through short-range hydrophobic attraction. (b) and (c) show the fluorescence image of
colloidal kagome lattice and its FFT image (inset). Bottom of (c) illustrates the orientation of Janus
particles. Scale bar: 4ptm. Figure reprinted with permission from ref. [40], C 201 INPG.
1.2 Challenges and Opportunities
Although all these methods differ significantly, the basis is generating building
blocks with well-defined shape, size and composition in large quantities. As we can see
from previous examples, the ability to self-assemble largely depends on how many
"patches" one building block has. Current research focuses mainly on Janus particles
with two faces. In addition, colloid particles are primarily synthesized based on droplet
break-off in microfluidics. Colloids generated by this approach are shaped by both
surface tension and channel geometry, thus having simple and limited shapes. Besides,
the device lacks tunability and adaptability, making it cost-ineffective in certain
applications. Last but not least, chemical anisotropy benefited from co-flow is very
limited, usually two or three components in parallel. Therefore, a new approach is
desired, which generates geometrically and chemically anisotropic particles with highthroughput. In addition, this approach should be able to generate particles with complex
composition, and preferably have adaptability.
A building block fabrication system promising to satisfy all these requirements is
developed, based on projection micro-stereolithography (PpISL) and stop-flow
lithography (SFL).
18
1.2.1 Projection Micro-StereoLithography (PpSL)
Stereolithography refers to a technique patented in 1986 by Charles W. Hull. It is
described as an additive manufacturing method where patterned ultraviolet (UV) light
polymerizes or crosslinks a thin layer of photocurable material into solid and construct a
3D structure in a layer-by-layer fashion. Inspired by it, our group has previously
developed the technique into Projection Micro-Stereolithography (PpSL) where a spatial
light modulator (SLM) is used instead of a series of physical photomasks [42,43]. A SLM
is a device that modulates incoming light spatially. With the SLM as a dynamic mask that
reconfigures pattern digitally, there are no more needs for replacing physical photomasks
and doing alignments for each layer.
Fig 1.6 schematically shows PptSL and its process flow. Surface geometry of
complex 3D structure is stored in an stl (stereolithography) file and is then digitally sliced
into a series of cross-sectional images by an open-source software (CreationWorkshop).
Each cross section image functions as a digital photomask to fabricate the corresponding
layer. After an image is sent to SLM, the UV LED is turned on for a proper period of
time. The UV light is reflected off the SLM to be patterned and then projected onto the
surface of the photocurable resin, crosslinking the illuminated area and generating a thin
patterned layer corresponding to the image sent to SLM. After the layer is formed, the
substrate which holds the sample lowers itself from the surface of photocurable resin by
the thickness of the next layer. Fresh photocurable resin comes in and covers the area
above the polymerized structure and the next image is sent to the SLM so as to
polymerize the next layer on top of the previous one. A 3D structure is fabricated in this
layer-by-layer fashion until all layers are complete.
2014 AAAS.
permission from ref. [43],
.
A state-of-the-art SLM is capable of displaying high-definition images, which
means there are more than 1,000,000 pixels on the SLM. However, the pixel size is
usually ~10pm by 10pm, much larger than the resolution required in micro fabrication.
Fortunately, during the projection process, a reduction of lateral resolution can be
achieved. With a 10:1 projection, the theoretical resolution is reduced from 10pm to 1pm.
Taking into account the practical limit imposed by point spread function (PSF) of the
optical system, the image experiences certain degree of blurriness and the real resolution
19
is usually worse than the naive calculation given above. Detailed characterization of the
optical system for SFL is given in the next Chapter.
1.2.2 Stop-Flow Lithography (SFL)
Stop-flow lithography is first developed in Patrick Doyle's group for highthroughput generation of colloid particles [44]. A stream of photocurable material flows
inside a microfluidic channel which is held on the sample stage of a microscope. A
transparency photomask is inserted between the UV illumination light and objective, and
is focused into the microfluidic channel. The flow is first stopped before it is exposed to
patterned UV light. The illuminated part of photocurable material is crosslinked
instantaneously and forms the desired geometry. The synthesized polymers are then
flushed away before the next 'stop-polymerize-flow' cycle is performed (see Fig 1.7).
Considering the high flow rate (driven by compressed air) and short exposure time
(usually ~ 100ms), the overall throughput is very high, generally ~ 104-106 particles/h.
With laminar nature of microfluidics, complex composition/chemical anisotropy is
introduced through co-flow. Last but not least, it is convenient to switch materials inside
the microfluidic channel.
(a)
3-way
microfluidic
channel
valve
miroscope
---
open
(h)
s
top
(
cl
sed
Polymerize
)
WA)Kt:
sed
Flow
[44],6C 2007 RSC.
1.2.3 Incorporation of PpSL and SFL
Both PpSL and SFL are based on patterning a photocurable resin by UV light.
Incorporation of the two technologies possibly combines high-throughput from SFL,
20
high-resolution from lithography and flexibility from dynamic mask of PpSL. For
brevity, we refer to this new system as digital SFL without ambiguity. The combination
of SFL and P[tSL provides unprecedented fabrication freedom and capability. On the one
hand, we can fabricate chemically and geometrically anisotropic polymers with high
throughput and resolution; On the other hand, we can dynamically change the shape of
polymers. One example is particle sorting introduced in Chapter 4. With a CCD camera
monitoring the microfluidic channel, we can see the microfluidic environment in real
time and generate specific patterns according to the in situ situation. With a home-written
program in LabVIEW, we are able to analyze particles suspended in the fluid and
encapsulate different particles accordingly. A simple geometric consideration makes the
encapsulated particles to enter different outlet. This method is also applicable to cell
sorting and generating particles with more complex composition.
In addition, all photocurable materials used in this study are easily functionalized,
making them good candidates for biomedical studies. Furthermore, a selection of these
materials belong to "smart" materials, which means they have properties that responds to
external stimuli in a controlled fashion, such as stress, temperature, hydration, pH. To
name a few, shape memory polymer made of poly(ethylene glycol) dimethacrylate
(PEGDMA) and benzyl methacrylatc (BMA), which responds to temperature and stress;
poly(N-isopropylacrylamide) (PNIPAAm) which responds to temperature and hydration
and poly(ethylene glycol) diacrylate (PEGDA) which responds to hydration only. The
combination of PpSL and SFL provides the opportunity of designing and fabricating
building blocks made from a combination of these smart materials. This allows extra
freedom in self-assembly, reconfigurable and tunable devices.
1.3 Dissertation Organization
In Chapter 2, a detailed description of the SFL system is given, together with
optical characterization; Chapter 3 describes basic functions, featuring fabrication of
geometrically anisotropic functional materials and chemically anisotropic particles at
high throughput, suitable for generating building blocks for self-assembly; Chapter 4
demonstrates the adaptability, with graphical encoding as an example. This function
enables fabrication of building blocks with cores that cannot be synthesized directly with
SFL, greatly extending our material selection pool; Chapter 5 presents another research
project solid-state superionic stamping (S4), which is independent of SFL. This
fabrication approach is based on ionic gels, and is specifically designed for direct
patterning on metals, supplementary to SFL which is suitable for colloid polymers.
Chapter 6 summarizes the work introduced in this thesis and outlooks possible directions
for applying our technique.
21
2. Stop-Flow Lithography with Dynamic Mask
2.1 Experimental Setup and Procedure
The experimental setup and operation procedure mainly follows that of Stop-Flow
Lithography (SFL) as shown in Fig 1.7. The unique feature in our system is the
introduction of a dynamic mask instead of physical mask, which allows instantaneous
change of projection patterns. Not only does it enable fast switching between different
projection patterns, but also it makes possible the adaptive graphical encoding, which will
be introduced in Chapter 4. This Chapter covers all the major components making up our
SFL system.
2.1.1 Optical Setup
Figure 2.1 Experimental setup of digital SFL platform. The setup is based on an inverted optical
microscope. A DSLR camera is used as the CCD sensor. The Liquid Crystal on Silicon (LCoS) chip from a
Canon projector is hack-jacked and serves as the dynamic mask.
As introduced in Chapter 1, the system is a combination of Stop-Flow
Lithography (SFL) and Projection Micro-StereoLithography (PjtSL), featuring high
throughput and adaptability. As shown in Fig 2.1, the whole setup is based on an inverted
optical microscope. On the right hand side, 405nm light from a high power UV LED
passes through a collimator, a polarizing beam-splitter and finally spatial light modulator
(SLM). The patterned light then enters the microscope through an input port and is
22
focused into the microfluidic channel by the objective. During the fabrication cycle,
microfluidic flow is first stopped by shutting off the compressed air pressure. Then, a UV
pattern is projected into the channel and polymerization occurs. Finally, the pressure
resumes and synthesized particles are flushed away and collected. This cycle is repeated
to generate large quantities of colloid particles at high throughput. Detailed description of
key components is given below.
2.1.1.1 Dynamic Mask
SLM is a device that configures light spatially. There are two major mechanisms
for SLM, one based on micro-mirrors and another on liquid crystals. The first kind is
referred to as digital micro-mirror device (DMD, Texas Instrument), the core of digital
light processing (DLP) projection technology. It has an array of micro-mirrors on the
chip, each representing one pixel. Each mirror tilts 120 individually when bias voltage is
applied, passing or deflecting away light [45]. Liquid crystal display works on both
reflection and transmission modes. Of interest is the reflection one, usually referred to as
liquid crystal on silicon (LCoS) with a highly reflective silicon backplane. Liquid crystal
is a special form of matter between conventional liquid and solid crystal. It flows like a
liquid, but has domains within which molecules are well oriented, as inside a solid
crystal. The Nobel Prize in Physics 1991 was awarded to Pierre-Gilles de Gennes at the
Universite Paris-Sud "for discovering that methods developed for studying order
phenomena in simple systems can be generalized to more complex forms of matter, in
particular to liquid crystals and polymers". Liquid crystal molecules generally have
permanent electric dipole moments. Under the influence of external electric field, these
electric dipole moments tend to align parallel to the electric field, reorienting different
domains. The degree of "alignment" is controllable through the strength of the field.
When electrical bias is applied, the whole liquid crystal acts effectively as a uniaxial
crystal. Light with proper polarization decomposes its components into ordinary light and
extraordinary light, experiences an optical phase difference during the passage of liquid
crystal and then recombines again into one beam. Depending on the phase difference,
linearly polarized light becomes elliptically polarized light, circularly polarized light, or
linearly polarized light in a different direction. Amplitude modulation is achieved when a
linear polarizer is utilized to filter the outgoing light. Inside a projector, there are usually
three LCoSs for red, green and blue colors respectively. By modulating each primary
color and combining them, we obtain a true color projection image. Also noted, that the
contrast ratio is determined by the polarizers. A typical contrast ratio for LCoS based
projectors is 1000:1. However, 405nm UV light is concerned in this study and most
materials start to show strong absorption in the UV spectrum. As a result, commercial
laminated thin film polarizers usually give less than 100:1 extinction ratio.
On the other side, each pixel on a DMD chip is a micromirror that can be tilted
individually by 120. When the optics is aligned appropriately, the tilting of these
micromirrors decides the ON and OFF state of single pixels where incident light is
reflected to the detector or deflected away. Since 120 is small and inconvenient for
compact optical alignment, commercial projectors use only one DMD chip. A
mechanically spinning color wheel is harnessed to fast switch outgoing light between red,
23
green and blue. When the switching is fast enough, the so-called "rainbow effect" makes
human perceive true color. Unlike LCoS, which modulates light intensity (or grayscale)
analogously, DMD controls only On and OFF states. The grayscale is achieve by rapidly
hopping between these two states and control the ratio of ON and OFF period (duty
cycle). Similarly to the true color, people perceive grayscale rather than flashes.
For photo-polymerization, both respond fast enough. DMD provides much better
contrast in the UV while LCoS is capable of controlling grayscale for individual pixels.
In experiments, a contrast of 2 is good enough for all practical purposes because of the
existence of a threshold for polymerization. In this study, we choose LCoS for its
grayscale control. The LCoS is taken from a commercial projector, Canon Realis SX50
with native resolution 1400x 1050. The size of LCoS is 0.7 inch, rendering a pixel size of
~ 10pm by 1 Om. In fact, the pixel is slightly rectangular- rather than square-shaped, as
will be covered in the "scalability" subsection. PC recognizes and automatically
configures the projector as a second monitor. Sending an image to the second monitor
configures three LCoS accordingly. The LCoS in charge of blue color is chosen for the
setup, as the dispersion between UV and blue light is minimal.
2.1.1.2 UV LED
One fabrication cycle period consists of stop time (details in section 3.2.1),
polymerization time and flow time. To achieve high throughput, each time should be
minimized. Stop time depends on the channel geometry and material viscosity. Flow time
depends on the pressure, viscosity and travel distance, and can be minimized by
increasing the air pressure. Polymerization time is inversely proportional to light intensity
according to
D
t = Dead
(2.1)
iap,
Where t is the polymerization time, D proportional to the reactivity between photocurable
material and photoinitiator, I the light intensity, a and ap1 the absorption coefficient of
resin and photoinitiator, and d the cured depth.
A high power 405nm UV LED (LumiBrightTM UV 2600N-700, Innovations in
Optics) is selected since its maximum optical power is 11.5W.
The forward voltage for this LED is between 3.OV-4.8V. Maximum drive current
is 30A. Agilent E3633A power supply is used to drive the LED. Fig 2.2 shows optical
power vs. drive current provided by the vendor. Over the accessible range in our
experiment (0-20A), the optical power is in good linear relation with drive current. In
turn, the projection intensity can be controlled well.
24
Optical Power vs. Drive Current
12
4
0
0
5
10
1
20
25
30
Drive Current(A)
Figure 2.2 Optical power versus drive current of the UV LED. Below 20A, it shows good linear relation,
indicating good control over projection intensity through drive current. Figure reproduced with permission
from ref. [46], C 2014 Innovations in Optics.
2.1.1.3 UV Light Beam Shaping
The UV light coming out of the LED is significantly inhomogeneous and
divergent. Accordingly, the optics of our SFL systems consists of a homogenizer,
collimator, LCoS and projection part (see Fig 2.3). In this section, emphasis is on the first
three elements.
A " engineered diffuser (Dl -C20-MD, Thorlabs) works for 400nm-700nm and
has a divergence angle of 200. Due to the large divergence of LED, light coming out of
the diffuser has a divergence angle much large than 200. To collect the most of light, an
achromatic doublet of 75mm focus (AC508-075-A-ML, Thorlabs) is followed
immediately after the diffuser. Following the doublet, we have a bi-convex lens of 60mm
focus (LB1723, Thorlabs), a ring-activated iris diaphragm (SM2D25D, Thorlabs) and a
bi-convex lens of 100mm focus (LB1676, Thorlabs) spaced 0.5", 0.5" and 5" away,
respectively. The iris diaphragm is at the Fourier plane, thus being able to adjust the
components of light in the k space. At the expense of light intensity, we obtain highly
collimated light with small aperture. By measuring the light spot projected on a screen 2
meters away, a simple calculation shows a divergence angle less than 6.8 minutes.
25
Figure 2.3 Optics of SFL platform. Inside the 2" tube, the UV light is homogenized by a diffuser,
collimated by a series of lens and a pinhole. Then it goes through a polarizer and hits the polarizing beam
splitter. The reflected light then gets patterned and reflected from the LCoS chip, transmits through the
beam splitter. Unwanted polarization component is further filtered by a polarizer before the patterned UV
light enters the microscope.
