Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Alternative replica molding methods for polymer based microfluidic channels Coquinco, Bernard 2014

Warning
You are currently on our download blacklist and unable to view media. You will be unbanned within an hour.
To un-ban yourself please visit the following link and solve the reCAPTCHA, we will then redirect you back here.

Item Metadata

Download

Media
24-ubc_2014_september_coquinco_bernard.pdf [ 1.66MB ]
Metadata
JSON: 24-1.0166937.json
JSON-LD: 24-1.0166937-ld.json
RDF/XML (Pretty): 24-1.0166937-rdf.xml
RDF/JSON: 24-1.0166937-rdf.json
Turtle: 24-1.0166937-turtle.txt
N-Triples: 24-1.0166937-rdf-ntriples.txt
Original Record: 24-1.0166937-source.json
Full Text
24-1.0166937-fulltext.txt
Citation
24-1.0166937.ris

Full Text

Alternative Replica Molding Methods for Polymer Based Microfluidic Channels by  Bernard Coquinco  BASc, The University of British Columbia, 2006  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF APPLIED SCIENCE in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Biomedical Engineering)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  August 2014  © Bernard Coquinco, 2014 ii  Abstract  Microfluidics provides an opportunity to create low cost devices that can potentially contain many elements of a diagnostics lab on a single chip.  While the cost of the finished product may be low, a common method of fabricating microfluidic devices such as soft lithography can be expensive to prototype due to the use of photolithography equipment meant for the semiconductor industry.  In addition, the majority of microfluidic research has been done using rectangular channels but in some cases the ability to make circular cross-section channel microfluidic devices would be very useful.  For areas such as modelling cardiovascular flows, investigating micro flow cytometry and inertial particle focusing, the ability to create circular channels could provide improvement over the use of rectangular channels.  To address these issues, an ultra low cost method of making silicon molds patterned with SU-8 has been developed as well as a method to create circular microfluidic channels via hot embossing and double casting techniques in both thermoplastic materials and PDMS. This hot embossing based method to create round channels allows for the rapid creation of straight and curving round channels in PMMA and other plastics as well as a method to create PDMS round channels using soft lithography.   iii  Preface  For all chapters the design of the research program, all testing, analysis of data, and collection of images was performed by me except as specified below.   A portion of Chapter 2 was published as part of the CanCam 2011 proceedings:  B. Coquinco and E. Lagally, "Ultra Low-cost Microfluidic Fabrication Technologies," in 23rd Canadian Congress of Applied Mechanics 2011,Vancouver, B.C., pp  280-283.  All experimental work for the CanCam 2011 proceedings was performed by me as well as half the writing. The other half of the writing was done by Dr. Eric Lagally.  All work and images on the 3D printed channels created using the Stratasys Objet 30 3D printer in Chapter 3 was performed by Anna Lee.  Designs for the channels tested were based on previous designs created by Michael Winer.  SEM images in Chapter 3 (Figures 17, 18, 19) were taken by Derrick Horne of the UBC BioImaging Facility. iv  Table of Contents  Abstract .............................................................................................................................. ii Preface ............................................................................................................................... iii Table of Contents ............................................................................................................. iv List of Tables ................................................................................................................... vii List of Figures ................................................................................................................. viii List of Abbreviations .........................................................................................................x Glossary ............................................................................................................................ xi Acknowledgements ......................................................................................................... xii Dedication ....................................................................................................................... xiii Chapter 1: Snapshot of Microfluidic Fabrication Technologies ...................................1 1.1 Polymer Microfluidic Device Fabrication ...........................................................1 1.1.1 Photolithography ..............................................................................................3 1.1.2 Soft Lithography ..............................................................................................4 1.1.3 Hot Embossing .................................................................................................5 1.2 Why Circular Cross-section Channels? ...............................................................7 1.3       Thesis Goals and Research Objectives ................................................................9 Chapter 2: Alternative Fabrication Techniques ...........................................................10 2.1 SU-8 Mold Fabrication ......................................................................................10 2.1.1 Procedure .......................................................................................................10 2.1.2 Results ............................................................................................................14 2.2 Deep-UV Etching of PMMA .............................................................................15 v  2.2.1 UV-Ozone Etching of PMMA .......................................................................17 2.3 Cast Epoxy Molds ..............................................................................................21 Chapter 3: Fabrication of Circular Cross-Section Microfluidic Channels ................24 3.1 Hot Embossing of Wire Based Microfluidic Channels......................................27 3.1.1 Fabrication .....................................................................................................28 3.1.1.1 Straight Channels ...................................................................................29 3.1.1.2 Serpentine Channels...............................................................................32 3.1.1.3 Spiral Channels ......................................................................................33 3.1.1.4 Variable Width Channels .......................................................................34 3.1.1.5 Joining Channels ....................................................................................34 3.2 Bonding of Round Channel Microfluidic Devices ............................................35 3.2.1 Alignment Techniques ...................................................................................35 3.2.2 Bonding Method ............................................................................................37 3.3 Fabrication of PDMS Round Channels..............................................................40 3.3.1 Emboss and Cast Method...............................................................................40 3.4 Challenges ..........................................................................................................43 3.4.1 Levelness of Samples .....................................................................................43 3.4.2 Achieving Sharp Edges on Embossed Features .............................................43 3.4.3 Reducing Cracking During Bonding .............................................................48 3.4.4 Increasing Long Term Bonding of Channels .................................................50 3.5 Results ................................................................................................................51 3.5.1 Embossed Channel Geometry ........................................................................51 3.5.1.1 Wire Geometry.......................................................................................51 vi  3.5.1.2 Channel Cross Sectional Profile ............................................................52 3.5.1.3 Tangent Chord Angle .............................................................................55 3.5.1.4 Channel Surface Quality ........................................................................55 3.5.1.5 Bonding Reproducibility and Alignment Accuracy ..............................58 3.5.1.6 Cast PDMS Round Channel Cross Sectional Profile .............................61 3.5.1.7 Curved Channel Final Geometry ...........................................................62 3.5.1.8 Spiral Channels ......................................................................................65 3.5.1.9 Variable Width Channels .......................................................................67 3.5.1.10 Joined Channels .................................................................................68 Chapter 4: Conclusion .....................................................................................................71 4.1 Strengths and Limitations ..................................................................................71 4.2 Potential Applications ........................................................................................74 4.3 Future Research .................................................................................................75 References .........................................................................................................................77 Appendix A Multiple Height PCB Based Molds ...........................................................82  vii  List of Tables Table 1 Resulting thicknesses for the various spin times and photoresists used .................... 14 Table 2 Summary of embossing experiments to find optimal embossing parameters ........... 47 Table 3 Different feature heights on two circuit boards ......................................................... 84  viii  List of Figures Figure 1 Photolithography process flow ................................................................................... 3 Figure 2 Soft lithography process flow ..................................................................................... 4 Figure 3 Hot embossing process flow ....................................................................................... 6 Figure 4 DC fan with silicon wafer mounted for photoresist spinning................................... 11 Figure 5 Finished patterned SU-8 on using the ultra low cost method on a 2" silicon wafer 13 Figure 6 Example showing the general deep-UV etching process ......................................... 16 Figure 7 Example process flow for creating an epoxy mold and embossing ......................... 21 Figure 8 PDMS channels created using a Stratasys Objet 30 ................................................. 27 Figure 9 Methods of embossing a wire to create a microfluidic channel ............................... 28 Figure 10 Process to assemble the straight round channel sample for embossing ................. 29 Figure 11 Material stack used when embossing a round channel ........................................... 31 Figure 12 Process to assemble a curved channel sample prior to embossing ......................... 33 Figure 13 Example depicting how a spacer material would be used to coil a wire around a cylinder to create a spiral ........................................................................................................ 34 Figure 14 Process for building the straight wire with alignment marker stack ...................... 36 Figure 15 Vacuum assisted solvent bonding setup ................................................................. 39 Figure 16 Process flow for double casting an embossed PMMA sample into PDMS ........... 42 Figure 17 Images of channels embossed with various parameters ......................................... 47 Figure 18 A wire end being imaged for measuring the wire diameter ................................... 52 Figure 19 Image indicating how the amount of corner rounding is calculated ...................... 54 Figure 20 Example of a tangent chord angle measurement. ................................................... 55 ix  Figure 21 SEM image of a bonded round channel that shows an unusual pattern on the interior of the channel ............................................................................................................. 56 Figure 22 SEM images of an embossed PMMA half and the wire used to create it. ............. 57 Figure 23 SEM image of a cleaned 250 µm wire.   ................................................................ 57 Figure 24 Bonded PMMA round channel that was embossed using 101.6°C 14 MPa for 20 minutes .................................................................................................................................... 60 Figure 25 Bonded PMMA round channel that was embossed using 96.1°C, 14 MPa for 20 minutes .................................................................................................................................... 60 Figure 26 Bonded PDMS round channel ................................................................................ 61 Figure 27 Curved PDMS round channels ............................................................................... 63 Figure 28 Image of the 1 mm 3D printed forms clamped together under a microscope ........ 64 Figure 31 Variable width channel formed by protecting a wire with an etch resist pen and etching. .................................................................................................................................... 68 Figure 33 Channel joined by overlaying wires ....................................................................... 70 Figure 34 Process flow to go from the initial embossing process to mass produced polymer round channels ........................................................................................................................ 72 Figure 35 Proposed embossing procedure to quickly create molds for embossing PMMA or casting PDMS round channels ................................................................................................ 76 Figure 36 Possible layer combinations when using a PCB as a multilayer mold ................... 84  x  List of Abbreviations CAD Computer Aided Design COC Cyclic olefin copolymer  EGDMA Ethylene glycol dimethacrylate DI H2O  Deionized water IPA Isopropanol PC Polycarbonate PCB Printed circuit board PDMS Polydimethylsiloxane PID Proportional Integral Derivative PMMA Polymethyl methacrylate PLA Polylactic acid PUMA Polyurethane methacrylate ROHS Reduction of hazardous substances RPM Revolutions per minute TCA Tangent chord angle Tg Glass transition temperature TPE Thermoset polyester UVO Ultra violet ozone  xi  Glossary FR4 - A type of glass reinforced epoxy laminate that is typically used to create circuit boards Platen - The metal plates in a hot embossing system that are used to apply heat and pressure to the sample   xii  Acknowledgements  I would like to thank my supervisor Dr. Karen Cheung for the opportunity to complete this research as well as her guidance, patience, and support throughout this adventure.  