Process Characterization of Fabricating Plastic Microfluidic Devices

fredrickson_c.pdf

fredrickson_c_Page_23.txt

fredrickson_c_Page_28.txt

fredrickson_c_Page_46.txt

fredrickson_c_Page_01.txt

fredrickson_c_Page_83.txt

fredrickson_c_Page_27.txt

fredrickson_c_Page_55.txt

fredrickson_c_Page_66.txt

fredrickson_c_Page_40.txt

fredrickson_c_Page_09.txt

fredrickson_c_Page_51.txt

fredrickson_c_Page_79.txt

fredrickson_c_Page_14.txt

fredrickson_c_Page_49.txt

fredrickson_c_Page_30.txt

fredrickson_c_pdf.txt

fredrickson_c_Page_35.txt

fredrickson_c_Page_78.txt

fredrickson_c_Page_80.txt

fredrickson_c_Page_26.txt

fredrickson_c_Page_31.txt

fredrickson_c_Page_61.txt

fredrickson_c_Page_77.txt

fredrickson_c_Page_81.txt

fredrickson_c_Page_43.txt

fredrickson_c_Page_50.txt

fredrickson_c_Page_13.txt

fredrickson_c_Page_05.txt

fredrickson_c_Page_73.txt

fredrickson_c_Page_63.txt

fredrickson_c_Page_48.txt

fredrickson_c_Page_56.txt

fredrickson_c_Page_64.txt

fredrickson_c_Page_17.txt

fredrickson_c_Page_68.txt

fredrickson_c_Page_16.txt

fredrickson_c_Page_25.txt

fredrickson_c_Page_20.txt

fredrickson_c_Page_15.txt

fredrickson_c_Page_53.txt

fredrickson_c_Page_74.txt

fredrickson_c_Page_33.txt

fredrickson_c_Page_03.txt

fredrickson_c_Page_69.txt

fredrickson_c_Page_70.txt

fredrickson_c_Page_18.txt

fredrickson_c_Page_32.txt

fredrickson_c_Page_11.txt

fredrickson_c_Page_65.txt

fredrickson_c_Page_60.txt

fredrickson_c_Page_38.txt

fredrickson_c_Page_29.txt

fredrickson_c_Page_52.txt

fredrickson_c_Page_75.txt

fredrickson_c_Page_12.txt

fredrickson_c_Page_44.txt

fredrickson_c_Page_62.txt

fredrickson_c_Page_04.txt

fredrickson_c_Page_47.txt

fredrickson_c_Page_22.txt

fredrickson_c_Page_07.txt

fredrickson_c_Page_67.txt

fredrickson_c_Page_57.txt

fredrickson_c_Page_42.txt

fredrickson_c_Page_10.txt

fredrickson_c_Page_71.txt

fredrickson_c_Page_41.txt

fredrickson_c_Page_21.txt

fredrickson_c_Page_58.txt

fredrickson_c_Page_37.txt

fredrickson_c_Page_72.txt

fredrickson_c_Page_34.txt

fredrickson_c_Page_36.txt

fredrickson_c_Page_54.txt

fredrickson_c_Page_19.txt

fredrickson_c_Page_82.txt

fredrickson_c_Page_06.txt

fredrickson_c_Page_39.txt

fredrickson_c_Page_76.txt

fredrickson_c_Page_45.txt

fredrickson_c_Page_08.txt

fredrickson_c_Page_24.txt

fredrickson_c_Page_59.txt

fredrickson_c_Page_02.txt

PROCESS CHARACTERIZATION OF FABRICATING PLASTIC MICROFLUIDIC
DEVICES













By

CARL FREDRICKSON


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


2005

































Copyright 2005

by

Carl Fredrickson















ACKNOWLEDGMENTS

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

CHAPTER

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
2.2.1.1 Photom ask design....................................... ......................... 16
2 .2 .1.2 Silicon m aster............. .................................... ........ ... ........... 17
2.2.1.3 G lass m asters .................................... ...................... ........... .... 18
2.2.1.4 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
2.3.1.1 Effects of temperature and pressure on device thickness...............26
2.3.1.2 Transfer of ridge detail ............. .............................. ...............28









2 .3.1.3 C razing ................................................ ............... 32
2.3.1.4 G lass-based e-form .................................. .......................... 35
2.3.2 L am nation ......................................................... .. ....... ... 35
2.3.2.1 Sandw ich m material ........................................ ........ ............... 36
2.3.2.2 R oller pressure........................................ ............ ............ 37
2.3.2.3 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

APPENDIX

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








v
















LIST OF TABLES

Table p

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


Figure p

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
DEVICES

By

Carl Fredrickson

December 2005

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.














CHAPTER 1
INTRODUCTION

1.1 History

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 [1]. 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 [2].

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

infancy.

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 [11]. 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 [12].

In the early 1990s, articles again began to appear giving the theoretical basis for

development of lab on a chip [13]. 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 [13].

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 [20]; 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. [21] 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 [23]). PDMS shows a good ability to replicate features [24], 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

cure.

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 [25].

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 [26]. 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.
Modulus of
Elasticity Tg Water Vapor Water
(GPa) (C) Permeability Absorption, 7 day
Material [27:778] [27:801] (g/m2 day) [28] immerse (%)
Polymethyl
methylacrylate 2.24-3.24 3 -13 1.7 [29]
(PMMA)

Polycarbonate
Polycarbonate 2.38 150 15 0.45 [29]
(PC)
Polystyrene
Polystyrene 2.28-3.28 100 N/A <0.1 [30]
(PS)
Polyethylene
terephthalate 2.8-4.1 70 -3 0.1
(PET)

Topas 8007
2.6 80 0.023 <0.01
(COC) [31]

Topas 6013
3.2 140 0.035 <0.01
(COC) [31]
PDMS 62 -123 N/A 0.1 [32]
Glass
so 69 N/A N/A N/A
(soda lime)
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

section.

1.4.1 Interconnect

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

separate module?

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

silicon.




* This section has been published with other materials in [33]: 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

conditions.

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
difficult.

* Reliability. Interconnects should be leak free at standard operating conditions for
the application.

* 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
efficiency.

* 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
microscopy.

* 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

failure occurs.

1.4.2 Propulsion

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 [13].

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

relationship [10:20-29]:

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

Se
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)
6xy7 r

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.

Voltage
Applied i




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
(pI).

1.5 Objectives

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.














CHAPTER 2
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 [36], silicon wafers [36], silicone [37,38], quartz [39], 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 [42]. Increasing the embossing

temperature yields features in the plastic devices much closer to the mold dimensions

[39].

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

mold [45].

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 [46]. Lesser utilized methods for creating

microfluidic devices include using a mill to cut channels [47], using ultraviolet light to

polymerize a liquid in a mold [48,49] and laser ablation [50].

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 [50]. 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 [53].









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.

2.2.1.1 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.

2.2.1.2 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.

2.2.1.3 Glass masters

The process used to fabricate the glass masters was adapted from known methods

[15]. 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

115 oC.









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.


4




6

Figure 2-3. Process diagram for glass masters.

2.2.1.4 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.




















A B

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.

SA B

Figure 2-5. Expected profiles of the molds used in this study. A) silicon-based
electroforms. B) glass slide master.

2.2.2 Molding

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.

Glass






E-form Spacer

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.

2.2.3 Milling

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
milling.

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.

2.2.4 Lamination

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

software.

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.




Mirror F,
d under
ND Filter p :1 Bandpass Filter
(535+/- 25nm)

Mirror
Laser Lens

Channel

Figure 2-9. Optical detection setup used for this investigation [54].

2.3 Results and Discussion

2.3.1 Molding

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.

2.3.1.1 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.

1.70
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-

1.20 -

1.10

1.00
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.









2.3.1.2 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 [55].

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
the corner.










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 .__

25 25888
25__ ON
200,000
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)









I E









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
30 09O









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

fabrication process.

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.

40
o Unselected #1 o Unselected #2

35
E
3 o
00
30
"3
0
03
25- n


20
0 1 2 3 4 5 6
Location


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

Ridge Widths
Average 29.9 29.0 29.8 29.7 30.7
St. Dev. 0.9 0.9 1.0 0.8 0.9

Ridge Separations
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.

2.3.1.3 Crazing

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



Optical Bench



Counter-


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.


45
40 -
35
S30-
S25
0
a 20 -
S15-
10 -
5
0
105 110 115 120 125 130 135
Temperature (OC)


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 2.3.1.1 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.

2.3.1.4 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.









A B

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.

2.3.2 Lamination

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

below.

















Figure 2-18. Laminator with preheating hotplate.

2.3.2.1 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.

2.3.2.2 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.


A B

Figure 2-19. Comparison of devices laminated at different roller settings. A) 1/16". B)
1/8".









2.3.2.3 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

observed.

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.

0.014
-120 C
0.012 ----115C
0.010 -------110C
0.010 / \ __ 105C
105 C
0.008 -

0.006 I

0.004 --' 1 .---- .

0.002 '" --

0.000
0.0 2.0 4.0 6.0 8.0 10.0 12.0
Time (s)


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

120 oC.

4.0
3.5
-3.0
U2.5

2.0 -
0
LL 1.5
.. 1.0
0.5
0.0
105 C 110 C 115 C 120 C
Roller Temperature

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
lamination.

2.3.3 Milling

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











C D

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.

4.5
-21 mm
4.0 -32 mm
3.5 43 mm
-54 mm
3.0










0 500 1000 1500 2000 2500
2.4 Conclusion 2.5
2.0
1.5
1.0
0.5
0.0
0 500 1000 1500 2000 2500
Pixel


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
V/cm respectively.

2.4 Conclusions

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.















CHAPTER 3
DESIGN, FABRICATION, AND TESTING OF A DEVICE FOR PROTEIN
EXPRESSION*

3.1 Introduction

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

screening applications.