After the light is collimated, it goes through a thin film polarizer (Tiffen 52mm
linear polarizer) and then a 1" polarizing beam-splitter (PBS25-405-HP) that is optimized
for 405nm light. The light is totally reflected to the LCoS, modulated and reflected back
to the beam splitter (see arrows in Fig 2.3). Only the light in white region changes its
polarization, thus able to pass through the polarizing beam splitter and to enter the optical
microscope. For better contrast, a thin film linear polarizer (PRINZ 37mm linear
polarizer) is inserted behind the beam-splitter. Due to the size mismatch between
collimated light and polarizing beam splitter, a fraction of light is scattered. This stray
light in general has a random polarization and cannot be filtered by the polarizing beamsplitter. As a result, the first polarizer has very limited contribution to contrast and a weak
polarizer is selected for higher transmission. The second polarizer eliminates most stray
light and allows only those from LCoS, deciding the system's contrast. As we will see
later in section 2.2, the experimentally acquired contrast is 5:1, limited by the linear
polarizer and stray light (from both the LED and environmental light from above the
sample).
2.1.1.4 Optical Microscope
The patterned light then goes into an inverted optical microscope (Olympus IX7 1)
through a port designed for fluorescence output. An output port is selected because it is
26
preferred to observe the microfluidic channel simultaneously when project patterns into
it. The focus of projection and imaging/observation should coincide for both alignment
and observation purposes. For an inverted optical microscope, this means the optical path
from the sample to the CCD and to LCoS must be the same. As fluorescence output and
observation shares the optical path from sample through objective to a movable mirror
for the right side port (Fig 2.4), their focal planes stay coincide when adjusting focus. To
be able to observe and project simultaneously, the reflection mirror for right side port is
replaced by an 80:20 beam-splitter. Another advantage of sharing the objective is, no
matter how small the projection image is on the sample, a projection pattern of original
size is observed. However, sharing the optical path has its disadvantage. Any UV portion
causes polymerization in the microfluidic channel, including the one from illumination
light. To minimize unwanted polymerization and prevent the channel from being
blocked, an UV filter is inserted between illumination light source and the sample. Since
the optical path for projection passes the color filter for reflection mode illumination,
reflection mode is not suitable for digital SFL. To allow for the projection UV light, only
transmission mode is used. In this mode, illumination light from a condenser focuses onto
a small region of the sample which covers the objective's numerical aperture. However,
inverted optical microscopes are usually designed for biological purposes, and
condensers have short working distance. In our case, it is 27mm, well below the height of
the whole microfluidic device. As a result, the condenser must be raised to accommodate
the device, making the illumination light blurring out. On one hand, the blurred light
causes unwanted polymerization in a large area; On the other hand, to achieve a clear
view in high magnification, excess illumination light is needed, which promotes
unwanted polymerization process of all photocurable materials inside the channel,
limiting experimental time.
27
Figure 2.4 Schematic of the inverted optical microscope (Olympus IX71). UV projection enters the
microscope from the right side port. The mirror for the right side port is replaced by an 80:20 beam splitter
for simultaneous projection and observation. Figure reproduced with permission from ref. [47], Olympus
America Inc.
2.1.1.5 CCD camera
In our system, projection and imaging processes share the same optical path of the
inverted optical microscope and the projection area appears the size of LCoS in the
camera. As a result, standard CMOS cameras like DFK 72AUC02 (Imaging Source) are
not large enough to cover the whole projection area. One way is to insert a reduction lens
to fit the projection area onto the CMOS camera's active area. Another way is to use a
large CCD camera. In our case, a DSLR camera (Canon EOS 60D) is used which has an
APS-C CCD sensor (22.20mmx 14.80mm, 5184x3456 resolution). In addition to the high
end CCD, the viewfinder provides 96% coverage of the imaging area with 1056x704
resolution.
2.1.1.6 Optical Alignment
It is crucial to make sure that the projection plane is parallel to the sample stage
and the light passes through the center of objective. However, as discussed above, the
projection image passes through a beam-splitter and objective in the optical microscope
before it hits the sample. We have no control over the objective and beam splitter,
28
making it hard to adjust optics. An alternative way is having a laser pointer (Laseraim,
LTM2HDK) hanging above the sample stage. We adjust the laser pointer so laser light
enters the microscope through the center of objective and the reflected light spot returns
to the laser pointer. Now, the laser beam is vertical to the sample stage. Then, we adjust
the polarizing beam splitter and LCoS so the laser beam hits them vertically. Finally, the
collimator is adjusted so the laser beam goes out of its center. Thanks to optical
reversibility, images then project vertically to the sample through the center of objective.
To check the performance, two images are taken when projection pattern is focused on
the top and bottom surfaces of a glass slide with a 20X objective, respectively (not shown
here). The displacement of the two images taken is within 2 pixels, showing a deviation
from vertical projection smaller than 3.4 minutes.
Next, we coincide the focus of projection and imaging. This is done through
imaging while projecting grid patterns onto a 1mm stage micrometer with I Om divisions
(Thorlabs, RIL3S2P). By adjusting the distance from LCoS to optical microscope, it is
possible to have the projection and imaging in focus simultaneously, as shown in Fig 2.5.
Figure 2.5 The observation and projection planes coincide with each other. It means the optical path from
LCoS chip to the sample is the same as from the sample to CCD sensor, and they will remain the same
when adjusting focus. Therefore, observation of the synthesized particles in real time without changing
focus is possible. Scale bar: 200ptm.
2.1.2 Microfluidic Channel
2.1.2.1 Channel Preparation
All microfluidic channels in this study are prepared by standard photolithography
and soft lithography. For the photolithography process, a silicon wafer is cleaned with
aceton and isopropanol. SU-8 2000 series photoresist (Microchem) is spin-coated on the
silicon wafer for 30s. A spinning speed of 2000 rpm is recommended for 40[tm thick
0
0
SU8-2025 film. The wafer is then baked for 3min at 65 C and 6min at 95 C on a hotplate.
Next, the microfluidic device is patterned in hard contact mode (MA4, Karl Suss) with a
transparency mask (designed by SolidWorks). The wafer is post baked for I min and 6
min at 65 0 C and 95 0C, respectively. Finally, the patterned wafer is developed with aceton
until no white film is produced when rinsed with isopropanol. This results in 37tm thick
29
channels as characterized by surface profiler (Dektak 150, Veeco). Parameters for other
film thickness can be found from microchem.com. For the soft lithography process, a
10:1 mix of polydimethylsiloxane (PDMS) silicone elastomer (Sylgard 184, Dow
Corning) and curing agent is first poured onto the SU-8 mold. After baking at 60"C for 2
hours, the PDMS is not fully cured but is geometrically stable. Then, microfluidic devices
arc cut from the PDMS and punched at all inlets and outlets. Next, a glass slide coated
with PDMS is partially cured at 60"C for 23min. Finally, each microfluidic device is
bonded to a coated glass slide and baked at 60"C overnight. This partial curing method is
superior to plasma bonding in two ways: firstly, the bonding is stronger [48]; secondly,
PDMS is no longer oxygen-permeable when plasma treated. As all photopolymerization
processes concerned in this study are initiated by free-radicals, which terminate at the
contact of oxygen [49], oxygen-permeability prevents synthesized colloid particles from
attaching to the walls, and ensures high throughput.
2.1.2.2 Flow control in microfluidic channel
Unlike conventional microfluidics experiments, where a micro syringe pump
controls the flow rate, here we use compressed air instead. For one thing, syringe pump
provides volumetric flow that does not stop immediately while pressure flow does; for
the other, high pressure from compressed air gives higher flow rate and therefore
throughput. As shown in Fig 2.6, the compressed air is connected to a precision air
pressure regulator (Omega PRG200-25) and gauge (Omega DPGOOOB-30G) which are
controlled by a digital relay (R410 Pro RS232 relay). The outlet of the valve is branched
into 4 channels. Pressure inside each channel can be regulated independently from zero to
the pressure provided by the regulator through a pressure relief valve (Swagelok, BORS2). Every channel is responsible for an individual inlet of microfluidic device and
therefore they could be controlled independently. By changing the relative pressure for
each inlet, the relative stream widths vary. With the same projection pattern, the
composition of materials within that projection area is changed. In this way, fabrication
of chemically anisotropic particles is achieved, as will be covered in detail in Chapter 3.
30
Figure 2.6 Pressure control system of the digital SFL platform. Compressed air is connected to a 3-way
pressure regulator. Flow is on when the regulator is turned on and off when the regulator is connected to
the atmosphere. The 4 pressure relief valves are connected to the regulator and could output lower pressure
individually. Each valve is connected to a inlet of a microfluidic device, giving independent control of the
inlet pressure and correspondingly stream width inside the channel.
2.1.3 Materials
Most photocurable precursors used in this study belong to multifunctional acrylate
molecules which form tightly crosslinked network upon free-radical induced
polymerization. Typical examples of precursors are poly(ethylene glycol) diacrylate
(PEGDA, molecular weight 250, 575 or 700, Sigma Aldrich) and 1,6-hexanediol
diacrylate (HDDA, Sigma Aldrich), which are classified as oligomers and monomers,
respectively. Poly(ethylene glycol) dimethacrylate (PEGDMA, Sigma Aldrich) and
benzyl methacrylate (BMA, Sigma Aldrich) copolymerize to a shape-memory polymer
(SMP). The ratio between PEGDMA and BMA sets glass transition temperature and its
mechanical properties. The more PEGDMA, the lower the glass transition temperature
and higher the modulus in rubbery state. N-isopropylacrylamide (NIPAAm, Acros
Organics) polymerizes to poly(N-isopropylacrylamide) (PNIPAAm), a temperature
agent N,N'with cross-linking
crosslinked
When
polymer.
responsive
31
methylenebis(acrylamide) (MBAm, Sigma Aldrich), it fornis a temperature responsive
hydrogel which dehydrates above its lower critical solution temperature around the
human body temperature. There is one vinyl group inside each acrylate, methacrylate or
acrylarnide group, which opens up and crosslinks with each other in a free-radical rich
environment. Since there are two acrylate groups inside each diacrylate, dimethacrylate
and bisacrylamide group, the precursors crosslink into a network rather than a long chain,
resulting in a three-dimensional gel. Free-radicals are introduced through illuminating on
photoinitiators. In this study phenylbis(2,4,6-rimethylbenzoyl)phosphine oxide (Sigma
Aldrich) is selected as photoinitiator for its relatively large absorption in the nearUV/visible range. A typical 2%wt of photoinitiator is added, ensuring thorough crosslink
in illuminated regions. The monomer HDDA is most tightly crosslinked, exhibiting high
modulus. The average molecular weight of oligomers determines the average chain
length between neighboring crosslinks, and thus modulus and toughness of resulting gel.
The higher the molecular weight, the softer and more stretchable the gel is.
For added porosity and stretchability, poly(ethylene glycol) (PEG, molecular
weight 200, Sigma Aldrich) and deionized water are added to the precursors. As PEG and
water do not participate in the polymerization process, they leave voids inside the
resulting polymer, increasing average chain length between neighboring crosslinks.
For visualization purposes, a 0.05%wt Rhodamine B (Sigma Aldrich), Sudan I
(Sigma Aldrich) or other food color is added to the solution. These dyes also serve as
photo absorber, which enhance the decay of light along its propagation direction. When
illuminated long enough, this determines how thick particles are independent of exposure
time [42,50].
2.1.4 Control through LabVIEW
LabVIEW stands for Laboratory Virtual Instrument Engineering Workbench. It is
a graphical programming language designed by National Instrument. LabVIEW takes
advantage of the computing power of PC. Unlike traditional instruments, which embed
computing unit inside the instrument and are designed for specific tasks, virtual
instrument connects hardware to a computer. Hardware takes charge of measurements
and computer processes data according to different applications. For better performance,
PXI and VXI are used instead of PC. As a result, the same set of hardware serves
different purposes and functions. A brief introduction to how to control each component
in LabVIEW is listed below.
2.1.4.1 Projecting Patterns
As no LabVIEW driver for the projector is available, an alternative way is taken.
The projector (LCoS) is recognized and configured as a second monitor by Windows. In
LabVIEW, an image is displayed in a window. The window's frame, title bar and scroll
bars are set invisible and background black. It is then fit to a size of 1400x 1050
corresponding to the LCoS's resolution and sent to exactly where the second monitor is.
Equivalently, an image is sent to the LCoS. Fig 2.7(a) shows the LabVIEW code for this
process.
32
2.1.4.2 LED through power supply
The power supply (Agilent E3633A) acts both as driver and controller of the
LED. A LabVIEW driver is available. Fig 2.7(b) shows the LabVIEW code to control the
power supply. In our case, output range is set to 0-8V, 20A, forward voltage fixed at
4.2V if not mentioned elsewise and drive current regulated.
2.1.4.3 Relay
No LabVIEW driver is available for the relay. However, LabVIEW VISA
provides universal I/O interface, which includes RS232 serial port. The code is attached
in Fig 2.7(c), which turns on and off the relay given its address, baud rate, data bits and
other specifications.
2.1.4.4 DSLR camera
Canon EOS camera LabVIEW control (Ackermann Automation) based on
Canon's Digital Camera Software Developers Kit (SDK) is used. Fig 2.7(d) shows a
snapshot of the LabVIEW code, which runs parallel with the main body. In most of time,
the low-resolution electronic viewfinder is used for real-time observation and image
analysis.
(a)
I
K311313 K] 13 LE U
P.&SLALRA K3 0
0
0
0
a
a
12 U U "
131313
IF.
----------
I
U
i
enot Out
ento
MOM01
in (no em
L05.0i
R~rrwvrrr~rrrlr
LIrw
a0a""aUU"
MR-M
MR-M-MMOM B
a aa 0a 0
(b)
VISA rae
-M
Current Limit 0 A)
n(aVISA
33
ource
name
out
(c)
r1, Default
le
'71
254
e
1
MW
LwQ
termination cha
r
A
=\n' = LF)
Enable Termt nation Char (T)
timeou t (10sec)
Th
(tnone)
stop bits (10:1 bit)
flow cont rol (8:none)
I
LZZiJ~1~
lwrie binmry deta out to seriaL
baud ri
data bits (8)
JUTT in (no WWr)J
lose seral port (ifOuit flag is set to true).I
(d)
sm
9"I~ewt
Ti"
_0<
-
t~w~
Il
Figure 2.7 Automatic control of the SFL platform through LabVIEW. (a) An alternative way of sending
images to the LCoS chip. The LCoS is recognized as a second monitor by windows. The image is first
displayed in a window and moved to the second monitor. (b) Control of the power supply through its
LabVIEW driver. The voltage is fixed while electric current regulated. The range is 0-20A. (c) Control of
the RS232 relay through a typical LabVIEW VISA application, given the address, baud rate and other
specifications. (d) Control of the Canon DSLR camera through a third-party LabVIEW driver. Shown here
is the real-time display of viewfinder.
2.1.4.5 Microscope sample stage and focus
ProScan III motorized stage system from Prior Scientific is used. Though no
LabVIEW driver is available, it allows for control through Terminal. A LabVIEW code
emulating hyper terminal is adapted and embedded into the main body. The code controls
X, Y motion of the stage and focal plane of the objective.