It would not have been possible to complete this without her help.  I would also like to thank Dr. Eric Lagally for letting me just go and try things.  I would also like to thank my other committee members Dr. Hong Ma and Dr. Konrad Walus who served on my defense committee for their critical reading of this thesis.  I would also like to thank Dr. Hong Ma for allowing me to use the hot embossing equipment.  A large portion of this work would not have been possible without it.  Last but not least, I would also like to thank all the various students, staff, and professors  I have met along the way.  Without you all this would not have been as interesting and rewarding as it has been. xiii  Dedication    To my parents for their unconditional love and encouragement 1  Chapter 1: Snapshot of Microfluidic Fabrication Technologies Microfluidics provides the opportunity to reduce costs by reducing the volume of the reagents needed and provides the potential to shrink an entire lab onto a chip.  Various fluidic components such as pumps, valves and mixers have already been designed and many diagnostic techniques such as electrophoresis and electrochemical detection have been proven to work on these devices [1]. However, while the end result can lead to low cost diagnostic tools, the upfront costs in equipment, maintenance and tooling to build these chips can be quite expensive as many of these come from other fields such as the semiconductor industry.    1.1 Polymer Microfluidic Device Fabrication There are many different processes to create microfluidic devices such as injection molding, soft lithography, hot embossing, roller embossing, conventional machining, laser ablation and wet etching that are used depending on the desired substrate material and other factors such as cost, speed to fabricate, and availability of equipment. Many disposable microfluidic devices are made of polymers, and fabrication methods can be grouped into three main categories: Polymer thermoforming, polymer ablation and polymer casting [2].   Polymer ablation encompasses techniques such as laser ablation and mechanical milling where the polymer substrate is directly altered.  While these methods can be used to for very quick rapid prototyping, mass production is limited due to the need to process each sample serially.  Polymer thermoforming includes techniques such as hot embossing and injection molding.  Injection molding typically has high tooling and equipment cost but is very high 2  throughput while hot embossing is more widely used for rapid prototyping as the same type of molds used in polymer casting can be used. Polymer casting includes the most commonly used technique of soft lithography which does not need as much equipment as hot embossing and injection molding.  What both hot embossing and soft lithography have in common is that they both typically start with making a mold using photolithography. [2]    3  1.1.1 Photolithography Photolithography is a method taken from the semiconductor industry that works by exposing a type of photopolymer called a photoresist to light in order to create a pattern.  In the case of microfluidics, typically a pattern is created in a computer aided design (CAD) tool and is then printed in high resolution onto a photomask.  A silicon wafer is then uniformly coated with a photoresist and the thickness of the photoresist coating dictates the height that the final microfluidic channel.  The photoresist is then patterned with the photomask using UV light.  Once the exposure step is complete, the silicon wafer is placed in a developer solution to form the pattern.  If positive photoresist is used, the areas that have not been exposed to UV light will remain after development and the opposite will occur for negative resist. For soft lithography, SU-8 (Microchem Corp.) is a commonly used.  Figure 1 Photolithography process flow   SubstrateCoat with photoresistAlign mask and exposeResult for positive resistResult for negative resist4  1.1.2 Soft Lithography Once the mold has been made, uncured PDMS is now poured on the mold, degassed and baked.  Once this is complete, the PDMS piece can be peeled off and bonded to a substrate such as glass or even another piece of PDMS to form the final channel [3].  While PDMS is the typical material used in soft lithography, other materials such as thermoset polyester (TPE), polyurethane methacrylate (PUMA) and Norland Optical Adhesives 81 have been used to provide different mechanical, chemical, and optical properties [2].  Another way to take advantage of different material properties and reduce fabrication costs would be to use plastics using a method such as hot embossing.  Figure 2 Soft lithography process flow   Mold masterPour PDMS on mold and cureRemove PDMS from moldBond PDMS to substrate5  1.1.3 Hot Embossing As with soft lithography, hot embossing starts with making a mold.  Typically these molds are silicon molds made in the same manner as soft lithography, polymer or epoxy based molds [4] but other types of molds such as metal molds [5] can be used as well.  Unlike soft lithography, hot embossing relies on heat and pressure to form thermoplastic pieces to the shape of the mold.  To do this, the mold and the sample are placed between the platens of the embossing system and everything is heated to the desired embossing temperature which is typically above the glass transition point (Tg) of the polymer to be embossed.  Once the temperature reaches the target, the substrate is compressed by the platens with the desired amount of force while the temperature is held constant.  In more expensive systems, vacuum pressure is used simultaneously to ensure the mold is filled completely. Once the specified dwell time is completed, with the force still applied, the substrate is cooled to below Tg and then the sample can be removed.[6] 6   Figure 3 Hot embossing process flow  Using hot embossing to create microfluidic channels can greatly increase the speed at which devices can be made as the thermoplastics used do not require degassing and curing steps as with soft lithography.  However, as both soft lithography and hot embossing require the use of a mold, the use of photolithography hinders the creation of other channel geometries such as round channels.    Mold master placed in embossing system with thermoplastic substrateCompressed with heat and forceCool down then force removedBondSubstrate removed from mold7  1.2 Why Circular Cross-section Channels? Due to the nature of the photolithographic process, it creates rectangular channels.  Circular cross-section channels, otherwise called round channels, have different properties that make them useful in a wide variety of specific applications.  One of the advantages of a round channel is that it is the optimal geometry for generating fixed flow rates at minimal pressure compared to a rectangular channel of similar dimensions [7].  This, along with the circular geometry, allows for immobilizing cells while limiting the chance of cell lysis [8].  Round channels are also very useful in creating microvessel scaffolds  as the round shape creates a similar velocity profile to real microvessels that enables uniform oxygen and nutrient transport to the cells being grown [9].  It is also useful for mimicking cardiovascular flow conditions as it replicates the round profile of blood vessels.  Besides creating similar velocity profiles, a round channel also is able to mimic how cells will conformally plug a vessel when the cells are larger than the dimensions of the blood vessel [10].  Microvascular networks also benefit from round channels as rectangular channels generate varying shear stresses depending on the cell position in the channel [11].Rectangular channels also influence cell growth as the geometry differs from physiological conditions.  These differences in geometry and shear stresses can influence gene expression, alignment and differentiation [8].  Besides biological applications, round microchannels have many other uses.  One of these uses is to increase light transmission efficiency through on-chip waveguides [8]. Another application is to provide better microfluidic valve sealing [12]. A microfluidic valve has a 8  both a control layer and a flow layer with a membrane sandwiched in between.  When the control layer is pressurized, the membrane is deformed to block the flow layer.  As the channel is rectangular and the deformed membrane is round, there is a geometric mismatch between the two and the best way to make a better seal would be to use a half round channel[12].  Circular cross-section channels could also be used for testing particle flow focusing geometries. The ability to focus particles is highly desired as it would allow for particle separation based on particle size [13] and simplify flow cytometry by creating focused particle streams with defined positions.  Particle focusing and cell separation have been shown in straight [14], serpentine [15][16] (or curved) channels, and spiral channels [17], with rectangular and trapezoidal cross-sections [18], but particle focusing in circular cross-section channels have not been widely studied beyond the original work by Segré and Silberberg [19] . While straight round channels can be tested in a relatively straightforward manner due to the availability of glass capillaries and the wire pulling method described in [20], curved geometries are much more difficult to create.  By utilizing a wire to create the form, many channel configurations can be tested including spirals and curving serpentine channels. As long as a method can be developed to bend the wire to the desired shape it is possible.      9  1.3 Thesis Goals and Research Objectives The aims of this thesis are as follows:  To investigate the development of lower cost methods of fabricating microfluidic devices  To develop a method to create polymeric, circular cross-section microchannels  The main novel contributions in this thesis are the development of an ultra low cost photolithography method and the development of a hot embossing based method of creating round channels in both poly methyl methacrylate (PMMA) and PDMS.  Other contributions are the improvements to the reproducibility of the ethylene glycol dimethacrylate (EGDMA) based PMMA bonding method and investigating the use of fabricated printed circuit boards (PCB) as molds. 10  Chapter 2: Alternative Fabrication Techniques 2.1 SU-8 Mold Fabrication The integrated circuit industry provides the conventional microfabrication equiment used which is expensive to purchase, operate and maintain.  Furthermore, the level of precision provided by these machines is far higher than what is needed for microfluidics as typical geometries are in the order of a 100µm rather than in the nanometer scale for integrated circuits.  The typical large features sizes of tens of microns also permit the use of low-cost laser-printed transparencies as photomasks.  As a result of this, it was decided to look at the feasibility of using relatively common pieces of equipment to achieve the same result to avoid the use of a spinner and mask aligner.  2.1.1 Procedure If one looks critically at what is needed to make an SU-8 mold on a silicon wafer it requires a method to spin the photoresist on the wafer and expose it to UV through a mask.  To replace the spinner, in the ideal case it would utilize a brushless motor to limit spark generation in case the device is used in an enclosed chamber containing solvent vapor and would have high maximum RPM to match the capabilities of a spinner.  While there are relatively low cost high speed brushless DC motors such as those used for building quadcopters, this would require creating a housing and mount for the device capable of spinning at high speed and the electronics would be more complex as well.  To keep things as simple as possible, a simple two pin computer fan was used instead.  Computer fans are already brushless DC and in the simplest case can be controlled by just the input voltage.  The fan is already optimized to run with minimal vibration at its rated speed and the center cylinder can be used to mount the 11  silicon wafer.  If a two pin fan is used as in this case, the only indication of the spin speed is voltage which may not be linear due to how the fan is made and load that is placed on top.  Three pin fans provide RPM information and four pin fans provide both RPM and the ability to adjust spin speed via a 5V PWM waveform.  With either the three or four pin fans, it would be possible to provide full control of the spin parameters given the appropriate hardware.    SU-8 2075 is very viscous and during initial testing it was found that during spinning the SU-8 would cause the fan to stop by clogging up the area between the fan blades and the fan housing. To prevent this, the fan (Sunon KD1208PTB3-6) is clipped from its housing and attached to a base that is chemically resistant to both the solvents found in the photoresists used and the clean up process such as metal or as in the picture FR4.  During the spin process the entire fan spinner assembly is placed in an covered container or a barrier is simply placed around the sides of the device to aid in the cleanup process.  Figure 4 DC fan with silicon wafer mounted for photoresist spinning.  12  As no clean room is used in this process, the area must be kept relatively clean of dust.  Furthermore if any dust is found on the substrate it is simply blown off using an N2 stream.  First the silicon wafer is centered and attached to the fan using either double sided tape or simply a rolled piece of flat tape instead of the vacuum suction on a normal spinner.  Next we apply SU-8 2075 (MicroChem Corp.), cover the fan assembly, then spin for 60 seconds at 12 V. With the two pin fan used, 12 V under no load conditions should generate 2500 RPM. Next an opaque cover is placed on the wafer to limit dust and light exposure.  The SU-8 is baked using the same procedure as with traditional methods consisting of a 3 minute, 65°C prebake, a 10 minute, 95°C baking period and then 1 minute at 65°C as a film relaxation step.   Next the mask is placed on the SU-8 and is sandwiched by two glass slides and clamped down with two binder clips.  When this is done, it can be observed that the rainbow like interference pattern is present that people use as an indicator that they have sufficient contact when using a mask aligner.  A low cost 4W UV lamp from uvlp.ca (model H DC UVL 5/2) is assembled with the filtered long wave bulb that emits 356 nm light and is mounted 6 cm above where the sample is to go.  For convenience, the battery contacts on the wire have been soldered with wires so that the light can be controlled outside the box. The sample is then placed underneath the light, the box is closed and the sample is exposed to three cycles of having the lamp on for 60 seconds and 30 seconds off.  Afterwards, the substrate is covered and baked for 3 minutes at 95°C then 1 minute at 65°C.  Then the substrate is washed in developer until properly developed which usually takes 13  between 60 to 150 seconds.  Figure 5 as shown below shows the completed SU-8 mold as completed during this process flow.  