Study of Small Scale EHD Driven Flow Distribution

Control and Understanding of the Effect of Temperature on

EHD Conduction Pumping Performance

A Dissertation
Submitted to the Faculty
of the
WORCESTER POLYTECHNIC INSTITUTE

In partial fulfillment of the requirements for the

Degree of Doctor of Philosophy
In
Mechanical Engineering

By
Michal Talmor

May 2022

Advisory Committee
Dr. Jamal Yagoobi, Major Advisor
Dr. John Blandino, Committee Member
Dr. Burt Tilley, Committee Member
Dr. Yan Wang, Committee Member
Dr. Pratap Rao, Graduate Committee Representative
 In Memory of Michal Talmor

Michal was due to graduate with her PhD degree under my advising in May 2021. She passed away
suddenly due to medical complications on March 29th, 2021. This loss came as a terrible shock to me and
her lab-mates as well as many classmates, professors, and staff members within the Mechanical and
Materials Engineering Department and at WPI. I extend my deepest sympathy to Michal’s husband and
her family and friends.

Michal was a very bright and inspirational individual. It was an honor for me to be her PhD advisor. Her
understanding of the fundamentals of thermal sciences was very solid. As shown in Michal’s dissertation,
her PhD work dealt with electrohydrodynamic (EHD) pumping of dielectric fluids in micro- and macro-
scale. One of her major contributions to the field was to fundamentally describe the impact of temperature
on the performance of EHD conduction pumping. This understanding is very critical for the design of EHD
pumps to effectively function in the outer space environment. Michal’s research resulted in eight journal
publications (including three to be submitted), one book chapter, and eight conference papers.

Michal was a great role model to many PhD and undergraduate students at WPI. Her dream was to help
lead efforts toward establishing a more permanent human presence in space. In the final year of her PhD,
Michal joined Aurora Flight Sciences Corporation, a Boeing Company, working on innovative technologies
for the aerospace industry, primarily with a focus on space applications.

Several of my former and current PhD students who worked closely with Michal have shared the following
memory quotes.

Jamal Yagoobi

Dr. Yasmin Khakpour: Michal will be always remembered as a special bright individual with great
passion for learning, understanding the unknown and transferring her knowledge to others. We will always
think of her as a kind heart and social soul.

Dr. Viral Patel: I met Michal in 2012 and worked alongside her for 3 years. During this time, I was
meant to mentor her as a senior lab member, but I also learned a great deal from her. She was so kind,
intelligent, funny and passionate about science. She was always willing to help her fellow students and
researchers and come up with solutions to problems together. It truly was a privilege to work with her and
to call her my friend, and I will always remember her.
Dr. Mengqiao Yang: Michal was extremely thoughtful and friendly, and her devotion to family and science
was admirable. She was nice and knowledgeable. I still remember the days she introduced WPI and the

i
surroundings to me when I came to the lab at the first time. Whenever I needed help, I know she would be
the person to rely on. My deepest sympathies for losing of Michal and I will always cherish the memories I
have of her.
Alexander Castaneda: Michal was the first person I began to work with when I first started the PhD
program. The first thing I noticed was how kind and helpful she was, especially when helping me learn
about the field of electrohydrodynamics and how the lab operates. Throughout my first few years at WPI
she always was ready to help and offer assistance whenever needed. The other thing I noticed was how
bright and sharp she was. I always got a clear and concise answer to my questions, which made working
so much easier and comfortable. She is greatly missed, and is one of the best colleagues I have had the
pleasure of working with at WPI.
Zahra Noori O’Connor: During the short time of getting to know Michal in the beginning of my PhD, I
found her a very ambitious and strong woman who calls herself a nerd. Her self-confidence and her
knowledge about the research she was doing was always inspiring. Something always admirable was that
she was always ready to give a professional speech about her research. I remember her saying I had to
work so hard to get to where I am but I fight so our daughters don't have to. She has a special place in the
hearts that have been touched by her and she will be missed.

Nathaniel O’Connor: Michal Talmor was an excellent mentor and friend when I joined the Multiscale
Heat Transfer Lab in 2018. I worked closely with her at the beginning of my Ph.D. and she taught me many
of the fundamental experimental skills I needed to succeed as a graduate student. Most importantly, she
taught me how to fundamentally approach a problem, identify shortcomings or issues, and how to seek a
solution. The knowledge she imparted on me will remain with me for the rest of my life and I am forever
grateful for the time I was able to spend working with her.

ii
 Acknowledgments

Michal Talmor PhD research was sponsored by the prestigious NASA Space Technology Research
Fellowship. In her absence, I would like to thank Professor Christophe Louste of Institut PPRIME,
Département Fluide-Thermique-Combustion, CNRS-University of Poitiers-ESMA, France and Professor
P.A. Vazquez of Departamento de Física Aplicada III, Escuela Técnica Superior de Ingeniería, Universidad
de Sevilla, Spain, for their collaborations with Michal and I, pertinent to her dissertation. I would also like
to thank Michal’s PhD advisory committee members, Professors John Blandino, Pratap Rao, Burt Tilley,
and Yan Wang for their support.

Without any doubt, Michal would have utterly thanked her devoted husband, Alexandria Tilley, for all the
priceless support that she received from him during her PhD study at WPI. Of course, she would have
acknowledged the love of her parents, Ron and Shoshana (Rosie) Talmor, her in-laws, Karen and
Joe Tilley, and her cat Chicha as well.

iii
 Abstract of the Dissertation

Study of Small Scale EHD Driven Flow Distribution Control and Understanding of the Effect of
Temperature on EHD Conduction Pumping Performance
by
Michal Talmor
Ph.D. Candidate
Mechanical and Materials Engineering Department
Worcester Polytechnic Institute, 2022
Research Advisor: Dr. Jamal Yagoobi

Electrohydrodynamic (EHD) conduction pumping technology offers a unique way to control flow
distribution in multi-scale environments. In EHD conduction pumping, the interaction between an applied
electrical field and dissociated electrolyte species in a dielectric fluid generates a net body force within the
fluid resulting in a net flow in the desired direction. EHD conduction pumps have remarkable potential due
to their lack of moving parts, simple designs, light weight, low power consumption, and ability to operate
in microgravity. The suitability of these pumps increases at small scales and they have been previously
proven effective for heat transfer enhancement, with possible applications in electronics cooling and more,
both terrestrially and in space.

In this study, the single-phase flow distribution control with EHD conduction pumps between two parallel
micro-channels is experimentally investigated. In EHD conduction pumping, a strong electric field is
applied via asymmetric submerged electrodes in a dielectric liquid. The field enhances the dissociation of
electrolytic impurities present within the fluid, generating ions that migrate to form heterocharge layers
over each electrode. This study numerically investigates the heterocharge layer morphology of EHD
conduction pumping used in flow distribution control between parallel micro-scale branches, using different
pumping orientations and accounting for flow inertia effects. The results are qualitatively compared with
available experimental data and serve to explain observed behaviors.

Most experiments performed using EHD conduction pumps have focused on global flow rate and pressure
generation measurements, but have not provided measurements of the actual flow velocity fields generated
by these pumps. While these flow velocity fields can be simulated numerically, both qualitative flow

iv
visualization and quantitative measurements of the true flow velocity vectors are difficult to accomplish for
EHD conduction due to the presence of the strong electric field. Few studies have therefore attempted to
perform any kind of flow visualization of EHD conduction pumping in general, and fewer still offered
velocity measurements for these devices. This study provides a comprehensive set of measurements of the
flow velocity fields generated by a multi-electrode EHD conduction pump, measured using particle imaging
velocimetry, with unique insulating particles as the visualization elements. These measurements are taken
for several flow conditions that provide insight into the effect of external flow inertia on the EHD
conduction pumping mechanism - static pressure generation with no net flow through the loop, open flow
through the loop, and in the presence of externally applied flows supplied by a mechanical pump.

Previous experimental works have shown that, in some cases, increasing the temperature of the fluid
decreases the performance of EHD conduction pumps, while in other cases it is the opposite. In this study,
these seemingly contradictory behaviors are explained by taking into account the different working regime
(i.e., ohmic regime and saturation regime) of the EHD conduction pumps. Specifically, the effect of
temperature on the pressure generation performance of EHD conduction pumping is fundamentally
investigated and described based on the corresponding regime. In the ohmic regime there is always two
heterocharge layers next to the electrodes and an electroneutral bulk, while in the saturation regime the
heterocharge layers overlap and there is no electroneutral bulk. These considerations are relevant for the
design of EHD conduction pumps, especially in small scales.

v
 Contents
In Memory of Michal Talmor ........................................................................................................................ i
Acknowledgments........................................................................................................................................ iii
Abstract of the Dissertation ......................................................................................................................... iv
List of Figures ............................................................................................................................................ viii
List of Tables ............................................................................................................................................... xi
Chapter 1 – Introduction ............................................................................................................................... 1
Reference .................................................................................................................................................. 1
Chapter 2 - PIV Flow Field Measurements of Electrohydrodynamic Conduction Pumping ........................ 2
2.1. Introduction ........................................................................................................................................ 2
2.2. Experimental Setup ............................................................................................................................ 7
2.3. Particle Imaging Velocimetry .......................................................................................................... 10
2.3.1 Results & Discussion ................................................................................................................. 12
2.4. Conclusion ....................................................................................................................................... 23
References ............................................................................................................................................... 23
Chapter 3 - Flow Distribution Control in Micro-scale via Electrohydrodynamic Conduction Pumping ... 26
3.1. Introduction ...................................................................................................................................... 26
3.2. Experimental Setup .......................................................................................................................... 31
3.3. Results and Discussion .................................................................................................................... 33
3.4. Conclusion ....................................................................................................................................... 39
References ............................................................................................................................................... 39
Chapter 4 - Numerical Study of Micro-scale EHD Conduction Pumping: The Effect of Pump Orientation
and Flow Inertia on Heterocharge Layer Morphology and Flow Distribution Control .............................. 41
4.1. Introduction ...................................................................................................................................... 41
4.2. Mathematical Model ........................................................................................................................ 45
4.3. Domains and Boundary Conditions ................................................................................................. 51
4.4. Simulation Cases ............................................................................................................................ 55
4.5. Results and Discussion .................................................................................................................. 58
4.5.1 Flow redistribution – forward and reverse orientation ............................................................... 59
4.5.2 Maldistribution recovery – forward and reverse orientation ...................................................... 65
4.5.3 Heterocharge layer morphology................................................................................................. 67
4.5.4 Qualitative comparison with experimental trends...................................................................... 73
4.6. Conclusions ...................................................................................................................................... 74
Nomenclature Table ................................................................................................................................ 77

vi
 References ............................................................................................................................................... 77
Chapter 5 - Influence of Temperature on the Performance of Electrohydrodynamic Conduction Pumps . 80
5.1. Introduction ...................................................................................................................................... 80
5.2. Experimental Results ....................................................................................................................... 83
5.2.1 Experimental Study .................................................................................................................... 83
5.3. Physical Model and Conduction Regimes ....................................................................................... 89
5.3.1 Physical Model........................................................................................................................... 90
5.3.2 Analysis of the Temperature Equation ....................................................................................... 94
5.3.3 Conduction Regimes .................................................................................................................. 96
5.4. Dependence of the Generated Pressure on the Temperature.......................................................... 105
5.4.1 Influence of Temperature in the Dissociation-Recombination Process ................................... 105
5.4.2 Comparison of Theoretical Model with Experimental Results ................................................ 110
5.5. Conclusions .................................................................................................................................... 114
References ............................................................................................................................................. 117
Appendix ................................................................................................................................................... 119
List of Publications ................................................................................................................................... 119

vii
 List of Figures
Figure 2.1 EHD Conduction Pumping Mechanism ...................................................................................... 3
Figure 2.2 EHD Conduction Pumping Dye Visualization [19] .................................................................... 5
Figure 2.3 Simulated EHD Conduction Velocity Streamlines [20] .............................................................. 6
Figure 2.4 Experimental Setup Schematic .................................................................................................... 7
Figure 2.5 Electrode Configuration Schematic ............................................................................................. 8
Figure 2.6 Test Section with EHD Conduction Pump .................................................................................. 9
Figure 2.7 PIV Measurement Schematic and LaVision PIV System.......................................................... 10
Figure 2.8 All Cases at 20kV, Flow Waves ................................................................................................ 13
Figure 2.9 Flow Fields, EHD Conduction Pump at 2kV ............................................................................ 14
Figure 2.10 Flow Fields, EHD Conduction Pump at 8kV .......................................................................... 14
Figure 2.11 Flow Fields, EHD Conduction Pump at 20kV ........................................................................ 15
Figure 2.12 Primary and Secondary Vortices for (a) Dye Visualization [19], (b) PIV Measurements ...... 16
Figure 2.13 Vortex Directions for Electrodes with Reversed Polarities, (a) numerical [20], (b)
experimental [19] ........................................................................................................................................ 17
Figure 2.14 Static Case Maximum Velocities ............................................................................................ 20
Figure 2.15 Dynamic Case Maximum Velocities ....................................................................................... 21
Figure 2.16 Opposed Flow Case Maximum Velocities .............................................................................. 22
Figure 3.1 EHD Conduction Pumping Mechanism .................................................................................... 28
Figure 3.2 Meso-scale Flow Rate Redistribution with Initially equal 1.5 mL/min Distribution [12] ......... 30
Figure 3.3 EHD Conduction Pumping Mechanism [12] ............................................................................. 30
Figure 3.4 Schematic of Parallel Microchannel Experimental Setup ......................................................... 31
Figure 3.5 Distribution Control EHD Conduction Pump [13] .................................................................... 32
Figure 3.6 Flow Starvation in Inactive Branch, with 42 kg/m2s Initially Equal Distribution .................... 34
Figure 3.7 Pressure Drop and Current Load for 1.35 mL/min Initially Equal Distribution ........................ 34
Figure 3.8 Flow Redistribution, with 2.1 mL/min Initially Equal Distribution .......................................... 35
Figure 3.9 Pressure Drop and Current Load for 2.1 mL/min Initially Equal Distribution .......................... 36
Figure 3.10 Flow Redistribution, with 4.5 mL/min Initially Equal Distribution ........................................ 37
Figure 3.11 Pressure Drop and Current Load for 4.5 mL/min Initially Equal Distribution ........................ 37
Figure 4.1 Schematic of EHD Conduction Mechanism .............................................................................. 43
Figure 4.2 Isothermal Panel, which Integrates Fluid Channels into the Rib Structure and Face-sheet [18]
.................................................................................................................................................................... 45
Figure 4.3 Simulation Domains for the Fluid (filled shape) and Walls (patterned shapes) - Not to Scale 52

viii
Figure 4.4 EHD Pump Electrodes and Boundary Conditions in One Branch - Not to Scale...................... 54
Figure 4.5 The “forward” (a) Co-directional Configuration, vs. the “reverse” (b) Configuration ............. 56
Figure 4.6 Overview of the Simulation Domain for (a) Flow Redistribution, (b) Maldistribution Recovery
.................................................................................................................................................................... 57
Figure 4.7 Redistribution of Initially Equal (a) 0.5 mL/min, (b) 1 mL/min, (c) 2 mL/min, Forward
Orientation .................................................................................................................................................. 59
Figure 4.8 : Flow Fields of Forward Orientation Flow Redistribution, (a) with and (b) without
Recirculation ............................................................................................................................................... 61
Figure 4.9 Forward vs. Reverse Configurations – Redistribution of (a) 0.5 L/min, (b) 1 mL/min, (c)
2mL/min...................................................................................................................................................... 62
Figure 4.10 Flow Velocity Fields for the Forward (a,d) and Reverse (c,d) Configurations, EHD Pump at
1.5 kV.......................................................................................................................................................... 64
Figure 4.11 Total x-direction Force Generated by Forward and Reverse Pumping Orientations ............... 65
Figure 4.12 Forward vs. Reverse Orientation Maldistribution Recovery of (a) 1 mL/min, (b) 2 mL/min,
(c) 2 mL/min with a Wider Separation, and (d) 4 mL/min Incoming Flow Rates ...................................... 66
Figure 4.13 Net Space Charge Surface Plot (background coloring), Overlayed with Outlines of each
Individual Ionic Species’ Maximum Travel Distance, λ(x), and Adjusted λs for the Heterocharge Layers
.................................................................................................................................................................... 69
Figure 4.14 Heterocharge Layers for a Forward Configuration Pump Activated at 1500V for 1 mL/min (a)
and 8 mL/min (b) Inlet Flow Rates ............................................................................................................. 70
Figure 4.15 Heterocharge layers for a Reverse Configuration Pump Activated at 1500V for 1 mL/min (a)
and 8 mL/min (b) Inlet Flow Rates ............................................................................................................. 70
Figure 4.16 Heterocharge Layer Widths over the (a) First and (b) Second Electrode Pairs ....................... 72
Figure 4.17 Qualitative Comparison with Flow Redistribution Experimental Data from [23] for 0.5
mL/min (a) and 2 mL/min (b) ..................................................................................................................... 73
Figure 5.1 Illustration of the EHD conduction mechanism for a cylindrical EHD conduction pump. The
HV electrode allows the liquid to go through. The heterocharge layers locations are indicated in blue and
red. The corresponding polarity w/ respect to the electrode ....................................................................... 81
Figure 5.2 : Lateral view of a EHD electrode pair arrangement. The liquid flows from left to right [25]. 84
Figure 5.3 Top view of the individual EHD pump components. The liquid flows into the page [25]. ...... 86
Figure 5.4 Schematic of the Experimental Setup [25]. ............................................................................... 87
Figure 5.5 Comparison of Static Pressure Generation at Different Temperatures [25]. ............................. 88
Figure 5.6 Comparison of Electric Current at Different Temperatures [25]. .............................................. 89
Figure 5.7 Axisymmetric computational domain for the simulations with no fluid motion. The lengths
correspond to the pump described in Section 2.1, scaled with L2. ............................................................. 97
Figure 5.8 The dimensionless electric force along the axis generated in the cylindrical pump vs. C0 for
several values of β assuming that the liquid at rest, and the dependence of the dimensional force vs.
applied electric field. ................................................................................................................................... 98

ix
Figure 5.9 Dimensionless charge distribution for β = 0.1. In the first plot it C0 = 4 (ohmic regime), in the
second plot it is C0 = 0.1 (intermediate regime), in the third plot it is C0 = 0.002 (saturation regime).
Computations performed with the liquid at rest........................................................................................ 101
Figure 5.10 Axisymmetric computational domain for the simulations taking into account the motion of
the liquid. The lengths correspond to the pump described in Section 2.1, scaled with L2. ...................... 102
Figure 5.11 Dimensionless electric force vs. C0 when the liquid motion is considered, for different values
of β. The dots correspond to the electric force computed for the configurations corresponding to the
temperatures used in the experimental results. ......................................................................................... 103
Figure 5.12 This plot shows the net electric charge computed as qt =n+ −n−, and the streamlines for β =
0.1 and three different values of C0. The first plot C0 = 2, ohmic regime. The second plot C0 = 0.1,
intermediary regime. The third plot C0=0.01, saturation regime. ............................................................ 104
Figure 5.13 Computed values of neq and C0 computed from experimental data from Table 5.4. Both
magnitudes are scaled by their values for the first temperature. ............................................................... 110
Figure 5.14 Experimental vs. predicted pressure generation for R-123 with C0 = 3 and β = 3. Solid lines
correspond to a = 8A ̇ ................................................................................................................................ 114
Figure 5.15 Experimental [19] vs. predicted pressure generation for R-134a, with C0 = 0.02 and β = 0.06.
Solidlines correspond to a = 8A ̇ ............................................................................................................... 115

x
 List of Tables

Table 2.1 EHD Conduction Pump Dimensions ............................................................................................ 8

Table 2.2 HFE-7000 Properties [24] ............................................................................................................. 9
Table 3.1 Branch Pumps Electrode Dimensions [13] ................................................................................. 32
Table 3.2 Systematic Measurement Errors ................................................................................................. 32
Table 3.3 HCFC-123 Properties at 25°C [15], [16] .................................................................................... 33
Table 3.4 Comparison of Scale Dependent Mass Flux Redistribution Capabilities at Max. Voltage ........ 38
Table 4.1 EHD Pump Dimensions .............................................................................................................. 54
Table 4.2 Study Cases ................................................................................................................................. 55
Table 4.3 Working Fluid Properties ............................................................................................................ 58
Table 5.1 Component Dimensions for EHD Pump [25]. ............................................................................ 85
Table 5.2 R-123 physical properties at ambient temperature...................................................................... 88
Table 5.3 Typical values of physical properties along with the ratios of typical times to the advection time
in Jeong and Didion [19] (1) and [20] (2). Here, τJ and τv are computed taking ∆T = 1◦C. In both cases cp
≈ 2×103 J/kg·K. .......................................................................................................................................... 96
Table 5.4 R-123 Temperature Variation of Properties.............................................................................. 106

xi
 Chapter 1 – Introduction

In 2018, Talmor and Yagoobi jointly published a long chapter entitled, “Electrohydrodynamically
Augmented Internal Forced Convection” in the Handbook of Thermal Science and Engineering,
referenced below [1]. Thus, it is not necessary to repeat the same in-depth literature review on this
topic in this dissertation.

