Embed Size (px)
Citation preview
UNIVERSITY OF CALIFORNIA
Los Angeles
Manipulation of Microscopic Gas Bubbles by Using Surface Tension:
Capturing, Venting and Pumping
A dissertation submitted in partial satisfaction of the
requirements for the degree Doctor of Philosophy
in Mechanical Engineering
by
Dennis Desheng Meng
2005
ii
The dissertation of Dennis Desheng Meng is approved
Chih-Ming Ho
Harold Monbouquette
Pirouz Kavehpour
Chang-Jin Kim, Committee Chair
University of California, Los Angeles
2005
iii
To my wife, Melanie, our parents and my late grand parents
iv
TABLE OF CONTENTS
TABLE OF CONTENTS iv
LIST OF FIGURES vii
LIST OF TABLES x
ACKNOWLEDGEMENTS xi
VITA xiii
ABSTRACT xiv
Chapter 1. Introduction 1
1.1 Microscopic Gas Bubbles as Powerful Tools
1
1.2 Scaling Effect of Surface Tension 4
1.3 Control of Surface Tension 6
1.4 Overview of This Dissertation 9
References 11
Chapter 2. Bubble-Traps and Bubble Capturing Potential 15
2.1 Background 15
2.2 Bubble-Traps: Definition and Qualitative Explanation 15
2.3 Bubble Capturing Potential: Φbc 18
2.4 Quantitative Analysis of Bubble-Traps 20
2.5 Simulation Results 25
v
2.6 Experimental verifications 26
2.7 Summary and Future Directions 28
References 29
Chapter 3. Hydrophobic Venting 30
3.1 Introduction 30
3.1.1 Bubble Clogging of Microchannels 30
3.1.2 μDMFC: a System with Continuous Bubble Generation 34
3.2 Distributed Hydrophobic Venting 37
3.3 Silicon-Based Distributed Breather 41
3.4 Membrane Sandwiched Breather 45
3.5 Gas-Permeable Microchannels 52
3.6 Summary and Future Directions 58
References 59
Chapter 4. Micropumping Enabled by
Hydrophobic Venting of Bubbles 63
4.1 Introduction 63
4.2 Comparative Study of Electrolysis and Boiling for Actuation 66
4.2.1 Definitions and Assumptions 66
4.2.2 Electrolysis for Bubble-Driven Actuation 67
4.2.3 Boiling for Bubble-Driven Actuation 69
4.2.4 Comparison between Electrolysis and Boiling 72
4.3 Electrochemical-Bubble Driven Pump 74
vi
4.3.1 Essential Components 74
4.3.2 Pumping Concept 75
4.3.3 Pump Loop Configuration and Fabrication 78
4.3.4 Preparation of Test 79
4.3.5 Verification of Liquid Circulation in Pump Loop 80
4.3.6 Characterization of Pump Loop 82
4.3.7 Characterization in Open Loop 85
4.4 Exploration of More Pumping Approaches 88
4.5 Summary and Future Directions 91
References 93
Chapter 5. Conclusion and Outlooks 100
6.1 Conclusion 100
6.2 Outlooks 103
6.2.1 Bubble-Powered μTAS 103
6.2.2 Innovative Designs of μDMFC 104
References 105
vii
LIST OF FIGURES
Figure Description Page
Figure 1-1 Bubbles under various pressures. 2
Figure 1-2 A microcapillary lens for X-rays. 3
Figure 1-3 In vivo non–slice-selective coronal images of rat vasculature acquired using 3He microbubbles suspended in Hexabrix. 3
Figure 1-4 Fundamental testing of EWOD principle on a sessile droplet. 8
Figure 1-5 Basic microfluidic functions performed by EWOD. 9
Figure 2-1 Contact angle of a gas bubble’s three-phase interface. 16
Figure 2-2 Bubbles’ status in a liquid-filled container 17
Figure 2-3 A gas bubble on a flat surface. 20
Figure 2-4 A gas bubble in a concave conic pit. 21
Figure 2-5 Bubble shape at different contact angles 24
Figure 2-6 Simulation result of Φbc on flat or concave surfaces with different contact angles. 26
Figure 2-7 Gas bubbles captured on an array of type I bubble-traps. 27
Figure 2-8 Gas bubbles captured on an array of type II bubble-traps. 27
Figure 3-1 Bubble formation during priming of two liquid samples. 31
Figure 3-2 “Channel in channel” design to prevent bubble clogging. 33
Figure 3-3 Microchannel with a bubble-trapping region. 33
Figure 3-4 Working principle of DMFC (Direct Methanol Fuel Cell). 35
Figure 3-5 Gas bubble filter for nozzle-diffuser bubble pump. 37
viii
Figure 3-6 Handling of picoliter liquid samples by using HMCV (Hydrophobic MicroCapillary Venting). 38
Figure 3-7 Micro-degassing for portable dialysis system. 39
Figure 3-8 Breathing and liquid holding mechanism in a hydrophobic venting hole. 39
Figure 3-9 Process flow for the first-generation venting plate. 42
Figure 3-10 Experimental setup for distributed venting. 43
Figure 3-11 Venting experiment in the first-generation distributed breather. 44
Figure 3-12 Configuration of the second-generation venting plate 46
Figure 3-13 Venting experiment in the second-generation breather. 47
Figure 3-14 Determination of leakage onset pressure. 48
Figure 3-15 Flow-pressure curves for DI water in porous-membranes-covered microchambers. 49
Figure 3-16 An irregular venting hole in porous membrane. 50
Figure 3-17 SEM pictures of porous membranes. 51
Figure 3-18 Alignment and bonding setup for the gas permeable microchannel. 52
Figure 3-19 The gas permeable microchannel with an on-chip bubble injector. 53
Figure 3-20 Venting of a bubble in a gas permeable microchannel. 54
Figure 3-21 A typical time-space diagram of a venting bubble. 55
Figure 3-22 The venting threshold. 56
Figure 3-23 Hypothetical explanations for the venting threshold. 57
Figure 4-1 Schematic view of the test chips for bubble-driven actuations. 67
Figure 4-2 Speed control of electrochemical bubble actuation. 69
Figure 4-3 Model for the thermal bubble actuation. 70
ix
Figure 4-4 Comparison between electrochemical and thermal bubble actuation. 73
Figure 4-5 The virtual check valve for gas bubbles. 75
Figure 4-6 Pumping by directional growth and hydrophobic venting of gas bubbles: the concept. 77
Figure 4-7 Configuration of pump loop. 79
Figure 4-8 Bubble motion in the pumping section (area A in Figure 4-7). 81
Figure 4-9 Fluid uptake from the reservoir (area B in Figure 4-7). 82
Figure 4-10 μ-PIV to determine the flow rate (area C in Figure 4-7). 83
Figure 4-11 Open loop test setup. 86
Figure 4-12 The flow rate vs. the pressure head in open loop test. 87
Figure 4-13 Pumping by gas injection-venting. 89
Figure 4-14 Visualization of pumping effect by gas injection-venting. 90
Figure 5-1 Figure 5-1. Ultra-compact μDMFC with integrated fuel stack. 105
x
LIST OF TABLES
Table Description Page
Table 3-1 Leakage onset pressure: calculated and measured value 50
Table 4-1 Experimental data for electrochemical actuation 68
Table 4-2 Experimental data for thermal actuation 71
Table 4-3 Control of the volumetric flow rate in pump loop 85
xi
ACKNOWLEDGEMENTS
I would like to give my sincerest appreciation to my advisor, Professor Chang-Jin
Kim, for his guidance, encouragement and support, which has made my experience in
UCLA enjoyable and fruitful. The conversations with him are always enlightening and
inspiring. He himself is a fine example of researcher with creativity, integrity and
responsibility, from which I will benefit far beyond the scope of this dissertation.
I am thankful to Professor Chih-Ming Ho for not only his valuable comments on
my research but also his kind suggestions on my career. My thanks are also due to
Professor Harold Monbouquette and Professor Pirouz Kavehpour for serving in my
committee and giving valuable advices.
Appreciation also goes to our collaborators in micro direct methanol fuel cell
(μDMFC) project. This work could not have been accomplished without the valuable
inputs from Professors Chih-Ming Ho, Xiang Zhang, Xiaolin Zhong, Chaoyang Wang, Drs.
Thomas Cubaud, Ta-Jen Yen, Mahidhar Tatineni and Guoqiang Lu.
My former and current colleagues in Micromanufacturing Lab deserve great
gratitude for providing me with a friendly and encouraging environment. Drs. Da-Jeng
Yao, Sung Kwon Cho, Yen-Wen Lu, Uichong Yi, Joonwon Kim, Wenjiang Shen and
Shih-Kang Fan have endured my endless questions about microfabrication. Brian Van Dyk,
Annie Lee, James Jenkins, Hisang-Wei Lu and Gaurav Shah have assisted me to revise my
research writing. Rihui He, Hyejin Moon Jane Tsai, Fardad Chamran, Prosenjit Sen, Jian
xii
Gong, Chang-Hwan Choi and all the other present or past lab members have given me
spiritual support and generous help during my research and dissertation writing.
Special thanks extend to the MEMS and nanotechnology community in UCLA. I
benefited from MEMS courses offered by Professor Jack Judy, molecular biology
courses in biomedical engineering department, guest lecture given by Dr. Sung Kwon
Cho and numerous seminars organized by California NanoSystems Institute (CNSI) and
the Institute for Cell Mimetic Space Exploration (CMISE). This enriched environment is
such a blessing for students and young researchers.
I would also like to express my thanks to my beloved family. My parents, brother
and late grand parents have put so much hope on me and sacrificed a lot. My wife,
Melanie, gave up her own career dream in China and came to U.S. to realize our dream
together.
xiii
VITA
1974 Born, Bei’an, Heilongjiang Province of China
1998 B.S. in Mechanical Engineering Tsinghua University Beijing, China
2001 M.S. in Mechanical Engineering Tsinghua University Beijing, China
2001-2005 Graduate Student Researcher Mechanical and Aerospace Engineering Department University of California, Los Angles
PUBLICATIONS AND PRESENTATIONS
D. D. Meng, J. Kim, and C.-J. Kim, "A Distributed Gas Breather for the Micro Direct Methanol Fuel Cell," Proc. The 16th IEEE Int. Conf. on Micro Electro Mechanical Systems, Kyoto, Japan, Jan., 2003, pp. 534-7.
D. D. Meng, T. Cubaud, C.-M. Ho, and C.-J. Kim, "A Membrane Breather for
Micro Fuel Cell with High Concentration Methanol," Tech. Dig. Solid State Sensor, Actuator and Microsystems Workshop, Hilton Head Island, SC, Jun., 2004, pp. 141-4.
D. D. Meng and C.-J. C. Kim, "Self-aligned Micro Bubble Arrays by Using
Surface Tension," 2004 ASME Int. Mechanical Engineering Congress and Exposition, Anaheim, CA, Nov., 2004, CD: IMECE 2004-62182.
D. D. Meng and C.-J. Kim, "Micropumping by Directional Growth and
Hydrophobic Venting of Bubbles," Proc. The 18th IEEE Int. Conf. on Micro Electro Mechanical Systems, Miami, FL, Jan., 2005, pp. 423-6.
D. D. Meng, Y. Ju, and C.-J. Kim, "A Comparative Study of Electrolysis and
Boiling for Bubble-Driven Microactuations," Tech. Dig. The13th Int. Conf. on Solid-State Sensors, Actuators and Microsystems, Seoul, Korea, Jun., 2005.
D. D. Meng, "Manipulation of microscopic gas bubbles", UCLA Mechanical and
Aerospace Engineering Department Seminar, May 12, 2005
xiv
ABSTRACT OF THE DISSERTATION
Manipulation of Microscopic Gas Bubbles by Using Surface Tension:
Capturing, Venting and Pumping
by
Dennis Desheng Meng
Doctor of Philosophy in Mechanical Engineering
University of California, Los Angeles, 2005
Professor Chang-Jin Kim, Chair
Microscopic gas bubbles have recently been recognized as powerful tools for a
variety of applications, such as micro-lenses, visualization particles, spacers, actuation
pistons and pressure sensors. However, reliable manipulation of these marvelous bubbles
is still challenging, hindering their employment in the real-world devices. In this study,
surface tension force is exploited to manipulate gas bubbles, because it is inherently
dominant over other forces in submillimeter scale.
Bubble capturing potential is first introduced to quantify the immobilization of a
gas bubble on a solid surface. Reliable formation of bubble arrays in a liquid environment
is demonstrated. A universal gas removal approach termed “hydrophobic nanoporous
venting” is developed, which can promptly breathe out virtually any kind of gas bubbles.
The leakage onset pressure of up to ~35psi ensures its applications in practical
microfluidic devices.
xv
Bubble actuations in closed-loop microfluidics were traditionally restricted to
thermal bubbles using the energy-hungry boiling process and limited by their slow
condensation. The two basic manipulations, capturing and venting, have enabled new
bubble generation approaches, such as electrolysis, injection and chemical reactions, for
microactuations. The comparative study shows that electrolysis improves actuation power
efficiency by 2-3 orders of magnitude while exhibiting better controllability, bio-
compatibility and miniaturization potential, compared with traditional boiling actuation in
a similar setup. By combining a virtual check valve with bubble capturing and venting, a
new paradigm of micropumps is developed. Fluid circulation in a closed-loop is achieved
by using electrolytic gas bubbles (H2 and O2) with 10-100 times higher power efficiency
over traditional thermal-bubble-driven micropumps. The flexibility of the bubble source
provides an opportunity to optimize the micropumping mechanism for a particular
application and address its specific concerns.
Active manipulation of microscopic gas bubble is promising, considering the recent
progresses in the electrical control of surface tension by electrowetting-on-dielectric
(EWOD). The reliable manipulation of microscopic gas bubbles is expected to contribute
greatly to the research on micro total analysis systems (μTAS) and micro power
generators (e.g. micro direct methanol fuel cell or μDMFC), and facilitate their
contribution to the field of MEMS and nanotechnologies.
1
CHAPTER 1
INTRODUCTION
1.1 Microscopic Gas Bubbles as Powerful Tools
Although gas bubbles are notoriously inconsistent and fragile in macroworld,
scaling law makes microscopic gas bubbles much more stable in microscale. Their
unique properties can therefore be employed to provide powerful tools for various
applications.
The Laplace-Young equation [1] reveals a property of gas bubbles in a liquid
environment: the pressure difference across the liquid/gas interface of a bubble is
inversely proportional to its radius:
RP vl /2γ=Δ (1-1)
where ΔP stands for the pressure difference; γvl stands for surface free energy of the
liquid/gas interface and R stands for the radius of the bubble.
Therefore, the environmental pressure change can be measured by observing
the size variation of gas bubbles [2], as Figure 1-1 shows. When a large amount of
microscopic gas bubbles are introduced in a flow, simultaneous pressure and velocity
measurement over the entire flow domain can be provided by imaging [3].
2
Figure 1-1: Bubbles under various pressures [3].
Another insight provided by the Laplace-Young equation is that a smaller gas
bubble (with a radius of R) generates an even larger pressure difference ΔP. This is
known as the scaling effect of surface tension, due to which surface tension has been
exploited as an actuation method since the very beginning of MEMS technology. An
early example is the printhead of inkjet printers [4], which has been commercialized
successfully. Attracted by this relatively huge force in microscale, researchers came up
with a broad range of ideas to make use of surface tension for microvalve [5],
micropump [6], microassembly [7] and even the creation and manipulation of
individual liquid droplets [8]. The electrochemical bubble-driven micropump by
hydrophobic venting [9] in this dissertation provides a latest example of actuation by
microscopic gas bubbles.
Microscopic gas bubbles can also be employed in optical applications, owing
to the refractive index deference between the gas bubble and its surrounding. By
trapping gas bubbles in a glass capillary filled with glue, an X-ray microlens of
hundreds of micron in diameter has been demonstrated [10]. Several lenses can be
formed and solidified in the same capillary, as Figure 1-2 shows. Similar ideas can be
found in Philips fluid lenses [11, 12] or Varioptic tunable lenses [13, 14], although oil
droplets instead of gas bubbles are used in these electrowetting-controlled lenses.
0.33 bar 1.00 bar 1.34 bar1mm
3
Figure 1-2: A microcapillary lens for X-rays [10].
Laser-polarized noble gas bubbles with smaller size (tens of microns) can be
used in MRI (Magnetic Resonance Imaging) to provide a strong signal source. Since
these gases are innocuous, high-resolution in vivo MRI images of human/animal
tissues like Figure 1-3 can be obtained [15]. Blood flow velocity and tissue perfusion
can also be measured quantitatively.
Figure 1-3: In vivo non–slice-selective coronal images of rat vasculature acquired
using 3He microbubbles suspended in Hexabrix [15].
A more complex and exciting phenomenon is sonoluminescence [16], which
was first found in the 1940s and is still not fully understood even today. It is observed
that ultrasonic sound at a certain frequency can agitate a micron-sized bubble
suspended in water, expand it to ~50 micron and then allow it to implode rapidly into
450μm
glue air
4
a submicron bubble. Since the bubble is enormously compressed (~10-6 in volume),
the gas inside reaches extreme temperature and pressure. Light (mostly ultraviolet) is
emitted. The temperature during light emission is believed to be 10,000~20,000K [17],
even hotter than the surface of the sun. Some researchers claimed that a fusion
reaction is taking place inside this violently collapsing bubble [18, 19]. Even though
the so-called “tabletop sonofusion” is still being debated fiercely, the cavitation-
actived bubble definitely provided a microreactor with extreme condition inside,
without the involvement of expensive high-energy equipments. Both fundamental
sciences and engineering applications can be expected with further investigation of
these bubbles.
1.2 Scaling Effect of Surface Tension
Considering the great potential of microscopic gas bubbles, little has been done
to turn them into reliable devices. Since macroscopic bubbles have gained the name of
inconstancy and frangibility, it is hard to draw an analogy between their microscopic
counterparts and a solid rotator, a glass lens or a beaker, which can be controlled
definitely and repeatedly in the macro-world. However, the scaling effect makes the
gas bubbles an even more feasible solution for micro sensors and actuators than
miniaturized replicas of macro-devices.