34
2.1.5 Scalability
One fundamental limit to photolithography is scaling up. Industry takes a stepand-repeat process to print fine features on large silicon wafers. In research, however, it
is sometimes good enough to scale the size up with coarser resolution. In SFL, switching
objectives of different magnification scales the size up. One pixel on LCoS is about
10pmx I0[tm. Without the consideration of point spread function, which will be discussed
in section 2.2.1, a 5:1 projection results in 2pmx2ptm and a total projection area of
2.8mmx2.Imm; a 100:1 projection gives 0.1pRmx0.Ip m pixel size and 0.14mmxO.105mm
projection area in a linear way. Fig 2.8 provides experimentally acquired relationship
between pixel size and objective used. When doing the measurement, a feature pattern is
projected onto a Imm stage micrometer (Thorlabs, R1L3S2P). An image is taken
containing the projected pattern and the stage micrometer by the CCD camera. A
homemade image analysis code then extracts the features from the image and compares
distances between these features with the stage micrometer. An original pixel size of
11.2pmx 10.5p.im is deduced for the LCoS. As a result, the lithographic area scales up all
2
the way from 0.157mmx0.111mm to 3.16mmx2.22mm, or 0.0174mm to 6.02mm2. It is
even possible to have magnified projection with reduction lens instead of magnification
objective.
Projected Pixel Size vs Magnification
2.5
short side
long side
2
'
1.5
0
0
20
60
40
Magnificafion(X)
80
100
Figure 2.8 Experimentally measured pixel size versus objectives with different magnification. Blue line
corresponds to pixel size of the long side and red line of the short side. The LCoS pixel is thus rectangular
rather than square, with an estimated size of 11 .2pmx 10.5pm.
2.2 Optical Characterization
2.2.1 Point Spread Function measurement
As the name suggests, point spread function (PSF) denotes how a point input
spreads out after passing through a system. The major contribution to PSFs in real
systems is from diffraction. As light from a point in the object passes through any optical
35
aperture, it diffracts due to the finite size of aperture. In ideal case, the diffraction pattern
is an Airy disk. When two Airy disks move closer and closer, it becomes more and more
difficult to resolve them. Lord Rayleigh defined the famous "Rayleigh criterion" that
when the maximum of the first Airy disk coincides with the first minimum of the second,
they can be considered resolved. The separation between the two maximum is thus
defined as the resolution of this system.
Obviously, the real image is the convolution of input and PSF.
0(x, y) =
if
I(u, v) - PSF(u - x, v - y)dudv
(2.2)
Mathematically, Fourier Transform converts convolution into multiplication and
brings the system from spatial domain to frequency domain.
o(ffy) = I(f, fy)-MTF(f,,fy)
(2.3)
MTF in the equation denotes modulation transfer function, which is the Fourier
Transform of PSF. The amplitude of MTF is sometimes referred to as optical transfer
function (OTF).
To characterize optical performance of SFL, the OTF and PSF are experimentally
determined. Random noise images are used as input to excite all frequency components
(Fig 2.9(a)). The projections are recorded by the CCD camera (Fig 2.9(b)). MTF is
obtained by inserting the Fourier Transform of input image and recorded projection into
equation (2.3). PSF is obtained by taking the inverse Fourier Transform of OTF. This
process is repeated 5 times with different input images for each objective. The resultant
PSFs are averaged and given in Fig 2.9(c). Since the LCoS pixel is not perfectly squared,
there is a little difference in PSF along the long and short side of pixel. According to the
Rayleigh criterion, the resolution along both sides is about twice the LCoS pixel size
under that magnification. Fig 2.9(d) shows a representative comparison of PSF along the
long and short sides, under 50X magnification. Although the minimums occur at the
same positions, the PSFs along the short side generally have lower side lobes, causing
less cross-talk between adjacent pixels and thus higher resolution.
According to the measurement, submicron fabrication is possible with high
magnification objectives (50X and IOOX). Note that, for IOOX, the distance from peak to
the first minimum of PSF along the long side of LCoS (225nm) reaches the diffraction
limit (405nm/2N.A.(0.9)~225nm). This does not mean we have a diffraction-limited
optical system. The first minimum does not reach zero along the long side of LCoS and
there is no minimum along the short side, which restricts the performance from being
ideally diffraction limited. Also noted, however, the measurement here is PSF for both
the projection part and the observation part. The limit of fabrication is determined solely
by the projection part. As a result, the real PSF for fabrication might be slightly better
than what we measured in this experiment in the sense that it has a lower minimum.
36
(a
(0
.
(d)
Point Spread Function
Comparison of Point Spread Fusnetions aogthe long and short sides
V
08
I
I-lox
I-oxt
/
11 i
IsI
05
0
0
4Ln~na
PI-~.
1
~ ~ .4-
n
5~~
b.
0
It1,111 L41a PIIII'noL
Figure 2.9 Characterization of the point spread function. (a) and (b) show the projection and observation of
one random noise pattern, respectively. Random noise is used to excite all frequency components to avoid
mathematical infinities during the calculation of MTF. Five such measurements are averaged to obtain the
MTF for each objective. The OTF is obtained by taking the amplitude of MTF. (c) The inverse Fourier
Transform of OTF gives PSF. Since the LCoS is of rectangular shape, the PSFs along the long (bottom
part) and short (top part, lifted by 0.45) side are slightly different. The short side has lower side lobes. (d) A
representative comparison of the PSFs along the long side (blue dashed line) and short side (red dash dot
line). The minimums occur at the same position, while the short side has lower side lobes, meaning less
cross talk between adjacent pixels and thus higher resolution.
2.2.2 Grayscale Projection for Cross-Link Density Control
As mentioned earlier, one advantage of dynamic mask is it modulates light
intensity directly, rather than controls the duty cycle and generates illusion of grayscale.
The photoinitiator releases free-radicals upon absorbing UV light [51]. Since we
generally have an over-dose of photoinitiator, the brighter the illumination is, the more
free-radicals and crosslinks we have. With gray-scaled projection image, it is possible to
manipulate crosslink density spatially, and thus modulus and swelling behavior in a
simple fashion [52,53].
37
Fig 2.10(a) shows the experimental relationship between input grayscale and
projection intensity. The projection grayscale is measured by averaging the intensity of a
CCD image, and then normalized by the 255 projection measurement. This result shows
good linearity over the grayscale range of 17-204, demonstrating our ability to
continuously and precisely control light intensity. Fig 2.10(b) shows discoids synthesized
inside microfluidic channel with different grayscales. The contrast with surrounding
uncured materials reflects their refractive index difference, an indicator of cross-linking
density.
(a)
(b)
Grayscale Response of LCoS Chip
.
100
80-
60-
40-
20'-
0
0
50
100
150
Grayscale(0-255)
200
250
Figure 2.10 Grayscale projection for cross-link density control. (a) Experimentally measured relationship
between input grayscale and output light intensity (normalized by the intensity at 255 grayscale projection)
by CCD. The curve shows good linearity. (b) Discoids synthesized by grayscale 255, 230, 205, 180, 155,
130 inside a microfluidic channel. The contrast with surrounding material shows difference in refractive
index, an indicator of cross-link density. Inset shows the corresponding lithographic pattern. Scale bar:
50im.
38
3. Basic Functions of Digital Stop-Flow Lithography
3.1 Experiment Setup
3.1.1 Microfluidic channel setup
To setup the experiment, 200tL pipettes are cut to fit the inlets of microfluidic
device. These pipettes are then loaded with photocurable resins and inserted into the
inlets of microfluidic device as shown in Fig 3. 1(a). One rubber tube then seals the top of
a pipette by elasticity and connects it to the air pressure controller. In experiments, the
sealing works well up to 10psi. Once air pressure is turned on, the pressure imbalance
drives fresh resin inside the microfluidic channel at high speed.
3.1.2 Residue Pressure Adjustment
However, as the liquid column at inlets and outlet are usually of different height,
there exists a hydrostatic pressure p=pgh, which drives the flow. If the inlet fluid is
higher, there is forward flow; if it is lower, there is backward flow. In both cases, the
formed structures are smeared out due to the flowing stream and finite exposure time,
which is the same issue faced by continuous flow lithography [44,54]. This sets a limit of
practical resolution. As the fabrication process goes on, more resins enter the outlet side
and the pressure imbalance keeps growing.
One way to alleviate this process is to increase the cross sectional area of outlet
reservoir, so that the elevation of liquid surface will be minimal. As shown in Fig 3.1(a),
a centrifuge tube is punched and connected to the outlet of microfluidic device through a
steel tube. The centrifuge tube is preloaded with Dl water so that the pressure imbalance
is minimal. One advantage of this setup is that a leaning centrifuge tube maximizes its
cross section and the elevation of liquid surface is negligible; it's also convenient to slide
the centrifuge tube along the steel tube to initiate the pressure balance. Another advantage
is that all synthesized colloid particles are collected in the centrifuge tube and are ready
for a centrifugal separation. This avoids transfer of particles from the reservoir to a
centrifuge tube and increases the collection yield.
3.2 High through-put fabrication of complex structures
3.2.1 Modeling of the stop-flow process
All devices in this study are PDMS channels of rectangular cross section bonded
to a cover glass coated with PDMS. With a 37pm height and 200dm width, these
channels belong to low aspect ratio channels, tending to bulge under imposed pressure.
How fast PDMS responds to drive pressure and recovers to its initial state determines the
stop time of a stop-flow-lithography cycle. This also sets the time scale for initializing the
flow. The framework of the modeling follows [44, 55].
39
3.2.1.1 Elasticity of PDMS
The top wall of the device is a few millimeters thick, and can be considered semiinfinite considering channel height of only a few tens of microns. The bottom wall is a
thin layer of PDMS on a rigid cover glass (Young's modulus EPDMS- IMPa,
Egiass~62GPa), whose deformation is negligible compared to the top wall.
We assume that throughout the experiment, PDMS stays in its linear elastic
region and satisfies o=&E, where 5 is the stress, P the strain and E the Young's modulus.
To calculate strain, a characteristic length scale is needed. Since the material is semiinfinite, strain vanishes at large distances, rendering a characteristic length scale being the
channel width W, rather than device thickness or channel height H [56]. In addition, for
low aspect ratio channels, the deformation in the lateral direction is proportional to height
H, which is much smaller than the corresponding deformation in height proportional to
width W. This deformation can thus be neglected in the analysis [55]. Therefore, the
strain scales with the deformation of the channel height Ah, channel width W, external
pressure P and Young's modulus of PDMS according to (see Fig 3.1 (b) [44] for a
schematic of the channel geometry)
Ah P
(3.1)
W
E
As a result, the channel height deformation is proportional to local pressure
according to equation (3.1). In addition, as the pressure drops along the channel, the
deformation/strain decreases accordingly. The largest deformation in turn occurs at the
entrance of the channel
(3.2)
Ahmax~PW/E
(a)
(b)
PDMS
deformed PDMS
relaxed POMS
dii PowS
glass Support
Figure 3.1 (a) Microfluidic channel setup with a micro pipette loaded with photocurable resin at the inlet.
The rubber tube delivers pressure and seals the pipette through elasticity. The outlet is connected to a
centrifuge tube to balance the residue pressure. (b) shows the simplified 2D model of the deformed
microfluidic channel, where the height is averaged across width at all cross sections. Inset shows the bulged
cross-sectional view of the channel when the external pressure is turned on. b Figure reprinted with
permission from ref. [44],
2007 RSC.
3.2.1.2 Laminar flow
To simplify the three-dimensional fluid problem to a two-dimensional problem,
the average, of channel height is taken over the channel width at all cross-sections along
the length of the microfluidic device to obtain h(x), as shown in Fig 3.1(b). In
microfluidic channels, the Reynolds number is much smaller than 1, allowing for the
40
lubrication approximation and the Navier-Stokes equation simplifies to
dP
0 2v
-- =
yX
2 (3.3)
ax
az 2
where p is the viscosity of photocurable material of interest. The continuity equation
allows us to calculate v.
= 0
(3.4)
ax
az
By assuming small curvature of the top wall compared to the length of the
channel L, boundary conditions at the top and bottom walls become
-+
vZ(z = h(x)) = -h at
vz(z = 0) = 0
Combining equation (3.3) with boundary conditions, we have
1 dP
VX = 1-a- [z2 - zh(x)]
(3.5)
(3.6)
(3.7)
3.2.1.3 Coupling of elasticity and fluid flow
When the external pressure is turned off, elasticity of bulged PDMS channel
forces it back to its original shape, squeezing excess fluid towards inlets and outlets. The
characteristic timescale associated with this process T, measures the stop time of a stopflow-lithography cycle.
During the process, the pressure exerted on the fluid by the channel is assumed to
be proportional to the local strain according to equation (3.1).
(3.8)
h(x)-H
Ah(x)
=E(3)
P(x) =E
W
W
Differentiation of the above equation (3.8) and combining it with equation (3.7)
lead to
VX =
E dh
-[zz
21 W ax
- zh(x)]
(3.9)
Building a relation between equation (3.9), (3.4) and boundary condition (3.5) we
have
ah
at
h 2 E [(h
-=-
41M
+hd 2 h(
-dx
ax
+d3 ax2
(3.10)
With a scaling analysis commonly used in fluid mechanics, we have channel
height h(-H), deformation Ah(~PW/E), x(-L) and t(~'t). For the partial derivatives in
(3.10), the scaling reads
dh PW
t
(3.11)
2
2
(dh
PW
(3.12)
ax)
(EL
)2h\
PW
(
hEL2 (3. 10) gives
Plugging the above scalingsPuggn
into equation
41
(3.13)
~
1
-~
(3.15)
H 2E
W
PW H
-- + -(3.14)
Since the deformation is small compared to channel height, the first term in the
bracket is negligible and the final result reads
12pW Lz
H
sEH3
Plugging in the channel geometry and material properties W=200pm, L=lcm,
H=37pm, E=lMPa, p~57cP (PEGDA 575, 25C), we have rs~300ms. In experiments, a
500ms stop time is enough for the flow to stabilize, which agrees well with the model.
As will be discussed later in 3.2.4, the fabrication throughput is inversely
proportional to the time period of one fabrication cycle. For current SFL platform and
microfluidic channel geometry, one cycle is -500ms while the stop time is -300ms,
making the dominant contribution. According to equation (3.15), to decrease the stop
time, one either decreases the fluid viscosity p., channel width W and channel length L, or
increases channel's elasticity E and channel height H. E is determined by curing
condition of PDMS, which could hardly change. Viscosity p. can be lowered by switching
from PEGDA 575 and PEGDA 700 to PEGDA 250 [57]. Further reduction is reached by
mixing PEGDA with PEG (average molecular weight 200) and DI water. In addition, the
stop time is related to the square of L and cube of H, making it very sensitive to channel
geometry. A reduction in channel length and increase in channel height easily decrease
the stop time by one order, making it a minor contribution to the fabrication cycle period.
With the exposure time and flow time fixed, the throughput can twice as high.
3.2.2 Geometrical anisotropy
The most basic function of our digital SFL platform is to fabricate large quantities
of colloid particles with almost arbitrary 2D structures, similar to that of traditional SFL
[44]. During the fabrication process, a digital image is sent to the LCoS chip. The chip
reconfigures itself and serves as a photomask. The patterned UV light then focuses into
the microfluidic channel and generates particles of exactly the same shape. As has been
demonstrated in Chapter 2, the PSF of our system is about two LCoS pixel size under the
corresponding magnification. Any pattern with spacing/gap below 2 pixels is
meaningless.
