|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
PROCESS CHARACTERIZATION OF FABRICATING PLASTIC MICROFLUIDIC
A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF SCIENCE
UNIVERSITY OF FLORIDA
First off, thanks go to my family for their constant words of support and occasional
infusions of the comforts of home.
For the device optimization work presented in Chapter 2, I appreciate the efforts of
Zheng Xia for fabricating the silicon and glass masters and arranging to get the SEM
pictures from the Department of Materials Science and Engineering's Major Analytical
Instrumentation Center, undergraduates Fernando Tavares and Ryan Ferguson for helping
press the hundreds of devices that were used for all facets of this research, Champak Das
for his work on the on-chip protein separation process, a portion of which is summarized
here, and John Klingman of Ticona for providing the resin and film.
For the toxin-identification device work presented in Chapter 3, I thank Qian Mei
for running the tests for adhesive biocompatibility and protein expression and helping
enhance my understanding of the processes.
This work was supported in part by the startup fund from the University of Florida,
the National Science Foundation, and the National Aeronautics and Space Administration
(NASA) via the UF Space Biotechnology and Commercial Applications Program and the
UCF-UF Space Research Initiative.
I appreciate Dr. Roger Tran-Son-Tay and Dr. Gregory Sawyer for taking the time
to serve on my committee. Finally, and most importantly, I thank Dr. Hugh Fan for
chairing the committee and hosting and supporting this work.
TABLE OF CONTENTS
A C K N O W L E D G M E N T S ......... ......... .. ..................................................................... iii
LIST OF TABLES ........ .................... .......... .......................... vi
LIST OF FIGURE S ......... ..................................... ........... vii
ABSTRACT .............. .......................................... ix
1 IN TR OD U CTION ............................................... .. ......................... ..
1.1 H history ..................................................................... 1
1.2 The C ase for M iniaturization...................................................................... ... 4
1.2 Common Materials for Lab-on-a-Chip Devices..................................................5
1.4 D design Attributes .................. .................................... ................ .8
1.4.1 Interconnect .................. ............................ .. ....... ................ .9
1.4 .2 P propulsion ......................................................................................... ......... 11
1.4.3 Isoelectric F focusing ........................................................ ............... 12
1 .5 O bjectiv e s ................................................................13
2 CHARACTERIZATION OF FABRICATION PROCESS ........................................14
2.1 Previously Disclosed M ethods of Fabrication .................................................14
2.2 Fabrication Process and M ethods of Testing.....................................................16
2.2.1 M aster F abrication .......................................................... ............... 16
184.108.40.206 Photom ask design....................................... ......................... 16
2 .2 .1.2 Silicon m aster............. .................................... ........ ... ........... 17
220.127.116.11 G lass m asters .................................... ...................... ........... .... 18
18.104.22.168 Electroform fabrication ........................................ ............... 19
2 .2 .2 M o ld in g ................................................... ................ 2 0
2 .2 .3 M illin g ................................................................................................. 2 2
2.2.4 L am ination................ .... ............................. .............. .............. 24
2.2.5 Demonstration Protein Separation............................................. 25
2.3 R results and D discussion .............................................. .............................. 26
2.3.1 M holding ............................................... ............... 26
22.214.171.124 Effects of temperature and pressure on device thickness...............26
126.96.36.199 Transfer of ridge detail ............. .............................. ...............28
2 .3.1.3 C razing ................................................ ............... 32
188.8.131.52 G lass-based e-form .................................. .......................... 35
2.3.2 L am nation ......................................................... .. ....... ... 35
184.108.40.206 Sandw ich m material ........................................ ........ ............... 36
220.127.116.11 R oller pressure........................................ ............ ............ 37
18.104.22.168 R oller tem perature................................................. ...... ......... 38
2.3.3 Milling .................................................... 40
2.3.4 Actual U sage Protein Separations .............................................. 43
2.4 Conclusions ............................................. 44
3 DESIGN, FABRICATION, AND TESTING OF A DEVICE FOR PROTEIN
EXPRESSION ............. .... ................. ......... .............. .......... .. 46
3.1 Introduction ............. ..... ...................... 46
3.2 Synthesis of P roteins.......... .......................................................... .... .... ..... 46
3 .3 E x p erim mental ...............................................................4 8
3 .3 .1 A d h e siv e s ............................................................................................. 4 8
3.3.2 D evice F abrication.......................................................... ............... 49
3.3.3 P rotein E expression .......................................................... ............... 50
3.4 R results and D discussion .............................................. .............................. 51
3.4.1 D vice D designs ........... .............................. .... ...... ................. 51
3.4.2 A dhesive B iocom patibility ................................... ........... ................... 53
3.4.3 P rotein E expression .......................................................... ............... 56
3.5 C conclusions ............................................. 56
4 CONCLUSIONS AND FUTURE RECOMMENDATIONS ...................................58
4.1 Compression M holding Process Optimization .............. ............. ....................58
4.2 D evice for Protein Expression .................................... .......................... ......... 60
4.3 O overall C onclusions......... ...................................................... ... ... .... .....61
CONSTRUCTION AND CALIBRATION OF LOAD CELL USED FOR PEEL
T E S T S .......................................................................... 6 2
A. 1 Design................................................. 62
A.2 Construction.................... ................................. 64
A .3 C alib ration ................................................................6 6
L IST O F R E FE R E N C E S ..................................................................... ..... ...................67
B IO G R A PH IC A L SK E TCH ..................................................................... ..................73
LIST OF TABLES
1-1 Properties of polymeric materials compared to glass and silicon. ...........................8
2-1 Width and separation of ridges shown in Figure 2-13...................... ...............32
LIST OF FIGURES
1-1 Possible embodiment of hand-held LOC system. ....................................................... 5
1-2 Basic process of isoelectric focusing with carrier ampholytes..............................13
2-1 Basic patterns used in this investigation........................................... ...............17
2-2 Process diagram for silicon m aster ................................................. ....... ........ 18
2-3 Process diagram for glass m asters. ........................................ ......................... 19
2-4 E-forms used for this research.. ........................................ ................... 20
2-5 Expected profiles of the molds used in this study. .............................................. 20
2-6 Side view of pressing operation ..................................................... ............... 21
2-7 Device exhibiting discoloration ............ ........................................... ............... 22
2-8 Method used to evaluate milling process. ...................................... ............... 23
2-9 Optical detection setup used for this investigation..................................................26
2-10 Observed change in device thickness as a function of temperature .......................27
2-11 SEM picture of part of a plastic device.. ........................................ ............... 28
2-12 Typical profilom etry results.. ............................. ................................................ 29
2-13 Variation in ridge height as a function of temperature and location ......................30
2-14 Feature size data taken from two other silicon wafers that were candidates to be the
m aster. ............................................................................. 3 1
2-15 Dependence on crazed area with reference to cooling conditions.........................33
2-16 Crazing as a function of temperature after cooling in a 60 C oven.........................34
2-17 Devices molded against the glass-based e-form .....................................................35
2-18 Lam inator w ith preheating hotplate....................................................................... 36
2-19 Comparison of devices laminated at different roller settings ...............................37
2-20 Peel test results grouped by roller temperature. ........................................... ........... 38
2-21 Average value of the average in the first two seconds of the test ............................39
2-22 Blockage that extends beyond burr location sometimes observed after lamination.40
2-23 Post-lamination from drilling and milling using a 2 mm-diameter end mill............42
2-24 Results from using a razor blade for burr removal................................................43
2-25 IEF for different channel lengths under a constant 500 V................... ............44
3-1 Illustrations of w ell-device. .......................................................................... ... .... 48
3-2 O original concept for device. .............................................. ............................. 52
3-3 Evolution of dual chamber design. ........................................ ......................... 52
3-4 PCR results for adhesives considered in this study. ................................................54
3-5 Demonstration of incomplete curing of epoxy during insert fabrication .................55
3-6 Comparison of expression of CAT between microcentrifuge tube and device..........56
A-i Illustration of load cell.................... .... ................. ..... .........62
A-2 Circuit diagram for Wheatstone bridge used to collect data............... .................63
A-3 Dimensions used for load cell in millimeters. ................................ .................64
A-4 Completed load cell. ..................................... ........ ................. 65
A-5 Calibration curve obtained for load cell. ...................................... ............... 66
Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Science
PROCESS CHARACTERIZATION OF FABRICATING PLASTIC MICROFLUIDIC
Chair: Z. Hugh Fan
Major Department: Mechanical and Aerospace Engineering
Running chemical analysis in microfluidic devices, sometimes referred to as lab-
on-a-chip technology, can increase accuracy while reducing the analysis time, sample
size, and the amount of reagents consumed. While glass, silicon, and plastic can be used
to make these devices, plastics have the advantage of lower costs, resulting in disposable
devices that eliminate the risk of cross-contamination. Applications where these devices
can be used include, but are not limited to, drug discovery, point-of-care or field testing,
and forensic analysis.
Most plastic devices are created using an embossing or injection molding process,
and usually only the conditions used to fabricate the particular device being discussed are
disclosed; how changing those variables affects the final product is not investigated in
detail. Compression molding is a compromise of these fundamental techniques, using
embossing technology to mold polymer resin instead of preformed plaques. A
characterization of a process for creating microfluidic devices by compression molding a
chip, milling reservoirs and edges, and then sealing by lamination is presented; variables
associated with each sub-process are characterized, and settings are recommended.
In addition, the development of the method used to fabricate a two-chambered
device for the specific purpose of protein expression is discussed; greater amounts of
protein were found to be expressed in this device than in standard microcentrifuge tubes
under identical conditions.
Despite the recent growth of microchemistry in both applications and methods, the
need for analyzing small quantities of materials can be dated back to more than a century
ago, when the novelty of radioactive materials was at the forefront of science. In an 1898
paper, Pierre and Marie Curie announced the discovery of a substance more radioactive
than uranium, and suggested the name polonium . Though they were able to
determine it was chemically similar to bismuth, the full strength of its radioactivity could
not be determined because they ran out of sample (the amount of material they started
with is not mentioned). Six years later, summarizing her work and observations on
radium, Marie notes that the element could be recovered from pitchblende, a uranium ore,
at a rate of between 0.2 and 0.3 grams per ton .
In the late 1930s or early 1940s, the field of ultramicrochemistry was developed.
Though a professor named Andon Alexander Benedetti-Pichler was considered to be the
leader of the movement, perhaps the best summary of the work describes the efforts of
Glenn Seaborg's plutonium research in 1942:
With the special tools of ultramicrochemistry the young chemists could work on
undiluted quantities of chemicals as slight as tenths of a microgram... They would
manage their manipulations on the mechanical stage of a binocular stereoscope
adjusted to 30-power magnification. Fine glass capillary straws substituted for test
tubes and beakers; pipettes filled automatically by capillary attraction; small
hypodermic syringes mounted on micromanipulators injected and removed
reagents from centrifuge microcones; miniature centrifuges separated precipitated
solids from liquids. The first balance the chemists used consisted of a single quartz
fiber fixed at one end like a fishing pole stuck from a riverbank inside a glass
housing that protected it from the least breath of air. To weigh their Lilliputian
quantities of material they hung a weighing pan, made of a snippet of platinum foil
that was itself almost too small to see, to the free end of the quartz fiber and
measured how much the fiber bent, a deflection which was calibrated against
standard weights "It was said," notes Seaborg, "that 'invisible material was
being weighed with an invisible balance."' [3:409-410]
These researchers had the advantage of analyzing radioactive materials, which
could be characterized with the use of Geiger counters and x-ray film. But many of the
sensitive techniques we use today had either not yet been developed or were just in their
Mass spectroscopy (MS) was first developed in the early 1900s. Initial separations
and results were crude, with the greatest growth in the technology coming in the 1940s-
1960s. Though a number of subfields have developed, the general approach is to send
particles through an electromagnetic field, which then separates them by their
charge/mass ratio. Originally, only atoms and small molecules could be analyzed using
this technique; more recently, larger molecules including proteins have been successfully
analyzed with this approach [4,5]. Portable MS systems have been developed, with a unit
prices starting at approximately $75,000 [6,7].
Chromatography is a very powerful analysis tool. In general, a sample is carried in
a mobile phase over or through a stationary phase. As the sample travels through the
media, its constituents travel at different rates based on their affinities to each of the
phases, and the elution times of each constituent are measured by a detector at the end
[8:628-631,9]. The two broadest categories are gas chromatography (GC) and liquid
chromatography (LC), with several subdivisions existing for each technique [8:609-610].
Another separation technology is electrophoresis, which uses an electric potential
to separate charged particles. It is frequently used to analyze biological molecules, such
as proteins and DNA, since these molecules have an inherent charge [10:1].
The main problem with these systems is limited throughput. As the separation area
can only be occupied by one experiment at any time, analyses must be run serially (space
for housing these instruments, the requirement of skilled operators, and overall system
costs may be additional problems, but any comprehensive analysis system will most
likely encounter these issues). One way to address this challenge is to miniaturize these
instruments, as it offers the opportunity to run many experiments in parallel.
The first widely-recognized attempt at miniaturization was reported in 1979, when
researchers at Stanford University developed a gas chromatograph where the 1.5 meter
long separation channel fit on a 2"-diameter silicon wafer . For a variety of reasons,
development of additional micro analysis systems languished in 1980's, although
examples of individual components, such as microvalves, were reported .