Figure 5 Finished patterned SU-8 on using the ultra low cost method on a 2" silicon wafer    14  2.1.2 Results Samples made using SU-8 2075 with a 60 second spin time had film thicknesses across the wafer of between 43-45 µm.  Samples made using SU-8 2025 using 30 and 60 second spin times had an uneven thickness that measured 15-30 µm and 9-18 µm respectively from side to side.  The variation in film thickness could be attributed to the fan being slightly off balance due to the use of tape to mount the wafer.   This work shows that it is possible to replace the use of a spinner and mask aligner with inexpensive components and perform SU-8 mold fabrication without the use of a clean room.  Table 1 Resulting thicknesses for the various spin times and photoresists used Photoresist  Spin Time (s)  Thickness (µm)  SU-8 2075  60  43-45  SU-8 2025  30  15-30  SU-8 2025  60  9-15     15  2.2 Deep-UV Etching of PMMA Typically making a polymer based microfluidic device would require utilizing techniques such as hot embossing, laser ablation, injection molding.  Alternatively, as demonstrated in [21], it is possible to directly etch channels into the PMMA utilizing a 254 nm UV light source that can be acquired at a relatively low cost for low power lights and reduces the need to purchase additional costly equipment.    As described, PMMA samples are sputtered with 100 nm of gold for use as a hard mask for the deep-UV exposure.  Then standard photolithographic techniques are used to pattern and etch the gold film such that the etched areas correspond to the areas that will be etched in the PMMA. Samples are exposed to the deep-UV using a Stratalinker 2400 UV crosslinker for 15 hour intervals up to 60 hours and etched for up to 60 minutes in 7:3 isopropyl alcohol (IPA):H2O at 28°C.  The purpose of the UV exposure is to break the chemical bonds in the PMMA.  By breaking bonds of the main polymer chain, the solubility of the PMMA in the developer solution is increased due to the reduction in molecular weight. Using these parameters they were able to achieve depths of up to 100 µm. [21]  16   Figure 6 Example showing the general deep-UV etching process  While this method does work, it has a strange combination of properties that make it neither well suited for rapid prototyping or for volume production.  Rapid prototyping typically trades manufacturing speed for cost while high volume production techniques such as injection molding and hot embossing have high upfront tooling costs in return for low per unit costs.  Compared to rapid prototyping techniques, this method is very slow as it requires exposure times of up to 60 hours, requires the same photolithographic equipment as making a PDMS chip, and a method to create the metal film is still required.  Compared to a hot embossing or injection molding setup, while the tooling costs are fairly low, masking every piece and the long exposure time would make the per unit cost very high.  Furthermore, hot embossing systems such as [22] can be created at fairly low cost further reducing the benefit  of this method. PMMAMask and exposePost exposure resultAfter mask removaland etching17  2.2.1 UV-Ozone Etching of PMMA Due to these drawbacks, enhancements to the method were investigated to reduce the time needed to etch a sample.  Even with the development of low cost hot embossing systems, it  is still another tool that would need to be purchased or built.  Given that some labs would have UV-Ozone cleaners either for removing organic contaminants on substrates or for use to bond PDMS microfluidic devices to glass, utilizing a UV-Ozone cleaner to enhance the deep-UV etching method was tried.  Unlike the Stratalinker 2400 UV crosslinker used in [21], UV-Ozone cleaners use high wattage lamps to generate 184 nm and 254 nm UV light to generate atomic oxygen to clean the substrate from molecular oxygen and ozone respectively [23].  For deep-UV etching of the PMMA, it is expected that the high dose of 254 nm radiation would speed up the patterning process.  As a metal surface or another material is needed that can block large quantities of UV light from the surface of the PMMA is needed.  While PMMA with aluminum film layer is readily available in the form of mirrored acrylic sheets, none of the manufacturers were able to supply the sheets without a permanently adhered backing layer that is used to protect the aluminum film.  Companies that provide vacuum deposition of aluminum can be readily found as the process is used very frequently to provide electromagnetic shielding in plastic electronic enclosures.  While a sample with a deposited metal film would remove the evaporation step, patterning and etching a mask would still be required.  In an attempt to 18  remove the photolithographically patterned mask entirely, aluminum tape was tried as the mask.  Aluminum tape was decided upon as it contains an actual metal film, can be easily obtained at any hardware store, readily adheres to the PMMA and could be cut to create microfluidic patterns using a cutting plotter as is done in xurography [24].  Other tapes were avoided as degradation due to the high levels of UV may permit exposure to the PMMA underneath.  While UV resistant tapes could be found, using the aluminum tape for initial testing ensured that UV degradation of the polymer would not be an issue.  Even with the aluminum tape, degradation to the exterior polymer layers was noticeable. To try this, PMMA pieces (Plaskolite Optix Acrylic) were rinsed with IPA and H2O and dried with N2.  Aluminum tape (Nashua 322) was cut into 0.5 mm wide strips and adhered to the PMMA pieces.  An initial test was done to see if it would be possible to remove the aluminum tape from the PMMA and while it adhered fairly strongly, it was possible to remove the tape and the adhesive residue.  Using a Jelight Model 42 UV-Ozone cleaner the taped samples were exposed to UV light to a maximum of 90 minutes.  For development, two groups of samples were tested: one where the aluminum tape would be left on until development was complete and one where it would be removed prior to development.  Development was done in a 28°C bath of 7:3 IPA:de-ionized (DI) H2O for 90 minutes.  For the samples where the aluminum tape was left on during development, the tape became incredibly difficult to entirely remove.  For samples where the aluminum tape was removed prior to development, removal was more difficult but 19  still feasible.  Measurements showed that depths of 90-100 µm were achievable for channel widths of approximately 0.5-2mm.  As removing the aluminum tape on larger quantities of samples would be very time consuming, it was decided to try a different method of patterning the surface.  Rather than create a mask on every piece, a patterned metal film on a piece of glass would serve as the mask so it could be replicated on multiple samples.    Standard glasses such as  what is used for windows [25], BK7 borosilicate glass [26], and even Pyrex [27] are unable to transmit 254 nm UV light.  As a result more expensive quartz or fused silica needs to be used [26].  2" round 1/16" thick quartz pieces (McMaster-Carr 1357T12) were purchased, had a gold layer with a chromium adhesion layer patterned with 150 µm wide channels through photolithography and etching.  The quartz glass mask was placed on a cleaned PMMA piece with the gold layer in contact with the PMMA surface to mimic having the metal layer patterned on the surface of the PMMA.  A piece of tape was attached to the bottom of the PMMA and wrapped around the edges of the glass mask in order to secure the mask to the sample.  Samples were exposed for 90 minutes and then etched for 60 minutes using 7:3 IPA:DI H2O at 28°C with agitation.  Profiling the samples showed high levels of non uniformity with channel depths of ranging from 30-57 on a single chip using the glass mask while depths ranging from 90-100 µm were measured when using the aluminum tape.   The difference in depth is likely caused by the UV transmission losses through the quartz glass.  The nonuniformity is thought to be caused by the quartz mask not being held in position properly.  The UV Ozone cleaner vibrates a fair amount due to the ozone scrubber motor and the eventual degradation of the tape used to hold the pieces together could potentially lead to the mask lifting.  In some cases, there has even been 20  evidence of multiple channels where there should be one which indicates the mask is shifting.  For future testing it would be better to try clamping everything together using binder clips as in the low cost photolithography method in section 2.1.  While these problems could likely be solved with more experimentation, it was decided that research on this method would be stopped.  As a patterned quartz mask is now required in order to avoid having to deposit and pattern each individual sample, it now requires more time, resources, and equipment than simply fabricating chips using soft lithography or hot embossing without any additional benefit.  Furthermore, this method relies on shortening the polymer chains to make it easier to dissolve the PMMA in the channels.  As a result, with methods such as solvent bonding, it is expected that the channel will soften and clog more readily due to the shortened chains. Preliminary bonding tests using ethanol and isopropanol as solvents showed this as well.  While these results could be confounded by sagging in low aspect ratio channels (500-2000 µm wide and 90-100 µm tall), the lack of a clear benefit for using this method meant our efforts would be redirected to methods such as hot embossing.    21  2.3 Cast Epoxy Molds For hot embossing, while silicon molds with patterned SU-8 molds can be used, they can potentially have limited lifespans of as low as 5 cycles.  As a result, cast epoxy molds were created by double casting an SU8 mold into PDMS and then using the PDMS as a mold to cast the epoxy.   By creating these epoxy molds Koerner et al. were able to create a low cost durable mold that would last more than 50 cycles and still be able to use the original SU-8 master to create more epoxy molds as needed. [28]   Figure 7 Example process flow for creating an epoxy mold and embossing  Pattern Si waferCast PDMSCast EpoxyEmboss PMMA22  Following the procedure described in [28] to replicate a 4" silicon wafer with SU-8 structures did not produce the result as expected. We found that there were multiple issues that were not fully detailed in the paper.  The first was that in order to get a usable epoxy mold the area that is intended to be used for curing both the PDMS and the epoxy must be extremely level.  Failure to do so would lead to errors that could compound with each cast and end with an uneven epoxy mold. This could cause problems during either the embossing stage or the sealing stage as the force would be distributed unevenly. While a silicone piece could be used to help distribute the force more evenly [22], it is best to prevent the problem if possible. As a result, an aluminum plate was attached to one rack of an unheated oven, PDMS was poured into it, and then the temperature was changed to 80°C to cure.  Due to the flowable nature of the uncured PDMS, once it is cured the resulting bed will be level.  Another issue that occurred is that if a thicker piece of PDMS is cast (approximately 0.5 cm), when the epoxy is curing the PDMS will curl while the epoxy is still not fully cured rendering it unusable. It is believed that this is caused by stresses developed in the epoxy during the final stages of curing.  During the first few hours of the curing process no curling is seen but near the end of the curing process curling occurs. Changes to the PDMS can be seen as the PDMS becomes permanently translucent as the curing process proceeds. It is possible that this process changes the composition of the epoxy in the mold via absorption of compounds into the PDMS such that the region in contact with the PDMS cures at a different rate than epoxy further away. As a result, if a thicker piece of PDMS is desired for handling purposes, a glass wafer can be adhered to the back using UV-Ozone treatment or any other method prior to pouring the epoxy.   23   The third issue that occurred was that due to the surface tension of the epoxy, the epoxy near the sides of the mold will not be level.  So features need to be sufficiently far from the edges in order to be properly embossed.  Once these three issues were taken care of, it was possible to make epoxy molds as intended.  24  Chapter 3: Fabrication of Circular Cross-Section Microfluidic Channels For mimicking biological systems such as vein networks, circular channels are very important [7].  However from standard lithographic processes the fabrication process creates rectangular rather than circular channels.  As a result, [29] utilized small aperture sizes and backside diffused exposure to create semi-round channels while [9] reflowed negative resist to create semicircular channels as well.  Later, under ideal conditions Huang et al.[11] were able to make round channels by reflowing positive AZ P4620 photoresist.  These conditions are difficult to control as it involves material evaporation, out gassing, the nature of the resist, variation in substrate and many others [11].   If only a single straight channel is required, simply using a round capillary tube would be sufficient.  However, if any other configuration is desired such as a curved channel for flow focusing or making a microvessel network is desired it would not be possible.  Also, by using materials such as PDMS, PMMA and other various polymers the mechanical and optical properties as well as chemical resistance, biocompatibility and gas permeability can be adjusted to meet the application.  By utilizing less conventional techniques,  straight round channels have been created by casting a wire and pulling it through PDMS [20] and by modifying a rectangular channel [10][8].  Work has also been done to make two dimensional channels by optimizing surface wetting parameters [7] and  three dimensional channels by forming and melting a wire [30], sucrose fibres [31], while [32] used pressure assisted deformation of poly lactic acid (PLA) channels. 25  The work of [29], [9], and [11] create channels that may not necessarily be perfectly round but the resulting molds enable the creation of replicated samples rather than a single use mold.  On the other hand, [30], and [31] are able to create round channels but all of these designs result in a unique single sample that potentially may not be reproducible.   While it may be possible with [32], no indication has been given if the pressure deformation step is reproducible between samples. The process described in [20] is capable of creating round channels reproducibly due to the use of a jig and a wire but that method is limited to making straight wires.    Circular cross section channels have also been created using a micromilling process using a 4 flute ball nose end mill [33] to cut a semicircular mold followed by a double casting process to create a circular cross section channel.  In this work, the machined semicircular halves were cut and had an average width of 1041 ± 8 µm and depth of 532.9 ± 8 µm using a 508 nominal radius ball end mill.  Trapezoidal channels have also been machined with widths of 600 um [18] and rectangular channels of between 250-1000 µm [34].  