The marvelous characteristics of microscopic gas bubbles, which cannot be
found in their macroscopic counterpart, provide a good example of the scaling effect:
objects with the same material but different size behave differently. These phenomena
can be explained by examining the relationships between the scale of an object and its
properties, or the scaling effect of physical variables. Trimmer and Stroud [20]
5
provided several examples, such as the scaling effect of force, energy, acceleration and
transit time.
As far as the manipulation of an object is concerned, the dominant force on this
object should be investigated, because there are usually several forces acting on an
object simultaneously. For example, both gravity and capillary force act on the water
column inside a glass tube. If two large glass tubes (with diameters larger than several
millimeters) are connected at the bottom to form a u-shaped tube, the gravitational
hydraulic pressure will dominate over surface tension. The meniscuses in the two
tubes will be kept at the same height, even if the size of one tube is substantially larger
than the other. However, surface tension (capillary pressure) will no longer be
neglected if the inner diameter of one tube shrinks to the microscale (e.g. 100
micrometer or less). This is the mechanism that trees use to “pump” water up to tens
of meters. A force can play different roles according to the scale of the object it works
on.
Most forces decrease if the size of its relevant object shrinks. This effect can be
described by a force scaling equation: nLF ∝ (1-2)
where L stands for the characteristic length of the object in question.
Equation 1-2 is not an accurate measure of any specific force, but it can
evaluate the relative weight of different forces according to the characteristic length
and reveal the scaling effect of forces. The scaling effect of several ordinary forces is
summarized as following:
Surface tension: n = 1
Electrostatic, pneumatic, muscular: n = 2
Magnetic: n = 2~4
6
Gravitational: n = 3~4
It is necessary to notice that the value of n depends on the assumptions in some
cases. If the gravity of an object (gravitational force between this object and the Earth)
is considered, n equals to 3. Gravitational force between two astronomical objects, on
the other hand, scales with an index of 4, if the density remains constant. The scaling
of magnetic forces depends upon how the current density scales [20].
Smaller n indicates a slower decrease with shrinking size, thus the forces with
smaller n will emerge as dominant microscopic forces. As the only common force
with first order scaling, surface tension surely dominates over most other forces in the
sub-millimeter scale [21, 22].
The dominance of surface tension keeps the integrity and stability of
microscopic gas bubbles and makes them behave consistently and work reliably. More
importantly, the most powerful way to manipulate these bubbles is obviously the
passive or active control of surface tension.
1.3 Control of Surface Tension
Several methods have been developed to control surface tension. The most
straightforward way is to decrease liquid-gas surface tension by adding surfactant into
the liquid. The apparently irreversible process can be actively controlled by changing
the property of the surfactant electrochemically [23] or optically [24].
Thermocapillary pumping [TCP] has also been reported to actuate discrete bubble or
droplet in microchannels [6, 25].
Since surfactants and overheating are rarely acceptable for biomedical liquid
samples, control of solid-liquid surface tension has been studied to provide more
7
generic approaches. This can be realized by a self-assembled monolayer (SAM)
coating, which can be reversibly changed by electrical potential [26] or light [27].
A more energetically efficient and flexible way to control surface tension is
electrowetting, with electrowetting-on-dielectric (EWOD) [8] leading the way. By
applying an electrical potential through a dielectric layer, the solid-liquid surface
tension can be changed according to Lippmann’s equation [1]: 2
00, )(2
)( VVcV SLSL −−= γγ , (1-3)
where c (F/m2) is the specific capacitance (capacity per unit area) of the dielectric
layer; V is the electric potential applied across the interface. The contact angle can be
changed accordingly, as shown in Figure 1-4 and described quantitatively as 2
0 2cos)(cos VcV
LGγθθ =− , (1-4)
8
(a) Testing setup. Without applying voltage, contact angle is θ0.
(b) When applying voltage, contact angle becomes θ (V).
Figure 1-4 Fundamental testing of EWOD principle on a sessile droplet [8].
By using EWOD, basic manipulation of droplets, such as creating,
transportation, cutting and merging can be realized to implement reconfigurable
microfluidic circuits, as Figure 1-5 demonstrates.
V
θ (V)
Dielectric layer V
θ0
Probe needle or wire
Conductive substrate
t
9
(a) Creating (b) Transportation (moving)
(c) Cutting (d) Merging
Figure 1-5 Basic microfluidic functions performed by EWOD [8].
Although not included in this dissertation, active and programmable
manipulation of microscopic gas bubble is promising by using the above approaches,
especially EWOD.
1.4 Overview of This Dissertation
In this dissertation, three functions will be reported to manipulate gas bubbles
in microreactors by surface tension.
Chapter 2 describes the theory and experiments to capture bubbles onto a
patterned surface in a liquid environment. Guided by surface free energy, bubbles can
automatically attach to the energetically favorable locations (bubble-traps) and align
into a prescribed array pattern. Bubble capturing potential Φbc is proposed as the
parameter to quantitatively evaluate the surface’s “affinity” with gas bubbles. A
bubble-trap can therefore be viewed as an area with locally maximum Φbc. Simulation
10
and experiments are carried out to verify the concept. The bubble traps analyzed in
this chapter can be used to collect gas bubbles inside a distributed breather (chapter 3)
and improve the performance of a bubble-driven micropump (chapter 4).
Chapter 3 introduces a universal gas bubble removal approach, termed
hydrophobic venting, to reduce or eliminate gas bubbles from microfluidic devices.
Micro/nano-meter sized hydrophobic venting holes are used to hold liquid, while
allowing the gas to pass through freely. Distributed bubble-capturing breathers are
shown to be able to collect gas bubbles from microchambers and vent them out. Gas
permeable microchannels are demonstrated to remove gas bubbles from gas/liquid
two-phase flow and avoid bubble clogging of microchannels. The leakage prevention
ability makes the distributed breathers and gas permeable microchannels ideal for
portable microfluidic devices.
Based on the bubble capturing and venting technology, an electrochemical
bubble pump is developed in chapter 4, which combines directional bubble growth and
symmetric bubble removal to obtain a net liquid flow. The ability to remove insoluble
gas bubbles such as O2, H2, N2 and CO2 has enabled new bubble sources for
micropumps, such as electrolysis of water, injection, chemical reaction and ultrasonic
cavitation. Using electrolysis as an example, power efficiency of the bubble pump is
demonstrated to be improved 10~100 times over that of the traditional thermal-bubble-
driven pumps. A comparative study suggests the potential for even greater power
saving. Better controllability is also achieved by replacing boiling with electrolysis as
the actuation method.
Micro direct methanol fuel cell (μDMFC) is used as an example of
microreactors in this dissertation. For the venting mechanism, the methanol
component in fuel presents a challenge for leakage prevention. The 35psi or higher
11
leakage onset pressure for 10M methanol fuel makes hydrophobic venting feasible for
the next generation μDMFC with concentrated fuel. The new two-phase flow
management schemes have enabled innovative designs of μDMFC.
References
[1] A. W. Adamson, "Physical Chemistry of Surfaces", 5th Ed.: New York: John
Wiley & Sons, Inc, 1999.
[2] B. Ran and J. Katz, "The Response of Microscopic Bubbles to Sudden Changes in
the Ambient Pressure," Journal of Fluid Mechanics, 1991, vol. 224, pp. 91-115.
[3] A. Akonur and A. K. Prasad, "Methodology and Implementation of Bubble-Image
Barometry," Measurement Science and Technology, 2000, vol. 11, pp. 398-405.
[4] F. Tseng, C.-J. Kim, and C.-M. Ho, "A High Resolution High Frequency
Monolithic Top-Shooting Microinjector Free of Satellite Drops: Part I. Concept,
Design and Model," Journal of Microelectromechanical Systems, 2002, vol. 11, pp.
427-36.
[5] A. P. Papavasiliou, A. P. Pisano, and D. Liepmann, "High-Speed and Bi-Stable
Electrolysis-Bubble Actuated Gate Valves," Tech. Dig. The 11th Int. Conf. on
Solid-State Sensors, Actuators and Microsystems, Munich, Germany, 2001, pp.
940-3.
[6] T. K. Jun and C.-J. Kim, "Valveless Pumping Using Traversing Vapor Bubbles in
Microchannels," Journal of Applied Physics, 1998, vol. 83, pp. 5658-64.
12
[7] X. R. Xiong, Y. Hanein, W. H. Wang, D. T. Schwartz, and K. Bohringer, "Multi-
Batch Micro-Self-Assembly via Controlled Capillary Forces," 2001 IEEE/RSJ
International Conference on Intelligent Robots and Systems, 2001, pp. 1335-42.
[8] S.-K. Fan, "Digital Microfluidics by Cross-Reference EWOD Actuation: Principle,
Device and System," Ph. D. Thesis, University of California at Los Angeles, 2003
[9] D.-S. Meng and C.-J. Kim, "Micropumping by Directional Growth and
Hydrophobic Venting of Bubbles," Proc. The 18th IEEE Int. Conf. on Micro
Electro Mechanical Systems, Miami, Florida, 2005, pp. 423-6.
[10] Y. I. Dudchik and N. N. Kolchevsky, "A Microcapillary Lens for X-rays,"
Nuclear Instruments and Methods in Physics Research (A), 1999, vol. 421, pp.
361-4.
[11] S. Kuiper and B. H. W. Hendriks, "Variable-Focus Liquid Lens for Miniature
Cameras," Applied Physics Letters, 2004, vol. 85, pp. 1128-30.
[12] "Philips' fluid lenses bring things into focus,"
http://www.research.philips.com/newscenter/archive/2004/fluidlenses.html.
[13] C. Gabay, B. Berge, G. Dobillaire, and S. Bucourt, "Dynamic Study of a
Varioptic Variable Focal Lens," SPIE Vol. 4767(2002), Seattle, July 8-9, 2002, pp.
159-65.
[14] "Technology presentation of Varioptic tunable lenses,"
http://www.varioptic.com/v3/technology.php.
[15] M. S. Chawla, X. J. Chen, H. E. Moller, G. P. Cofer, C. T. Wheeler, L. W.
Hedlund, and G. A. Johnson, "In Vivo Magnetic Resonance Vascular Imaging
13
Ising Laser-Polarized 3He Microbubbles," Proc. Natl. Acad. Sci., 1998, vol. 95, pp.
10832-5.
[16] S. J. Putterman, "Sonoluminescence: Sound into Light," Scientific American,
1995, pp. 46-51.
[17] H. Metcalf, "Sonoluminescence: that Flashing Sound," Science, 1998, vol. 279,
pp. 1322-3.
[18] R. P. Taleyarkhan, C. D. West, J. S. Cho, R. T. Lahey, Jr., R. I. Nigmatulin, and
R. C. Block, "Evidence for Nuclear Emissions During Acoustic Cavitation,"
Science, 2002, vol. 295, pp. 1868-73.
[19] R. P. Taleyarkhan, J. S. Cho, C. D. West, R. T. Lahey, Jr., R. I. Nigmatulin, and
R. C. Block, "Additional Evidence of Nuclear Emissions during Acoustic
Cavitation," Physical Review E, 2004, vol. 69, pp. 036109-11.
[20] W. Trimmer and R. H. Stroud, "Chapter 2: Scaling of Micromechanical Devices,"
in the MEMS handbook, M. Gad-el-Hak, Ed.: RCR Press, 2002, pp. 2.1~2.9.
[21] C.-J. Kim, "MEMS Devices Based on the Use of Surface Tension," Proc. Int.
Semiconductor Device Research Symp., Charlottesville, VA, Dec., 1999, pp. 481-4.
[22] R. S. Fearing, "Survey of Sticking Effects for Micro-Parts," IEEE Int. Conf.
Robotics and Intelligent Systems (IROS '95), Pittsburgh, PA, August 7-9, 1995.
[23] B. S. Gallardo, V. K. Gupta, F. D. Eagerton, L. I. Jong, V. S. Craig, R. R. Shah,
and N. L. Abbott, "Electrochemical Principles for Active Control of Liquids on
Submillimeter Scales," Science, 1999, vol. 283, pp. 57-60.
14
[24] J. Y. Shih and N. L. Abbott, "Using Light to Control Dynamic Surface Tensions
of Aqueous Solutions of Water Soluble Surfactants," Langmuir, 1999, vol. 15, pp.
4404-10.
[25] T. A. Sammarco and M. A. Burns, "Thermocapillary Pumping of Discrete Drops
in Microfabricated Analysis Devices," AIChE Journal, 1999, vol. 45, pp. 350-66.
[26] J. Lahann, S. Mitragotri, T.-N. Tran, H. Kaido, J. Sundaram, I. S. Choi, S. Hoffer,
G. A. Somorjai, and R. Langer, "A Reversibly Switching Surface," Science, 2003,
vol. 299, pp. 371-4.
[27] K. Ichimura, S.-K. Oh, and M. Nakagawa, "Light-Driven Motion of Liquids on a
Photoresponsive Surface," Science, 2000, vol. 288, pp. 1624-6.
15
CHAPTER 2
BUBBLE-TRAPS AND BUBBLE CAPTURING POTENTIAL
2.1 Background
The marvelous characters of microscopic gas bubbles make them potentially
great tools for both scientific research and engineering applications. However, reliable
manipulation of these bubbles is necessary so that their unique properties can be
deployed to perform specific tasks. Trapping bubbles onto pre-determined locations
(e.g. arrays) is one of the basic manipulations. Immobilization also provides a basic
function for other more complex manipulations, such as bubble-driven micropumps.
It had been realized that surface tension can guide gas bubbles to form regular
arrays in polymer solution [1]. Hydrophobic microwells on a solid surface were also
reported to be able to capture bubbles [2]. But quantitative analysis of the bubble
capturing structure (i.e. bubble-traps) is still absent. This study proposes bubble
capturing potential (Φbc) as the quantity to evaluate a bubble-trap’s ability to capture a
gas bubble onto it from a liquid environment, so as to provide a design guideline for
relevant microfluidic devices.
2.2 Bubble-Traps: Definition and Qualitative Explanation The bubble capturing mechanism is based on the multiphase fluidic system’s
tendency to minimize its total surface energy. If the total energy can be minimized
when the bubble attaches onto a specific pattern on a heterogeneous surface, this
pattern is called a bubble-trap. When a bubble gets a chance to move around, it will
16
tend to stay on a bubble-trap so that the total system energy is minimized. The energy
here is surface free energy, whose three components can be found in Young’s equation
of contact angle [3], illustrated in Figure 2-1:
slsvvl γγθγ −=cos , (2-1)
where γvl, γsv and γsl are the surface free energy of liquid-vapor, solid-vapor and solid-
liquid interfaces respectively, and θ is contact angle.
Figure 2-1: Contact angle of a gas bubble’s three-phase interface.
The total surface energy of the system can then be defined as:
slslsvsvvlvl AAAAE γγγγ +⋅+⋅=⋅= ∑ , (2-2)
where Avl, Asv and Asl are the surface areas of liquid-vapor, solid-vapor and solid-liquid
interfaces, respectively. Bubble-trap is a location where total surface energy is
minimized when a bubble is attached onto it. Figure 2-2 is a schematic drawing of the
two kinds of proposed bubble-traps.
liquid
gas bubble
γvl
γsv γsl
θ
17
Figure 2-2: Bubbles’ status in a liquid-filled container
a: floating; b,c and d: attached.
Type I bubble-trap is a hydrophobic pattern on a flat hydrophilic surface. To
simplify the argument but still understand bubble capturing qualitatively, we assume
the liquid-vapor surface is kept constant whether a floating bubble (bubble “a” in
Figure 2-2) attaches to a hydrophilic flat surface (bubble “b”) or hydrophobic flat
surface (bubble “c”). Then the bubble attachment process can be viewed as
substituting a solid-liquid interface with area ΔA by a solid-vapor interface with the
same area, resulting in an increase of system energy by ΔΕ = ΔA(γsv -γsl) . When θ is
smaller than 90o (hydrophilic), or γsv > γsl, ΔΕ >0, i.e. the formation of a solid-vapor
interface increases the system energy. This is not favorable, so the bubbles tend to
detach from the hydrophilic surface. Vice versa, when θ is larger than 90o
(hydrophobic), or γsv < γsl, ΔΕ <0, i.e. the formation of a solid-vapor interface
decreases the system energy. This is favorable, so the bubbles tend to attach to the
hydrophobic surface. This somewhat oversimplified model explains why a
hydrophobic spot on a hydrophilic surface can serve as a bubble-trap to capture the
bubbles from a liquid environment and hold them.
A hydrophobic concave pit on a hydrophilic surface can serve as an even better
bubble-trap (type II). Two factors contribute to this geometrically enhanced bubble-
liquid
hydrophilic hydrophobic
a
b cdbubble-trap
type I
bubble-trap type II
18
trap. First, a larger interface (ΔA) is “exchanged” during bubble capturing (bubble d of
Figure 2-2), compared with a flat surface. This larger area ΔA entails a larger energy
reduction in ΔΕ = ΔA(γsv -γsl) and promotes the capturing. Second, the liquid-vapor
interface of a bubble can also be reduced significantly in the attaching process, and
thus is the total energy.
2.3 Bubble Capturing Potential: Φbc The qualitative explanation above can help understand the bubble capturing
phenomena. But it is not accurate, because the bubbles will deform during this attach-
detach process, so the liquid-vapor surface area is also changing. Moreover, the total
energy in equation 2-2 depends on the bubble size and absolute value of liquid surface
tension, so it cannot be used as an indicator for the bubble-trap’s bubble capturing
ability.
In order to eliminate the influence of bubble size and liquid properties, the
scale to evaluate the surface’s “affinity” for gas bubbles is defined as bubble capturing
potential:
( ) vlbc LEEΦ γ⋅−−= 20 / , (2-3)
where E0 is the total surface energy of a three-phase system with a floating bubble (i.e.
bubble a in Figure 2-2); E is the total surface energy of a three-phase system with an
attached bubble (bubble b, c or d in Figure 2-2); L = V1/3 is defined as the characteristic
length of the bubble; and γvl stands for the surface free energy on vapor-liquid
interface.