In the early 1990s, articles again began to appear giving the theoretical basis for
development of lab on a chip . Harrison et al. [14,15] introduced a device formed
from a glass plate and appears to be the harbinger for modern research in the field.
Since then, an uncountable number of microfluidic devices have been reported, and
the field is continuing to grow. Devices have been developed or proposed to detect a
number of substances, including illegal drugs and explosives residues, biological and
chemical warfare agents, and environmental pollutants. They have also been developed
for growing cells, reproducing DNA via polymerase chain reaction (PCR), and
immunoassays to determine the efficacy of drugs and/or toxins [16-19]
1.2 The Case for Miniaturization
The general idea behind micro total analysis systems ([tTAS), also known as lab-
on-a-chip (LOC), is to scale down chemical reactions, biological assays, cell
manipulations, and other analytical processes as much as possible. These operations,
which include mixing, separations, and chemical reactions, are typically integrated into
planar chips. Some chips also embed sample preparation, propulsion, and sensing into
the device, though this is not a requirement.
One of the original driving forces behind miniaturization was accuracy and speed.
In 1990, Manz et al. argued that if you could design a system that was limited by mass or
thermal diffusion, analysis time would be a function of the flux rate, which has units of
area per time. If the diameter of a channel, d, was proposed as an arbitrary length scale,
the time required for analysis would be a function of the area, or d2. Thus, by reducing
the diameter of a channel by a factor of 10, the analysis time could be reduced by a factor
of 100. This reduction in length scale would lead to better and faster electrophoretic and
chromatographic separations .
A significant consequence of miniaturization is that the amount, and thus the cost,
of reagents required to complete a study is also significantly reduced. This is especially
important to applications where materials are expensive and/or rare and many
experiments must be done, as in the cases of immunoassays and drug discovery.
Another potential result is that that overall footprint required for each
instrument/test could shrink, depending on the requirements of the system. One of the
potential outcomes is systems the size of or smaller than personal digital assistants
(PDAs), similar to the illustration in Figure 1-1.
Figure 1-1. Possible embodiment of hand-held LOC system.
As with any technology, there are (and should) be concerns with its use. As
miniaturization requires smaller samples, those samples are more sensitive to
contamination. For instance, the Seattle Post-Intelligencer ran a series of articles in 2004
examining the Washington State Patrol Crime Laboratories ; it was noted that even
something as seemingly harmless as talking while collecting a sample has apparently
resulted in sample contamination, most likely due to errant bits of spittle that were
disbursed during the conversation. With increasing reliance on this technology for
identifying and prosecuting offenders, this is an issue of supreme importance, and can be
overcome by training, shields, and rigid adherence to procedure.
1.2 Common Materials for Lab-on-a-Chip Devices
Materials that are used to create functional microfluidic devices fall into one of
three categories: silicon, glass, and polymers; each has its own set of advantages and
disadvantages, and no one material has been found to be ideal for all scenarios.
Silicon is a crystalline solid most commonly sold in the form of wafers. As noted
earlier, the first microfluidic device was the gas chromatograph on a 2"-diameter silicon
wafer. Other silicon devices include the systems developed by Gray, et al.  Silicon
was a natural first choice since micro-scale fabrication had been well characterized
during the growth of the electronics industry [22:184-185] It is crystalline, and can be
etched isotropically (same etch rate in all directions, resulting in rounded features) or
anisotropically (different etch rate in different directions, resulting in features with high
aspect ratios and well-defined angles). Being less than a millimeter in thickness, wafers
tend to be brittle and can be expensive when compared to glass or polymers. Silicon is
also conductive, meaning that additional steps are required to insulate the channels if
using electrokinetic pumping for propulsion or electrochemical sensing.
Glass is an amorphous solid with excellent chemical and thermal resistance
properties. The devices reported by Harrison et al. were created by etching channels into
glass plates, and then using another plate through which holes had been ultrasonically
drilled to seal the channels [14,15]. On a laboratory/prototyping scale, it can be difficult
to achieve consistent channel dimensions, through-holes, and bonding.
Polymers can be split into two basic groups: thermoplastics and thermosets.
Thermoplastic polymers have two functional groups on each monomer unit, resulting in
long chains of material. These chains then intercalate amongst each other and fold upon
themselves to form structures. These chains can be unlocked by heating the polymer,
which allows them to be formed and reformed into usable products, but also allows them
to be melted and destroyed. Thermosets, on the other hand, are made up of mixture of
monomers, some of which have more than two functional groups. As a result, the
structure is cross-linked, meaning that it is essentially one very large molecule. This
structure gives thermoset materials excellent temperature stability because they won't
melt, but also makes it difficult to reprocess the material.
With the exception of the wide use of polydimethyl siloxane (PDMS), it appears
that thermoset polymers are not used to make functional chips (though epoxies have been
used for packaging ). PDMS shows a good ability to replicate features , but its
elastomeric nature makes it inappropriate for many applications. Use typically entails
mixing two liquids (base and curing agent), pouring it into a mold, and allowing it to
In the mid- to late 1990s, many investigators started making chips from
thermoplastics. Polymethyl methacrylate (PMMA) appears to be the most commonly
used plastic, with occasional references to other materials. A few groups work with
copolymers, usually in the form of cyclic olefin copolymers (COC); copolymers are
created by mixing two or more types of monomer units; changing the ratio these units are
combined in alters the mechanical and thermal properties of the final product. Though
more expensive, COCs have exhibited greater chemical resistance and optical properties
than other polymers .
Table 1-1 summarizes some of the important properties of the materials discussed
above, as well as other polymers that may arise as options for microfluidic systems.
From a commercialization standpoint, thermoplastic polymers are the best material
for most microfluidic devices. They are relatively inexpensive, and much, if not all, of
the molding and sealing process can be automated, resulting in low manufacturing costs
and high reproducibility, which in turn leads to the option of disposable devices. Existing
commercial examples are created from polymers, such as Gyros' LabCD . However,
in a research or prototyping environment where few chips will be made, fabrication of
thermoplastic chips requires much more in terms of equipment and effort than PDMS.
Table 1-1. Properties of polymeric materials compared to glass and silicon.
Elasticity Tg Water Vapor Water
(GPa) (C) Permeability Absorption, 7 day
Material [27:778] [27:801] (g/m2 day)  immerse (%)
methylacrylate 2.24-3.24 3 -13 1.7 
Polycarbonate 2.38 150 15 0.45 
Polystyrene 2.28-3.28 100 N/A <0.1 
terephthalate 2.8-4.1 70 -3 0.1
2.6 80 0.023 <0.01
3.2 140 0.035 <0.01
PDMS 62 -123 N/A 0.1 
so 69 N/A N/A N/A
Silicon (111) 187 N/A N/A N/A
1.4 Design Attributes
When creating a microfluidic device, a number of issues must be considered and
overcome. Fairly obvious is channel and reservoir design to accomplish desired tasks,
which could be conveyance of material from one area of chip to another, mixing, or
chemical reaction. Other features, such as heating elements or sensing electrodes, are
only included on an application-specific basis. A summary of every known technology
that could be included in a microfluidic device is beyond the scope of this thesis.
However, it is necessary to provide an overview of the design attributes and technologies
incorporated into the devices discussed in later chapters, and that is the purpose of this
When the macro-to-micro interface is being designed, one needs to ask a few
questions. What does the interface need to be capable of doing? Is it simply introducing
sample to the system, or does it have to also include a preparation step such as filtering or
atomization (e.g., for interfacing with mass spectrometry)? Will pneumatic pumping be
used, requiring the use of leak-proof interconnects? In applications involving EOF, how
will the electrodes be included? Is the entire interconnect to be disposable or just a certain
part of it? Is it advantageous to integrate interconnects into the device or leave it as a
After understanding the needs, one should examine the issues related to the device
design [34,35]. First, there is a lack of features to adhere to and align with, since most
devices tend to be planar in nature. Even if the manufacturing process is rigorously
controlled, aligning to an edge isn't always a failproof method of creating a connection.
Second, the dimensions and tolerances of the device are small; it is beneficial to have the
dimensions of the interface agreeing with the existing industrial standards for the desired
application. The distance between ports in some devices is less than a millimeter. Third,
the properties of the materials for interconnects must be compatible to the device, since
the device can be fabricated from a number of materials, including glass, plastics, or
* This section has been published with other materials in : Fredrickson, C.K and Fan, Z.H., 2004,
"Macro-to-micro interfaces for microfluidic devices," Lab Chip, 4, pp. 526-533.
Materials used in interconnects include metals such as steel and aluminum, glass,
silicon, and polymers such as PEEK, PTFE, and acrylic. Rubbers or silicones are often
used for gaskets, and mechanical means, such as screws and bolts, may be used to hold
interconnects and device together. As often practiced, interconnects can also be held
together through the use of adhesives such as epoxies. Each of these materials has its own
set of properties and limitations, effectively forming an envelope of acceptable operating
After answering the fundamental questions and addressing the design obstacles, it
may be prudent to modify the design to incorporate desired characteristics. These
characteristics may include the following.
* Ease of assembly. Interconnects should align to the ports on the microfluidic device
easily and with minimum need for supporting jigs. If an interconnect consists of
many components, they should go together quickly with minimal use of tools.
Replacement of consumables, such as capillary tubes and gaskets, should not be
* Reliability. Interconnects should be leak free at standard operating conditions for
* Chemical compatibility. The materials used in the assembly will not unduly
influence or react with the samples being examined.
* Minimal dead volume. Interconnects should be free of areas where fluid does not
circulate. Dead volumes increase reagent usage and reduce precision and
* Maximum field of view. An interconnect should allow the user to gather data when
the analysis will be conducted optically, such as by laser induced fluorescence or
* Minimal pressure drop when the flow is pneumatically driven. Pressure drop is
caused by the geometry of the system as well as the constriction.
* Ability to operate over a range of flow rates.
* Ability to be automated. Certain experiments require repetitive steps, such as
dispensing, which should be automated to increases precision and reduce the
amount of time required for the operation.
* Low cost.
After designing and fabricating interconnects, data should be collected to test if
they successfully meet the desired operational parameters. Leakage can be tested by
filling the channels and then using a pump to pressurize it until failure is observed. The
liquid may be dyed to facilitate detection. Flow rates and pressures are measured during
operation. Mechanical strength may also be tested by pulling the capillary or tube until
One of the easiest ways to propel fluids through microfluidic devices is to create a
pressure gradient through the use of a pump. The problem with these systems is that as
the channel dimensions are reduced, the pressure drop over a given distance increases
significantly, resulting in decreased flow or the need for stronger pumps .
This obstacle is frequently sidestepped by using electroosmotic flow (EOF) to drive
the liquids: if the channel is filled with a buffer solution, some of the disassociated ions
will adsorb onto the channel surface and give the wall an inherent charge. At the same
time, a "diffuse double layer" of opposite charges will develop in the solution adjacent to
the wall. When voltage is applied, a shear plane will develop between the opposite
charges, and the material in the middle of the channel will move.
The observed velocity of a particle in this system can be given by the following
VOBS = E(IEOF + tEP) (1)
where voBSis the velocity in cm/s, E is the electric field strength in V/cm, and UEOF and
,/EP are the electroosmotic and electrophoretic mobilities, respectively, in units of cm2/V.
The electroosmotic mobility is a measure of the speed of neutral material in the channel,
and can be found by the relationship
PEOF -- (2)
where Sis the thickness of the diffuse double layer in cm, e is the charge per unit surface
area (coulomb/cm2), and r is the viscosity in g/cm s. It is desirable to reduce or eliminate
EOF in some instances, and this can be accomplished by increasing the viscosity of the
liquid in the channel.
The electrophoretic mobility is a measure of the speed of charged particles in the
channel (which are more attracted to the anode or cathode than a neutral particle), and is
related to a particle's charge in coulombs, q, and radius in cm, r, by the equation
/EP q (3)
1.4.3 Isoelectric Focusing
Isoelectric focusing is a method to separate and analyze proteins. As mentioned
earlier, electrophoresis relies on an electric field to separate molecules and ions.
However, nothing in the process stops the motion of the analyte; if the voltage were
applied for an infinite amount of time, eventually all of the charged molecules will end up
at the electrodes. Thus the technique is dependent on duration, composition of the
separation media, and strength of the applied electric field.
A related technique that does not face this problem is isoelectric focusing, which
requires the introduction of a pH gradient [10:80-83]. Proteins are amphoteric, meaning
that they will accept or donate protons, and thus charge, depending on the pH of the local
solution. While charged, they are subject to the electric field passing through the solution,
and move towards the electrode of opposite charge; at their isoelectric point (pI), they are
neutral and the field has no effect. If the sample to be separated contains only proteins,
only a few discrete pH values exist in the path between the electrodes, resulting in a step-
like gradient and inconsistent results. The gradient can be smoothed out by immobilizing
one along the separation path or by adding a mixture of amphoteric molecules called
ampholytes, which will establish a gradient when voltage is applied. As long as the
voltage, and thus the gradient, is maintained indefinitely, proteins will eventually reach
an equilibrium position and then remain stationary.
Figure 1-2. Basic process of isoelectric focusing with carrier ampholytes. The carrier
ampholytes (circles) align in order of pH, indicated by the shading. Proteins
(stars) move along the pH gradient until they reach their isoelectric point
The objective of this thesis work is to develop and study fabrication processes for
microfluidic chips. Chapter two presents the characterization of a compression molding
process for fabricating COC-based microfluidic devices, and reports how changing
process variables affects the final results; though the devices are used to separate proteins
via isoelectric focusing, the emphasis is on the process, and is sufficiently generic that it
will apply to a range of device applications. Chapter three details the development of a
device for the specific purpose of expressing proteins.