As mentioned in section 2.1.1.2, light intensity, exposure time, curing depth
satisfy
(3.16)
= exp (ad)
where t is the polymerization time, D proportional to the reactivity between photocurable
material and photoinitiator, I the light intensity, a and apj the absorption coefficient of
resin and photoinitiator, and d the cured depth. In our experiments, the light intensity and
exposure time result in a depth larger than the height of microfluidic channel. The
thickness of synthesized particle is therefore determined by the height of channel and the
inhibition layer [44,49].
In this subsection, fabrication of standard geometrically anisotropic particles is
demonstrated with different materials, including functional materials (smart materials).
D
D
42
The access to a selection of materials with varied functions is vital for designing and
synthesizing building blocks preprogrammed for self-assembly.
3.2.2.1 PEGDA
Poly(ethylene glycol) diacrylate (PEGDA) is the most common material used for
SFL. It is based on poly(ethylene glycol) structure, with two ends replaced with acrylate
groups, shown in Fig 3.2(a) and (b). It has the fastest response to UV exposure among all
compatible materials we have investigated. In our setup, when the UV LED is set to
4.2V, 12A, a 100ms exposure under 20X objective cures pure PEGDA and leaves it
visually observable. Fig 2.10(b) shows PEGDA700 cured with grayscale at 4.2V, 15A,
SOOms, 20X. Fig 3.2 (c)shows test pattern cured at the same condition. Inset shows the
photomask. As we can clearly see, horizontal lines are discernable up to 2 pixel pitches,
while verticle lines blur out before 3 pixel pitches. This agrees well with our PSF result in
Chapter 2, where under 20X magnification, the first minimum occurs two pixels away
from the maximum. The PSF measurement also shows different performance along the
long and short sides of LCoS pixels. The short side, which corresponds to horizontal
lines, has lower side lobes in its PSF and thus better resolution than vertical lines. Fig 3.2
(d) shows examples of identification numbers synthesized inside the microfluidic
channel, proving it a good candidate for geometrically anisotropic particle synthesis.
Another advantage of PEGDA is its sensitivity to UV exposure. Compared to other
materials, PEGDA requires the least optical power and cures fast. Moreover, PEGDA of
20% or lower concentration shows good cell viability when cures upon UV exposure
[12]. Its biocompatibility makes it possible for biological and biomedical applications.
(a)
H
C
(b)
0
n
H2C
OH
0
0
0*
0
CH2
(d)
Figure 3.2 Structural formulas for (a) PEG and (b) PEGDA. (c) Resolution test pattern, with pitch size of
10, 5, 3, 2 ,J pixels, respectively. Inset: photo mask; (d) SEM image of identification numbers. Scale bar:
50ptm.
43
3.2.2.2 HDDA
1,6-hexanediol diacrylate (HDDA) is a monomer commonly used in
stereolithography. Unlike PEGDA, which is hydrogel and swells in contact with water,
HDDA is stiff compared with other gels, since it has much shorter chains between
acrylate groups (Fig 3.3(a)) and is hard to stretch. As a result, HDDA shows better
performance when it comes to fabrication resolution. Fig 3.3(b) shows HDDA tested with
the same lithography pattern as in Fig 3.2(c). Compared with PEGDA, HDDA
successfully discerns 2 pixel pitches for both horizontal and vertical lines, reaching the
limit set by PSF and Rayleigh criterion. Fig 3.3(c)-(e) demonstrates HDDA's capability
of fabricating complex structures and control through grayscale.
Pure HDDA requires slightly lower optical power than pure PEGDA. In our
setup, 4.2V, 10A, 100ms under 20X magnification synthesizes visually observable
particles. In addition, unlike PEGDA, HDDA is not very sensitive to increased optical
power. For PEGDA, when illumination is increased from 4.2V, 12A, lOOms to 4.7V,
15A, 100ms under 20X, PEGDA is fully cured. While for HDDA, 4.7V, 20A, 100ms
under 20X won't fully solidify the flow. HDDA thus has a robust working range, making
it suitable to form Janus particles with other materials, as will be covered later.
(a)0
O)C2
O
H 2C
0
Figure 3.3 (a) Structural formula of HDDA. (b)Resolution test pattern synthesized with the same photo
mask as shown in Fig3.l(c). (c) Micro-gear and (d) Identification numbers. (e) Grayscale projection with
the same photo mask as used in Fig 2.10(b) synthesized at a more robust range. The grayscale follows 255,
230, 205, 180, 155, 130. Scale bar: 50pm.
44
3.2.2.3 Shape-memory polymer
Shape memory polymers (SMPs) are a class of "smart" materials capable of
switching configurations in response to environmental stimuli such as temperature,
hydration, pH, specific molecules, magnetic or electric field, among others. SMPs with
the shape memory effect triggered by change of temperature have been investigated most
and are referred to as "thermally triggered SMPs". There are two mechanisms for these
thermally triggered SMPs: glass transition and crystallization. The former one is widely
used and is also the mechanism for the SMP used in this work. In some sense, all
chemically cross-linked amorphous polymers belong to this category. To achieve the
shape memory effect from glass transition, we first deform an SMP sample at a
temperature higher than its glass transition temperature Tg, when the SMP is at its
rubbery state and has polymer chains with extreme mobility; then, we decrease the
temperature to a point below its T. while maintaining the external load that causes the
deformation. At this low temperature, the polymer enters its glassy state. Its chains lose
their mobility and numerous physical crosslinks are formed. After release of the external
load at low temperature, most deformation is kept which forms a temporary shape. The
sample recovers its permanent shape upon heating to the original high temperature, as the
deformed polymer chains regain their rapid chain mobility and trend to return to their
energetically most favorable conformation due to entropic elasticity. This large
deformation behavior spanning glass transition temperature can be qualitatively
explained by a three mechanisms model as shown in Fig 3.4(a) [58,59,60].
The first micromechanism accounts for intermolecular resistance. The spring
represents intermolecular energetic bond-stretching while the dashpot accounts for the
thermal-activated plastic flow like rotation of local chain segments.
The second and third micromcchanisms accounts for intramolecular resistance. In
the second one, the springs denotes the response of network due to changes in freeenergy while the dashpot also means thermally- activated plastic flow, but this time
slippage of polymer chains.
Note here that, below glass transition temperature, the material is in its glassy
state and there is negligible plastic flow. We assume the dashpot allows no plastic flow so
that micromechanism 2 is purely elastic. Then we may find that micromechanisms 2 and
3 can be combined to one spring. Two micromechanisms describe the main features of
thermosets under glass transition temperature (or thermoplastics spanning glass transition
temperature). As a result, the behavior of thermosets and thermoplastics are very similar
below glass transition temperature. However, above the glass transition temperature, the
material enters its rubbery state and becomes a viscous fluid. Plastic flow is allowed in
this regime and with higher temperature, the mobility of polymer chains increase
significantly. As a result, the resistance to plastic flow drops off quickly above glass
transition temperature.
The third micromechanism represents resistance due to chemical cross-links.
These cross-links do not slip and thus there is no dashpot. As a result, thermosets show
no permanent-set above glass transition temperature since the elastic response of
chemical cross-links is enough to bring the material back to its permanent shape when the
resistance to plastic flow due to mechanical cross-links is negligible.
However, not all chemically cross-linked amorphous polymers are of interest in
45
the sense of application. Only those with a reasonable glass transition temperature,
especially when the glass transition temperature is close to our body temperature are of
particular interest. The shape-memory polymer (SMP) used here is based on benzyl
methacrylate (BMA) with poly(ethylene glycol) dimethacrylate (PEGDMA) as
crosslinker, illustrated in Fig 3.4(b) and (c). Fig 3.4(d) shows its dynamic mechanical
analysis (DMA) results for minigels with different constitution of BMA and PEGDMA
with 5% photo initiator. The data were measured by cooling the test sample from 120'C
to 0 0 C at rate of 2 0 C/min with 1Hz loading frequency. As can be seen, the glass transition
temperature can be tuned from below body temperature to about 50*C by changing the
ratio of BMA and PEGDMA. The modulus in glassy state in all occasions is at least one
magnitude higher than that in rubbery state. This particular kind of SMP requires high
optical power, more than 3 seconds at 4.7V, 20A to be visually observable, making it a
poor selection to combine with other materials. PEGDMA particles synthesized inside a
microfluidic channel is shown in Fig 3.4(e)(f). The resin is slightly overcured for
visualization purposes. One of the future work is to characterize the mechanical behavior
of SMP in the form of microgel.
(d)
(a)
Dy=unic Mechmaical Aamlysis
10
-MA70%+fl*3MA30%
10
(b)
0
~C~oO
O~10
20
()
C3
0
40
60
80
too
120
Tcperamwe(C)
Figure 3.4 (a) Three-micromechanism model of springs and dashpots that qualitatively explains the
thermo-mechanically-coupled large-deformation behavior of amorphous polymers in a temperature range
spanning their glass transition temperature. (b) and (c) show the structural formulas of BMA and
PEGDMA, respectively. In (d), DMA results for different constitution of BMA and PEGDMA with 5%
photo initiator are given. The data were measured by cooling the test sample from 120'C to 0*C at a rate of
2 0C/min with 1Hz loading frequency. (e) and (f) show high resolution particles of SMP synthesized inside
the microfluidic channel (slightly overcured for visualization purposes). Figure a reprinted with permission
from ref. [60], 0 20 10 PERGAMON. Scale bar: 50tm.
3.2.2.4 Temperature-responsive hydrogel
Unlike PEGDA, which is a neutral hydrogel, there exists functional hydrogels.
46
Poly(N-isopropylacrylamide) (PNIPAAm), for example, performs temperature-dependent
swelling behavior. PNIPAAm hydrogel is polymerized from N-isopropylacrylamide
(NIPAAm, Acros Organics) and cross-linking agent N,N'-methylenebis(acrylamide)
(MBAm, Sigma Aldrich). Their structural formulas are illustrated in Fig 3.5(a)(b). The
resulting temperature-responsive hydrogel reversibly undergoes a coil-to-globule
transition with temperature, where PNIPAAm macromolecule collapses from an
expanded coil state through an ideal coil state to a collapsed globule state. The transition
is directly related to PNIPAAm's solubility behavior upon heating. At low temperature,
PNIPAAm orders its polymer chains to form hydrogen bond with water/ethanol
molecules. This causes water/ethanol to rearrange themselves and lower the total entropy
of the system. PNIPAAm thus absorbs water/ethanol and dissolves in the solvent in this
temperature range. At temperatures above its lower critical solution temperature (LCST),
however, PNIPAAm becomes hydrophobic and dehydrates itself. With three different
monomers, it is also possible to terpolymerize temperature- and pH-sensitive PNIPAAm
[61].
0
CH 3
H2C
CH 2
H2C
H
0
0
(d
(c
Teaqrahx-Sensitive Respoie of PNIPAAm
*
17
16
13
12
11
10
15
20
25
30
35
40
45
50
55
TePAe(*c)
Figure 3.5 (a) and (b) show structural formulas of NIPAAm and MBAm, respectively. (c) and (d) compare
PNIPAAm discoids synthesized by our SFL platform before and after swelling in ethanol. (e) gives the
temperature-sensitive swelling behavior of PNIPAAm discoids in water. All PNIPAAm hydrogel here is
polymerized from a mix of 100 mL ethanol, 70g NIPAAm, lg MBAm and 5g photoinitiator.
Fig 3.5(c) and (d) show PNIPAAm discoids synthesized by our SFL platform
undergoing large swelling in ethanol. Fig 3.5(e) is the experimentally measured
temperature-sensitive free-swelling behavior of PNIPAAm in water. Swelling ratio is
measured by comparing the diameter of PNIPAAm discoid before and after swelling. The
47
result shows a -5 times volume change, which agrees with other report [62]. The
PNIPAAm used here is prepared by mixing 100 mL ethanol, 70g NIPAAm, 1g MBAm
and 5g P.I. Polymerized particles are visually available at experimental condition of
4.7V, 15A, 100ms.
3.2.3 Chemical anisotropy
From individual studies of above four materials, PEGDA and HDDA share
similar curing condition. As a result, it is possible to cure PEGDA and HDDA across
their interface to form chemically anisotropic particles, or Janus particles.
Until now, the advantage of our digital SFL platform is the introduction of SLM
and its unprecedented ability to dynamically reconfigure high-definition lithographic
patterns. Microfluidic device serves as a means to flush out synthesized particle with
fresh photocurable resin. It contributes mostly to the throughput of our system. During
the fabrication of Janus particles, however, the laminar nature of microfluidics and
control of external pressure is the key to the synthesis and manipulation of Janus particles
with varying composition.
In a multi-inlet microfluidic device, the relative pressure of each inlet controls
how wide the stream is. By changing the external pressure, it is convenient to manipulate
the width of each stream, as shown in Fig 3.6 (a)-(c). The same projection across the
interface as indicated by the white dashed box results in Janus particles of different ratio
between material A, B, C, D. Limited by material selection, we demonstrate Janus
particles made of PEGDA and HDDA. As discussed before, both materials polymerize
through recombination of vinyl groups in diacrylate, making it possible for them to
chemically crosslink together. The same optical power works for both, making it
experimentally feasible. The curing condition for PEGDA is taken as HDDA is more
robust upon illumination. Fig 3.6 (d)-(h) show the same star pattern synthesized inside a 2
inlet 1 outlet device. Inlet 1 flows HDDA (transparent) and inlet 2 PEGDA 700 (red).
During the demonstration, the ratio between P1 and P2 is gradually changed, resulting in
composition of star particle from 1:0 ratio of HDDA to PEGDA to 0:1.
(bi
(d)
)
Figure 3.6 (a)-(c) conceptually show tuning chemical composition by controlling inlet pressure
individually. When (a) Pl=P2=P3=P4, (b) P2=P4>PI=P3 (c) P4>PI=P2=P3, the hypothesized particles in
the white dashed box have different chemical composition, though the projection patterns remain the same.
(d)-(h) experimentally validates the concept. Inlet I is transparent HDDA and inlet 2 red PEGDA 700. The
same star pattern is projected with varying pressure ratio between inlets, rendering star patterns of various
compositions. White dashed line indicates the interface between HDDA and PEGDA. Scale bar: I 00pim.
48
This fabrication technique, together with the geometrically anisotropic synthesis
capability and a pool of functional materials makes our digital SFL platform appropriate
for the fabrication of heterogeneous metamaterials in self-assembly.
3.2.4 Fabrication speed
For basic functions, there is no need for imaging. One cycle consists of stop, flow
and lithography. As discussed above, stop time is -300ms, lithography time -100ms. A
flow time of 100ms is more than sufficient with ipsi external pressure. Thus, one cycle
period is about 500ms. In a 200pm wide channel, as in our case, it is convenient to
synthesize -10 micro-particles in one exposure, resulting in a throughput of 10 5 particles
per hour. In extreme case, when higher magnification objectives are used, or when highresolution physical photo mask is used, 100 micro-particles can be synthesized in one
exposure, giving a throughput of 106 particles/h.
As talked about at the end of 3.2.1.3, by switching to low viscosity materials,
reduce the channel length and increase channel height, the stop time could be easily
lowered by one order to -30ms. This improves the throughput by a factor of two.