According to the definition, a surface with positive Φbc means system energy
will decrease during bubble attaching, which is energetically favorable. So a bubble
tends to be captured onto a surface with positive Φbc. An area with larger Φbc
represents a stronger tendency to retain bubbles on it. A bubble-trap can therefore be
viewed as an area with locally maximum (and positive) Φbc.
19
Here, we assume that the patterns or pits are big enough to accommodate the
bubbles. Or conversely, the bubble is small enough to be accommodated in a single
bubble-trap. In this case, we will show that Φbc only depends on surface topology and
contact angle in the following deduction.
The volume of the floating bubble (bubble a in figure 2-2) is 3
34 RV ⋅= π . So
the characteristic length of this bubble is RVL ⋅== 33
34 π . Therefore the surface
area (vapor-liquid interface) is: 2320 364 LRAvl ⋅=⋅= ππ .
Therefore, the introduction of a floating bubble into the three-phase system
leads to a total surface energy increase of:
vlvlvl LAE γπγ ⋅⋅=⋅= 2300 36 (2-4)
However, the introduction of an attached bubble into the three-phase system
causes two area changes. Firstly, a surface area is dried, or a solid-liquid interface is
replaced by a solid-vapor interface. Secondly, the area of vapor-liquid interface is
changed. The total surface energy of an attached bubble is hence:
sldrysvdryvlvl AAAE γγγ ⋅−⋅+⋅= (2-5)
Since θγγγ cos⋅=− vlslsv (equation 1), equation 5 becomes:
vldryvl AAE γθ ⋅⋅+= )cos( (2-6)
Substitute (4) and (6) into equation 3, the bubble capturing potential:
( )
θπ
θπθ
cos36
cos36
3
223
dryvl
dryvlbc
AA
LA
LA
−−=
−−=Φ, (2-7)
where Avl and Adry stand for vapor/liquid and vapor/solid interface area of the attached
bubble respectively. They are typically proportional to L2, if the particular pattern can
accommodate the bubble completely (i.e. bubble is small enough). The corresponding
normalized area Avl and Adry are functions of surface topology and contact angle only.
20
Consequently, Φbc is independent of the absolute value of surface tension and bubble
size. Elimination of these two variables makes Φbc valid to evaluate the bubble-traps’
bubble capturing ability, if only the surface topology and contact angle are known. In
other words, Φbc is a mere property of the surface for a given liquid environment.
If a big bubble attaches onto an area covering two or more patterns, the bubble
capturing potential will be the weighted average Φbc of all the patterns that it touches.
However, Φbc is more meaningful for microscopic gas bubbles, because surface
tension of large bubble is less significant and can be neglected in a lot of circumstance.
2.4 Quantitative Analysis of Bubble-Traps Equation 2-7 provided the foundation to calculate bubble capturing potential of
any surface structure, including bubble-traps, a surface area with locally maximum
(and positive) Φbc. Bubble capturing potential of flat surface and the concave conic pit
will be analyzed to evaluate the two kinds of bubble-traps mentioned before.
On a flat surface as Figure 2-3 represents, the height of a attached bubble is:
Rh ⋅+= )cos1( θ .
Figure 2-3: A gas bubble on a flat surface.
So the relationship between characteristic length L and radius R of this
attached bubble can be obtained by means of its volume:
liquid
θ θ
hR
21
32
23
)cos2()cos1(3
)3(3
R
hRhVL
⋅−⋅+=
−⋅⋅==
θθπ
π
(2-8)
Then the vapor-liquid interface area Avl and vapor-solid interface area Adry can
be expressed by characteristic length L. 2)(2 LARhA vlvl ⋅== θπ (2-9)
22 )()sin(2 LARA drydry ⋅== θθπ (2-10)
Both Avl(θ)and Adry(θ) here are mere functions of contact angle θ , which
means that Φbc on flat surface is a mere function of contact angle.
Similar calculations can be applied to concave conic pits, as Figure 2-4
illustrates.
Figure 2-4: A gas bubble in a concave conic pit.
The angle between the horizontal level and the normal radius (r) of three-phase
line is:
αθβ −= , (2-11)
2α
θ
h
d
R
90ο−α
90ο− θ rβ
22
where α stands for the conic angle of the pit and θ stands for the contact angle.
Knowing this angle, the volume of upper spherical cap [4] can be calculated as:
)3(3
2 hRhVs −⋅⋅=π , (2-12)
where, Rh ⋅−= )sin1( β is the height of the cap. What need to be noticed is that this
volume Vs should be considered negative when β > 90o and the spherical cap is
concave (refer to Figure 2-5). We can rewrite equation 12 as:
⋅−⋅⋅= )3(3
2 hRhVsπ sign )(cosβ , (2-13)
The sign function here is defined as:
, (2-14)
The volume of the partial conic pit [5] is:
drVc ⋅⋅= 2
3π , (2-15)
where the radius of the cone bottom is βcos⋅= Rr and the depth of the conic pit is
αtan/rd = .
Then we can get the relationship between R and L as: 33 ),( RVVVVL cs ⋅=+== αθ . (2-16)
The dimensionless parameter V(θ, α) here is determined by equations 2-13~2-16. It is
a function of contact angle θ and conic angle α. Therefore, once θ andα are fixed, R
is proportional to L. Other critical dimensions (e.g., h, r and d) are thereby
proportional to L too.
Then the vapor-liquid interface area Avl [4] and vapor-solid interface area Adry
[5] can be expressed by characteristic length L. 2),(2 LARhA vlvl ⋅== αθπ (2-17)
222 ),(2 LArdrA drydry ⋅=+⋅= αθπ (2-18)
sign(x) = 1, if x > 00, if x = 0-1, if x < 0
23
Again, both Avl(θ, α)and Adry(θ, α) here are mere functions of contact angle θ
and conic angle α, which means that Φbc of concave conic pits is a function of contact
angle and conic angle.
The origin of the sign function is explained as following. Bubble shape
changes according to variable contact angle, illustrated in Figure 2-5. Supposing the
conic angle α is fixed and contact angle θ changes from 0o to 180o, the sign of volume
and surface area can be analyzed. When θ < α + 90o (i.e., β = θ − α ranges from
−α to 90o), the bubble bulges to form a convex liquid/gas interface. During the
increase of contact angle, the spherical cap keeps flattening from larger half to
hemisphere (when β =0) and then smaller half, although the volume of cap (Vs) is
positive. However, at the point that θ equals to α + 90o (i.e., β = θ − α = 90o), a flat
liquid/gas interface is assumed. The volume of spherical cap (Vs) becomes 0 in this
case. If the surface becomes more hydrophobic (i.e., β = θ− α ranges from 90o to
180o−α), the liquid/gas interface turns concave, Vs should be subtracted from the total
volume (negative). Therefore the sign of Vs is coincidently same as the sign of cosβ.
This coincidence happened because h should be considered to share the same sign
with sin(90o−β). The other volume and surface area values (Vc, Avl and Adry ) are all
kept positive for all physically possible contact angle.
24
Figure 2-5: Bubble shape at different contact angles
2α
θ
d
Rβ < 0
2αd
R90ο− θ
β
hθ
2α
θh
dR
90ο−α
90ο− θ β
contact angle: θ0o α α + 90ο 180o
β = θ − α :
− α 0ο 90ο 180o−α
Vs:+ + −
Vc, Avl, Adry + + +
h
25
2.5 Simulation Results Matlab® simulation results of flat surface and conic pit are shown in Figure 2-6,
assuming a contact angle ranging from 0o to 180o. The conic angle is set to 40o to
represent KOH-etched silicon pits, which have a maximum tilt angle of 45o and a
minimum tilt angle of 35o. KOH-etching is a convenient way to get relatively
hydrophobic concave pits. KOH-etched pits are also used in our experimental
verification and application example (i.e., bubble-capturing breathers), to be explained
in this study.
Two interesting aspects can be found in these curves: Firstly, a surface doesn’t
need to be strictly hydrophobic in order to capture bubbles. For both the flat surface
and conic pit, Φbc turns positive at contact angle value as small as ~20o. Accordingly,
given a contact angle larger than 20o, a bubble would rather attach onto a hydrophilic
surface than float in the liquid. This can be confirmed by the observation that bubbles
can form on the wall of a water-filled glass beaker (contact angle around 20o) and stay
there, when the beaker is heated. However Φbc increases dramatically with increasing
contact angle after it is larger than ~80o. Secondly, the simulation results suggest a
substantially higher Φbc for the concave structures, compared to a flat surface with the
same contact angle for most of the range of θ (i.e. θ > 20o). This predicts a stronger
bubble attraction of the Type II bubble-traps. For example, Φbc of a conic pit at 80o
contact angle is 1.91 - almost 3 times of Φbc for the flat surfacewith the same contact
angle (0.69).
26
Figure 2-6: Simulation result of Φbc on flat or concave surfaces with different contact
angles.
2.6 Experimental verifications In order to verify the bubble capturing concept experimentally, a hydrophilic
sample is prepared by thermally growing silicon dioxide on a bare silicon wafer
(contact angle: θ ~20 o). HMDS is vapor-coated and patterned by lift-off process to
provide relatively hydrophobic spots (contact angle: θ ~80o) in a square-grid pattern.
The sample is then immersed in 5% H2SO4 aqueous solution. When hydrogen and
oxygen gas bubbles are generated by electrolysis and brought to the sample surface by
buoyancy, the gas bubbles preferentially attached to the hydrophobic spots, as Figure
2-7 demonstrates. Around 60% of the flat hydrophobic patterns successfully captured
gas bubbles onto them.
Φbc
: Bub
ble
capt
urin
g po
tent
ial
contact angle o
KOH pit flat surface
0 20 40 60 80 100 120 140 160 180
7
6
5
4
3
2
1
0
-1
27
Figure 2-7: Gas bubbles captured on an array of type I bubble-traps.
Type II bubble-traps are implemented by paramedic pits etched into a (100)
silicon wafer by 30% KOH with SiO2 as a mask. The bare silicon, with a contact angle
of ~ 80o, serves as type II bubble-traps (hydrophobic concave pits) on a hydrophilic
SiO2 surface. Under similar experimental conditions, the KOH-etched pits provide
better bubble capturing performance than the HMDS flat pits did, as shown in Figure
2-8. Around 90% of the KOH-etched pits in this experiment successfully captured gas
bubbles onto them.
Figure 2-8: Gas bubbles captured on an array of type II bubble-traps.
28
The experiment confirmed that both kinds of bubble-traps can be used to
capture bubbles and form bubble arrays. Better bubble capturing performance is
demonstrated by Type II bubble-traps. The result agrees with the simulation and
supports the proposition to use Φbc as an indicator of the surface’s ability to capture
bubbles.
2.7 Summary and Future Directions
Bubble capturing by using surface tension is described in this chapter. Φbc is
proposed as the quantitative parameter to evaluate a bubble-trap’s tendency to capture
bubbles. The merit of this definition is that Φbc can be expressed in term of two
measurable and controllable variables: surface topology and contact angle θ.
Therefore Φbc is independent of the absolute value of surface tension and bubble size,
and can be considered as a property of the surface. Simulation suggests a distinct
performance enhancement for type II bubble-traps (hydrophobic concave pits) over
type I bubble-traps (flat hydrophobic patterns), which is confirmed by experiments.
It is concluded in this study that both surface property and geometry can play
important roles in bubble immobilization. A complete survey on all kinds of surface
structures can lead to the design of more effective bubble-traps. The concept of Φbc
can also be applied to evaluate other possible bubble-traps than the two mentioned in
this study.
The passive immobilization of gas bubble discussed here can form bubble
arrays without any energy input (as Figure 2-7 and 2-8 show). Active control of
surface tension [6] or topology [7] will add more functions to this manipulation, which
can find more applications in the future. Again, bubble capturing potential or similar
concepts will provide a guide line for the design of such kind of devices.
29
References
[1] M. Srinivasarao, D. Collings, A. Philips, and S. Patel, "Three-Dimensionally
Ordered Array of Air Bubbles in a Polymer Film," Science, 2001, vol. 292, pp. 79-
83.
[2] E. Ostuni, C. S. Chen, D. E. Ingber, and G. M. Whitesides, "Selective Deposition
of Proteins and Cells in Arrays of Microwells," Langmuir, 2001, pp. 2828-34.
[3] A. W. Adamson, "Physical Chemistry of Surfaces", 5th Ed.: New York: John
Wiley & Sons, Inc, 1999.
[4] "Geometry of spherical cap," http://mathworld.wolfram.com/SphericalCap.html.
[5] "Geometry of cone," http://mathworld.wolfram.com/Cone.html.
[6] S. K. Cho, H. Moon, and C.-J. Kim, "Creating, Transporting, Cutting, and Merging
Liquid Droplets by Electrowetting-Based Actuation for Digital Microfluidic
Circuits," Journal of Microelectromechanical Systems, 2003, vol. 12, pp. 70-80.
[7] B. He and J. Lee, "Dynamic Wettability Switching by Surface Roughness Effect,"
Proc. The 16th IEEE Int. Conf. on Micro Electro Mechanical Systems, Kyoto,
Japan, Jan. 19-23, 2003, pp. 120-3.
30
CHAPTER 3
HYDROPHOBIC VENTING
3.1 Introduction
3.1.1 Bubble Clogging of Microchannels
Fluid flow in Microchannels [1] is essential for modern microfluidic devices,
such as micro Total Analysis System (μTAS) or micro Direct Methanol Fuel Cell
(μDMFC). As the Reynold’s number decreases [2, 3] and channel size shrinks below
capillary length (~1mm for most aqueous solutions) [4], many phenomena are found
different from the flow in macrochannels. Bubble clogging problem in microchannels
is one example, which was realized nearly a decade ago [5, 6]. It is observed that the
presence of gas bubbles will increase the flow resistance of microchannel by
introducing an additional counter-flow pressure (i.e. clogging pressure) [7, 8], or even
completely block the flow in some cases. In spite of the simple appearance of this
problem, many different physical effects are involved [9], some of which have not
been understood well and cannot be predicted precisely. The clogging can be
attributed to friction [10], contracting geometry of the channel [11] and contact angle
hysteresis. In reality, the problem can be complicated by noncircular channel shape,
compressibility of gas bubble, surface uncertainty of channel wall and kinetics [12].
The bubble clogging problem can therefore disturb the measurement of flow in an
unpredictable manner and bring additional resistance to the actuation (pumping) of
fluid in microchannel. A significant burden is thus put on the micropumps, which is
31
usually powered by a limited power source. Attaching onto the surface of the
microchannel, gas bubbles can isolate the reactant from the catalyst, electrodes or
sensing components so as to ruin the highly regarded high area-volume-ratio of
microreactors. In a sealed microfluidic device, gas bubbles generation can accumulate
the pressure and damage the device.
Unfortunately, the existence of gas bubbles in microchannels is prevalent or
even inevitable for most microfluidic devices. Priming is a major source of the
accidentally introduced gas bubbles. Figure 3-1 illustrates a bubble formed when two
liquid samples are introduced in to the confluence of two microchannels subsequently.
Figure 3-1: Bubble formation during priming of two liquid samples [13].
Although “careful” or negative-pressure (vacuum) priming were usually
described in the laboratorial practices to avoid this problem, they are not practical for
the reliable operation of a commercial product. The specific scenario shown in Figure
3-1 was proposed to be solved by synchronizing the two flows [13] with abrupt
hydrophobic necks as retarding valves. But the pressure of both flows has to be
controlled well, which is still too demanding for practical systems.
32
Fluctuation of pressure or temperature can also introduce gas bubbles with an
even more unpredictable and unnoticeable way. As is permissible in certain
applications, the sample may be boiled or treated by ultrasonic wave to decrease the
solubility of gas before priming. However, most biomedical liquid samples cannot
tolerate this kind of treatments. Gas bubble can also be introduced by electrochemical
reaction (e.g. electrolysis), acoustic cavitations or other chemical reactions, which are
not as common as the previous sources but must be handled properly to ensure reliable
operations. The bubble clogging problem is expected to be even more severe when
individual microreactors are brought together to construct a complex lab-on-a-chip,
where the chance to introduce accidental gas bubbles are higher.
Since the prevention of gas bubble formation in microchannel is not
dependable, special designs of channel shape were proposed to increase the channel’s
tolerance of bubbles. For example, a smaller channel parallel to the “major
microchannel” can provide a “bypass” for pure liquid flow [14], because gas bubbles
tend to stay in larger channel to minimize surface free energy. Flow resistance can
thus be reduced.
33
Figure 3-2: “Channel in channel” design to prevent bubble clogging [14].
Planar “bypass” as illustrated in Figure 3-3 was also studied to function
similarly, with a designed structure to trap the bubble in a certain position [9], instead
of allowing it to block the flow completely in a contracting part of the microchannel.
Figure 3-3: Microchannel with a bubble-trapping region [9].
Even though those designs can relieve the bubble clogging problem more or
less, they cannot remove gas bubble completely from the microchannel. The
34
uncertainty and blocking tendency are still left in the microchannel, especially those
complex ones with corners, turns or abrupt changes.
3.1.2 μDMFC: a System with Continuous Bubble Generation
Compared with accidentally introduced gas bubbles, continuously generated
gas bubbles bring much more severe bubble clogging problems to microfluidic devices.
Micro direct methanol fuel cell (μDMFC) [15, 16] is one example.
Micro fuel cells have been considered as the next generation of power sources
for potable electronics like cell phones or laptop computers. With μDMFC leading the
way, micro fuel cells feature higher energy capacity over most existing solutions (e.g.
several folds higher than lithium-based thin film batteries). The working principle of
DMFC is elucidated in Figure 1. The methanol-filled anodic chamber and air-filled
cathodal chamber are separated by membrane electrode assembly (MEA), which
consists of two electrode layers, two catalyst layers and one layer of PEM (Proton
Exchange Membrane). The PEM allows protons to transport from anode side to
cathode side and react with oxygen there. However, the electrons cannot transport
through this nonconductive membrane. Instead, they are collected by the anodic
electrode and provide continuous current for an external circuit.