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 [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.









transcription phase, ribonucleic acid (RNA) is copied from selected regions of

deoxyribonucleic acid (DNA) [57]. 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 [59] and picoliters [60] 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.

3.3 Experimental

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

reaction chamber.


.

t[1 insert or tray
insert or tray reaction chamber
1) i membrane---

well plate feeding chamber
A B

Figure 3-1. Illustrations of well-device. A) Three dimensional B) cross sectional.

3.3.1 Adhesives

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

luminometer.

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 [62], respectively,

which in turn would allow finalized arrays to be analyzed using existing microplate

readers.








Reaction chamber




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 [63], 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.

bp bp
-2,645

-1,605 -2,645
-1,198 -1,605
-1,198
-676
-517
-676
-350 -517
460
-222 -396
-179 -350
-126
-75 -222
-179
-126
-75


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)
epoxies.


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


problematic.


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 [64].


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

labeled #3.










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.

18000
*Tray Device
15000 A Microcentrifuge tube

S12000

9000

= 6000
0)
3000

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.

3.5 Conclusions

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.














CHAPTER 4
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

channels.

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 [45].

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 [65] 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

peel strengths.

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

the three.

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

factors.

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

feeding solution.

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.














APPENDIX
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 [66].

A.1 Design

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 [67]









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.



R, R2


vs V0

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

R2R,
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,.

A.2 Construction

Mr. Nakka has conveniently provided a spreadsheet [68] 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.






4756

23.81
RiO.41


31.75


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.














A B

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.3 Calibration

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.

5.0

4.5 y = 1.1043x + 2.399
R2 = 0.9962
4.0


S3.5

3.0

2.5

2.0
0 0.5 1 1.5 2
Force (N)


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,"
http://masspec.scripps.edu/information/history/perspectives/borman.html, last
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.
2623-2636.

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
November 2005.

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
Raton.

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-
5291.

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
2005.

31. Topas Product Brochure 351 D BR 08.2000, August 2000, Ticona GmbH,
Frankfurt, Germany.

32. Information about High Technology Silicone Materials, 1991, Dow Coming,
Midland, MI.

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,
pp. 3638-3645.

42. Xu, J., Locascio, L., Gaitan, M., and Lee, C.S., 2000, "Room-temperature
imprinting method for plastic microchannel fabrication," Anal. Chem., 72, pp.
1930-1933.

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.
411-412.

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.
3689-3693.

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,
pp. 1935-1941.

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
Diagnostics, Indianapolis.

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
ANSI/SBS 4-2004.

63. Eschbach, L., 2000, "Nonresorbable polymers in bone surgery," Injury, 31(D), pp.
22-27.

64. USP class 6 approved, implants, Epoxy Technologies,
http://www.epotek.com/subapplications.asp?ID=17&SID=18. Last accessed 4
November 2005.

65. Products, Airtech International, Inc.,
http://www.airtechonline.com/products/index.htm. Last accessed 4 November
2005.

66. Strain gage load cell for thrust measurement, Richard Nakka's Experimental
Rocketry Site, http://www.nakka-rocketry.net/strainlc.html. Last accessed 4
November 2005.

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
2005.















BIOGRAPHICAL SKETCH

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.

PAGE 1

PROCESS CHARACTERIZATION OF FABRICATING PLASTIC MICROFLUIDIC DEVICES By CARL FREDRICKSON A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2005

PAGE 2

Copyright 2005 by Carl Fredrickson

PAGE 3

ACKNOWLEDGMENTS First off, thanks go to my family for th eir 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 Engineerings Major Analytical Instrumentation Center, undergraduates Fernan do 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 biocompa tibility and protein expression and helping enhance my understanding of the processes. This work was supported in part by the star tup fund from the University of Florida, the National Science Foundation, and the Natio nal 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. iii

PAGE 4

TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................iii LIST OF TABLES .............................................................................................................vi LIST OF FIGURES ..........................................................................................................vii ABSTRACT .......................................................................................................................ix CHAPTER 1 INTRODUCTION........................................................................................................1 1.1 History ....................................................................................................................1 1.2 The Case for Miniaturization .................................................................................4 1.2 Common Materials for Lab-on-a-Chip Devices ....................................................5 1.4 Design Attributes ...................................................................................................8 1.4.1 Interconnect .................................................................................................9 1.4.2 Propulsion ..................................................................................................11 1.4.3 Isoelectric Focusing ...................................................................................12 1.5 Objectives ............................................................................................................13 2 CHARACTERIZATION OF FABRICATION PROCESS........................................14 2.1 Previously Disclosed Methods of Fabrication .....................................................14 2.2 Fabrication Process and Methods of Testing .......................................................16 2.2.1 Master Fabrication.....................................................................................16 2.2.1.1 Photomask design ............................................................................16 2.2.1.2 Silicon master ..................................................................................17 2.2.1.3 Glass masters ...................................................................................18 2.2.1.4 Electroform fabrication ...................................................................19 2.2.2 Molding.....................................................................................................20 2.2.3 Milling.......................................................................................................22 2.2.4 Lamination .................................................................................................24 2.2.5 Demonstration Protein Separation ..........................................................25 2.3 Results and Discussion ........................................................................................26 2.3.1 Molding.....................................................................................................26 2.3.1.1 Effects of temperature and pressure on device thickness ................26 2.3.1.2 Transfer of ridge detail ....................................................................28 iv

PAGE 5

2.3.1.3 Crazing ............................................................................................32 2.3.1.4 Glass-based e-form ..........................................................................35 2.3.2 Lamination .................................................................................................35 2.3.2.1 Sandwich material ...........................................................................36 2.3.2.2 Roller pressure .................................................................................37 2.3.2.3 Roller temperature ...........................................................................38 2.3.3 Milling.......................................................................................................40 2.3.4 Actual Usage Protein Separations ..........................................................43 2.4 Conclusions ..........................................................................................................44 3 DESIGN, FABRICATION, AND TEST ING OF A DEVICE FOR PROTEIN EXPRESSION............................................................................................................46 3.1 Introduction..........................................................................................................46 3.2 Synthesis of Proteins ............................................................................................46 3.3 Experimental ........................................................................................................48 3.3.1 Adhesives ...................................................................................................48 3.3.2 Device Fabrication .....................................................................................49 3.3.3 Protein Expression .....................................................................................50 3.4 Results and Discussion ........................................................................................51 3.4.1 Device Designs ..........................................................................................51 3.4.2 Adhesive Biocompatibility ........................................................................53 3.4.3 Protein Expression .....................................................................................56 3.5 Conclusions ..........................................................................................................56 4 CONCLUSIONS AND FUTURE RECOMMENDATIONS....................................58 4.1 Compression Molding Process Optimization ......................................................58 4.2 Device for Protein Expression .............................................................................60 4.3 Overall Conclusions .............................................................................................61 APPENDIX CONSTRUCTION AND CALIBRATION OF LOAD CELL USED FOR PEEL TESTS .........................................................................................................................62 A.1 Design.................................................................................................................62 A.2 Construction ........................................................................................................64 A.3 Calibration ..........................................................................................................66 LIST OF REFERENCES ...................................................................................................67 BIOGRAPHICAL SKETCH .............................................................................................73 v

PAGE 6

LIST OF TABLES Table page 1-1 Properties of polymeric materi als compared to glass and silicon. ...............................8 2-1 Width and separation of ridges shown in Figure 2-13. ...............................................32 vi

PAGE 7

LIST OF FIGURES Figure page 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 master. ............................................................................18 2-3 Process diagram for glass masters. .............................................................................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 th ickness as a function of temperature. ........................27 2-11 SEM picture of part of a plastic device.. ................................................................28 2-12 Typical profilometry results.. ...................................................................................29 2-13 Variation in ridge height as a function of temper ature and location. .......................30 2-14 Feature size data taken from two other silic on wafers that were candidates to be the master. ......................................................................................................................31 2-15 Dependence on crazed area with reference to cooling conditions. ..........................33 2-16 Crazing as a function of temper ature after cooling in a 60 C oven. ........................34 2-17 Devices molded against the glass-based e-form. ......................................................35 vii

PAGE 8

2-18 Laminator with preheating hotplate. .........................................................................36 2-19 Comparison of devices laminated at different roller settings...................................37 2-20 Peel test results gr ouped 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 a nd 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 well-device. ........................................................................................48 3-2 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 cu ring of epoxy during insert fabrication. ..................55 3-6 Comparison of expression of CAT be tween microcentrifuge tube and device. .........56 A-1 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 viii

PAGE 9

Abstract of Thesis Presen ted 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 DEVICES By Carl Fredrickson December 2005 Chair: Z. Hugh Fan Major Department: Mechanic al and Aerospace Engineering Running chemical analysis in microfluidic devices, sometimes referred to as labon-a-chip technology, can increase accuracy while reducing the analysis time, sample size, and the amount of reagents consumed. Wh ile glass, silicon, and plastic can be used to make these devices, plastics have the adva ntage of lower costs, re sulting in disposable devices that eliminate the risk of cross-contamination. App lications where these devices can be used include, but are not limited to, dr ug discovery, point-of-care or field testing, and forensic analysis. Most plastic devices are cr eated using an embossing or injection molding process, and usually only the conditions used to fabricate the particular devi ce being discussed are disclosed; how changing those variables affect s the final product is not investigated in detail. Compression molding is a compromise of these fundamental techniques, using embossing technology to mold polymer resi n instead of preformed plaques. A characterization of a process for creating microfluidic de vices by compression molding a ix

PAGE 10

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 met hod used to fabricate a two-chambered device for the specific purpose of protein e xpression is discussed; greater amounts of protein were found to be expressed in this de vice than in standard microcentrifuge tubes under identical conditions. x