The objectives of this PhD study are summarized as follows,

Chapter 2: The objective of this study is to provide a comprehensive set of measurements of the
flow velocity fields generated by a multi-electrode EHD conduction pump, measured using particle
imaging velocimetry, with unique insulating particles as the visualization elements.

Chapter 3: The objective of this study is to experimentally investigate the single-phase flow
distribution control with EHD conduction pumps between two parallel micro-channels.

Chapter 4: The objective of this study is to numerically investigate the heterocharge layer
morphology of EHD conduction pumping used in flow distribution control between parallel micro-
scale branches, using different pumping orientations and accounting for flow inertia effects.

Chapter 5: The objective of this study is to fundamentally understand the effect of temperature
on the pressure generation performance of EHD conduction pumping.

Reference
[1]. Talmor, M. and Seyed-Yagoobi, J., “Electrohydrodynamically Augmented Internal Forced Convection”, a
Chapter in Handbook of Thermal Science and Engineering, Edited by Francis A. Kulacki, Springer Publishing
Co., New York, New York, ISBN 978-3-319-28573-3, 2018 (The e-book version of the Handbook of Thermal
Science and Engineering ).

1
 Chapter 2 - PIV Flow Field Measurements of
Electrohydrodynamic Conduction Pumping

Note: This chapter is based on the following technical paper. Talmor conducted particle imaging
velocimetry (PIV) flow field measurement of EHD conduction pumping at the PPRIME Institute,
CNRS-University of Poitiers-ESMA located in France, under the guidance of Professor Christophe Louste.

Talmor, M, Louste, C., and Yagoobi, J., “PIV Flow Field Measurements of Electrohydrodynamic
Conduction Pumping”, Electrostatics Society of America, E-Proceedings, Boston, Massachusetts,
June 2018.

2.1. Introduction
Electrohydrodynamic (EHD) pumping utilizes the interaction between Coulomb body forces,

generated by electric fields applied on fluids, and charges within fluids in order to generate and

control fluid flow. This state-of-the-art technology achieves significant flow motion without

mechanical means or moving parts and can be easily miniaturized. These properties make EHD

pumping advantageous for applications such as electronics cooling and heat transfer enhancement

[1]. Three types of EHD pumping mechanisms have been investigated, known as ion drag

pumping, induction pumping, and conduction pumping. These mechanisms differ from each other

by the method each uses to introduce charges into the working fluid.

Ion drag pumping, or injection pumping, uses a corona discharge from a sharp emitter electrode

in order to inject charges into the fluid [2]. Induction pumping depends on the existence of a

gradient in the electrical conductivity of the fluid, due to an interface between different fluids or

phases, or the presence of temperature gradients. Naturally occurring charges within the fluid

accumulate in the locations of such property gradients and can be moved via a travelling AC

2
electrical field due to the difference between the charge transit time and the charge relaxation time

[3].

Conduction pumping, the mechanism studied in this work, utilizes strong electrical fields in

order to enhance the dissociation reaction of naturally occurring electrolytic impurities within

dielectric liquids. In equilibrium conditions, in the absence of a strong electric field, these naturally

occurring impurities dissociate and recombine at equal rates, as shown in the following equation

[4],

𝐷𝑖𝑠𝑠𝑜𝑐𝑖𝑎𝑡𝑖𝑜𝑛
→
𝐴𝐵 𝐴+ + 𝐵 − (1)
←
𝑅𝑒𝑐𝑜𝑚𝑏𝑖𝑛𝑎𝑡𝑖𝑜𝑛

Applying a strong electric field between two electrodes in contact with the working fluid

increases the dissociation rate exponentially, while the rate of recombination remains relatively

constant. The positive and negative charges formed in this manner are then attracted to the

oppositely charged electrodes and migrate to them due to the electric field. A new equilibrium is

then formed, in which the ions retain their charge as they travel within thin layers near each

oppositely charged electrode, known as the heterocharge layers, while recombining and losing

their charge outside of these layers [5].

Figure 2.1 EHD Conduction Pumping Mechanism

3
 The greater the size of these space charge layers, and the greater the concentration of charge

within them, the greater the Coulomb forces applied on the fluid. Since, for a given electrode pair,

the two layers generate opposing forces on the fluid, if the wetted surface areas of the high voltage

and ground electrodes are the same only mixing will occur. Therefore, EHD conduction pumps

are designed with asymmetric electrodes, such that a net force is applied on the fluid, which creates

a net flow in a desired direction. The EHD conduction pumping mechanism is illustrated in Figure

2.1, for electrodes flush with the wall.

Flow visualization of fluids under the effect of EHD forces has been previously performed

primarily for ion drag pumping of gases, also known as the corona ionic wind. Most of the

techniques used for such flow visualization have focused on exciting inserted fluorescent

molecules via lasers [6-7], the Schlieren photography method for visualization of density changes

[8], or gas phase Particle Imaging Velocimetry (PIV) [9-10].

Additional researchers have performed flow visualization experiments on ion drag pumping in

the liquid phase using the Schlieren method for fluids, which relies on local changes in light

refraction rather than density changes [11], as well as other techniques such as Laser Doppler

Velocimetry (LDV) [12], Laser Doppler Anemometrya (LDA) [13], and liquid phase PIV [14-15].

LDA was also used for velocity measurements of EHD induction pumping of a stratified fluid film

[16]. Other than the PIV and LDV/LDA methods, the other flow visualization techniques yielded

only qualitative flow pattern images, rather than precise measurements of the flow velocities. In

addition, the LDV/LDA method capture single measurements of velocities at discreet pre-

determined points (probe volumes), rather than obtaining the full velocity field all at once. Moving

the laser to measure different probe volumes allows for measurements to be made throughout the

flow field, but it is a long process, with the resolution of the flow field depending on how many

4
probe volumes are investigated. As such, these methods provide an accurate quantitative

measurement of the flow, but at comparatively low spatial resolution.

Studies of two-phase flow visualization in the presence of an electric field have also been

performed for liquid flows driven by interfacial forces exerted by ion drag pumping in a stratified

medium, using the PIV technique [17], or micro PIV technique [18] for thin fluid films. Flow

visualization of EHD conduction pumping either in single or two-phase flows, however, has not

been well studied. The first study appearing in the literature is by Hemayatkhah et al. [19], who

used dye infusion to generate a qualitative flow pattern map, as shown in Figure 2.2.

Figure 2.2 EHD Conduction Pumping Dye Visualization [19]

In this figure, the EHD conduction pump electrode widths are 4mm and 12mm for the ground

and high voltage electrodes, respectively, with a 2mm gap between said electrodes. The image

shows a sub-section of the pump, which contained six electrode pairs. This study showed the

characteristic vortices associated with EHD conduction pumping using flush electrodes, which had

been previously only observed in numerical studies [20], such as in Figure 2.3. This figure shows

the velocity streamlines generated by EHD conduction in a nondimensional simulation.

5
 Figure 2.3 Simulated EHD Conduction Velocity Streamlines [20]

Some studies using PIV flow field measurements have also been performed by Daaboul et al.

[21] on a confined EHD conduction pump inside a rectangular test chamber, and by Chirkov et al.

[22] on an enclosure containing a barrier restriction. In both cases, only a single pair of electrodes

was used, and the flow was circulating inside of an enclosure, without any net generated flow or

flow loop. Lastly, Siddiqui and Seyed-Yagoobi [23] measured mean flow velocities for the overall

flow generation in a liquid film by a multi-electrode EHD conduction pump, at different film

heights. In that study, a rudimentary particle observation system with a strong light and a high

speed camera was used. However, those measurements did not yield the full flow field for the

liquid film.

The purpose of this study is to provide a more comprehensive set of PIV based velocity field

measurements for a multi-electrode EHD conduction pump driving a fluid film inside a flow loop,

in the presence and absence of externally applied flows.

6
2.2. Experimental Setup

Figure 2.4 Experimental Setup Schematic

A schematic of the experimental loop used in this work is shown in Figure 2.5. The 25cm x 13cm

x 5cm test section housed the EHD conduction pump, and had highly polished, clear acrylic on

three sides and a matt black polypropylene back on the fourth side. The laser sheet was projected

into the test vessel through one of the clear acrylic 5cm wide sides, while external shielding placed

against the opposite side prevents undue laser refraction. The high speed camera was facing the

clear acrylic 25cm x 13cm side, and used the black side as a backdrop so that the particles could

be captured clearly. Two valves were used to seal the test section from the rest of the loop. An

ISMATEC BVP-Z model gear pump with a suction-shoe type pumping head able to deliver

102mL/min at 6000rpm was used to both apply an external flow on the EHD conduction pump,

and to recirculate the particles to prevent sedimentation.

7
 The EHD conduction pump assembly design is shown in Figure 2.6. The pump is comprised

of three electrode pairs, designed as interlocking combs, with the electrode widths and spacing

given in Table 2.1.

Table 2.1 EHD Conduction Pump Dimensions

Ground Electrodes
(3mm)
High voltage
Electrodes (9mm)

Ultem base

Spring-loaded connections impinge on these tabs

Figure 2.5 Electrode Configuration Schematic

The electrodes were press-fit into an Ultem substrate, which served as electrical insulation

between the electrodes, structural support for the assembly, and location setting for the electrodes.

The tab shown in Figure 2.6 on each set of electrodes was connected to electrical feedthrough

inputs via spring-loaded connectors to ensure a secure contact. These feedthroughs were inserted

from the bottom of the housing. To ensure the tabs did not significantly affect the generated

electrical field of the pump, and to maintain the minimum distance between all grounded parts and

all parts connected to high voltage, the space between the first and second electrode pairs was set

to 25mm instead of the 15mm quoted in Table 2.1.

8
 Figure 2.6 Test Section with EHD Conduction Pump
Figure 2.7 shows the final assembly, where the pump is inserted into a track in the housing and

pressed in on both sides by the front and back plates.

Power was supplied to the electrodes via a Spellman SL10 power supply with a maximum

voltage and current ratings of 30kV 0.33mA. The working fluid used in this study was the 3M

HFE-7000 engineering fluid. This fluid’s properties are shown in Table 2.2. The film height chosen

for this experiment was 12mm, which represents a sufficient height for PIV measurements to

provide a good resolution of the velocity profile. The EHD conduction pump was tested both in

the “static” case, where the valves isolated it from the rest of the loop, and the “dynamic” case

where flow was allowed to circulate through the loop. In addition, the mechanical pump was used

in order to oppose the EHD conduction pump, by applying an opposing flow rate of 15mL/min,

which translates to an average opposing flow velocity of 0.5 m/s.

Table 2.2 HFE-7000 Properties [24]

9
2.3. Particle Imaging Velocimetry
PIV is a well-established velocity measurement technique. Its working principle is briefly

illustrated in Figure 2.4, alongside the LaVision system used in this study. The LaVision PIV

acquisition system was comprised of a category 4, double pulsed Twins Ultra Nd: YAG laser and

a LaVision Imager Intense high speed CCD digital camera with a spatial resolution of 1376×1040

pixels. The velocity fields were computed via the Lavision Davis 8.0 software package, using

cross-correlation with a 50% overlap for an interrogation window of 16×16 pixels. Measurements

were taken only for steady state flows, defined as allowing 90 seconds for the EHD conduction

pump’s current to reach a steady value. These measurements were then time averaged over a set

of 1000 captures.

Figure 2.7 PIV Measurement Schematic and LaVision PIV System

The reflective seeding particles used were from a PTFE Teflon powder with an average particle

diameter of 200µm. The size and composition for these particles were selected to provide an

accurate representation of the flow in the presence of a strong electric field.

In applications without a strong electric field controlling for the size of the particles is sufficient

for quality measurements. However, due to the presence of the strong electric field in this work, it

was essential to avoid charging of the seeding particles. Charged, the seeding particles would

migrate along the electric field lines, instead of following the fluid flow field [25]. In addition,

charged particles could become attracted to the EHD conduction pump’s electrodes, either
10
obscuring the electrode surface and reducing the pump’s performance, or bridging the gap between

electrodes and causing a spark. Therefore, particles made of an insulating material were chosen

for this study. As an additional measure for avoiding charge effects, the particles were circulated

through the test loop between tests, with the EHD pump turned off, in order to relax any charge

accumulated during each test.

The error for these PIV measurements was computed based on the Correlation Statistics [26]

recommended by a recent comparison of different techniques [27]. Since the measured quantity is

velocity, the error is comprised of a spatial error and a time error. In the case of the former,

considering the camera’s A/D conversion rate of 4𝑒 − and the interrogation window size, the error

due to noise is 0.05px. Assuming a sizeable out of plane motion (OOPM) due to the test chamber’s

dimensions, and considering that the average particle image takes up 3.44 pixels, the spatial error

is compounded by another 0.55px. Adding the error due to random Brownian particle motion,

using an assumed seeding density of 0.2 based on an average of five particles captured in each

interrogation window, the spatial error grows by another 0.05px. The PIV software removes bias

errors from the final results such that the correlation statistics’ peak is symmetrical. Therefore, the

spatial error for a single measurement is dominated by the out of plane motion error:

2 2 2
𝜎𝑠𝑝𝑎𝑡 = √𝜎𝑛𝑜𝑖𝑠𝑒 + 𝜎𝑂𝑂𝑃𝑀 + 𝜎𝐵𝑟𝑜𝑤𝑛𝑖𝑎𝑛 = 0.55𝑝𝑥

Accounting for the inherent variability of the Correlation Statistics’ algorithm, this value can

vary by up to ±32% based on the assumption of five particles per interrogation window. However,

the overall error is tempered by the time averaging of each set of 1000 measurements:

𝜎𝑠𝑝𝑎𝑡_𝑡𝑜𝑡𝑎𝑙 = 1.32 ∗ 𝜎𝑠𝑝𝑎𝑡 /√1000 = 0.023𝑝𝑥

11
 The time error for the type of laser used in this study is generally on the order of 50ns [28],

much lower than the pulse interval range used in this study, which is on the order of milliseconds.

Therefore, the overall velocity uncertainty is dominated by the spatial uncertainty. The reference

length used for this analysis is taken to be one pixel. Therefore, in terms of percentages, the overall

uncertainty is:

𝑢 = ±2.3%

2.3.1 Results & Discussion

Figure 2.8 shows the behavior of the liquid film as captured by the high speed camera, at the

maximum applied voltage of 20kV for all three cases studied: (a) the static case where the test

section was isolated from the loop using the inlet and outlet valves, (b) the dynamic case where

the valves were open and flow was allowed to circulate in the test loop while bypassing the

mechanical pump, and (c) the opposed flow case where the mechanical pump was also connected

to the loop and was actively opposing the EHD conduction pump’s flow. In these images, the net

flow direction is toward the right. The electrode pairs are highlighted using blocks overlaying the

images. The acquisition system sets the X-axis such that the mid-point of the acquisition window

is set as X=0. Figure 2.8c is shifted slightly toward the right, since achieving the opposing flow

required moving the test setup, which necessitated recalibrating the PIV acquisition system, and

its X-axis was shifted similarly to facilitate the comparison with the other cases. In addition, while

the fitting on the left in this figure appears larger than in the other images, that is merely due to

refraction making the fully submerged fittings appear smaller than they truly are, whereas in Figure

2.8c the fitting on the left is only partially submerged.

12
 These images provide a qualitative view of a few of the characteristic differences between

these cases that will be shown quantitatively in the velocity field measurement plots further on in

this section. The static case presented in Figure 2.8a shows a waveform almost like a standing

wave, since the flow is confined to the test section. Whereas in the dynamic case shown in Figure

2.8b the waves appear to be more irregular and there is a subtle increase in the overall wave

amplitude in the direction of flow. This is also seen in Figure 2.8c, which shows a significantly

more choppy set of waves and the increase in amplitude in the positive x direction is more apparent.

(a) Static Case Net Flow Direction

Net Flow Direction

(b) Dynamic Case

Net Flow Direction

(c) Opposed Flow Case

Figure 2.8 All Cases at 20kV, Flow Waves

These qualitative trends are shown quantitatively in Figures 2.9-2.11 which present the full

flow field measurements taken at different applied voltages for each of the three cases, with the

shift for the opposing flow case corrected. It should be noted that Figures 2.9-2.11 do not fully

show the interface region, making it appear as if the surface of the liquid was flat in all cases. As

13
Figure 2.8 shows, that is clearly not the case. However, meaningful velocity data was difficult to

obtain from that region of the flow since the interface motion was not uniform across the thickness

of the chamber, such that waves near the front of the sight window would obscure the offset waves

behind them, where the laser sheet was projected. The flow profiles in Figures 2.9-2.11 have been

cleared of significant outliers resulting from those obstructions, but a fully accurate capture of the

velocity profiles of the constantly moving interface was unfeasible.

Figure 2.9 Flow Fields, EHD Conduction Pump at 2kV

Figure 2.10 Flow Fields, EHD Conduction Pump at 8kV

14
 Figure 2.11 Flow Fields, EHD Conduction Pump at 20kV

The discussion of Figures 2.9-2.11 therefore focuses on the flow fields within the bulk of the

liquid. For the sake of ease of comparison, all of these figures show the flow fields up to the

nominal height of the liquid film, 12mm, regardless of the location or shape of the moving

interface.

Several general trends can be discerned from observing Figures 2.9-2.11. First, the formation

of vortices within the liquid. These vortices originate from the non-uniform electric field between

the electrodes in each electrode pair. This leads to the application of non-uniform Coulomb forces

in both the x and y directions, and subsequently to rotational flow fields. Figure 2.12 shows a

comparison between Figure 2.3 from Hemayatkhah et al [19] and the results shown here. Since

they reported no net flow in their experiment, and the voltage applied was the maximum allowed

for their working fluid, the comparison is made with the static case at 20kV shown in Figure 2.11.

Since the electrodes in Figure 2.3 are in the reverse order, such that any pump flow would have

flowed from right to left, the PIV image was flipped for the sake of an easier visual comparison.

The streamlines for the vortices have been highlighted in Figure 2.12, and the direction of

15
circulation of each vortex was clearly marked. As shown, both the locations and directions of

circulation for each vortex are similar between the two experimental cases. Of special note is the

comparison between the middle pair in Figure 2.12b and the highlighted pair in Figure 2.12a,

where we see, from right to left, an incoming counterclockwise “secondary” vortex from the

previous pair, a clockwise “primary” vortex forming over the electrode pair and another

counterclockwise “secondary” vortex forming off of the trailing edge of the high voltage electrode.

The “primary vortices” are those that are generated directly by the Coulomb force’s vertical and

horizontal components and span across both electrodes in a pair, whereas the “secondary vortices”

are caused by the shear stresses induced in the liquid by the primary vortices’ flow.

Figure 2.12 Primary and Secondary Vortices for (a) Dye Visualization [19], (b) PIV Measurements

One difference of note between the two figures below is the inclusion of visualization of the

primary vortex over the first electrode pair in Figure 2.12b. No such visualization was provided in

[19]. As shown, this first primary vortex is almost exactly centered over the gap between the

electrodes, whereas the subsequent pair’s primary vortex is centered closer to the ground electrode.
16
It also appears that the second pair’s primary vortex is wider. Hemayatkhah et al. [19] equated the

size of the primary vortex with augmented flow, at least for liquids whose negative ionic mobilities

are greater than the negative ionic mobilities. Since this is true for dielectric liquids used in EHD

conduction in general, this provides an additional confirmation for the additive effect of electrode

pairs in EHD conduction pumps in such liquids.

Figure 2.13 Vortex Directions for Electrodes with Reversed Polarities, (a) numerical [20], (b) experimental
[19]

It should be noted that the numerical results from Yazdani and Yagoobi [20] shown in Figure

2.3 show the same direction of circulation for the primary and secondary vortices, despite having

opposite polarities defined for the electrodes. Hemayatkhah et al. [19] showed that such a reverasal

of polarities should result in vortices rotating in the opposite directions in the same dielectric liquid,

since the ionic mobilities of the negative and positive ions in the liquid are different. A comparison

between these two cases is shown in Figure 2.13 with the highlighted vortex directions. The reason

for this discrepancy is likely due to the assumption of equal ionic mobilities made by Yazdani and

Yagoobi [20]. The current study did not include cases for reversed electrode polarity, so it remains

a point of interest for future studies.