As a result, since 4 flute ball nose end mills can be found with diameters down to 250 µm [35] it is expected that much smaller diameter channels would be achievable as well.  Typically, 4 flute end mills provide better surface finish than 2 flute versions [36] but if 2 flute end ball nose end mills are found to provide an acceptable result for microscale machining, then diameters down to 75 µm are available [37].  3D printers have been also used to create microfluidic channels with channels of 500 x 500 µm [38] utilizing an Objet Connex 350 Multimaterial Printer and even enclosed 250x250 µm 26  channels and 200 µm channel width molds using a MiiCraft stereolithographic printer [39]. While this method has enough resolution to create 200 µm channel molds, it is still not well suited for creating circular channels as the x-y pixel size of the sterolithographic printer is still just 56 µm x 56 µm with a pixel height of 50 or 100 µm. As a result, this would still only create a very pixilated round channel.  The Stratasys Objet 30 has better X x Y x Z resolution of 42 x 42 x 16 µm so it was used in an attempt to make 3D printed round channel molds.  Utilizing a Stratasys Objet 30 printer, tests were performed that included creating ellipses, circles, and square channels but the results differed from what was expected.  Due to the spreading of the plastic, the height would be lower and the edges of the channels would be more spread out than intended.  The result of this was that by attempting to print a square channel a more curved or rounded profile could be created, while ellipses and more circular features would create shorter and wider channels.  Due to the resolution limits of the 3D printer, larger channels would more accurately create circular channels when cast into PDMS however there would still be significant tapering of the corners as seen in the images below.  27   Figure 8 PDMS channels created using a Stratasys Objet 30. Left: 500x500 µm 3D printed channel cast in PDMS. Right 400x400 µm 3D printed channel cast in PDMS.  Channels and images were created by Anna Lee.  3.1 Hot Embossing of Wire Based Microfluidic Channels While these methods to make round channels are capable of making channels in PDMS, some methods such as the machined channels could be used to make thermoplastic devices via hot embossing while others such as the 3D printing based method would require double casting to make a more durable mold.  Other methods such as those based on wires cannot be used, as the halves cannot be separated to double cast the PDMS. As a result, a new method utilizing metal wire to form the mold was devised.  By utilizing hot embossing, keeping a sample level across the length of the material is easier to achieve and does not need a jig in order to do it.  It is also expected to be easier to keep a curved wire in its desired position as PDMS does not need to be poured which can potentially alter the wire's position during the pouring process.  This method can also be used as a rapid prototyping technique as no mold needs to be made in order to create a microchannel.     28  Furthermore, as wires down to thicknesses of 18 µm are readily available for processes such as wire bonding integrated circuits [40] and electrical wire, creating different diameters is as simple as purchasing a different wire.  As wires are made using a drawing process [41] it is believed that all these different wires would be round due to the assumed relative simplicity of making a round die.   3.1.1 Fabrication Typically hot embossing replicates the features of the mold plates onto the top and bottom of the sandwiched substrate.  When using a wire as a mold, two different methods are possible.  The conventional one sided mold and the more interesting case where the wire is sandwiched between two pieces of substrate.  For the purposes of fabricating round channels, only the latter method will be investigated.  Figure 9 Methods of embossing a wire to create a microfluidic channel  This allows for the creation of different channel geometries, such as round channels as it allows for both sides of the wire to be formed. The hypothesis is that if the two substrate pieces have identical mechanical properties and everything is held in an isothermal state, the Wire embossed around two substrate piecesWire embossed with one piece of substrate29  normal force exerted on both sides of the wire will be identical, thereby molding equal halves of the wire in each side.   3.1.1.1 Straight Channels To make straight channels of 250 and 150 µm, Alphawire 1061 and NTE WH22-00-100 stranded wires were stripped in lengths greater than 5 cm and debraided. The unbraided strands of the NTE and Alphawire wires are 250 and 150 µm in diameter respectively. To straighten the wire, an unbraided strand is simply clamped on both ends with pairs of pliers and pulled until one side breaks.  The wire is adhered to a 4 cm x 5 cm piece of PMMA (Plaskolite Optix) using a low temperature hot glue gun (Surebonder GM-160C) on the edges.  Once the glue has cooled, a 3.5 cm x 4 cm piece of PMMA (Plaskolite Optix) is placed on top.  In both cases, the PMMA used is 1.59 mm thick.   Figure 10 Process to assemble the straight round channel sample for embossing  30  For the initial samples, the wires were checked using a microscope to check that the wire diameter was the same before and after straightening the wire.  Furthermore, according to [42], deformation is uniform until necking occurs which will start at a defect on the surface.  As the wire is clamped down with pliers, the most likely spot for deformation is where the pliers come in contact with the wire.  This corresponds with what is observed as the wires usually break where the pliers are.  In addition, if the wires break elsewhere they are simply discarded.  The Carver Inc. 4386 heated lab press was used as the embossing system and is heated up to 101.67 °C (215°F) for both plates.  Once the plate temperature has stabilized, the PMMA pieces are sandwiched by mirror finished copper plates and placed inside. The heated platens are brought together close but without pressing down on the sample.  Once the temperature has stabilized again, the samples are pressed down with 14 MPa of pressure and the pressure is brought back to 14 MPa three times in 5 minute intervals as the system does not provide constant force output. Typically samples are embossed in groups of two or four to assist in balancing the load distribution in the embossing system. Afterwards, while still under pressure, the heaters are turned off, the samples are cooled to less than 65°C and then they are removed from the system. 31   Figure 11 Material stack used when embossing a round channel  Note that stripping a single long length of wire is preferred over multiple short segments as it limits the chance of damage to the wire. If the straightened wires need to be trimmed, flush cutters need to be used for the straightened otherwise the wire may slip between the gap of the shearing faces and bend the wire.    PlatenMirror finished copperPlastic substrate32  3.1.1.2 Serpentine Channels Besides making round channels, this method allows for the creation of curved round channel geometries as well.  Two methods of bending wire were proposed: developing a system similar to the open source DI Wire [43] or using a 3D printer to create forms.  Preliminary work was done on a DI Wire type device using a sewing needle mounted on a servo motor and found that while feasible, limitations would arise as the wire diameter decreases and the size of the pattern to be bent increases.  As the diameter of the wire decreases, stiffness is lost as the cross sectional area of the wire decreases.  During the wire bending process, the wire must support its own weight and resist changing from its desired shape.  As the pattern gets larger, more weight must be supported until the process is complete.  If this is not possible  the end result will not be as expected.   In addition, as access to 3D printers and availability of 3D printing services are steadily rising and would be cost effective for the small forms needed for this, the 3D printing method was pursued.   A Stratasys Objet 30 was used to create mated forms and the wires are simply placed in between and sandwiched together to crimp it. Once the wire has been crimped, the wire is mounted and embossed to PMMA pieces as in section 3.1.1.1. 33   Figure 12 Process to assemble a curved channel sample prior to embossing  3.1.1.3 Spiral Channels Spiral channels could be made using by simply using a cylinder, a spacer material such as a dissolvable tape or dry photoresist, and a wire.  To do this, the tape or photoresist is folded enough times to create the desired wire spacing.  Afterwards the wire is straightened and placed on the folded tape.  Once this is done, the tape and wire assembly is carefully rolled around the cylinder and once complete it is placed in developer until the spacer material dissolves.  Once it dries, the wire is placed on a piece of PMMA the outer end is pulled past one edge of the PMMA piece to make an outlet and the center section is pulled to roughly center it to give more clearance when holes need to be made.  After this is complete, the top piece of PMMA is added and the wire is embossed as in section 3.1.1.1.  34   Figure 13 Example depicting how a spacer material would be used to coil a wire around a cylinder to create a spiral  3.1.1.4 Variable Width Channels Variable width circular channels are desirable in order to compare them to rectangular channels that expand and contract while focusing [44]. Variable width channels can be created simply using an MG Chemicals etch resist pen (416-RP).  Using a straight wire, the areas to be protected are simply rolled along the tip of the pen.  Once all areas to be protected are covered, it is etched in ferric chloride until the desired width is achieved.  Alternatively, the wire can be sprayed with an airbrush through a solid mask.  The areas that get covered with paint are protected from the ferric chloride.  Once the desired pattern is achieved, the wire is embossed as in section 3.1.1.1.   3.1.1.5 Joining Channels Up until now only single wire designs have been discussed.  By joining multiple segments of wire, more complicated designs can be created.  This can be done by soldering wires together or by simply overlaying them.  To create the smoothest solder joint, this is done using solder SpacerWire35  paste and a hot air soldering iron or by adding soldering flux to an existing joint and reflowing it using a hot air soldering iron.  3.2 Bonding of Round Channel Microfluidic Devices Aligning a microfluidic device that needs to be bonded in an embossing system is much more challenging than aligning two PDMS channels.  The main problem is that these devices cannot be checked under a microscope as they are bonded.  While they can be aligned with a microscope prior to transferring the sample to the embossing system, the use of solvents for bonding makes it more likely that the sample will shift during transport. As a result, a better alignment technique than simply aligning things by eye is required.  3.2.1 Alignment Techniques Alignment markers for microfluidic devices can be made in multiple ways but most involve using manual observation or complex fabrication methods [45].  In the case of the embossed round channels, methods based on visual observation would be difficult to use as alignment needs to be maintained as the device is placed in the embossing system for bonding.  Using a post and a mating feature [46] in this case would entail making a raised feature on one half which would be difficult to do as both substrates are being embossed.  Furthermore, if it was accomplished, tolerances in the fabrication of both features would introduce additional error in the alignment process.  As a result of these difficulties, a new method of alignment was devised.  In order to be compatible with the embossing method used to make the round channels, an alignment 36  marker is added to the device at the same time as the wire is being embossed.  The object used to create the alignment marker is left in the device even during the bonding process.  This creates an alignment marker in the system with minimal tolerance as the object used to create the mating features is used for the alignment process. Objects that have sharper sidewalls such as the inner ring of a washer work better as the sharp edges make it apparent that the device is mated together.  However, smaller objects such as wires tend to emboss better due to the smaller volume of material that needs to be displaced during the embossing process.  As a result, the 250 µm wire is bent into right angles and adhered using hot glue in the same manner as the wire in section 3.1.1.1.  These wires are left embedded in the PMMA for use in the bonding process.   Figure 14 Process for building the straight wire with alignment marker stack  37  3.2.2 Bonding Method For polymers, with the exception of chemical surface modification and adhesives, bonding methods for plastics typically are either thermal bonding methods or solvent bonding methods.  With thermal bonding, bonding occurs simply by applying heat, typically above Tg, and pressure to let the polymer chains diffuse across the boundary.  Solvent bonding methods are essentially the same except that a solvent is used to help the polymer chains move more readily at lower temperatures and in some cases down to room temperature.  Solvent bonding methods typically provide higher bond strengths than thermally bonded samples.  In addition, with thermal bonding, unless the bonding parameters are highly optimized, there is a high chance for deformation of the channel.  Due to the benefits of increased bond strength and the higher chance of deformation with thermal bonding, only solvent bonding methods were investigated.  While there are have been many different bonding methods shown in literature for bonding PMMA based microfluidic devices, few were actually usable in this application and fewer still worked upon testing. Some methods such as one that required soaking one piece in chloroform [47] could not be used as it is meant to be used on a blank cover slip; in the work here, both sides are embossed, and the solvent would swell and deform the embossed piece. Likewise, solvent imprinting [48] is incompatible with the process as well.  [49] used water to form an ice barrier to prevent channel collapse but this method would be difficult to use.  The ice would likely break as the halves are pulled apart as the ice would form a protrusion since the channel is on both sides. Methods that involved simply heating and clamping solvent mixtures such as [50] and [5] resulted in clogged channels, even within 30 seconds.  38  It should be noted though that the low clamping force needed in these methods is difficult to produce with the embossing system used so forces used in testing are higher than what those papers describe.  Solvent assisted bonding using EGDMA [51] was tested on a standard microfluidic design and was discovered that the method as described is highly operator dependent as the solvent is blown off manually using an air jet.  Modifying the method to use solvent removal via spinning resulted in a reproducible method.  