PAGE 11

CHAPTER 1 INTRODUCTION 1.1 History Despite the recent growth of microchemistry in both applications and methods, the need for analyzing small quantities of material s can be dated back to more than a century ago, when the novelty of radioactive materials wa s at the forefront of science. In an 1898 paper, Pierre and Marie Curi e announced the discovery of a substance more radioactive than uranium, and suggest ed the name polonium [ 1 ]. 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 sa mple (the amount of material they started with is not mentioned). Si x 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 [ 2 ]. In the late 1930s or early 1940s, the fiel d 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 Seaborgs plutonium research in 1942: With the special tools of ultramicroch emistry the young chemists could work on undiluted quantities of chemicals as slight as tenths of a microgram They would manage their manipulations on the mechan ical stage of a binocular stereoscope adjusted to 30-power magnification. Fine glass capillary straws substituted for test tubes and beakers; pipettes filled automa tically by capillary attraction; small hypodermic syringes mounted on micromanipulators injected and removed reagents from centrifuge microcones; mini ature centrifuges separated precipitated solids from liquids. The first balance the ch emists used consisted of a single quartz fiber fixed at one end like a fishing pole stuck from a riverbank inside a glass 1

PAGE 12

2 housing that protected it from the least brea th of air. To we igh 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 de flection which was calibrated against standard weights . It was said, note s 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 infancy. Mass spectroscopy (MS) was first developed in the early 1900s. Initial separations and results were crude, with the greatest gr owth in the technology coming in the 1940s1960s. Though a number of subfields have developed, the general a pproach is to send particles through an electromagnetic fiel d, which then separates them by their charge/mass ratio. Originally, only atoms and small molecules could be analyzed using this technique; more recently, larger molecu les 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 ra tes based on their affinities to each of the phases, and the elution times of each constitu ent 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 subdi visions existing for each technique [ 8 :609-610].

PAGE 13

3 Another separation technology is electrophores is, which uses an electric potential to separate charged particles. It is frequently used to an alyze biological molecules, such as proteins and DNA, since these mol ecules 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 expe riment 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 a ddress this challenge is to miniaturize these instruments, as it offers the opportunity to run many experiments in parallel. The first widely-recognized attempt at mi niaturization was reported in 1979, when researchers at Stanford University devel oped a gas chromatograph where the 1.5 meter long separation channel fit on a 2-diameter silicon wafer [ 11 ]. For a variety of reasons, development of additional micro analysis systems languished in 1980s, although examples of individual components, such as microvalves, were reported [ 12 ]. In the early 1990s, articles again began to appear giving the theoretical basis for development of lab on a chip [ 13 ]. Harrison et al. [ 14 15 ] introduced a device formed from a glass plate and appears to be the harb inger for modern research in the field. Since then, an uncountable number of microf luidic devices have been reported, and the field is continuing to grow Devices have been developed or proposed to detect a number of substances, includi ng illegal drugs and explosiv es residues, biological and chemical warfare agents, and environmental po llutants. They have also been developed for growing cells, reproducing DNA via polymerase chain reaction (PCR), and immunoassays to determine the effi cacy of drugs and/or toxins [ 16 19 ]

PAGE 14

4 1.2 The Case for Miniaturization The general idea behind micro total analysis systems ( TAS), also known as labon-a-chip (LOC), is to scale down chemical reactions, biological assays, cell manipulations, and other analy tical processes as much as possible. These operations, which include mixing, separations, and chemical reactions, are typically integrated into planar chips. Some chips also embed samp le 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 ar bitrary length scale, the time required for analysis would be a function of the area, or d 2 Thus, by reducing the diameter of a channel by a factor of 10, th e analysis time could be reduced by a factor of 100. This reduction in length scale would l ead to better and fast er electrophoretic and chromatographic separations [ 13 ]. A significant consequence of mi niaturization is that the amount, and thus the cost, of reagents required to comple te a study is also significantly reduced. This is especially important to applications where materi als 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 i llustration in Figure 1-1.

PAGE 15

5 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 [ 20 ]; 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 functi onal microfluidic devi ces fall into one of three categories: silicon, glass, and polyme rs; 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 th e gas chromatograph on a 2-diameter silicon wafer. Other silicon devi ces include the systems deve loped by Gray, et al. [ 21 ] Silicon was a natural first choice since micro-scal e fabrication had been well characterized

PAGE 16

6 during the growth of th e electronics industry [ 22 :184-185] It is crystalline, and can be etched isotropically (same etch rate in all directions, resu lting in rounded features) or anisotropically (different etch rate in different directions, re sulting in features with high aspect ratios and well-defined a ngles). 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 insula te the channels if using electrokinetic pumping for propulsi on or electrochemi cal sensing. Glass is an amorphous solid with excel lent chemical and thermal resistance properties. The devices reporte d by Harrison et al. were crea ted 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/prototypi ng scale, it can be difficult to achieve consistent channel di mensions, through-holes, and bonding. Polymers can be split into two basi c groups: thermoplastic s 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 ch ains 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 wont melt, but also makes it difficult to reprocess the material.

PAGE 17

7 With the exception of the wide use of polydimethyl siloxane (PDMS), it appears that thermoset polymers are not used to ma ke functional chips (though epoxies have been used for packaging [ 23 ]). PDMS shows a good ability to replicate features [ 24 ], but its elastomeric nature makes it inappropriate fo r many applications. Use typically entails mixing two liquids (base and curing agent), pouring it into a mol d, and allowing it to cure. In the midto late 1990s, many inve stigators 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 mechan ical and thermal properties of the final product. Though more expensive, COCs have exhibited greater chemical resistance and optical properties than other polymers [ 25 ]. Table 1-1 summarizes some of the important properties of the materials discussed above, as well as other polymers that may ar ise as options for microfluidic systems. From a commercialization standpoint, therm oplastic polymers are the best material for most microfluidic devices. They are rela tively inexpensive, and much, if not all, of the molding and sealing process can be auto mated, 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 [ 26 ]. However, in a research or prototyping environment wher e few chips will be made, fabrication of thermoplastic chips requires much more in term s of equipment and effort than PDMS.

PAGE 18

8 Table 1-1. Properties of polymeric mate rials compared to glass and silicon. Material Modulus of Elasticity (GPa) [ 27 :778] Tg ( o C ) [ 27 :801] Water Vapor Permeability (g/m 2 day) [ 28 ] Water Absorption, 7 day immerse (%) Polymethyl methylacrylate (PMMA) 2.24-3.24 3 ~13 1.7 [ 29 ] Polycarbonate (PC) 2.38 150 ~ 15 0.45 [ 29 ] Polystyrene (PS) 2.28-3.28 100 N/A <0.1 [ 30 ] Polyethylene terephthalate (PET) 2.8-4.1 70 ~ 3 0.1 Topas 8007 (COC) [ 31 ] 2.6 80 ~ 0.023 <0.01 Topas 6013 (COC) [ 31 ] 3.2 140 ~ 0.035 <0.01 PDMS 62 -123 N/A 0.1 [ 32 ] Glass (soda lime) 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 ba sis. 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 overvie w of the design attri butes and technologies

PAGE 19

9 incorporated into the devices discussed in late r chapters, and that is the purpose of this section. 1.4.1 Interconnect 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 in clude 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 appl ications involving EOF, how will the electrodes be included? Is the entire in terconnect to be disposable or just a certain part of it? Is it advantageous to integrate in terconnects into the device or leave it as a separate module? 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 manuf acturing process is rigorously controlled, aligning to an edge isnt always a failproof me thod of creating a connection. Second, the dimensions and tolerances of the de vice 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 interconnect s must be compatible to the device, since the device can be fabricated from a number of materials, including glass, plastics, or silicon. This section has been published with other materi als in [33]: Fredrickson, C.K and Fan, Z.H., 2004, Macro-to-micro interfaces for mi crofluidic devices, Lab Chip, 4 pp. 526-533.

PAGE 20

10 Materials used in interconnects include me tals such as steel and aluminum, glass, silicon, and polymers such as PEEK, PTFE, a nd 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 conditions. After answering the fundamental questions and addressing the design obstacles, it may be prudent to modify the design to in corporate desired characteristics. These characteristics may include the following. Ease of assembly. Interconnects should ali gn to the ports on the microfluidic device easily and with minimum need for supporting jigs. If an interconnect consists of many components, they shoul d go together quickly with minimal use of tools. Replacement of consumables, such as capi llary tubes and gaskets, should not be difficult. Reliability. Interconnects should be leak free at standard operating conditions for the application. 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 efficiency. Maximum field of view. An interconnect shou ld allow the user to gather data when the analysis will be conducted optically, such as by laser induced fluorescence or microscopy. Minimal pressure drop when the flow is pneumatically driven. Pressure drop is caused by the geometry of the syst em as well as the constriction. Ability to operate over a range of flow rates.

PAGE 21

11 Ability to be automated. Certain experiment s 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 fabricati ng interconnects, data should be collected to test if they successfully meet the desired operat ional 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. Fl ow rates and pressures are measured during operation. Mechanical strength may also be tested by pulling the capillary or tube until failure occurs. 1.4.2 Propulsion One of the easiest ways to propel fluids th rough 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 pr essure drop over a given distance increases significantly, resulti ng in decreased flow or the need for stronger pumps [ 13 ]. 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 th e wall an inherent charge. At the same time, a diffuse double layer of opposite char ges will develop in the solution adjacent to the wall. When voltage is applied, a sh ear 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 relationship [ 10 :20-29]: EP EOF OBSEv (1)

PAGE 22

12 where v OBS is the velocity in cm/s, E is the electric field strength in V/cm, and EOF and EP are the electroosmotic and electrophoretic mobilities, respectiv ely, in units of cm 2 /V. The electroosmotic mobility is a measure of the speed of neutral material in the channel, and can be found by the relationship eEOF (2) where is the thickness of the diffuse double layer in cm, e is the charge per unit surface area (coulomb/cm 2 ), and 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 particles charge in coulombs, q, and radius in cm, r by the equation r qEP 6 (3) 1.4.3 Isoelectric Focusing Isoelectric focusing is a method to separa te and analyze proteins. As mentioned earlier, electrophoresis relies on an electric field to separate molecules and ions. However, nothing in the process stops the mo tion of the analyte; if the voltage were applied for an infinite amount of time, eventu ally 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

PAGE 23

13 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 pa ssing 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 sa mple to be separated contains only proteins, only a few discrete pH values exist in the pa th between the electrode s, resulting in a steplike 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. Voltage Applied Figure 1-2. Basic process of isoelectric focusing with carri er 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 (pI). 1.5 Objectives The objective of this thesis work is to develop and st udy fabrication processes for microfluidic chips. Chapter two presents the characterization of a compression molding process for fabricating COC-based microf luidic 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.