17
 Another trend of note relates to the maximum local flow velocities recorded for each applied

voltage and each experimental configuration. As expected, in the static cases, increasing the

applied voltage increases the maximum local velocity recorded. The net flow here is still

nonexistent, but the vortices churn faster at higher voltages, as can be seen in Figures 2.9a, 2.10a,

and 2.11a. The same kind of increase in local flow velocities with increasing voltage can also be

seen for the dynamic cases in Figures 2.9b, 2.10b, 2.11b, and for the opposing flow cases in Figures

2.9c, 2.10c, 2.11c. However, some unique observations can be made when comparing between

flow configurations at the same applied voltages. For instance, in the 2kV and 8kV dynamic cases

the maximum local velocity is lower than in the corresponding static cases, yet in the 20kV cases

the maximum dynamic local velocity is slightly higher than the static one. The tipping point

according to the full data set seems to be at 10kV. This corresponds to the voltage at which the

force applied on the liquid by the EHD conduction pump begins to increase exponentially, and can

better overcome the inertial resistance of the liquid.

Interestingly, the highest local flow velocities by far are shown in the opposing flow cases. In

these cases it would be reasonable to assume that the opposing force would weaken the local flow

velocities in the primary vortices of the EHD conduction pump. However, the data shows that

these primary vortices are both larger and rotate faster than their counterparts in the static and

dynamic cases. In the 20kV cases, the maximum local velocity is almost twice that of the dynamic

case, for instance. This is due to the effect of the incoming opposing flow on the formation of the

heterocharge layers over the electrodes. The opposing flow serves to push and lift the heterocharge

layers over the electrodes, such that they propagate further into the bulk of the working fluid [29].

Due to the asymmetry of the electrode designs, this causes the layer over the high voltage electrode

to increase in size more significantly than its counterpart on the ground electrode, making the

18
difference in space charge larger than it was before and thereby increasing the net Coulomb force

the pump applied on the liquid.

Figures 2.14-2.16 provide additional insights into these trends for the static, dynamic, and

opposing flow cases, respectively. In these figures, surface velocities are defined as the horizontal

flow velocities within the top 1mm of the liquid, whereas bulk or core velocities are defined as the

horizontal flow velocities below this surface “layer”. Each figure presents the average surface and

core velocities, as well as the maximum positive and negative magnitudes of these velocities.

Figure 2.14a shows that in the static case, while the average surface velocity remains positive

in the pump’s flow direction and slowly increases with applied voltage, the average core velocity

shifts significantly. Up to 10kV, the core actually flows backwards on average from the expected

flow direction for the EHD conduction pump. At higher voltages the core and surface move in the

same direction. In addition, when the core velocity becomes more negative, the surface velocity

becomes more positive and vice versa, but the surface flow never reverses. Figure 2.14b shows the

maximum negative core velocity clearly increasing up to 10kV before aligning more closely with

the maximum positive one and being overtaken by it. This asymmetrical circulation implies that

the flow in the core is bouncing back in the test volume with a greater force than the pump can

generate at low voltages but overcomes this counter-flow at higher voltages. This is made more

apparent when considering Figure 2.15b, in which the maximum and negative core velocities

remain roughly the same at all applied voltages – implying that the circulation is more symmetrical

due to the lack of confinement. Figure 2.14c shows that, overall, the maximum positive and

negative surface velocities are the same, since the fluid has nowhere to go, but diverge at the

maximum applied voltage, as the pump forcibly shoves more liquid against the edge of its confined

space.

19
Figure 2.14 Static Case Maximum Velocities

20
 Figure 2.15 Dynamic Case Maximum Velocities

Figure 2.15a shows that in the absence of confinement, in the dynamic open loop case, the

core’s average velocity does not dip as low as it does in the static case, though it still becomes

more negative up to 10kV. At higher voltages its average velocity becomes positive rapidly,

representing the full shift to the net circulating flow. Figure 2.15c shows that in this dynamic case

the film’s maximum positive and negative surface velocities are still roughly the same up to 10kV,

much like in the static case. However, there is a point of divergence at 10kV where the maximum

positive surface velocity overtakes the negative one. In addition, comparing Figures 2.15b and

2.15c shows that the positive surface velocity is higher than the positive core velocity for the same

voltages. This differs from the static case in Figures 2.14b and 2.14c, where the core’s positive

velocity is almost always higher than the positive surface velocity. This means that, overall, the

surface of the film in the dynamic case moves faster in the pumping direction than the core.

21
 Figure 2.16 Opposed Flow Case Maximum Velocities

Comparing these observations with the opposing flow case shown in Figure 2.16, we see

several significant differences. In Figure 2.16a, the film’s average surface velocity is no longer

always higher than the core’s average velocity – past 10kV the core accelerates significantly.

Figure 2.16b shows that the circulation in the core is asymmetrical at these higher voltages, with

the maximum positive velocity deviating significantly than the maximum negative one. Yet,

Figure 2.16c shows a very similar trend for the maximum surface velocities as Figure 2.14c for

the static case. Overall, Figure 2.16 shows that the opposing flow has a similar effect on the local

flow velocities as confining the flow to an enclosed space, but that the EHD conduction pump

manages to generate much higher positive velocities overall in this case. Affirming the idea that

the opposing flow actually improves the EHD conduction pump’s performance.

22
2.4. Conclusion
The Particle Imaging Velocimetry (PIV) flow visualization and velocity field measurement

technique was used on a multi-electrode EHD conduction pump, pumping a 12mm film in different

flow configurations. The results show that EHD conduction flows are highly circulatory, with

vortices emerging in the vicinity of the electrodes and generating waves on the film’s surface. PIV

imaging of the flow field provided a high resolution method of measuring the film’s surface

velocities and core velocities. The results show the additive effect of multiple electrode pairs on

the shape and strength of these vortices. In addition, a quantitative analysis of the flow velocities

shows that the velocities at the core of the liquid determine the overall strength of the flow and its

ability to push against obstruction – be it a wall or opposing flow. However, in a dynamic open

loop case with no obstructions, the film’s surface is the fastest moving portion of the flow. Lastly,

when applying an opposing flow to the EHD conduction pump’s generated flow, the result is a

direct and significant enhancement of the generated flow velocities throughout the liquid. This is

due to the underlying space charge and Coulomb force formation mechanisms, which are enhanced

by the presence of an opposing flow.

References
[1] Seyed-Yagoobi, J., “Electrohydrodynamic pumping of dielectric liquids”, Journal of Electrostatics, vol. 63,
pp.861-869 (2005).
[2] Stuetzer, O.M., “Ion drag pumps”, Journal of Applied Physics, vol. 31(1), pp.136-146 (1960).
[3] Melcher, J. R., “Traveling‐Wave Induced Electro-convection”, The Physics of Fluids, vol. 9(8),
pp. 1548-1555 (1966).
[4] Atten, P. and Seyed-Yagoobi, J., “Electrohydrodynamically induced dielectric liquid flow through pure
conduction in point/plane geometry”, IEEE Transactions on Dielectrics and Electrical Insulation, vol. 10(1),
pp. 27-36 (2003).
[5] Pontiga, F. and Castellanos, A., “Electrical conduction of electrolyte solutions in nonpolar liquids”, IEEE
Transactions on industry applications, vol. 32(4), pp.816-824 (1996).
[6] Ohyama, R., Aoyagi, K., Kitahara, Y. and Ohkubo, Y., “Visualization of the local ionic wind profile in a DC
corona discharge field by laser-induced phosphorescence emission”, Journal of visualization, vol. 10(1), pp.
75-82 (2007).

23
[7] Kitahara, Y., Aoyagi, K. and Ohyama, R., “An experimental analysis of ionic wind velocity characteristics
in a needle-plate electrode system by means of laser-induced phosphorescence”, Proceedings of the Annual
Conference on Electrical Insulation and Dielectric Phenomena, pp. 529-532 (2007).
[8] Sosa, R., Arnaud, E., Memin, E. and Artana, G., “Schlieren Image Velocimetry applied to EHD flows”,
Proceedings of the International Symposium on Electrohydrodynamics, pp. 331-334 (2006).
[9] Nakamura, H. and Ohyama, R., “An image analysis of positive ionic wind velocity under the DC corona
discharge in needle-cylinder electrode system”, Proceedings of the IEEE Conference on Electrical Insulation
and Dielectric Phenomena, pp. 192-195 (2009).
[10] Kocik, M., Podlinski, J., Mizeraczyk, J. and Chang, J.S., “Particle image velocimetry measurements of wire-
nonparallel plates type electrohydrodynamic gas pump”, IEEE Transactions on Dielectrics and Electrical
Insulation, vol. 16(2), pp. 312-319 (2009).
[11] Ohyama, R., Inoue, K. and Chang, J.S., “Schlieren optical visualization for transient EHD induced flow in a
stratified dielectric liquid under gas-phase ac corona discharges”, Journal of Physics D: Applied Physics, vol.
40(2), p.573 (2007).
[12] Daaboul, M., Louste, C. and Romat, H., “LDV measurements of liquid velocity induced by charge injection
in Diesel oil in a blade-plane-slit geometry” Journal of Physics: Conference Series, vol. 142(1), p. 012041
(2008).
[13] McCluskey, F.M.J. and Perez, A.T., “The electrohydrodynamic plume between a line source of ions and a
flat plate-theory and experiment”, IEEE Transactions on electrical insulation, vol. 27(2), pp.334-341 (1992).
[14] Daaboul, M., Louste, C. and Romat, H., “Transient velocity induced by electric injection in blade-plane
geometry”, Journal of Electrostatics, vol. 67(2-3), pp.359-364 (2009).
[15] Yan, Z., Louste, C., Traoré, P. and Romat, H., “Velocity and turbulence intensity of an EHD impinging
dielectric liquid jet in blade–plane geometry”, IEEE Transactions on Industry Applications, vol. 49(5),
pp.2314-2322 (2013).
[16] Wawzyniak, M., J. Seyed-Yagoobi, and G. L. Morrison. "An experimental study of electrohydrodynamic
induction pumping of a stratified liquid/vapor medium." Journal of heat transfer, vol. 122(1), pp. 200-203
(2000).
[17] Ohyama, R., Kaneko, K. and Chang, J.S., “Flow visualization and image analysis of gas-phase AC corona
discharge induced electrohydrodynamic liquid flow in a stratified fluid”, IEEE transactions on dielectrics
and electrical insulation, vol. 10(1), pp. 57-64 (2003).
[18] Qin, J.J., Yeo, L.Y. and Friend, J.R., “MicroPIV and micromixing study of corona wind induced
microcentrifugation flows in a cylindrical cavity”, Journal of Microfluidics and Nanofluidics, vol. 8(2), pp.
231-241 (2010).
[19] Hemayatkhah, M., Gharraei, R. and Esmaeilzadeh, E., “Flow pattern visualization of liquid film conduction
pumping using flush mounted electrodes”, Experimental Thermal and Fluid Science, vol. 35(6), pp.933-938
(2011).
[20] Yazdani, M. and Seyed-Yagoobi, J., “Electrically induced dielectric liquid film flow based on electric
conduction phenomenon”, IEEE Transactions on dielectrics and electrical insulation, vol. 16(3), pp. 768-
777 (2009).
[21] Daaboul, M., Traoré, P., Vázquez, P. and Louste, C., “Study of the transition from conduction to injection in
an electrohydrodynamic flow in blade-plane geometry”, Journal of Electrostatics, vol. 88, pp.71-75 (2017).
[22] Chirkov, V.A., Komarov, D.K., Stishkov, Y.K. and Vasilkov, S.A., “Comparative analysis of numerical
simulation and PIV experimental results for a flow caused by field-enhanced dissociation”, Journal of
Physics: Conference Series, vol. 646(1), p. 012033 (2015).
[23] Siddiqui, M.A. and Seyed-Yagoobi, J., “Experimental study of pumping of liquid film with electric
conduction phenomenon”, IEEE Transactions on industry applications, vol. 45(1), pp.3-9 (2009).
[24] 3M, “3M™ Novec™ 7000 Engineered Fluid”, 10316HB datasheet, issued November 2014.
[25] Daaboul, M., Louste, C. and Romat, H., “PIV measurements on charged plumes-influence of SiO 2 seeding
particles on the electrical behavior” IEEE Transactions on Dielectrics and Electrical Insulation, vol. 16(2),
pp.335-342 (2009).
[26] Wieneke, B., “PIV uncertainty quantification from correlation statistics”, Measurement Science and
Technology, 26(7), p.074002 (2015).

24
[27] Sciacchitano, A., Neal, D.R., Smith, B.L., Warner, S.O., Vlachos, P.P., Wieneke, B. and Scarano, F.,
“Collaborative framework for PIV uncertainty quantification: comparative assessment of methods”,
Measurement Science and Technology, 26(7), p.074004 (2015).
[28] Bardet, P.M., André, M.A. and Neal, D.R., “Systematic timing errors in laser-based transit-time
velocimetry”, proceedings of the 10th International Symposium on Particle Image Velocimetry—PIV13
(Delft, The Netherlands, July 2013).
[29] Talmor, M., Seyed-Yagoobi, J., “Numerical Study of Micro-Scale EHD Conduction Pumping: The Effect of
Pump Orientation and Flow Inertia on Heterocharge Layer Morphology and Flow Distribution Control”,
manuscript submitted for publication to the Journal of Electrostatics

25
 Chapter 3 - Flow Distribution Control in Micro-scale via
Electrohydrodynamic Conduction Pumping

Note: This chapter is based on the following technical paper.

Talmor, M., Yang, L., Larkin, T.R., Kamat, O.K., Dancy, T.J., and Yagoobi, J.S., “Flow
Distribution Control in Micro-Scale via Electrohydrodynamic Conduction Pumping”, Proceedings
of Electrostatic Society of America Annual Meeting, Lafayette, Indiana, June 2016.

3.1. Introduction
Flow distribution control is an essential capability for modern day fluid based thermal control

systems, where the needs of the system can vary from one moment to the next. Smart, automated

control and near-immediate response times are desirable characteristics for systems that can

redistribute flow to areas of need in real time. Much like electric smart grids, such a system would

also be able to minimize the power costs of controlling the flow during times when the heat loads

are reduced. Longevity, ease of operation and ease of maintenance are similarly desirable to reduce

the costs and increase the robustness of the overall flow control system. In macro-scale flows,

there are many flow distribution solutions available on the market using traditional fluid pumps.

Yet, as innovations lead to further miniaturization of technology, fluid-based thermal control

systems for high power electronics can no longer effectively utilize traditional solutions involving

large, vibrating machinery with many moving elements. The need for a flow distribution control

systems at the meso-, micro- and even nano-scales requires non-traditional, innovative flow

control solutions that are robust yet simple, consume less power, are easily maintainable and are

still effective at manipulating flows when scaled down to these smaller scales.

An example of such a non-traditional innovation is an electrohydrodynamic (EHD) conduction

based flow generation and distribution control system. Electrohydrodynamics is the study of the

26
interactions between applied electric fields and fluid flow fields. In all EHD phenomena, fluid

motion is caused by charges in the working fluid migrating under the effect of an applied electric

field, and carrying the surrounding fluid medium with them, thereby generating fluid motion. In

EHD conduction pumping, this fluid motion is achieved by applying a strong electric field between

electrodes submerged within a dielectric working fluid. In dielectric fluids, in the absence of a

strong electric field, naturally present electrolyte impurities dissociate into charged ions and

recombine back into neutral particles at an equal rate [1]. The general reaction is given in (1).

𝐷𝑖𝑠𝑠𝑜𝑐𝑖𝑎𝑡𝑖𝑜𝑛
→
𝐴𝐵 𝐴+ + 𝐵 − (1)
←
𝑅𝑒𝑐𝑜𝑚𝑏𝑖𝑛𝑎𝑡𝑖𝑜𝑛

where 𝐴+ is the positive ionic species, 𝐵 − is the negative ionic species, and 𝐴𝐵 is the neutrally

charged electrolyte particle. Under the effect of a strong electric field, above a critical threshold

on the order of 1kV/cm, the rate of dissociation increases while the rate of recombination remains

relatively constant. The sudden excess of ionic charges are attracted to the nearby oppositely

charged electrodes and form uniformly charged layers over each of the electrodes, referred to as

the heterocharge layers. These layers represent local regions of space charges near the electrodes.

Opposing Coulomb forces are then applied onto the working fluid due to the presence of these

space charges. If the high voltage and ground electrodes have identical wetted surface area

(symmetrical electrodes), and the ionic mobilities are the same for both negative and positive

species, the opposing Coulomb forces exerted on the fluid are equal in magnitude for both the

negative and positive space charges, and only flow circulation occurs between the electrodes.

However, purposeful asymmetry in the electrode design causes a force imbalance that generates a

net flow [2].

27
 Figure 3.1 EHD Conduction Pumping Mechanism

Figure 3.1 schematically shows this mechanism for a single electrode pair with the

heterocharge layers illustrated over each electrode surface. Since the magnitudes of the applied

forces on the fluid depend on the electrode dimensions, the force attracting the negative charge

carriers to the high voltage electrode would be larger than the force attracting positive charge

carriers to the ground electrode in this figure, leading to a net force and a net flow as shown.

Numerical investigations by Yazdani and Yagoobi [3] have confirmed that, under the assumption

of equal charge mobility for both positive and negative ions [4], the net flow direction is always

toward the electrode with higher wetted surface area, regardless of the polarity of the individual

electrodes.

Due to this flow generation mechanism, EHD conduction pumps contain no moving parts and

their performance is entirely controlled by the applied voltage between the electrodes. Although

this voltage is high, in order to generate the strong electric field, the consumed current is only on

the order of μA, resulting in little power consumption - on the order of a few Watts or less. EHD

conduction pumps can contain several electrode pairs, spaced out such that the electric field of

each pair does not interfere with neighboring pairs. Jeong and Seyed-Yagoobi [5] have shown

improvement in performance increases quasi-linearly with the number of electrode pairs in a single

pump. Theoretical studies of EHD conduction by Atten and Seyed-Yagoobi [6] have shown that

EHD pumps are capable of generating significant pressure and flow rates. Experimental work by

Mahmoudi et al., Pearson et al. and Patel et al. [7-9] have confirmed that these devices are capable

28
of generating useful pressure and fluid flow in macro-, meso-, and micro- scales. A comparative

study performed by Pearson and Seyed-Yagoobi [10] showed that EHD conduction performance

significantly improves with decreasing size scales, and that careful selection of the electrode

geometries can significantly change the pressure and flow rate generation performance of the final

pump. These properties make EHD conduction pumps highly customizable and scalable to the

performance needs for any given thermal management challenge. However, these single phase

studies have focused only on a single branch, without studying flow distribution control between

multiple branches. Feng and Seyed-Yagoobi [11] previously studied single phase flow generation

and distribution control in a macro-scale, simple two branch system. In their experimental study,

a single EHD pump in one branch (the “active” branch) was able to redirect the flow generated by

an upstream mechanical pump from the other branch (the “inactive” branch), until the inactive

branch was entirely starved of flow at the maximum operational voltage of the active branch’s

pump. More recently, Yang et al. [12] studied and characterized the performance of an EHD

conduction driven, single phase flow distribution system in meso-scale in a three-branch system.

The results of that study showed that EHD conduction driven flow distribution works well in meso-

scale and in more complex systems, and that both flow redistribution and maldistribution recovery

is possible using EHD conduction pumps. In addition, that study showed that the trends of behavior

for such a system could be reasonably predicted via simplified simulations based solely on pre-

obtained performance curves for the EHD conduction pumps to be used. Representative results

from that study, for flow redistribution and maldistribution recovery, are shown in Figure 3.2 and

Figure 3.3, respectively.

The current study aims to show that EHD driven, single phase flow distribution control is

effective at the micro-scale, which is relevant for many applications in electronics cooling, and to

29
characterize the performance of the distribution system at this size scale. The work presented here

is the preliminary characterization of EHD driven flow distribution control between twin parallel,

500 micron tall microchannels.