Another improvement could be made by increasing number of particles
synthesized in one exposure. Currently most experiments are carried out under 20X
objective. However, a 200pim wide channel covers -1/3 of the projection area of our
LCoS chip. For lower magnification, which is preferred in many cases, the number of
particles fabricated in one exposure is even smaller. The utilization of a higher
magnification objective or a wider channel solves the issue. Although increase in channel
width linearly prolongs stop time, this could be compensated for by further increase the
channel height. At low magnification, the focal depth is large enough that no deformation
should be observed with deep channels. Whole coverage of the projection area gives a
three times enhancement to throughput.
In addition, an increase in optical power could lower the lithography time and
thus increase throughput. Current UV LED provides -3W under 4.8V, 20A. However,
after the diffuser, collimator and polarizers, we expect ->90% attenuation. A newly
purchased DLP projection module (Star-07 DLP® Projection module, VIALUX)
provides >2W output power of the entire system with a response time of -1 Os, which is
10 times higher than our current output. This makes the lithography time -Oms.
The flow time could be easily reduced to -<30ms by slightly increase external
pressure. This together with the decrease in lithography time and stop time gives a cycle
period of-OOms, or a five times increase in throughput.
Taking all the above into account, the SFL platform has a potential throughput of
~107 particles/h, which is -15 times higher than the current one. The next step for this
project will be experimenting with different concentration of photocurable materials to
find out the optimum that minimize the total of exposure (increase with lower
concentration) and stop time (decrease with lower concentration/viscosity). Also we will
implement the DLP projection module to our setup to realize higher contrast and optical
power. This should help us get close to one order of magnitude of increase in throughput.
49
4. Adaptive Graphical Encoding
Utilization of a dynamic mask not only allows us to reconfigures projection
pattern at negligible cost, it makes possible reconfiguration in real-time, according to in
situ information. When the stream flowing inside a microfluidic channel is not pure
photocurable resin, but a suspension of particles/cells in the resin, it is possible to detect
the particles/cells through processing the image taken by CCD camera. The positions and
even categories of these particles/cells are decided, and encapsulation patterns are
generated and sent to the LCoS chip. The pattern then precisely encapsulates the
suspensions, forming colloidal shells outside them. On the one hand, this extends our
material selection pool for the building blocks for self-assembly, as materials
incompatible with direct SFL fabrication could be accessed by encapsulation; on the
other hand, this capability applies to single cell encapsulation and graphical encoding. In
this chapter, graphical encoding is taken as an example to demonstrate the adaptability of
our SFL system.
4.1 Rotation, off-set and scale adjustment
Encapsulation of cells inside functional hydrogels is an important research effort
for drug delivery and tissue engineering. Conventional methods include lithographically
synthesizing hydrogels with cells in them. However, this approach works at extremely
low throughput. When SFL was first introduced, researchers applied the technique
instead of lithography for high throughput cell encapsulation [12]. With different physical
photomasks, the throughput is as high as ~10 3-10 4 particles/h. However, the
encapsulation process is based on statistics. Before characterizing the hydrogel under
fluorescence microscope, we don't know how many cells are inside one hydrogel particle
and where they are. With the spatial light modulator of our digital SFL, however, it is
possible to dynamically reconfigure lithography pattern according to the amount and
positions of cells within projection area and encapsulate exactly one cell in each hydrogel
particle. This adaptive encapsulation capability is the major advantage of our digital SFL
platform. It generates projection patterns according to the visual information inside the
microfluidic channel. To achieve this, the locations of beads are recorded by the CCD
camera, and the corresponding locations on the LCoS are calculated. In other words, a
one-to-one mapping from the CCD pixels to the LCoS pixels is desired. A two-step
determination of such mapping is done in experiment. Firstly, a rotation of the CCD
camera ensures its pixel array is parallel with that of LCoS. Secondly, the mapping is
measured experimentally for later adaptive projection.
4.1.1 Rotation
As stated above, the mapping from CCD pixels to LCoS pixels is required for
adaptive projection. To determine the mapping experimentally, a simple relation is
preferred, where the pixel array of CCD camera is parallel to that on LCoS when focused
on the sample stage. In this way, the position functions xLCos(xCCD,yCCD),
yLCoS(XCCD,yCCD) simplifies to XLCoS(yCCD), yLCoS(XCCD). To achieve this, a rotation
alignment is performed before experiment. A grid pattern is projected, imaged and
50
displayed in the LabVIEW front panel. Reference lines are given in LabVIEW. The
camera is rotated until the grid lines are parallel to the reference lines. In this way, the
pixel arrays of CCD camera and LCoS are parallel when focused inside the microfluidic
channel. The mapping from CCD pixel to LCoS pixel is neat.
4.1.2 Offset and Scale Adjustment
Then, the mapping relation is experimentally measured. With the objective of
interest mounted, a prescribed feature pattern is projected onto a clean glass slide. A
digital image is taken by the CCD camera, which is already rotation aligned. LabVIEW
then reads the image and calls MATLAB to extract those features and determine their
locations. The locations are then related to the original projection image, which finally
determines the mapping relation.
4.2 Image acquisition and analysis
With the mapping relation known, the next step is to acquire, process, analyze and
understand what is going on in the microfluidic channel with the images taken by CCD
camera, and decide what actions to take (in terms of projection pattern in our case). This
falls into the field of computer vision. Image processing and image analysis are powerful
tools for 2D images, which transform one image into another by means of contrast
enhancement, feature extraction etc. These techniques do not depend highly on prior
knowledge of the image content, providing robust results for our experiments. Parameters
are optimized for better performance in specific experiments given solution, suspensions
and photoabsorber/fluorescent dye.
4.2.1 Image acquisition and processing
In our SFL platform, observation and projection shares the same optical path in
the inverted optical microscope. To allow for UV light of projection and minimize UV
light from illumination, only transmission mode with an UV filter is used. However, in
high magnification, strong illumination is needed for CCD camera, and unwanted
polymerization is initiated by the excess illumination light. This is not an issue when
generating large quantities of prescribed building blocks, as demonstrated in Chapter 3,
where the illumination is not necessary and can be turned off. However, in experiments
where adaptability is important, this imposes a time limit for the experiment. In general, a
DSLR camera's CCD sensor provides higher sensitivity (ISO) than the viewfinder,
allowing less intense illumination and longer experimental time. Besides, the CCD sensor
has higher resolution (5184x3456 versus 1056x704), which is extremely helpful in
distinguishing different micro-particles. The trade-off is that high ISO setting also
amplifies noise and produces low signal-to-noise ratio (SNR). In addition, the CCD
sensor does not provide real-time monitoring as the viewfinder does. It takes and saves
images only at designated time. High-resolution image also takes much longer time for
processing, decreasing throughput significantly. As the efficiency of image processing
and analysis scales up with the size of images, for high resolution images taken by the
CCD sensor, it takes ~2-3 seconds to undergo all the process from image acquisition to
51
pattern generation. In cases where the flow does not stop well, this long processing time
causes mismatch between the projection pattern and actual locations/distribution of
particles. Low resolution images taken by the viewfinder (1056x704, comparable to that
of LCoS), on the other hand, pushes the same image analysis process within 0.2 second,
maintaining the high throughput of SFL. Taking all the above into account, the
viewfinder is used for image acquisition in most cases. High resolution CCD sensor is
only used for particle sorting where resolution plays an important role. The viewfinder is
always for real-time observation.
To avoid distractions like the channel walls or stains, a background subtraction is
performed. 10 images are taken, each after a flush of fresh photocurable materials. The
average is subtracted as background before further image processing. Suspensions in the
stream like micro beads are averaged out, while distractions on the sample like stains and
channel walls remain and are subtracted. Fig 4.1(a) and (b) compare digital images before
and after background subtraction. For visualization purposes, a dusty substrate is used to
demonstrate the effect of background subtraction. After the background subtraction,
almost all the stains are removed. Only a small fraction of the channel walls is remained,
and can be removed during further processing.
4.2.2 Image analysis
There are two key requirements for the whole image analysis process: speed and
accuracy. To trade off between the two, our image analysis is not targeted at specific
shapes or features, and does not have high recognition accuracy as machine learning
does. Based on the binary image from the previous step, a general analysis is performed
to measure connected components (Image Processing Toolbox, MATLAB). Generic
information like area (A) and perimeter (p) for each connected component helps to decide
shape: for spheres or discoids, Ap 2/4n; for squares A- p2 /16; for equilateral triangles,
A~ V3p 2 /36 etc. Other information like eccentricity and Euler number may be helpful in
identifying complex structures.
Before the above analysis, the connected regions are extracted first. For faster
process in consistence with our choice of viewfinder, edge analysis is avoided. On the
one hand, edge analysis has to run for every image; on the other hand, convolution takes
much longer time than subtraction in our case. In our experiment, the maximum
grayscale of the resulting image is multiplied by a dynamic coefficient and then used as
threshold to cut off all pixels below it. The dynamic coefficient is optimized by analyzing
the maximum area of connected regions in an image which includes suspensions. This
process is iterated with increased coefficient until a reasonable value of the maximum
area of connected regions is achieved. This takes -3 seconds at the beginning of
experiment together with background subtraction and the coefficient applies directly for
later adaptive projection. The image is then transformed into black-and-white binary
image as shown in Fig 4.1(c). For comparison, binary images without optimized
processing are shown in the inset. In the top left inset (green box), the image goes
through neither background subtraction, nor optimization of the dynamic coefficient. The
top right one (red box) is processed with the optimum dynamic coefficient without
background subtraction. In both cases, the dominant bright regions are channel walls and
stains due to strong contrast. These white regions impose distractions to the identification
52
process and could cause incorrect decisions. Thus, the above two steps greatly enhance
the signal-to-noise ratio.
Fig 4.1 (d) identifies connected regions in Fig 4.1 (c) and decides based on their
perimeter and area if they are micro beads, remaining dirt or noise. According to the
detected size/area difference, a further decision could be made to categorize those micro
beads. The positions and categories of these identified micro beads are then used to
generate adaptive projection patterns which accomplish specific tasks like creating
tracking information or sortin
((a)
Figure 4.1 Process flow of the adaptive projection. (a) Original digital image of the microfluidic channel
with suspension of PS beads. There are a lot of stains on the substrate. The stains together with the channel
walls provide strong contrast with the surroundings and are major distractions for the identification of
beads. (b) Digital image after background subtraction. The grayscale is adjusted for visualization purposes.
Almost all stains and channel walls are removed. It is superior to (a) in terms of identification. (c)
Remaining distractions are removed by subtracting the image with an optimized dynamic coefficient. The
result is transformed into black-and-white image. Inset compares the output with (top left green box) binary
image without background subtraction or optimization of dynamic coefficient, and (top right red box)
binary image with background subtraction but no optimization of coefficient. The signal-to-noise ratio is
greatly enhanced by these two steps. (d) A decision is made based on the area, perimeter etc. of the
connected regions to distinguish beads (circled out by red) from noise. (e) An adaptive projection pattern is
generated (inset) and encapsulates the detected beads. Scale bar: 100pm.
53
4.2.3 Adaptive projection pattern generation
Once the decisions are made and positions on the CCD are known, the
experimentally acquired mapping is applied to obtain corresponding positions on the
LCoS. Projection pattern is generated according to the positions, categories and
application of interest. When the flow is stopped well, the illuminated pattern should
encapsulate all detected particles. Fig 4.1(e) shows the encapsulated micro bead. Inset
shows the corresponding projection pattern. Suspended micro beads are precisely
encapsulated with hydrogel particles of well-defined shape and size. Not only is there one
bead inside one hydrogel particle, but also the positions of encapsulated beads are
prescribed.
4.3 Graphical Encoding
In a fundamental level, researchers have long been using fluorescent dyes or
proteins to label targets. However, this spectroscopy based encoding method suffers from
a limited selection of indicating material with minimal spectral overlap and delicate
fabrication process [63,64,65,66]. Besides, with each fluorescent exciter and detector
added, the cost goes up. Graphical encoding is advantageous in that it is based on
optically detectable patterns on the micro particle surface and has a relatively large
encoding capacity [63,67,68]. In addition, the visual information in graphical encoding
does not require special training or instrument for decoding. As demonstrated in [69], a
portable decoder composed of a smart phone and an objective is sufficient. This
straightforward and low cost decoding process makes it promising for not only research
purposes, but also everyday use, like drug authentication check at home.
4.3.1 Identification Numbers
As a national identification (ID) number is used by many governments to track
their residents of work, tax, health care etc., ID number can also be applied in the micro
regime to track micro particles of their locations, reactions etc. In the fundamental case, a
unique series of digits is reserved for one particular micro particle, as shown in Fig 4.2(b).
Here, particles are detected and indexed, and ID numbers are generated based on the
indices. These ID numbers are attached to the suspensions as they undergo further
reactions. One application will be to study the dynamics of mesoscopic assembly.
Compared to spectroscopic encoding, which is binary, graphical encoding can be decimal,
which is convenient for non-specialists. To encode a 3-digit ID number as in our case
with spectroscopic approach, 9-10 distinct fluorescent dyes are required simultaneously,
which is inaccessible. For graphical encoding, however, there is no fundamental limit on
the encoding capacity except for size. Thus, graphical encoding is ideal for informationrich applications, especially when a wide spectrum of information needs to be tracked.
54
Figure 4.2 Encoding ID numbers to micro beads. (a) Digital image of suspended beads in a mix of
PEGDA700, PEG and ethanol. (b) The same beads encoded with ID numbers. These IDs are attached to
them through further physical or chemical processes and they are easy to readout for non-specialists. Scale
bar: 50pm.
4.3.2 Barcode
4.3.2.1 1D Barcode
ID numbers are good for small-scale experiments, as they provide relatively large
encoding capacity and user-friendly readout. However, for the ID tag introduced above, it
is not convenient for machine readout. As a result, it is not suitable for large-scale
automated experiments or production. Besides, the information density encoded is not
highest possible. In that case, one decimal digit takes 27x41=1 107 pixels.
To overcome the above issues, I D barcode is used. The barcode is a machinereadable representation of numbers related to the object it is attached to. Specifically, we
choose the Universal Product Code (UPC) as the basis of our encoding and decoding
system. UPC data structures are part of the Global Trade Item Numbers based on
international standards. They are widely used for tracking goods in stores, similar to our
purpose. For feasible optical readout, UPC barcode consists of black bars and white
spaces, representing 12 digits, which is a unique ID number for trade goods. The pattern
of the scannable area of a UPC is SLLLLLLMRRRRRRE, where S (start), M (middle), E
(end) guard bars are the same in every UPC. The 6 Ls (left) and 6 Rs (right) represent 12
unique digits. The last R is an error detecting check digit that corrects errors happened
during scanning or manual entry. All digits have two representations, L with odd parity
and R with even parity. This helps to decide the direction of the barcode in addition to the
S and E guard bar [70].