PAGE 24

CHAPTER 2 CHARACTERIZATION OF FABRICATION PROCESS This section focuses on the characterization of a generalized fabrication process. The basic approach employed is to compre ssion 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. Bu t before this method is investigated, it is necessary to review the methods that have b een 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 literat ure for patterning channels into polymeric substrates. The most fr equently reported method is embossing, which consists of pressing a pre-formed plastic plat e against a mold, which in turn have been made from wires [ 36 ], silicon wafers [ 36 ], silicone [ 37 38 ], quartz [ 39 ], 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 [ 42 ]. Increasing the embossing temperature yields features in the plastic devices much clos er to the mold dimensions [ 39 ]. Injection molding requires melting the materi al, and then injecti ng it into a cavity. While used frequently in industry to make pl astic parts, it is used less frequently in laboratory/prototype setti ngs due to the costs of the inje ction molder and molds, though 14

PAGE 25

15 some groups have employed this method [ 43 44 ]. One example of note is the work of Ahn et al, who custom-an made injection mo lding machine and used an interchangeable nickel electroform insert similar to the type th at will be used in this investigation as their mold [ 45 ]. Compression molding is a compromise be tween embossing and injection molding, placing resin (pellets) on the mo ld 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 [ 46 ]. Lesser utilized methods for creating microfluidic devices include using a mill to cut channels [ 47 ], using ultraviolet light to polymerize a liquid in a mold [ 48 49 ] and laser ablation [ 50 ]. Sealing is frequently reported as being accomplished through a thermal annealing process, which is essentially heating the devi ce 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 C for 2 hours, and then allowed it to cool from 30 C to room temperature overnight in an oven [ 50 ]. 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 channe ls. 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 temperat ure (Tg) as an adhesive [ 47 52 ] Use of a heat-activated adhesive to bond the film and devi ce has also been reported [ 53 ].

PAGE 26

16 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 thos e that are commercially available. 2.2 Fabrication Process and Methods of Testing 2.2.1 Master Fabrication The masters were created using a photolithogr aphic process similar to that used to fabricate microchips for the electronics industr y. 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. 2.2.1.1 Photomask design Figure 2-1 shows the general patterns used in this investigation. The -D device consisted of a number of horizontal channels crossed by a single vertical channel of identical width. Two versions of this patt ern 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 m wide and separated by 90 m gaps; 87 channels could be fit into the device. However, it al so was designed with the intention that the silicon wafer would be the mold, thus the chan nels 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 m wide and separated by 340 m gaps on the photomask, with the understandi ng that the channels would grow during the isotropic etching process, undercutting prot ective masking layers and resulting in a Dshaped profile; only 29 horizontal cha nnels could be fit into the same area.

PAGE 27

17 The -channel design consists of six channels 10, 21, 32, 43, 54, and 65 mm long, and was intended for glass. Each cha nnel 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 m wide, again with the understanding that the channels would grow during the isotropic etching process. 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. 2.2.1.2 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 th e packaging and shipping process, and each subsequent chemical removes residue of th e previous. Shipley 1813 photoresist (Shipley, Marlborough, MA) was decanted onto the wafer, spun at 4000 rpm to a thickness of approximately 2 m (step 1 Figure 2-2), and soft bake d for 30 minutes at 85-95 C. The 2-D pattern was exposed (step 2) using an MJB 3 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 C.

PAGE 28

18 The wafers were etched using deep reactive ion etching (DRIE) (Surface Technology Systems, Ltd., Newport, UK) (ste p 4). DRIE creates high-aspect ratio structures by alternating between etching a nd the application of a passivation layer to protect the areas that shouldnt be etched. Material that remained after the etching process (both photoresist and passivation layer) was rem oved using a five-minute O 2 plasma etch (step 5). ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~1 2 3 4 5 4 5 Figure 2-2. Process diag ram for silicon master. 2.2.1.3 Glass masters The process used to fabricate the gla ss masters was adapted from known methods [ 15 ]. Glass microscope slides (Fisher Scientif ic, Atlanta, GA) were annealed (step 1 in Figure 2-3) in a programmable furnace (Bar nstead 4500, Barnstead Scientific, Dubuque, Iowa) to reduce internal stresses created dur ing the manufacturing pr ocess, resulting in smooth channel edges during the etching steps. To protect the glass during the etching process, 300 of chrome and 1000 of gold (both 99.99+%, Alfa Aesar) were vapor deposited via an electron-beam process (chrom e serves as an adhesion promoter between glass and gold). Positive photor esist (Shipley 1813) was decant ed onto the slide, spun at 4000 rpm to a thickness of approximately 2 m, and soft baked for 30 minutes at 85-95 C (step 2). The patterns were exposed (s tep 3) using 405 nm light on an MJB3 mask aligner, developed (AZ Mif 312 1:1:2) (step 4), and then ha rd baked for 30 minutes at 115 C.

PAGE 29

19 To define the channels in the glass slid es (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% HNO 3 / 66% H 2 O 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. ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ 1 2 3 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ 1 2 3 4 5 6 4 5 6 Figure 2-3. Process diagram for glass masters. 2.2.1.4 Electroform fabrication The masters were shipped to Optical El ectro 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 wa s unrealized that silver was still on the mold when it was returned, and it lifted off dur ing the first few pressing operations. The discolorations observed in the edges are rema ining silver residue. It was also observed that bumps form on the back of the e-form wh ere the nickel accumulated faster during the electroplating process; these were ground off after receipt.

PAGE 30

20 A B Figure 2-4. E-forms used for this resear ch. A) silicon master; the odd coloration (especially around edges) is due to imp roper removal of th e 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 m wide separated by 90 m gaps. Figure 2-5 B shows the profile of the glass-based e-forms, from which we e xpect D-shaped channels approximately 30 m deep and 120 m wide. A B Figure 2-5. Expected profiles of the mold s used in this study. A) silicon-based electroforms. B) glass slide master. 2.2.2 Molding 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 hydr aulic press (Carver MH, 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 un til the glass plate contacted the upper platen, allowed to sit for five minutes to warm the resin, and then compressed using a

PAGE 31

21 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 un it 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 tr avel from full-up to full-down in 10, 30, or 60 seconds; a graduated stick was used as a refe rence). Spacers were used to achieve the desired thickness of the device. This final c onfiguration 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. platen platen Spacer Glass E-form 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 transferre d from e-form to device, these samples were metrologically characterized using a profilo meter (Dektak IIa, Veeco Instruments, Woodbury, NY); scan distance was 250 m 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 ex treme 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 sca nner (Canoscan 30, Canon Inc.) in grayscale

PAGE 32

22 mode. The devices were covered with a piece of blue vinyl to provide contrast. Images were analyzed using th e threshold function in ImageJ ( http://rsb.info.nih.gov/ij/ ). Figure 2-7. Device exhi biting discoloration. As this first batch of devices show ed no relationship be tween 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 me tal 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 overl apping set was recast using the same force and compression rate, bu t 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. 2.2.3 Milling 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

PAGE 33

23 chips cast against the glass-based 2-D pattern using a 2-mm diameter, twin fluted squareend mill with titanium nitrate (TiN) coating (Richards Mi cro 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 ra tes varying from 225 to 300 mm/min. By making each cut in a single pass, one si de 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 immediat ely behind or in front of the cutting tool, depending on the directions of rotation and f eed; standard milling ejects the material to the front, while climb milling ejects it to the rear). climb standard climb standard 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 diff erence between climb and standard milling. Patterns were cut with the channels on t op (facing cutting tool ) and bottom (facing table); an acrylic plate was secured to the tabl e to keep the tool from hitting the table. Debris was removed by blowing air across th e device; an air line was fitted with an adjustable valve (Campbell Hausfeld, Harris on, OH) equipped with a pressure gauge set to either 0 (off), 10, or 20 psi. A 12 flex ible hose (Loc-line, Lockwood Products, Lake Oswego, OR) equipped with a 1/16 diameter nozzle was connected to the outlet of the 25 50 75 100 125 150 175 200

PAGE 34

24 valve; the nozzle was positioned 3-5 cm from th e cutting tool, and directed to blow at the tool-device interface at an angl e of about 35 from horizontal. 2.2.4 Lamination Topas 8007 film (4 mil, or approximately 100 m 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 minut es. The detergent was disposed of and the pieces were rinsed at least twice by submergi ng in ultra-purified wa ter in an ultrasonic bath for 5-10 minutes. Devices were air-dr ied 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 te st 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 a nd 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 create d. 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 su rface 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

PAGE 35

25 film with it. Voltage data was collected fr om a Wheatstone bridge containing the strain gage via a digital multimeter (Model 72-6870, Tenma) and Tenmas Data Logger software. 2.2.5 Demonstration Protein Separation Isoelectric focusing of proteins in a channe l 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) (M n = 10,000, Aldrich, St. Louis, MO) and 2% 2-hydroxyethyl cellulose (HEC) (M v = 90,000, Aldrich) in water at 60 C u nder 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 l of this solution was combined with 10 l of 80% w/w glycerol (Fisher Scientific, Atlanta, GA) and 5 l of carrier ampholytes for pH range 3-10 (Bio-Rad Laboratories, Hercules, CA). Recombin ant green fluorescent protein (GFP) (BD Biosciences Clontech, Palo Alto, CA) and R-Phycoerythrin (RPE) (Cyanotech, Kailuakona, HI) were added such that the final protein concentration was 5g/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 remain ing 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 pow er supply (Glassman High Voltage, High Bridge, NJ) interfaced via a Labview SCB-68 card to a computer running an in-house