However, it was unfortunately discovered that the chips still had sufficient chemical residue after rinsing and two weeks in the fume hood that some members of the lab would feel slightly ill near them.  As a result, while very promising, a different method needed to be found.  In the end a solvent displacement technique [52] utilizing vacuum pressure to displace solvent in the channel with water was found to work reliably.  Typically this method requires the additional step of creating a jig that contains the inlet and outlet ports as the ports are typically on the top of the sample. However, in this case the most convenient place to put the ports is in plane by simply hot embossing around the metal inlet tubes which removes the need for the additional jig.  In order to bond using this method, a few modifications are made to the embossing process previously described in sections 3.1.1.1 and 0.  After the wire is straightened, 25 gauge syringe tips (McMaster-Carr 75165A686, ID 0.305 mm OD 0.508 mm) with the plastic melted off are threaded onto the wire on both ends.  At this point, if curves are desired the 39  wire is crimped in the middle.  Assembly is done in this order to limit the amount of handling once the wire is crimped.  The syringe tips are hot glued to the PMMA substrate and alignment markers are added.  After the embossing process is complete, the samples are annealed at 80°C for a minimum of 90 minutes.  The inlet tubing is primed with water and one end is left in a water reservoir. The other end and the vacuum tube is attached as shown below.  The sample is placed on copper mirror polished plates for surface finish, 250 µL of solvent is placed between the halves and the sample is clamped with 1.4MPa.  Once heating is started, vacuum is pulled using a syringe such that a slow steady drip is achieved. Once the heat reaches 75°C a timer is started for 5 minutes then the sample is cooled down to 30°C before removal. Vacuum is only removed at this point.   Figure 15 Vacuum assisted solvent bonding setup     WaterreservoirTo syringe or vacuum sourcePlatten Mirror finished copperEmbedded syringe tip40  3.3 Fabrication of PDMS Round Channels While being able to make polymer chips has it uses, the ability to create round channels out of PDMS is still highly desirable.  For biological applications it has the benefits of being biocompatible, transparent, gas permeable, and has low auto fluorescence [53].  Furthermore, while there are drawbacks such as the absorption of small molecules [54][55] it is still very commonly used in research due to the ease with which devices can be prototyped and tested.  3.3.1 Emboss and Cast Method In this method, the wire design is embossed as in sections 3.1.1.1 and 0 as needed.  This ensures that the channel is level and that the channel design kept as desired during the PDMS casting process. Once the embossed pieces are made, they are used as the initial mold for a double casting process to make the final PDMS pieces. To make the mold, multiple different materials and processes were tried.  It was found that a durable urethane mold could not be cast from the PMMA pieces as the PMMA was attacked by the urethane.  Embossing Sculpey polymer clay (Part no. S302 1113) from PMMA was attempted but removal prior to curing the clay would result in incomplete channels and removal after curing resulted in bonding.  Oil based Teflon enhanced lubricant Triflow (Part no. TF20006) did not work as a mold release and neither did Armor-all despite the recommendation on [56].  Embossing an Armor-all (Part no. 10228) coated PMMA piece into JB Weld Kwik Plastik, curing, then removing it worked well. However, even curing the epoxy for a week still resulted in the PDMS not curing locally around the epoxy.  This occurred with curing the PDMS at both 80°C and room temperature.  41  As a result, the PDMS double casting method was tried.  First the embossed PMMA halves are taped down to a plate, cast using Sylgard 170, degassed and cured for 15 minutes at 70C.  Sylgard 170 was used rather than Sylgard 184 as it was discovered that it would attack the PMMA slightly when cured at 80°C for two hours. While untested, it was thought that repeated castings of Sylgard 184 on the PMMA pieces could potentially result in compromising the channel geometry as half submerging pieces of PMMA in both Sylgard 184 and Sylgard 170 and curing at 80°C revealed an interesting result.  The PMMA piece that was partially submerged in Sylgard 184 exhibited slight optical distortion in the submerged section and there was a visible line where the two halves met. With the Sylgard 170, there was no noticeable difference between both sides of the sample.  While there is limited information on the composition of both Sylgard 184 and 170, the MSDS of the Sylgard 184 shows that there is a small amount of ethylbenzene in the base while there is none in the Sylgard 170 components.  It is believed that this may be the cause of the distortion in the PMMA by Sylgard 184.  Besides the potential for distortion, Sylgard 170 degasses more readily, cures much faster, and is a dark gray colour which helps with identifying the mold and cast sample.   Afterwards, the PMMA pieces are removed and the edges of the mold are trimmed with a knife to make a shallow angle to aid in releasing the second cast PDMS from the mold later on.  The resulting PDMS mold is placed in the oven at 100°C for 48 hours minimum as described in [57].  Once this is complete, the second cast of Sylgard 184 is poured in, degassed and cured at 100°C for 15 minutes and removed immediately.  This is imperative as 42  it was noticed that even with the aging process for the mold, if the Sylgard 184 was left for an hour it became impossible to remove.    Figure 16 Process flow for double casting an embossed PMMA sample into PDMS  Even with removing the PDMS immediately, it is still a difficult process as the thermal aging results in the mold losing some of its elasticity. So to aid in the removal process, the mold is soaked in ethanol. As the sample begins to lift from the mold more ethanol is added between the layers.  If a region is adhered very well more time is provided to allow the ethanol to loosen the sample.  Once the halves have been removed, they are plasma treated and then methanol is placed between the halves as a lubricant to help with the alignment process.  The pieces are aligned under a microscope then baked at 70°C for a minimum of two hours.    Emboss PMMACast Sylgard 170Cast Sylgard 184Bake 48Hrs at 100°C43  3.4 Challenges 3.4.1 Levelness of Samples Many challenges were faced trying to get the round channels fabricated but the main challenge was to ensure that both the top and bottom platens are parallel at all times; especially as the pressures are increased.  Regardless of whether the hot embossing process is meant to yield a single sided embossed part or dual embossed parts having the platens parallel is imperative.  If they are not parallel, the embossed sample can exhibit a slant in the bulk substrate. This can cause difficulties in bonding unless the bonding setup can compensate for the slant.  In the case of the Carver Inc. hot embossing system used, the top plate is fixed but the bottom plate is lifted by a center mounted hydraulic cylinder. Unfortunately, the bottom plate is not constrained in the horizontal plane.  If the sample(s) are not evenly distributed, as the force increases the bottom plate would develop a tilt which would be transferred to the embossed sample.  Depending on the material, temperature, pressure and time the slant may not be noticeable by eye.  But when the sample is then bonded there will be a region exposed to less force due to the slanted nature of the sample.  While some people have sandwiched the piece with compliant materials to compensate [22], this would only work if the tilt was not severe.  The best way to solve these problems is to simply ensure that the samples are evenly distributed and centered both during the embossing and bonding processes.  3.4.2 Achieving Sharp Edges on Embossed Features Achieving sharp edges on channels is actually somewhat difficult to achieve when using the double sided casting method needed to create the round channels.   Typically to achieve 44  proper filling of the mold there are multiple options which can be used in conjunction with each other: increasing temperature, force, time, and applying vacuum.  When trying to do things low cost, using vacuum is not really an option as it is a feature typically only on more expensive devices as it essentially requires making the hot embossing system or at least the platens into a vacuum oven. [6]   Of the remaining three, the fact that the double substrate molding involves two identical substrates creates a few more limitations that do not exist when doing traditional hot embossing with a mold and a substrate.  If the combination of force and temperature is too high, the wire can be moved out of its intended position due to outward movement of the material.  Furthermore, as the force required to bend the wire decreases proportionately as the wire diameter decreases, the effect gets worse as the desired channel diameter decreases.  In addition, thermal bonding can take place if the combination of time, temperature and force allows it.  In some cases, the level of bonding can be very minor and the pieces will be possible to remove but in other cases it will not be.      In comparison, when a single substrate and mold are used these restrictions do not occur.  Increasing temperature and force has no effect unless it is high enough to cause deflection of the mold material at the specified force.  With a metal mold this would not be an issue but with SU-8 and epoxy based molds it can be.  When using PMMA, as the glass transition temperature is typically between 95-100°C depending on the composition, this is not likely going to be an issue.  45  As a result, the process parameters described in section 3.1.1.1 were found by performing an approximately weighted binary search of temperature, time, and force.  The straightness of the channel, the level of deformation in the bulk material, the sharpness of the corners when sectioned, and presence of thermal bonding were used to determine the direction the search would take.  Furthermore, as the slowest step in this process is by far heating the platens to an identical temperature, a cooling step was not performed prior to removal so some deformation of the channels was expected.  Instead, once the parameters were narrowed down to what was expected to be the optimal results, additional samples were run with the full cooling process.  It should also be noted that it is expected that there are a multitude of parameter combinations that would be acceptable as all three parameters influence each other.  Initial experiments were performed with parameters of 96.1°C, 14 MPa for 20 minutes so a test at 96.1°C, 24.5 MPa for 20 minutes was tried and produced a channel with much lower levels of rounding at the corners of each semicircular half.  Later tests of bonding at this temperature using both ethanol and IPA exhibited significantly higher levels of cracking after annealing than the original bonding trials at 96.1°C, 14 MPa for 20 minutes so different embossing parameters needed to be found.  A higher temperature of 112.8°C was arbitrarily chosen as a higher bound.  All tests at this temperature were fully thermal bonded except when a lower setting of 3.5 MPa for 5 minutes was used.  Even with this short embossing time, thermal bonding was quite strong so there is 46  a high risk of breaking samples. Furthermore, the straight wire segments became bowed due to the embossing process so a lower temperature is still needed.     Tests were then accidentally performed at an incorrect temperature for the binary search process of 101.8°C, 14 MPa for 20 minutes.  However as the results were promising, work was continued at this temperature.  On these samples a small amount of thermal bonding was evident as prying the halves apart now presented an audible tearing sound.  Additional tests were done at this temperature using both 24.5 MPa and 19.25 MPa for 20 minutes but both exhibited bowing of the straight wire so could not be used.    As can be seen from the images below, the end result parameters of 101.8°C, 14 MPa, 20 minutes are much better than the initial samples that used the parameters of 96.1°C 14 MPa for 20 minutes in terms of corner sharpness. However, there is potentially room for further optimization of the parameters to reduce the needed embossing time.    47    96.1C 14MPa, 20 minutes  96.1C 24.5MPa, 20 minutes  101.8C, 14MPa, 20 minutes  112.8C, 3.5MPa, 5 minutes  Figure 17 Images of channels embossed with various parameters. From top left clockwise 96.1°C 14 MPa 20 min, 96.1°C 24.5 MPa 20 min, 112.8°C 3.5 MPa 5 min, 101.8°C 14 MPa 20 min  Table 2 Summary of embossing experiments to find optimal embossing parameters Temperature (°C) Pressure (MPa) Embossing Time (min) Observations 96.1 14 20 Initial settings with large amount of rounding in the corners 96.1 24.5 20 Corners are much sharper but significant cracking during bonding 112.8 14 20 Aborted early as sample was immediately compressed significantly, fully bonded together 112.8 7 10 Very strong thermal bond, cannot pull apart 112.8 3.5 5 Strong thermal bonding, can pull apart sometimes 101.8 14 20 Corners are fairly sharp, small amount of thermal bonding 101.8 24.5 20 Cannot use as bowing of wire was seen 101.8 19.25 20 Cannot use as bowing of wire was seen  48  3.4.3 Reducing Cracking During Bonding Cracking during bonding turned out to be a significant hurdle in the development process besides finding a method that would not clog the channels.  Once the vacuum assisted solvent bonding method was discovered to work, work was undertaken to attempt to limit the formation of cracks during the bonding process as a significant number of cracks formed during the initial testing.  The first step of this process was to anneal all samples for at least 90 minutes at 80°C.  Even with annealing there was still significant cracking at times which was worse in the samples using the optimized parameters (101.8°C, 14 MPa, 20 min) compared to the ones using 96.1°C, 14 MPa, 20 min. With standard hot embossing, one of the options is to increase the embossing temperature high enough such that the material relaxes to limit stress.  In this case though it is not possible due to the presence of thermal bonding as the temperature increases further as was evident with both the samples embossed at  both 101.8°C and 112.8°C.    Blank PMMA samples of the same dimensions were bonded using process parameters of 65°C, 7 MPa, 30s without cool down using isopropanol, ethanol, and methanol using an approximate volume of 1 ml.  It was discovered that in all three cases significant amounts of cracking would occur if too much solvent was used such that the solvent spilled over the edges of the sample.  It was also noticed that the amount of cracking from least to most was consistently methanol, ethanol then isopropanol.  What sounded like cracking noises were also audible when the force was removed so tests were then rerun with cooling down to 35°C to see if the lack of a cool down period was the issue.  