4
Exp. Inactive 1

Flow Rate [mL/min]

3
Exp. Inactive 2
2 Exp. Branch
1 Sim. Inactive 1

0 Sim. Inactive 2

-1 Sim. Branch
0 500 1000 1500
Applied EHD Voltage [V]

Figure 3.2 Meso-scale Flow Rate Redistribution with Initially equal 1.5 mL/min Distribution [12]

2.5 Exp. Inactive 1

Flow Rate [mL/min]

2 Exp. Inactive 2

Exp. Active
1.5
Sim. Inactive 1
1
Sim. Inactive 2
0.5 Sim. Active
0 500 1000 1500
Applied EHD Voltage [V]

Figure 3.3 EHD Conduction Pumping Mechanism [12]

30
3.2. Experimental Setup

Figure 3.4 Schematic of Parallel Microchannel Experimental Setup

The schematic for the experimental setup used in this study is shown in Figure 3.4. The flow rate

was supplied to the parallel branches via a Cole Parmer Model 75211-10 mechanical pump. A

manifold split the flow generated by the supply pump into two branches. Each branch contained a

circular channel EHD conduction pump with 20 electrode pairs and a flow channel diameter of

1mm. Downstream of the EHD pump was a sample chip assembly containing the parallel, separate

microchannels. The branches converged again at another manifold, which fed back into the supply

pump, completing the loop. The branch distribution control pumps were highly reliable, meso-

scale EHD conduction pumps with a well-known performance based on previous studies [13]-[14].

These pumps had an operational voltage between 0kV and 1.5kV. The electrode design for the

pumps is depicted in Figure 3.5, with electrode dimensions given in Table 3.1. Each EHD pump

was instrumented with a Validyne DP15 pressure transducer and a micro-flow Sensirion SLQ-

HC60 flow meter. The systematic measurement error, calculated based on the information

provided by the manufacturers of the sensors and power supply, is shown in Table 3.2.

31
 Figure 3.5 Distribution Control EHD Conduction Pump [13]

Table 3.1 Branch Pumps Electrode Dimensions [13]

Table 3.2 Systematic Measurement Errors

Precise turn valves on each branch allowed for control of the initial distribution. This ensured

that the baseline distribution could be made to either be equal across all branches or purposefully

unequal – in order to observe the effectiveness of the distribution control and flow stabilization

capabilities of the system. Due to power supply constraints, all EHD pumps were connected to the

32
same voltage input, such that each pump that was connected to the supply received the same

voltage. However, each pump could be disconnected from the communal voltage independently

from the others via a switching board specifically constructed for use in this experiment. The data

acquisition system was comprised of two National Instruments boards, to accommodate the many

differential sensors involved, the necessary inputs for voltage and current monitors and the

necessary control outputs. The primary board was a National Instruments NI-PCI 6024e and the

secondary board was a National Instruments NI-USB-6009. Both were connected to a LabVIEW

program on a dedicated computer. The working fluid was the HCFC-123 refrigerant (commonly

referred to as R-123). The electric [15] and thermodynamic [16] properties of this working fluid

are shown in Table 3.3.

Table 3.3 HCFC-123 Properties at 25°C [15], [16]

3.3. Results and Discussion

The preliminary results presented here are for flow redistribution of an initially equal distribution,

with different externally applied flow velocities. Figure 3.6 shows a case with an initial 1.35

mL/min in each branch, corresponding to a mass flux of about 90 kg/m2s to the system. As one

EHD pump is activated from 0V to 1500V, the flow begins to diverge, with the divergence

initiation point being almost immediate, at 200V. This case shows that flow redistribution at this

33
size scale was very effective for this flow rate. Flow starvation of the inactive microchannel was

also still possible at this size scale and with this initial flow distribution. This shows that despite

the greatly increased pressure load due to the reduced channel size to 500 microns, the EHD

conduction pumps were still able to have a significant effect on the flow distribution.

2.5 75

Mass Flux [kg/m2s]

Flow Rate [mL/min]

2 60
1.5 45
1 30 Active Branch
0.5 15 Inactive Branch
0 0
-0.5 -15
0 500 1000 1500
Applied EHD Voltage [V]

Figure 3.6 Flow Starvation in Inactive Branch, with 42 kg/m2s Initially Equal Distribution

20 80
Pressure Drop [Pa]

10
Active
Current [μA]

0 60 Branch
Pressure
-10
40 Inactive
-20
Branch
-30 20 Pressure
-40 EHD
Current
-50 0
0 500 1000 1500
Applied EHD Voltage [V]

Figure 3.7 Pressure Drop and Current Load for 1.35 mL/min Initially Equal Distribution

Figure 7 shows the pressure drops in each branch, as well as the consumed current for the

active EHD conduction pump. As previously mentioned, the current expectedly remained on the

order of tens of μA, making the maximum power consumption consumed by the EHD pump for

this case only 0.1 W at the highest applied voltage. The pressure drop is shown to significantly

decrease in the active branch due to the pressure generation of the EHD conduction pump. The

34
smaller decrease in the pressure drop in the inactive branch is attributed to the reduction in flow

rate through that branch, which reduced the frictional pressure losses in that branch.

Figure 3.8 shows a case with a higher initial flow rate distribution of 2.1 mL/min in each branch

was used, corresponding to a total mass flux of 130 kg/m2s to the system. In this case, the EHD

conduction pump was able to obtain a separation of 76 kg/m2s, with the active branch achieving

104 kg/m2s at the maximum applied voltage. Flow starvation was not observed, and the point of

divergence in the flow rates between the two branches was delayed to 600V, as shown. This

matches the trend observed by Yang et al. [12] for meso-scale EHD conduction driven flow

distribution. This delayed divergence effect is attributed to both the greater frictional losses that

arise from the increased flow rate, as well as the effects of the incoming flow velocity on the

formation and layer thicknesses of the heterocharge layers. In this figure we also observe a slight

bowing of the flow rate curves, wherein the flow rate in the active branch decreases before

increasing and the flow rate in the inactive branch increases before decreasing.

3.5 110
Flow Rate [mL/min]

Mass Flux [kg/m2s]

3
88
2.5
2 66
1.5
Active Branch
44
1 Inactive Branch
22
0.5
0 0
0 500 1000 1500
Applied EHD Voltage [V]

Figure 3.8 Flow Redistribution, with 2.1 mL/min Initially Equal Distribution

Figure 3.9 shows the corresponding pressure drops and current. The pressure drop is shown to

grow in the active branch before the EHD pump’s pressure generation reverses the trend and flow

divergence begins. The pressure drop also grows in the inactive branch due to the temporary

35
increase in flow rate, and therefore the frictional losses in the inactive branch, before being reduced

due to the flow being redirected to the active branch.

50 80
40 Active

Pressure Drop [Pa]

Branch
30 60

Current [μA]
Pressure
20
10 40 Inactive
0 Branch
-10 20 Pressure
-20 EHD
-30 0 Current
0 500 1000 1500
Applied EHD Voltage [V]

Figure 3.9 Pressure Drop and Current Load for 2.1 mL/min Initially Equal Distribution

Figure 3.10 shows a near-limiting case for the capability of the EHD conduction pump to

diverge the flow using this micro-scale configuration. In this figure, an initially equal distribution

of 4.5 mL/min was used, corresponding to a total mass flux of 280 kg/m2s. Under this condition,

the EHD pump was not able to have a noteworthy positive effect on the flow divergence until

1200V, and the final divergence at the maximum applied voltage of 1500V was only 20 kg/m2s.

Additional experiments have shown that the EHD pump ceases to have an effect on the flow

distribution at 5mL/min in each branch. This is lower than the limit shown for the meso-scale

experiments [12], but still within the same order of magnitude. This implies that while the greater

frictional losses due to the microchannels’ size have a limiting effect on the flow rate range at

which EHD conduction can be used to control the flow distribution, the performance of the EHD

conduction driven flow distribution system in micro-scale is still comparable to that at larger

scales. The curve bowing shown in Figure 3.10 is more significant than in Figure 3.8, showing that

this behavior becomes severe at this size scale near the effective flow rate limit, due to the

additional pressure losses from the increase in flow rates.

36
 6 180

Flow Rate [mL/min]

Mass Flux [kg/m2s]

5
135
4
3 90
Active Branch

2 Inactive Branch
45
1
0 0
0 500 1000 1500
Applied EHD Voltage [V]

Figure 3.10 Flow Redistribution, with 4.5 mL/min Initially Equal Distribution

100 80
Active
Pressure Drop [Pa]

80
60 Branch

Current [μA]
Pressure
60
40 Inactive
40 Branch
20 Pressure
20
EHD
0 0 Current
0 500 1000 1500
Applied EHD Voltage [V]

Figure 3.11 Pressure Drop and Current Load for 4.5 mL/min Initially Equal Distribution

Figure 3.11 shows the pressure drops and current for this case. The pressure drop in the active

branch does not dip below its original value until the point of divergence, before which no pressure

generation occurs. The pressure drop is also shown to increase in the active branch at the voltage

levels prior to divergence, implying that activating the pump at lower voltages near its operational

limit was working against the flow. Therefore, even though divergence does occur in this case, for

the majority of the applied voltage range the EHD pump was actually working against the flow

rather than aiding it, making operating near the divergence limit overall undesirable for this sort

of system. Understanding how to minimize this unwanted effect near the flow rate limit will be of

great value for any thermal control systems at these scales that relies on EHD conduction pumping

technology.

37
 A comparative analysis between the limits of performances, in terms of mass fluxes, for the

EHD conduction driven flow distribution systems studied thus far at different size scales is shown

in Table 3.4. In this table, the second column describes the minimum mass flux at which flow

starvation of the inactive branch was achieved, while the third column shows the mass flux value

at which the EHD driven system is no longer able to diverge the flow between the branches. For

the macro-scale study this value had to be estimated based on the trends of behavior observed at

the other size scales, since the original study did not investigate this limiting case. The data

suggests that although the flow channel sizes reduce by an order of magnitude between the

different size scales, the EHD driven flow distribution system still had comparable performance at

all scales. This shows that such a system is capable of directing small-scale flows as effectively as

large-scale flows, and its relative performance improves at smaller size scales.

Table 3.4 Comparison of Scale Dependent Mass Flux Redistribution Capabilities at Max. Voltage

38
3.4. Conclusion
The work presented here has demonstrated preliminary results for the performance

characterization of an EHD conduction driven, single phase flow distribution system for parallel

microchannels. The results show that EHD conduction driven flow distribution is still effective in

a micro-scale system and has comparable performance in terms of mass fluxes to similar systems

at larger size scales that have been previously investigated. This confirms the potential for effective

EHD conduction driven thermal control systems at the micro-scale, making this technology

desirable for small-scale electronics cooling. To obtain the full characterization of the flow

redistribution and recovery from maldistribution capabilities of this system, as well as an

understanding of what occurs near the flow rate limit at small scales, further studies still need to

be performed.

References
[1] J. Seyed-Yagoobi, “Electrohydrodynamic pumping of dielectric liquids”, Journal of Electrostatics, vol.
63, pp. 861-86, Jun. 2005.
[2] A. Castellanos, "Conduction models in dielectric liquids," in Electrohydrodynamics, Springer Vienna,
1998, pp. 103-120.
[3] M. Yazdani and J. Seyed-Yagoobi, “Electrically Induced Dielectric Liquid Film Flow Based on Electric
Conduction Phenomenon”, IEEE Transactions on Dielectrics and Electrical Insulation, vol. 16, pp.
768-77, Jun. 2009.
[4] M. Yazdani and J. Seyed-Yagoobi, “Effect of Charge Mobility on Dielectric Liquid Flow Driven by
EHD Conduction Phenomenon”, Journal of Electrostatics, vol. 72, pp. 285-294, Aug. 2014.
[5] S. I. Jeong, and J. Seyed-Yagoobi, “Innovative Electrode Designs for Electrohydrodynamic Conduction
Pumping,” IEEE Transactions on Industry Applications, Vol. 40, No. 3, pp. 900-904, May 2004.
[6] P. Atten, and J. Seyed-Yagoobi, “Electrohydrodynamically Induced Dielectric Liquid Flow through
Pure Conduction in Point/Plane Geometry”, IEEE Transactions on Dielectrics and Electrical
Insulation, vol. 10, pp. 27-36, Feb. 2003.
[7] S. R. Mahmoudi, K. Adamiak and G.S.P. Castle, “Prediction of the Static Pressure Generation for an
Electrohydrodynamic Conduction Pump”, presented at the 17th IEEE International Conference on
Dielectric Liquids, Trondheim, Norway, 2011.
[8] M. R. Pearson and J. Seyed-Yagoobi, “Experimental study of EHD conduction pumping at the meso-
and micro-scale”, Journal of Electrostatics, vol. 69, pp. 479-485, Dec. 2011.
[9] V. K. Patel, and J. Seyed-Yagoobi, “Dielectric fluid flow generation in meso-tubes with micro-scale
electrohydrodynamic conduction pumping”, presented at the 17th IEEE International Conference on
Dielectric Liquids, Trondheim, Norway, 2011.

39
[10] M. R. Pearson, and J. Seyed-Yagoobi, “Advances in Electrohydrodynamic Conduction Pumping”,
IEEE Transactions on Dielectrics and Electrical Insulation, vol. 16, pp. 424-434, Apr. 2009.
[11] Y. Feng and J. Seyed-Yagoobi, , “Control of Liquid Flow Distribution Utilizing EHD Conduction
Pumping Mechanism”, IEEE Transactions on Industry Applications, Vol. 42, No. 2, pp. 369-377, Mar.
2006.
[12] L. Yang, M. Talmor, K. S. Minchev, C. Jiang, B. C. Shaw and J. Seyed-Yagoobi, "Flow distribution
control in meso scale via electrohydrodynamic conduction pumping." presented at the Annual Meeting
of the IEEE Industry Applications Society, Dallas, TX, USA, 2015.
[13] V. K. Patel, F. Robinson and J. Seyed-Yagoobi, “Terrestrial and Microgravity Experimental Study of
Microscale Heat-Transport Device Driven by Electrohydrodynamic Conduction Pumping”, IEEE
Transactions on Industry Applications, vol. 49, pp. 2397-2401, Nov. 2013.
[14] V. K. Patel and J. Seyed-Yagoobi, "Long-Term Performance Evaluation of Microscale Two-Phase Heat
Transport Device Driven by EHD Conduction." IEEE Transactions on Industry Applications, vol. 50,
pp. 3011-3016, Sep. 2014
[15] J. E. Bryan, "An experimental study of ion-drag pumping in a vertical axisymmetric configuration,"
Ph.D. dissertation, Department of Mechanical Engineering, Texas A&M University, College Station,
TX, 1990.
[16] M. O. McLinden, et al., “NIST thermodynamic and transport properties of refrigerants and refrigerant
mixtures,” NIST, Gaithersburg, MD, Rep. REFPROP 6.01, Jan. 1998

40
 Chapter 4 - Numerical Study of Micro-scale EHD
Conduction Pumping: The Effect of Pump Orientation and
Flow Inertia on Heterocharge Layer Morphology and Flow
Distribution Control

Note: This chapter is based on the following technical publication.

Talmor, M. and Seyed-Yagoobi, J., “Numerical Study of Micro-scale EHD Conduction Pumping: The
Effect of Pump Orientation and Flow Inertia on Heterocharge Layer Morphology and Flow Distribution
Control”, Journal of Electrostatics, Vol. 111, pp. 1-13, May 2021
(https://doi.org/10.1016/j.elstat.2020.103548).

4.1. Introduction
Electrohydrodynamic (EHD) conduction pumping uses the interaction between a strong,

externally applied electric field and dissociated electrolyte impurities within a dielectric working

fluid to generate fluid motion. Naturally occurring electrolyte impurities present in all dielectric

fluids undergo a reversible chemical process of dissociation into ions and recombination back

into neutral species. In the absence of a strong electric field, the equilibrium reaction rates for the

dissociation and recombination reactions are equal, such that the overall fluid is considered

electroneutral [1]. This reversible reaction is denoted as,

41
 When a strong electric field is applied between submerged electrodes in the working fluid,

the rate of the dissociation reaction exceeds that of the recombination reaction, leading to a

significant increase in the amount of available ionic charges. These charges migrate toward the

oppositely charged electrodes under the influence of the electric field, forming thin heterocharge

layers over the wetted surface area of each electrode. While the aforementioned equilibrium and

electroneutral conditions are still valid in the bulk of the fluid, these heterocharge layers

represent a different equilibrium in the near-electrode region, in which the ions retain their charge

while transiting across the region of the applicable heterocharge layer before recombining with

an oppositely charged ion [2]. These heterocharge layers provide significant regions of space

charge, which translate to the presence of opposing Coulomb forces on the ions. The magnitude

of each force derives directly from the morphology of the heterocharge layers, and therefore

from the geometry of the submerged electrodes. During their transit within the heterocharge

layers, the ions drag the fluid molecules within their solvation sheath, translating the electric

force onto the bulk of the fluid [3]. Under the assumption of equal ionic mobilities, if the

oppositely charged electrodes have equal wetted surface areas, these forces will be equal in

magnitude and only flow circulation in the gap between the electrodes will be observed. To

achieve a net flow, the electrodes must be designed asymmetrically such that a net force is

generated pointing in the desired flow direction [4]. As a matter of convention for EHD

conduction pumps, the high voltage electrode is selected to have a higher wetted surface area,

and the net flow direction is subsequently set to go from the ground electrode toward the high

voltage one. While the simulation model herein does not take this into account, a design for a

physical pump that follows this convention takes advantage of experimental evidence that the

negative ionic current is stronger than the positive one for non-polar dielectric liquids [5,6]. The

42
EHD conduction process is shown qualitatively in Figure 4.1, for the simple case of electrodes

flush with the flow channel walls. In this figure, the high voltage electrode is depicted as having

a higher wetted surface area (width) than the ground electrode, thereby generating a greater

Coulomb force and a net flow toward the right hand side. It should be noted that asymmetry

between the wetted surface areas of the electrodes can be achieved in many ways [7], most

notably involving the use of porous media [8], such that width alone is an insufficient descriptor

for the role of each electrode in an EHD conduction pump. Following the aforementioned

convention, it is typical to refer to the electrodes by their applied voltage, with the flow direction

implied by these definitions.

Figure 4.1 Schematic of EHD Conduction Mechanism

EHD conduction has been investigated theoretically and numerically by several

researchers over the last few decades. Based on fundamental work on dissociation and

conduction of charges in a dielectric medium done by Pontiga and Castellanos [9], Atten and

Seyed-Yagoobi [10] first presented the mathematical model for the EHD conduction pumping

mechanism and established the key nondimensional parameters related to EHD conduction

pumping performance. Yazdani and Seyed-Yagoobi [11] performed the first numerical

simulations using this nondimensional model for the EHD conduction mechanism, providing

the first simulated morphology profiles for the heterocharge layers for an EHD driven liquid

43
film flow.

Additional numerical studies by Yazdani and Seyed-Yagoobi [12] and Chirkov et al.

[13] investigated the effect of the ratio of the positive and negative ionic mobilities on EHD

conduction pumping performance, since in realistic applications equal ionic mobilities are not

guaranteed. Gharraei et al. [14] performed similar simulations on a dimensional model and

compared the results to experimental data, showing a good prediction of the effect of changes

to mobility using true mobility data, and investigating the relationship between this effect and

the effect of the electrode asymmetry ratio on the pumping performance. Mahmoudi et al. [15]

showed that these numerical models could also be used for the prediction of pressure generation

performance of an EHD conduction pump. Studies of the effect of charge convection due to the

generated flow on the heterocharge layer were conducted numerically by Feng and Seyed-

Yagoobi [16]. The aforementioned numerical studies have verified the general validity of the

mathematical model as a predictive tool of EHD conduction performance, but have focused on

simulating only a single pair of electrodes using periodic boundary conditions to simulate the

remainder of the electrode pairs. However, realistic EHD conduction pumps contain multiple

electrode pairs in order to enhance the overall pumping performance based on the non-linear

additive effect on pressure and flow rate generation that was experimentally observed by and

were experimentally observed by Patel and Seyed-Yagoobi [17].

Recent studies have also examined the use of EHD conduction pumps for flow

distribution control between multiple parallel flow branches. Sinnamon [18], for example,

proposed the use of an embedded EHD conduction driven thermal control system in a satellite

structural panel with distributive channels, as shown in Figure 4.2. His work was based on the

successful experimental demonstrations of EHD conduction driven flow distribution control by

44
Feng and Seyed-Yagoobi [19, 20] for both single and two-phase flows in macro-scale.

Figure 4.2 Isothermal Panel, which Integrates Fluid Channels into the Rib Structure and Face-sheet [18]

Yang et al. [21,22] also experimentally investigated flow distribution control in single and

two-phase flows in meso-scale, and compared the results with rudimentary numerical

simulations that were based on conduction pumps, this study also observes the heterocharge

layer morphology over multiple electrode pairs. In addition, flow distribution control

performance using the forward and reverse pumping orientations previously discussed is also

observed and explained.