PAGE 36

26 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 illumi nated. A schematic of this setup is shown in Figure 2-9. Mirro r Mirro r Beam expander Bandpass Filter (535+/25nm) Lens ND Filte r Laser Channel Figure 2-9. Optical detection set up u s ed f o r this investig ation [ 54 ]. 2.3 Results and Discussion 2.3.1 Molding Though referred to as a resin, the plastic bei ng used for this inve stigation is actually in pellet form. Since it is not heated prior to application to the m o ld, there is a strong possibility that air can b ecom e entrappe d between the pellets during the com p ression step, resulting in bubbles in the chip. In working with the m a terials and press before this investigation, it was found that by slowing the rate at which the ram controlling lever was cycled, the inclusion of bubbl es could be m i ni m i zed. 2.3.1.1 Effects of temperature a nd pressure on device th ickness Since spacers were used, as shown in Figur e 2-6, the thickness of every chip should have all been the sam e with an expected device thickness of about 1 mm However, in

PAGE 37

27 the course of handling the devices, differen ces were noted. The overall thickness of the pieces was measured using dial calipers, w ith the data presented in Figure 2-10. 1.00 1.10 1.20 1.30 1.40 1.50 1.60 1.70 80 100 120 140 160 Platen Temp. (C)Thickness (mm) 5 kip, 10 sec. 5 kip, 30 sec. 5 kip, 60 sec. 10 kip, 10 sec. 10 kip, 30 sec. 10 kip, 60 sec. Figure 2-10. Observed change in device thickness as a function of temperature. The data suggests that de vice 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 tr ansition state where both solid and liquid properties are exhibited. In th is 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 fo rces (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.

PAGE 38

28 2.3.1.2 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-ba sed 2-D pattern. As mentioned in section 2.2.1, the expected chip would be the negative of an actual device; ridges 30 m wide separated by 90 m gaps were expected. S ections of a device were observed under a scanning elec tron microscope, and Figure 211 shows typical results. In general, the plastic devices appear to be accep table. 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 [ 55 ]. Another characteristic to note is the appearan ce of slight deformations along the top edges of the features; this is particularly noticeable corners as shown by the inset. Figure 2-11. SEM picture of part of a plas tic device. Exploded view shows evidence of ability to transfer scallops from mold to device, along with deformation in the corner.

PAGE 39

29 Profilometry data was collected from each of the cast devices to quantify the feature sizes. Typical profilometer re sults are shown below in Figure 2-12 A. One of the first features to note in the profiles is the presence of large hump s on the tops of the ridges. This provides additional evidence of deformations along the top edges of the features, as observed in Figur e 2-11. These bumps were not observed in the wafers, as shown in Figure 2-12 B. A 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 de termine the ridge dimensions are shown in Figure 2-12 A. Ridge heights were meas ured 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 ve rtical. Ridge widths and separations were considered to be the distance between two ridge edges; edges were de fined as the location where the stylus stopped rising or began to fall during the course of th e scan; in the event a hump was present, the edge was the locati on on the separation-si de of the hump where it crossed the horizontal line pa ssing through the point used to determine the height of the ridge. As the profilometer analysis soft ware had a horizontal incrementation of 2 m in

PAGE 40

30 the mode used for measurement, it was not al ways possible to get the cursor to align exactly with these points. Feature height was measured at five loca tions spaced evenly across the pattern, and was taken in areas without crazing whenever pos sible. 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 Fi gure 2-13. The measurements varied from approximately 33 m near the ridge ends to 29 m in the center, or about 13% difference. 13 4 5 2 Location 2 27.0 28.0 29.0 30.0 31.0 32.0 33.0 34.0 35.0 90100110120130140150160 Temperature (C)Ridge Height (um) 10 sec., 5 kip 30 sec., 5 kip 60 sec., 5 kip Location 4 27.0 28.0 29.0 30.0 31.0 32.0 33.0 34.0 35.0 90100110120130140150160 Temperature (C)Ridge Height (um) 10 sec., 5 kip 30 sec., 5 kip 60 sec., 5 kip Location 1 27.0 28.0 29.0 30.0 31.0 32.0 33.0 34.0 35.0 36.0 90100110120130140150160 Temperature (C)Ridge Height (um) 10 sec., 5 kip 30 sec., 5 kip 60 sec., 5 kip Location 3 27.0 28.0 29.0 30.0 31.0 32.0 33.0 34.0 35.0 90100110120130140150160 Temperature (C)Ridge Height (um) 10 sec., 5 kip 30 sec., 5 kip 60 sec., 5 kip Location 5 27.0 28.0 29.0 30.0 31.0 32.0 33.0 34.0 35.0 90100110120130140150160 Temperature (C)Ridge Height (um) 10 sec., 5 kip 30 sec., 5 kip 60 sec., 5 kip 13 4 5 2 13 4 5 2 Location 2 27.0 28.0 29.0 30.0 31.0 32.0 33.0 34.0 35.0 90100110120130140150160 Temperature (C)Ridge Height (um) 10 sec., 5 kip 30 sec., 5 kip 60 sec., 5 kip Location 4 27.0 28.0 29.0 30.0 31.0 32.0 33.0 34.0 35.0 90100110120130140150160 Temperature (C)Ridge Height (um) 10 sec., 5 kip 30 sec., 5 kip 60 sec., 5 kip Location 2 27.0 28.0 29.0 30.0 31.0 32.0 33.0 34.0 35.0 90100110120130140150160 Temperature (C)Ridge Height (um) 10 sec., 5 kip 30 sec., 5 kip 60 sec., 5 kip Location 4 27.0 28.0 29.0 30.0 31.0 32.0 33.0 34.0 35.0 90100110120130140150160 Temperature (C)Ridge Height (um) 10 sec., 5 kip 30 sec., 5 kip 60 sec., 5 kip Location 1 27.0 28.0 29.0 30.0 31.0 32.0 33.0 34.0 35.0 36.0 90100110120130140150160 Temperature (C)Ridge Height (um) 10 sec., 5 kip 30 sec., 5 kip 60 sec., 5 kip Location 3 27.0 28.0 29.0 30.0 31.0 32.0 33.0 34.0 35.0 90100110120130140150160 Temperature (C)Ridge Height (um) 10 sec., 5 kip 30 sec., 5 kip 60 sec., 5 kip Location 5 27.0 28.0 29.0 30.0 31.0 32.0 33.0 34.0 35.0 90100110120130140150160 Temperature (C)Ridge Height (um) 10 sec., 5 kip 30 sec., 5 kip 60 sec., 5 kip Location 1 27.0 28.0 29.0 30.0 31.0 32.0 33.0 34.0 35.0 36.0 90100110120130140150160 Temperature (C)Ridge Height (um) 10 sec., 5 kip 30 sec., 5 kip 60 sec., 5 kip Location 3 27.0 28.0 29.0 30.0 31.0 32.0 33.0 34.0 35.0 90100110120130140150160 Temperature (C)Ridge Height (um) 10 sec., 5 kip 30 sec., 5 kip 60 sec., 5 kip Location 5 27.0 28.0 29.0 30.0 31.0 32.0 33.0 34.0 35.0 90100110120130140150160 Temperature (C)Ridge Height (um) 10 sec., 5 kip 30 sec., 5 kip 60 sec., 5 kip Figure 2-13. Variation in ridge height as a function of temperature and location. The center sketch shows approximately wh ere in the device the measurement was made; each of the graphs shows the ridge height as a function of temperature. The scales are identical on all 5 graphs. Comparison to the e-form itself was impossible because the profilometer stylus was too large to fit into the 30 m channels. There was only one measurement done of the

PAGE 41

31 wafer used to make the e-form, and th e wafer was destroyed during the e-form fabrication process. 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 condi tions, resulting in the variance in feature height. Nevertheless, it is clearly seen that the center of the pattern is noticeably shallower than the ends. 20 25 30 35 40 0123456 LocationRidge height (um) Unselected #1 Unselected #2 Figure 2-14. Feature size data taken from two other silicon wafers that were candidates to be the master. Each was processe d using slightly different conditions, resulting in differences betw een 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 ar e able to exhaust faster near the ends and deeper etching is observed [ 22 :107]. Much like the data presented in Figure 213, the average values for the ridge widths and separations showed very little differe ntiation with respect to temperature or

PAGE 42

32 compression rate. To simplify presentation, th ey were averaged by location, and listed in Table 2-1. Given the incrementation issues discussed earlier and th at the device was not always lined up exactly perpendicular to the st ylus 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 Ridge Widths Average 29.9 29.0 29.8 29.7 30.7 St. Dev. 0.9 0.9 1.0 0.8 0.9 Ridge Separations 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, widt h, and separation, there is little variation within each of the locations. From this, we can conclude that macro feature replication is independent of casting temper ature and compression rate. 2.3.1.3 Crazing An unexpected observation during the casti ng 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 e liminated. Initial re sults (not shown) indicated that there was no clear relati onship 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

PAGE 43

33 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 eq uilibrates at an established temperature; as the former method can lead to transient results, the latt er 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 (rela tively insulated surface), allow the device to cool all the way to room temperatur e on the metal surface of an optical bench (relatively conductive surface), or allow it to eq uilibrate 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. 020406080100 Counter Optical Bench Oven, 10 min %age of area 125 C 115 C 020406080100 Counter Optical Bench Oven, 10 min %age of area 125 C 115 C Figure 2-15. Dependence on crazed area with reference to cooling conditions. Devices were molded at temperatures shown; ove n was set to 60 C. Pictures are of representative examples used to measure crazing. Its 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,