CHARACTERIZATION OF FABRICATION PROCESS
This section focuses on the characterization of a generalized fabrication process.
The basic approach employed is to compression mold the bulk device and channels.
Reservoirs will be milled in the device so that the channels can be accessed, and then the
channels will be sealed by lamination. But before this method is investigated, it is
necessary to review the methods that have been used to make some of the multitude of
devices that have been reported to date.
2.1 Previously Disclosed Methods of Fabrication
A number of techniques have been reported in the literature for patterning channels
into polymeric substrates. The most frequently reported method is embossing, which
consists of pressing a pre-formed plastic plate against a mold, which in turn have been
made from wires , silicon wafers , silicone [37,38], quartz , or nickel
electroforms [40,41]. While relatively easy to implement, the method does have its
drawbacks. For example, Xu et al. showed when pressing at room temperature, the
resulting feature size is a function of both pressing time and pressure, and that the
channel depths were 50-70% of that of the mold . Increasing the embossing
temperature yields features in the plastic devices much closer to the mold dimensions
Injection molding requires melting the material, and then injecting it into a cavity.
While used frequently in industry to make plastic parts, it is used less frequently in
laboratory/prototype settings due to the costs of the injection molder and molds, though
some groups have employed this method [43,44]. One example of note is the work of
Ahn et al, who custom-an made injection molding machine and used an interchangeable
nickel electroform insert similar to the type that will be used in this investigation as their
Compression molding is a compromise between embossing and injection molding,
placing resin (pellets) on the mold instead of a preformed plate, and was selected in an
attempt to increase the fidelity of features that could be transferred during the molding
process without much additional expense. Liu et al. have used this method to form
plastic devices from a silicon wafer . Lesser utilized methods for creating
microfluidic devices include using a mill to cut channels , using ultraviolet light to
polymerize a liquid in a mold [48,49] and laser ablation .
Sealing is frequently reported as being accomplished through a thermal annealing
process, which is essentially heating the device to a temperature near the softening point
of the material for a prolonged period of time. For example, Bowden clamped a cover
sheet to a PMMA-based device, baked it at 110 oC for 2 hours, and then allowed it to cool
from 30 C to room temperature overnight in an oven . While an annealing time of
hours appears to be more frequently used, times as short as 10 minutes have been
reported [48,51]. However, by heating the material to a temperature close to its softening
point, there is a risk of deforming the channels. To allay this, some have made their
devices from one grade of material, and then used a thin layer of another grade with a
lower glass transition temperature (Tg) as an adhesive [47,52] Use of a heat-activated
adhesive to bond the film and device has also been reported .
For this investigation, it is hoped that by working with a commercial laminator,
sealing will be accomplished faster than the thermal annealing method. In addition, the
devices would more closely simulate those that are commercially available.
2.2 Fabrication Process and Methods of Testing
2.2.1 Master Fabrication
The masters were created using a photolithographic process similar to that used to
fabricate microchips for the electronics industry. Patterns of interest were designed using
a computer-aided design (CAD) program (AutoCAD 2002, Autodesk, San Rafael, CA).
Each pattern was designed to fit on a standard 3" x 1" microscope slide.
22.214.171.124 Photomask design
Figure 2-1 shows the general patterns used in this investigation. The "2-D" device
consisted of a number of horizontal channels crossed by a single vertical channel of
identical width. Two versions of this pattern were created; one for silicon and one for
glass. Since silicon can be anisotropically etched using deep reactive ion etching (DRIE),
features were designed to be 30 rtm wide and separated by 90 |tm gaps; 87 channels
could be fit into the device. However, it also was designed with the intention that the
silicon wafer would be the mold, thus the "channels" were in fact ridges that would form
channels when a suitable polymer was applied to the pattern. Glass can only be
isotropically etched, so features were designed to be 20 rtm wide and separated by 340
|jm gaps on the photomask, with the understanding that the channels would grow during
the isotropic etching process, undercutting protective masking layers and resulting in a D-
shaped profile; only 29 horizontal channels could be fit into the same area.
The "6-channel" design consists of six channels 10, 21, 32, 43, 54, and 65 mm
long, and was intended for glass. Each channel had a 4-mm long arm protruding to the
side; the arm itself was placed 4-mm from the end of the channel. On the mask, each
channel was 20 |tm wide, again with the understanding that the channels would grow
during the isotropic etching process.
IA A B
Figure 2-1. Basic patterns used in this investigation. A) 2-D. B) 6-channel.
The file with the pattern was sent to a vendor (Photo Sciences; Torrance, CA), and
a chrome-on-glass photomask was created. This mask was used to transfer the patterns to
the master materials, as explained in the next two sections.
126.96.36.199 Silicon master
Figure 2-2 shows the fabrication process for the silicon masters. Prime grade
silicon wafers (100-orientation, University Wafer, South Boston, MA) were washed with
tetrachloroethylene (TCE), acetone, methanol, and deionized water before use; the TCE
removes any oils that accumulated during the packaging and shipping process, and each
subsequent chemical removes residue of the previous. Shipley 1813 photoresist (Shipley,
Marlborough, MA) was decanted onto the wafer, spun at 4000 rpm to a thickness of
approximately 2 |tm (step 1 Figure 2-2), and soft baked for 30 minutes at 85-95 C. The
2-D pattern was exposed (step 2) using an MJB3 mask aligner (Karl Suss, Germany) with
405 nm light, developed (step 3) (AZ Mif 312 1:1:2, Clariant Corp., Somerville, NJ), and
then hard baked for 30 minutes at 115 oC.
The wafers were etched using deep reactive ion etching (DRIE) (Surface
Technology Systems, Ltd., Newport, UK) (step 4). DRIE creates high-aspect ratio
structures by alternating between etching and the application of a passivation layer to
protect the areas that shouldn't be etched. Material that remained after the etching
process (both photoresist and passivation layer) was removed using a five-minute 02
plasma etch (step 5).
2 1 I
31 l 5 r i
Figure 2-2. Process diagram for silicon master.
188.8.131.52 Glass masters
The process used to fabricate the glass masters was adapted from known methods
. Glass microscope slides (Fisher Scientific, Atlanta, GA) were annealed (step 1 in
Figure 2-3) in a programmable furnace (Barnstead 4500, Barnstead Scientific, Dubuque,
Iowa) to reduce internal stresses created during the manufacturing process, resulting in
smooth channel edges during the etching steps. To protect the glass during the etching
process, 300 A of chrome and 1000 A of gold (both 99.99+%, Alfa Aesar) were vapor
deposited via an electron-beam process (chrome serves as an adhesion promoter between
glass and gold). Positive photoresist (Shipley 1813) was decanted onto the slide, spun at
4000 rpm to a thickness of approximately 2 itm, and soft baked for 30 minutes at 85-95
C (step 2). The patterns were exposed (step 3) using 405 nm light on an MJB3 mask
aligner, developed (AZ Mif 312 1:1:2) (step 4), and then hard baked for 30 minutes at
To define the channels in the glass slides (step 5), the exposed gold and chrome
were first etched away by submerging in gold and chrome etches in sequence (Type TFA
and 1020 respectively, both of Transene Company, Danvers, MA) until the exposed
metal areas were removed. The slides were then submerged in a 20% HF/14% HNO3/
66% H20 solution for approximately 7 minutes. Once the channels were defined, the
remaining photoresist was rinsed away with acetone, and then the gold and chrome etch
procedures were repeated (step 6) to remove the rest of the metal mask.
Figure 2-3. Process diagram for glass masters.
184.108.40.206 Electroform fabrication
The masters were shipped to Optical Electro Forming (Clearwater, FL), who
created the nickel electroforms (e-forms) by an electroplating process. They were
secured to a metal backer plate and a thin layer of silver was applied to increase the
conductivity. Voltage was applied for approximately 24 hours, yielding the e-forms
shown in Figure 2-4.
In the case of the silicon-based mold, it was unrealized that silver was still on the
mold when it was returned, and it lifted off during the first few pressing operations. The
discolorations observed in the edges are remaining silver residue. It was also observed
that bumps form on the back of the e-form where the nickel accumulated faster during the
electroplating process; these were ground off after receipt.
Figure 2-4. E-forms used for this research. A) silicon master; the odd coloration
(especially around edges) is due to improper removal of the conductive silver
layer. B) glass slide masters. Pennies were used for size reference.
Figure 2-5 shows the expected topology of the e-forms used to create devices in
this study. Figure 2-5 A shows the profile of the silicon-based e-forms, from which we
expect ridges 30 |tm wide separated by 90 |tm gaps. Figure 2-5 B shows the profile of
the glass-based e-forms, from which we expect D-shaped channels approximately 30 |tm
deep and 120 |tm wide.
Figure 2-5. Expected profiles of the molds used in this study. A) silicon-based
electroforms. B) glass slide master.
An appropriate mass (- 6 g for the silicon-based e-form; 8 g for the glass-based
e-form) of Topas 8007 resin (Ticona, Florence, KY) was placed on the desired pattern
on the e-form and covered with a 6" square glass plate. This sandwich was placed
between the platens of a hydraulic press (Carver VMH, Carver, Inc., Wabash, WI) that
were heated to the desired temperature (100-130 C in approximately 5 C increments,
and 150 C). The lower platen was raised until the glass plate contacted the upper platen,
allowed to sit for five minutes to warm the resin, and then compressed using a
predetermined rate until the desired force (either 5,000 or 10,000 lbs; referred to as 5 and
10 kip, respectively) was reached (as the ram unit is closed by manually cycling a lever,
with each cycle raising the lower platen about 0.8 mm, the closure rate was established
by setting the desired time for the lever to travel from full-up to full-down in 10, 30, or 60
seconds; a graduated stick was used as a reference). Spacers were used to achieve the
desired thickness of the device. This final configuration is shown in Figure 2-6. Except
for the devices cast at 100 and 105 C, three devices were cast for each combination of
temperature, force and downstroke time.
Figure 2-6. Side view of pressing operation.
After five minutes at force, the mold was removed from the press and allowed to
cool to room temperature before the devices were removed from the mold. To determine
the amount of detail that was being transferred from e-form to device, these samples were
metrologically characterized using a profilometer (Dektak IIa, Veeco Instruments,
Woodbury, NY); scan distance was 250 |tm and speed was set to the "slow" setting on
the machine. Many devices exhibited white regions in the areas of the channels, a
behavior called crazing; an extreme example is shown in Figure 2-7; wherever possible,
the scans were not done in these areas.
To quantify the extent to which the crazing occurred, each device was scanned at a
resolution of 300 dpi using a flatbed scanner (Canoscan 30, Canon Inc.) in grayscale
mode. The devices were covered with a piece of blue vinyl to provide contrast. Images
were analyzed using the threshold function in ImageJ (http://rsb.info.nih.gov/ij/).
Figure 2-7. Device exhibiting discoloration.
As this first batch of devices showed no relationship between crazing and
temperature, a selection of devices was recast using 5 kip of force, a 10-second
downstroke, and temperatures of 115 and 125 C, with the cooling conditions controlled:
allowed to cool to room temperature on a countertop (relatively insulated surface),
allowed to cool to room temperature on metal table (relatively conductive surface), or
allowed to sit for 10 minutes in a 60 C oven. When it was found that the devices cooled
in the oven showed the smallest amount of crazing, a second overlapping set was recast
using the same force and compression rate, but over the temperature range of 110 C to
130 C in 5 C increments, and cooled in the oven to look for temperature dependence of
the behavior. Three devices were cast at each temperature and cooling condition. All of
these devices were scanned and analyzed as described in the previous paragraph.
Reservoirs and final chip outlines are defined using a miniature computer
numerical controlled (CNC) mill (Flashcut 2100; Flashcut, Menlo Park, CA). To find the
optimum milling conditions, the test pattern shown in Figure 2-8 A was milled into the
chips cast against the glass-based 2-D pattern using a 2-mm diameter, twin fluted square-
end mill with titanium nitrate (TiN) coating (Richard's Micro Tool, Plymouth, MA).
Spindle speed was varied from 750 to 2000 rpm, but held constant for each device, while
one hole and one slot was cut for vertical and horizontal feed rates varying from 25
mm/min to 200 mm/min as shown; for the spindle speeds of 1750 and 2000 rpm, an
additional device was cut with feed rates varying from 225 to 300 mm/min.
By making each cut in a single pass, one side of the cut was climb milled while the
other was standard milled, as shown in Figure 2-8 B (when milling a free edge, the
material can be ejected into the area immediately behind or in front of the cutting tool,
depending on the directions of rotation and feed; standard milling ejects the material to
the front, while climb milling ejects it to the rear).
climb sta ndard
25 50 75 100 125 150 175 200 A B
Figure 2-8. Method used to evaluate milling process. A) Pattern used to test different
combinations of feed rate and spindle speed. Numbers indicate feed rates in
mm/min. B) Illustration showing difference between climb and standard
Patterns were cut with the channels on top (facing cutting tool) and bottom (facing
table); an acrylic plate was secured to the table to keep the tool from hitting the table.