35
Figure 3-4: Working principle of DMFC (Direct Methanol Fuel Cell).
The electrochemical reaction is:
CH3OH + H2O → 6e- + 6H+ + CO2 ↑ (anodic side)
1.5O2 + 6e- + 6H+ → 3H2O (cathodal side)
CH3OH +1.5O2 → 2H2O + CO2 (overall reaction)
According to this reaction, DMFCs generate CO2 gas bubble intrinsically. The
small bubbles in large-scale DMFC may not cause much trouble. They get enough
space to flow around. The pump is strong enough to push them along with the fuel
flow to a downstream external gas/liquid separator, essentially an open tank. Gas
bubbles can be easily released there. But problems occur in their microscale
counterparts, where channels may be as small as individual bubbles. Bubble clogging
problem is an obvious one, noticing that extra power consumption to overcome the
increased flow resistance will undermine the performance of a micro power source
Membrane ElectrodeAssembly (MEA)
anodechamber
cathode
e-
CO2
methanolsolution
air
e-
e- e- e- e-
e-e-
e-
waterdroplets
H+
H+
H+
e-
H+
H+
H+
H+
H+
H+ H+
H+
H+
e-
e-
e-
e-
e-
H+
anode
36
dramatically. As CO2 bubbles fill the microchannel, the fuel will be isolated from
catalyst and electrode. The reaction rate will be decreased accordingly. An even worse
possibility is that the pressure buildup inside a sealed device can aggravate fuel cross-
over of PEM and eventually damage the device.
It is clear that these problems cannot be completely solved by temporary
bubble restriction techniques provided by [9] or [14]. Reliable bubble removal is
necessary instead.
A downstream opening is the most straightforward solution for bubble removal.
One example has been shown in a thermal-bubble-driven micropump [17], where
condensation itself is not fast enough to remove all of the gas bubbles introduced by
boiling. So a “gas bubble filter” as Figure 3-5 shows is employed to filter out gas
bubbles and generate bubble-free liquid flow for downstream process. The principle is
that gas bubble prefer larger hydrophilic channel instead of small ones (filter channels)
for sake of minimum surface free energy. But the liquid can pass through these small
hydrophilic channels easily to form a bubble-free stream in the liquid outlet. The gas
bubbles can be released from the bubble outlet.
37
Figure 3-5: Gas bubble filter for nozzle-diffuser bubble pump [17].
This bubble-filter can be employed in a stationary or semi-stationary system.
However, the big opening for bubble outlet puts the system under the risk of leakage,
which is not acceptable for portable systems, such as micro fuel cells, the potential
power sources for portable electronics. A venting method, which can withstand certain
internal pressure, is desired for μDMFC, as well as any upcoming microfluidic devices
with continuous gas generation inside.
3.2 Distributed Hydrophobic Venting
In this study, we propose a universal gas removal method for portable
microfluidics, termed distributed hydrophobic venting.
38
A precursor of this technique can be found in a droplet handling system [18],
illustrated in Figure 3-6. The hydrophobic microcapillaries (Figure 3-6-A) connect
liquid channel to pneumatic channel, through which both positive and negative
pressure can be applied. Since the size of microcapillaries is very small (3μmx5μm),
they can prevent the intrusion of liquid if the pressure difference is not too high. When
proper patterns of pressure are applied through pneumatic channels, the liquid can be
positioned (Figure 3-6-B) or metered (Figure 3-6-C).
Figure 3-6: Handling of picoliter liquid samples by using HMCV (Hydrophobic
MicroCapillary Venting) [18].
A similar idea was proposed to degas liquid sample (dialysate) for portable
dialysis system [19], as Figure 3-7 shows. The dissolved gas is driven out by
ultrasonic wave to form bubbles, which can then be vented out through hydrophobic
venting channels. Gas concentration in the dialysate can therefore be controlled
without the vacuum system, which is difficult to be integrated into a portable device.
39
Figure 3-7: Micro-degassing for portable dialysis system [19].
Both of these devices use hydrophobic venting holes (i.e. hydrophobic venting
capillary or channel) to hold liquid while allowing gas to pass through freely. The
principle of this liquid holding mechanism is illustrated in Figure 3-8.
Figure 3-8: Breathing and liquid holding mechanism in a hydrophobic venting hole.
At the entrance corner of a hydrophobic capillary, the meniscus of liquid can
change its shape corresponding to the pressure difference it withstands. The varying
P0
θa
r
Pl w
40
transmeniscus pressure can therefore be balanced according to Laplace-Young
equation:
rPP l
l)180cos(2
0ασ −⋅
=− (3-1)
where Pl is the pressure inside liquid, P0 is the ambient pressure, α is the angle
between meniscus and the capillary wall, at the entrance of the hydrophobic capillary,
σl is the surface tension of fluid, r is diameter of capillary. When Pl increases, α
increases to accommodate the pressure change. However, when α exceeds the
maximum value possible for the capillary-air-liquid interface, the meniscus can no
longer hold the liquid. Leakage occurs in this case. Therefore, the maximum pressure
difference that the hydrophobic capillary can withstand (leakage onset pressure) is:
( )r
PP lleak
)180cos(2 maxmax
ασ −⋅=Δ= (3-2)
If the transmeniscus pressure (Pl -P0) is kept lower than this leakage onset
pressure, the gas can be released without liquid loss except trivial amount of
evaporation through the tiny liquid/air interface. Here we note that the maximum
contact angle αmax is the dynamic advancing angle θa, which can be very different
from the equilibrium contact angle of single-component liquid droplet on an ideal flat
surface. In other words, surface topography and fluid composition can strongly
influence the advancing contact angle, and thus the leakage pressure.
Distributed hydrophobic venting is based on the same liquid holding
mechanism as the previous works. However, the hydrophobic venting holes are
fabricated in the channel wall instead of using in-plane venting capillaries/channels in
the prior arts. The gas bubbles can therefore be removed promptly, close to where
there are generated. Bubble traps introduced in chapter 2 can be used to collect the gas
41
bubbles from the two-phase flow to the vicinity of hydrophobic venting holes. The
two functions add up to a distributed bubble-capturing breather for portable
microfluidic devices.
3.3 Silicon-Based Distributed Breather
The 1st-generation distributed breather is fabricated to prove the concept. The
venting holes are etched by deep reactive ion etching (DRIE) and coated hydrophobic
with Teflon®. Hydrophobic patterns are also formed at the vicinity of holes to capture
gas bubbles (type I bubble trap). The process flow of the breathing plate is shown in
Figure 3-9. A (100) silicon wafer is partially thinned down to about 150μm at specific
locations by KOH etching. A breathing hole (50μm in diameter) is etched through in
each thinned section by DRIE from the other side. A SiO2 layer (~ 0.1μm) is grown on
the sample surface by thermal oxidation to make it hydrophilic. The sample is then
immersed into 0.2% Teflon® solution to coat a hydrophobic layer onto its surface.
Immediately after being taken out of the Teflon® solution, the sample is blow-dried by
a strong nitrogen flow perpendicular to the surface, in order to clear the breathing
holes from being blocked. The hydrophobic layer is then patterned by oxygen RIE at
200mTorr and 200W for 5min with a 1.6μm thick AZ5214 PR mask.
42
Figure 3-9: Process flow for the first-generation venting plate.
The finished sample is then cut into 30mm×50mm chips with a diamond saw
and packaged to complete a device for venting test, as shown in Figure 3-10. The
microchamber is formed by positioning a venting plate on top, a transparent glass slide
below, and a spacer (1.5mm thick) in between. Chemical reaction is used to emulate
distributed gas bubble generation in a multiphase microfluidic system such as μDMFC.
Sodium bicarbonate (NaHCO3) solution and weak sulfuric acid (H2SO4) are injected
into the microchamber by two individual syringes sequentially. The chemical reaction
generates carbon dioxide (CO2) gas bubbles:
1. Partially thin down by KOH
50μm
2. Open breathing holes by DRIE
3. Hydrophilic treatment by oxidation
4. Hydrophobic coating and patterning
(100) Silicon wafer900μm
150μm 200μm
550μm
SiO2(~1000Å)
200μmHydrophobicpatterns
43
H2SO4 + 2NaHCO3 = Na2SO4 + 2H2O + 2CO2↑
These CO2 gas bubbles are then vented out through the vertical venting holes,
under proper conditions.
Figure 3-10: Experimental setup for distributed venting.
Through the glass base indicated in Figure 3-10, the breathing process was
recorded by a CCD camera, as shown in Figure 3-11.
~1.5mm
H2SO4 NaHCO3
CO2
pipe(flow outlet) syringe
(flow inlet1)
syringe (flow inlet2)
microchamber
venting plate spacer
epoxy
glass
camera(observe from bottom)
a) Top view
b) Cross-section View
Si
epoxy
44
Figure 3-11: Venting experiment in the first-generation distributed breather.
Chemical-reaction-induced gas bubbles grew and peaked in the microchamber
27 seconds after the two chemical solutions are introduced and mixed, corresponding
to the 6th frame of Figure 3-11. Subsequently, the bubbles started to shrink, eventually
leaving most of the surface free of bubbles, corresponding to the 10th frame (50
second). When the venting plate was substituted with a bare silicon chip (i.e. no
breathing holes) for the same experiment, the unchecked growth of bubbles covered
most of the surface and persisted there without visible change for hours.
The bubble-capturing effect was not observed as clearly as Figure 2-7 shows.
One reason is that the flow rate is intentionally kept extremely slow after the whole
chamber is filled with chemicals in order to avoid leakage. Therefore, most bubbles do
not have enough mobility to move to hydrophobic patches (e.g. the small bubbles in
the 10th frame). In addition, venting is very fast. Once bubbles are captured, they are
vented out quickly and will not show up clearly on the hydrophobic patches as in
Figure 2-7. Both reasons originate from the relatively large venting holes (50μm),
which directly leads to poor pressure and flow-rate tolerance for the liquid. We
frame 1 (00:00) original breather surface
frame 2 (00:10) introduce NaHCO3
frame 3 (00:19) introduce H2SO4
frame 4 (00:20) merging
frame 5 (00:22) CO2 Generated
frame 6 (00:27) gas bubbles peak
frame 7 (00:32)
frame 8 (00:35) bubbles shrinking
frame 9 (00:40) frame 10 (00:50) bubble-reduced surface
45
observed leakage frequently during our experiments with this breather. Although not
measured directly, the theoretical leakage pressure is calculated from equation 3-2 to
be 3×103 Pa (=0.44 psi). Here, we assumed the liquid surface tension γ =72.8
dynes/cm, advancing contact angle of Teflon® θadv= 122o, and the radius of breathing
holes r =25μm.
In this sense, leakage prevention is the main concern of the first-generation
distributed breather, because leakage is unacceptable in the practical microfluidic
devices. Smaller venting holes are necessary for higher leakage onset pressure. It is
possible to refine this fabrication approach and fabricate smaller (e.g., several micron)
venting holes by DRIE. However, fabrication of submicron venting holes is difficult,
limited by both lithography and aspect-ratio of DRIE. Considering venting holes of a
particular size, the aspect-ratio determines the maximum wall thickness and thus the
strength of the structure. Nonstandard microfabrication process, such as porous silicon
etching should be exploited to fulfill these requirements. Nevertheless, hydrophobic
coating inside submicron holes is another challenge. The holes tend to be blocked and
the uniformity of hydrophobic coating is hard to be ensured.
A simpler and cheaper solution is provided by hydrophobic nanoporous
membrane, without delicate microfabrication process.
3.4 Membrane Sandwiched Breather
Hydrophobic porous membranes have been developed for sample preparation
of X-ray spectrochemistry [20], high performance liquid chromatography (HPLC) [21]
and ultrafiltration [22]. The typical pore diameter of 0.2~ 3μm is adequate for the
venting application. No additional coating is needed, because the material is
46
intrinsically hydrophobic. The membrane can be sandwiched to construct the second-
generation venting plate as Figure 3-12 shows.
Figure 3-12: Configuration of the second-generation venting plate
With 2000 Å SiO2 as the mask, two identical silicon chips are etched through
by KOH, making 200μm square openings at the bottom of the pits. Porous
polypropylene film with ~0.2μm-diameter pores [20] is sandwiched between the two
chips. The pits, with relatively hydrophobic bare silicon surfaces, can serves as type II
bubble sinks as stated in chapter 2. This second-generation venting plate is fitted into
the experiment setup shown in Figure 3-10 to substitute the first-generation venting
plate.
Again, sodium bicarbonate (NaHCO3) solution and weak sulfuric acid (H2SO4)
are injected into the microchamber sequentially and generate carbon dioxide (CO2) gas
bubbles by chemical reaction. The results of this experiment are shown in Figure 3-13.
In this experiment, large CO2 bubbles shrink into small ones confined within the
two phase flow
Teflon® coating
SiO2
Si
KOH pit
epoxy
~200 μm
hydrophobic porous membrane
47
KOH-etched pits. No leakage is observed during the entire experiment. We attribute
the better leakage-prevention performance to both smaller pore size (~0.2 μm in
diameter, as specified by the manufacturer) and uniform hydrophobic nature of the
membrane material.
Figure 3-13: Venting experiment in the second-generation breather.
In order to measure the leakage onset pressure, a piece of hydrophobic porous
membrane is fixed into a dead-end microchamber as illustrated in Figure 3-14. A
3mm×3mm square area of the membrane is exposed after sandwiching. The
microchamber is DIRE-etched on a piece of 400μm-thick silicon chip, which is
anodically bonded to a piece of Pyrex® glass. Then membrane and tubing are glued by
epoxy.
frame 1 (00:00) NaHCO3-filled chamber
frame 2 (00:07) introduce H2SO4
frame 3 (00:33) CO2 generated
frame 4 (00:46) gas bubbles peak
frame 5 (01:02)
frame 6 (01:09) captured bubbles
frame 7 (01:15) frame 8 (01:25) bubbles shrink
frame 9 (01:40) frame 10 (02:06) bubble-reduced surface
48
Figure 3-14: Determination of leakage onset pressure.
The liquid is pressurized by a gas tank through a reservoir. The pressure is
increased gradually from 0psi by adjusting a regulator. The liquid leakage flow rate
corresponding to the pressure is monitored by a flow meter with pressure sensor. Once
leakage occurs, a steady flow rate can be read out of the flow meter. Liquid droplets
can be eventually observed on the outer surface of the membrane. The leakage onset
pressure can thereby be recorded as the point where the pressure starts to increase.
Figure 3-15 shows the flow-pressure curve for DI water in the porous-membrane-
covered microchambers. If the pressure is reduced after obvious leakage, the flow rate
decreases linearly, following Darcy’s Law. This indicates that the leakage is
irreversible. Noticeably, no leakage has been observed for the porous polypropylene
film until the pressure reached 35psi, at which pressure the membrane breaks.
Pyrex® glass
membrane holder
epoxy
hydrophobic porous membrane
Si microchamber
3mm
3mm
hydrophobic porous membrane
flow meterwith pressure
sensor
regulator
liqui
d re
serv
oir
gas tank
inlet opening
tubing
49
Figure 3-15: Flow-pressure curves for DI water in porous-membranes-covered
microchambers.
Similar curves are also obtained for methanol (10M) aqueous solution in order
to verify that the membrane breathers also work for μDMFC. High concentration fuel
is desired in μDMFCs to reduce reservoir size and enhance energy density. Advances
in PEM technology are expected to substantially relieve the cross-over problem and
increase the feasible fuel concentration from current 0.5-2M to 8-10M. Due to the
absence of proper flow meter for Methanol, leakage onset pressure of Methanol is
estimated visually. Nevertheless, a comprehensive study on liquid entry pressure (i.e.
leakage onset pressure) of aqueous alcohol solutions [23] gives the data of same order
of magnitude, as table 3-1 summarizes. Again, the porous polypropylene film breaks
around 35psi. The leakage onset pressure of 35psi or higher can be considered safe for
μDMFC, where the working pressure inside fuel stack rarely exceeds 1psi.
membrane breaks
leakage onset pressure for p-PTFE
porous PTFE (d = 3μm) porous polypropylene (d= 0.2μm)
liqui
d le
akag
eflo
w ra
te
d: nominal diameter of pores (provided by manufacturers )
pressure (psi)
50
Table 3-1. Leakage onset pressure: calculated and measured value
Porous polypropylene Porous PTFE nominal
diametercalculated
value measured
value nominal diameter
calculated value
measuredvalue
10M Methanol 22psi > 35psi * 1.5psi ~5psi DI water 0.2μm 110psi > 35psi * 3μm 7.4psi 16psi
*membrane breaks at 35 psi
The leakage onset pressure is even higher than the value calculated from the
nominal diameter specified by the manufacturers. This means the estimation by the
model described in the Figure 3-8 is too conservative for the porous hydrophobic
membrane. The reason can be attributed to the irregular holes in porous membrane and
their rough surface. So the new model of the breathing hole is illustrated in Figure 3-
16.
Figure 3-16: An irregular venting hole in porous membrane.
From this model, the leakage onset pressure of a complex single hole is:
P0
θeff
reff
PF w
51
]/)[cos(2 effefffhole rMaxP θπσ −⋅= , 3-3
which means leakage occurs in the most constricted part of a hole. The leakage onset
pressure of the whole membrane can be described as:
]min[ holemembrane PP = , 3-4
which means leakage occurs at the largest hole of a membrane. Although reff and θeff
are difficult to measure, the tendency is obvious: both smaller reff and larger θeff
increase leakage onset pressure. This explained the unexpected experimental results.
The rough surface and irregular holes are confirmed by SEM pictures, shown in
Figure 3-17.
(a) porous polypropylene (b) porous PTFE
Figure 3-17: SEM pictures of porous membranes.