PAGE 44

34 countertop cooling yields unacceptable re sults. Cooling on the optical bench significantly reduces the amount of area that is crazed, but con centrates it to the area near the vertical channel. Coo ling 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 temperat ure dependent, as shown in Fi gure 2-16. Thus crazing is a function of molding temperature and the method in which the device is cooled after removal from the press. 0 5 10 15 20 25 30 35 40 45 105110115120125130135 Temperature (C)%age of Area 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 2.3.1.1 also probably allows greater penetration into the scallops shown in Figure 2-11, whic h in turn creates greater amounts of damage when the device is removed from the mold. Seeing that all temperat ures that had been examined by profilometry yielded consistent ridge heights, it is recommended that the

PAGE 45

35 temperature be kept as low as possible. Th e 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. 2.3.1.4 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 am ount of concavity in it as a result of the manufacturing process (the plate that held the slides dur ing 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 de vice by increasing the amount of resin from six to eight grams a nd the increasing the cas ting 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. A B Figure 2-17. Devices molded against the glass-base d e-form. A) Typical pattern of crazing from glass-based e-form when it did occur. B) Optimized glass device showing no crazing. 2.3.2 Lamination The first attempts at lamination, usin g the laminator as supplied by the manufacturer, were not good: complete bonding could only be accomplished running the device through twice at the maximum temperat ure and lowest speed. However, there was

PAGE 46

36 concern that the channels would be deform ed during the second pass, and there was no room to alter conditions should this not wor k. 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 below. Figure 2-18. Laminator with preheating hotplate. 2.3.2.1 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 bond ing 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 ha s a Tg of approximately 70 C; as a result, the material is in it s rubbery state during the lami nation process, which may limit its ability to hold the metallic coating). While the metallic material can be removed by

PAGE 47

37 scrubbing with an acetone-soaked swab or tissue, the chips can be scratched in the process, leaving the desire for another soluti on. Other materials that were unsuccessfully tried were brass sheet, aluminum sheets and fo il, paper, and overhead transparencies; all imparted undesired textures into the devices or did not effectively transfer heat transfer during passage through the rollers. 2.3.2.2 Roller pressure The laminator has a total of six roller sp acing settings; two mean t strictly for paper or other thin objects, and four meant to accomm odate 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 st ill 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 va ry 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 us ually received using the 1/8 setting; narrower settings would occasionally result in sealed channel ends, as shown in Figure 219, while the widest setting did not always provide enough pre ssure. However, given the limited options in roller spacing available, it may be advisabl e to come up with a system that has finer adjustability. A B Figure 2-19. Comparison of devices laminated at different roller sett ings. A) 1/16. B) 1/8.

PAGE 48

38 2.3.2.3 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 inst antaneously, 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; howev er, a large amount of temporal variation is observed. The variation itself is likely due to a co mbination 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 me thod of attaching the sample to the load cell should be identified. 0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.02.04.06.08.010.012.0 Time (s)Voltage (V) 120 C 115 C 110 C 105 C Figure 2-20. Peel test results grouped by roller temperature. The rectangle indicates the range of data averaged to assemble Figure 2-21.

PAGE 49

39 The peel strength for each roller temperature was determined by averaging the first two seconds of the averages reported in Figur e 2-20, as indicated by the rectangle; these averages and their standard deviations ar e reported in Figure 2-21. There is a clear differentiation between the first three temp eratures tested, though not between 115 and 120 C. 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 105 C110 C115 C120 C Roller TemperaturePeel Force (N/cm) 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 actu al effects of lamination process on the devices. The roller temperature of 120 C cau sed some bulk deformation (bending) of the device, indicating that the ro ller and/or preheat temper atures may be too high. In addition, channels were sometimes bl ocked 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 th e milling process since they never occur in the middle of the channels, the easiest solution has been to reduce the preheat temperature to ~ 65 C 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 C increments, mo re precise control of temperatures is not

PAGE 50

40 possible using the current setup. Neither th e penalties in lamination strength nor the potential requirement of increasing rolle r temperature have been studied. Figure 2-22. Blockage that extends be yond burr location someti mes observed after lamination. 2.3.3 Milling 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 prototyp es, 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 m, even the smallest of burrs could create ma jor problems in the usability of the device. To determine which, if any, milling cond itions 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 i nvestigate from 25 to 200 mm/m in 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.

PAGE 51

41 During the cutting process, the channels can e ither be facing the cu tting tool (up) or the table (down). Both methods in theory have their advantages and drawbacks: having the channels down minimizes the time in c ontact with the cutting tool, but risks blowout 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 cha nnels, where they have the opportunity to weld back on and cause a blockage. Remova l 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 a nd 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 slot s 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 m illing may prove problematic as the channels provide exhaust paths for extruded plastic.

PAGE 52

42 Burrs were never completely eliminated fo r either the drilling or milling operations, though they were minimized in the orientati ons 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 differe nce 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 in terface is unknown, but solution was able to enter the channels of the laminated devices. A B C D Figure 2-23. Post-lamination from drilling and milling using a 2 mm-diameter end mill. Air pressure was 10 psi, spindle spee d 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.

PAGE 53

43 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 met hod our lab has selected to fill the channels prior to isoelectric focusing to add liquid th at is to fill the channel to one well, use a vacuum pump to draw the liquid through the channel to a second we ll, and then fill the wells with the anolyte or catholyte. However, when excess liquid is removed prior to addition of the appropriate elect rolyte, 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 requir ed 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-Phyc oerythrin (RPE) were selected for the separation in a pH gradient gel; their natura l fluorescence allows them to be tracked by optical detection under proper illumination.

PAGE 54

44 Figure 2-25 shows the separations that were achi eved using a constant 500 V, but different channel lengths. This in turn result s in different electric fields, with the shorter channels getting stronger fields. The result s show that the separations of these two proteins can be achieved on shor ter channels, though some of the finer detail may be lost due to compression of the peaks. 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 0 5001000150020002500 PixelIntensity (A.U.) 21 mm 32 mm 43 mm 54 mm Figure 2-25. IEF for differen t channel lengths under a cons tant 500 V. Channel lengths were 2.1, 3.2, 4.3, and 5.4 cm, resulting in fields of 238, 156, 116, and 93 V/cm respectively. 2.4 Conclusions There are a number of va riables and conditions that must be optimized to consistently produce usable microfluidic de vices. Overall device thickness and crazing are temperature dependent, thought the craz ing 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-fo rm 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 sec onds per downstroke and force of 5 kip are recommended as these allow the fastest mold ing of devices under this method.

PAGE 55

45 Milling was used to define the wells and edges of the devices used in this study. Best results when drilling with an end mill ar e achieved when the channels are facing the cutting head and air is used to remove de bris; 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 tabl e, 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 th e material and patterns used in this investigation; different materials, such as acrylics, polycarbonate, and other COCs, will have different thermal and mechanical prope rties, 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 characterizi ng the multitude of process variables has been presented, and these results should pr ovide a starting point for optimization of microfluidic devices created usi ng a compression molding process.

PAGE 56

CHAPTER 3 DESIGN, FABRICATION, AND TESTI NG OF A DEVICE FOR PROTEIN EXPRESSION 3.1 Introduction The successful expression of proteins in a laboratory setting is vital to such fundamental research as determining a protei ns 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 qu ick expression of pr oteins 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 crea ted that applies produc tion-type technology to screening applications. The original intent was to apply the resu lts 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 [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. 46

PAGE 57

47 transcription phase, ribonucle ic acid (RNA) is copied fr om selected regions of deoxyribonucleic acid (DNA) [ 57 ]. 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 ar e typically done with small quantities of material where the need exists to run ma ny experiments very quickly. For example, Roches RTS 100 system is for screening, and advertises up to 20 g 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 [ 59 ] and picoliters [ 60 ] have also been reported. While having the advantage of bei ng fast, they also have low yields, limiting the number of proteins that can be expre ssed. Production is accomplished using larger quantities that take longer to express, but with quantities much too large to run costeffective assays: Roche's RTS 500 system adve rtises a much more productive 5 mg per 1 ml of solution in 24 hours [ 58 :119]. This variance in yield is reliant on th e environment in which the proteins are grown: in closed environments such as mi crocentrifuge tubes and well plates, nutrients needed to continue the reac tion are slowly depleted and toxic byproducts of the reaction build up; as a result, 2-4 hours is the maximu m period of time experiment an experiment can endure. In vitro protein production, on the other ha nd, uses two chambers that are separated by a semi-permeable memb rane: a reaction chamber containing DNA templates, ribosomes, and an initial concen tration of nutrients, and a feeding chamber that houses additional nutrient s; exchange of nutrients and waste products across the membrane prolongs experiment duration and in creases yield. This technology has been licensed for use in the United States by Roche [ 58 :12, 61 ].