These simulated cases represent a comprehensive overview of the common operational

conditions EHD conduction driven flow control systems need to operate under at the meso- and

micro-scale. A com- parison with available experimental data is also provided qualitatively in

this study, using the dimensional simulation results.

4.2. Mathematical Model

The EHD conduction pumping mechanism for an isothermal dielectric fluid relies on three core

45
physics phenomena whose mathematical models must be coupled and solved together to arrive

at a solution. These three models describe the electrical field and charge density, the fluid flow,

and the charge conservation within the working fluid. The electrical field is described by the

familiar electrostatics Poisson Equation,

where 𝐸̅ is the electric field, V is the electric potential, ε is the electric permittivity, and

ρe is the total charge density as defined by the ionic charge densities pe and ne. The fluid flow is

described by the standard mass continuity and Navier-Stokes equations for the conservation of

the rate of change of momentum, for an isothermal, incompressible fluid’s internal flow in

steady state in a horizontal branch,

where 𝑢̅ is the fluid velocity, ∇P is the pressure gradient, and μ is the fluid viscosity.

These equations assume that the working fluid is incompressible and has constant properties.

The final term in Equation (4) is the Coulomb force, which couples the electrostatics and fluid

flow models together,

The charge concentration conservation is often described by the conservation of current

46
density for the positive and negative ionic currents,

where J is the current density, Di is the diffusion coefficient for the charge carrier, b is

the electric mobility for the charge carrier, zi is the charge number, and ne / pe is the charge

density for the charge carrier.

For this numerical study, however, it was useful to use an alternate formulation of

Equation (6) for the concentration of the charged species rather than charge density. This

formulation is known as the Nernst- Planck equation for conservation of concentration of

charged species, and is shown here for both the positive and negative charges as separate

equations,

The n / p terms stand for the concentrations of negative and positive charges, respectively.

The concentration and charge density are simply related via the unit charge, e0,

To make matters simple for the simulation model, the ionic mobility term is considered

47
to be a constant, such that the mobilities of each ionic species are the same (b+ = b- = b). While

it is generally known that the real ionic mobilities for the positive and negative species

in dielectric liquids are not the same, the effect of charge mobility is beyond the scope of this

study and has been extensively studied in the literature [[12], [13], [14]]. The diffusion rate

coefficient for the working fluid is obtained from the Stokes-Einstein relation for charged

particles [24],

The last term in Equations (7) and (8) represents the reaction rates for the generation of

charges. This term is the same for both concentration conservation equations,

In Equation (11), neq is the equilibrium concentration of the neutral species in the absence

of a strong electric field. In such an equilibrium, Ri = 0. The recombination rate constant, kR,

and the electric field enhanced dissociation rate constant, kD can be expressed as [25],

where the constant expression for kR is the upper bound for the recombination rate as

originally given by Langevin [26], who showed that for ions whose ionic radius is small

compared with the mean ionic separation, as is the case in nonpolar fluids such as the dielectric

liquids used in EHD conduction, that the dissociation rate is independent of the applied electric

field and has the maximum value given in (12). kD,0 in (13) is the equilibrium dissociation

48
rate in the absence of a strong electric field, and F ωe is known as the Onsager field enhancement

function [27]. This function is defined using the Onsager parameter, ωe ,

The equilibrium dissociation rate is obtained by evaluating (11) for neq2 = np,

the equilibrium condition. This yields the following:

Putting (12) – (13) into (11) then results in the following expression for the overall reaction

balance between dissociation and recombination. This is used as a source term for added

concentration of charged species in Equation (7) when the strong electric field is applied.

Equations (2), (4), (7) and (8) form the fully coupled equations that must be solved together

to obtain EHD conduction driven flow simulations. These equations can be nondimensionalized

using the following parameter conversions,

Noting that the velocity and pressure terms are nondimensionalized using the ion drift

velocity, ue = bV/d and V0 is the applied potential between the electrodes. The following set of

49
nondimensional equations is then obtained, where several nondimensional parameters now

appear in the formulation,

ReEHD is the electric Reynolds number commonly used in EHD formulations [28], which

has the same form as the traditional Reynolds number but the characteristic ion drift velocity,

ue, is used instead of the mean fluid velocity. α is the familiar diffusivity parameter, and C0,

M0 are the EHD conduction characteristic parameters,

where n0 is the equilibrium charge density, n0 = e0neq. M0 represents the ratio between the

fluid’s electroconvective mobility, which is dependent only on the fluid’s physical properties,

and the true ionic mobility of the charged species in the fluid. As such, this parameter

represents a ratio between the potential velocity the fluid would achieve solely via the

application of the electric field on it, and the actual velocity of the charged species

50
generating the fluid motion [29]. In realistic applications, this parameter does not vary

significantly.

C0 is a more critical parameter for EHD conduction, and can converted to a more useful

form using the following relations:

where τ is the time it takes an ion to traverse its mean free path without recombining with

another ion back into a neutral species, known as the charge relaxation time [30], and where tT

is the time it takes an ion to cross a distance equal to the characteristic distance between

electrodes Therefore, known as the ionic transit time. Therefore, the

converted C0 in (29) represents the ratio between these two characteristic times.

The simulation results shown in this paper are dimensional for convenience, and for

qualitative comparisons of trends with available experimental data. Therefore, the

nondimensional model is given here for the reader's reference.

4.3. Domains and Boundary Conditions

The 2D domains in which equations (18)–(21) were solved are shown schematically in Figure

4.3, with the important dimensions given and 𝑛̂ being a boundary surface’s normal vector.

51
The fluid domain, shown as the filled area, represents two parallel branches, each 7 mm long,

which receive an incoming flow from the inlet and distribute the flow between them before the

flow exits through the outlet. Note that since this is a 2D domain, these channels are not assumed

to have a circular cross section. Figure 4.3 also illustrates the inlet and outlet boundary

conditions used for the conservation of momentum and concentration equations.

Figure 4.3 Simulation Domains for the Fluid (filled shape) and Walls (patterned shapes) - Not to Scale

The reaction from equation (16) was also given as a condition in the entire fluid domain,

with an initial condition for each species concentration equal to the neutral species

concentration, neq, computed for zero applied voltage via the equilibrium relation, σ = neqeb.

This goes against the assumption that few charges are available in the liquid at zero applied

voltage, but allows the solution of the Nernst-Planck models to easily converge. Another way of

implementing the simulation would set those initial conditions at zero and add another model for

computing the concentration of the neutral combined species with initial condition neq, with the

reaction from equation (16) tying all three models together. However, the relevant results shown

herein do not change between these two methods and an additional model would only add

computation time. All other boundaries for the fluid domain were treated as walls with a no-slip

52
 condition in the conservation of momentum equation, in order to capture boundary layer effects.

The flow redistribution is performed by EHD conduction pumps containing two electrode pairs

in each branch, as shown.

The inlet and outlet were made wider than the distributive branches to eliminate any

pressure drop effects from those sections. In addition, both the inlet and outlet are extended a

length of 8 mm away from the distributive branches, in order to make sure the incoming and

outgoing flows to and from the branches are fully developed. Since the domain is 2D and the

channels are not considered circular, an analysis of the required entrance length for the relevant

flow rates was done based on a comparison between different models for flow between parallel

plates [31]. This analysis showed that, at a flow rate of 10 mL/min at the inlet, the required

entrance length for fully developed flow is 4 mm at most. Therefore, the entrance length chosen

for the simulation domain had a safety factor of 2 over the minimum entrance length.

The electrostatic distribution is computed in the solid dielectric wall domains for two main

reasons. First, the configuration being modeled represents the actual construction of EHD

conduction pumps, wherein the metallic, flush electrodes have a single wetted surface colinear

with the flow channel walls, with all other surfaces encased in an electrically insulating material

whose exterior is grounded. In addition, all exterior surfaces of the flow channel's walls are

grounded as well. Second, the value of the potential or charge distribution at solid wall boundary

of the flow channels is unknown, such that only a symmetry Neumann boundary condition can be

applied there for the electrostatics equations. The added wall domain allows for

a Dirichlet boundary to be applied at a location where the potential is known – the grounded outer

surfaces of the walls. Subsequently, the unknown charge distribution at the flow channel walls

can be computed as part of the solution. Therefore, the inclusion of the wall domain fully defines

53
the boundary value problem in a manner that better represents the real assembly of an EHD

conduction pump.

To further save on computation time, only the electrostatics equation was solved for the

wall domains. The thickness of the wall was 1 mm around the branches and 2.5 mm in the inlet

and outlet regions. The connection between the electrostatics equations of the fluid domain and

the wall domains was done via the symmetry boundary condition, as shown in the figure.

The EHD conduction pump design for each branch is shown in Figure 4.4, with the

relevant dimensions given in Table 4.1. These dimensions were chosen based on an EHD

conduction pump extensively studied by Patel and Seyed-Yagoobi [32]. Figure 4.4 also provides

the boundary conditions for the electrostatics and conservation of concentration equations for the

electrodes. Although this figure shows the electrodes as having a certain thickness, in the

simulation the electrodes were only represented by segments of the fluid domain's boundaries and

not as actual domain areas.

Figure 4.4 EHD Pump Electrodes and Boundary Conditions in One Branch - Not to Scale

Table 4.1 EHD Pump Dimensions

54
4.4. Simulation Cases
The simulated cases are listed in Table 4.2. The first set of cases considered were the

redistribution of an initially equal incoming flow rate to each branch. This was done by setting

the inlet boundary condition to a uniform velocity profile matching the quoted flow rate in the

table, which would then be split into the two branches equally. The outlet pressure was set to

zero as a reference pressure. This simulated the existence of an external pump, upstream of the

fluid domain, generating a constant flow rate. The EHD conduction pump in one of the two

branches was then activated in order to redistribute the flow.

Table 4.2 Study Cases

The next set of cases were for maldistribution recovery. The same boundary conditions

on the inlet and outlet were used here as in the prior cases, but a valve was added to the domain

in order to constrain the flow in one of the branches. This represented a situation where some

damage or manufacturing error caused one branch to have a constriction that stifled the flow.

The EHD pump in the constrained branch was then activated in order to attempt to recover an

equal flow distribution be- tween the branches.

55
 The “forward” and “reverse” configurations are shown in Figure 4.5. Activation of the

pump in the “forward” case should increase the flow rate in the active pump’s branch and

decrease it in the inactive pump’s branch, while in the “reverse” case, activation should result

in the opposite flow rate behavior. For ease of running all cases using a single domain with

parametric sweeps, the bottom branch’s pump was set in the “forward” orientation, while the

top branch’s pump was set in the “reverse” one, as shown.

Figure 4.5 The “forward” (a) Co-directional Configuration, vs. the “reverse” (b) Configuration

The simulation cases were run in COMSOL Multiphysics 5.0. The physical model

equations (18)–(21) and the boundary conditions shown in Table 4 . 2 were represented in the

software using the Electrostatics, Laminar Flow, and Transport of Diluted Species modules.

The solver allowed error on both the residual of the solution and the iteration error was set to

10−5 to ensure accurate results within a reasonable time frame. Post-processing of the results

was done via the built-in Matlab Livelink feature.

The actual geometry used for the simulation is shown in Figure 4.6. The darker region

was used as the fluid domain, while the lighter regions were used for the wall domains. For the

initial maldistribution cases, a constriction was introduced as shown in Figure 4.6b, representing

the valve. The dimensions of the constrictions were variable, so that different maldistribution

for the same inlet flow could be studied.

56
 Figure 4.6 Overview of the Simulation Domain for (a) Flow Redistribution, (b) Maldistribution Recovery

The mesh for this domain was comprised of 48,810 elements with an average element

quality of 0.67. The mesh was optimized to provide results at a resolution of 1 μm in the relevant

areas of charge transport and flow boundary layers, while utilizing coarser elements in less critical

areas to keep the computational time and memory consumption down. The mesh for the secondary

wall domain, for instance, only contains 9200 elements. This optimization process was done

incrementally, to minimize any errors related to the mesh size and resolution.

In order the have qualitative comparisons with experimental data, the numerical model was

solved dimensionally. For this purpose, the fluid properties shown in Table 4.3 were used for the

working fluid in the simulation. These properties correspond to those of HCFC-123, a common

refrigerant frequently used with past EHD conduction experiments.

57
 Table 4.3 Working Fluid Properties

4.5. Results and Discussion

This section provides an overview of the most significant results for this simulation study. These

results are divided into sections covering each family of test cases shown in Table 4.2. The primary

results include the flow rates in each individual branch for each set of cases, which showcase the

flow distribution behavior between the branches in each case. In addition, fluid flow fields are

shown in the near-electrode region of each branch, as well as overall in the domain, as appropriate.

Lastly, these flow behaviors are explained using the simulated heterocharge layer morphology for

these cases.

The first cases considered in this study were for redistribution of an initially equally

distributed single-phase liquid flow between the parallel branches. Both forward and reverse

pumping orientations were considered for these cases. In both configurations only one branch's

pump was activated, while the other branch's pump remained inactive. These cases represented a

scenario where a system may want to increase the flow rate in one branch as a response to some

58
sensor input or other condition, at the acceptable expense of the flow in another branch. Figure

4.7 shows the simulated results for initially equal flows of 0.5 mL/min, 1 mL/min and 2 mL/min

flowing into each branch. As the applied EHD voltage of the active pump is increased, the flow

distribution begins to diverge, leading to an intentionally unequal distribution between the two

branches. The maximum divergence is at 1500 V, which is the maximum voltage that was used in

the simulations. This maximum voltage corresponds to a nominal electric field strength of

1 kV/cm and is generally accepted as the maximum safe applied voltage for the electrode design

chosen here.

Figure 4.7 Redistribution of Initially Equal (a) 0.5 mL/min, (b) 1 mL/min, (c) 2 mL/min, Forward Orientation

4.5.1 Flow redistribution – forward and reverse orientation

In the case of the initial 0.5 mL/min in each branch, Figure 4.7a shows the flow in the inactive

branch reducing down to zero and then going negative. This represents the situation where the

flow in the inactive branch is completely starved of flow, when the entire flow is diverted into the

active branch, followed by flow recirculation, when the flow in the inactive branch reverses and

59
feeds into the active branch. The maximum divergence for this case was 2.4 mL/min at 1500V. In

the case of the initial 1 mL/min in each branch, Figure 4.7b shows how increasing the initial flow

rate pumped into each branch reduces the overall ability of the pump to diverge the flow, with a

maximum achieved flow divergence of 1.8 mL/min for this case at 1500V. In addition, while the

inactive branch almost reaches flow starvation in this case, at the maximum applied voltage, it

never reaches recirculation. Figure 4.7c shows the case for an initial distribution of 2 mL/min into

each branch. In this case, the trend of reduced maximum divergence continues, at only a

0.8 mL/min final difference between the branches at 1500V. In addition, Figure 4.7c shows a delay

in the onset of meaningful divergence, meaning the point at which the EHD conduction pump is

able to create a perceptible difference between each branch's flow rates, compared to Figure 4.7a

and b. In the 0.5 mL/min and 1 mL/min cases, the flow begins to diverge immediately with an

application of a nonzero voltage. However, in Figure 4.7c divergence only begins at 300V. This

shows that the incoming flow inertia has a significant effect on the performance of the pump. This

effect will be further discussed in the section on the heterocharge layer morphology.

The overall domain flow fields for the cases shown in Figure 4.7a and b are presented

respectively in Figure 4.8a and Figure 4.8b for the maximum applied voltage of 1500V. In these

figures, the flow velocity color map is overlaid with the streamlines for the flow. The arrows

represent the local flow direction, with their magnitude representing the logarithmic magnitude of

the local velocity vectors. These figures clearly show both the generation of flow in the active

branch, on the bottom branch of each image, as well as the flow direction in the inactive branch,

on the top of each image. The flow fields near the electrodes themselves are unique and are shown

in more detail further on in this section. It should be noted that the vortices seen in the inactive

branch in Figure 4.8b are primarily due to local force balances in the momentum equation between

60
the shear stress at the walls and upstream inertia. In addition, the domain was sufficiently small

that a weak electrical field was able to form between the active bottom branch's high voltage

electrodes and the top inactive branch's grounded electrodes. While this field was too weak to

affect the dissociation rates, some existing dissociated charges could migrate into the inactive

branch along it.

Figure 4.8 : Flow Fields of Forward Orientation Flow Redistribution, (a) with and (b) without Recirculation

Figure 4.9 shows the results of the simulations for both forward and reverse

configurations for the three cases from Figure 4.7 – initially equal distributions of 0.5 mL/min

(7a), 1 mL/min (7b), and 2 mL/min (7c) in each branch. From this set of results, we see that the

reverse configuration is better able to diverge the flow overall at the maximum applied voltage

of 1500V. In the 0.5 mL/min case, the forward configuration is initially able to obtain a wider

separation up to 1000V, as is evident by the “hump” shown in Figure 4.9a. The forward

configuration also achieves flow starvation in the inactive branch before the reverse
61
configuration, at 400V versus 550V respectively, and therefore initiates circulation sooner.

However, the slope of the flow rate for the forward configuration does not continue to rise as it

does for the reverse configuration as the applied voltage is increased. As shown, the reverse

configuration overcomes the performance of the forward configuration starting at 1000V. The

final divergences, the difference between the flow rates in each branch, are 2.7 mL/min for the

forward configuration and 3.0 for the reverse one. While the difference between the

configurations in this case is not very significant, the cases for the higher initial flow rates show

that the reverse configuration easily outpaces the forward one.

Figure 4.9 Forward vs. Reverse Configurations – Redistribution of (a) 0.5 L/min, (b) 1 mL/min, (c) 2mL/min

In the 1 mL/min case, the forward and reverse configuration behave nearly the same up to

850V, at which point the reverse configuration continues to significantly diverge the flow rates

between the branches while the forward configuration shows the same diminished slope as in the

previous case. Here, the forward configuration is unable to reverse the direction of flow in its

inactive branch, while the forward configuration reverses the flow in the active branch and initiates

62
flow circulation starting at 1150V. The final divergences for this case are 1.82 mL/min for the

forward configuration and 2.85 mL/min at 1500V, an increase of over 50%. In the last case shown

in Figure 4.9, for the initial distribution of 2 mL/min in each branch, the reverse configuration

categorically outperforms the forward configuration for all applied voltages. The final divergences

for this case are 0.8 mL/min for the forward configuration and 2.4 for the reverse. Therefore, unlike

the forward configuration, the reverse configuration does not seem to have a diminishing

effectiveness with increasing incoming flow rates, and actually becomes more effective with such

an increase in incoming flow.

The flow velocity fields for the forward (left) and reverse (right) configurations for each

flow rate case are shown in Figure 4.10, with increasing incoming flow rates from top to bottom.

As in Figure 4.8, the black arrows show the flow directions, and they are sized based on

a logarithmic scale of the magnitude of the velocity vectors they represent. The first thing to note

about these flow maps is the formation of vortices in the vicinity of the electrodes. For the forward

configuration case, the ground electrodes are located at x = 0.762 mm and x = 2.54 mm, while the

high voltage electrodes are located at x = 1.02 mm and x = 2.79 mm. For the reverse

configurations, the high voltage electrodes are located at x = 0.762 mm and x = 2.54 mm, while

the ground ones are located at x = 1.27 mm and x = 3.05 mm. Therefore, the vortices’ placement

and size directly correspond with both the location and lengths of the respective electrodes.

Previous studies by Yazdani and Yagoobi [11] have shown that these vortices are to be expected

for EHD conduction generated flows, however this is the first time that they are shown under the

effect of an incoming flow.

63
 Figure 4.10 Flow Velocity Fields for the Forward (a,d) and Reverse (c,d) Configurations, EHD Pump at 1.5
kV
As shown in the forward configuration cases, the first vortex from left to right circulates

the flow over the ground electrode toward the gap between the electrodes, where the flow enters

the second vortex over the high voltage electrode and accelerates downstream. The second vortex

is therefore elongated in the direction of the flow. The vortices on the subsequent electrode pair

are smaller, and the velocity magnitudes there are smaller than for the first set of vortices.

Conversely, for the reverse configuration, the first vortex from left to right is over the high voltage

electrode, and the flow is slingshot through the gap between the electrodes and back out toward

the left, upstream. The smaller, elongated vortex over the ground electrode has much smaller

velocities and primarily feeds flow into the gap between the electrodes. The vortices on the

subsequent electrode pair in this case are larger than those on the first pair, and the velocity

magnitudes there are not lessened. As the incoming flow rate is increased, the vortices over the

second electrode in the first pair (high voltage for the forward configuration, and ground for the

reverse) lose their shape and are incorporated into the main flow downstream.