The leakage onset pressure of 35psi or higher is considered safe for most
multiphase microfluidic applications. For example, in μDMFC, the pressure inside
fuel stack rarely exceeds 1psi. The successful demonstration of gas bubble removal
from water/methanol mixture verified that this breather can be used in μDMFC. It is
1μm 0.3 μm
52
more difficult than pure water case because both contact angle and surface tension are
decreased.
3.5 Gas-Permeable Microchannels
Distributed bubble-capturing breathers can be used to collect gas bubbles and
remove them from microchambers. The same mechanism can be used to remove gas
bubbles from microchannels too. A gas permeable microchannel is fabricated as
Figure 3-18 shows.
Figure 3-18. Alignment and bonding setup for the gas permeable microchannel.
The breather chip and membrane holder are both fabricated from a same
400μm-thick (100) silicon wafer by DRIE etching. A cross-shaped gas-bubble
generator [7] is also designed on the microchannel chip to produce adjustable
liquid/gas two-phase flow, corresponding to μDMFC’s fuel flow with CO2 bubbles.
membrane(Semi-transparent)
illuminate from below
alignment marks
Pyrex®
glass
membrane holder chip
microchannel chip
Si
Si
53
Parts of the microchannel on the breather chip are protected by polyimide tape after
the DRIE etching has reached the desired depth. After DRIE and subsequent Piranha
cleaning, the breather/bubble generator chip is anodically bonded to a piece of Pyrex®
glass. Then the membrane is sandwiched between breather chip and membrane holder
chip with epoxy to complete a gas-permeable microchannel, analogous to μDMFCs’
anodic microchannel. During the epoxy adhesive bonding, through-holes on both
chips are used as alignment marks. The alignment is assisted by strong illumination
from below.
Figure 3-19. The gas permeable microchannel with an on-chip bubble injector.
fuel inlet
gas inlet
to pressure sensor
hydrophobic porous membrane
pressure buffering channels
bubble generator breather pressure buffer
outlet
b) Cross-section view membrane holder
microchannel
epoxy gas bubbles
Si
Pyrex ® glass
membrane
a) Top view
54
The finished device (as Figure 3-19 shows) can then be connected to the flow
meter, pressure sensor, liquid reservoir and gas tank, through a set of Upchurch®
fitting and tubing apparatus, similar as Figure 3-14 shows. The gas and liquid are
injected into the microchannel via the tubing to form two-phase flow. Pressurized CO2
is injected directly and the liquid is driven indirectly via a liquid reservoir (not shown
in the Figure). Flow meters and pressure sensors are used to monitor the flow rate and
pressure inside the microchannel. The experimental data are collected by a computer.
DI water and 10M methanol are tested to verify the venting function of gas
permeable microchannel. The transmembrane pressure (Pl –P0) is controlled to be 0.5-
2psi. Reliable venting is observed in all the tests. Figure 3-20 shows a typical bubble
venting process. The gas permeable microchannel is formed by porous polypropylene
membrane. The working fluid is 10M methanol, under a transmembrane pressure of
0.8 psi.
00s 01s 02s 03s 04s 05s
06s 07s 08s 09s 10s 11s
Figure 3-20. Venting of a bubble in a gas permeable microchannel.
55
Based on the experimental video, a time-space diagram of a venting bubble can
be generated, as Figure 3-21 demonstrates. The position of the two meniscuses, as
well as the bubble length d, is shown as a function of time. The flow direction affects
the breathing rate geometry of the bubble. The upstream bubble meniscus (left side)
moves faster than the downstream bubble meniscus (right side). The “stick and slip”
motion of the bubble boundaries in the microchannel suggests a complex interplay
between dynamic contact angle and hysteresis, shear, breathing, and geometry.
Figure 3-21. A typical time-space diagram of a venting bubble.
Although bubble removal is observed for both DI water and 10M methanol, the
two-phase flow in gas permeable microchannel behaves different in the two kinds of
working fluid. In DI water, once the gas bubble touches the unblocked porous
membrane, it starts to be vented out, as Figure 3-22-a demonstrates. However, a
bubble train is formed in 10M methanol while the leading bubble is vented out. The
56
bubbles in the bubble train essentially keep intact until they become the leading bubble.
The bubble train can be pretty long, as Figure 3-22-b shows. This phenomenon is
named as venting threshold because it suggests that bubbles need to break a certain
barrier before they can really be vented out freely.
a).DI Water
PF – Po =0.8psi
c).10M MethanolPF–Po =0.5psi
gas bubble
leading bubble
b).10M MethanolPF – Po =0.8psi
leading bubble
long bubble train
shorter bubble train
a).DI Water
PF – Po =0.8psi
c).10M MethanolPF–Po =0.5psi
gas bubble
leading bubble
b).10M MethanolPF – Po =0.8psi
leading bubble
long bubble train
shorter bubble train
Figure 3-22. The venting threshold.
Two reasons can account for the venting threshold. One hypothesis is that a
tiny liquid droplet can be trapped inside the venting hole and block the breathing of
gas (Figure 3-23-a). Due to contact angle hysteresis, the contact angle at the head and
tail of this droplet (θhead and θtail) can be different. This introduces a corresponding
pressure difference that can hold this droplet in the venting hole temporarily. However,
57
the eventual success of venting suggests that the tiny droplet in this hydrophobic nano-
pore is unstable. It can finally be removed to clear the venting hole. Another
hypothesis is that a thin liquid film is first formed between the bubble and the venting
holes (Figure 3-23-b). It takes certain time for the liquid film to break. Then the gas
bubble can be vented out freely.
Figure 3-23. Hypothetical explanations for the venting threshold.
Both of the hypotheses can be supported by two facts. First, experiments with
DI water, whose contact angle hysteresis is much smaller than that of 10M methanol,
demonstrate less or no venting threshold at all. Smaller contact angle hysteresis can
increase the instability of both tiny bubble and liquid film. Second, the bubble train is
observed to be much shorter when the flow rate is reduced, as Figure 3-22-c shows.
Slower flow gives longer time and more opportunities to eliminate the trapped droplet
or break the liquid thin film. Nevertheless, more investigation is necessary to confirm
these hypotheses or find new explanation for the phenomenon of venting threshold.
Since the bubble train will evidently increase the flow resistance in the microchannel,
a). trapped droplet b). liquid thin film
P0 θhead r
PF w PB
d
θtail
P0 r
PFw PB
d
58
theoretical analysis and corresponding design modification should be made in the
future to minimize the threshold in the working μDMFC.
3.6 Summary and Future Directions
The existence of gas bubbles presents a major challenge for modern
microfluidic devices, such as μTAS and μDMFC. The necessity of rapid bubble
removal becomes even more urgent in μDMFC, where gas bubbles are continuously
generated.
Distributed bubble-capturing breathers are developed to collect the gas bubbles
from two-phase flow and vent them out by hydrophobic venting. The concept is
proven by 50μm silicon-micromachined holes with Teflon® coating. Despite the low
level of leakage prevention, the device helped to identify the hole-size and coating
uniformity as key points of the further improvements. Compared with the silicon-
micromachined breathing holes, the leakage onset pressure has been improved
significantly by hydrophobic nanoporous membrane because of their uniform
hydrophobicity and small pore size. Successful venting from 10M methanol fuel with
35psi or higher pressure tolerance has fulfilled the requirement of the next-generation
μDMFC.
Characterization of the porous membranes shows higher leakage onset pressure
than the theoretical value calculated from the nominal diameters, by a straight and
smooth capillary model. The unexpected performance can be explained better by an
irregular capillary model, which is supported by SEM pictures of the membranes.
Gas/liquid two-phase flow is generated by bubble injectors and released from a
gas permeable microchannel successfully. Both DI water and 10M methanol can be
59
degassed by this hydrophobic nanoporous venting mechanism. However, the venting
threshold is found for gas permeable microchannel filled with methanol aqueous
solution. Trapped droplets and liquid thin film are proposed to explain the venting
threshold.
Gas permeable microchannel opens up a new research direction of both
application importance and fundamental interest. Venting threshold deserves more
investigation because use of organic solvent is prevalent in microfluidic devices and
the bubble trains will evidently increase the flow resistance in the microchannel.
Theoretical analysis and corresponding design modification should be made in future
to minimize the venting threshold.
Further study on the venting rate is also expected to elucidate some
fundamental issues on the behaviors of microscopic bubbles/droplets inside the
microchannel. A design guideline can therefore be provided to the relevant
microfluidic devices.
References
[1] C.-M. Ho and Y.-C. Tai, "Micro-Electro-Mechanical-Systems (MEMS) and Fluid
Flows," In: Annual Review of Fluid Mechanics, 1998, vol. 30, pp. 579-612.
[2] J. Evans, D. Liepmann, and A. P. Pisano, "Planar Laminar Mixer," Proc. The 10th
IEEE Int. Conf. on Micro Electro Mechanical Systems, Jan. 19-23, 1997, pp. 96-
101.
60
[3] Y. K. Lee, P. Tabeling, C. Shih, and C. M. Ho, "Characterization of a MEMS-
Fabricated Mixing Device," Proc. 2000 ASME Int. Mechanical Engineering
Congress and Exposition, Orlando, Florida, 2000, pp. 505-11.
[4] H.-C. Chang, "Chapter 11: Bubble/Drop Transport in Microchannels," in the
MEMS handbook, M. Gad-el-Hak, Ed.: RCR Press, 2002, pp. 11.1~.3.
[5] H. Matsumoto and J. E. Colgate, "Preliminary Investigation of Micropumping
Based on Electrical Control of Interfacial Tension," Proc. IEEE Micro Electro
Mechanical System Workshop, 1990, pp. 105-10.
[6] P. Gravesen, J. Branebjerg, and O. S. Jensen, "Microfluidics-a Review," Journal of
Micromechanics and Microengineering, 1993, pp. 168-82.
[7] T. Cubaud and C.-M. Ho, "Transport of Bubbles in Square Microchannels,"
Physics of Fluids, 2004, vol. 16, pp. 4575-85.
[8] A. Kawahara, P. M.-Y. Chung, and M. Kawaji, "Investigation of Two-Phase Flow
Pattern, Void Fraction and Pressure Drop in a Microchannel," International
Journal of Multiphase Flow, 2002, vol. 28, pp. 1411-35.
[9] M. J. Jensen, "Bubbles in Microchannels," M.Sc. Thesis, Technical University of
Denmark (DTU), 2002
[10] M. A. Tenan, S. Hackwood, and G. Beni, "Friction in Capillary Systems,"
Journal of Applied Physics, 1982, vol. 53, pp. 6687-92.
[11] M. J. Jensen, G. Goranović, and H. Bruus, "The Clogging Pressure of
Bubbles in Hydrophilic Microchannel Contractions," Journal of Micromechanics
and Microengineering, 2004, pp. 876-83.
61
[12] A. P. Ershov, Z. M. Zorin, V. D. Sobolev, and N. V. Churaev, "Kinetics of Gas
Bubble Motion in a Capillary," Colloid Journal, 2001, vol. 63, pp. 179-85.
[13] S.-J. Kim, Y. B. Shin, D.-S. Lee, H. Yang, K. Kim, S. Park, and Y. T. Kim,
"Capillary-Driven Passive Retarding Microvalve Based on an. Aspect Ratio
Concept," The 7th International Conference on Micro Total Analysis Systems,
Squaw Valley, California USA, October 5-9, 2003, pp. 175-8.
[14] J. Kohnle, G. Waibel, R. Cernosa, M. Storz, H. Ernst, H. Sandmaier, T. Strobelt,
and R. Zengerle, "A Unique Solution For Preventing Clogging Of Flow Channels
by Gas Bubbles," Proc. The 15th IEEE Int. Conf. on Micro Electro Mechanical
Systems, Las Vegas, Nevada, 2002, pp. 77-80.
[15] W. Y. Sim, G. Y. Kim, and S. S. Yang, "Fabrication of Micro Power source
(MPS) Using a Micro Direct Methanol Fuel Cell (μDMFC) for the Medical
Application," Proc. The 14th IEEE Int. Conf. on Micro Electro Mechanical
Systemson, Interlaken, Switzerland, 2001, pp. 341-4.
[16] T. J. Yen, N. Fang, X. Zhang, G. Q. Lu, and C. Y. Wang, "A Micro Methanol
Fuel Cell Operating At Near Room Temperature," Applied Physics Letters, 2003,
vol. 83, pp. 4056-8.
[17] J.-H. Tsai and L. Lin, "Active Microfluidic Mixer and Gas Bubble Filter Driven
by Thermal Bubble Micropump," Sensors and Actuators A, 2002, vol. 97-98, pp.
665-71.
62
[18] K. Hosokawa, T. Fujii, and I. Endo, "Handling of Picoliter Liquid Samples in a
Poly(dimethylsiloxane)-Based Microfluidic Device," Analytical Chemistry, 1999,
vol. 71, pp. 4781-85.
[19] Z. Yang, S. Matsumoto, and R. Maeda, "A Prototype of Ultrasonic Micro-
Degassing Device for Portable Dialysis System," Sensors and Actuators A, 2002,
vol. 95, pp. 274-80.
[20] "ChemplexTM's X-Ray Spectrochemical Sample Preparation,"
http://www.chemplex.com/.
[21] A. Weston, "HPLC and CE: Principles and Practice": W. B. Saunders Company,
1997.
[22] "FluoroporeTM Membrane Filters,"
http://www.millipore.com/catalogue.nsf/docs/C254.
[23] M. C. Garcia-Payo, M. A. Izquierdo-Gil, and C. Fernandez-Pineda, "Wetting
Study of Hydrophobic Membranes via Liquid Entry Pressure Measurements with
Aqueous Alcohol Solutions," Journal of Colloid and Interface Science, 2000, vol.
230, pp. 420-31.
63
CHAPTER 4
MICROPUMPING ENABLED BY HYDROPHOBIC VENTING OF BUBBLES
4.1 Introduction
Recently, much attention has been given to the research of microfluidic devices,
with μTAS as a well-known example. Microfluidics has provided exciting tools for
ink-jet printing [1-4], micro power generation [5-8], as well as biomedical researches
and practices [9, 10], such as genomics, proteomics, clinic diagnostics, pharmaceutics
and drug delivery. The micropump is an essential component to mobilize the fluid for
microfluidic devices. Without a proper pumping mechanism, most functions of the
microfluidics, including mixing, separation, and detection become either impossible or
less effective. Syringe pumps and off-chip pressure sources can be employed at the
early stage of research and development. However, a proper pumping mechanism
must be integrated with the practical microfluidic products, so that the advantages of
microfluidics can be fully realized.
Micropumps can be classified into two categories, depending on whether they
have solid mechanical moving parts or not. Micropumps with solid mechanical
moving parts [11-15] are straightforward in their working mechanisms, with clear
analogues in macroscopic world. Low leakage rates and high pressure heads can also
be achieved by using mechanical check valves [16]. However, the fabrication of such
kind of micropumps is usually complex, and large-scale integration is typically
64
difficult. Long-term reliability is also a concern due to the much more severe wear and
tear experienced by these micropumps as compared to their macro-counterparts.
On the other hand, micropumps without solid mechanical moving parts [14, 15,
17-19] typically utilize the unique properties of liquid in microchannel. These
properties are usually negligible factors for macroscopic flow. However, they
outweigh other factors and become dominant in microchannel. The surface tension [20]
serves as a good example of this kind of factors. Generally speaking, these
micropumps allow simpler fabrication, easier integration and more reliable long-term
operation. Nevertheless, electroosmotic or electrohydrodynamic micropumps are
usually sensitive to the nature of the liquid or surface.
Typically independent of solid mechanical moving parts, the bubble-driven
micropumps [17-19] are less demanding on the fluid properties. Nevertheless, they
also suffer major drawbacks, which can be attributed to the thermal generation of gas
bubbles (boiling), the most common bubble-actuation approach. First, the required
heat flux of boiling increases dramatically with shrinking size [21], which makes the
thermal bubble generation in microscale an “energy hungry” process [22]. Secondly,
boiling is a complex thermal-physical phenomenon. The onset of boiling can be
significantly affected by surface property, which is prone to fabrication variation and
contamination. The heat transfer rate can be considerably impacted by boundary
conditions, which is sensitive to environmental changes. Therefore, it is difficult to
precisely predict or control the bubble growth rate (actuation speed), which can show
significant discrepancies due to the fabrication process and environmental changes.
Thirdly, overheating denatures most biological large molecules (e.g. DNA and protein)
and affects biomedical liquid samples irreversibly. Fourthly, there is a tradeoff
between the generation (boiling) and collapse (condensation) of thermal bubbles. Slow
65
heat dissipation is preferred to decrease the power consumption of boiling. However,
rapid condensation of thermal vapor bubbles ironically requires quick heat dissipation
in the same device. Since the natural condensation is usually far slower than boiling,
bubble collapse represents the limiting step. The power efficiency has to be
compromised to increase the actuation frequency or average pumping rate.
Other bubble generation approaches, such as electrolysis, injection and
chemical reaction, have been tried. However, removal of insoluble gas bubbles from a
sealed device is even harder and slower than condensation, if possible at all [23, 24].
As a result, most bubble-driven pumps are made open (i.e., more like dispensers), so
that the bubbles are expelled with the liquid [1, 2, 25]. Even condensation itself cannot
completely remove the entire vapor bubble residue within a practical time scale. And a
downstream opening is usually necessary to avoid bubble clogging of the fluid loop
[26]. There is no bubble-driven micropump suitable for a closed-loop fluidic device in
a portable system like a micro Direct Methanol Fuel Cell (μDMFC) [5] or circular
chromatographic [27].
In order to explore these bubble sources for micropumping, a breakthrough in
universal and rapid gas removal is required. The hydrophobic nanoporous venting
discussed in chapter 3 represents such a mechanism. In addition, the venting rate
associated with nanoporous membrane can be faster than natural condensation of
vapor bubbles. This universal gas removal actually enables virtually all the existing
bubble generation approaches for micropumping, as well as bubble actuation in
general. By choosing the proper bubble generation approach, the concerns of
individual applications (e.g. energy efficiency, thermal sensitivity, bio-compatibility,
adjustable flow rate), can be addressed specifically.