PAGE 58

48 Creating a fluid exchange mechanism in a well plate-sized chamber should improve the protein yield without si gnificantly increasing reagent consumption, allowing the use of systems that are difficult to expr ess the microarray format, such as -glucuronidase (GUS), for assays, and fabricating such a device is the goal of this chapter. 3.3 Experimental After examining several designs, we sett led upon one having two wells, one nested inside the other. The cont ents 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 Fi gure 3-1. The smaller, inner chamber would contain the DNA vectors, RNA, proteins, and an initial ch arge of nutrients, and be referred to as the reaction chamber. Th e 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 dialys is membrane to the upper piece to create the reaction chamber. A insert or tray well plate reaction chamber feeding chamber membrane B Figure 3-1. Illustrations of well-device. A) Three dimensional B) cross sectional. 3.3.1 Adhesives Three brands of cyanoacrylate adhesives were initially considered: Krazy Glue (Elmers Products, Columbus, OH), Super Glue (Loctite, Avon, OH), and Vetbond (3M

PAGE 59

49 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 th e autoclaving process, two types of epoxy, 302-3M and 353ND-T (bot h of Epoxy Technologies, Billerica, MA), were acquired, mixed and cured per the manufacturers 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 manufacturers recommendations. A small piece of each dried adhesive was placed in a microcentrifuge tube, covered with 50 l 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 in to 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 sl ab, which was then placed in the separation cell and voltage applied. When separation wa s completed, the gel was removed from the cell, de-stained and rinsed to remove ex cess ethidium bromide, and imaged using a luminometer. 3.3.2 Device Fabrication As shown in Figure 3-1, the finalized desi gn 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 In ternational, 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

PAGE 60

50 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 cure d for 20 minutes at 95 C 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 acryl ic (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 diam eter, 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 (C AT) was expressed using an RTS 500 kit (Roche Diagnostics GmbH, Mannheim, Germany) Reaction and feeding solutions were prepared as recommended by the manufacturer. 7 l of reaction solution were added to each of six chambers in an insert and five microcentrifuge tubes, while 70 l 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 evaporat ion. The tray and tubes were placed on a

PAGE 61

51 shaker at room temperature, with samp les being taken at after 2, 6, 10, 14, 20, and 24 hours until samples were depleted. Th e samples were separated via sodium dodecylsulfate-polyacrylmide gel electrophores is (SDS-PAGE), and then analyzed using western blotting. To summarize the blotting process, polyvinyl di fluoride (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: n itro-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 tack le 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 contai n 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 Screenings standa rds for 384and 96-well plates [ 62 ], respectively, which in turn would allow finalized arrays to be analyzed using existing microplate readers.

PAGE 62

52 Buffer Reaction chamber Buffer Reaction chamber A B Figure 3-2. Original concept for device. A) C onceptual 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 occu r, we decided to minimize the channel dimensions to both reduce the amount of solution used and en able the use of electroosmotic flow to drive the convective process. Consistently milling channels of adequate smallness proved to be diffi cult, and this approach was abandoned. The second iteration was to create small r eaction 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 fa bricated 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 thes e donuts into each well of a standard 96-well plate, the donuts in theory could also be placed in a large common bath containing

PAGE 63

53 feeding solution. However, the stability of these devices was less than desirable; the second version added a flange whic h could rest on a lip. Extens ion 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, th e materials used to make the device had to meet a number of requirements. First, they had to be ab le to endure aqueous environments for prolonged periods of time. Second, they had to surv ive the ster ilization process (originally intended to be autocl aving, 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 us ed in contact lenses and bone cement [ 63 ], so its biocompatibility was assumed, along with that of the dialysis membrane. However, the biocompatibility of adhesives is less doc umented, necessitating our own testing. PCR is a method of quickly replicati ng DNA segments, and was selected to measure biocompatibility because the m echanism 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 act ually expressing the pr oteins. 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 la ne, and serves as a ruler to calibrate the molecular size of any products th at appear; the number of base pairs in each band appears to the left of each picture. The other lane s 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 th e amount of PCR product present. As a result, a dimmer signal indicated that the glue inhibits the reaction mo re, indicating that it should be avoided for device fabrication (in order to maximize the protein yields, it is

PAGE 64

54 imperative that the device itself has minimal, if any, negative effect on the transcription or translation process). It was found that all adhesives test ed exhibited some amount of inhibition to the PCR results. -179 -396 -517 bp -2,645 -1,605 -460 -222 -676 -1 198 -126 -350 -75 DNA Positive Krazy Glue Vetbond Super Glue A -179 -396 -517 bp -2,645 -1,605 -460 -222 -676 -1,198 -126 -350 -75 DNA Marker Positive 302-3M 353ND-T B Figure 3-4. PCR results for adhe sives considered in this st udy. A) cyanoacrylates. B) epoxies. Of the three cyanoacrylate glues tested, Vetbond inhibite d the reaction the least, indicating that it would be the best of those choices. Unfortunately, none of these adhesives survived the autoclaving process (use d for sterilization at the time), and the low viscosity of the adhesive also made a pplication of consistent amounts of glue problematic. Two types of epoxies were examined as replacement adhesives. 302-3M has reportedly met U.S. Pharmacope ia (USP) class VI standards for biological compatibility; 353ND-T has not been met those standard s, but a related system, 353ND, has [ 64 ]. Neither of these adhesives completely inhibi ted 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

PAGE 65

55 limiting the circulation and mixing during the PCR process. We proceeded with 353NDT due to its viscous nature, which al lowed the formation of a better film. One design issue of note is that despit e our doubling of the manufacturers cure recommendation for that temperature, th e epoxy doesnt reach full cure until after sterilization. Figure 3-5 show s 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 turn s to a dark amber colo r; this color is an indication of level of cure. 1 23 Figure 3-5. Demonstration of incomplete cu ring of epoxy during inse rt fabrication. 1) shows results after initial cure and 2) shows results after UV sterilization. 3) Bulk epoxy that underwent same cure conditi ons as 1), but no sterilization. The fact that #1 is not fully cured is belie ved to be a result of the quantity of epoxy being too small: epoxy cures via an exotherm ic reaction, and harvests energy both from the external environment and the heat genera ted by the reaction its elf; 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 c onditions as #1; the resulting product is labeled #3.

PAGE 66

56 3.4.3 Protein Expression The primary goal of this system was to e xhibit 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 inte nsities are similar after two hours, but where the microcentrifuge tubes 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 conve ntional microcentrifuge tubes. This shows that the device does indeed produce higher yields than the closed environments currently used for microarrays. 0 3000 6000 9000 12000 15000 18000 0481216202428 Incubation time (h)Signal intensity (A.U.) Tray Device Microcentrifuge tube Figure 3-6. Comparison of expre ssion of CAT between microcentrifuge tube and device. 3.5 Conclusions A small-scale, dual chamber device for pr otein 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 stronge r than the conventional means of closed reaction vessels (microcentrifuge tu bes or conventional well plates).

PAGE 67

57 The system presented uses a total of 77 l of solution; while this compares favorably with the Roche 500 system which consum es 1 ml, is still at least two orders of magnitude larger than the single chambere d nanoand 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 reduc tion in the volume may be possible (i.e. to 384or 1536-well plate standa rds), liquid migration by tr aveling along the wetted out surfaces of the tray and well plate will proba bly become a limiting factor of this design.

PAGE 68

CHAPTER 4 CONCLUSIONS AND FUTURE RECOMMENDATIONS The applications and needs for micr ofluidic lab-on-a-chip technology are increasingly growing. This thesis has developed and ch aracterized two fabrication methodologies: a compression molding process that can be applied to any laboratoryscale exploratory work of new designs, and a machining process used to create a device for the specific application of protein expression. Presente d here the overa ll conclusions, as well as recommendations for aspect s that may deserve further study. 4.1 Compression Molding Process Optimization A compression molding process for fabrica ting thermoplastic microfluidic devices has been characterized. The to tal process consists of moldi ng the device, using a mill to define reservoirs and edges, and laminating with film of identical material to seal the channels. During the molding process, replication of bulk features (ridges /channels) appears to be independent of molding conditions ove r the range that was tested, but device thickness and degree to which cr azing occurred are not. Wh ile 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 temper ature 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 mi nimize this behavior. One aspect that wasnt 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, 58

PAGE 69

59 the milling of edges and holes results in a final device containing approximately three grams of material. That means for every devi ce created, at least half of the material will be wasted, assuming it isnt being recycled. One possible solution is to modify the mold to create a set of walls around the pattern th at 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 [ 45 ]. In milling operations, burr formation was ne ver 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 sp eeds and feed rates tested. Milling channellike reservoirs worked better when the channe ls were down (faci ng the table); spindle speed and feed rates should be high, and clim b milling should be avoided. In all cases, active removal of debris should be empl oyed; 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 materi al to ~10-15 C below the Tg of the material. The rollers had a tendenc y to impart a texture in the material, which was best alleviated by sandwiching the devi ce 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 [ 65 ] used in the composite industry may provide suitable replacements. The strength of the bond between film and device was tested by measuring

PAGE 70

60 the force required to peel the film off, a nd higher roller temperatur es resulted in higher peel strengths. Sagging of the film into the channels during the lamination process was not investigated. This behavior could significan tly decrease the volume of the channels. The degree to which this occurs is probably depe ndent on the preheat temperature and roller temperature and pressure. Investigati on 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 temp erature, and then optimizing the roller temperature and pressure using only that optimized value, due to the inter-relationship of the three. 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 re action 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 96well plates could be substituted. This allows the use of commercially-available wellplate readers to measure the protein expre ssion yield directly in the device should a suitable labeling system be developed. Well sizes and alignments corresponding to 384and 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 factors. The presented device was shown to exhibit significantly greater levels of protein production than the closed environments of microcentrifuge tubes. The speeds of

PAGE 71

61 expression were comparable in both the open an d closed systems; further research should be devoted to speeding up this process, perhaps through adjusting the frequency and displacement of shaking or usi ng an oscillating pressure to more actively circulate the feeding solution. If this general design is kept, future iterations should have pins and corresponding sockets to assist in the alig nment of the two pieces. Us e of an adhesive to bond the reaction chamber insert to the feeding chambe r 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 di mensions, which correspondingly have small tolerances for error. While th e 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.