Currently, the pattern is simplified to SLRE in our system to make the tag
compact, as shown in Fig 4.3. At the same width, one digit takes 7x41=287 pixels
compared to 1107 pixels in the ID number case, giving an increase of 3.86 in information
density. Besides, the UPC system provides some degree of error-detecting capability
even in the SLRE case. The consistency between S and the parity of the first digit that
follows, together with E and parity of the last digit checks the correctness of the detected
barcode. For more complicated experiments or delicate applications, more formal pattern
could be taken, which includes the error detecting check digit. It will not only detect the
existence of an error, but also correct it under certain circumstances. Most importantly,
these barcodes are machine-readable, and are suitable for large-scale experiments where
machines instead of researchers keep track of all the targets. This function can also be
implemented in the SFL LabVIEW program to enhance the adaptive projection. For
example, different projection patterns are generated according to the detected barcodes in
real-time. In addition, compared to the ID number, barcodes are better suited for SFL in
55
case of backward/forward flow. As mentioned in Chapter 3, a centrifuge tube is used to
minimize the residue pressure and eliminate the backward/forward flow. However, as the
fabrication process goes on, this balance is broken. For the ID number case, overlap of
micro beads and the numbers makes the tag unable to read. For the barcode, however, as
long as the micro bead does not cover the whole bar, it is still readable, making it robust
in experiment condition.
4.3.2.2 2D Barcode
Besides ID/tracking number in the micro regime, another interesting topic is
fighting against counterfeit drugs. Currently, authentication is on the package, in forms of
printed marks, RF or optical tags [71]. However, these methods are susceptible to
imitation. Further, they guarantee only the genuineness of the package, rather than the
drug itself. Recently, microtaggants in drug formulation has been introduced as an ondose authentication (ODA) technology [72,73]. For one thing, ODA labels drug on the
drug formulation level, making it difficult for imitation; for the other, it guarantees the
authenticity of drug itself, rather than the package. Therefore, it is an appealing approach
for the pharmaceutical industry to fight against counterfeit drugs [74].
1 D barcode demonstrated above is a good candidate for ODA. However, it suffers
from several disadvantages. First of all, it encodes only numerical digits, limiting its
authentication ability. Secondly, as the name suggests, it works only in one dimension.
To form stable barcode tags, as shown in Fig 4.3, a certain width is needed. However, it
lowers the information density, making it inappropriate for ODA.
2D barcode, developed from its 1 D version, addresses the above issues. It takes
advantages of 2D space, representing more data in unit area. For numeric information,
one character occupies only 3 /3 pixels, compared to hundreds in 1 D barcode. Moreover,
it carries not only numeric, but also alphanumeric, byte/binary and even Japanese
characters (kanji) [75]. The most popular 2D barcode is Quick Response Code (QR code).
Sangkwon Han et al. have already implemented this type of code for anti-counterfeit drug
purposes [74]. However, because they lack the adaptive projection capability, they
fabricated the authentication QR code separately and mix it with drugs. Strictly speaking,
this is not on-dose authentication, as it is not synthesized at the drug formulation level. It
is prone to imitation just like the authentication on package. Here we demonstrate ondose authentication by precisely attach information-rich QR code with a micro bead (in
position of drugs). An open source QR code generator from the ZXing project [76] is
called in MATLAB to create QR code patterns for our experiments. For structural
stability, the black regions are adjusted to gray so the tag crosslinks as a whole.
Otherwise isolated parts in the QR code will float away. As shown in Fig 4.4(a), micro
beads are encapsulated with QR code containing tracking number, size and production
date. Fig 4.4(b) and (c) show machine readout of the encoded information with a smart
phone. In this case, the detected size is encoded in the QR code, being 66 and 203,
respectively. The significant difference indicates the accuracy of categorizing 10ptm and
20ptm beads. In addition to its high information density, QR code is intrinsically error
correctable. Using the Reed-Solomon error correction algorithm, corrupted QR code
could restore as much as 30% of codewords with the highest error correction level. This
56
could help QR codes survive in tough conditions, making it suitable for ODA. In our
demonstration experiments, level L error correction was used, which restores ~7%.
Figure 4.3 1D barcode encoded micro beads. 10pm beads are attached with simplified UPC barcodes for
tracking. The barcodes are of SLRE pattern. Inset: digital image of the beads right before adaptive
encoding. Scale bar: 50im.
C~ ScuWle
u
SanMWO
Scawrd Mw h
I &%
*s
C. SCWvWW
Samowm
Scanned Was fI~
ago
"W
4 U714
.Aa
.
(b ) "r
20712
Sha data
1!
Shme data
Figure 4.4 Conceptual demonstration of QR code ODA. (a) QR code ODAs attached to micro beads (in
position of drugs). Black regions of QR code is adjusted to gray so the structure crosslinks as a whole.
Otherwise isolated parts will float away. The error correction level is L, which means 7% of the codeword
could be restored when the QR code is corrupted. (b) and (c) show machine readout of the two QR codes in
(a). The three numbers indicate tracking number, size of the micro bead and production date. Scale bar:
50ptm.
4.3.3 Particle/Cell caging
In all the above demonstrations, the targets are encapsulated in hydrogel particles.
Although the hydrogel is porous, it physically limits activities of the target encapsulated,
especially when it is living protein or cell. One alternative way is to build a cage for the
target, a micro-sized counterpart of petri dish. The target has activities as usual without
restrictions, except being confined in a defined area.
However, our digital SFL could only synthesize 2D structures, making it difficult
to build a cage that confines target in all three dimensions. One alternative way is to
harness the ceiling and floor of microfluidic device. After identifying the positions of
targets, circular ring patterns are created which encloses the targets. Long exposure
57
makes sure the ring rises from floor to the ceiling, and cages the target as shown in Fig
4.5.
Figure 4.5 Caging targets inside a microfluidic channel. Ring patterns are created to confine targets inside
while allowing activities. Dashed white circle is representative of the shape and size of cage generated in
this case. Scale bar: 100ptm.
4.3.4 Particle sorting
As discussed before, during the image analysis, we could extract information of
targets like area, perimeter, eccentricity etc. This information helps to categorize different
targets, being either micro particles or cells. One more step would be to sort these
categorized particles based on SFL.
Existing sorting techniques include fluorescence-activated cell sorting (FACS)
[77,78] and magnetically-activated cell sorting (MACS) [79]. Both methods rely on
target labels, either in the form of fluorescence or magnetism. Label-free techniques exist
[80,81]. They are advantageous in their independence on pre-existing labels. However,
unlike FACS or MACS, where one device works for different particles as long as they are
labeled properly, most of these label-free sorting mechanisms are based on the geometry
of particles. Thus, one design of sorting device applies only to specific particles. In other
works, they are not robust to scaling.
SFL on the other hand, synthesizes particles of sizes and shapes at will. Even for
geometry based sorting mechanism, one device could work for different particles. As
shown in Fig 4.6, with a 3 inlets 2 outlets microfluidic channel, it is convenient to focus
the middle inlet to the connection point of outlets. If suspensions are only in the middle
inlet, an angle could direct itself upward or downward due to geometric restrictions. For
visualization purposes, a mix of l0pm and 20ptm beads is distributed on a glass slide.
Digital image of the original beads and beads encapsulated with sorting angles are shown
in Fig 4.7. Due to a lack of contrast, it is difficult to discern some 10pjm beads after
encapsulation. The inset shows projection pattern. With this setup, the same microfluidic
device works for any combination of two particles of different sizes as long as they are
discernable with CCD camera.
58
Figure 4.6 Sorting mechanism based on geometric restrictions. A 3 inlet 2 outlet microfluidic device is used.
Three streams flow from right to left and the middle stream is focus to the connection point of outlets. (a)
Upward and downward angles synthesized in the middle stream of the microfluidic channel; (b) and (c)
Due to geometric restrictions, the angles direct themselves into different outlets. Scale bar 100p m.
Figure 4.7 Categorizing and encapsulating particles with sorting patterns. (a) Digital image of intact 10pm
and 20pm beads. (b) The same beads encapsulated with sorting patterns. The angles direct themselves
towards designated outlets in microfluidic channel. Inset: projection pattern. Scale bar: I 00pm.
59
5. Solid-State Superionic Stamping
5.1 Introduction
All the previous chapters focus on high throughput fabrication of heterogeneous
structures made of functional polymers. There are also situations where metallic
micro/nano features are required for applications in nano-electronic, photonic,
nanoelectromechanical devices and nanoscale chemical sensors and transducers [82,83].
However, the current metallic micro-/nano-patterning approach is indirect: patterning on
photoresist using lithography or nanoimprinting [84,85,86] methods, followed by metal
evaporation and lift-off process [84]. This methodology involves multiple steps and is
highly sensitive to ambient conditions, making it time and money consuming. An
alternative method has been proposed to pattern metallic features directly, called
electrochemical micromachining (ECM) [87,88]. However, ECM suffers from limited
lateral extension of features due to diffusion of reacting materials. A solid-state
electrochemical imprinting method is introduced to address this issue [89,90,91,92]. A
solid electrolyte or superionic conductor stamp, which conducts ion usually at room and
low temperatures, is at the heart of this process. It is thus referred to as solid-state
superionic stamping (S4). A stamp made of a superionic conductor with a mobile cation
is patterned with inverse features through focused ion beam milling (FIB) or hot
embossing. The stamp is then brought into contact with the metal substrate. An electrical
bias is applied with the metal substrate as anode and an electrode at the back of the stamp
as cathode. At the contact points, electrochemical reaction takes place: at the anode side,
metal ions are oxidized and become mobile cation. These cations migrate through the
superionic conductor, recombine with electron and deposit at the cathode side. The stamp
thus dissolves excess metal at contact region, patterning the metal substrate progressively.
In this S4 process, the charge coupled mass transport occurs only at regions with physical
contact, making it ideal for high resolution patterning compared to ECM. Fig 5.1 shows
the schematic of S4 experiment.
There are challenges for these cases, where Ag 2 S, Cu 2 S, Agl-AgPO
3
are used as
the stamp materials. For Ag-S and Cu2 S, the stamp size cannot exceed a few hundred
microns as it is prepared by microtome cutting. The stamp surface is patterned by FIB or
mechanical imprinting, which restrict the stamp size to ~one hundred microns and a few
millimeters, respectively. Besides they are brittle and high cost. These issues leave them
unsuitable for large scale manufacturing. The formation of AgI-AgPO 3 stamps, on the
other hand, involves a relatively large temperature change, causing residue stress and
deformation. A new form of material is desired, which is easy to mold, low cost, provides
high fidelity and works under room temperature.
Solid-state electrolytes based on polymers and polymer blends with copper(II)
salts are investigated recently [93,94]. High plasticizer content makes these polymer gel
good ionic conductors at ambient temperature combined with good mechanical properties
[95]. These room temperature superionic polymers offer potential for S4 process. It's
convenient to pattern large-area polymer stamps either by directly casting the polymer on
a hard master mold or embossing process. Polymer stamps generally provide better
conformal contact and could reach submicron resolution. Two separate efforts are made
for copper and aluminum, respectively, and will be discussed in detail.
60
Figure 5.1 Schematic of S4 process. (a) The initial state. The stamp and metal to be patterned are connected
through electrical bias, with the stamp as cathode and metal as anode. (b) When the stamp is in physical
contact with metal surface, electrochemical reaction happens at the contact area and the anode metal is
dissolves progressively. (c) S4 gives complementary pattern to the pattern on the stamp. Figure reprinted
with permission from ref. [89], c 2007 ACS.
5.2 Copper
5.2.1 Materials
Solid-state polymer electrolytes are designed based on the concept of trapping
solutions in polymer cages. Several such electrolyte incorporating
electrolyte
liquid
lithium salt solutions are investigated by plasticizing poly(acrylonitrile) (PAN) or
poly(methylmethacrylate) (PMMA) [96,97]. Copper ion conducting PAN-based polymer
electrolyte is then studied as candidate anode material for solid-state batteries [93,95].
Polyacrylonitrile (PAN)-dimethyl sulfoxide (DMSO)-dimethylformamide (DMF)copper(II) trifluoromethanesulfonate (CuTf2) is chosen as our stamp material. DMSO and
DMF serve as plasticizers to prevent PAN from forming crystalline-like structures,
leaving room for ion transport. In addition, excess DMF acts as solvent for casting and
stamp formation purposes.
Figure 5.2 Solution for preparing the superionic polymer stamp. (a) A mix of 3.3g PAN, 8g DMSO and
lOg DMF. (b) 2.37g CuTf2 dissolves in IOg DMF. (c) A mix of (a) and (b), which is finally cast into S4
stamp.
61
When preparing the polymer solution, 3.3g PAN (Sigma Aldrich), 8g DMSO
(Thermo Scientific) and lOg DMF (Sigma Aldrich) are fully mixed by stirring with a
magnetic stirrer on a hot plate set to 50'C as shown in Fig 5.2(a). Another 2.37g CuTf2
and lOg DMF are mixed to obtain blue transparent solution shown in Fig 5.2(b). All
beakers are sealed to avoid water vapor. After both solutions are fully mixed, they are
mixed on the hot plate. After -1h, the mix becomes transparent and green as shown in
Fig 5.2(c).
5.2.2 Stamp Formation
A hard master mold is prepared by electron beam lithography on a silicon wafer.
The pattern is replicated by PDMS in the same way as replicating microfluidic channels.
The PDMS mold is then placed on a silicon wafer to avoid deformation. A 500 C polymer
solution prepared in the previous step is poured onto the PDMS mold and spin coated at
-350rpm for 60s. Uniform polymer film is achieved with a speed between 300rpm460rpm. The polymer film (S4 stamp) is then stored into an electronic desiccator cabinet
for 2 days (10%RH) and transferred to a substrate. Cu patterns of our interest are 1Im
L/S and pyramid pattern array. These features are of micrometer size and could be
checked qualitatively under optical microscope as shown in Fig 5.3.
Figure 5.3 S4 stamp for copper. (a) 6 inch polymer stamp with 400nm L/S pattern. The color comes from
light diffraction from the grating pattern. (b) Digital image of the L/S stamp taken under optical
microscope. Uniform pattern over a large area is transferred onto the stamp. (c) Polymer stamp with
pyramid array pattern transferred onto a glass slide. The stamp shows no crease when transferred properly.
(d) Pyramid pattern array observed under optical microscope. All pyramids from the hard master mold are
successfully transferred to the stamp (I inchx 1 inch). The square bottom of pyramid can be seen from the
figure.
62
5.2.3 Results and discussion
Cu samples are prepared by e-beam evaporating a 10nm Cr adhesive layer on
glass slide at lA/s and Cu at 2A/s. The S4 stamp on glass slide is connected by a
conductive copper tape on the side to the cathode and fixed on a tilting stage to the whole
setup, as shown in Fig 5.4. The stamp is brought down close to the Cu substrate by a
manual translational stage. The tilting angle of the stamp is calibrated to ensure
parallelism between the stamp and Cu substrate. The stamp is then brought down by a
motorized translational stage with an accuracy of 1plm. A load cell measures the force
applied from the top of the tilting stage. Once contact is established, a non-zero
compression force is displayed. The stage moves further down until ~1OkPa is applied
between the stamp and the Cu substrate. The whole setup then sits still for 10min for the
stress in stamp to relax. Electrical bias is then applied and electrochemical etching starts.
Quantitative results are obtained by coating a thin layer of gold on top of the etched
copper substrate and observe under scanning electron microscope (SEM) as shown in Fig
5.5. The line edge roughness for imprinted lpm L/S structure is ~30nm. By transferring
the pattern from a hard silicon master mold, to PDMS mold, to the S4 stamp and finally
to Cu structures, the ratio between peak and valley change from 1:1 to -9:7. For the
pyramid pattern, -0.6 inch 2 is finally transferred to Cu as shown in the top inset of Fig
5.5(b). Aspect ratio is roughly halved due to deformation of the polymer stamp.