66
4.2 Comparative Study of Electrolysis and Boiling for Actuation
A proper bubble source is essential for the proposed bubble-driven pumping
mechanism. First, electrolysis of water is chosen as the bubble generation approach to
be employed in these proposed micropumps. Electrolysis can be simply achieved by
two electrodes in nonspecific aqueous solution, without the trouble of PZT patterning
for cavition bubbles. Compared with boiling, electrolysis possesses apparent
advantages. An electrolysis-bubble actuated valve [23] has been reported to consume
power of four orders of magnitude less than a similar thermal-bubble actuated valve
[28]. In addition, electrolysis bubble actuation has recently been used to manipulate
living cells [29], which is difficult for thermal bubble actuation. And yet, a systematic
study has not been conducted to compare the two bubble actuation approaches. With
universal gas removal, the time is ripe for this kind of study. A design guideline can
thus be provided by comparing electrochemical bubbles with thermal bubbles for the
actuation of microdevices, including micropumps.
4.2.1 Definitions and Assumptions
Considering the diverse applications of bubble actuators, it is imperative to
focus on the most important factors in a simplified scenario. It is decided that bubble
generation in a bulk liquid environment (DI water) by electrolysis and boiling will be
analyzed first. In spite of the simplifications, the results will provide information for
the study of more complex configurations, such as bubble actuators in a microchannel.
Bubble collapsing will not be discussed here because it depends on the specific gas
removal techniques and shows little difference if a universal bubble removal approach
(i.e. hydrophobic venting) is applied.
67
The size of bubble-driven microactuators typically varies from tens of
micrometers to several millimeters. Circular electrodes and heater with a radius of
120μm are used for the first test of electrolysis and boiling bubble actuation,
schematically shown in Figure 4-1. Two kinds of substrate are tested: glass and
SiO2/Si (insulator thickness is 0.13μm).
The experiments always start with increasing voltage/power input from zero
until bubble generation is observed. The power is recorded as the minimum power
requirement for bubble generation. After that, several data points are acquired to
record the voltage, current and corresponding bubble growth rate in volume per
second. The measured bubble growth rate (actuation rate) will be put into the models
to calculate the theoretical power requirements, which can then be compared with the
experimental data to validate the theoretical models.
Figure 4-1. Schematic view of the test chips for bubble-driven actuations.
4.2.2 Electrolysis for Bubble-Driven Actuation
The average gas (H2 and O2) bubble growth rate (ΔV/Δt) is determined by the
current consumption of electrochemical bubble generation:
isolation layer electrodes
substrate
isolation layer heater
substrate
bubble
Electrolysis Boiling
water droplet
68
e
bubble
CI
tV
=Δ
Δ , (4-1)
where Ce = 5.2 x 106 A.s/m3 is a constant under the controlled experimental conditions
(1atm, 300K).
The minimum voltage input can be determined by the Nernst equation as
1.23V for DI water. The actual voltage is higher than this theoretical value, depending
on both ion concentration and electrode distance. The actual voltage is also
experimentally measured. Table 4-1 summarizes the experimental results and power
consumption calculated from the model.
Table 4-1. Experimental data for electrochemical actuation
voltage (volt) 3.2 4.0 5.0 6.0 8.0 current (μA) 4.9 16 43 70 140 power (μW) 16 64 220 420 1100 bubble growth rate (10-13m3/s) 2.0 5.2 18 26 49
theoretical power (μW) 3.3 11 48 84 210
*measured in glass substrate, Si substrate shows similar data
The measured power consumptions are always higher (~5 times) than the
theoretical predictions. This discrepancy is due to the large amount of power (~80%)
consumed by the liquid circuit between the two electrodes. This power loss can be
reduced by decreasing the electrode distance or increasing the ion concentration.
However, a linear relationship between the current and actuation speed is observed, as
shown in Figure 4-2. This implies that the actuation speed can be both measured from
and controlled by the current input. This unique feature of electrochemical bubble
69
actuation can be employed to achieve a stable actuation speed by using a constant-
current power source or a feedback control circuit.
(140, 49)
(70, 26)
(43, 18)
(16, 5.2)
(4.9, 2)0
10
20
30
40
50
60
0 50 100 150Current (μA)
Gro
wth
rate
( 10−
13m
3/s
)
Figure 4-2. Speed control of the electrochemical bubble actuation.
4.2.3 Boiling for Bubble-Driven Actuation
Although microscale thermal bubble generation has been studied previously
[22], there has not been a general model or systematic experiment to explain the
minimum power consumption and actuation speed. We use a very approximate model
shown in Figure 4-3 to estimate the power consumption of thermal bubble actuation
and examine the role of several essential factors, such as substrate material, heater size,
and isolator thickness.
70
Figure 4-3. Model for the thermal bubble actuation.
The total power consumption of boiling consists of three parts: vaporization
heat of water (Pevp), heat loss into substrate (Psb) and heat loss into water (Pwt). Pevp
can be estimated from the average bubble growth rate:
tVCP evpevp Δ
Δ⋅= , (4-2)
where 36 /1034.1 mJCevp ×= if the thermal actuation is conducted under the controlled
experimental conditions (1atm, 300K). Heat conduction into the substrate (Psb) and
water (Pwt) are considered separately. Psb can be estimated as:
)(0 ∞−⋅⋅= TTRKP hsbsb α , (4-3)
from two-dimensional steady state heat conduction analysis [30]. Here, Ksb is the
thermal conductivity of the substrate, and R0 is the heater radius. The proportionality
factor α is 4 if the substrate thickness is much greater than R0.
For a homogeneous substrate (e.g. glass), Psb can be calculated from Equation
4-3 directly. If there is an isolation layer on the substrate (e.g. SiO2 film on a Si
T∞ =27oC
R0
Th =100oC
h
water: Kwt
substrate: Ksb
isolator: Kis
heater
isolator/substrate interface: Tis-sb
R0>>h
71
substrate), the two-layer substrate can be modeled by assuming two thermal
resistances connected in series. Heat loss on this substrate can be calculated as:
)(0 ∞−⋅⋅′= TTRKP hsbsb , (4-4)
where sbissb KKR
hK απ
110
+=′ is the equivalent thermal resistance. Heat loss into the
water (Pwt) can be calculated by using an equation similar to Equation 4-3:
)(4 0 ∞−⋅⋅= TTRKP hwtwt , (4-5)
The total power consumption can then be estimated as:
wtsbevpboil PPPP ++= , (4-6)
Our model predicts that Psb > Pwt » Pevp for both glass and SiO2/Si substrates.
Psb is estimated to be of the order of 100 mW for glass substrates and of the order of
1000 mW for SiO2/Si substrates. These values are consistent with the experimental
results listed in Table 4-2. Pevp is much smaller and is on the order of several μW.
This suggests that power consumption in thermal bubble actuation is dominated by
heat loss into the substrate and not by phase change.
Table 4-2. Experimental data for the thermal bubble actuation
substrate Glass (1.5mm) voltage (V) 3.9 4.5 5.0 6 power (mW) 180 240 280 366 bubble growth rate (10-13m3/s) 14 16 17 28
substrate SiO2(0.13μm)+Si (500μm) voltage (V) 12.8 14 15 17 20 power (mW) 2500 3000 3400 4300 5900 bubble growth rate (10-13m3/s) 1.8 4.0 7.2 12 36
72
Our approximate model does not take into account transient heat transfer in the
water and cannot capture the actuation speed dependence. More detailed models must
take into account transient convective heat transfer in the water and finite superheating
in the heater. The amplitude and temporal shape of power input affect the temperature
distribution in the water and thus change the actuation speed, as observed in the
experiments. But this relationship is far less straightforward than the actuation-
speed/current relationship in electrolysis. More importantly, the boiling phenomenon
can be affected by both fabrication variations and environmental changes. Precise
control of the thermal bubble actuation speed is much more difficult than that of
electrolysis.
4.2.4 Comparison between Electrolysis and Boiling
Minimum power consumption: For the specific design of actuators, electrolysis gas
bubbles can be generated with tens of μW, while the thermal vapor bubbles require
hundreds of mW (glass substrate) or even several W (SiO2/Si substrate). Even if the
substrate is a near-perfect insulator (e.g., device on a membrane), thermal bubble
generation still consumes tens of mW to heat the water.
Actuation speed vs. power consumption: The experimental data indicate a wider
range of actuation speed for electrolysis, with only 10-5-10-3 of the power consumption
of boiling, as Figure 4-4 shows. It is also confirmed that the speed of electrochemical
bubble actuation can be both measured from and controlled by the current input.
Conversely, thermal bubble generation lacks this feature.
73
0
10
20
30
40
50
60
0.01 0.1 1 10 100 1000 10000Power (mW)
Gro
wth
rate
(10
-13 m
3 /s)
ElectrolysisBoiling on glass substrateBoiling on Si substrate
Figure 4-4. Comparison between electrochemical and thermal bubble actuation.
Scaling effect of power consumption: The simple models make it easier to predict the
scaling effect of power consumption. The existing data can therefore be used to
provide information for smaller or larger actuators. The characteristic length (L) in the
scaling analysis can be defined as the radius of electrodes or heater (R0).
Since V ∝ L3, Equation 4-1 gives the scaling effect of electrolysis current: I ∝
L3. If the same voltage (v) is used for different electrode sizes, the power consumption
is: Pelc = Iv ∝ L3. On the other hand, considering that Pevp can be neglected, Equations
4-2, 4-5 and 4-6 indicate the scaling effect of thermal bubble actuation: Pboil∝L.
Therefore, the efficiency advantage of electrolysis over boiling is expected to become
greater for even smaller actuation bubbles.
In term of power efficiency, controllability, bio-compatibility and
miniaturization potential, electrolysis actuation is a better approach than thermal
74
bubble actuation. However, a microactuation mechanism using electrolytic bubbles
most likely would require a provision to remove the bubbles, increasing the
complexity of devices. Aspects beyond power consumption should also be considered
to determine an appropriate bubble generation mechanism for a specific application.
4.3 Electrochemical-Bubble Driven Pump
Due to its advantages over other bubble generation approaches, electrolysis is
chosen as the first bubble source to be implemented in the new category of pumping
approach. The working principle of this micropump is based on two essential
components: a virtual check valve for asymmetric bubble growth and hydrophobic
nanoporous venting for bubble removal.
4.3.1 Essential Components
The flow direction of this pumping mechanism is controlled by a “virtual
check valve” [31] without any mechanical moving parts. The same principle has been
used in bubble-driven micropumps by thermal vapor [19] or injected gas [25]. The
structure of this virtual check valve is illuminated in Figure 4-5. For a bubble
generated on the boundary of an abrupt step in the microchannel, the menisci on either
side of the bubble can withstand a certain pressure difference. Given a square cross-
section for both sections of the microchannel, the maximum pressure difference that a
meniscus can withstand is:
( ) ili hP ασ cos2max ⋅=Δ , (4-7)
where σl is the surface tension of liquid, α is the contact angle, hi stands for the
channel depth (width), i=1,2 represents the two sections of the microchannel. Since h1
75
is smaller than h2, ΔP1max is larger than ΔP2
max. When the actual pressure generated
inside the growing gas bubble falls between these two values (ΔP1max > Pbubble
> ΔP2max), the left meniscus remains still, while the right meniscus will be moved
towards the right. The liquid in the larger channel can thus be pushed and form a net
rightward flow. This structure serves as a check valve by using surface tension,
without any mechanical moving parts.
Figure 4-5. The virtual check valve for gas bubbles.
The principle of hydrophobic venting has been explained in chapter 3. By using a
porous polypropylene membrane, venting is achieved for both DI water and 10M
methanol, the high-concentration fuel of μDMFC , with pressure tolerance as high as
35psi. Venting can be implemented in any orientation, free of any gravitational effects.
In addition, the hydrophobic venting technology works for both soluble and insoluble
gases, which makes it a universal gas removal approach. This study will also show
that the venting is usually faster than bubble generation by electrolysis. Thus, bubble
removal is no longer the bottleneck for pumping rate.
4.3.2 Pumping Concept The general pumping concept is schematically described in Figure 4-6. To
simplify the analysis, a pumping cycle with a single bubble is divided into three steps:
(1) When a bubble grows at a location close to the virtual check valve of a
h1 h2 bubble liquid
76
microchannel, expansion of the meniscus is hindered by the check valve on the left.
Consequently, the bubble only grows to the right and pushes the liquid rightward. (2)
The shape of hydrophilic microchannel can be designed to promote rightward bubble
motion. A diverging shape microchannel is specified here [32, 33]. However, pumping
is also observed in a simple straight channel. Therefore the diverging channel design
can help, but is not required, to complete the pumping. The design of the channel
shape can be further optimized to improve the pumping performance. Another factor
that facilitates the rightward motion of bubble is the surface free energy difference
between hydrophilic channel wall (SiO2) and hydrophobic membrane. This energy
gradient makes the membrane a “bubble trap” (as discussed in chapter 2), which
attracts the gas bubble to it automatically. Therefore, the bubble is drawn into the
membrane region once it reaches the hydrophobic nanoporous membrane. (3) Through
the breathing holes in the membrane, the bubble is vented out without any liquid loss,
providing a pressure difference (Pl–P0) less than Pleak (e.g. ~35psi for the device in this
paper). The liquid then fills into the section symmetrically to replace the vacancy left
by the gas bubble. A pumping cycle is thus completed and a net pumping to the right
is achieved.
77
Figure 4-6: Pumping by directional growth and hydrophobic venting of gas bubbles:
the concept.
Although this concept is illustrated with a single bubble, the coexistence of
multiple bubbles is acceptable for pumping as long as the venting rate of the
membrane is sufficient to remove all the bubbles promptly. Therefore, continuous
bubble generation can be used for the bubble-driven micropump reported here. This
differs from traditional thermal-bubble-driven micropumps, which use pulsed power
inputs to generate a vapor bubble and then turn off the heater to wait for bubble
collapse. Continuous bubble generation without precise modulation can significantly
simplify the driving circuit, reducing both the device complexity and the power
consumption. Tolerance of bubble generation pattern also enables the pump’s
Step 1: directional bubble growth
Step 2: built-in bubble transportation
Step 3: symmetric bubble collapse
Vt
Vgleft Vg
right
Vsleft Vs
right
virtual check valve
bubble generator
hydrophilic hydrophobic
porous membrane
78
applications to where precise regulation of bubble generation speed is difficult (e.g.,
chemical reaction).
4.3.3 Pump Loop Configuration and Fabrication
This pumping concept is first implemented in a closed pump loop in order to
demonstrate continuous liquid circulation, as required in a free-standing portable
microfluidic device, such as μDMFC and μTAS. The elimination of open ends also
prevents some uncertainties about the device, such as evaporation and the pressure
effect of the menisci.
The pump chip and membrane holders are all fabricated from the same 400
μm-thick <100> silicon wafer by DRIE etching. On the pump chip, the microchannels
of the breather, the reservoir and the connection port are etched through. Other parts of
the pump loop are protected by polyimide tape once the DRIE etching has reached the
desired depth. After DRIE and subsequent Piranha cleaning, the pump chip is
anodically bonded to a piece of Pyrex® glass. Then the venting membranes are
sandwiched between the pump chip and membrane holders. They are bonded together
by epoxy adhesive to form a breather. During the epoxy adhesive bonding, through-
holes on both chips are used as alignment marks. The alignment is assisted by strong
illumination from below. Details of this alignment and bonding technique are
described in chapter 3 and illustrated as Figure 3-18. Two platinum wires are inserted
into the “bubble source” position as the electrodes for electrolysis. The finished pump
loop (as illustrated in Figure 4-7) is subsequently connected to a syringe via a
tubing/fitting/adapter apparatus.
79
Figure 4-7: Configuration of pump loop.
4.3.4 Preparation of Test
The working fluid, Na2SO4 aqueous solution (~0.2M), is filled into the finished
device through the syringe. The presence of ions in the solution can lower the voltage
drop between the anode and cathode and therefore ease the electrochemical reaction.
During the priming process, it is usual for gas bubbles to be introduced into the loop
accidentally [34]. The reservoir is therefore covered by the venting membrane so that
these bubbles can be vented out automatically. The pump loop can thus be kept
bubble-free to avoid bubble-clogging problems [35-37]. Therefore, it is not necessary
to perform degassing of the liquid [38] or vacuum priming [39]. A mechanical valve is
used to isolate the device from the syringe, after the loop is filled with working fluid.
checkvalve membrane holder
epoxy
reservoir expanding bubble
Vgright
Vcleft
Pyrex Glass Vg
left
Vcright
bubble pathventingmembrane
inlet virtual check valve
bubble generator
net flow direction
outlet virtualcheck valve
Area C
Area B
connection port to syringe
membrane holder
shrinking bubble
venting membrane
Area A
breatherA-A View
B-B View
B-B B-B
A-A A-A
pumping section
pump chip
80
4.3.5 Verification of Liquid Circulation in Pump Loop
DC voltage is applied between the two platinum electrodes after the pump loop
is filled and ready for test. Although the theoretical minimum voltage for electrolysis
of water is 1.23V, substantial electrolysis is observed only when the voltage is above
10V. The main reason for this relatively high operation voltage is attributed to the
distance between the two electrodes (~2mm). Much lower operation voltage can be
expected if the electrodes are lithographically integrated into the device.
For thermal-bubble-driven micropumps (e.g., [18] or [19]), certain patterns of
pulsed input power is typically necessary because the boiling needs to be halted to
allow the condensation of vapor bubbles. The regulation of voltage/power input
requires a driving circuit, which will introduce additional complexity to the system.
Unlike the boiling/condensation cycle for thermal bubbles, electrochemical bubble
generation does not have to be stopped during venting of the leading bubble.
Accordingly, DC voltage is used to generate gas bubbles continuously. In this way,
pumping efficiency can be improved and the driving circuit can be simplified. This
feature represents considerable benefits for the portable microfluidic devices.