Cracking still occurred with the cool down period and the amount of cracking still was consistent with the type of solvent used. 49  As a result, tests on blank samples were performed using lower solvent quantities of 250, 150, and 100 µL. In all three cases for all three solvents two samples were tested and cracking was not seen except in some cases small cracks at the edges where the smaller top plate meet the bottom plate.  Prying apart the samples by lifting one corner was then attempted to roughly gauge the bond strength of the different solvents. Methanol bonded samples could be consistently peeled apart while ethanol bonded parts could be peeled apart approximately 0.5 to 1 cm from the edge before breaking.  Isopropanol bonded parts either could not be pried apart and would just immediately break.  As methanol had the weakest bond and isopropanol had the most cracking during bonding, it was decided that ethanol would be used going forward for bonding.   It is believed that the differences in the levels of cracking for the various solvents is related to the solubility of the PMMA in each.  The Hildebrandt solubility parameters for PMMA, isopropanol, ethanol, and methanol are  19.0, 23.8, 26.0, and 29.7 MPa1/2  respectively [58].  Closer solubility parameters indicate that materials are more miscible. So at the same temperature it is expected that isopropanol will cause more softening of the PMMA due to the solvent.  This could lead to cracking as the softened section presents a way for the stress to be released from the material.    Bonding with channels using ethanol showed some curious results.  While cracking was reduced overall, some types of cracking were not. In general, two types of cracks could be seen: random bulk material cracks that would be approximately up to 0.5-1cm in length and severe cracks that appeared to originate from the inlets or alignment markers that could 50  propagate up to the length of the chip. Bulk material cracks were minimized although there are usually still one to two while the severe cracks still occurred.     As a result bonding was tested on blank samples using 250 µL of ethanol, 65°C, 30s with cool down to 35°C and a reduced load of 1.75 MPa. While an even lower load would be ideal to test, due to the design of the embossing system, lower loads would not be very reproducible as 1.75 MPa is already very close to the bottom limit of the scale.  The bonded samples showed that the amount of cracking was minimal as with the previous group of samples using 7 MPa loads.  So a group of four samples with 250 µm round channels were embossed with the parameters in section 3.1.1.1, and then annealed at 80°C for 90 minutes followed by 1 week at 70°C.  These showed minimal bulk material cracking but 3 out of 4 still showed cracks originating from the inlets.  Due to the additional 1 week anneal, it is believed that this rules out further annealing to reduce the cracking and indicates that the alignment markers and inlets are the source of the cracking.  3.4.4 Increasing Long Term Bonding of Channels Due to problems with the various bonding methods in preventing channel clogging, the initial response was to limit the amount of time and lower the temperature the samples are heated and pressed.  With both the ethanol and isopropanol bonded samples (bonding settings of 65°C, 1.75 MPa, 30s, 250 µL of solvent)  it could be noticed even by eye that after a few days the edges of the channel would appear to be delaminating and occasionally in the bulk material.  As bonding was fairly consistent now, the process parameters were increased to 75°C at 1.75 MPa for 5 minutes with 250 µL ethanol with a syringe pulling vacuum and the 51  inlet connected to a water reservoir.  After bonding, the channel was tested to work and after one week no lifting at the edges or in the bulk material could be noticed unlike before.    3.5 Results 3.5.1 Embossed Channel Geometry 3.5.1.1 Wire Geometry After the wire was used for embossing a straight channel, it was taken to verify the geometry of the wire.  As the wire needs to be cut, a sample prior to embossing was not taken as it would not be in the embossed section.  Consistently, it could be seen that for the 250 µm wires when they are cut with a wire cutter one half would be significantly deformed but the other half would maintain the round cross section with minimal deformation.  For the 150 µm wires, some deformation is visible on both sides.  To take these images, a piece of PDMS was taken and a 0.5 mm hole was made in it.  It is placed on a glass slide and a small section of wire to be imaged (approximately 3 mm) is simply placed in it.  Due to the length of wire blocking the light source, taking an image of the wire end required significant overexposure and still it did not image very well.  Using Image J and the ThreePointCircularROI tool, the 250 µm wire was measured to be 231.8 ± 3.4µm in diameter and the 150 um wire was measured to be 149.8 ± 12.1 µm.  It should be also noted that there is potential that the wire could be a slight angle which could contribute to measurement error and the poor lighting could influence it as well.  52   Figure 18 A wire end being imaged for measuring the wire diameter. The wire has been placed vertically in a PDMS hole to allow for imaging the end  3.5.1.2 Channel Cross Sectional Profile Using the procedure described in section 3.1.1.1 150 and 250 µm wires were embossed and then pulled apart. The PMMA samples were then scored on the unpatterned sides and broken.  Care was taken to attempt to align the scoring marks between the top and bottom halves but as this is a difficult procedure it cannot be guaranteed.  Samples were then placed in a 3d printed jig to provide vertical alignment and glued to a piece of PMMA. Images were taken of both the top and bottom halves separately.  Using Image J the images were analyzed to determine the dimensions of the halves.  The width of the 150 µm and 250 µm halves were measured to be 134.5 ± 3.1µm and 233.1 ± 2.46 µm respectively. While the height of the 250 µm halves were 119.7 ± 8.6 µm and 68.9 ± 2.9 µm for the 150 µm wire.  This data comes 53  very close to matching what was expected from the measurements of the wires after embossing.  Due to the potential for distortion due to the cutting of the wire and the poor image quality, it is quite possible that the data for the wire and the channel dimensions are well matched.  One thing interesting to note is that for the 250 µm channels one side is consistently deeper than the other by a significant margin,  If the data is separated one half measures 127.1 ± 1.9µm while the other measures 112.4 ± 1.9µm.  For the 150µm channels, all the data lies within the standard deviation.  It is believed that this could potentially be caused by inaccurate temperature control of the of the heated platens.  During initial testing it was discovered that a temperature difference of a few degrees would result in the channel being embossed almost entirely on one side.  The best temperature resolution displayed on the embossing system is one degree Fahrenheit so even with both platens reading the same temperature, the potential combined errror is already one Fahrenheit without adding any additional measurement error from the thermistors and PID controllers.  If a few degrees is capable of causing the embossing to be mainly on one side then it is possible that a temperature difference of up to one Fahrenheight (0.556°C) could conceivably shift the channel slightly. 54   Figure 19 Image indicating how the amount of corner rounding is calculated. A circle for the channel and a line across the edge are drawn to constrain the area. The arrow points to the area to be calculated.  Image J was used to determine the amount of material missing from each semicircular half due to gap formed during the embossing process as described in section 3.4.2.  This was done first using the ThreePointCircularROI tool to estimate where the edge of the circle would be. Then using the surface of the material and the curved edge as a guide the area is calculated.  Measuring the original embossed pieces using the parameters of 96.1°C, 14 MPa for 20 minutes and the samples made at 101.6°C, 14 MPa for 20 minutes results in missing areas of 334.3 ± 16.8 µm2 and 4.0 ± 0.1 µm2.  This shows that the embossing parameters of 101.6°C, 14 MPa for 20 minutes are in fact very close to optimal and represents a 99% decrease in rounding compared to the initial embossing parameters.  55  3.5.1.3 Tangent Chord Angle In order to create a perfectly round channel, a tangent chord angle (TCA) of 90° is required across the flat edge of the semicircle [7].  As a result, the TCA can be used as a measure to indicate roundness of the channel.  The TCA was estimated by estimating the tangent on the edge of the channel halves versus the line formed by the surface of the substrate.  In all these images, there is a slight amount of rounding at the corners that will throw off the measurement.  As a result of this, the corners are estimated to be where the curve would meet the channel surface.  From the images of the TCA of the 250 µm and 150 µm channels were 85.2 ± 1.4 µm and 85.2 ± 2.6 µm respectively so they are close to the ideal.   Figure 20 Example of a tangent chord angle measurement. θ denotes the TCA to be measured  3.5.1.4 Channel Surface Quality During the initial stages of this research, an SEM microscope was used to image a channel as shown below in Figure 21 which exhibited a pattern on the surface of the channel. As a result, unbonded embossed samples were imaged as shown in Figure 22 which showed the pattern as well. It was thought that the patterns could be a residue from the wire 56  manufacturing process so 250 µm wires were cleaned in acetone then IPA then DI H2O for 5 minutes each.  The wires were imaged and the result is that the pattern actually from the wire itself.  As a result, it is believed that the cracking originates from the drawing process used to form the wires.   Figure 21 SEM image of a bonded round channel that shows an unusual pattern on the interior of the channel  57   Figure 22 SEM images of an embossed PMMA half and the wire used to create it. The left image shows the pattern on the walls of a embossed piece of PMMA. Right is an image of the wire used to emboss the PMMA   Figure 23 SEM image of a cleaned 250 µm wire.  The wire exhibited similar patterns to wires that were not cleaned  58  3.5.1.5 Bonding Reproducibility and Alignment Accuracy Eleven 250 µm embossed straight channels were bonded using the procedure specified in section 3.2.2.  Of those 4 had severe misalignment that could be attributed to difficulty placing the sample in the embossing system with all the tubing and plates required to complete the bonding process.   In these instances it is very apparent that the samples have moved from their initial starting positions during placement in the embossing system. When the solvent is initially placed and the pieces are put together it is easy to feel by hand and see visually that the halves have locked in place together. Placing the samples in the embossing system is quite difficult as there are two pieces of substrate, two metal plates and two long pieces of tubing that all freely move and need to be centered on the platen.  The addition of the solvent also complicates things as it acts as a lubricant, thereby allowing the two substrate halves to slide more easily. Of the remaining samples all were bonded successfully without any leaks matching the results described in [52].  Afterwards the bonded channels were all cut by scoring each side heavily with 20 passes of a utility knife and placed in a table mounted clamp between two pieces of rectangular FR4.  The FR4 is simply used to provide a sharp square corner to minimize the bend radius when snapping the sample.  The outer edge of the clamp faces and the FR4 pieces are all aligned such that they line up with the scored line.  With everything clamped together, the sample is held on both sides above the scored line and sharply snapped to break it.  While this method works quite well to break the pieces without separating the halves, it is not perfect.  The line where each half snaps is not always identical so it is not always possible to successfully image both halves simultaneously due to the limited depth of field of the microscope.  In 59  these cases, the alignment is measured with a lower power objective lens but it is not always possible due to shadows in the images.  For these alignment measurements the alignment is measured as the difference between the two halves along the seam where they meet.  Measurement of the 250 µm channels provided an alignment error of 10.9 ± 6.4µm.  Two 150 µm channels were also created and measured this way with a resulting alignment error of 17.6 ± 5.4 µm  For the bonded 250 µm samples the width was measured to be 237.8 ± 13.7 µm and the height was 225.0 ± 9.8 µm.  The width was measured along the seam to the farthest edges which means that the alignment error is added to this.  As a result, the width is not a very good measure to use. The misalignment translates into a height measurement error due to the fact the lowest point and the highest may not be aligned, it should be relatively small.  From this it can be seen that the channel is compressed slightly by the bonding process.  This could be due to multiple factors.  In some of the images was difficult to focus both halves of the sample due to how the sample broke but it could also be an indication that the channel is collapsing slightly and that there is still some optimization to be done in the bonding process.  60   Figure 24 Bonded PMMA round channel that was embossed using 101.6°C 14 MPa for 20 minutes   Figure 25 Bonded PMMA round channel that was embossed using 96.1°C, 14 MPa for 20 minutes   61  3.5.1.6 Cast PDMS Round Channel Cross Sectional Profile Channels were aligned under a microscope and sectioned with a knife.  Sectioned samples were simply placed vertically on a glass slide and imaged.  Images were analyzed in the same manner as section 3.5.1.2.and 3.5.1.3.  Measurements provided a width of 134.3 ± 1.1µm, height of 72.9 ± 1.3µm, a TCA of 84.1 ± 3.6° and missing rounded area of 3.3 ± 0.1 µm2.  These measurements match with the measurements for the embossed PMMA pieces which indicates that replica molding using Sylgard 170 is sufficient for this task.   Figure 26 Bonded PDMS round channel. The shadow is caused by the curved section behind it  Due to the compliance of PDMS, getting proper alignment between the halves is significantly more difficult. As the wire alignment markers tend to be lifted when cast in PDMS, they are not present in the PDMS molds.  Furthermore, due to the elasticity of PDMS, having alignment markers does not guarantee alignment.  As a result, this is a task that is highly dependent on operator skill.  Currently what has been achieved is to align the majority of the 62  center of the chip containing the curving channels but the straight sections cannot be aligned properly. Long straight PDMS samples can be aligned but these also have issues with aligning the ends.  It is suspected that the usage of hot glue to keep the wires in place is the issue.  While the hot glue is meant to only hold the wires on the outside of the chip, it occasionally gets in between the halves during the embossing processes.  While bonding PMMA pieces still works when this residue is present, it may interfere when making PDMS.  Regardless, since these are PDMS chips, the sections of the ends that are severely misaligned are simply removed.    