The last chart for the redistribution cases, Figure 4.11, shows the magnitude of the net force

in the x direction generated by the forward and reverse configurations for several upstream, initial

64
flow rates. This chart shows the evolution of the trend at higher initial branch flow rates than

previously shown. This total force was calculated by integrating the x components of the calculated

forces on each element in the domain over the entire simulation domain, thereby accounting for

the fluid's inertia and viscous stresses in the inlet, outlet and pumping sections. Only the x-direction

forces were included since they are the most dominant, as explained in the upcoming section on

heterocharge layer morphology. As shown, the total force significantly decreases with increasing

incoming flow rate for the forward configuration, while significantly increasing with increasing

incoming flow rate for the reverse one. Of special note is the force curve for the 4 mL/min case,

where at low voltages (below 300V) the force applied on the fluid becomes negative, meaning that

activation of the EHD conduction pump actually caused a reversal of the flow in its branch. Further

discussion of the cause for this difference in behaviors between the two configurations is shown

in the section on heterocharge layer morphology.

Figure 4.11 Total x-direction Force Generated by Forward and Reverse Pumping Orientations

4.5.2 Maldistribution recovery – forward and reverse orientation

The next cases to be considered in this study were for recovery from a maldistribution condition,

65
where the single liquid phase flow is initially unequally distributed between the two branches. In

these cases, the activation of the EHD conduction pump in the branches receiving less flow is

intended to equalize the distribution between the two branches. Figure 4.12 shows the

maldistribution recovery performance of both the forward and reverse configurations, in a similar

fashion to Figure 4.9. The cases in Figure 4.12a, b, and d were all generated using the same

constriction setting on the valve shown in Figure 4.6b, while Figure 4.12c shows a case with a

tighter constriction, and therefore a larger initial maldistribution between the two branches than

in Figure 4.12b.

Figure 4.12 Forward vs. Reverse Orientation Maldistribution Recovery of (a) 1 mL/min, (b) 2 mL/min, (c) 2
mL/min with a Wider Separation, and (d) 4 mL/min Incoming Flow Rates

From the results shown here, the performances of the forward and reverse configurations

are mostly comparable when equalizing the maldistribution of the flow for low flow rates. In fact,

for the case of the lowest incoming flow rate, Figure 4.12a shows that the forward configuration

was able to achieve equalization of the flow rates in the two branches before the reverse

66
configuration - at 400V versus 550V. However, past the point of flow equalization the behavior of

the two configurations matched what was shown in the previous section for flow redistribution.

Increasing the initial maldistribution difference in the flow rates between the branches has

significant effects, as can be seen in the difference between Figure 4.12b and c. As shown, the

performance of both configurations is similar up to 1200V. However, the reverse configuration is

able to achieve flow equalization sooner, at 1400V, as opposed to 1500V for the forward

configuration. Widening the initial maldistribution in this fashion is equivalent to increasing the

incoming flow rate into the domain, and so it is expected that at higher incoming flow rates the

reverse configuration will out-perform the forward one once again, as it did for the redistribution

cases. This is confirmed by Figure 4.12d, where the forward configuration is unable to equalize

the flow at all, while the reverse one achieves equalization at 1500V.

4.5.3 Heterocharge layer morphology

To explain the behaviors shown in all the previous sections, the shape and dimensions of the

heterocharge layers over the electrodes were investigated as well. Based on the analytical solution

to the governing equations (18), (19), (20), (21) shown by Atten and Seyed-Yagoobi [10], the

force generated by each of the heterocharge layers over each electrode is a function of the charge

density within the each of the layers. To obtain the pressure generated by each electrode, one can

integrate the Coulomb force term over the area of the layer. In the simple case of infinite parallel

plate electrodes and neglecting fluid motion, this integration can be done in one dimension. If the

electrodes are located at y=0 and y=d, the heterocharge layers will extend orthogonally from each

electrode surface toward the space between the electrodes. The thickness of the heteorcharge

67
layer, λ, will then be constant and uniform across the plane of the electrodes. Therefore, the

computation of pressure can be expressed simply as,

where this height is defined as the maximum orthogonal distance from the surface of the

electrode where the charge values become the same as in the rest of the bulk of the electroneutral

fluid, or the maximum distance ions can travel while retaining their charge before recombining

back into neutral species. In this simple configuration, the net difference between λ+ and λ- , for

the positive and negative electrodes, respectively, is therefore a measure of the net force applied

on the liquid. This difference can be designated as Δλ = λ+ - λ-. However, in other electrode

configurations, such as the one used here, and especially in the presence of fluid motion, the height

of the heterocharge layer on each electrode is no longer uniform along the electrode length. Instead,

it is a function, λ(x), whose maximum along the y-axis, the axis orthogonal to the plane of the

electrodes in the configuration discussed here, can serve as a measurement for λ. In addition, while

at larger size scales the general assumption of λ≪d holds, previous research has shown that for the

size scales depicted in this paper (sub-mm scale), λ > d and the heterocharge layers are expected

to overlap [34]. In this overlap region, recombination is possible due to the presence of both

species, yet the analysis that yields the definition for λ specifies that recombination must be

negligible within the heterocharge layers [10]. As such, λ+/ λ- must not include the overlap region.

Therefore, the easiest method for obtaining λ+/ λ- is to plot the net space charge, p−n, and

measure λ+/ λ- as the maxima of the n-dominated region and the p-dominated region, respectively.

Figure 4.13 shows an example of the overlapping layers. In this figure, the net space charge

is plotted as the background, with individual λ(x) outlined for each species. The figure also shows

68
where λ+/λ- would need to be measured as the maximum of λ(x) where recombination is still

negligible. Note also the region that seems empty of the presence of either charged species in the

space between the electrodes. In this area, the species repulse each other such that in the steady

state no ions remain in that region. This matches the saturation operational regime of EHD

conduction [34], where the assumption of λ>d holds true.

Figure 4.13 Net Space Charge Surface Plot (background coloring), Overlayed with Outlines of each
Individual Ionic Species’ Maximum Travel Distance, λ(x), and Adjusted λs for the Heterocharge Layers

An understanding of how the adjusted λs are affected by flow inertia will help explain the

observed behaviors both in the simulations shown here and in experiments. Figure

4.14 and Figure 4.15 both show the profiles of the heterocharge layers for flow redistribution of

initial 1 mL/min and 8 mL/min inlet flows, with only one branch's pump activated at

1500V. Figure 4.14 shows the layers for the forward configuration, while Figure 4.15 shows the

layers for the reverse configuration. At 1500V the heterocharge layers are at their maximum

dimensions and changes to them due to flow inertia are more apparent.

69
 Figure 4.14 Heterocharge Layers for a Forward Configuration Pump Activated at 1500V for 1 mL/min (a)
and 8 mL/min (b) Inlet Flow Rates

Figure 4.15 Heterocharge layers for a Reverse Configuration Pump Activated at 1500V for 1 mL/min (a) and
8 mL/min (b) Inlet Flow Rates

The dashed rectangles along the x-axis in both figures serve as a visual representation for

the locations of the electrodes in each figure. The outlines of the layers in these figures are set at

10% of the total positive or negative charge density, according to the scale on the right-hand side

of the figures. This value was chosen since the charge density concentration drops off

exponentially away from the electrodes, such that the 10% limit contains the vast majority of the

charge in these layers. In these figures, the “first” electrode pairs is on the left hand side of each

plot, while the “second” is on the right hand side, corresponding to the order in which the main

flow passes over them. The dashed white line in both figures is set at 42 μm for all cases, to better

visualize the changes in layer peak heights.

70
 In Figure 4.14, the positive charge layer over the ground electrode of the first pair of the

pump is relatively long in the x direction and flat in the y direction for the lower flow rate, while

at the higher flow rate this layer is shorter in the x direction and taller in the y direction. The higher

flow rate appears to push the ground electrode layer backwards and upwards so its propagation

into the fluid bulk is higher. Conversely, the negative charge layer over the high voltage electrodes

in the pump's first pair is shorter at the higher flow rate than at the lower flow rate, meaning that

this layer is flattened somewhat at the higher flow rate. The same trend is seen for the second

electrode pair of the pump, but with an added effect from the additional flow generated by the first

pair. Therefore, the downstream electrode pairs in the forward configuration always have a

smaller Δλ than the upstream pairs.

Figure 4.15, on the other hand, shows the opposite trend for the reverse configuration,

where the downstream pairs always have a greater Δλ than the upstream pairs. The locations and

lengths of the maximum heterocharge layer heights are denoted using the arrows in both of these

figures – black for the layers over the ground electrodes, and white for the layers over the high

voltage electrodes. As shown, not only do the lengths of the arrows change due to the flow inertia,

but so does their position relative to the electrode, as the layers are buffeted by the flow inertia.

The changes to this propagation height shown in the above figures may seem small, but they

correspond to large changes in the applied Coulomb forces and subsequent pumping performance.

The adjusted λ+/λ- thicknesses from all the flow redistribution cases studied are shown

in Figure 4.16a for the first pair of each pump configuration, and Figure 4.16b for the second pair.

These figures show the difference between the layer thicknesses more clearly than the previous

figures, as well as the trends from Figures 4.14 and 4.15 – the difference increases for the reverse

configuration as the incoming flow increases, while decreasing for the forward configuration. In

71
fact, in Figure 4.16b, for the 8 mL/min case, the layer over the second pair's ground electrode in

the forward configuration is actually taller than that for the high voltage electrode, making

its Δλ negative. This means a negative force, in opposition to the pump's desired pumping

direction, is being generated at this pair. This serves to explain the effect of diminishing returns

for adding electrode pairs that has been seen in experiments for forward configuration EHD

conduction pumps. While the addition of pairs increases the overall pump performance, the erosion

of the heterocharge layers of subsequent pairs by the flow generation of previous pairs leads to

there being a point where additional pairs are no longer of benefit to the pump, or even actively

hampering its performance. Conversely, adding pairs to the reverse configuration does not suffer

from this issue.

Figure 4.16 Heterocharge Layer Widths over the (a) First and (b) Second Electrode Pairs

It should be noted that the layers over each electrode differ in their width in the horizontal

direction between each flow rate case, and between the first and second pair of the pump.

However, those differences are much smaller than the ones in the vertical direction. In addition,

Equation (30) does not treat differences between the layers in any direction other than the one

orthogonal to the plane of the electrodes as significant when calculating pressure generation for

an electrode pair, so these differences are not shown in this paper.

72
4.5.4 Qualitative comparison with experimental trends

Figure 4.17 shows comparisons with representative experimental data gathered for flow

redistribution and maldistribution recovery cases with similar supply flow rates to those used in

the simulations, based on the recent work of Yang et al. [23] for both forward and

reverse pumping configurations in parallel meso-scale branches. While the working fluid used in

these recent experiments was HCFC-123, it should be noted that the exact mobilities for each

ionic species in the experiment likely differ from each other, unlike in the simulated results. In

addition, since the dimensions and number of electrode pairs used in the experiment do not match

those used in the simulation, these comparisons are presented here only to show the validity of

the simulated trends.

Figure 4.17 Qualitative Comparison with Flow Redistribution Experimental Data from [23] for 0.5 mL/min
(a) and 2 mL/min (b)

For instance, in Figure 4.17a, the experimental and numerical behaviors simulated and

measured results are clearly different, but the overall trend of comparable performances between

the forward and reverse pumping orientations is the same in both. However, the experimental

data shows that the forward orientation reached a higher flow rate in the active branch, and an

73
overall larger separation between the branches' flow rates, versus the reverse trends in the

simulations. The comparison between the experimental and simulated case in Figure 4.17b shows

better trend agreement, with the reverse pumping orientation out-performing the forward

orientation in both sets of curves. The experimental and simulated forward orientations also show

a generally symmetric redistribution of the flow in this case. However, the reverse pumping

orientation's experimental data shows a significantly skewed divergence, where the active branch

achieves complete flow starvation at 1500V, which was not observed in the simulations.

4.6. Conclusions
Numerical simulations of EHD conduction driven single liquid phase flow distribution control,

were performed on a micro-scale domain, using the fully coupled EHD conduction equations. The

simulated cases included both “forward” and “reverse” pumping configurations, for when the EHD

conduction pump was acting co-directional to the incoming flow or countering it, respectively.

Flow redistribution and maldistribution recovery cases were investigated, using a variety of

inlet supply flow rates. These cases represent a comprehensive set of simulations for the

performance characterization of flush electrode EHD conduction pumps at this size scales. The

results are summarized in the following items:

• The reverse pumping configuration was shown to have better performance than the

forward pumping configuration in both flow redistribution and maldistribution recovery

cases. The two configurations were only equivalent at very low inlet supply flow rates.

• Pressure generation in an EHD conduction pump of the flush electrode variety is

proportional to the difference between the maximum propagation into the bulk, or

thicknesses (λ’s), of the heterocharge layers on each of the electrodes in a given pair. This

74
 propagation is greater the higher the wetted surface area of the electrode. For the pump

studied in this paper, this difference, and pressure generation, is therefore: Δλ = λHV - λGrnd.

• Simulation solutions for the heterocharge layer morphology provided these maximum

thicknesses on the electrodes of both forward and reverse configurations. The results

showed that the externally applied flow (the inlet supply flow) pushes the leading

electrode's heterocharge layer downstream. At the same time, the charges in this layer are

repelled by the oppositely charged layer over the second electrode nearby, forcing the

leading electrode's layer to extend further into the bulk of the liquid than it would in the

absence of an external flow.

• Conversely, the charges in the layer over the second electrode are repelled by the

advancing, oppositely charged, leading layer. In addition, the applied flow stretches the

second electrode's layer further downstream. The result is a flattened heterocharge layer

that propagates less into the bulk than it would in the absence of an external flow.

• In the forward configuration, the leading electrode of the EHD conduction pump is the

ground electrode. Increasing λGrnd while decreasing λHV in this configuration means

that Δλ, is lower than in the absence of external flow. Subsequently, the pressure generated

by that electrode pair is lower as well.

• Conversely, in the reverse configuration, the leading electrode in the first pair of the pump

is the larger high voltage electrode, and the changes in layer thicknesses increase Δλ and

the pressure generated by that electrode pair in the presence of an external flow.

The heterocharge layer formation over subsequent electrode pairs in a multi-pair pump was also

studied and quantified, with these results:

75
 • In the absence of an externally applied flow on an EHD conduction pump in the forward

configuration, the flow velocity generated by any previous electrode pairs acts as an

externally applied flow for the subsequent electrode pairs. Therefore, while additional

electrode pairs enhance the overall pressure and flow rate generation performance of the

pump, this effect introduces diminishing returns on the amount of pressure generation each

new pair adds. This effect has been observed in experiments but has not been previously

quantified and simulated.

• In the absence of a significant external flow, the difference between the forward and

reverse configurations is meaningless, since the pump is the main generator of flow.

Therefore, the same diminishing returns effect will exist in both configurations, at the same

magnitudes, albeit with different generated flow direction.

• In the reverse configuration, the externally applied flow is slowed by each electrode pair

in the EHD conduction pump. However, as long as the flow opposing the pump is not

stopped entirely, each electrode pair benefits from the aforementioned enhancements

to Δλ. Once the flow is stopped over an electrode pair, that pair will generate its static

pressure, as if it were in an enclosed loop. Therefore, while adding additional electrode

pairs to this kind of pump will not detract from its performance, not all additional pairs

will benefit from the enhancement due to the external flow.

These results serve to explain the configuration-dependent behavior observed both in this

simulation study and in previous experimental studies. Although previous experimental studies

have used significantly different electrode geometries and characteristic branch dimensions,

qualitative comparisons between the experimental results and the results of this study can be made.

76
These comparisons show similar trends between the two, adding to the credibility of the simulation

results.

Multi-pair simulations of EHD conduction pumping, such as the ones performed here,

provide a more reliable basis for assessing the effects of adding electrode pairs to a baseline pump

design than the simple linear estimates used in previous studies.

Nomenclature Table

References
[1] R.M. Fuoss, Ionic association. III. The equilibrium between ion pairs and free ions, J. Am. Chem. Soc.
80 (19) (1958) 5059.
[2] J.P. Gosse, Electric conduction in dielectrics, in: The Liquid State and its Electrical Properties, NATO
ASI Series vol. 193, 1988, p.503.
[3] A.I. Zhakin, Solvation effects in liquid dielectrics, Surf. Eng. Appl. Electrochem. 51 (6) (2015) 540.
[4] J. Seyed-Yagoobi, Electrohydrodynamic pumping of dielectric liquids, J. Electrost. 63 (2005) 861.
[5] W.F. Schmidt, K.F. Volykhin, A.G. Khrapak, E. Illenberger, Structure and mobility of positive and
negative ions in non-polar liquids, J. Electrost. 47 (1–2) (1999) 83–95.
[6] I.S. Ivanishko, V.I. Borovkov, Comparison of the mobilities of negative and positive ions in nonpolar
solutions, J. Phys. Chem. B 114 (30) (2010) 9812–9819.
[7] M.R. Pearson, J. Seyed-Yagoobi, Advances in electrohydrodynamic conduction pumping, IEEE Trans.
Dielectr. Electr. Insul. 16 (2) (2009) 424–434.

77
[8] S.I. Jeong, J. Seyed-Yagoobi, Innovative electrode designs for electrohydrodynamic conduction
pumping, IEEE Trans. Ind. Appl. 40 (3) (2004) 900–904.
[9] F. Pontiga, A. Castellanos, Electrical conduction of electrolyte solutions in nonpolar liquids, IEEE
Trans. Ind. Appl. 32 (4) (1996) 816.
[10] P. Atten, J. Seyed-Yagoobi, Electrohydro-dynamically induced dielectric liquid flow through pure
conduction in point/plane geometry, Trans. Dielectr. Electr. Insul. 10 (2003) 27.
[11] M. Yazdani, J. Seyed-Yagoobi, Electrically induced dielectric liquid film flow based on electric
conduction phenomenon, IEEE Trans. Dielectr. Electr. Insul. 16 (3) (2009) 768.
[12] M. Yazdani, J. Seyed-Yagoobi, Effect of charge mobility on dielectric liquid flow driven by EHD
conduction phenomenon, J. Electrost. 72 (4) (2014) 285.
[13] V.A. Chirkov, Y.K. Stishkov, S.A. Vasilkov, Characteristics of electrohydrodynamic pump of the
dissociation type: low and high voltage ranges, IEEE Trans. Dielectr. Electr. Insul. 22 (5) (2015)
2709–2717.
[14] R. Gharraei, et al., Experimental investigation of electrohydrodynamic conduction pumping of
various liquids film using flush electrodes, J. Electrost. 69 (1) (2011) 43–53.
[15] S.R. Mahmoudi, K. Adamiak, G.S.P. Castle, Prediction of the static pressure generation for an
electrohydrodynamic conduction pump, in: Presented at the 17th IEEE International Conference
on Dielectric Liquids, Trondheim, Norway, 2011.
[16] Y. Feng, J. Seyed-Yagoobi, Electrical charge transport and energy conversion with fluid flow
during electrohydrodynamic conduction pumping, J. Phys. Fluid. 19 (5) (2007) 5710.
[17] V.K. Patel, J. Seyed-Yagoobi, A mesoscale electrohydrodynamic driven two-phase flow heat
transport device in circular geometry and in-tube boiling heat transfer coefficient under low mass
flux, J. Heat Tran. 137 (4) (2015) 41504.
[18] S. Sinnamon, Coolant Distribution Control in Satellite Structural Panels Using
Electrohydrodynamic Conduction Pumping, PhD dissertation, Department of Mechanical
Engineering, University of New Mexico, Albuquerque, NM, 2012.
[19] Y. Feng, J. Seyed-Yagoobi, Control of liquid flow distribution utilizing EHD conduction pumping
mechanism, IEEE Trans. Ind. Appl. 42 (2) (2006) 369–377.
[20] Y. Feng, J. Seyed-Yagoobi, Control of adiabatic two-phase dielectric fluid flow distribution with
EHD conduction pumping, J. Electrost. 64 (7) (2006) 621–627.
[21] L. Yang, et al., Flow distribution control in meso scale via electrohydrodynamic conduction
pumping, IEEE Trans. Ind. Appl. 53 (2) (2017) 1431–1438.
[22] L. Yang, M. Talmor, J. Seyed-Yagoobi, Flow distribution control between two parallel meso-scale
evaporators with electrohydrodynamic conduction pumping, in: Presented at the ASME 2016
International Mechanical Engineering Congress and Exposition, Phoenix, AZ, 2016.
[23] L. Yang and J. Seyed-Yagoobi, "Liquid-phase flow distribution control in meso- scale with
directionally reversed electrohydrodynamic conduction pumping configuration.", Proceedings of
the 2017 Annual Meeting of the Electrostatics Society of America, Ottawa ON, Canada.
[24] P.H. Rieger, Electrochemistry, first ed., Champan & Hall, 1994.
[25] A. Castellanos, A. P´erez, Electrohydrodynamic Systems", Springer Handbook of Experimental
Fluid Mechanics, Springer Berlin Heidelberg, 2007, pp. 1317–1333.
[26] P. Langevin, Recombinaison et mobilites des ions dans les gaz, Ann. Chem. Phys. 28 (433) (1903)
122.
[27] L. Onsager, Deviations from Ohm’s law in weak electrolytes, J. Chem. Phys. 9 (1934) 599.
[28] O.M. Stuetzer, Magnetohydrodynamics and electrohydrodynamics, Phys. Fluid. 5 (5) (1962) 534–
544.
[29] A. Castellanos, Coulomb-driven convection in electrohydrodynamics, IEEE Trans. Electr. Insul.
26 (6) (1991) 1201–1215.
[30] F. Pontiga, A. Castellanos, Electrical conduction of electrolyte solutions in nonpolar liquids, IEEE
Trans. Ind. Appl. 32 (4) (1996) 816–824.