Debris was removed by blowing air across the device; an air line was fitted with an
adjustable valve (Campbell Hausfeld, Harrison, OH) equipped with a pressure gauge set
to either 0 (off), 10, or 20 psi. A 12" flexible hose (Loc-line, Lockwood Products, Lake
Oswego, OR) equipped with a 1/16" diameter nozzle was connected to the outlet of the
valve; the nozzle was positioned 3-5 cm from the cutting tool, and directed to blow at the
tool-device interface at an angle of about 350 from horizontal.
Topas 8007 film (4 mil, or approximately 100 |tm thick) was acquired and cut into
strips one inch wide and three inches long. Film and milled devices were inserted in a
custom-made acrylic rack and immersed in a 1% Alconox (Alconox, White Plains, NY)
solution in an ultrasonic bath for 5-10 minutes. The detergent was disposed of and the
pieces were rinsed at least twice by submerging in ultra-purified water in an ultrasonic
bath for 5-10 minutes. Devices were air-dried in a laminar hood. Film and chip were
sandwiched between two layers of a 2 mil-thick metallized Mylar film (Hydrofarm, Inc.,
Petaluma, CA), heated for 90 seconds on a 70 C hotplate, and then run through a
laminator (Catena 35, GBC, Northbrook, IL) at a rate of approximately 30 cm/min.
A peel test was devised to test the bond strength between film and device, as peel is
the method that is most likely to lead to device failure. Three test specimens were
created for each roller temperature using the lamination procedure described above. A
load cell was created from a piece of acrylic and a strain gage as outlined in the appendix,
and mounted in the CNC mill. Delamination was initiated by a razor blade, and a flap of
film approximately 1.25 cm in length was created. The flap was attached to the metal T
on the load cell by a piece of Scotch Tape (3M, Minneapolis, MN). Tape was selected
because it took up negligible space, appeared to be more secure than clamps or clips of
that size, and allowed the forces to be spread out over the entire surface of the film being
held, rather than focus them along a line or points as other methods of clamping may
have done. The gage was pulled 50 mm vertically at a rate of 300 mm/min, pulling the
film with it. Voltage data was collected from a Wheatstone bridge containing the strain
gage via a digital multimeter (Model 72-6870, Tenma) and Tenma' s Data Logger
2.2.5 Demonstration Protein Separation
Isoelectric focusing of proteins in a channel was selected as a demonstration of the
effectiveness of the device. To suppress electroosmotic flow, a stock solution of
separation media of linear polymers was created by dissolving 5.5% w/w hydroxypropyl
cellulose (HPC) (Mn = 10,000, Aldrich, St. Louis, MO) and 2% 2-hydroxyethyl cellulose
(HEC) (My = 90,000, Aldrich) in water at 60 oC under constant agitation by a magnetic
stirrer; The stock was further diluted with water to form a solution that was 1.83% HPC
before use. 85 [tl of this solution was combined with 10 [tl of 80% w/w glycerol (Fisher
Scientific, Atlanta, GA) and 5 [tl of carrier ampholytes for pH range 3-10 (Bio-Rad
Laboratories, Hercules, CA). Recombinant green fluorescent protein (GFP) (BD
Biosciences Clontech, Palo Alto, CA) and R-Phycoerythrin (RPE) (Cyanotech, Kailua-
kona, HI) were added such that the final protein concentration was 5[g/mL. The
separation media and proteins were loaded into one of the wells, drawn through to the
other well by a vacuum pump, and the remaining material in the well was removed. The
end wells were then filled with solutions of 15 mM acetic acid (Fisher Scientific) and 15
mM ethanolamine (Sigma, St. Louis, MO) to serve as anolyte and catholyte, respectively.
To study the effectiveness of different channel lengths, the 21, 32, 43, and 54 mm
long channels were loaded. Platinum wire electrodes were placed into the end wells, and
500 V was applied from a high voltage power supply (Glassman High Voltage, High
Bridge, NJ) interfaced via a Labview SCB-68 card to a computer running an in-house
developed Labview program. Protein position was monitored via a charge-coupled
device (CCD) (Apogee Instruments, Auburn, CA) mounted directly above the chip.
Illumination was provided by a 488-nm laser which was passed through a beam expander
and lens so that the entire channel was illuminated. A schematic of this setup is shown
in Figure 2-9.
ND Filter p :1 Bandpass Filter
Figure 2-9. Optical detection setup used for this investigation .
2.3 Results and Discussion
Though referred to as a resin, the plastic being used for this investigation is actually
in pellet form. Since it is not heated prior to application to the mold, there is a strong
possibility that air can become entrapped between the pellets during the compression
step, resulting in bubbles in the chip. In working with the materials and press before this
investigation, it was found that by slowing the rate at which the ram-controlling lever was
cycled, the inclusion of bubbles could be minimized.
220.127.116.11 Effects of temperature and pressure on device thickness
Since spacers were used, as shown in Figure 2-6, the thickness of every chip should
have all been the same, with an expected device thickness of about 1 mm. However, in
the course of handling the devices, differences were noted. The overall thickness of the
pieces was measured using dial calipers, with the data presented in Figure 2-10.
S5 kip, 10 sec.
1.60 o 5 kip, 30 sec.
A 5 kip, 60 sec.
1.50 10 kip, 10 sec.
E m a 10 kip, 30 sec.
S 1.40 lA 10 kip, 60 sec.
I 1.30-- t-
80 100 120 140 160
Platen Temp. (C)
Figure 2-10. Observed change in device thickness as a function of temperature.
The data suggests that device thickness is a function of both pressing temperature
and pressure, but independent of compression rate. The probable explanation is that,
while the temperatures being considered are above the materials Tg, they are well below
its melt point, leaving it in a rubbery transition state where both solid and liquid
properties are exhibited. In this viscoelastic state, it experiences creeping flow under the
constant pressure applied by the platens [27:478-481]. While creep is normally associated
with elongation under tensile forces (the standard example being the deformation of
plastic coat hangers after several years of use), it stands to reason that as a compressive
force is applied, additional compaction of the sample would be observed as the
experiment proceeded; the extent of compaction would depend on both the temperature
and initial force.
18.104.22.168 Transfer of ridge detail
Transfer of ridge detail from the mold to the devices was tested by pressing a
number of devices under different conditions against the silicon-based 2-D pattern. As
mentioned in section 2.2.1, the expected chip would be the negative of an actual device;
ridges 30 |jm wide separated by 90 |jm gaps were expected. Sections of a device were
observed under a scanning electron microscope, and Figure 2-11 shows typical results. In
general, the plastic devices appear to be acceptable. Other than some minor debris that
was probably introduced through various handling, the horizontal surfaces look smooth
and flat. It appears that a high degree of fidelity is being transferred from e-form to
device, as noted by the presence of scallops on the vertical surfaces; these were expected,
a result of the alternating etching as passivation steps in the DRIE process .
Another characteristic to note is the appearance of slight deformations along the top
edges of the features; this is particularly noticeable covers as shown by the inset.
Figure 2-11. SEM picture of part of a plastic device. Exploded view shows evidence of
ability to transfer scallops from mold to device, along with deformation in
Profilometry data was collected from each of the cast devices to quantify the
feature sizes. Typical profilometer results are shown below in Figure 2-12 A. One of the
first features to note in the profiles is the presence of large humps on the tops of the
ridges. This provides additional evidence of deformations along the top edges of the
features, as observed in Figure 2-11. These bumps were not observed in the wafers, as
shown in Figure 2-12 B.
-j- ..... _....._______.358 .__
I/i / 1 -- 6
1--150,00I I f10
-_ 4 _1.eee 00j 5800
I .6- J _
a 10 15 2N -0 L5O8 150 ;0@ 250
R CUR 31 A e i&6u M CURSOR = 237 R CUR 0 r 8, O M CUFSLiO 249
M CUR: 17,916 A @ 237uM SLOAN DEKTAK II A M CUR: 5,845 A 249.M SLOAN DEKTAK II B
Figure 2-12. Typical profilometry results. A) molded device B) silicon wafer. Arrows
in A) indicate dimensions measured to compile Figure 2-13.
The various points on a profile used to determine the ridge dimensions are shown
in Figure 2-12 A. Ridge heights were measured from the middle of the humps to the
bottom of the area between the channels, as shown by the arrow. The sloped surfaces
between the top and bottom of the ridges are an artifact of the conical shape of the stylus,
and are believed to be vertical or near vertical. Ridge widths and separations were
considered to be the distance between two ridge edges; edges were defined as the location
where the stylus stopped rising or began to fall during the course of the scan; in the event
a hump was present, the edge was the location on the separation-side of the hump where
it crossed the horizontal line passing through the point used to determine the height of the
ridge. As the profilometer analysis software had a horizontal incrementation of 2 ptm in
the mode used for measurement, it was not always possible to get the cursor to align
exactly with these points.
Feature height was measured at five locations spaced evenly across the pattern, and
was taken in areas without crazing whenever possible. The original intent was to take
these five values and average them; however, it was found that the centers of the channels
were significantly shallower than the ends. So instead, the results were grouped by
location across the pattern, as shown in Figure 2-13. The measurements varied from
approximately 33 |tm near the ridge ends to 29 |tm in the center, or about 13% difference.
Location 1 Location 3 Location 5
30 35 0 6-------------- p 35 0
S e0 sec 5 kip i L
5 0 0- 30 5-klp-340 se
340 ^1 f 330. 0 -^4- ----
Temperature (C) Temperature (C) Temperature (C)
Temperature (oC) Temperature ("C)
Figure 2-13. Variation in ridge height as a function of temperLocatonure and location. The
Comparison to the e-form itself was impossible because the profilometer stylus was
Sle e 30 es. ee s e esee e
227 0 30-------5---k--2-----0--28--i 270 5---k-----]---21----0--------
90 100 110 120 130 140 150 160 0 100 1 10 20 130 140 150 160 go 1 120 130 140 150 160
Te p t C) Temperature (C) Temperature (C)
Temperature. The scales are identical on all 5 graphs.t 4
5Comparison to the e-form itself was impossible because the profilometer stylus was0
too large to fit into the 30 sm channels. There was only one measurement done of the
was 0 ae eaho3h2rah 0hw h rdehih s ucino
31 praue 0h clsaeietclo l rps
Comprisn t thee-frm tsel wa imossile ecase te pofiometr sylu3wa
wafer used to make the e-form, and the wafer was destroyed during the e-form
However, the results were compared to two other candidate wafers, and the same
variation in feature height, as shown in Figure 2-14. The three wafers (master and two
candidates) were each etched using slightly different processing conditions, resulting in
the variance in feature height. Nevertheless, it is clearly seen that the center of the
pattern is noticeably shallower than the ends.
o Unselected #1 o Unselected #2
0 1 2 3 4 5 6
Figure 2-14. Feature size data taken from two other silicon wafers that were candidates
to be the master. Each was processed using slightly different conditions,
resulting in differences between each other and the master.
This variation is explained by "RIE-lag": during the plasma etch process, gases
must escape between etch cycles; the gases exhaust much more readily from larger
channels or surfaces. Since the ridges are adjacent and open to the large area surrounding
them, which is also being etched, the gases are able to exhaust faster near the ends and
deeper etching is observed [22:107].
Much like the data presented in Figure 2-13, the average values for the ridge widths
and separations showed very little differentiation with respect to temperature or
compression rate. To simplify presentation, they were averaged by location, and listed in
Table 2-1. Given the incrementation issues discussed earlier and that the device was not
always lined up exactly perpendicular to the stylus path, these values match well with the
expected dimensions across the entire device.
Table 2-1. Width and separation of ridges shown in Figure 2-13.
Location: 1 2 3 4 5
Average 29.9 29.0 29.8 29.7 30.7
St. Dev. 0.9 0.9 1.0 0.8 0.9
Average 90.6 91.2 90.7 90.6 89.9
St. Dev. 1.0 1.0 0.9 0.9 0.8
It is observed that for ridge height, width, and separation, there is little variation
within each of the locations. From this, we can conclude that macro feature replication is
independent of casting temperature and compression rate.
An unexpected observation during the casting of the test devices against the
silicon-based e-form was the appearance of white areas after removal from the mold.
This behavior was deemed to be crazing, the localized yielding of plastics when stressed
[27:483-484] and needed to be minimized or eliminated. Initial results (not shown)
indicated that there was no clear relationship between crazing and pressure or
temperature. It was then considered that this was the result of cooling conditions, which
were originally not closely controlled.
Cooling the devices adequately is an important part of the molding process; if the
pieces are removed from the mold too early, there is a risk that the channels or even the
devices themselves will deform. There are a myriad of methods available to cool the
device. Once the mold is removed from the press, it can be cooled at room temperature,
set on a hotplate or placed in an oven set at a temperature below the Tg but above room
temperature, or placed in a refrigerator. Once the mold is placed in the cooling
environment, it can be left for a set amount of time so that it cools but does not reach
equilibrium, or it can be left until the mold and device equilibrates at an established
temperature; as the former method can lead to transient results, the latter is preferred. For
simplicity, the only three methods examined here were to allow the device to cool all the
way to room temperature on a countertop (relatively insulated surface), allow the device
to cool all the way to room temperature on the metal surface of an optical bench
(relatively conductive surface), or allow it to equilibrate in a 60 C oven (20 C below the
Tg of the material). Figure 2-15 shows the amount of crazing detected for each of the
three cooling conditions tested.