Length of the structure can also be a large factor in how difficult the bonding process will be.  With larger structures it is not possible to view the whole device underneath the microscope simultaneously.  Since the device is flexible, moving one section into alignment can potentially move some sections to be misaligned that are out of the field of view while leaving others further away still aligned.  This is vastly different from a rigid polymer as the entire sample will move around a single pivot point.  In the case of PDMS, theoretically there are an infinite amount of positions that can be moved slightly out of position in any direction.   3.5.1.7 Curved Channel Final Geometry What is very noticeable even prior to embossing the wire is that while the wire does conform to the form, due to handling and the springiness of the wire it is not necessarily as straight as the original form.  Furthermore, in both the PMMA and the PDMS chips it can be seen that in some sections of the embossed channel the wire has been stretched as part of the forming process.  Both curves shown below were made using the 3D printed forms.  The curve 63  showing stretching comes from roughly the middle of the form.  The presence of the stretching is not entirely surprising.  If the wire was to be crimped simply by immediately sandwiching the wire, consistently the wire would have a tendency to break near the center of the form.  To prevent this, the forms were brought together with a rocking motion in order to limit the stretching of the wire.  As a result, it is believed that a better way to do this would be to crimp one section of a curve at a time followed a final crimp pressing down on everything.  This would limit the amount of stretch in the wire as one end would not be held down.  One way of doing this would be to make the two forms mesh together like gears.   Figure 27 Curved PDMS round channels.  From left to right a PDMS curved channel without deformation and a curved channel with deformation that was caused by stretching of the wire while being crimped to form the curves  It should be noted that the 3D printed forms used to create the channels in Figure 27 as shown in Figure 28 do not actually match the original 3D model. The original 3D model simply had a repeated wave consisting of 1 mm semicircle curves. The long fingers on the other side also were supposed to be round 1mm semicircles as well. While the Objet 30 used 64  to print these objects specifies X x Y x Z resolution of 600 x 600 x 900 dpi (42 x 42 x28 µm) and a geometry dependent best case accuracy of 0.1 mm it is clear that it cannot make proper 1 mm semicircles.   Figure 28 Image of the 1 mm 3D printed forms clamped together under a microscope  65  To compare the fabricated PDMS channel to the form, images were taken of both of them individually and overlayed using the Gimp Image Editor as shown below in Figure 29.  From this it can be seen that the formed channel does conform to the shape of the wire.     Figure 29 Overlaid image of a PDMS round channel and the 3D printed form  3.5.1.8 Spiral Channels Creating spiral channels by winding the wire around a form is possible however some limitations have been discovered.  Due to the flexibility of the wire, keeping an exact spacing between the neighboring spirals is very difficult.  As a result, an approximate 0.8 mm spacer material was made from Eternal E9200 dry photoresist by folding the photoresist repeatedly which was then used to coil the wire.  Paper was also tried as a spacing material but it turned out to be very difficult to remove the paper without moving the wire out of position.  By using the dry photoresist, the spacer is simply removed by dissolving it in acetone.  While the proper developer could be used, it was found that acetone worked far faster than the 66  developer in removing the photoresist. One problem that occured was that occasionally pieces of photoresist would not dissolve in the acetone.  So as a compromise, the initial dissolution of the photoresist is done using acetone.  Once the wire falls off the cylinder the wire is then placed in the developer solution to remove any other traces of photoresist from the wire.  After embossing, it could be seen that the channel did not have the uniform spacing that was expected and the resulting spiral has a much larger spacing.  It is believed that the elasticity in the wire must be overcome if the spiral is desired to keep the spacing set by the photoresist similar to how the wires were pulled to make them straight.  However as the photoresist is quite sticky, the wire does not have the ability to slide as needed which is essential if the wire is to be pulled tight. Furthermore, this is also influenced by how well the wire can be wound to always be in the exact same position as the previous winding.  Variability in the turn to turn positioning of the wire while winding means that there is additional wire that will need to be displaced once it is flattened during embossing that further changes the spacing.  As a result, unless the spacing is very wide the loops of wire will collide. Due to the variability in height, the spiral did not turn out to be as circular as desired but instead as elongated.   67   Figure 30 Embossed spiral channel.  Wire has been left inside for imaging purposes.  3.5.1.9 Variable Width Channels As can be seen in Figure 31 below, etching a wire to create a variable width channel is possible.  However manually using a pen to perform the masking resulted in a highly non uniform channel.  While dry photoresist was tried as well a problem arose.  It is difficult to get the photoresist wrapped around the wire such that after development there would be no gaps for the etchant to seep in.  It is proposed that a better way to protect the wire would be to dip the wire in photoresist similar to how one would make a candle. 68   Figure 31 Variable width channel formed by protecting a wire with an etch resist pen and etching.    3.5.1.10 Joined Channels  Joined channels could be formed by embossing in two ways. Soldering wires together as can be seen in Figure 32 and overlayed channels in Figure 33.  The quality of soldered channels is highly dependent on the skill of the operator so while it is possible, it may not be the best option.  Using this method, the wires segments of various diameters could be crimped into the desired shape using 3D printed forms or other methods and then soldered or welded together.     Simply overlaying wires works as well however embossing near the intersection is not done as well as in the surrounding areas.  Rounding of the area near the intersection can be seen in the image below which indicates that it was not fully embossed near the intersection when using the optimized embossing parameters for the straight and curved channels.  This could 69  possibly be fixed by altering the embossing parameters.  As with the soldered channels, wires of various diameters could be crimped into the desired shape and then overlayed for embossing.  One potential issue with this method is that the area where the wires cross will be a different height than the surrounding area.  However, the final height actually lower than the height given by the simple addition of the heights of the two wires as the embossing process deforms the wires to fit together.   Figure 32 Channel joined by soldering 70   Figure 33 Channel joined by overlaying wires 71  Chapter 4: Conclusion As has been demonstrated, there are many alternative low cost methods that can be used to make microfluidic devices that each come with their own strengths and potential weaknesses.  4.1 Strengths and Limitations For the ultra low cost fabrication method described in section 2.1 the main benefit of it is simply the reducing the cost of performing photolithography for creating microfluidic devices.  With the equipment as it was developed there are limitations to the size of device that can be created.  Due to the use of a computer fan for spinning it is suspected that the lift force of the fan could cause problems for larger wafers.  The fan blades were not removed as it could potentially throw off the balance of the motor.  Furthermore, there is a limitation on the max rotational speed achievable and weight of the sample may slow the fan down as it is not designed for this purpose.  A better option would be to use a hard drive spindle motor as many consumer units can reach 7200RPM and some contain a hole in the middle which could potentially be used to clamp the sample down using vacuum pressure.  Another issue is that the UV lamp used would not be suitable for exposing large wafers as the intensity would be much lower on the edges of a 4" wafer however this could be easily remedied by using multiple lamps.  The wire based hot embossing method is the most unique out of all the methods demonstrated.  On its own it is a low cost method that at the minimum would require just a hot embossing system which could also be built at low cost if desired.  It is also a very fast method to make an initial prototype as the mold is a simple off the shelf wire.  By using the 72  embossed PMMA to cast PDMS, this becomes a method that can create multiple copies of the same round channel device unlike many other methods. This method can be taken a further step by using PDMS to make epoxy molds.  By doing this it would be possible to replicate the channels in PMMA.  However, as is the alignment markers would not be in position as they tend to be lifted out of position by the PDMS. Furthermore, with this method it is possible to make curved round channels unlike with glass capillaries.  Figure 34 Process flow to go from the initial embossing process to mass produced polymer round channels  This method is not without limitations though.  Decreasing the target diameter means selecting a thinner wire which makes handling more difficult.  With the currently selected embossing parameters, there is still some slight rounding of the corners and in some cases the two sides are not equally round and the channel heights are not equal as expected.  Furthermore, unlike the sucrose based channels [31] and the solder based channels [30], making PDMS channels is difficult due to difficulties with the alignment process. Emboss PMMACast PDMSCast PDMSCast EpoxyEmboss PMMA73   In all there is no perfect fabrication method.  Instead it is simply a case of selecting the best method for the application at hand.   74  4.2 Potential Applications One application would be for the use in vein modeling.  Using the wire based hot embossing method described here would allow for the creation of reproducible vein model chips in a variety of materials.  While there is potential difficulty in making the initial wire design, once the first sample is made the rest can be replicated from that.  To preserve the original embossed pieces and allow for embossing in a wide variety of polymers, after the initial casting in PDMS an epoxy mold can be created.  Furthermore, by joining various diameters of wire different size veins can be modeled in the same chip.  Another application would be for flow focusing. While much work has been done on rectangular channels, not much has been performed on circular cross section channels.  With the work done here, it is possible to create straight, serpentine, and spiral round channels that can be reproduced in a variety of materials as needed. Furthermore, the channel diameters can be easily altered by simply purchasing different diameters of wire.  Rapid prototyping is another potential use for the wire embossing method.  Regardless if PMMA is placed on one side or both, channels can be formed relatively quickly and cheaply. No photolithography equipment or masks are required and if polymers are being used, simple solvent or thermal bonding methods can be used.     75  4.3 Future Research For the circular cross section channels, one area of research would be embossing in higher temperature materials.  Since PMMA is used to cast the PDMS, it limits the use of the alignment markers as the wires tend to move out of position during the casting.  If a higher temperature material was used to make the initial mold such as cyclic olefin co-polymer (COC), it could be used to mold a lower temperature polymer such as PMMA.  In this manner the alignment markers could be preserved as there is no way for them to move out of position from the second cast.  At this point PDMS could be cast from the PMMA and then either used or again used to make a durable epoxy mold.  Furthermore, if another material that was chemically resistant to the PDMS was found to replace the use of PMMA, the double casting step could be removed entirely.  That would remove approximately two days from the mold making process and it would most likely be more durable than the PDMS mold as the aging process makes it brittle.  Alternatively, if another polymer is used it could be feasible to keep all the molding steps in polymers as shown in the figure below.  If this is done, reproducible samples can be made very fast as the limiting would be simply how fast the hot embossing system can be heated an cooled.    76   Figure 35 Proposed embossing procedure to quickly create molds for embossing PMMA or casting PDMS round channels  Another idea is to combine the hot embossed round channel with another fabrication method such as laser ablation, CNC machining or even photolithography.  The purpose of this is to be able to combine the reproducibility of making these round channels with a more traditional microfluidic channel fabrication method on one or both of the halves.  The reason for this is while this method of making round channels works, the process of joining different channels together is a time consuming manual process. So as the complexity of the device increases this process would get infinitely harder.  It is likely that only critical sections of the device would require round channels while for other sections it would not matter. So by doing this the complexity required to manufacture the device is reduced and samples made in this method can still be replicated in both PDMS and thermoplastics by using the double casting method.  Emboss COCEmboss PCEmboss PMMA orcast PDMS77  References [1] G. S. Fiorini and D. T. Chiu, “Disposable microfluidic devices: fabrication, function, and application,” BioTechniques, vol. 38, no. 3, pp. 429–446, 2005. [2] E. Sollier, C. Murray, P. Maoddi, and D. Di Carlo, “Rapid prototyping polymers for microfluidic devices and high pressure injections,” Lab. Chip, vol. 11, no. 22, p. 3752, 2011. [3] D. B. Weibel, W. R. DiLuzio, and G. M. Whitesides, “Microfabrication meets microbiology,” Nat. Rev. Microbiol., vol. 5, no. 3, pp. 209–218, Mar. 2007. [4] R. K. Jena, C. Y. Yue, Y. C. Lam, P. S. Tang, and A. Gupta, “Comparison of different molds (epoxy, polymer and silicon) for microfabrication by hot embossing technique,” Sens. Actuators B Chem., vol. 163, no. 1, pp. 233–241, Mar. 2012. [5] Y.-C. Hsu and T.-Y. Chen, “Applying Taguchi methods for solvent-assisted PMMA bonding technique for static and dynamic μ-TAS devices,” Biomed. Microdevices, vol. 9, no. 4, pp. 513–522, Aug. 2007. [6] Y. Qin, Micromanufacturing Engineering and Technology. William Andrew, 2010. [7] M. D. Ville, P. Coquet, P. Brunet, and R. Boukherroub, “Simple and low-cost fabrication of PDMS microfluidic round channels by surface-wetting parameters optimization,” Microfluid. Nanofluidics, vol. 12, no. 6, pp. 953–961, May 2012. [8] M. Abdelgawad, C. Wu, W.