Under an operation voltage of 20 Vdc, three areas in the pump loop (A, B and
C in Figure 4-7) are observed to verify the circulation of liquid. Bubble motion in the
pumping section (area A in Figure 4-7) is demonstrated in Figure 4-8. Fine bubbles
generated on the electrodes merge into individual large bubbles first (Figure 4-8.a).
Since the left side of the pumping section is blocked by the meniscus next to the
virtual check valve, the bubbles only move toward the right (Figure 4-8.b). Both the
shape of microchannel and surface energy difference facilitate the rightward bubble
motion (Figure 4-8.c) and deliver the leading bubble to the venting membrane (Figure
81
4.d). Once the leading bubble touches the venting membrane, it starts to be vented out
and collapses rapidly (Figure 4-8.c-e). Meanwhile, bubbles are continuously generated
by the electrodes to start the subsequent pumping cycles. The liquid is pushed
rightward along with the bubbles during the bubble growth and transportation phases,
introducing a net rightward flow.
a. 0s
b. 2.5s
c. 3.3s
d. 3.4s
e. 4.5s
f. 8.1s
Figure 4-8. Bubble motion in the pumping section (area A in Figure 4-7).
In order to verify the liquid circulation more concretely, the fluid uptake from
the reservoir was observed in area B of Figure 4-7. Fluorescent particles (4μm in
diameter) are mixed into the working fluid to visualize the flow. Figure 4-9 shows the
video sequence of fluid uptake captured by a fluorescent microscope. The flow close
to the inlet of the check valve was found to be essentially unidirectional with
occasional stops. Unidirectional fluid uptake implies that fresh liquid reactant from the
check valve electrodes breather
hydrophilic hydrophobic
82
reservoir can be supplied to the microreactor by using this pumping approach. For
example, in μDMFC, this means a constant supply of fresh methanol fuel and stable
fuel concentration in anodic channel due to sufficient mixing. The liquid circulation of
the whole pump loop is thus confirmed by the proper bubble motion mode and definite
fluid uptake from the reservoir.
a. t = 0s
b. t = 0.2s
c t= 0.4s
d. t= 0.5s
Figure 4-9. Fluid uptake from the reservoir (area B in Figure 4-7).
4.3.6 Characterization of Pump Loop
The microscopic particles in the fluid can also be employed to quantitatively
characterize the flow in the pump loop, on a simplified μ-PIV (micro Particle Image
Velocimetry) [40]. Area C in Figure 4-7 is chosen to perform velocimetry. The
volumetric flow rate can be calculated by multiplying the flow velocity with cross-
50μmcheck valves
net flow direction
83
sectional area (600μm × 300μm). Figure 4-10 demonstrates the particle motion under
20Vdc input voltage. Particle clusters are used to measure the flow velocity because
they can be distinguished easily. As expected, the flow is pulsatile and bidirectional in
nature, since brief backflow is observed in each pump cycle, as Figure 4-10.c-e shows.
However, definite net flow to the designed direction is verified, which further
confirms liquid circulation.
a. t = 0s, x = 0mm
b. t = 1s, x = 0.2mm
c. t = 2s, x = 0.6mm
d. t = 2.4s, x = 0.3mm
e. t = 19s, x = 1.3mm
f. t = 82.4s, x = 1.0mm
g. t = 84.4s, x = 3.0 mm
Figure 4-10. μ-PIV to determine the flow rate (area C in Figure 4-7).
It is noticed that the bidirectional flow pattern of area C is different from the
unidirectional flow in area B. This different flow pattern in the same fluidic loop is
reasonable because the flexible venting membrane buffers the interaction of the
isolated segments of liquid. For the flow in the reservoir (Figure 4-9, or area B),
1mm
net flow direction
84
leftward bubble growth is effectively blocked by the virtual check valve.
Consequently, only a brief stop of the flow is observed during “bubble growth” phase.
While the leading bubble is in either the “transportation” or “collapse” phase,
rightward flow is introduced. In area C, however, the flow is affected directly by the
right meniscus of leading bubble, which undergoes a brief retreat during the “bubble
collapse” phase. Therefore, a minor backflow can be observed in area C. Even though
the pump intrinsically provides pulsatile flow, the clockwise liquid circulation is both
theoretically justified and experimentally confirmed.
Different operation voltages are used to characterize the performance of the
pump loop by determining the relationship between driven voltage/power and
volumetric flow rate. The results are summarized in Table 4-3. This result indicates
that the flow rate is well controlled by power input and suggests a flow rate adjustable
bubble-driven micropump with a broad range of flow rates. The reason for this feature
is that the bubble generation rate is theoretically proportional to the current. High
repeatability of this correlation has been observed during the experiments, which
suggests that the pumping rate can be both measured from and controlled by the
current. This unique feature can be employed to stabilize the average pumping rate by
using a constant-current power source or feedback control circuit. In comparison,
generation of thermal bubbles (boiling) is complicated by both surface properties and
heat transfer boundary conditions. Precise control of thermal-bubble-driven
micropumps is much more difficult.
85
Table 4-3. Control of the volumetric flow rate in pump loop
Voltage (V) 10 20 30 40
Average Current (mA) 0.20 0.69 1.32 2.13 Average Power (mW) 2.0 13.8 39.6 85.2
Particle Velocity (μm/s) 25 33 50 75 Volume flow rate* (nL/s) 4.5 5.9 9.0 13.5
4.3.7 Characterization in Open Loop
Another important characteristic of a micropump is the pressure head and its
relationship to the flow rate. Direct measurement of pressure head requires integrated
pressure sensors, which will increase the complexity of the device. In this study, the
pressure head is measured in an open loop setup to reduce the device complexity and
simplify the fabrication process. The open loop pump is fabricated with a similar
procedure as its closed loop counterpart. The closed loop is thereby replaced by a
straight channel with the same pumping section, as Figure 4-11 illustrates. Through-
holes are etched at the two ends of this straight channel, with two glass tubes attached
on by epoxy. Working fluid (water with Na2SO4) is introduced slowly from the top of
the inlet tube by syringe. After the meniscus of the outlet tube rises to a certain height
and stabilizes, DC voltage is applied to start the pumping. The movement of the
inlet/outlet menisci during the whole process is recorded by a digital video system.
86
Figure 4-11. Open loop test setup.
The velocity of outlet meniscus is measured from the video clips. The
volumetric pumping rate can be calculated by multiplying the meniscus velocity by
the cross-sectional area of the outlet tube. Since the cross-section area of the inlet tube
is much larger than the outlet tube, the inlet meniscus remains roughly same during
pumping. The position of outlet meniscus can therefore be used to determine the
pressure head at any given time. Under an operation voltage of 20Vdc, the relationship
between the flow rate and the pressure head is revealed experimentally and shown in
Figure 4-12. When the pump just starts to work, the pressure head is close to zero, the
maximum flow rate is achieved at this point. However, the first two data points are
discarded because the pumping is yet not stabilized yet. Before the first bubble is
vented out, the flow rate only reflects the bubble growth rate. The practical maximum
pumping rate of ~60nl/s is obtained with ~50psi back pressure. One data point is
recorded whenever the height of outlet water column increases 1mm (equivalent to
11.8Pa pressure). After the outlet meniscus stays at the same level for more than 5 min,
the flow rate is considered to be zero. The maximum pressure head of ~140 Pa is
Pyrex® glass
inlet outlet
B B
A A
A-A View
pressure head (mmWater)
B-B View
87
obtained at this point. Two factors contribute to the relatively low pressure head.
Firstly, the channel size of virtual check valve (50μm) is still large. The pressure
difference that it can withstand is limited. Secondly, the square cross-section of the
hydrophilic microchannel allows the liquid to fill the corners, known as wedging
effect [41]. The liquid can therefore leak through the corners with a flow rate related
to the back pressure (pressure head). The total flow rate therefore reaches zero before
the virtual check valve fails. This is confirmed by the observation that the pumping
section (area A of Figure 4-7) still works properly when the pumping rate drops to
zero. Miniaturizing and rounding of the microchannels should be able to improve the
practical pressure head [42].
Figure 4-12. The flow rate vs. the pressure head in open loop test.
Under 20Vdc operation voltage, the average current during pumping is
measured as 0.7mA. Therefore, the power consumption of this experiment is 14mW.
0
10
20
30
40
50
60
70
30 50 70 90 110 130 150Pressure (Pa)
Flow
rate
(nL/
s)
maximum flow rate: ~60nL/s
maximum pressure head: ~140 Pa
88
In order to achieve the similar maximum flow rate in thermal bubble-driven
micropumps, power consumption above several hundred mW [12, 18, 19, 22] is
usually required. In other words, the efficiency is improved approximately 10~100
times in this electrochemical-bubble-driven pump.
4.4 Exploration of More Pumping Approaches
The electrochemical-bubble-driven pump shows several significant
improvements over traditional thermal-bubble-driven pumps. However, there are still
many other bubble generation approaches to be explored so as to take full advantage
of hydrophobic nanoporous venting and enable a series of bubble-driven micropumps.
By choosing the best bubble generation approach for an application, it is possible to
address the concerns of this particular application appropriately, such as energy
efficiency, thermal sensitivity, bio-compatibility or flow rate adjustability.
One example of the potential bubble generation approaches is “gas injection”,
which is more generic than boiling or electrolysis because it puts no limit on the nature
and composition of liquid to be pumped. More importantly, injection of an inert gas
does not chemically affect the liquid as heating or electrical fields can. Therefore,
biocompatible gas injection actuation should be highly regarded for the handling of
biomedical liquid samples. A proof-of-concept device is schematized in Figure 4-13.
The configuration of pumping section is similar to that of the electrochemical-bubble-
driven pump, with virtual check valve for directional growth of bubbles and a venting
membrane to vent them out. The gas bubbles are injected from an external nitrogen
gas tank through the gas injection channel. An additional virtual check valve is placed
between the gas injection channel and the pumping section to prevent flooding of the
89
gas injection channel. An external mechanical valve is used to control gas injection
manually. Two larger openings are etched into the pump chip as inlet and outlet
reservoirs. The surfaces of reservoirs are coated with HMDS (contact angle ~80o), so
that the capillary effect of the reservoirs is minimized.
Figure 4-13. Pumping by gas injection-venting.
Before testing, DI water droplet is deposited at the inlet reservoir by pipette.
The hydrophilic microchannel in the pumping section is filled with liquid quickly. The
front meniscus of liquid usually stops at the breather part because the membrane is
hydrophobic. Then the mechanical valve is opened to inject gas into the pumping
section. The gas bubble expands as Figure 4-14.b and c show. At the right-hand side
of this gas bubble, the liquid is pushed to the outlet reservoir. Meanwhile, the left
meniscus of gas bubble is stopped at the inlet virtual check valve. No liquid is pushed
towards the left. After the front meniscus of gas bubble reaches the venting membrane
of the breather, the valve is closed. The bubble starts to shrink immediately because
gas inlet
pressurized gasmechanical valve
virtual check valve
PyrexTM Glass B B
A A
inlet reservoir
outlet reservoir
breatherA-A View
B-B View pumping section
90
the gas is venting out (Figure 4-14.d). Since the hydrophobic membrane of breather
also works as a bubble trap, the bubble is held to the breather part while being vented
(Figure 4-14.e and f). Liquid fills in spontaneously as the left meniscus of bubble
moves toward the right. Once the bubble is completely vented out, the valve is opened
again to start the next pumping circle (Figure 4-14. g). Figure 4-14.i shows that the
outlet reservoir is completely filled with water after 4 pumping circles.
a. t = 0 s
b. t = 0.5s
c. t = 1 s
d. t = 2 s
e. t = 3 s
f. t = 5 s
g. t = 6 s
h. t = 22 s
i. t = 23 s
Figure 4-14. Visualization of pumping effect by gas injection-venting.
expanding bubble
600 μm
breather
close valve
close valve
shrinking bubble
openvalve
open valve
reservoir filled with water
91
Two additional observations confirm that pumping is induced by gas
injection/venting instead of capillary or hydraulic pressure. Firstly, if the mechanical
valve is closed during any time of the pumping phases, the liquid delivery to outlet
reservoir will be stopped accordingly. Secondly, most of the liquid in the inlet
reservoir can be pumped to outlet, leaving the inlet reservoir almost completely dry,
except in the corner (as illustrated in Figure 4-13).
4.5 Summary and Future Directions
Actuation by solid mechanical moving parts in the microscale is challenged by
many issues, such as elaborate fabrication processes, stiction problem, long-term
reliability and large-scale integration. Accordingly, actuation without mechanical
moving parts, such as bubble-driven actuation, has attracted significant interest in
MEMS. The interest has further been amplified by the demands to handle biomedical
liquid samples in micro total analysis systems (μTAS).
Thermal generation of vapor bubbles (i.e., boiling) has been commonly used
in bubble-driven micropumps due to its simple structure (electrical heater) and
convenient bubble reduction (natural condensation). However, several drawbacks limit
the application of thermal bubble actuation in microfluidic devices, such as huge
energy consumption, difficulty in controlling the bubble growth rate, denaturization of
large biological molecules (e.g. DNA and protein) by overheating and the heat
dissipation trade-off between condensation and boiling.
Hydrophobic nanoporous venting provides a rapid universal gas removal
mechanism and enables virtually any bubble generation method to be employed in
92
micropumping, such as electrolysis, injection, chemical reaction, ultrasonic cavitation
and those to be developed for specific applications.
By combining a virtual check valve, bubble capturing and hydrophobic
nanoporous venting, a new paradigm of micropump is developed. Electrolysis is
chosen as the first bubble generation approach to implement this category of
micropumps.
In order to understand the potential performance improvement by switching the
bubble source from boiling to electrolysis, the two bubble-driven actuation approaches
are studied comparatively. Compared with boiling in a similar setup, electrolysis
improves actuation power efficiency by 100-1000 times, while exhibiting better
controllability, bio-compatibility and miniaturization potential.
Electrolysis of water is employed as the bubble source to demonstrate liquid
circulation in a closed-loop. Compared with thermal-bubble-driven micropumps, the
electrochemical-bubble-driven micropump demonstrated in this study achieved similar
volume flow rate with 10-100 times greater power efficiency. The power efficiency
has the potential to be further improved if the electrodes are micromachined and the
distance between them is decreased. The pumping rate is correlated to electrical
current, which enables a micropump with variable pumping rate, directly controlled by
current input. Electrolysis also shows better bio-compatibility compared to boiling,
making the pump more promising for lab-on-a-chip applications. A preliminary
investigation has also proved the concept of a gas injection-venting micropump with
even less impact on biomedical liquid samples. This suggests that other bubble sources
are also able to be incorporated into the pumping concept to optimize the performance
of microfluidic devices.
93
Further investigation is necessary to integrate the gas injection-venting
micropump into μTAS. The recent development of on-chip gas generators [43, 44]
make it possible to eliminate the external gas tank. By fully integrating the gas bubble
generators with a pumping section, a bio-compatible micropumping approach is very
promising for lab-on-a-chip.
Hydrophobic nanoporous venting itself can also be employed in bubble-driven
micropumps with mechanical valves [12] to provide better short-term performances
(e.g. pressure head) than valve-less micropumps. By replacing boiling with other
bubble generation approaches, all the corresponding benefits can be applied to bubble-
driven mechanical micropumps, just as they do for valve-less micropumps.
References
[1] F.-G. Tseng, C.-J. Kim, and C.-M. Ho, "A High Resolution High Frequency
Monolithic Top-Shooting Microinjector Free of Satellite Drops: Part I. Concept,
Design and Model," Journal of Microelectromechanical Systems, 2002, vol. 11, pp.
427-36.
[2] F.-G. Tseng, C.-J. Kim, and C.-M. Ho, "A High Resolution High Frequency
Monolithic Top-Shooting Microinjector Free of Satellite Drops: Part II:
Fabrication, Implementation, and Characterization," Journal of
Microelectromechanical Systems, 2002, vol. 11, pp. 437-47.
[3] K. Hoshino, T. Nagai, Y. Mita, M. Sugiyama, K. Matsumoto, and I. Shimoyama,
"Active Ink-Jet Nozzles Equipped with Arrayed Visual Sensors for Parallel
94
Alignment Control," Proc. The 18th IEEE Int. Conf. on Micro Electro Mechanical
Systems, Miami, FL, Jan., 2005, pp. 48-51.
[4] H. Ago, K. Murata, M. Yumura, J. Yotani, and S. Uemura, "Ink-Jet Printing of
Nanoparticle Catalyst for Site-Selective Carbon Nanotube Growth," Applied
Physics Letters, 2003, vol. 82, pp. 811-3.
[5] T. J. Yen, N. Fang, X. Zhang, G. Q. Lu, and C. Y. Wang, "A Micro Methanol Fuel
Cell Operating at Near Room Temperature," Applied Physics Letters, 2003, vol. 83,
pp. 4056-8.
[6] S. K. Chaudhuri and D. R. Lovley, "Electricity Generation by Direct Oxidation of
Glucose in Mediatorless Microbial Fuel Cells," Nature Biotechnology, 2003, vol.
21, pp. 1229-32.
[7] M. Chiao, K. B. Lam, and L. Lin, "Micromachined Microbial Fuel Cells," Proc.
The 16th IEEE Int. Conf. on Micro Electro Mechanical Systems, Kyoto, Japan,
Jan., 2003, pp. 383-6.
[8] J. S. Boland, J. D. M. Messenger, H. W. Lo, and Y.-C. Tai, "Arrayed Liquid Rotor
Electret Power Generator Systems," Proc. The 18th IEEE Int. Conf. on Micro
Electro Mechanical Systems, Miami, FL, Jan., 2005, pp. 618-21.
[9] D. R. Meldrum and M. R. Holl, "Microfluidics: Microscale Bioanalytical
Systems," Science, 2002, vol. 297, pp. 1197-8.
[10] P.-A. Auroux, D. Iossifidis, D. R. Reyes, and A. Manz, "Micro Total Analysis
Systems. 2. Analytical Standard Operations and Applications," Analytical
Chemistry, 2002, vol. 74, pp. 2637 -52.