Oven, 10 min
S125 C 0 20 40 60 80 100
115 OC O/oage of area
Figure 2-15. Dependence on crazed area with reference to cooling conditions. Devices
were molded at temperatures shown; oven was set to 60 C. Pictures are of
representative examples used to measure crazing.
It's apparent that the method used to cool the device affects the amount of area that
is crazed, though the behavior could not be completely eliminated. Obviously,
countertop cooling yields unacceptable results. Cooling on the optical bench
significantly reduces the amount of area that is crazed, but concentrates it to the area near
the vertical channel. Cooling in the oven further reduces the amount of area that is
crazed, but also distributes it across the entire device, and appears to show the pattern of
individual resin pellets.
When devices were cast at different temperatures, and cooled using the oven at 60
C (trying to minimize the amount of crazing that occurred), it was found that the amount
of crazing was indeed temperature dependent, as shown in Figure 2-16. Thus crazing is a
function of molding temperature and the method in which the device is cooled after
removal from the press.
a 20 -
105 110 115 120 125 130 135
Figure 2-16. Crazing as a function of temperature after cooling in a 60 C oven.
The susceptibility to increased flow at higher temperatures that results in the
thinner devices as discussed in section 22.214.171.124 also probably allows greater penetration
into the scallops shown in Figure 2-11, which in turn creates greater amounts of damage
when the device is removed from the mold. Seeing that all temperatures that had been
examined by profilometry yielded consistent ridge heights, it is recommended that the
temperature be kept as low as possible. The occurrence of this behavior would be most
easily reduced or eliminated by optimizing the etching conditions to minimize the
formations of the scallops on the master.
126.96.36.199 Glass-based e-form
The glass-based e-form was not subjected to as intensive of a characterization. The
e-form itself was found to have a slight amount of concavity in it as a result of the
manufacturing process (the plate that held the slides during the electroforming process
was not completely flat). Large voids were observed in the upper surface of the device,
though these could be pushed to the edges or scrap areas of the device by increasing the
amount of resin from six to eight grams and the increasing the casting temperature to 125
C. Crazing was not nearly as much of an issue due to the relatively smoothness of the
features; any crazing that did occur tended to exhibit itself as a scalloped pattern across
the top of the entire device, as shown in Figure 2-17.
Figure 2-17. Devices molded against the glass-based e-form. A) Typical pattern of
crazing from glass-based e-form when it did occur. B) Optimized glass
device showing no crazing.
The first attempts at lamination, using the laminator as supplied by the
manufacturer, were not good: complete bonding could only be accomplished running the
device through twice at the maximum temperature and lowest speed. However, there was
concern that the channels would be deformed during the second pass, and there was no
room to alter conditions should this not work. To bypass this, a hot plate was added to
preheat the material, as shown in Figure 2-18, with a surface temperature of 70 C, just
below the Tg of the material. A metal cover sheet helps bridge the gap between the
hotplate and rollers; 0.016" brass sheet was used for the data collected and analyzed
Figure 2-18. Laminator with preheating hotplate.
188.8.131.52 Sandwich material
The silicone rollers on the laminator tended to impart a leather-like texture into the
device; therefore, a material was needed to sandwich the chip between for the lamination
process. Metallized Mylar was found to yield the best bonding results of materials tested
to date, but has a tendency to release metallic material onto the device, which is visually
obtrusive, especially when using laser light for detection (this may be caused by the fact
that Mylar is polyethylene terephthalate, which has a Tg of approximately 70 C; as a
result, the material is in it's rubbery state during the lamination process, which may limit
it's ability to hold the metallic coating). While the metallic material can be removed by
scrubbing with an acetone-soaked swab or tissue, the chips can be scratched in the
process, leaving the desire for another solution. Other materials that were unsuccessfully
tried were brass sheet, aluminum sheets and foil, paper, and overhead transparencies; all
imparted undesired textures into the devices or did not effectively transfer heat transfer
during passage through the rollers.
184.108.40.206 Roller pressure
The laminator has a total of six roller spacing settings; two meant strictly for paper
or other thin objects, and four meant to accommodate thicker materials. These last four
settings are labeled as 1/32" (0.79 mm), 1/16" (1.59 mm), 1/8" (3.18 mm), and 3/16"
(4.76 mm), and indicate the thickness of the material to be sealed; as the material needs to
be drawn through the laminator, the rollers still touch each other for these settings. The
actual pressure the laminator is able to convey to the device at each of these settings was
not measured, and in any case would vary with device thickness and width.
For the chips used in the optimization of lamination, which were approximately 1.5
mm in thickness, acceptable results were usually received using the 1/8" setting;
narrower settings would occasionally result in sealed channel ends, as shown in Figure 2-
19, while the widest setting did not always provide enough pressure. However, given the
limited options in roller spacing available, it may be advisable to come up with a system
that has finer adjustability.
Figure 2-19. Comparison of devices laminated at different roller settings. A) 1/16". B)
220.127.116.11 Roller temperature
The most likely method of failure of the bond between film and device is peel. In
theory, the entire area of film could debond instantaneously, but that is rather unlikely. A
peel test was developed using the CNC mill to pull the film at a constant rate of 300
mm/min, negating any dynamic effects that may be imparted by using a falling weight to
pull the film. The results recorded from the strain gage/Wheatstone bridge are shown in
Figure 2-20 grouped by roller temperature. Ideally, the recorded values would remain
constant along the length of the device; however, a large amount of temporal variation is
The variation itself is likely due to a combination of irregularities along the length
of the device and, in the case of the higher roller temperatures, failure in the tape used to
secure the devices to the load cell. Should higher lamination temperatures or more secure
methods of bonding be tested, a more secure method of attaching the sample to the load
cell should be identified.
0.010 / \ __ 105C
0.004 --' 1 .---- .
0.002 '" --
0.0 2.0 4.0 6.0 8.0 10.0 12.0
Figure 2-20. Peel test results grouped by roller temperature. The rectangle indicates the
range of data averaged to assemble Figure 2-21.
The peel strength for each roller temperature was determined by averaging the first
two seconds of the averages reported in Figure 2-20, as indicated by the rectangle; these
averages and their standard deviations are reported in Figure 2-21. There is a clear
differentiation between the first three temperatures tested, though not between 115 and
105 C 110 C 115 C 120 C
Figure 2-21. Average value of the average in the first two seconds of the test (points
inside rectangle in Figure 2-20).
Not apparent in these results are the actual effects of lamination process on the
devices. The roller temperature of 120 oC caused some bulk deformation (bending) of the
device, indicating that the roller and/or preheat temperatures may be too high.
In addition, channels were sometimes blocked off beyond where burrs that resulted
from the milling process ended, as shown in Figure 2-22. While these blockages are
partially attributed to burr formation during the milling process since they never occur in
the middle of the channels, the easiest solution has been to reduce the preheat
temperature to 65 oC and replacing the brass with 0.125"-thick aluminum, which was
found to have less temperature variability across the surface. As the hotplate temperature
controller adjusts only in 5 oC increments, more precise control of temperatures is not
possible using the current setup. Neither the penalties in lamination strength nor the
potential requirement of increasing roller temperature have been studied.
Figure 2-22. Blockage that extends beyond burr location sometimes observed after
In a full-production setting, the molds would probably be designed to create the
wells during the molding process to eliminate this step. However, in a laboratory
creating a relatively small number of prototypes, using a CNC mill to create the access
holes/reservoirs in the device makes sense as it will give a high degree of repeatability
and consistency from device to device and within each device. Tooling paths can be
created by writing a program from scratch or importing a pattern designed using a CAD
programs, and are generally trouble-free once optimized.
The biggest problem encountered is the formation of burrs which end up blocking
the channels. Since the channels being considered only have a depth of about 30 |tm,
even the smallest of burrs could create major problems in the usability of the device.
To determine which, if any, milling conditions resulted in the smallest burrs,
spindle speeds were investigated over the range of 750 and 2000 rpm in increments of
250, while feed rates were investigate from 25 to 200 mm/min in increments of 25.
Faster spindle speeds are capable of working at higher feed rates, so the spindle speeds of
1750 and 2000 were also tested at feed rates up to 300 mm/min.
During the cutting process, the channels can either be facing the cutting tool (up) or
the table (down). Both methods in theory have their advantages and drawbacks: having
the channels down minimizes the time in contact with the cutting tool, but risks "blow-
out" that can occur with the tool coming through the surface; having the channels up
eliminates blow-out, but maximizes the time spent in contact with the cutting tool
(resulting in heat production which could lead to melting), and also requires that all
material removed during the cut pass the channels, where they have the opportunity to
weld back on and cause a blockage. Removal of debris can be enhanced by blowing a
fluid across the cutting surface. Air is the cleanest option, and was used in this case.
Results indicate that using an end mill to drill through-holes appears to work better
when the channels are facing up, apparently due to the blow-out issue described earlier.
Horizontal milling showed the opposite behavior: having the channels up appears to lead
to greater burr production than having the channels down. Examples of these behaviors
are shown in Figure 2-23.
This raises the possibility that if holes and slots must be milled into a device, either
the tool paths have to be written to include only one type of cut, or the drilling of holes
and milling of slots may require two separate steps. In addition, it is observed that the
climb side (left side of slots) exhibits heavier burr formation than the standard side; since
climb milling ejects material into the area behind the mill, this probably results from the
extruded material welding back on to the cut areas. Should slots need to be cut in devices
to access many channels, the cut should be made in two passes so that both sides receive
standard milling. In any case, horizontal milling may prove problematic as the channels
provide exhaust paths for extruded plastic.
Burrs were never completely eliminated for either the drilling or milling operations,
though they were minimized in the orientations discussed above. In general, milling
results tend to look the same in the optical microscope once the feed rate is above 125
mm/min regardless of feed rate; little difference is noted in the drilling results across the
range of spindle speeds and feed rates tested. Use of air to remove debris also helped
enormously, though little difference was noted between 10 and 20 psi. The extent to
which burrs block the channel-reservoir interface is unknown, but solution was able to
enter the channels of the laminated devices.
A / B
Figure 2-23. Post-lamination from drilling and milling using a 2 mm-diameter end mill.
Air pressure was 10 psi, spindle speed was 1750 rpm, and feed rates were
100 mm/min in all cases. A & B): channels up (facing mill) C&D)
channels down (facing table). The left sides of the slots were climb milled,
while the right sides were standard milled.
Burr removal was attempted by scraping the inner rim of the wells with a razor
blade after milling, but before the pre-lamination cleaning; a representative result is
shown in Figure 2-24 A. While this method greatly increases the chances of getting a
usable device, it is inconvenient to have to scrape each hole.
In addition, this method also can lead to a different usage issue, the formation of
bubbles at the ends of the channels. The method our lab has selected to fill the channels
prior to isoelectric focusing to add liquid that is to fill the channel to one well, use a
vacuum pump to draw the liquid through the channel to a second well, and then fill the
wells with the anolyte or catholyte. However, when excess liquid is removed prior to
addition of the appropriate electrolyte, the meniscus has a tendency to retreat into the
scraped area, as shown in Figure 2-24 B. If the second well was then filled improperly, a
bubble would form in the scraped area, preventing the circuit required for electrophoresis
and electroosmotic flow.
A bubble B
Figure 2-24. Results from using a razor blade for burr removal. A) picture of well edge
after scraping. B) profile of well after scraping and filling
2.3.4 Actual Usage Protein Separations
Green Fluorescent Protein (GFP) and R-Phycoerythrin (RPE) were selected for the
separation in a pH gradient gel; their natural fluorescence allows them to be tracked by
optical detection under proper illumination.
Figure 2-25 shows the separations that were achieved using a constant 500 V, but
different channel lengths. This in turn results in different electric fields, with the shorter
channels getting stronger fields. The results show that the separations of these two
proteins can be achieved on shorter channels, though some of the finer detail may be lost
due to compression of the peaks.
4.0 -32 mm
3.5 43 mm
0 500 1000 1500 2000 2500
2.4 Conclusion 2.5
0 500 1000 1500 2000 2500
Figure 2-25. IEF for different channel lengths under a constant 500 V. Channel lengths
were 2.1, 3.2, 4.3, and 5.4 cm, resulting in fields of 238, 156, 116, and 93
There are a number of variables and conditions that must be optimized to
consistently produce usable microfluidic devices. Overall device thickness and crazing
are temperature dependent, thought the crazing can be minimized by employing proper
cooling conditions. For the silicon mold, a temperature of 115 C was found to be the
best compromise, while the glass-based e-form required 125 C due to imperfections in
the mold. There was no dependence on closure rate or pressure under the conditions
tested, therefore a closure rate of 10 seconds per downstroke and force of 5 kip are
recommended as these allow the fastest molding of devices under this method.
Milling was used to define the wells and edges of the devices used in this study.
Best results when drilling with an end mill are achieved when the channels are facing the
cutting head and air is used to remove debris; beyond that, performance appears to be
independent of feed rate and spindle speed. Defining slots is best done with standard
feeding, the channels against the table, and feed rates above 125 mm/min.
Although best lamination results have been obtained using 2 mil metallized
Mylar, it is recommended that another material be found to prevent the deposition of
metallic flakes. For Topas 8007, the preheat surface should have a temperature in the
range of 60-65 C and roller temperature should be between 110 C and 115 C.