Figure 5.4 (a) S4 setup. A manual stage is installed on a motorized translational stage. Screwed to the
manual stage is a load cell connected with a tilting stage and the stamp holder. (b) Close-up view of the
experimental setup. The bias is applied through a copper conductive tape.
63
Figure 5.5 SEM images of (a) I pm L/S pattern and (b) pyramid pattern. (a) The L/S pattern is successfully
transferred over most area. There are small fraction of area where missing/connected lines indicated
over/under etching. The line edge roughness is ~30nm. The ratio between peak and valley change from 1:1
of the hard master mold to -9:7 of etched Cu. (b) Top view of the pyramid array pattern under SEM. The
square bottom of pyramid is clearly seen. Top inset: Pyramid arrays etched on Cu sample. The spacing
between adjacent pyramid arrays indicate successful transfer of pattern over large area. The patterned
region is ~0.6 inch 2. Bottom inset: close-up view of the 3D pyramid structure. The square bottom is
precisely transferred with a side length of -2pm. However, the aspect ratio is significantly smaller than
hard master mold due to deformation of the stamp. Scale bars: a, 10pm; b, 20pm, bottom inset: 2pm.
Compared with previous results using Ag 2 S, Cu2 S, AgI-AgPO 3, this polymer
stamp does not provide as high resolution. However, it enables unprecedentedly large
stamping area under room temperature because the thin film stamp provides conformal
contact when applied properly. The hard master mold can be replicated with little damage
and the formation of stamp is convenient. It also provides opportunities in non-planar
fabrication due to the flexibility of the thin film stamp.
64
Figure 5.6 During the imprinting process, the thin film stamp is a soft interface sandwiched by two hard
substrate, causing wrinkling. (a) Optical microscope and (b) SEM images of a I pm L/S sample showing
wrinkling pattern. Scale bar: 20pm.
However, compared to the stamp holder and Cu substrate, the S4 stamp has
largest compliance. Thus, from the viewpoint of solid mechanics, the imprinting process
is actually a soft thin film sandwiched between two hard substrates, where wrinkles occur
[98] as shown in Fig 5.6. Possibilities for other deformation exist depending on the
pressure and geometry of the stamp [99]. The search for softer stamp holder and stiffer
stamping material could be the next step of this research.
On the basis of the S4 process, a preliminary attempt on selective deposition is
investigated. By simply reversing the electrical bias, Cu from the cathode transfers
through the stamp and deposits on the substrate. Fig 5.7 shows the schematic and
preliminary result of this concept of selective deposition. For the L/S pattern, long-range
ordering is observed despite randomly scattered deposits. For the pyramid pattern, a
deposited array is clearly seen under the optical microscope.
65
layer of metal to be deposited is attached to the ionic stamp as anode. The substrate is in contact with the
patterned surface of stamp and connected to the cathode. When an electrical bias applies, the anode metal is
dissolved and the ions transport through the stamp and deposit
onto the contact regions of substrate. When
the deposition is finished, the stamp is removed from the substrate and the selectively deposited metals are
left on the substrate, forming the same pattern as that on the stamp (compared to the reverse pattern in the
S4 case). (b) and (c) show preliminary result by selectively depositing Cu (b) 1Im L/S and (c) pyramid
array pattern onto a Cu substrate. In (b), long range ordering is observed and in (c), the array is clearly
seen.
5.3 Aluminum
Unlike silver or copper which belongs to inert metals, aluminum (Al) is active and
there is no Al ion in aqueous solution. It is possible to transport Al in its ionic form, via
ionic liquids [100,101,102]. However, ionic liquid is generally unstable in ambient
conditions and is extremely expensive for industry. An alternative way for Al is selective
anodization. Anodic aluminum oxide (AAO) has been widely investigated and used for
over 100 years [103]. However, to our knowledge, no one has demonstrated
nanostructure patterning with AAO.
5.3.1 Materials
The electrochemical reaction involved in our experiment is
2A1+3H 2 0=A1 203+3H 2
(5.1)
To control the contact region, a sulfonated tetrafluoroethylene based
fluoropolymer-copolymer called Nafion is used as the stamp material. Nafion has good
thermal and mechanical stability thanks to its Teflon backbone. It shows ionic
conductivity, especially for proton and receives attention for application in fuel cells. To
incorporate the electrochemical reaction with Nafion's ionic property, a weak acid
condition is provided. Fig 5.8 conceptually shows the selective anodization process.
66
(a)
Figure 5.8 (a) Schematic of the selective anodization. A Nafion film exchanges proton and promotes
growth of AAO at contact areas. (b) SEM image of AAOs formed by this electrochemical process. Scale
bar: 500nm.
5.3.2 Stamp formation
Hot embossing is chosen for patterning Nafion 117 (Sigma Aldrich) as the stamp.
The procedure follows [104]. A piece of Nafion 117 film is first cleaned in DI water,
placed on a silicon wafer and dried in N2 stream to remove observable water droplets.
The unpatterned Nafion is then placed on top of the master mold and inserted to the hotembossing machine. A pressure of -650 psi is applied to prevent the film from buckling
due to the loss of moisture when heated to 150C. After -5min, heating is turned off
while the pressure remains, and the temperature slowly drops due to heat exchange with
the environment. After -30min, the load is removed and the patterned stamp separated
from the master mold. Examples of hot embossed Nafion stamps are shown in Fig 5.9.
(b)
(a)
Figure 5.9 SEM images of (a) 100nm and (b) 65nm L/S Nafion stamp formed by hot embossing under
-650 psi at 150'C. Scale bars: a, 2pm; b, 2ptm.
5.3.3 Results and discussion
Following the schematic shown in Fig 5.8(a), selective anodization of aluminum
is achieved with Nafion stamps. Fig 5.10 shows SEM images of 150nm and 100nm L/S
pattern on Al. Inset is the close-up view of the nanostructure. From the inset in Fig
5.10(b), it is clear that the pattern is formed by formation of AAO nano pores. However,
as in the 150nm sample, over-anodization occurs. It is thus vital to control the
67
experimental condition. Until this stage, the end of patterning process is determined by
the color change of Al substrate, which is not accurate enough. Another future step would
be to minimize the pore size. From Fig 5.8(b), the pore sizes are not uniform in our case.
Large pores impose issue to high-resolution fabrication. In our case, the finest half pitch
size achieved is 1 00nm. No L/S pattern is observed from 65nm Nafion stamp.
Figure 5.10 SEM images of selectively anodized aluminum with (a) 150nm and (b) 100nm L/S pattern.
Inset shows close-up views of the nanostructures. As can be told from the inset of (a), over-anodization
occurs The inset of (b) shows appropriate anodization condition, where AAO nanopores form the desired
structure. Scale bars: a, 5tm, inset: 500nm; b, 2prm, inset: 500nm.
68
6. Summary and Outlook
6.1 Summary
We built a digital Stop-Flow Lithography (SFL) system by incorporating a
standard SFL setup with a spatial light modulator (SLM) as dynamic mask, aiming at
expanding capabilities of fabrication geometrically and chemically anisotropic colloidal
particles for self-assembly.
Firstly, the basics of the SFL fabrication platform are introduced, ranging from
hardware like the experimental setup and microfluidic channel to the control program
coded in LabVIEW. Optical characterization shows both potential in grayscale control
and limits in resolution due to point spread function.
Secondly, the crucial function is presented, being fabricating geometrically and
chemically anisotropic colloidal particles at high throughput. A wide spectrum of
materials is compatible with the fabrication process, including biocompatible material,
neutral hydrogel, shape memory polymer and temperature and pH sensitive hydrogel. A
discussion is presented on the limiting factors of throughput, showing potential of one
order improvement.
Thirdly, the unique capability of adaptive projection is demonstrated. Our system
detects suspensions and encapsulates them with prescribed patterns. On the one hand, this
enlarges material selection pool, because lots of material of the powder form could be
incorporated by this technique; on the other hand, the adaptive encapsulation on the
single cell/particle level is useful for biological, biomedical and pharmaceutical purposes.
Provided with the capabilities above, large quantities of functional geometrically
and chemically anisotropic colloid particles could be synthesized in short time,
facilitating research on self-assembly of colloids.
To add to the fabrication of colloid particles, which are mostly gels, we
investigate solid-state superionic stamping (S4), which is a direct metal patterning
technique. It features one-step, large fabrication area, low cost and working in ambient
conditions. Direct patterning on copper and aluminum is demonstrated as examples.
6.2 Outlook
Future work could be branched into two directions: further improvement of the
SFL fabrication platform and application to self-assembly. Potential pathways are listed
below
6.2.1 Characterization of micro-gels
Most of the work demonstrated here is proof-of-concept. An important follow-up
is to characterize the behavior of these synthesized microgels. For example, we will study
the swelling behavior of PEGDA, or the mechanical response of SMP to change of
temperature in the micro regime.
69
6.2.2 Increase of throughput
As discussed in 3.2.4, the throughput could be further increased by one order.
Redesign of channel geometry makes a major contribution, with shorter length, larger
height and optimized width trading off between longer stop time and more particles
fabricated in one exposure.
Switching from our home-made projection module to off-shelf projection module
gives a ten times enhancement in optical power, shortening the exposure time, thus
contributing to the throughput.
We expect ~107 particles/h after all these improvements and optimization.
6.2.3 Finer resolution
In the sub-micron regime, many short range interactions play an important role,
one of which being hydrophobic attraction working in 10-100nm [40]. These forces
provide extra pathways for designing and programming self-assemblies. To harness short
range forces, building blocks cannot be too much larger than the interaction range.
However, current resolution of our SFL is ~1Am depending on the choice of objective,
inaccessible to these interactions. In this case, we are restricted by the finite aperture of
the system. One possible solution is to do a deconvolution on the desired pattern with
measured point spread function and use the result as projection pattern. This in principle
pushes the resolution towards diffraction limit.
6.2.4 Flow in three dimensions
Until now, we are harnessing laminar flow in two dimensions (lateral directions).
As a result, the chemical anisotropy is restricted to stripes in 2D. Hydrodynamic focusing
lithography, on the other hand, synthesizes particles with chemical anisotropy in the
vertical direction (channel height direction) by stacking flows in a two-layer microfluidic
channel [105]. This technique is compatible with our setup, differing in channel design. It
is possible to redesign our channel to incorporate laminar flows not only in the lateral
direction but also in the vertical direction. In this way, the chemical anisotropy would
have 3D pattern, adding to the function of a building block for self-assembly or on-dose
authentication.
6.2.5 Additive manufacturing
As mentioned before, the adaptive encapsulation capability enlarges our material
selection pool. Polystyrene or metal oxides, which could not be synthesized directly
through SFL, can be loaded inside a colloidal shell, empowering building blocks. These
colloidal shells could also act as micro carriers and self-assemble loaded with materials
of interest. A sintering process could remove the shell and leave the cores forming
exactly the same structure [43]. Moreover, the processed encapsulations can be collected
70
and sent to a larger channel for a second encapsulation with different material or
modification of the surface, very similar to additive manufacturing.
6.2.6 3D printing inside the microfluidic channel
The lithographic apparatus of our SFL platform is very similar to those of PpiSL,
except for the lack of a translational stage. However, a translational stage is meaningless
if we could fabricate one thin layer directly. It has been demonstrated through tight
focusing with high magnification objective [106]. The shallow focal depth leaves resin
out of focus uncured, solidifying a thin layer at positions (in the height direction) at will.
Incorporating this and our adaptive technique, it is possible to reproduce PpSL inside the
microfluid channel by detecting the unfinished 3D structure, adjusting focal plane and
printing one layer on top of the structure.
6.2.7 Self-assembly
The ultimate goal of our SFL is to generate powerful building blocks to realize
designs of self-assembly. The first step could be study of assembly by Janus particles
hydrophobic and hydrophilic on two sides. Candidates are PEGDA and HDDA or
PNIPAAm and HDDA. After the assembly is achieved, functions like tuning lattice
constant can be investigated by swelling of hydrogel, which composes one end of the
building block. One possible application is tunable acoustic metamaterial.
71
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
D. H. Everett, Basic Principles of Colloid Science, Royal Society of Chemistry,
London, 1988. B
Russel, W.B., Brownian Motion of Small Particles Suspended in Liquids. Ann.
Rev. Fluid Mech., 13, 425-455 (1981)
F. Caruso et al, Colloids and Colloid Assemblies: Synthesis, Modification,
Organization and Utilization of Colloid Particles, Wiley, 2004
Thomas Graham, "On Liquid Transpiration in Relation to Chemical
Composition", Phil. Trans. R. Soc. Lond., 151, 373-386 (1861)
Rnst A. Hauser, "The history of colloid science: In memory of Wolfgang
Ostwald", J. Chem. Educ., 32, 2 (1955)
Robert Shepherd (2010). Microfluidic Assembly and Packing Dynamics of
Colloidal Granules, PhD thesis, UIUC
Sharon C. Glotzer, "Some Assembly Required", Science, 306, 419 (2004)
Steve Granick, Shan Jiang and Qian Chen, "Janus Particles", Physics Today, 62,
68 (2009)
Alonzo D. Cook et al., "Characterization and development of RGD-peptidemodified poly(lactic acid-colysine) as an interactive, resorbable biomaterial",
Journal of Biomedical Materials Research, 35, 513-523 (1997)
Ben Alsberg et al., "Engineering growing tissues", Proc. Natl. Acad. Sci., 99,
12025 (2002)
Sven Halstenberg et al., "Biologically Engineered Protein-graft-Poly(ethylene
glycol) Hydrogels: A Cell Adhesive and Plasmin-Degradable Biosynthetic
Material for Tissue Repair", Biomacromolecules, 3, 710-723 (2002)
Priyadarshi Panda et al., "Stop-flow lithography to generate cell-laden microgel
particles", Lab Chip, 8, 1056-1061 (2008)
Lily Nari Kim et al., "Single exposure fabrication and manipulation of 3D
hydrogel cell microcarriers", Lab Chip, 11, 48 (2011)
David C. Appleyard et al., "Bar-coded hydrogel microparticles for protein
detection: synthesis, assay and scanning", Nature Protocols, 6, 1761 (2011)
Liang Feng et al, "Xenopus Laevis as a Novel Model to Study Long Bone
Critical-Size Defect Repair by Growth Factor-Mediated Regeneration", Tissue
Engineering A, 17, 691 (2011)
Robert Langer and David A. Tirrell, "Designing materials for biology and
medicine", Nature, 428, 487 (2004)
Andreas Lendlein and Steffen Kelch, "Shape-Memory Polymers", Angew. Chem.