-Y. Chien, W. R. Geddie, M. A. S. Jewett, and Y. Sun, “A fast and simple method to fabricate circular microchannels in polydimethylsiloxane (PDMS),” Lab. Chip, vol. 11, no. 3, p. 545, 2011. [9] G.-J. Wang, K.-H. Ho, S. Hsu, and K.-P. Wang, “Microvessel scaffold with circular microchannels by photoresist melting,” Biomed. Microdevices, vol. 9, no. 5, pp. 657–663, Oct. 2007. [10] L. K. Fiddes, N. Raz, S. Srigunapalan, E. Tumarkan, C. A. Simmons, A. R. Wheeler, and E. Kumacheva, “A circular cross-section PDMS microfluidics system for replication of cardiovascular flow conditions,” Biomaterials, vol. 31, no. 13, pp. 3459–3464, May 2010. [11] Z. Huang, X. Li, M. Martins-Green, and Y. Liu, “Microfabrication of cylindrical microfluidic channel networks for microvascular research,” Biomed. Microdevices, vol. 14, no. 5, pp. 873–883, Oct. 2012. [12] A. Zahra, G. de Cesare, D. Caputo, and A. Nacetti, “Formation of Round Channel for Microfluidic Applications,” Int. J. Math. Comput. Phys. Quantum Eng., vol. 8, no. 7, pp. 932–936, 2014. [13] L. Wu, G. Guan, H. W. Hou, A. A. S. Bhagat, and J. Han, “Separation of Leukocytes from Blood Using Spiral Channel with Trapezoid Cross-Section,” Anal. Chem., vol. 84, no. 21, pp. 9324–9331, Nov. 2012. [14] A. A. S. Bhagat, S. S. Kuntaegowdanahalli, and I. Papautsky, “Inertial microfluidics for continuous particle filtration and extraction,” Microfluid. Nanofluidics, vol. 7, no. 2, pp. 217–226, Aug. 2009. [15] D. Di Carlo, J. F. Edd, D. Irimia, R. G. Tompkins, and M. Toner, “Equilibrium Separation and Filtration of Particles Using Differential Inertial Focusing,” Anal. Chem., vol. 80, no. 6, pp. 2204–2211, Mar. 2008. 78  [16] D. D. Carlo, D. Irimia, R. G. Tompkins, and M. Toner, “Continuous inertial focusing, ordering, and separation of particles in microchannels,” Proc. Natl. Acad. Sci., vol. 104, no. 48, pp. 18892–18897, Nov. 2007. [17] S. S. Kuntaegowdanahalli, A. A. S. Bhagat, G. Kumar, and I. Papautsky, “Inertial microfluidics for continuous particle separation in spiral microchannels,” Lab. Chip, vol. 9, no. 20, p. 2973, 2009. [18] G. Guan, L. Wu, A. A. Bhagat, Z. Li, P. C. Y. Chen, S. Chao, C. J. Ong, and J. Han, “Spiral microchannel with rectangular and trapezoidal cross-sections for size based particle separation,” Sci. Rep., vol. 3, Mar. 2013. [19] G. Segré and A. Silberberg, “Behaviour of macroscopic rigid spheres in Poiseuille flow Part 2. Experimental results and interpretation,” J. Fluid Mech., vol. 14, no. 01, pp. 136–157, 1962. [20] A. Morarka, S. Agrawal, S. Kale, A. Kale, S. Ogale, K. Paknikar, and D. Bodas, “Quantum dot based immunosensor using 3D circular microchannels fabricated in PDMS,” Biosens. Bioelectron., vol. 26, no. 6, pp. 3050–3053, Feb. 2011. [21] M. Haiducu, M. Rahbar, I. G. Foulds, R. W. Johnstone, D. Sameoto, and M. Parameswaran, “Deep-UV patterning of commercial grade PMMA for low-cost, large-scale microfluidics,” J. Micromechanics Microengineering, vol. 18, no. 11, p. 115029, Nov. 2008. [22] Z. Chen, L. Zhang, and G. Chen, “A spring-driven press device for hot embossing and thermal bonding of PMMA microfluidic chips,” ELECTROPHORESIS, vol. 31, no. 15, pp. 2512–2519, Jul. 2010. [23] “UV Ozone Cleaners | Compact and Low Cost UV-Ozone Cleaning Systems.” [Online]. Available: http://www.uvotech.com/uv_ozone_cleaners/. [Accessed: 15-Jul-2014]. [24] D. A. Bartholomeusz, R. W. Boutte, and J. D. Andrade, “Xurography: rapid prototyping of microstructures using a cutting plotter,” J. Microelectromechanical Syst., vol. 14, no. 6, pp. 1364–1374, Dec. 2005. [25] “UV FAQs | International Ultraviolet Association.” [Online]. Available: https://www.iuva.org/uv-faqs. [Accessed: 15-Jul-2014]. [26] “Optical Materials.” [Online]. Available: http://www.newport.com/Optical-Materials/144943/1033/content.aspx. [Accessed: 15-Jul-2014]. [27] “Pyrex® (Corning 7740).” [Online]. Available: https://www.escooptics.com/material-data/pyrex-corning-7740.html. [Accessed: 15-Jul-2014]. [28] T. Koerner, L. Brown, R. Xie, and R. D. Oleschuk, “Epoxy resins as stamps for hot embossing of microstructures and microfluidic channels,” Sens. Actuators B Chem., vol. 107, no. 2, pp. 632–639, Jun. 2005. [29] N. Futai, W. Gu, and S. Takayama, “Rapid Prototyping of Microstructures with Bell-Shaped Cross-Sections and Its Application to Deformation-Based Microfluidic Valves,” Adv. Mater., vol. 16, no. 15, pp. 1320–1323, Aug. 2004. [30] S.-H. Song, C.-K. Lee, T.-J. Kim, I. Shin, S.-C. Jun, and H.-I. Jung, “A rapid and simple fabrication method for 3-dimensional circular microfluidic channel using metal wire removal process,” Microfluid. Nanofluidics, vol. 9, no. 2–3, pp. 533–540, Aug. 2010. 79  [31] J. Lee, J. Paek, and J. Kim, “Sucrose-based fabrication of 3D-networked, cylindrical microfluidic channels for rapid prototyping of lab-on-a-chip and vaso-mimetic devices,” Lab. Chip, vol. 12, no. 15, pp. 2638–2642, Jul. 2012. [32] J.-H. Huang, J. Kim, Y. Ding, A. Jayaraman, and V. M. Ugaz, “Embedding Synthetic Microvascular Networks in Poly(Lactic Acid) Substrates with Rounded Cross-Sections for Cell Culture Applications,” PLoS ONE, vol. 8, no. 9, p. e73188, Sep. 2013. [33] M. E. Wilson, N. Kota, Y. Kim, Y. Wang, D. B. Stolz, P. R. LeDuc, and O. B. Ozdoganlar, “Fabrication of circular microfluidic channels by combining mechanical micromilling and soft lithography,” Lab. Chip, vol. 11, no. 8, p. 1550, 2011. [34] A. Sodemann, M. Li, R. Mayor, and C. R. Forest, “Micromilling of molds for microfluidic blood diagnostic devices,” Proc Annu Meet Am Soc Precis Eng, pp. 4–9, 2009. [35] “Carbide Micro End Mill, Carbide Micro End Mills, Micro EndMill.” [Online]. Available: https://www.american-carbide.com/MicroEndMills/MEBallProducts.php. [Accessed: 15-Aug-2014]. [36] “Types and Characteristics of End Mills.” [Online]. Available: http://www.endmill.com/pages/training/types.html. [Accessed: 15-Aug-2014]. [37] “Performance Micro Tool - Carbide Ball Nose Mills.” [Online]. Available: http://www.pmtnow.com/end-mill/TR-BN. [Accessed: 15-Aug-2014]. [38] J. L. Erkal, A. Selimovic, B. C. Gross, S. Y. Lockwood, E. L. Walton, S. McNamara, R. S. Martin, and D. M. Spence, “3D printed microfluidic devices with integrated versatile and reusable electrodes,” Lab. Chip, vol. 14, no. 12, p. 2023, 2014. [39] A. I. Shallan, P. Smejkal, M. Corban, R. M. Guijt, and M. C. Breadmore, “Cost-Effective Three-Dimensional Printing of Visibly Transparent Microchips within Minutes,” Anal. Chem., vol. 86, no. 6, pp. 3124–3130, Mar. 2014. [40] “Wire Bonding Services | Gold Wire Bonding | Copper Wire Bonding | Quik-Pak.” [Online]. Available: http://www.icproto.com/capabilities-services/ic-assembly/wire-bonding/. [Accessed: 06-Jul-2014]. [41] “How copper cable is made - Cabling Install.” [Online]. Available: http://www.cablinginstall.com/articles/print/volume-5/issue-9/contents/design/how-copper-cable-is-made.html. [Accessed: 06-Jul-2014]. [42] D. Roylance, “Stress-strain curves,” Mass. Inst. Technol. Study Camb., 2001. [43] “Arduino-Powered Wire Bending Machine | MAKE.” [Online]. Available: http://makezine.com/2012/05/03/arduino-powered-wire-bending-machine/. [Accessed: 06-Jul-2014]. [44] J. Zhang, M. Li, W. H. Li, and G. Alici, “Inertial focusing in a straight channel with asymmetrical expansion–contraction cavity arrays using two secondary flows,” J. Micromechanics Microengineering, vol. 23, no. 8, p. 085023, Aug. 2013. [45] J.-C. Lu, W.-H. Liao, and Y.-C. Tung, “Magnet-assisted device-level alignment for the fabrication of membrane-sandwiched polydimethylsiloxane microfluidic devices,” J. Micromechanics Microengineering, vol. 22, no. 7, p. 075006, Jul. 2012. [46] J.-S. Kim and D. R. Knapp, “Microfabrication of Electrospray Ionization Emitters in Polydimethylsiloxane Microfluidic Devices Using a Self-Aligning Assembly System,” in Micro Total Analysis Systems 2001, J. M. Ramsey and A. van den Berg, Eds. Springer Netherlands, 2001, pp. 333–334. 80  [47] M. T. Koesdjojo, C. R. Koch, and V. T. Remcho, “Technique for Microfabrication of Polymeric-Based Microchips from an SU-8 Master with Temperature-Assisted Vaporized Organic Solvent Bonding,” Anal. Chem., vol. 81, no. 4, pp. 1652–1659, Feb. 2009. [48] X. Sun, B. A. Peeni, W. Yang, H. A. Becerril, and A. T. Woolley, “Rapid prototyping of poly(methyl methacrylate) microfluidic systems using solvent imprinting and bonding,” J. Chromatogr. A, vol. 1162, no. 2, pp. 162–166, Aug. 2007. [49] M. T. Koesdjojo, Y. H. Tennico, and V. T. Remcho, “Fabrication of a Microfluidic System for Capillary Electrophoresis Using a Two-Stage Embossing Technique and Solvent Welding on Poly(methyl methacrylate) with Water as a Sacrificial Layer,” Anal. Chem., vol. 80, no. 7, pp. 2311–2318, Apr. 2008. [50] S. H. Ng, R. T. Tjeung, Z. F. Wang, A. C. W. Lu, I. Rodriguez, and N. F. de Rooij, “Thermally activated solvent bonding of polymers,” Microsyst. Technol., vol. 14, no. 6, pp. 753–759, Jun. 2008. [51] F. Umbrecht, D. Müller, F. Gattiker, C. M. Boutry, J. Neuenschwander, U. Sennhauser, and C. Hierold, “Solvent assisted bonding of polymethylmethacrylate: Characterization using the response surface methodology,” Sens. Actuators Phys., vol. 156, no. 1, pp. 121–128, Nov. 2009. [52] S. H. Ng, R. T. Tjeung, and Z. Wang, “Formation of Embedded Microchannels by a Solvent Displacement Bonding Technique,” in Electronics Packaging Technology Conference, 2007. EPTC 2007. 9th, 2007, pp. 211–214. [53] “PDMS in biology researches: A critical review on PDMS soft lithography for biological studies | Elveflow microfluidic instruments.” . [54] K. J. Regehr, M. Domenech, J. T. Koepsel, K. C. Carver, S. J. Ellison-Zelski, W. L. Murphy, L. A. Schuler, E. T. Alarid, and D. J. Beebe, “Biological implications of polydimethylsiloxane-based microfluidic cell culture,” Lab. Chip, vol. 9, no. 15, pp. 2132–2139, Aug. 2009. [55] M. W. Toepke and D. J. Beebe, “PDMS absorption of small molecules and consequences in microfluidic applications,” Lab. Chip, vol. 6, no. 12, pp. 1484–1486, Nov. 2006. [56] “Polymer Clay & Release Agents.” [Online]. Available: http://www.polymerclayweb.com/tools/supplies/releaseagents.aspx. [Accessed: 09-Jul-2014]. [57] K. Zió\lkowska, K. Żukowski, M. Chudy, A. Dybko, and Z. Brzózka, “Enhancing efficiency of double casting prototyping by thermal aging of poly (dimethylsiloxane),” in Proc. MicroTAS, 2011, pp. 1164–1166. [58] H. J. Vandenburg, A. A. Clifford, K. D. Bartle, R. E. Carlson, J. Carroll, and I. D. Newton, “A simple solvent selection method for accelerated solvent extraction of additives from polymers,” Analyst, vol. 124, no. 11, pp. 1707–1710, Jan. 1999. [59] A. P. Sudarsan and V. M. Ugaz, “Printed Circuit Technology for Fabrication of Plastic-Based Microfluidic Devices,” Anal. Chem., vol. 76, no. 11, pp. 3229–3235, Jun. 2004. [60] C.-W. Li, C. N. Cheung, J. Yang, C. H. Tzang, and M. Yang, “PDMS-based microfluidic device with multi-height structures fabricated by single-step photolithography using printed circuit board as masters,” Analyst, vol. 128, no. 9, pp. 1137–1142, Aug. 2003. 81  [61] “OSH Park ~ Design Submission Guidelines.” [Online]. Available: https://www.oshpark.com/guidelines. [Accessed: 16-Jul-2014]. [62] “ITEAD Studio 2Layer Green PCB 5cm x 5cm Max.” [Online]. Available: http://imall.iteadstudio.com/open-pcb/pcb-prototyping/im120418001.html. [Accessed: 16-Jul-2014].    82  Appendix A  Multiple Height PCB Based Molds Another method that could be used to make molds without the leveling problems of the epoxy molds is to simply etch a printed circuit board (PCB) [59] or even etch with undercuts to make multilayer molds with PCBs [60].  Blank copper clad circuit boards can be easily purchased with copper thicknesses of 17.2, 34.3,68.6 and 137.2 µm which are then patterned and etched to form the mold [59].  Alternatively, due to the availability of services providing low cost PCB fabrication, it is possible to simply make a design in CAD and receive a board to use as a mold provided that one follows the company's design rules for trace widths.  Using low cost services such as [61] or [62] one can get 35 µm heights with 150 µm minimum widths for less than 25 USD.  More expensive services can create widths of 75 µm but the high setup fees of upwards of 300 USD make these services less feasible.  Using an actual PCB fabrication service creates an opportunity to make molds for multiple height microfluidic channels cheaply if the PCB design constraints are acceptable.  While multiple height molds have been made with PCBs [60], this would allow for simply designing and purchasing the board without needing to do any fabrication work. Completed PCBs usually have an additional two layers to the copper layer: the solder mask and the silk screen.  Solder mask is used to protect the copper from solder wherever it is not desired such as on the circuit traces. Silkscreen on the other hand is simply a coloured ink that is used to label circuit boards.  Furthermore, with the introduction of reduction of hazardous substances (ROHS ) requirements in Europe, circuit boards made to meet that standard have a Tg of a minimum of 180°C in order to tolerate the temperatures needed for ROHS compliant soldering which makes it ideal for hot embossing.  However, because there are different ways 83  to apply the solder masks and silkscreen, the thicknesses of these two layers may potentially vary between suppliers.  On a circuit board there are different material combinations that are possible.  Solder mask can be on top of bare FR4 as well as on top of a copper trace.  Silkscreen however is always on top of solder mask. While it would be possible to place silkscreen on top of bare FR4, the fabrication shop used would not guarantee that it would actually stick to the board. Using two different circuit boards from Itead Studio the boards were profiled to determine feature heights as specified in the table below.    Different circuit boards that were purchased at different times were used as the PCB manufacturing process typically makes boards in large quantities on a single panel and then routed to size. Purchasing boards at different times ensures that the samples are not from the same manufacturing run so that some measure of batch to batch variability can be seen.  More boards than what is ordered are typically made to compensate for potential board failure during electrical short circuit testing.  If the same board is ordered both times, if there were excess functional boards from the previous run those may be provided when the second batch is purchased. 84   Figure 36 Possible layer combinations when using a PCB as a multilayer mold   Table 3 Different feature heights on two circuit boards Material Combination Height of Board 1 (µm) Height Board 2 (µm) Copper over FR4 37 37 Soldermask over FR4 16 10 Soldermask over copper 14 12 Silkscreen over soldermask 12 14 Silkscreen over soldermask on top of copper 13 12  As can be seen from the data, the two boards have very similar thicknesses for the various layer combinations.  If these limitations are acceptable, purchased circuit boards can be used to make multilevel microfluidic devices relatively cheaply.    Copper over FR4Copper over FR4 with soldermaskCopper over FR4 with soldermask and silkscreenSilkscreenSoldermaskCopperFR4Legend

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0166937/manifest

Comment

Related Items