78
[31] R.K. Shah, A.L. London, Laminar Flow Forced Convection in Ducts: a Source Book for Compact
Heat Exchanger Analytical Data, Academic press, 2014.
[32] V.K. Patel, F. Robinson, J. Seyed-Yagoobi, Terrestrial and microgravity experimental study of
microscale heat-transport device driven by electrohydrodynamic conduction pumping, IEEE
Trans. Ind. Appl. 49 (2013) 2397–2401.
[33] J.E. Bryan, J. Seyed-Yagoobi, An experimental study of an ion-drag pump in a vertical and
axisymmetric configuration, IEEE Trans. Ind. Appl. 28 (2) (1992) 310–316.
[34] P.A. Vazquez, M. Talmor, J. Seyed-Yagoobi, P. Traor´e, M. Yazdani, In-depth description of
electrohydrodynamic conduction pumping of dielectric liquids: physical model and regime
analysis, Phys. Fluids 31 (11) (2019) 113601.

79
Chapter 5 - Influence of Temperature on the Performance of
Electrohydrodynamic Conduction Pumps
Note: This chapter is based on the following technical paper.

Talmor, M., Yagoobi, J.S., Vazquez, P.A., and Traoré, P., “Influence of Temperature on the Performance of
Electrohydrodynamic Conduction Pumps”, Physics of Fluids (to be submitted).

5.1. Introduction
Electrohydrodynamic Conduction Pumping (EHDCP) of weakly conducting liquids is one of the

most interesting practical applications of Electrohydrodynamics (EHD) [1]. Compared to

traditional pumping techniques it has several advantages: EHDCP devices have simple designs

with no moving parts, are easy to miniaturize and to control. Previous works have shown that

EHDCP works well for sizes ranging from mm to hundreds of µm [2–4], even in absence of gravity

[5]. This last feature makes these pumps especially interesting for aerospace and satellite

applications. EHDCP and similar phenomena has been studied by several groups around the world

both for fundamental studies [6,7] and applications: heat transfer enhancement [8–11], liquid

extraction [12], flow control [13–15], microfluidics [16], falling film flows [17].

EHD flows are generated by electric forces acting on electric charges present in a liquid. In

EHDCP these charges are generated by dissociation of a weak electrolyte in the liquid [18]. When

a electric field is applied in the liquid layers of net electric charge appears near each electrode,

with opposite polarity to the electrode. These are the heterocharge layers. They arise from the

balance between electric drift and recombination of the ionic species. In EHDCP the electric field

acting upon these heterocharge layers produces the motion of the fluid. Figure 5.1 shows a model

of an axisymmetric EHD conduction pump along with the placement of the heterocharge layers.

The pump is cylindrical, with the symmetry axis indicated in the figure. The high voltage electrode

is a grid that allows the liquid to go through. The grounded electrode is flushed against the

80
cylindrical wall. Due to the different size of the heterocharge layers the net electric force on the

liquid is non zero. The experimental results discussed in this paper has been obtained with a pump

of this type.

Figure 5.1 Illustration of the EHD conduction mechanism for a cylindrical EHD conduction pump. The HV
electrode allows the liquid to go through. The heterocharge layers locations are indicated in blue and red. The
corresponding polarity w/ respect to the electrode

One of the main applications of EHDCP is the enhancement of heat exchange. Therefore,

understanding the influence of the temperature of the working fluid in the performance of the

pumps is critical to optimize its design. This is true for both space and terrestrial applications,

where the devices to be cooled or exchange heat may face considerable fluctuations in their

corresponding ambient temperatures. This is the case, for example, for aerospace applications or

the cooling of high power, high temperature electronics. Another example would be a conduction

81
pump installed in a satellite which could experience a range of temperature (from low to high) that

could impact its operation.

Temperature variations on EHDCP performance has been investigated experimentally by some

researchers who have reported seemingly contradictory results. Jeong and Didion [19] observed

that pressure generation and resultant mass flow rate decreased with increased temperature.

However, Nourdanesh and Esmaeilzadeh [20] studied EHDCP of thin films of R-123 using flush

electrodes and found that the generated flow rate increases with the working fluid temperature.

This is also the case in experiments performed by the authors and discussed in this paper. Although

these are not recent experiments, to the best of the authors’ knowledge, a rigorous description of

the cause for these observations and characterizations of the effect of temperature on EHDCP have

not been presented to date.

In a previous paper [21] the authors have presented a detailed physical model of EHDCP. This

model, based on the Taylor-Melcher leaky dielectric model [22], takes into account the

enhancement of dissociation by the applied electric field, known as Onsager-Wien effect (OW)

[23,24]. It specifies also the appropriate boundary conditions to be imposed on non-metallic

substrates, needed for the study of EHDCP devices of small size. In that work the authors

performed a detailed analysis of the dependence of the generated pressure on the applied electric

field for EHDCP. We have adapted and improved this analysis to the geometry used in the

experiments reported in this paper to show that the explanation of the dependence of the generated

pressure on temperature lies in the different working regimes of the pumps. There are two limit

regimes in EHDCP: ohmic and saturation. The working regime can be characterized by two

dimensionless numbers: the conduction number C0 and a new number β introduced in that paper.

82
This number β is useful to characterize a given pump, as it does not depends on the applied electric

field. Evaluating these numbers for a given [3] pump allows to determine its working regime.

The plan of this work is as follows. In Section 2, the results from experiments performed to

study the dependence of the generated pressure on the temperature of the fluid are presented. In

Section 3, the physical model presented in [21] is presented and the model to the geometry of the

pump used in the previous section is adopted. The regime analysis taking into account the motion

of the liquid is improved. Then, in Section 4, the reason of the different behavior of the generated

pressure with temperature reported in the literature and this chapter are described. Finally, in

Section 5, the main conclusions of this work are presented.

5.2. Experimental Results

In this section, we discuss some experimental results, already partially discussed in [25]. We show

that experiments with different EHDC pumps show contradictory dependence of the generated

pressure with temperature. Although the experiments discussed here have been performed some

time ago, the explanation of this contradictory behavior has not been yet explained in the literature.

5.2.1 Experimental Study

To further investigate the effect of temperature on the EHD conduction mechanism a simple

experimental setup was constructed and tested. The experimental setup contained one macroscale

EHD conduction pump consisting of only a single electrode pair. Figure 5.2 shows a lateral view

of pump, with dimensions given in Table 5.1. In this pump, the liquid flows from left to right. The

cylindrical space where the liquid flows, that we call flow channel, is formed out of the electrodes.

The high voltage electrode was a stainless steel, porous-permeable disc with pores of 10µm in

83
diameter (Figure 5.3b). The ground electrode was a stainless steel flush ring that formed a part of

the flow channel wall (Figure 5.3a). Figures 5.3c and 5.3d show the thin and thick spacers. The

10mm flow channel is shown as the central hole in Figure 5.3a, 5.3c and 5.3d. The porous-

permeable disk in Figure 5.3b allowed the liquid to pass through. Two conductive bus-line rods

were run through the electrode pair in the small holes shown at the top of Figure 5.3a and 5.3b,

and through the spacers via similar holes shown in Figure 5.3c and 5.3d. The bus-lines were

extended beyond the sleeve so that connections could be made with the high voltage input and the

grounded setup.

Figure 5.2 : Lateral view of a EHD electrode pair arrangement. The liquid flows from left to right [25].

84
 Table 5.1 Component Dimensions for EHD Pump [25].

This electrode design generated an asymmetric electric field designed to yield a high pressure

generation based on a comparative study of different electrode geometries [26]. The high pressure

generation range was helpful for showcasing the temperature effect. Careful manufacturing and

smoothing of the electrodes minimized the existence of sharp edges in order to avoid unwanted

charge injection into the working fluid, to guarantee that EHDCP was the only mechanism present

in the setup. The spacers were made of Polytetrafluoroethylene (PTFE). This material was chosen

due to its properties as an electrical and thermal insulator. The thin separator was carefully

manufactured in order to control the separation distance between the electrodes, since it is an

important parameter to the overall pump performance due to the electric field strength being

inversely related to this distance. The flow channel diameter was 10mm and the outer diameter of

the pump assembly was 25.4mm.

The pump assembly was placed within an Aluminum 6061-T6 block with machined channels.

The block was 152.4 mm by 152.4 mm square and 50.8 mm wide. To electrically insulate the

pump from this block, the pump assembly was sheathed inside a protective PTFE sleeve 28.6 mm

in diameter that fit precisely within the channel housing the pump. The block was electrically

grounded, allowing the pump’s ground bus-line to simply be connected to the inner wall of the

channel the pump was housed in. In addition to providing grounding and leak containment, the

aluminum block served as a conductive heat exchanger between the coolant, which was pumped

via a chiller into another channel in the block, and the working fluid in the EHD pump channel.
85
Heavy insulation around this block ensured that ambient temperature conditions did not affect the

internal temperature conditions of the setup. Based on a one-dimensional radial conduction

analysis, the heat loss from the housing unit was less than 0.1W for an insulation thickness of 7cm,

using NOMEX felt insulation and still air gaps.

An OMEGA® RTD sensor was inserted into the EHD pump channel as well in order to

measure the working fluid temperature. Since only static pressure generation was of interest to this

study, the EHD pump channel was closed off with valves at either end. The pressure was measured

using a Validyne DP15-32 differential pressure transducer with a measurement range of 0-5 kPa.

National Instruments USB-6211 and USB-9219 data acquisition boards were used to transmit the

pressure, temperature, voltage monitoring and current monitoring data into a LabVIEW Virtual

Instrument program. Post processing of the data was performed via MATLAB. The maximum

experimental uncertainties associated with the pressure, temperature, voltage and current

measurements were ±5.5 Pa, ±0.2 ◦C, ±15 V and ±0.03 mA, respectively. A schematic of the

experimental setup is shown in Figure 5.4. The chiller connections and insulation were omitted for

sake of clarity.

Figure 5.3 Top view of the individual EHD pump components. The liquid flows into the page [25].

86
 Figure 5.4 Schematic of the Experimental Setup [25].

The working fluid was the refrigerant R-123, also known as HCFC-123. This fluid has been

used extensively in EHD conduction research and its properties are well documented for different

temperatures. The properties of this fluid at standard ambient conditions are given in Table 5.2.

The experiments were conducted at four different working fluid temperatures, −20◦C, 0◦C,

10◦C and 20◦C. At each temperature the EHD pump voltage was incrementally increased from

zero to its maximum operational voltage of 20 kV. This maximum was selected to be just below

the breakdown voltage of the fluid based on the size of the space between the high voltage and

ground electrodes. The pump was allowed to settle until a stable current measurement, within ±5%

variations, was observed for at least 15 s, signifying that the pump was operating in steady state.

This typically resulted in a total time step of 30 s between each voltage increase.

87
 Table 5.2 R-123 physical properties at ambient temperature

From this experimental study, the static pressure generation as well as the corresponding

electric current versus applied voltage are shown in Figure 5.5 and 5.6, respectively. These results

show that a maximum pressure generation of 1500 Pa was achieved with this single electrode EHD

conduction pump, at its maximum operational voltage of 20 kV, with a current of 0.16 mA, at a

temperature of 20◦C. The pressure and current both increased as the applied voltage increases at

all temperature levels, as was expected. However, Figure 5.5 clearly shows that the pressure

significantly increases with an increase in the working fluid temperature. In this figure, the

pressure generation at 20 kV is 39.4% higher at 20◦C than at −20◦C, for instance. This trend

matches the behavior observed by Nourdanesh and Esmaeilzadeh [20] with increasing temperature

but contradicts the observations by Jeong and Didion [19].

Figure 5.5 Comparison of Static Pressure Generation at Different Temperatures [25].

88
 Figure 5.6 Comparison of Electric Current at Different Temperatures [25].

The electric current plot in Figure 5.6 shows a generally clear trend of higher EHD currents at

higher temperature as well. The data shown for −20◦C contradicts this trend past 5 kV. However,

when considering the maximum measurement uncertainty of ±0.03mA, it is possible that the

measurement error for this case was at the higher bounds of uncertainty. In the higher voltages,

the overlap between the current values can also be attributed to this measurement uncertainty.

Nevertheless, the differences with temperature are apparent even when the uncertainty is taken

into account. It is interesting to note that, similarly to pressure generation, it was reported in [19]

that the current in their experimental setup went down with temperature.

5.3. Physical Model and Conduction Regimes

In this section we adapt the physical model and regime analysis presented in [21] to the geometry
of the pump described in Section 2.

89
5.3.1 Physical Model

We recall briefly the physical model described in [21]. The model considers a reversible

dissociation-recombination process of a neutral species into univalent positive and negative ions.

This process is described as,

Here, n+eq and n−eq are the equilibrium volume densities of positive and negative ions,

respectively. Their units are ions/m3. In equilibrium they have the same value n0eq. The

magnitudes kD0 and kR are the dissociation and recombination constants, respectively. Both type

of ions have the same ionic mobility, K, and diffusion constant D. The application of an electric

field enhances dissociation, this is the OW effect [23]. The dissociation constant becomes

dependent on the magnitude of the electric field E,

Here, F is the Onsager function and w(|E|) is the enhanced dissociation rate coefficient,

90
 Here I1 is the modified Bessel function of the first kind and order one, e0 is the elementary

electric charge, ε is the absolute permittivity of the liquid, kB is the Boltzmann constant and T is

the absolute temperature.

For univalent electrolytes, assuming that negative and positive ions have the same mobility

K, the conductivity of the liquid with no enhanced dissociation can be estimated as [21],

When the OW effect is taken into account, the conductivity depends on the magnitude of the

electric field as,

The equations of the model include the Poisson equation for the electric potential, the

transport equations for both species, and the Navier-Stokes equations for the fluid motion. The

scales for each physical magnitude are,

Here, d is a typical length characterizing the size of the system, ρm is the fluid mass density

and E0 = Φ0/d is the applied electric field, Φ0 being the applied electric potential. With these

scales the dimensionless equations are,

91
 The first term in the right hand side of equation (9) corresponds to the dissociation of ionic

species, described by the dissociation constant in (2). The Onsager function, written in

dimensionless way, becomes,

The second term in the right hand side of (9) describes the recombination of ionic species.

This term is the same for both positive and negative species. The last term in (10) is the Coulomb

force acting on the liquid. It is the only relevant EHD force in this system. The dimensionless

parameters in these equations are,

92
 The conduction number C0 can be understood as the ratio of ionic transit time and the ohmic

recombination time. In the liquids used in EHDCP only a small fraction of the electrolyte

molecules are dissociated. Hence, the experimental value of the ions density, n0eq, is not easily

accessible. Using (4) an alternative expression of C0 can be derived,

The diffusion number α describes the relevance of charge diffusion. The electric Reynolds

number, ReE, is a Reynolds number built with the ionic drift velocity. The mobility number M

is the hydrodynamic mobility [29]. It depends only on the physical properties of the fluid. The

Onsager number O characterizes the OW effect. As it is discussed in [21], it can be written as,

When O ≥ 1 the enhanced dissociation by the electric field becomes important. This allows

to estimate the value of the electric potential where the OW effect is relevant. For the

experimental setup discussed in Section 2 it is,

We have used the lowest and highest value of temperature used in the experiments and the

values,
93
of the physical magnitudes of the working liquid given in Table 5.4.

The number β depends only on operational values of the EHDCP device. In this way, the

Onsager function can be computed as,

The parameter β, along with C0, can be used to identify the operating regime of a given EHD

conduction pump.

5.3.2 Analysis of the Temperature Equation

In this model the fluid is assumed to be isothermal. In order to justify this, let us consider the

temperature equation. For non-polar liquids it is [1],

Here, cp is the specific heat at constant pressure, κ is the thermal conductivity, J’ is the current

density seen by an observer moving with the flow and Φv is the viscous dissipation. It is,

where τ = µ(∇u+∇uT) is the viscous stress tensor. The second term of the left hand side in

(18) represents the advection of temperature by the fluid while the terms on the right hand side

94
describe diffusion, Joule heating and viscous dissipation, respectively. We can define typical

times associated to each one of these processes,

Here, d is the typical size of the system, U0 is the typical velocity of the fluid, κD = κ/ρcp is

the thermal diffusivity and E0 is the typical value of the applied electric field. Time τu is the time

that a parcel of fluid takes to travel across the EHD pump, τd is the time that a difference of

temperature takes to diffuse, and τJ and τv are the typical time that Joule heating and viscous

dissipation take to increase the temperature of the fluid by ∆T. The latter is always much higher

that the others by several orders of magnitude. Thus, viscous heating is negligible. Table 5.3

shows the typical times obtained from the data in Jeong and Didion [19] and Nourdanesh and

Esmaeilzadeh [20]. The values of τJ and τv are computed fixing ∆T = 1◦C. This is the order of

magnitude of the highest temperature change in the experiments. We can see that the advection

time τu is always the smaller one by at least a factor 6. Then, the liquid does not stay long enough

in the pump section to increase significantly its temperature. We will assume that, on these

experiments, the liquid remains isothermal, and the electric properties of the liquid take the

values corresponding to the working temperature. Let us stress that this paper consider EHD

conduction pumps configurations where there is no heat source or sink. If this were the case the

dependence of the properties of the fluid with temperature should be included in the model,

along with the energy equation.

95
 Table 5.3 Typical values of physical properties along with the ratios of typical times to the advection time in
Jeong and Didion [19] (1) and [20] (2). Here, τJ and τv are computed taking ∆T = 1◦C. In both cases cp ≈
2×103 J/kg·K.

5.3.3 Conduction Regimes

In [21] the authors discuss in detail the different regimes in EHDCP and the dependence of the

generated pressure on the dimensionless numbers in each regime. There are two limit regimes

in EHDCP: ohmic and saturation. In the ohmic regime the ions recombine in the liquid before

they can reach the electrodes. In this case the structure of the charge distribution is given by two

heterocharge layers near the electrodes and an electroneutral bulk. As it is discussed in [21], in

this regime the heterocharge layers are boundary layers with typical dimensionless thickness

given by λH ∝ 1/C0pF(w(E0)), with the functions F and w defined in (3). In the saturation regime,

the ions typically get to the electrodes before they can recombine. There, they neutralize

interchanging electrons with the metallic electrodes. In this case the two heterocharge layers are

not boundary layers. They overlap and there is no electroneutral bulk. In this regime, in the limit,

the liquid can get almost completely depleted of ions.

5.3.3.1 Computations with the Liquid at rest

In [21] the regimes were characterized numerically studying a simple assembly of two parallel

infinite electrodes. The electric force in half the domain was computed assuming that the liquid

was at rest. We have performed the same computation for the pump used in the experiments

described in Section 2.1 We solve the set of equations (7)-(9) taking out the contribution of the

fluid velocity in (9). The computational model shown in Figure 5.7 reproduces the configuration

96
depicted in Figure 5.1. The left border corresponds to the axisymmetric axis. The lengths are of

the different elements are taken from Table 5.1, scaled with L2. The boundary conditions at each

type of surface are,

Figure 5.7 Axisymmetric computational domain for the simulations with no fluid motion. The lengths
correspond to the pump described in Section 2.1, scaled with L2.