Int. Ed., 41, 2034-2057 (2002)
Vikas Srivastava, Shawn A. Chester, Lallit Anand, "Thermally actuated shapememory polymers: Experiments, theory and numerical simulations", Journal of
the Mechanics and Physics of Solids, 58, 1100-1124 (2010)
H. M. Wache, D. J. Tartakowska, A. Hentrich, M. H. Wagner, "Development of a
polymer stent with shape-memory effect as a drug delivery system", Journal of
Materials Science: Materials in Medicine, 14, 109-112 (2003)
72
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
C. M. Yakacki, R. Shandas, C. Lanning, B. Rech, A. Eckstein, K.Gall,
"Unconstrained recovery characterization of shape-memory polymer networks for
cardiovascular applications", Biomaterials, 28, 2255-2263 (2007)
G. M. Baer et al., "Fabrication and in vitro deployment of a laser-activated shapememory vascular stent", Biomedical Engineering Online, 6, 43 (2007)
Andreas Lendlein and Robert Langer, "Biodegradable, Elastic Shape-Memory
Polymerfor Potential Biomedical Applications", Science, 296, 1673 (2002)
S. H. Park, D. Qin and Y. Xia, "Crystallization of mesoscale particles over large
areas", Advanced Materials, 10, 1028-1031 (1998)
S. H. Park and Y. Xia, "Assembly of mesoscale particles over large areas and its
application in fabricating tunable optical filters", Langmuir, 15, 266-273 (1999)
B. Gates, D. Qin and Y. Xia, "Assembly of nanoparticles into opaline structures
over large areas", 11, 466-469 (1999)
Jie Sun et al., "Fabrication of Centimeter-Sized Single-Domain Two-Dimensional
Colloidal Crystals in a Wedge-Shaped Cell under Capillary Forces", Langmuir,
26, 7859 (2010)
Victor Canalejas-Tejero et al., "Qualitative and Quantitative Analysis of
Crystallographic Defects Present in 2D Colloidal Sphere Arrays", Langmuir, 28,
161-167 (2012)
N. Boechler et al., "Interaction of a Contact Resonance of Microspheres with
Surface Acoustic Waves", Physical Review Letters, 112, 068103 (2013)
Yu Lu, Yadong Yin and Younan Xia, "Three-Dimensional Photonic Crystals with
Non-spherical Colloids as Building Blocks", Advanced Materials, 13, 415 (2001)
Egon Matijevic, Paul Scheiner, "Ferric hydrous oxide sols: III. Preparation of
uniform particles by hydrolysis of Fe(III)-chloride, -nitrate, and -perchlorate
solutions", J. Colloid Interface Sci., 63, 509-524 (1978)
Yadong Yin, Yu Lu, Byron Gates and Younan Xia, "Template-Assisted SelfAssembly: A Practical Route to Complex Aggregates of Monodispersed Colloids
with Well-Defined Sizes, Shapes and Structures", J. Am. Chem. Soc., 123, 87188729 (2001)
Gi-Ra Yi et al., "Generation of uniform photonic balls by template-assisted
colloidal crystallization", Synthetic Metals, 139, 803-806 (2003)
Jacinta C. Conrad et al., "esigning colloidal suspensions for directed materials
assembly", Current Opinion in Colloid & Interface Science, 16, 71-79 (2011)
Robert F. Shepherd et al., "Stop-Flow Lithography of Colloidal, Glass, and
Silicon Microcomponents", Advanced Materials, 20, 4734 (2008)
William E. Uspal and Patrick S. Doyle, "Self-organizing Microfluidic Crystals",
Soft Matter, 2014, DOI: 10.1039/C4SM00664J
William E. Uspal, H. Burak Eral & Patrick S. Doyle, "Engineering particle
trajectories in microfluidic flows using particle shape", Nature Communications,
4, 2666 (2013)
Stefano Sacanna et al., "Shaping colloids for self-assembly", Nature
Communications, 4, 1688 (2013)
S. C. Glotzer, M. J. Solomon and N. A. Kotov, "Self-Assembly: From Nanoscale
to Microscale Colloids", AIChE Journal, 50, 2978 (2004)
73
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
Dhananjay Dendukuri, T. Alan Hatton and Patrick S. Doyle, "Synthesis and SelfAssembly of Amphiphilic Polymeric Microparticles", Langmuir, 23, 4669-4674
(2007)
Qian Chen, Sung Chul Bae & Steve Granick, "Directed self-assembly of a
colloidal kagome lattice", Nature, 469, 381 (2011)
Ahmet F. Demirors et al., "Colloidal assembly directed by virtual magnetic
moulds", Nature, 503, 99 (2013)
C. Sun, N. Fang, D.M. Wu, X. Zhang, "Projection micro-stereolithography using
digital micro-mirror dynamic mask", Sensors and Actuators A, 121, 113-120
(2005)
Xiaoyu Zheng et al., "Ultralight, ultrastiff mechanical metamaterials", Science,
344, 1373 (2014)
Dhananjay Dendukuri, Shelley S. Gu, Daniel C. Pregibon, T. Alan Hatton and
Patrick S. Doyle, "Stop-flow lithography in a microfluidic device", Lab on a
Chip, 7, 818-828 (2007)
http://www.ti.com/lit/an/dlpa022/dlpa022.pdf
http://www.innovationsinoptics.com/products/datasheets/IOI_DataSheetUV260
ON-700.pdf
http://www.olympusamerica.com/files/seg_bio/seg_researchix71 -ix8 Ibro.pdf
Mark A. Eddings, Michael A. Johnson and Bruce K. Gale, "Determining the
optimal PDMS-PDMS bonding technique for microfluidic devices", J.
Micromech. Microeng., 18, 067001 (2008)
Christian Decker, Aubrey D. Jenkins, "Kinetic approach of oxygen inhibition in
ultraviolet- and laser-induced polymerizations", Macromolecules, 18, 1241-1244
(1985)
Howon Lee (2011). Three-Dimensional Micro Fabrication of Active Micro
Devices Using Soft Functional Materials, PhD thesis, UIUC
Urszula Kolczak et al., "Reaction Mechanism of Monoacryl- and
Bisacylphosphine Oxide Photoinitiators Studied by 3 1P-, "C-, and 'H-CIDNP and
ESR", JACS, 118, 6477-6489 (1996)
Chunguang Xia, Howon Lee and Nicholas X. Fang, "Solvent-driven polymeric
micro beam device", J. Micromech. Microeng., 20, 085030 (2010)
Howon Lee, Chunguang Xia and Nicholas X. Fang, "First jump of microgel:
actuation speed enhancement by elastic instability", Soft Matter, 6, 4342-4345
(2010)
Dhananjay Dendukuri et al., "Continuous-flow lithography for high-throughput
microparticle synthesis", Nature Materials, 5, 365-369 (2006)
T. Gervais et al., "Flow-induced deformation of shallow microfluidic channels",
Lab on a Chip, 6, 500-507 (2006)
T. Gervais (2006). Mass transfer and structural analysis of microfluidic sensors,
PhD thesis, Massachusetts Institute of Technology
http://www.fao.org/ag/agn/jecfa-additives/specs/Monographl/Additive-316.pdf
L. Anand, N.M. Ames, V. Srivastava, S.A. Chester, "A thermo-mechanicallycoupled theory for large deformations of amorphous polymers. Part I:
formulation", International Journal of Plasticity, 25, 1474-1494 (2009)
74
59.
60.
61.
62.
63.
64.
65.
66.
67.
68.
69.
70.
71.
72.
73.
74.
75.
76.
77.
78.
79.
N.M. Ames, V. Srivastava, S.A. Chester, L. Anand, "A thermo-mechanicallycoupled theory for large deformations of amorphous polymers. Part II:
applications", International Journal of Plasticity, 25, 1495-1539 (2009)
V. Srivastava, S.A. Chester, N.M. Ames, L. Anand, "A thermo-mechanicallycoupled large-deformation theory for amorphous polymers in a temperature range
which spans their glass transition", International Journal of Plasticity, 26, 11381182 (2010)
http://www.sigmaaldrich.com/materials-science/polymer-science/nipampolymers.html
Pradeep Pareek et al., "Tuning the Swelling Behavior of Chemisorbed Thin
PNIPAAm Hydrogel Layers by N,N-Dimethyl Acrylamide Content", Progress in
Colloid and Polymer Science, 132, 145-151 (2006)
Howon Lee et al., "Colour-barcoded magnetic microparticles for multiplexed
bioassays", Nature Materials, 9, 745-749 (2010)
R.J. Fulton et al., "Advanced multiplexed analysis with the FlowMetrixM
system", Clin. Chem., 43, 1749-1756 (1997)
M. Han et aL., "Quantum-dot-tagged microbeads for multiplexed optical coding of
biomolecules", Nature Biotechnology, 19, 631-635 (2001)
I.L. Medintz et al., "Quantum dot bioconjugates for imaging, labeling and
sensing", Nature Materials, 4, 435-446 (2005)
K. Braechmans, et al., "Encoding microcarriers by spatial selective
photobleaching", Nature Materials, 2, 169-173 (2003)
D.C. Pregibon et al., "Multifunctional encoded particles for high-throughput
biomolecule analysis", Science, 315, 1393-1396 (2007)
Jiseok Lee et aL., "Universal process-inert encoding architecture for polymer
microparticles", Nature Materials, 13, 524-529 (2014)
http://en.wikipedia.org/wiki/UniversalProductCode
R.Y. Shah, P.N. Prajapati and Y.K. Agrawal, "Anticounterfeit packaging
technologies", J. Adv. Pharm. Technol. Res., 1, 368-373 (2010)
R. Cockburn et al., "The Global Threat of Counterfeit Drugs: Why Industry and
Governments Must Communicate the Dangers", PLoS Med, 2, e100 (2005)
P.M.O. Wong, "Anti-counterfeiting", Tablets Capsules, 9, 28 (2011)
S. Han et aL., "Lithographically Encoded Polymer Microtaggant Using HighCapacity and Error-Correctable QR Code for Anti-Counterfeiting of Drugs",
Advanced Materials, 24, 5924-5929 (2012)
http://en.wikipedia.org/wiki/QR code
https://github.com/zxing/zxing/
L.A. Herzenberg et aL., "The History and Future of the Fluorescence Activated
Cell Sorter and Flow Cytometry: A View from Stanford", Clinical Chemistry, 48,
1819-1827 (2002)
Sherrif F Ibrahim and Ger van den Engh, "High-speed cell sorting: fundamentals
and recent advances", Current Opinion in Biotechnology, 14, 5-12 (2003)
J.D. Adams and H. Tom Soh, "Perspectives on utilizing unique features of
microfluidics technology for particle and cell sorting", JALA Charlottesv Va., 14,
331-340 (2009)
75
80.
81.
82.
83.
84.
85.
86.
87.
88.
89.
90.
91.
92.
93.
94.
95.
96.
97.
98.
L.R. Huang, E.C. Cox, R.H. Austin, J.C. Sturm, "Continuous Particle Separation
Through Deterministic Lateral Displacement", Science, 304, 987-990 (2004)
M. Yamada, M. Nakashima and M. Seki, "Pinched Flow Fractionation:
Continuous Size Separation of Particles Utilizing a Laminar Flow Profile in a
Pinched Microchannel", Anal. Chem., 76, 5465-5471 (2004)
M. Madou, Fundamentals of Microfabrication, 2 "d ed., CRC Press, New York,
2002
W.S. Ruska, Microelectronic Processing, McGraw-Hill, New York, 1987
S.Y. Chou, P.R. Krauss and P.J. Renstrom, "Nanoimprint lithography", J. Vac.
Sci. Technol. B, 14, 4129-4133 (1996)
S.Y. Chou, P.R. Krauss, W. Zhang, L. Guo and L. Zhuang, "Sub-10 nm imprint
lithography and applications", J. Vac. Sci. Technol. B, 15, 2897-2904 (1997)
S. Zankovych et al., "Nanoimprint lithography: challenges and prospects",
Nanotechnology, 12, 91-95 (2001)
A.L. Trimmer, J.L. Hudson, M. Kock, R. Shuster, "Single-step electrochemical
machining of complex nanostructures with ultrashort voltage pulses", Applied
Physics Letters, 82, 3327-3329 (2003)
B. Bhattacharyya, B. Doloi, P.S. Sridhar, "Electrochemical micro-machining: new
possibilities for micro-manufacturing", Journal of Materials Processing
Technology, 113, 301-305 (2001)
K.H. Hsu, P.L. Schultz, P.M. Ferreira and N.X. Fang, "Electrochemical
Nanoimprinting with Solid-State Superionic Stamps", Nano Letters, 7, 446-451
(2007)
A. Kumar, K.H. Hsu, K.E. Jacobs, P.M. Ferreira and N.X. Fang, "Direct metal
nano-imprinting using an embossed solid electrolyte stamp", Nanotechnology, 22,
155302 (2011)
P.L. Schultz, K.H. Hsu, N.X. Fang, P.M. Ferreira, "Solid-state electrochemical
nanoimprinting of copper", J. Vac. Sci. Technol. B, 25, 2419-2424 (2007)
K.E. Jacobs, K.H. Hsu, X. Han, A. Kumar, B.P. Azeredo, N.X. Fang and P.M.
Ferreira, "Solid-state superionic stamping with silver iodide-silver metaphosphate
glass", Nanotechnology, 22, 425301 (2011)
K. Perera, M.A.K.L. Dissanayake, P.W.S.K. Bandaranayake, "Copper-ion
conducting solid-polymer electrolytes based on polyacrylonitrile (PAN)",
Electrochimica Acta, 45, 1361-1369 (2000)
J. Tang, X. Chu, X. Gao and W. Zhao, "Electrochemical Synthesis of
Polyacrylonitrile (PAN)-Copper Composite Conductive Film", Journal of Applied
Polymer Science, 61, 1773-1779 (1996)
A. Lewandowski, I. Stepniak, W. Grzybkowski, "Copper transport properties in
polymer electrolytes based on poly(ethylene oxide) and poly(acrylonitrile)", Solid
State Ionics, 143, 425-432 (2001)
B. Scrosati, Applications of Electroactive Polymers, Chapman and Hall, London,
1993
B.V.R. Chowdari et al., Solid State Ionic Materials, World Scientific, Singapore,
1994
J. Yin et al., "Deterministic Order in Surface Micro-Topologies through
Sequential Wrinkling", Advanced Materials, 24, 5441-5446 (2012)
76
99.
100.
101.
102.
103.
104.
105.
106.
C.Y. Hui, A. Jagota, Y.Y. Lin, E.J. Kramer, "Constraints on Microcontact
Printing Imposed by Stamp Deformation", Langmuir, 18, 1394-1407 (2002)
T. Jiang et al., "Electrodeposition of aluminum from ionic liquids: Part Ielectrodeposition and surface morphology of aluminum from aluminum chloride
(AiC 3)-1-ethyl-3-methylimidazolium chloride ([EMIm]Cl) ionic liquids", Surface
& Coatings Technology, 201, 1-9 (2006)
T. Jiang et al., "Electrodeposition of aluminum from ionic liquids: Part 11-studies
on the electrodeposition of aluminum from aluminum chloride (AlCl3 )trimethylphenylammonium chloride (TMPAC) ionic liquids, Surface & Coatings
Technology, 201, 10-18 (2006)
T. Jiang et al., "Studies on the ALC13/dimethylsulfone (DMSO 2) electrolytes for
the aluminum deposition processes", Surface & Coatings Technology, 201, 63096317 (2007)
W. Lee et al., "Fast fabrication of long-range ordered porous alumina membranes
by hard anodization", Nature Materials, 5, 741-747 (2006)
A.D. Taylor, B.D Lucas, L.J. Guo, L.T. Thompson, "Nanoimprinted electrodes
for micro-fuel cell applications", Journal of Power Sources, 171, 218-223 (2007)
K.W. Bong et al., "Hydrodynamic Focusing Lithography", Angew. Chem. Int.
Ed., 49, 87-90 (2010)
W. Park, S. Han, H. Lee, S. Kwon, "Free-floating amphiphilic picoliter droplet
carriers for multiplexed liquid loading in a microfluidic channel", Microfluid
Nanofluid, 13, 511-518 (2012)
77