All of these results pertain solely to the material and patterns used in this
investigation; different materials, such as acrylics, polycarbonate, and other COCs, will
have different thermal and mechanical properties, thus requiring slightly different
processing conditions. Other device or film thicknesses may need to be substituted,
which would alter the lamination methods one would employ to get a working devices.
Nevertheless, a methodology for characterizing the multitude of process variables has
been presented, and these results should provide a starting point for optimization of
microfluidic devices created using a compression molding process.
DESIGN, FABRICATION, AND TESTING OF A DEVICE FOR PROTEIN
The successful expression of proteins in a laboratory setting is vital to such
fundamental research as determining a protein's structure or its biological purpose, and
can also be applied to drug discovery and toxin identification. There are two main
applications for which proteins are expressed in vitro: screening and production.
Screening experiments require quick expression of proteins in arrays to verify that a
process does or does not work, while production refers to the growth of proteins in a
chamber for extraction and use in another experiment. Due to limitations discussed
below, it is desirable that a device be created that applies production-type technology to
The original intent was to apply the results from the previous section and use the
same materials and processes to create this device. However, due to the unique
configuration and requirements, adoption of different materials and methods proved
easier to implement.
3.2 Synthesis of Proteins
Proteins are essentially long strands of amino acids. They are produced
biologically through two consecutive processes, transcription and translation. In the
* A part of this chapter has been published with other materials in : Mei, Q., Fredrickson, C., Jin, S.,
and Fan, Z.H., 2005, "Toxin detection by miniaturized in vitro protein expression array," Anal. Chem., 77,
transcription phase, ribonucleic acid (RNA) is copied from selected regions of
deoxyribonucleic acid (DNA) . Translation is the formation of proteins via
replication of certain regions of the RNA, and occurs in structures called ribosomes.
As mentioned earlier, screening studies are typically done with small quantities of
material where the need exists to run many experiments very quickly. For example,
Roche's RTS 100 system is for screening, and advertises up to 20 |tg of protein produced
per 50 [l of solution (equivalent to 0.4 mg/ml) in 2-4 hours [58:91-92]. Experimental
systems with volumetric capacities of nanoliters  and picoliters  have also been
reported. While having the advantage of being fast, they also have low yields, limiting
the number of proteins that can be expressed. Production is accomplished using larger
quantities that take longer to express, but with quantities much too large to run cost-
effective assays: Roche's RTS 500 system advertises a much more productive 5 mg per 1
ml of solution in 24 hours [58:119].
This variance in yield is reliant on the environment in which the proteins are
grown: in closed environments such as microcentrifuge tubes and well plates, nutrients
needed to continue the reaction are slowly depleted and toxic byproducts of the reaction
build up; as a result, 2-4 hours is the maximum period of time experiment an experiment
can endure. In vitro protein production, on the other hand, uses two chambers that are
separated by a semi-permeable membrane: a reaction chamber containing DNA
templates, ribosomes, and an initial concentration of nutrients, and a feeding chamber
that houses additional nutrients; exchange of nutrients and waste products across the
membrane prolongs experiment duration and increases yield. This technology has been
licensed for use in the United States by Roche [58:12, 61].
Creating a fluid exchange mechanism in a well plate-sized chamber should improve
the protein yield without significantly increasing reagent consumption, allowing the use
of systems that are difficult to express the microarray format, such as P-glucuronidase
(GUS), for assays, and fabricating such a device is the goal of this chapter.
After examining several designs, we settled upon one having two wells, one nested
inside the other. The contents of each chamber would be separated by a permeable
membrane, and several chambers would be linked together by a common flange. An
illustration and cross section are shown in Figure 3-1. The smaller, inner chamber would
contain the DNA vectors, RNA, proteins, and an initial charge of nutrients, and be
referred to as the reaction chamber. The larger, outer chamber would contain an
additional supply of nutrients, and serve as a sink to which byproducts would migrate.
To create the devices, the upper and lower pieces would have to be fabricated, and then a
suitable adhesive identified to bond the dialysis membrane to the upper piece to create the
t[1 insert or tray
insert or tray reaction chamber
1) i membrane---
well plate feeding chamber
Figure 3-1. Illustrations of well-device. A) Three dimensional B) cross sectional.
Three brands of cyanoacrylate adhesives were initially considered: Krazy Glue
(Elmer's Products, Columbus, OH), Super Glue (Loctite, Avon, OH), and Vetbond (3M
Health Care, St. Paul, MN). Samples of each adhesive were applied to a glass slide and
allowed to dry. When these were unable to consistently survive the autoclaving process,
two types of epoxy, 302-3M and 353ND-T (both of Epoxy Technologies, Billerica, MA),
were acquired, mixed and cured per the manufacturer's instructions.
Polymerase chain reaction (PCR) was used to test the biocompatibility of these
adhesives: PCR Master Mix (M7502, Promega, Madison, WI), a testing kit based on Taq
DNA Polymerase, was reconstituted per the manufacturer's recommendations. A small
piece of each dried adhesive was placed in a microcentrifuge tube, covered with 50 [tl of
the prepared solution, and run through 20 PCR cycles in a PTC-100 programmable
thermal controller (MJ Research, Waltham, MA). The PCR products were examined by
gel electrophoresis: Agarose (Fisher Scientific, Atlanta, GA) was dissolved in tris-boric
acid-EDTA (TBE) buffer solution and cast into a gel slab. When solidified, it was
immersed in TBE buffer, and a small amount of ethidium bromide dye was added. The
PCR products were added to wells in the slab, which was then placed in the separation
cell and voltage applied. When separation was completed, the gel was removed from the
cell, de-stained and rinsed to remove excess ethidium bromide, and imaged using a
3.3.2 Device Fabrication
As shown in Figure 3-1, the finalized design consisted of two parts. The upper
part, which will be referred to as the insert, was fabricated by milling through holes and a
common flange in 0.100" (2.5 mm) thick acrylic (Lucite-ES, Lucite International, Inc.,
Cordova, TN) using a CNC-mill (Flashcut 2100, Flashcut CNC, Menlo Park, CA).
Dialysis membrane (Spectra/Por 1, Spectrum Labs, Rancho Dominguez, CA) was cut
into small circles using a #5-tipped (4.37 mm diameter) leather punch (#3003-00, Tandy
Leather, Ft. Worth, TX,). 353ND-T epoxy and curing agent was mixed in a 10:1 ratio by
mass as recommended by the manufacturer on a glass plate. A doctor blade was created
by wrapping a piece of wire around each of a razor blade and used to create an epoxy
film of constant thickness (estimated to be 0.2 mm). Each insert was pressed into the
film and then removed, thus transferring a thin layer of epoxy on the bottoms of the
inserts. The precut membrane pieces were then individually applied to each well of the
insert, and the assembled inserts were cured for 20 minutes at 95 oC in an oven.
The bottom parts, referred to as well plates, were created by milling 4 mm deep
wells into a piece of 6.35 mm (0.25") thick acrylic (Lucite-ES). Both well plates and
completed inserts were sterilized by exposing to UV light for 30 minutes.
The insert chambers are 3 mm in diameter, and are surrounded by a 1 mm thick
wall, creating a structure with a net diameter of 5 mm. The well plate chambers are 7
mm in diameter. All well centers are spaced 9 mm apart.
3.3.3 Protein Expression
Chloroamphenicol acetyl-transferase (CAT) was expressed using an RTS 500 kit
(Roche Diagnostics GmbH, Mannheim, Germany). Reaction and feeding solutions were
prepared as recommended by the manufacturer. 7 [tl of reaction solution were added to
each of six chambers in an insert and five microcentrifuge tubes, while 70 [tl of feeding
solution was added to the chambers in the well plate (feeding solution was not added to
the microcentrifuge tubes so that an E. coli-based RTS 100 kit could be simulated). The
tray was covered with PCR tape (Costar 6524, Corning Inc., Corning, NY) and caps
closed on the PCR tubes to prevent evaporation. The tray and tubes were placed on a
shaker at room temperature, with samples being taken at after 2, 6, 10, 14, 20, and 24
hours until samples were depleted. The samples were separated via sodium
dodecylsulfate-polyacrylmide gel electrophoresis (SDS-PAGE), and then analyzed using
western blotting. To summarize the blotting process, polyvinyl difluoride (PVDF) film
was prepared by immersing it in methanol and then transfer buffer solution. The gel and
film were sandwiched between buffer-soaked pads, and voltage applied to transfer the
proteins from the gel to the PVDF film. After immersing the film in dilute reconstituted
milk to block non-targeted active sites, it was soaked in a solution containing a primary
antibody for about an hour, and then rinsed to remove excess antibody. Secondary and
labeling antibodies were applied in the same way. The film was developed by immersing
it in an NBT/BCIP (NBT: nitro-blue tetrazolium chloride; BCIP: 5-bromo-4-chloro-3'-
indolyphosphate p-toluidine salt) solution. It was scanned and the signals from the stains
were measured by ImageJ software.
3.4 Results and Discussion
3.4.1 Device Designs
The initial approach to tackle the problem was to build a device having three wells
connected by a single channel, as depicted in Figure 3-2. Proteins would be grown in the
center well, which would be separated from the channel by a dialysis membrane. The
outer two wells would contain solution with nutrients.
The centers between wells in each three-well set were separated by 4.5 mm, while
the spacing between centers of sets was set to 9 mm, in order to comply with the Society
for Biomolecular Screening's standards for 384- and 96-well plates , respectively,
which in turn would allow finalized arrays to be analyzed using existing microplate
Buffer A B
Figure 3-2. Original concept for device. A) Conceptual sketch. Buffer solution in each
of the outer wells would be driven through the channel via electroosmotic
flow (arrow). B) Constructed prototype.
Though the channel-membrane interface area should be maximized to provide the
largest area over which diffusion could occur, we decided to minimize the channel
dimensions to both reduce the amount of solution used and enable the use of
electroosmotic flow to drive the convective process. Consistently milling channels of
adequate smallness proved to be difficult, and this approach was abandoned.
The second iteration was to create small reaction chambers that would nest inside
larger feeding chambers. Figure 3-3 shows the evolution of this design philosophy.
Figure 3-3. Evolution of dual chamber design. From left to right, donut, flanged donut,
final array of feeding chambers, and fabricated device ready for use. Penny
is included for size reference.
The first iteration attempted was a simple donut-shaped device. Though the
original intention was to place one of these donuts into each well of a standard 96-well
plate, the donuts in theory could also be placed in a large common bath containing
feeding solution. However, the stability of these devices was less than desirable; the
second version added a flange which could rest on a lip. Extension of this flange to unite
several devices led to the final device configuration.
3.4.2 Adhesive Biocompatibility
In order for the device to be successful, the materials used to make the device had
to meet a number of requirements. First, they had to be able to endure aqueous
environments for prolonged periods of time. Second, they had to survive the sterilization
process (originally intended to be autoclaving, though exposure to UV radiation was later
used). Finally, as the intended use of the device is to express proteins, it is important that
they be biocompatible. Acrylic is used in contact lenses and bone cement , so its
biocompatibility was assumed, along with that of the dialysis membrane. However, the
biocompatibility of adhesives is less documented, necessitating our own testing.
PCR is a method of quickly replicating DNA segments, and was selected to
measure biocompatibility because the mechanism of DNA amplification closely
replicates the transcription phase of protein expression. Much of the PCR process is also
automated, making it easier to test than actually expressing the proteins. Figure 3-4
shows the resulting separation gels from running PCR in the presence of cured samples of
adhesive. The first column is a marker lane, and serves as a ruler to calibrate the
molecular size of any products that appear; the number of base pairs in each band appears
to the left of each picture. The other lanes contain either a control sample, which had no
adhesive in the reaction chamber, or one of the adhesive-containing samples, as labeled.
The brightness of the signal depends on the amount of PCR product present. As a
result, a dimmer signal indicated that the glue inhibits the reaction more, indicating that it
should be avoided for device fabrication (in order to maximize the protein yields, it is
imperative that the device itself has minimal, if any, negative effect on the transcription
or translation process). It was found that all adhesives tested exhibited some amount of
inhibition to the PCR results.
DNA Positive Krazy Vetbond Super DNA 302-3M 353ND-T Positive
Glue Glue A Marker B
Figure 3-4. PCR results for adhesives considered in this study. A) cyanoacrylates. B)
Of the three cyanoacrylate glues tested, Vetbond inhibited the reaction the least,
indicating that it would be the best of those choices. Unfortunately, none of these
adhesives survived the autoclaving process (used for sterilization at the time), and the low
viscosity of the adhesive also made application of consistent amounts of glue
Two types of epoxies were examined as replacement adhesives. 302-3M has
reportedly met U.S. Pharmacopeia (USP) class VI standards for biological compatibility;
353ND-T has not been met those standards, but a related system, 353ND, has .
Neither of these adhesives completely inhibited the reaction, as shown in Figure 3-4 B.
That the signal is reduced by about the same amount in each sample suggests that
inhibition may have resulted from a systemic factor, such as the solid piece of adhesive
limiting the circulation and mixing during the PCR process. We proceeded with 353ND-
T due to its viscous nature, which allowed the formation of a better film.
One design issue of note is that despite our doubling of the manufacturer's cure
recommendation for that temperature, the epoxy doesn't reach full cure until after
sterilization. Figure 3-5 shows devices at various stages of the fabrication process; #1
has been cured in an oven, while #2 has been UV sterilized. Note the color change: the
epoxy starts out yellow or light brown, but turns to a dark amber color; this color is an
indication of level of cure.