PAGE 72

APPENDIX 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 hol der of a CNC mill, and then use that automation to control the pull rate and dire ction. 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 [ 66 ]. A.1 Design 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 th e sides. As the load is applied over the notch, the intact edge elongate s. The ratio of change in le ngth over original length is called strain, a 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 [ 67 ] 62

PAGE 73

63 SRdRa (1) where dR is the change in resistance, S is the strain gage fact or (constant and unique to wire material) and R is the original resistance of the st rain gage. The change in resistance dR is typically very small, and is best meas ured via a Wheatstone bridge, which consists of three static resistors, on e 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. Vs VoR3R1R2R4 Vs VoR3R1R2R4 VoR3R1R2R4 Figure A-2. Circuit diagram for Wheats tone bridge used to collect data. Ideally, V o would equal zero at the start of the te st, which in turn requires that the ratios of R 1 / R 4 and R 2 / R 3i 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, V o and change in resistance is 3 2 32 2R RR RV Vi s o (2) (Several primers on this circuit have been written, so a development of this equation will not be included). Approximating R 3 as dR from the strain gage equation gives S RR RR VVa i i so2 32 32 (3)

PAGE 74

64 As strain a cannot be controlled and gage factor S is constant, the most sensitivity is achieved when the first two terms are maximized; V s is typically limited by the strain gage and power supply, while the resistance term reaches a maximum value of 0.25 when R 2 equals R 3i A.2 Construction Mr. Nakka has conveniently provided a spreadsheet [ 68 ] which will give the expected microstrain measurement after input ting the expected force, dimensions, and modulus of elasticity. This was used to determine the si ze 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-t hick acrylic using a CNC mill. A Tshaped piece with all legs approximately 0.25 wide was cut from 0.025-thick copper sheet and secured to the side opposite the stra in 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 m ills cutting head. The strain gage (cea-06250un-350, Vishay Micro Measurements) and 30-g auge magnet wire were attached using epoxy; the wire was soldered to the leads of the strain gage after the epoxy had set.

PAGE 75

65 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 mills 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. A B 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 350 350 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) inte rfaced with a personal computer running Data Logger (Tenma) software. For an actual peel test, the holder is attach ed to the tooling plate of the mill. The load cell is secured in the collet that norma lly 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 pr ocess, creating a flap

PAGE 76

66 of film between and long. The device is placed underneath the arms with the flap coming up from between them; the flap is secu red 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 a ddition to space constraints, clamps of the size needed here are likely to focus the for ces on small areas of the film, potentially leading to tearing). The load cell is then m oved straight up at the desired rate, pulling the peeling film with it. A.3 Calibration A cup was clipped to the copper tab, and wa ter was added 5 ml at a time. Voltage drop across the bridge was recorded. A calib ration curve relating voltage drop across the bridge and force was created assuming the de nsity of water was 1 g/ml; this is shown below in Figure A-5. y = 1.1043x + 2.399 R2 = 0.9962 2.0 2.5 3.0 3.5 4.0 4.5 5.0 0 0.5 1 1.5 2 Force (N)mV Figure A-5. Calibration curve obtained for load cell.

PAGE 77

LIST OF REFERENCES 1. Curie, P., Curie, Mme. P., and Bemont G., 1898, On a new, strongly radio-active substance contained in pitchblende, Comp tes rendus de l'Academie des Sciences, 127, pp. 1215-1217; translation fr om Romer, A., ed., 1970, Radiochemistry and the Discovery of Isotopes 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, http://masspec.scripps.edu/informati on/history/perspectives/borman.html last accessed 4 November 2005. 5. Borman, S., Dagani, R., Rawls, R.L., a nd Zurer, P.S., 12 January 1998, Chemistry crystallizes into modern 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, Prospect s for Miniaturization of Mass Spectrometry, Executive Summary, Arlington VA, http://www.nsf-mass-specmini-forum.umd.edu/final_report.html last accessed 4 November 2005. 8. Harris, D.C., 1987, Quantitative Chemical Analysis 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 separati on systems: a developing perspective in chemical monitoring, Adv. Chrom., 33, pp. 1-66. 67

PAGE 78

68 13. Manz, A., Graber, N., Widmer, H.M., 1990, M iniaturized 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 system s integrated on a planar glass chip, Anal. Chem., 64(17), pp. 1926-1932. 16. Verpoorte, E., 2002, Microfluidic chips fo r clinical and forensic analysis, Electrophoresis, 23, pp. 677-712. 17. Reyes, D.R., Iossifidis, D., Auroux, P.-A., and Manz, A., 2002, Micro total analysis systems. 1. Introduction, th eory, and technology, Anal. Chem., 74, pp. 2623-2636. 18. Auroux, P.-A., Iossifidis, D., Reyes, D.R. and Manz, A., 2002, Micro total analysis systems. 2. Analytical standa rd operations and applications, Anal. Chem., 2002, 74, pp. 2637-2652. 19. Vilkner, T., Janasek, D. and Manz, A., 2004, M icro total analysis systems. Recent developments, Anal. Chem, 76, pp. 3373-3386. 20. Teichroeb, R., 22 July 2004, Rare look insi de state crime labs reveals recurring DNA test problems. Seattle Post-Intelligencer, http://seattlepi.nwsource. com/local/183007_crimelab22.html Last accessed 4 November 2005. 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. 22. Madou, M., 2002, Fundamentals of Microfabrication 2nd Ed., CRC Press, Boca Raton. 23. Chen, H., Acharya, D., Gajraj, A., Me iners, J.-C., 2003, Robust interconnects and packaging for microfluidic elastomeric chips, Anal. Chem., 75(19), pp. 52875291. 24. McDonald, J.C., Duffy, D.C., Anderson, J.R., Chiu, D.T., Wu, H., Schueller, O.J.A., and Whitesides, G.M., 2000, Fabri cation of microfluidic systems in poly(dimethylsiloxane), Electrophoresis, 21, pp. 27-40.

PAGE 79

69 25. Kai, J., Sohn, Y.S., and Ahn, C.H., 2003, P rotein microarray on cyclic olefin copolymer (COC) for disposable protein lab-on-a-chip, in Proceedings of uTAS 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 Total Analysis 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-macplastics.net/data%20sheet s/celcon_data_sheet.htm Last accessed 4 November 2005. 31. Topas Product Brochure 351 D BR 08.2000, August 2000, Ticona GmbH, Frankfurt, Germany. 32. Information about High Technology Silic one Materials, 1991, Dow Corning, Midland, MI. 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 of SPIEThe International Society for Optical Engineering, 41 77(Microfluidic Devices and Systems III) Sept. 18-19, Santa Clara, CA, pp. 200. 36. Martynova, L., Locascio, L.E., Gaitan, M ., Kramer, G.W., Christensen, R.G., MacCrehan, W.A., 1997, F abrication 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 of uTAS 2003 Seventh International Conference on Mi cro Total Analysis Systems Oct. 5-9, Squaw Valley, CA, pp. 959-962.

PAGE 80

70 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 fa brication 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 embo ssing methods and their applications for DNA separation and detection, Sens. Act. B, 75, pp. 142-148. 40. Becker, H. and Heim, U., 2000, Hot embo ssing as a method for fabrication of polymer high aspect ratio st ructures, 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, pp. 3638-3645. 42. Xu, J., Locascio, L., Gaitan, M., a nd Lee, C.S., 2000, Room-temperature imprinting method for plastic micr ochannel fabrication, Anal. Chem 72, pp. 1930-1933. 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 micropatterning protocols, Anal. Chem., 77, pp. 5414-5420. 44. McCormick, R.M., Nelson, R.J., AlonsoAmigo, M.G., Benvegnu, D.J., Hooper, H.H., 1997, Microchannel elect rophoretic 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. 411-412. 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. 3689-3693. 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.

PAGE 81

71 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, CO 2 laser microfabrication of an integrated pol ymer microfluidic manifold for the determination of phosphorus, Lab Chip, 3, pp. 221. 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, pp. 1935-1941. 52. Bundgaard, F., Nielson, T., Nilsson, D., Shi, P., Perozziello, G., Kristensen, A., and Geschke, O., 2004, Cyclic olefin copolymer (COC/Topas ) an exceptional material for exceptional lab-on-a-chip systems, in Micro Total Analysis Systems 2004: Proceedings of UTAS 2004, 8th International Conference on Miniaturized Systems for 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 elect rophoretic 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., Ayn, A.A. Zhang, X., and Spearing, S.M., 2002, Effect of process parameters on the surface morphology a nd 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., a nd 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 Ce ll-Free Protein Expression 2003, Roche Diagnostics, Indianapolis. 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., K obayashi, M., Yamaura, S., Hasan, Q., Morita, Y., Nakano, H., Yamane, T., and Tamiya, E., 2004, A picoliter chamber array for cell-free protei n synthesis, J. Biochem 136, pp. 149-154.

PAGE 82

72 61. Alakhov, J.B., Baranov, V.I., Ovodov, S.J., Ryabova, L.A., Spirin, A.S., and Morozov, I.J., 1995, Method of preparing po lypeptides in cell-free translation system, U.S. Patent 5,478,730. 62. For microplates well positions, Society for Biomolecular Screening Standard ANSI/SBS 4-2004. 63. Eschbach, L., 2000, Nonresorbable polymers in bone surgery, Injury, 31 (D), pp. 22-27. 64. USP class 6 approved, implants, Epoxy Technologies, http://www.epotek.com/subappl ications.asp?ID=17&SID=18 Last accessed 4 November 2005. 65. Products, Airtech International, Inc., http://www.airtechonline.com/products/index.htm Last accessed 4 November 2005. 66. Strain gage load cell for thrust measurement, Richard Nakkas Experimental Rocketry Site, http://www.nakka-rocketry.net/strainlc.html Last accessed 4 November 2005. 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 Nakkas Experimental Rocketry Site, http://members.aol.com/nakkarocketry/softw.html Last accessed 4 November 2005.

PAGE 83

BIOGRAPHICAL SKETCH Carl Fredrickson grew up in Kent, Wash ington. In June of 2002, he earned a Bachelor of Science degree in chemical engi neering from the University of Washington. He enrolled in the University of Florid a to pursue a Master of Science degree in mechanical engineering in August 2003. 73