95
[11] S. Shoji and M. Esashi, "Microflow Devices and Systems," Journal of
Micromechanics and Microengineering, 1994, vol. 4, pp. 157-71.
[12] S. Zimmermann, J. A. Frank, D. Liepmann, and A. P. Pisano, "A Planar
Micropump Utilizing Thermopneumatic Actuation and in-Plane Flap Valves,"
Proc. The 17th IEEE Int. Conf. on Micro Electro Mechanical Systems, Maastricht,
The Netherlands, 2004, pp. 462- 5.
[13] J. Xie, J. Shih, Q. Lin, B. Yang, and Y.-C. Tai, "Surface Micromachined
Electrostatically Actuated Micro Peristaltic Pump," Lab on a Chip, 2004, vol. 4,
pp. 495-501.
[14] D. J. Laser and J. G. Santiago, "A Review of Micropumps," Journal of
Micromechanics and Microengineering, 2004, vol. 14, pp. R35-R64.
[15] N.-T. Nguyen, X. Huang, and T. K. Chuan, "MEMS-Micropumps: A Review,"
ASME Journal of Fluids Engineering, 2002, vol. 124, pp. 384-92.
[16] J. A. Frank and A. P. Pisano, "Low-Leakage Micro Gate Valves," Tech. Dig. The
12th Int. Conf. on Solid-State Sensors, Actuators and Microsystems, Boston, MA,
Jun., 2003.
[17] T. K. Jun and C.-J. Kim, "Valveless Pumping Using Traversing Vapor Bubbles in
Microchannels," Journal of Applied Physics, 1998, vol. 83, pp. 5658-64.
[18] J.-H. Tsai and L. Lin, "A Thermal-Bubble-Actuated Micronozzle-Diffuser
Pump," Journal of Microelectromechanical Systems, 2002, vol. 11, pp. 665-71.
96
[19] X. Geng, H. Yuan, H. N. Oguz, and A. Prosperetti, "Bubble-Based Micropump
for Electrically Conducting Liquids," Journal of Micromechanics and
Microengineering, 2001, vol. 11, pp. 270-6.
[20] C.-J. Kim, "MEMS Devices Based on The Use of Surface Tension," Proc. Int.
Semiconductor Device Research Symp., Charlottesville, VA, Dec., 1999, pp. 481-4.
[21] M. E. Steinke and S. G. Kandlikar, "An Experimental Investigation of Flow
Boiling Characteristics of Water in Parallel Microchannels," ASME Journal of
Heat Transfer, 2004, vol. 126, pp. 518-26.
[22] L. Lin, "Microscale Thermal Bubble Formation: Thermophysical Phenomena and
Applications," Microscale Thermophysical Engineering, 1998, vol. 2, pp. 71-85.
[23] A. P. Papavasiliou, A. P. Pisano, and D. Liepmann, "High-Speed and Bi-Stable
Electrolysis-Bubble Actuated Gate Valves," Tech. Dig. The11th Int. Conf. on
Solid-State Sensors, Actuators and Microsystems, Munich, Germany, Jun., 2001,
pp. 940-3.
[24] H. Suzuki and R. Yoneyama, "Integrated Microfluidic System with
Electrochemically Actuated on-Chip Pumps and Valves," Sensors and Actuators B,
2003, vol. 96, pp. 38-45.
[25] N. R. Tas, J. W. Berenschot, T. S. J. Lammerink, M. Elwenspoek, and A. v. d.
Berg, "Nanofluidic Bubble Pump Using Surface Tension Directed Gas Injection,"
Analytical Chemistry, 2001, vol. 74, pp. 2224-8.
97
[26] J.-H. Tsai and L. Lin, "Active Microfluidic Mixer and Gas Bubble Filter Driven
by Thermal Bubble Micropump," Sensors and Actuators A, 2002, vol. 97-98, pp.
665-71.
[27] S. Debesset, C. J. Hayden, C. Dalton, J. C. T. Eijkel, and A. Manz, "An AC
Electroosmotic Micropump for Circular Chromatographic Applications," Lab on a
Chip, 2004, vol. 4, pp. 396-400.
[28] A. P. Papavasiliou, D. Liepmann, and A. P. Pisano, "Fabrication of a Free
Floating Silicon Gate Valve," Proc. 1999 ASME Int. Mechanical Engineering
Congress and Exposition, Anaheim, CA, 1999, pp. 435-40.
[29] C.-T. Ho and C.-H. Liu, "Micro T-Switches For Cell Sorting Applications," 2004
ASME Int. Mechanical Engineering Congress and Exposition, Anaheim, CA, Nov.,
2004, CD: IMECE 2004-61427.
[30] F. P. Incropera and D. P. DeWitt, "Fundermentals of Heat and Mass Transfer",
Fifth: John Wiley&Sons, 2002.
[31] F. T. Brown, "Potential Building Blocks for Microhydraulic Actuators," the 1993
ASME winter annual meeting (micromechanical systems 1993), New Orlenas, LA,
1993, pp. 21-33.
[32] M. J. Jensen, G. Goranović, and H. Bruus, "The Clogging Pressure of
Bubbles in Hydrophilic Microchannel Contractions," Journal of Micromechanics
and Microengineering, 2004, vol. 14, pp. 876-83.
[33] U.-C. Yi and C.-J. C. Kim, "Soft Printing of Droplets Pre-Metered by
Electrowetting," Sensors and Actuators A, 2004, vol. 114, pp. 347-54.
98
[34] P. Gravesen, J. Branebjerg, and O. S. Jensen, "Microfluidics - A Review,"
Journal of Micromechanics and Microengineering, 1993, vol. 3, pp. 168-82.
[35] D.-S. Meng, J. Kim, and C.-J. Kim, "A Distributed Gas Breather for the Micro
Direct Methanol Fuel Cell," Proc. The 16th IEEE Int. Conf. on Micro Electro
Mechanical Systems, Kyoto, Japan, Jan., 2003, pp. 534-7.
[36] M. J. Jensen, "Bubbles in Microchannels," M.Sc. Thesis, Technical University of
Denmark (DTU), 2002
[37] A. Kawahara, P. M.-Y. Chung, and M. Kawaji, "Investigation of Two-Phase
Flow Pattern, Void Fraction and Pressure Drop in a Microchannel," International
Journal of Multiphase Flow, 2002, vol. 28, pp. 1411-35.
[38] A. Olsson, P. Enoksson, G. r. Stemme, and E. Stemme, "A Valve-Less Planar
Pump Isotropically Etched in Silicon," Journal of Micromechanics and
Microengineering, 1996, vol. 6, pp. 87-91.
[39] I. D. Johnston, M. C. Tracey, J. B. Davis, and C. K. L. Tan, "Microfluidic Solid
Phase Suspension Transport with an Elastomer-Based, Single Piezo-Actuator,
Micro Throttle Pump," Lab on a Chip, 2005, vol. 5, pp. 318-25.
[40] M. Raffel, C. Willert, and J. Kompenhans, "Particle Image Velocimetry":
Springer-Verlag Berlin, 1998.
[41] R. F. Probstein, "Hysiochemical Hydrodynamics - An Introduction", 2nd ed.:
Wiley-Interscience Publication, 1994.
99
[42] M.-C. M. Lee and M. C. Wu, "3D Silicon Transformation Using Hydrogen
Annealing," Tech. Dig. Solid-State Sensor, Actuator and Microsystems Workshop,
Hilton Head Island, SC, Jun., 2004, pp. 19-22.
[43] C.-C. Hong, S. Murugesan, S. Kim, G. Beaucage, J.-W. Choi, and C. H. Ahn, "A
Functional On-Chip Pressure Generator Using Solid Chemical Propellant for
Disposable Lab-on-a-Chip," Lab on a Chip, 2003, vol. 3, pp. 281 - 6.
[44] Y. H. Choi, S. U. Son, and S. S. Lee, "A Micropump Operating with Chemically
Produced Oxygen Gas," Sensors and Actuators A, 2004, vol. 111, pp. 8-13.
100
CHAPTER 5
CONCLUSION AND OUTLOOK
6.1 Conclusion
Due to their unique properties, microscopic gas bubbles can serve as
distinctive sensors and actuators. However, reliable manipulation is necessary to
eventually implement these functional gas bubbles in microfluidc devices. Surface
tension dominates over other forces in microscale and can be controlled chemically,
thermally or electrically. The strong and controllable surface tension is therefore
decided as the force to manipulate microscopic gas bubbles.
Bubble immobilization is studied first to provide the basics for more complex
manipulations. The energetically favorable sites on a surface are named as bubble-
traps and quantitatively evaluated by Φbc (bubble capturing potential). This definition
eliminates the effect of the absolute value of surface tension and bubble size. The
bubble capturing potential can thus be viewed as a property of the surface, determined
by surface topology and contact angle. Simulation suggests a distinct performance
enhancement with a concave hydrophobic structure (type II bubble-traps by
hydrophobic conic pits). The experimental results support the theoretical predictions.
Hydrophobic venting technology is developed to solve the bubble clogging
problem for microfluidic devices. For a microchamber, it is ideal to collect the gas
bubbles with bubble traps first and then release them through hydrophobic capillaries.
The concept is proven by using 50μm silicon micromachined venting holes coated
with Teflon®. It is then decided that the size of the venting holes has to be decreased,
101
and the uniformity of the hydrophobic layer needs to be improved, so that the leakage
onset pressure can be increased to a practical level for microfluidic devices.
Commercially available hydrophobic nanoporous membranes are found to be a
solution without delicate microfabrication processes. Benefitting from the
intrinsically hydrophobic material and submicron pore size, the porous polypropylene
membrane is demonstrated to be able to successfully vent out gas bubbles from liquid
environment with 35psi or higher pressure tolerance.
Gas permeable microchannels are demonstrated to remove gas bubbles from
two-phase flows of both DI water/gas and 10M methanol/gas. Unique venting
threshold phenomena are found for methanol/gas two-phase flow in a gas permeable
microchannel. The bubble train introduced by the venting threshold can be shortened
by proper conditions, e.g., slower flow. Further investigation is expected to clarify this
interesting phenomenon and minimize the flow resistance of two-phase flow inside the
gas permeable microchannels.
Bubble-capturing breathers and gas permeable microchannels can find its
applications in many microfluidic devices, especially the micro fuel cell. Significant
improvement in reliability and power efficiency can be anticipated by keeping these
microreactors free from the effect of gas bubbles, which can be either introduced
accidentally or generated intrinsically.
Combined with virtual check valves and bubble capturing, hydrophobic
nanoporous venting has enabled a new category of micropumping approaches. The gas
sources of bubble-driven micropumps are broadened to virtually any bubble
generation method, such as electrolysis, injection, chemical reaction, ultrasonic
cavitation and those to be developed for specific applications. The microfluidic
devices will benefit from the flexibility of the bubble generation approach, considering
102
the rapidly increasing interest in the miniaturization of both bio/chemical analysis and
power generation systems. Furthermore, the commercialization of existing
microfluidic devices will also bring more challenges to its pumping approaches. More
available bubble actuation methods would offer more choices to provide products with
better performance and lower prices.
A comparative study is conducted to understand the potential performance
improvement by employing electrochemical bubble-actuation to replace boiling, the
traditionally common approach. Both the theoretical model and experimental results
confirm that electrolysis improves actuation power efficiency by 100-1000 times
while exhibiting better controllability, bio-compatibility and miniaturization potential.
Known to be a bubble-driven actuation approach with so many advantages,
electrolysis is employed as the bubble source to demonstrate liquid circulation in a
closed-loop. Compared with thermal-bubble-driven micropumps, the electrochemical-
bubble-driven micropump in this study achieved similar volume flow rate with 10-100
times higher power efficiency. The pumping rate is also able to be controlled easily by
the input current. Better bio-compatibility over boiling makes the pump more suitable
for biomedical applications. A preliminary investigation has also proved the concept
of a gas injection-venting micropump with even less impact on biomedical liquid
samples. This suggests that other bubble sources are also able to be incorporated into
the pumping concept to optimize the performance of microfluidic devices. By
breaking the limits on bubble sources, a new category of micropumps is enabled by
hydrophobic nanoporous venting.
103
6.2 Outlooks
Considering the unique property and versatile applications of microscopic gas
bubbles, this study is only at the preliminary stage of bubble-powered MEMS. Further
research can be pursued in the following directions.
6.2.1 Bubble-Powered μTAS
Handling biomedical liquid samples properly is a constant concern of μTAS.
Biocompatible micropumping by injecting inert gas [1] represents a trend of future
continuous flow control elements of μTAS. Integrated with an on-chip gas generator,
the gas injection-venting pump proposed in this study can set an example for those
flow control elements. Electrolysis can also be expected to play a more important role
in bubble-driven microactuations, due to its simple fabrication, low power
consumption and good controllability.
Microscopic gas bubbles may also find applications in microfluidic mixing [2],
another challenging task for μTAS. Extra low Reynolds numbers and laminar flow is
common in μTAS [3]. Since molecular diffusion in a micro configuration is too slow
for biochemical reactions, active disturbances are usually necessary to speed up
mixing. Thermal bubble generation/collapse has been demonstrated to introduce
effective perturbations and promote mixing [4]. Other bubble-driven mixers can be
expected to join this trend.
Bubble-powered microparticle manipulation may also contribute to large-scale
genomic/proteomic analysis. The micropaticles with biochemical probes (such as
DNA or protein) immobilized on them can be recognized and analyzed automatically
by their shapes [5] or metallic barcodes [6, 7]. The bubble array formed by bubble
104
traps intuitively suggest large-scale self-assembly of microparticles by gas bubble.
Active control of surface tension may enable feedback control of individual
microparticles [8] by bubbles or lead to a bubble-powered conveyor.
6.2.2 Innovative Designs of μDMFC
The successful development of gas permeable microchannels has not only
improved the existing μDMFC technologies but also enabled innovative new designs.
For example, a passive fuel delivery mechanism could be achieved if a venting
membrane can be integrated into the sealed anodic microchannel of μDMFC. Bubbles
would then be purged where they are created. The generation/removal of gas bubbles
can facilitate the mixing of fuel in the anodic microchannel and maintain a uniform
methanol concentration even though the channel is relatively long. As the
electrochemical reaction consumes methanol, a concentration gradient from the fuel
reservoir to the chamber would be developed. Methanol can be driven passively from
reservoir to reaction channel by diffusion. In this way, no external pump will be
necessary for this μDMFC design. An ultra-compact μDMFC (Figure 5-1) can be
realized by repeating the fuel stack unit, which is stratified as venting membrane,
anode channel, membrane electrode assembly (or MEA, including electrode layers,
catalyst layers and PEM) and cathode channel. In order to achieve a practical μDMFC
system with this mechanism, several design factors should be considered and balanced
well, such as the fuel concentration in both reservoir and reaction microchannel, the
connection ports between reservoir and microchannel and proper isolation approach
when the μDMFC is not in working condition.
105
It is also possible to incorporate the pumping mechanism discussed in chapter
4 and control the fuel flow actively.
external pump
fuel cartridge
existing DMFC system
e-
e-
H+
H+
H+
e-
H+
H+
H+
H+
H+
H+ H
+
H+
e-
e-
e-
traditional fuel stack
fuel cartridge
integrated fuel stack
exploration: integrated μDMFC
gas/liquid separator
porous membrane
H+ H+H+
e- e-
cathode channel
anode channel MEA
one unit of fuel stack
external pump
fuel cartridge
existing DMFC system
e-
e-
H+
H+
H+
e-
H+
H+
H+
H+
H+
H+ H
+
H+
e-
e-
e-
traditional fuel stack
fuel cartridge
integrated fuel stack
exploration: integrated μDMFC
gas/liquid separator
porous membrane
H+ H+H+
e- e-
cathode channel
anode channel MEA
one unit of fuel stack
Figure 5-1. Ultra-compact μDMFC with integrated fuel stack
References
[1] N. R. Tas, J. W. Berenschot, T. S. J. Lammerink, M. Elwenspoek, and A. v. d.
Berg, "Nanofluidic Bubble Pump Using Surface Tension Directed Gas Injection,"
Analytical Chemistry, vol. 74, pp. 2224-8, 2001.
[2] N.-T. Nguyen and Z. Wu, "Micromixers - a Review," Journal of Micromechanics
and Microengineering, pp. R1-R16, 2005.
[3] C.-M. Ho and Y.-C. Tai, "Micro-Electro-Mechanical-Systems (MEMS) and Fluid
Flows," In: Annual Review of Fluid Mechanics, vol. 30, pp. 579-612, 1998.
106
[4] P. Deng, Y.-K. Lee, and P. Cheng, "Design and Characterization of a Micro Single
Bubble Actuator," Tech. Dig. Tech. Dig. 12th Int. Conf. on Solid-State Sensors,
Actuators and Microsystems, Boston, MA, 2003, pp. 647-50.
[5] J. G. F. Tsai, Z. Chen, B. Merriman, S. Nelson, and C.-J. C. Kim3, "Shape-
Encoded Particle For DNA Analysis," Tech. Dig. Tech. Dig. 12th Int. Conf. on
Solid-State Sensors, Actuators and Microsystems, Boston, MA, 2003, pp. 1239-42.
[6] S. R. Nicewarner-Pena, R. G. Freeman, B. D. Reiss, L. He, D. J. Pena, I. D.
Walton, R. Cromer, C. D. Keating, and M. J. Natan, "Submicrometer Metallic
Barcodes," Science, vol. 294, pp. 137-41, 2001.
[7] Z.-L. Zhi, Y. Morita, and E. Tamiya, "Microfabricated Addressable Particles for
Multianalyte Immunoassay," Tech. Dig. Proc. The 16th IEEE Int. Conf. on Micro
Electro Mechanical Systems, Kyoto, Japan, 2003, pp. 441-4.
[8] M. Armani, S. Chaudhary, R. Probst, and B. Shapiro, "Using Feedback Control
and Microfluidics to Steer Individual Particles," Tech. Dig. Proc. The 18th IEEE
Int. Conf. on Micro Electro Mechanical Systems, Miami, FL, 2005, pp. 423-6.