Figure 3-5. Demonstration of incomplete curing of epoxy during insert fabrication. 1)
shows results after initial cure and 2) shows results after UV sterilization. 3)
Bulk epoxy that underwent same cure conditions as 1), but no sterilization.
The fact that #1 is not fully cured is believed to be a result of the quantity of epoxy
being too small: epoxy cures via an exothermic reaction, and harvests energy both from
the external environment and the heat generated by the reaction itself; not having enough
epoxy means that more energy must be drawn from the environment, requiring warmer
temperatures and/or longer times in the oven. This was verified by curing a more
significant amount of epoxy using the same conditions as #1; the resulting product is
3.4.3 Protein Expression
The primary goal of this system was to exhibit higher protein yields than were
typical for existing, closed environment systems. Figure 3-6 shows the results from the
quantification of the western blot tests run to measure the amount of protein expressed in
each system. It is observed the signal intensities are similar after two hours, but where
the microcentrifuge tube's protein content remains fairly constant, samples from the tray
show significant growth over time, eventually exhibiting signal intensity approximately
an order of magnitude stronger than the conventional microcentrifuge tubes. This shows
that the device does indeed produce higher yields than the closed environments currently
used for microarrays.
15000 A Microcentrifuge tube
0 A A A A A
0 4 8 12 16 20 24 28
Incubation time (h)
Figure 3-6. Comparison of expression of CAT between microcentrifuge tube and device.
A small-scale, dual chamber device for protein expression has been developed by
milling a reaction chamber from acrylic, using epoxy to attach a dialysis membrane, and
then submerging it in a feeding chamber. This system has been shown to produce signals
approximately an order of magnitude stronger than the conventional means of closed
reaction vessels (microcentrifuge tubes or conventional well plates).
The system presented uses a total of 77 [tl of solution; while this compares
favorably with the Roche 500 system which consumes 1 ml, is still at least two orders of
magnitude larger than the single chambered nano- and picoliter systems reported by
others [59,60]. Though it only has six wells in its current form, well size and spacing
conforms to existing 96-well plate formats, allowing arrays of up to that size with
minimal additional effort. While further reduction in the volume may be possible (i.e. to
384- or 1536-well plate standards), liquid migration by traveling along the wetted out
surfaces of the tray and well plate will probably become a limiting factor of this design.
CONCLUSIONS AND FUTURE RECOMMENDATIONS
The applications and needs for microfluidic lab-on-a-chip technology are
increasingly growing. This thesis has developed and characterized two fabrication
methodologies: a compression molding process that can be applied to any laboratory-
scale exploratory work of new designs, and a machining process used to create a device
for the specific application of protein expression. Presented here the overall conclusions,
as well as recommendations for aspects that may deserve further study.
4.1 Compression Molding Process Optimization
A compression molding process for fabricating thermoplastic microfluidic devices
has been characterized. The total process consists of molding the device, using a mill to
define reservoirs and edges, and laminating with film of identical material to seal the
During the molding process, replication of bulk features (ridges/channels) appears
to be independent of molding conditions over the range that was tested, but device
thickness and degree to which crazing occurred are not. While device thickness can also
be modified by choosing different spacers, crazing is best minimized by reducing the
molding temperature and cooling the device in an oven to a temperature just below the
Tg (60 C oven for 80 C Tg in this case). Careful fabrication of the master to eliminate
surface roughness would also probably help minimize this behavior. One aspect that
wasn't discussed is the amount of waste that occurs in the compression molding process
being used. While it takes six to eight grams of material to mold an acceptable device,
the milling of edges and holes results in a final device containing approximately three
grams of material. That means for every device created, at least half of the material will
be wasted, assuming it isn't being recycled. One possible solution is to modify the mold
to create a set of walls around the pattern that limit the flow/expansion of the material
being formed. Another would be to convert to an injection molding machine similar to
the method used by the Ahn group .
In milling operations, burr formation was never eliminated, but could be reduced to
the point that functional devices were obtained. Drilling of hole-like reservoirs using an
end mill is best done with the channels "up" (facing the end mill); there appears to be
little difference in results for the spindle speeds and feed rates tested. Milling channel-
like reservoirs worked better when the channels were "down" (facing the table); spindle
speed and feed rates should be high, and climb milling should be avoided. In all cases,
active removal of debris should be employed; air was used in this case.
Sealing the channels was completed by using a commercial laminator. Suitable
results were acquired by using a hotplate to preheat the material to -10-15 C below the
Tg of the material. The rollers had a tendency to impart a texture in the material, which
was best alleviated by sandwiching the device and sealing film between two pieces of
metallized Mylar. Depositions of small amounts of metal from the Mylar could not be
eliminated, and while removal is possible using an acetone-soaked swab or tissue, this is
not recommended. Use of non-metallized Mylar of identical thickness or vacuum
bagging or release films  used in the composite industry may provide suitable
replacements. The strength of the bond between film and device was tested by measuring
the force required to peel the film off, and higher roller temperatures resulted in higher
Sagging of the film into the channels during the lamination process was not
investigated. This behavior could significantly decrease the volume of the channels. The
degree to which this occurs is probably dependent on the preheat temperature and roller
temperature and pressure. Investigation of this should probably proceed in a
combinatorial fashion by varying all three of those process parameters rather than, for
example, first finding the best preheat temperature, and then optimizing the roller
temperature and pressure using only that optimized value, due to the inter-relationship of
4.2 Device for Protein Expression
A two-part, "open system" tray device has also been developed for the purpose of
protein expression. The upper part, or insert, holds the reaction chambers for protein
synthesis; each well has a dialysis membrane on the bottom to retain the DNA vectors,
RNA, and produced proteins. The lower part holds the feeding solution, and simply
consists of wells; though we milled our own, the alignment of the insert is such that 96-
well plates could be substituted. This allows the use of commercially-available well-
plate readers to measure the protein expression yield directly in the device should a
suitable labeling system be developed. Well sizes and alignments corresponding to 384-
and 1536-well plates could also theoretically be developed, though issues related to
wetting out of insert and well plate material surfaces will probably serve as limiting
The presented device was shown to exhibit significantly greater levels of protein
production than the closed environments of microcentrifuge tubes. The speeds of
expression were comparable in both the open and closed systems; further research should
be devoted to speeding up this process, perhaps through adjusting the frequency and
displacement of shaking or using an oscillating pressure to more actively circulate the
If this general design is kept, future iterations should have pins and corresponding
sockets to assist in the alignment of the two pieces. Use of an adhesive to bond the
reaction chamber insert to the feeding chamber well plate to prevent cross contamination
of wells during normal handling of the device should also be considered.
4.3 Overall Conclusions
The successful fabrication of microfluidic devices is complicated due to the number
of variables present and small channel dimensions, which correspondingly have small
tolerances for error. While the settings for each variable will change with the particular
materials and tools used to complete the task, the fabrication and characterization
methods should carry over, and those present here will should assist the researcher
develop new and better devices.
CONSTRUCTION AND CALIBRATION OF LOAD CELL USED FOR PEEL TESTS
A load cell was required to measure the bond strength between the film and bottom
plate of the completed device. The anticipated use of the load cell apparatus was to
mount it in the cutting tool holder of a CNC mill, and then use that automation to control
the pull rate and direction. Commercial load cells and testing devices were not
considered for this work due to cost. An internet search turned up the design by Richard
Nakka that proved to be suitable .
The basic design for the load cell is shown in Figure A-1. It consists of a block of
material with a keyhole cutting through one of the sides. As the load is applied over the
notch, the intact edge elongates. The ratio of change in length over original length is
called strain, ga, and can be measured by strain gages.
Figure A-1. Illustration of load cell.
Strain gages are essentially wires that are bonded to a surface. As the wire
stretches, its resistance changes by the relationship 
dR = ESR (1)
where dR is the change in resistance, S is the strain gage factor (constant and unique to
wire material) and R is the original resistance of the strain gage. The change in resistance
dR is typically very small, and is best measured via a Wheatstone bridge, which consists
of three static resistors, one variable resistor that is sensitive to the property being
measured (strain gage), a constant voltage power supply, and a voltmeter. A diagram for
this circuit is provided in Figure A-2.
R4 ~ R3
Figure A-2. Circuit diagram for Wheatstone bridge used to collect data.
Ideally, Vo would equal zero at the start of the test, which in turn requires that the
ratios of R1/R4 and R2/R3, be equal; where the subscript i indicates that the initial value of
the strain gage resistance is being used. When this is true, the equation that relates the
voltage measured across the bridge, Vo, and change in resistance is
V, = ( R)2 AR3 (2)
(R2 + R32
(Several primers on this circuit have been written, so a development of this equation will
not be included). Approximating AR3 as dR from the strain gage equation gives
V, = V, R2R32 S (3)
(R2 + R3 2
As strain sa cannot be controlled and gage factor S is constant, the most sensitivity
is achieved when the first two terms are maximized; V, is typically limited by the strain
gage and power supply, while the resistance term reaches a maximum value of 0.25 when
R2 equals R3,.
Mr. Nakka has conveniently provided a spreadsheet  which will give the
expected microstrain measurement after inputting the expected force, dimensions, and
modulus of elasticity. This was used to determine the size of our cell; the dimensions
that were used are shown in Figure A-3.
Figure A-3. Dimensions used for load cell in millimeters.
The block was cut from a piece of 0.25"-thick acrylic using a CNC mill. A T-
shaped piece with all legs approximately 0.25" wide was cut from 0.025"-thick copper
sheet and secured to the side opposite the strain gage by a screw and epoxy, as shown in
Figure A-4 A. An end mill was inserted into a hole in the top of the block so that the
completed load cell could be attached to the mill's cutting head. The strain gage (cea-06-
250un-350, Vishay Micro Measurements) and 30-gauge magnet wire were attached using
epoxy; the wire was soldered to the leads of the strain gage after the epoxy had set.
To form the device holder, a 2.5" x 4" piece was cut from melamine-covered
particleboard. Holes were drilled to so that screws could secure it to the mill's tooling
plate as shown in Figure A-4 B. Broken end-mills were epoxied to the wood so that they
cantilevered out with 1/8" gap between them; a piece of 0.100"-thick acrylic was used to
support them while the epoxy set.
Figure A-4. Completed load cell. A) side view. B) shows the load cell and device
holder ready for use.
A Wheatstone bridge was assembled on a solderless breadboard. The initial
resistance of the strain gage was 350 Q; 350 Q resistors were made for the other three
legs by linking three smaller resistors in series. Supply voltage was provided by an
Agilent 3611 power supply operating at 5.0 volts. Measurements were picked up by a
digital multimeter (Model 72-6870, Tenma) interfaced with a personal computer running
Data Logger (Tenma) software.
For an actual peel test, the holder is attached to the tooling plate of the mill. The
load cell is secured in the collet that normally holds the cutting tool via the post sticking
out the top. The cell is situated so that the copper tab sits parallel to and between the two
cantilevered arms. A razor blade is used to start the delamination process, creating a flap
of film between 14" and 12" long. The device is placed underneath the arms with the flap
coming up from between them; the flap is secured to the copper tab with tape (although
tape is not the ideal method of securing, it provides the advantages of being easy to
implement and not taking up much space; in addition to space constraints, clamps of the
size needed here are likely to focus the forces on small areas of the film, potentially
leading to tearing). The load cell is then moved straight up at the desired rate, pulling the
peeling film with it.
A cup was clipped to the copper tab, and water was added 5 ml at a time. Voltage
drop across the bridge was recorded. A calibration curve relating voltage drop across the
bridge and force was created assuming the density of water was 1 g/ml; this is shown
below in Figure A-5.
4.5 y = 1.1043x + 2.399
R2 = 0.9962
0 0.5 1 1.5 2
Figure A-5. Calibration curve obtained for load cell.
LIST OF REFERENCES
1. Curie, P., Curie, Mme. P., and Bemont, G., 1898, "On a new, strongly radio-active
substance contained in pitchblende," Comptes rendus de l'Academie des Sciences,
127, pp. 1215-1217; translation from Romer, A., ed., 1970, Radiochemistry and the
Discovery of sotopes, Dover, New York.
2. Curie, Mme., 1904, "Radium and radioactivity," Century Magazine, pp. 461-466.
3. Rhodes, R., 1986, The Making of the Atomic Bomb, Simon and Schuster, New
York, pp. 409-410.
4. Borman, S., 1998, "A brief history of mass spectroscopy instrumentation,"
accessed 4 November 2005.
5. Borman, S., Dagani, R., Rawls, R.L., and Zurer, P.S., 12 January 1998, "Chemistry
crystallizes into modem science," Chem. & Eng. News, pp. 39-75.
6. Links to miniature mass spectrometer and mobile GC/MS,
http://www.gcms.de/index.html, last accessed 4 November 2005.
7. NSF workshop, Nov. 3-4, 2003, "Prospects for Miniaturization of Mass
Spectrometry," Executive Summary, Arlington VA, http://www.nsf-mass-spec-
mini-forum.umd.edu/finalreport.html, last accessed 4 November 2005.
8. Harris, D.C., 1987, Quantitative ChemicalAnalysis, 2nd Ed., W.H. Freeman and
Company, New York.
9. Oxtoby, D.W., Gillis, H.P., and Nachtrieb, N.H., 1999, Principles of Modern
Chemistry, 4th Ed., Saunders College Publishing, Fort Worth, pp. 366-368.
10. Baker, D.R., 1995, Capillary Electrophoresis, John Wiley & Sons, New York.
11. Terry, S. C., Herman, J. H., and Angell, J. B., 1979, "A gas chromatograph air
analyzer fabricated on a silicon wafer," IEEE Transactions on Electronic Devices,
ED-26(12), pp. 1880-1886.
12. Manz, A., Harrison, D.J., Verpoorte, E., Widmer, H.M., 1993, "Planar chips
technology for miniaturization of separation systems: a developing perspective in
chemical monitoring," Adv. Chrom., 33, pp. 1-66.
13. Manz, A., Graber, N., Widmer, H.M., 1990, "Miniaturized total chemical analysis
systems: a novel concept for chemical sensing," Sens. Act. B, 1, pp. 244-248.
14. Harrison, D.J., Fluri, K., Seiler, K., Fan, Z., Effenhauser, C.S. and Manz, A., 1993,
"Micromachining a miniaturized capillary electrophoresis-based chemical analysis
system on a chip," Science, 261, pp. 895-897.
15. Harrison, D.J., Manz, A., Fan, Z., Ludi, H., and Widmer, H.M., 1992, "Capillary
electrophoresis and sample injection systems integrated on a planar glass chip,"
Anal. Chem., 64(17), pp. 1926-1932.
16. Verpoorte, E., 2002, "Microfluidic chips for clinical and forensic analysis,"
Electrophoresis, 23, pp. 677-712.
17. Reyes, D.R., lossifidis, D., Auroux, P.-A., and Manz, A., 2002, "Micro total
analysis systems. 1. Introduction, theory, and technology," Anal. Chem., 74, pp.
18. Auroux, P.-A., lossifidis, D., Reyes, D.R. and Manz, A., 2002, "Micro total
analysis systems. 2. Analytical standard operations and applications," Anal.
Chem., 2002, 74, pp. 2637-2652.
19. Vilkner, T., Janasek, D. and Manz, A., 2004, "Micro total analysis systems. Recent
developments," Anal. Chem, 76, pp. 3373-3386.
20. Teichroeb, R., 22 July 2004, "Rare look inside state crime labs reveals recurring
DNA test problems." Seattle Post-Intelligencer,
http://seattlepi.nwsource.com/local/183007_crimelab22.html. Last accessed 4
21. Gray, B. L., Jaeggi, D., Mourlas, N. J., van Drieenhuizen, B. P., Williams, K. R.,
Maluf, N. I., and Kovacs, G. T. A., 1999, "Novel integration technologies for
integrated microfluidic systems," Sens. Act. A, 77, pp. 57-65.
22. Madou, M., 2002, Fundamentals ofMicrofabrication, 2nd Ed., CRC Press, Boca
23. Chen, H., Acharya, D., Gajraj, A., Meiners, J.-C., 2003, "Robust interconnects and
packaging for microfluidic elastomeric chips," Anal. Chem., 75(19), pp. 5287-
24. McDonald, J.C., Duffy, D.C., Anderson, J.R., Chiu, D.T., Wu, H., Schueller,
O.J.A., and Whitesides, G.M., 2000, "Fabrication of microfluidic systems in
poly(dimethylsiloxane)," Electrophoresis, 21, pp. 27-40.
25. Kai, J., Sohn, Y.S., and Ahn, C.H., 2003, "Protein microarray on cyclic olefin
copolymer (COC) for disposable protein lab-on-a-chip," in Proceedings ofuTAS
2003 Seventh International Conference on Micro Total Analysis Systems, Oct. 5-9,
2003, Squaw Valley, CA, pp. 1101-1104.
26. Jesson, G., Kylberg, G., and Andersson, P., "A versatile macro-to-micro dispensing
system," in Proceedings of uTAS 2003 Seventh International Conference on Micro
TotalAnalysis Systems, Oct. 5-9, 2003, Squaw Valley, CA, pp. 155-158.
27. Callister, Jr., W.D., 1997, Materials Science and Engineering: An Introduction, 4th
Ed., John Wiley & Sons, New York.
28. Styrolux Thermal Properties and Rheology (Information Sheet), June 2004, BASF,
Florham Park, NJ.
29. Zeonor Product Information Sheet 604ZR10030BP, 9 January 2004, Zeon
Chemicals LP, Louisville.
30. Celcon M90 Technical Data, K-mac Plastics, http://k-mac-
plastics.net/data%20sheets/celcon_datasheet.htm. Last accessed 4 November
31. Topas Product Brochure 351 D BR 08.2000, August 2000, Ticona GmbH,
32. Information about High Technology Silicone Materials, 1991, Dow Coming,
33. Fredrickson, C.K and Fan, Z.H., 2004, "Macro-to-micro interfaces for microfluidic
devices," Lab Chip, 4, pp. 526-533.
34. Nittis, V., Fortt, R., Legge, C.H., and de Mello, A.J., 2001, "A high-pressure
interconnect for chemical microsystem applications," Lab Chip, 1, pp. 148-152.
35. Galambos, P. and Benavides, G.L., 2000, "Electrical and fluidic packaging of
surface micromachined electro-microfluidic devices," in Proceedings ofSPIE The
International Society for Optical Engineering, 41 77(Microfluidic Devices and
Systems III), Sept. 18-19, Santa Clara, CA, pp. 200-207.
36. Martynova, L., Locascio, L.E., Gaitan, M., Kramer, G.W., Christensen, R.G.,
MacCrehan, W.A., 1997, "Fabrication of plastics microfluid channels by
imprinting methods," Anal. Chem., 69, pp. 4783-4789.
37. Carvalho, B.L., Schilling, E.A., Schmid, N. and Kellogg, G.J., 2003, "Soft
embossing of microfluidic devices," in Proceedings ofuTAS 2003 Seventh
International Conference on Micro Total Analysis Systems, Oct. 5-9, Squaw
Valley, CA, pp. 959-962.
38. Madou, M., Lee, L.J., Koelling, K.W., Daunert, S., Lai, S., Koh, C.G., Juang, Y.-J.,
Yu, L. and Lu, Y., 2001, "Design and fabrication of polymer microfluidic
platforms for biomedical applications," in ANTEC 2001 Plastics -- The Lone Star,
May 6-10, Dallas, TX, pp. 2534-2538.
39. Lee, G.-B., Chen, S.-H., Huang, G.-R., Sung, W.-C., and Lin, Y.-H., 2001,
"Microfabricated plastic chips by hot embossing methods and their applications for
DNA separation and detection," Sens. Act. B, 75, pp. 142-148.
40. Becker, H. and Heim, U., 2000, "Hot embossing as a method for fabrication of
polymer high aspect ratio structures," Sens. Act. A, 83, pp. 130-135.
41. Tan, W., Fan, Z.H., Qiu, C.X., Ricco, A.J., and Gibbons, I., 2002, "Miniaturized
capillary isoelectric focusing in plastic microfluidic devices," Electrophoresis, 23,
42. Xu, J., Locascio, L., Gaitan, M., and Lee, C.S., 2000, "Room-temperature
imprinting method for plastic microchannel fabrication," Anal. Chem., 72, pp.
43. Lee, D.-S., Yang, H., Chung, K.-H., Pyo, H.-B., 2005, "Wafer-scale fabrication of
polymer-based microdevices via injection molding and photolithographic
micropatteming protocols," Anal. Chem., 77, pp. 5414-5420.
44. McCormick, R.M., Nelson, R.J., Alonso-Amigo, M.G., Benvegnu, D.J., Hooper,
H.H., 1997, "Microchannel electrophoretic separations of DNA in injection-molded
plastic substrates," Anal. Chem., 69, pp. 2626-2630.
45. Choi, J.-W., Kim, S., Trichur, R., Cho, H.J., Puntambekar, A., Cole, R.L., Simkins,
J.R., Murugesan, S., Kim, K., Lee, JB, Beaucage, G., Nevin, J.H., and Ahn, C.H.,
2001, "A plastic micro injection molding technique using replaceable mold-disks
for disposable microfluidic systems and biochips," in Micro Total Analysis Systems
2001, J.M. Ramsey, and A. van den Berg, eds., Kluwer Academic Publishers, pp.
46. Liu, Y., Ganser, D., Schneider, A., Liu, R., Grodzinski, P., and Kroutchinina, N.,
2001, "Microfabricated polycarbonate CE devices for DNA analysis," Anal.
Chem., 73, pp. 4196-4201.
47. Stachowiak, T.B., Rohr, T., Hilder, E.F., Peterson, D.S., Yi, M., Svec, F., and
Frechet, J.M.J., 2003, "Fabrication of porous polymer monoliths covalently
attached to the walls of channels in plastic microdevices," Electrophoresis, 24, pp.
48. Muck, Jr., A., Wang, J., Jacobs, M., Chen, G., Chatrathi, M.P., Jurka, V., Vyborny,
Z., Spillman, S.D., Sridharan, G., and Schoning, M.J., 2004, "Fabrication of
poly(methyl methacrylate) microfluidic chips by atmospheric molding," Anal.
Chem., 76, pp. 2290-2297.
49. Zhou, W.X., and Chan-Park, M.B., 2005, "Large area UV casting using diverse
polyacrylates of microchannels separated by high aspect ratio microwalls," Lab
Chip, 5, pp. 512-518.
50. Bowden, M., Geschke, O., Kutter, J., and Diamond, D., 2003, "CO2 laser
microfabrication of an integrated polymer microfluidic manifold for the
determination of phosphorus," Lab Chip, 3, pp. 221-223.
51. Kameoka, J., Craighead, H.G., Zhang, H., and Henion, J., 2001, "A polymeric
microfluidic chip for CE/MS determination of small molecules," Anal. Chem., 73,
52. Bundgaard, F., Nielson, T., Nilsson, D., Shi, P., Perozziello, G., Kristensen, A., and
Geschke, 0., 2004, "Cyclic olefin copolymer (COC/Topas) an exceptional
material for exceptional lab-on-a-chip systems," in Micro TotalAnalysis Systems
2004: Proceedings of UTAS 2004, 8th International Conference on Miniaturized
Systemsfor Chemistry and Life Sciences, Malmo, Sweden, pp. 372-374.
53. McCormick, R.M., Nelson, R.J., Alonso-Amigo, M.G., Benvegnu, D.J., and
Hooper, H.H., 1997, "Microchannel electrophoretic separations of DNA in
injection-molded plastic substrates." Anal. Chem., 69, pp. 2626-2630.
54. Das, C., Xia, Z., Stoyanov, A., Fan, Z.H., 2005, "A laser-induced fluorescence
imaging system for isoelectric focusing," Inst. Sci. Tech., 33, pp. 379-389.
55. Chen, K.-S., Ay6n, A.A., Zhang, X., and Spearing, S.M., 2002, "Effect of process
parameters on the surface morphology and mechanical performance of silicon
structures after deep reactive ion etching (DRIE)," J MEMS, 11(3), pp. 264-275.
56. Mei, Q., Fredrickson, C., Jin, S., and Fan, Z.H., 2005, "Toxin detection by
miniaturized in vitro protein expression array," Anal. Chem., 77, pp. 5494-5500.
57. Guyton, A.C., and Hall, J.E., 2000, Textbook of Medical Physiology, 10th Ed.,
Elsevier, New York, pp. 24-32.
58. RTS Application Manual for Cell-Free Protein Expression, 2003, Roche
59. Angenendt, P., Nyarsik, L., Szaflarski, W., Glokler, J., Nierhaus, K.H., Lehrach,
H., Cahill, D.J., and Lueking, A., 2004, "Cell-free protein expression and
functional assay in nanowell chip format," Anal. Chem., 76, pp. 1844-1849.
60. Kinpara, T., Mizuno, R., Murakami, Y., Kobayashi, M., Yamaura, S., Hasan, Q.,
Morita, Y., Nakano, H., Yamane, T., and Tamiya, E., 2004, "A picoliter chamber
array for cell-free protein synthesis," J. Biochem., 136, pp. 149-154.
61. Alakhov, J.B., Baranov, V.I., Ovodov, S.J., Ryabova, L.A., Spirin, A.S., and
Morozov, I.J., 1995, "Method of preparing polypeptides in cell-free translation
system," U.S. Patent 5,478,730.
62. "For microplates well positions," Society for Biomolecular Screening Standard
63. Eschbach, L., 2000, "Nonresorbable polymers in bone surgery," Injury, 31(D), pp.
64. USP class 6 approved, implants, Epoxy Technologies,
http://www.epotek.com/subapplications.asp?ID=17&SID=18. Last accessed 4
65. Products, Airtech International, Inc.,
http://www.airtechonline.com/products/index.htm. Last accessed 4 November
66. Strain gage load cell for thrust measurement, Richard Nakka's Experimental
Rocketry Site, http://www.nakka-rocketry.net/strainlc.html. Last accessed 4
67. Wheeler, A.J., and Ganji, A.R., 1996, Introduction to Engineering
Experimentation, Prentice Hall, Upper Saddle River, pp. 194-196.
68. Rocketry software, Richard Nakka's Experimental Rocketry Site,
http://members.aol.com/nakkarocketry/softw.html. Last accessed 4 November
Carl Fredrickson grew up in Kent, Washington. In June of 2002, he earned a
Bachelor of Science degree in chemical engineering from the University of Washington.
He enrolled in the University of Florida to pursue a Master of Science degree in
mechanical engineering in August 2003.