Here, n is the outward unit normal vector.

The function of the small notches at both extremes of the grounded electrode is to avoid the

singularity that would appear when computing the electric field if a straight line would be used.

This is a similar to what is done in [30]. The dimensional radius of curvature of the notches is

100µm. The results computations are not affected by this value as long as it is not too small. The

figure is not at scale.

97
 The computations presented in this paper have been done with COMSOL Multiphysics,

which uses the finite element method. The details of the numerical schemes are described in

[21]. The mesh was made of triangular elements and it was thinner near the bottom, top and right

boundaries in,

Figure 5.8 The dimensionless electric force along the axis generated in the cylindrical pump vs. C0 for several
values of β assuming that the liquid at rest, and the dependence of the dimensional force vs. applied electric
field.
In Figure 5.8, the dots correspond to the electric force computed for the configurations

corresponding to the temperatures used in the experimental results. Below the dashed line the

Onsager number is less than one, and it is greater than one above. Grid independence tests have

been conducted to determine the most appropriate mesh for the computation. In the case of the

results shown in Figure 5.8 the mesh was refined until the differences of the typical values of

computed electric forces where of the order of 1%.

The first plot in Figure 5.8 represents the dimensionless longitudinal electric force as a

function of C0 for several values of β. The dimensionless value of α needed in (21) is α = 10−6.

The electric force is computed as,

98
 This figure is quite similar to the first plot in Figure 5.9 in [21]. For β = 0 (no OW effect)

the electric force is Fz in the ohmic regime (C 1), while it is Fz ∝ C0 in the saturation

regime (C For β > 0 the OW effect is not null. The dashed line corresponds to the values

of electric force computed for the value of C0O = β2 where the Onsager number is 1 (see (15)).

Therefore, in the region below this line the OW effect is not important, while it becomes relevant

in the region above.

For higher values of C0 (lower values of E0) the pump is in the ohmic region for all values of

β. However, when C0 becomes smaller (corresponding to E0 becoming higher) the behavior is

different for different values of β. For β = 1.0 the electric force dependence onC0 is similar to

the one corresponding to the ohmic regime, for all values of C0. The explanation is that, even if

the ions are able to get to the opposite electrode before recombining, the OW effect is able to

create enough ions and, therefore, there is always an electroneutral bulk between the electrodes.

When β is small (β = 0.1) the OW effect is not able to replenish the liquid with ions and a

saturation regime is observed when C0 decreases. However, for C0 small enough (E0 high

enough) the OW effect becomes important again and an ohmic like regime is attained. That is

why the curves in the left plot go upwards for C0 small enough. The value of C0 where the curves

turn up decreases with β. For intermediate values of β the curves diverge from the ohmic regime,

but the OW effect forbids the appearance of a saturation regime.

It is important to point out that not all the points of each curve for a given value of β are

accessible in the experiments. If E0 becomes too high (over 15Mv/m), charge injection from the

99
electrodes occur, and the physical model is no longer valid. In the experiments, high values of β

usually imply higher values of C0, and the pumps work in the ohmic regime. Small values of β

correspond to small values of C0, and the pumps work in or near the saturation regime.

The second plot shows the evolution of FzC0−1 vs. C0−1. The former magnitude is proportional

to the dimensional electric force, Fzdim, while the latter is proportional to the applied electric field

E0. In the ohmic regime the electric force scales as E02, while in the saturation regime it saturates

as E0 increases. The behavior described for the first plot can be seen here again.

In both plots the dots are obtained from computations performed with the values of the

dimensionless parameters corresponding to the experimental setup described in Section 2.1.

These points will be relevant in the understanding of the experimental dependence of the

generated pressure with temperature to be discussed in Section 4.

Figure 5.9 shows the non-dimensional charge distribution for three representative states. All

the simulations were made with β = 0.1. In the first plot it is C0 = 4, corresponding to the ohmic

regime. The two hetercharge layers area clearly separated by an electroneutral bulk. In the

second plot it is C0 = 0.1. This corresponds to an intermediate regime. The heterocharge layers

are larger and they overlap slightly. In the third plot it is C0 = 0.002. From Figure 5.8 we see that

this value is entering the saturation regime, right before the OW effect becomes important

regime. There is no clear structure of heterocharge layers, although, due to the geometry, it is

not the same distribution regime that the one obtained in the saturation regime in [21].

100
 Figure 5.9 Dimensionless charge distribution for β = 0.1. In the first plot it C0 = 4 (ohmic regime), in the
second plot it is C0 = 0.1 (intermediate regime), in the third plot it is C0 = 0.002 (saturation regime).
Computations performed with the liquid at rest.

5.3.3.2 Influence of the Motion of the Liquid

The simulations in the previous section were performed with the liquid at rest. However, the

structure of the heterocharge layers could be altered by the velocity of the liquid. In order to

address this point, we have performed a set of simulations including the liquid motion. We

consider the full set of equations (7)-(11). The computational domain is modified with the

addition of a chamber at the top. This is done to reproduce more closely the experimental setup.

The new computation domain is shown in Figure 5.10. In the experiments discussed in Section

2 the high voltage electrode is a porous disk that allows the liquid to go through but introduces

a high hydraulic resistance. We have not taken into account the pressure drop induced by this

porous electrode. Therefore, the conclusions obtained from these computations can only be

qualitative. We compute the electric potential and the ionic species only (A). Therefore, the

electric force acts only in this domain. The velocity of the fluid is computed in the whole domain

(A∪B). The boundary conditions are depicted in Figure 5.10. The values of the parameters ReE

101
and M corresponds to the values computed with the properties of the fluid corresponding to the

highest temperature in Table 5.4.

Figure 5.10 Axisymmetric computational domain for the simulations taking into account the motion of the
liquid. The lengths correspond to the pump described in Section 2.1, scaled with L2.
Figure 5.11 plots the same magnitudes than the first plot in Figure 5.8. Both plots are quite

similar for higher values of β. For C 1 we observe the ohmic regime for all values of β. As

β decreases, a saturation regime is attained for smaller values ofC0, although for small enough

values of C0 the OW effect becomes important enough to create enough ions to recover the

electorneutral bulk. Again, in the region below the dashed line the Onsager number is O < 1,

while it is O > 1 above. When the OW effect become important for lower values of β the lines

curve up, due to the reappearance of two separated heterocharge layers.

102
Figure 5.11 Dimensionless electric force vs. C0 when the liquid motion is considered, for different values of β.
The dots correspond to the electric force computed for the configurations corresponding to the temperatures
used in the experimental results.

However, the transition zone between the ohmic and saturation regimes is quite different for

low values of β. Figure 5.12 shows, the net electric charge in the stationary state, computed as

qt = n+ −n−, along with the stream lines for β = 0.1 and three different values of C0. In the first

plot it is C0 = 2, in the ohmic regime. The two heterocharge layers are well defined in both

electrodes. The positive heterocharge layer is slightly disturbed by the motion of the fluid in its

central part, but the overall structure is not affected. The flow pattern consists of four rolls, one

in the top chamber, another one around the HV electrode, a third one in front of the grounded

electrode and the last one at the bottom. In the second plot it is C0 = 0.1, in the intermediate

regime. The structure of the heterocharge layers is somewhat different, specially near the HV

electrode. Correspondingly, the flow pattern is different from the previous one, with one big top

roll, a small roll at the bottom and two smaller counter rolls near the grounded electrode. Even

more, in this region it is possible to found oscillatory patterns with no steady solution. In the

third plot it is C0 = 0.01 in the saturation regime before the OW effect becomes relevant. The

103
heterocharge layers have almost disappeared and the flow pattern has changed again. The

different structures of the heterocharge layers produce different patterns of electric force and

this explains the behavior of lines for lower values of β in Figure 5.11. For higher values of β

the OW effect keeps the heterocharge layers structure similar to the one in ohmic case.

Figure 5.12 This plot shows the net electric charge computed as qt =n+ −n−, and the streamlines for β = 0.1
and three different values of C0. The first plot C0 = 2, ohmic regime. The second plot C0 = 0.1, intermediary
regime. The third plot C0=0.01, saturation regime.

As it is the case in Figure 5.8, the dots in Figure 5.11 are obtained from computations

performed with the values of the dimensionless parameters corresponding to the experimental

described in Section 2.1. Therefore, for the experiments described the relevant line is still the

one corresponding to β = 1.0.

104
5.4. Dependence of the Generated Pressure on the Temperature
In this section we use the regime analysis performed in Section 3 to explain the different

dependence on temperature of the generated pressure reported in different experiments.

5.4.1 Influence of Temperature in the Dissociation-Recombination Process

In order to explain the different effect of temperature on the generated pressure in [19] and the

experimental results shown in Section 2.1, we discuss the influence of temperature on the

dissociation-recombination process. In particular, we will show that the conduction number C0

decreases with temperature, at least in the explored range of temperatures.

The parameters C0 and β determine how the generated pressure changes with the applied

electric field. The equilibrium value of C0 for the pump must be computed based on the starting

temperature conditions of the intended application in order to determine the regime the pump

operates in. However, for the variation of C0 with temperature, the change in the parameters for

the expression (14) with temperature is not significant enough to explain the dramatic effects

observed in the published experimental data and in the experiments reported in this work. For

instance, Table 5.4 summarizes the change of the relevant properties with temperature for R-

123, the fluid used in the experimental portion of this paper. Calculating the ratio of C0 values

for two temperatures using interpolated data from Table 5.4 yields that C0 decreases by 7% with

a rise in temperature between −20◦C and 20◦C. The decrease shows the correct general trend

which matches the overall behavior of the experimental studies. However, as it has been

demonstrated in the experimental study in Section 2.1, a change of 7% is much smaller than the

measured pressure generation performance change for this working fluid in that temperature

105
range. Similar information regarding the change in the electric conductivity of fluid R-134a at

the relevant temperature range is unfortunately unavailable in the literature, so a similar analysis

cannot be performed for the Jeong and Didion study [19]. Nevertheless, this underestimation of

the change in C0, and therefore pressure generation, versus what was observed in this study has

sparked an interest in taking a closer look at the assumptions that were made in the transition

from the original definition of C0 shown in (13) to the approximated expression in (14).

Table 5.4 R-123 Temperature Variation of Properties.

One problem of evaluating C0 using (14) is that the experimental value of the ionic mobility

K is usually not known. The true values of ionic mobility can only be obtained via careful

measurements of each ionic species present in the specific dielectric working fluid for the range

of relevant temperatures. This is a tedious task that is rarely undertaken by manufacturers. The

solution is to estimate the value of K through the Walden Rule [31]. Assuming that the electric

force acting upon an ion and the drag force described by the Sokes Drag are balanced we get

[32],

106
 where RH is the radius of the sphere representing the ionic atmosphere of the ion, µ is the

dynamic fluid viscosity, and v0 = KE is the constant drift velocity attained by the ion. Solving

for the mobility yields the traditionally used formula,

Here, the value RH ' 10−9 m has been used. However, Walden’s Rule has been shown to have

less relevance in the case of dielectric working fluids [33], and it is generally agreed that in real

applications involving dielectric liquids Walden’s Rule can only be used to provide crude order

of-magnitude assessments of the actual value of the ionic mobility [1]. Since the pressure

generation performance of an EHD conduction pump is a function of C0, and therefore a function

of the ionic mobility, knowing the value of the mobility only to within an order of magnitude is

insufficient to accurately predict how pressure generation performance will be affected by

temperature. Thus, a more fine-tuned assessment of the change in C0 and pressure generation is

required.

Instead of using the approximation in (14), we turn to the original definition of C0 shown in

(13). This form is based on the ionic concentration and the bulk permittivity and does not

involves the ionic mobility. The change of permittivity with temperature is easy to measure.

Therefore, we examine the behavior of the density of ions. In weak electrolytes this magnitude

is not directly measurable, as it is the case in strong electrolytes. However, we can deduce its

relative change with temperature. The equilibrium value of the concentration is obtained from

(1). The concentration of neutral species in the dielectric fluid is a constant under the assumption

that fluid expansion within the relevant temperature range is negligible. According to published

107
data [34], between -20 ◦C and 20 ◦C the liquid volume of R-123 and R-134a will expand 6% and

10%, respectively. However, the reduction in the value of concentrations induced by this

expansion is offset by the exponential increase of the OW effect, as it will be shown below. The

recombination rate constant and the dissociation rate constant of the equilibrium reactions are

obtained from the kinetic theory of ionic collisions [35],

In the above equations a is the minimum distance of approach between two oppositely

charged ions that leads to a collision and subsequent recombination, and U is the activation

energy of the reaction [35],

Equation (25) shows also the upper bound value for the recombination rate constant given

by Langevin [36], where it was assumed that e−U/kBT → 0. If the mobilities are considered equal

for both ionic species, we obtain the expression used in the model discussed in Section 3.

The OW effect does affect dissociation, not recombination. Applying (1) and (2) when there

is an electric field we get,

108
 The constant A = (3c0/4πa3)1/2 does not depend on temperature. The focus of this work is on

the change of the value of the concentration of ions between different temperatures,

r = neq(T1)/neq(T2). That is, we do not need to know the value of neq, only its relative change.

Then, the specifics of this constant are not important.

Equation (28) does not depend on ionic mobility at all and captures the full effect of

temperature on the density of ions while accounting for the field enhanced dissociation rate. The

dependence with temperature is driven by two physical magnitudes, the temperature T itself and

the permittivity ε(T). Both in [19] and in the experimental setup described in the precedent

section, ε(T) can be approximated as,

In the Onsager function the combination εT2 appears in the denominator of w. From (29),

this combination increases with T in the explored interval of temperature. Then, the value of the

Onsager function decreases always with temperature. In the exponential in (28) the combination

is εT. Given the value of γ in (29), it can be shown that, in the range of temperature in the

experiments, εT decreases with temperature. Therefore, the exponential is also a decreasing

function of T. Hence, the value of neq is a decreasing function of T. Physically, this is due to the

fact that when ε decreases the attraction between the ions of the electrolyte increases, and there

is less dissociation. When considering C0 in (13), the reduction of neq is exponential and

dominates over the slight reduction of the value of ε in the denominator. Thus, C0 decreases

when temperature increases. Figure 5.13 depicts the described dependence on temperature. For

the equilibrium density only, the relative change can be computed. We have taken for the

reference neqref the value corresponding at the lower temperature in Table 5.4. In order to

109
compute the values of C0, we have used (14) to compute C0ref, the value corresponding to the

lowest temperature. For the other temperatures we have used (13) to obtain,

Figure 5.13 Computed values of neq and C0 computed from experimental data from Table 5.4. Both
magnitudes are scaled by their values for the first temperature.

5.4.2 Comparison of Theoretical Model with Experimental Results

Computing the values of C0 and β for the experiments described in Section 2.1 we get β ≈ 1 and

C0 ∈[0.14,0.9], approximately. From Figure 5.8 we conclude this pump operated in the ohmic

regime. The dots in Figure 5.8 are obtained computing the electric force with no fluid motion

using the values of the properties of the fluid corresponding to the four temperatures in Figure

5.5. The dots in Figure 5.11 are computed in a similar way but taking into account the motion

of the liquid. In both cases, although these simulations are performed with different values of

temperatures, the computed electric force follows well the curve corresponding to β = 1. In the

110
ohmic regime Fz scales as C . Then, C0 decreases with increasing temperature, and, therefore,

Fz increases. This also explains the measured increase in current with temperature in this regime,

since a smaller value of C0 implies a smaller ratio of the ionic drift time and the ohmic

recombination time. Then, more ions are able to reach the electrodes before recombining.

The geometry of the EHD pumps used in Jeong and Didion [19] was different than the pump

used in the experiments discussed in Section 2. However, the similarity between Figure 5.8 and

Figure 5.9 in [21] shows that the conclusions from the regime analysis are quite general. Based

on the interelectrode spacing, the maximum applied voltage and the working fluid properties at

25◦C quoted in the study by, in their experiment they had β ≈ 0.03 and C0 ≈ 0.02. In this case Fz

decreases with C0. We have seen that increasing the temperature decreases the value of C0. Then,

the observed decrease of pressure with temperature in this regime is reasonable.

For sake of completeness, in the experiments described in Nourdanesh and Esmaeilzadeh

[20], they had β ≈ 1.7 andC0 ≈ 3.4. Then, their setup worked in the ohmic regime, and the

dependence of the generated pressure on temperature is similar to the one observed in our

experiments.

Some comments must be made about the effect of the temperature on the viscosity. From

Table 5.4, the viscosity decreases 50% approximately between the two extreme values of

temperature used in our experiments. It is true that this could affect the maximum velocity of

the flow. In order to take account of this aspect, a simulation of the complete geometry should

be performed, including the pressure drop induced by the porous electrode. However, the overall

structure of the heterocharge layers should not change and, as a consequence, our conclusions

about the dependence of the electric force on temperature will be valid.

111
 We have shown that our model explains qualitatively the observed dependence of the

generated pressure on the temperature. In order to get a quantitative explanation, there is still an

unknown parameter, the typical distance between ions, a. We can explore the influence of this

parameter on the value of the generated pressure. Our experimental setup works always in the

ohmic regime. In these conditions, while a change of temperature implies changes both in β and

C0, the change in β is not relevant, as in this regime the generated pressure depends only on C0

(see Figure 5.8). Figure 5.14 shows a plot of the experimental data from Section 2.1 (the

geometric shape symbols), along with the regions of prediction for pressure generation for each

temperature lower than 20◦C (shaded regions), based on values of minimum distance of approach

in the range of 6−10𝐴̇, and the dependence of generated pressure with C0 in the ohmic regime.

As shown in this figure, the upper bounds of the regions of prediction for each temperature

correlate to 10𝐴̇ whereas the lower bounds correlate to 6𝐴̇. The values for pressure generation

at 20◦C were therefore used as the baseline data, and the predicted values were calculated only

for the lower temperatures. These predicted values clearly show the variation in performance

due to the change in temperature, but also follow the trend of enhancing the change in

performance with increasing applied potentials that was previously discussed. Fewer values

were used in this figure out of the original experimental data shown in Figure 5.5 for the sake of

clarity. Most notably, Figure 5.14 shows that for the value 8𝐴̇ (solid lines) the prediction

matches remarkably well with the experimental data, with the largest error being 10%. As a

specific example, the predicted change in pressure generation performance between −20◦C and

20◦C at an applied potential of 20kV and using 8𝐴̇ was 41%, only 0.6% different from the

experimental value.

112
 Similar calculations and plots were made for the data from [19] and are shown in Figure

5.15. From Figure 5.8, and the value of β computed above, the non-dimensional generated

pressure is very approximately proportional to C0. Once more the experimental data (the

geometric shape symbols) is shown alongside the predicted change in pressure generation

performance for minimum distances of approach in the range of (6-10𝐴̇) (shaded regions). In

this figure, the upper bounds of the regions of prediction correlate to 6𝐴̇ while the lower bounds

correlate to 10𝐴̇ as shown. This reversal is due to the fact that the experiments in Figure 5.14

and 5.15 operated at different C0 regimes and their temperature based behavior is reversed from

each other, as previously explained. As was done for Figure 5.14, the data for 20◦C was used as

the baseline data and the pressure performance was predicted only for the lower temperatures.

Interestingly, using the 8𝐴̇ value for the minimum distance of approach as was used in Figure

5.14 still yields a remarkably good match between the predicted change in pressure generation

performance (solid lines) and the experimental data, with the largest error being 12%. One

apparent difference that can be noted between Figure 5.14 and 5.15 is that the predicted ranges

for each temperature in Figure 5.14 are narrow and do not overlap, unlike in Figure 5.15. This

is due to the change in permittivity for fluid R-134a being 32.5% between −20◦C and 20◦C,

versus 19.6% than it is for R-123 over the same temperature range. Hence, the less the change

in permittivity with temperature for the working fluid, the easier it would be to determine the

predicted pressure generation performance for that fluid as a result of temperature changes.

113
 Figure 5.14 Experimental vs. predicted pressure generation for R-123 with C0 = 3 and β = 3. Solid lines
correspond to a = 8A ̇

5.5. Conclusions
This fundamental study of the effect of temperature on the pressure generation performance of

EHD conduction pumping has allowed us to understand previously unexplained experimental

results. First, we have discussed experiments with a EHD pump where the generated pressure

increases with temperature. This agrees with experiments presented in [20], but it disagrees with

those discussed in [19]. Then, we have adapted the physical model and the analysis regime from

[21] to the geometry used in our experiments. In the ohmic regime there is always two

heterocharge layers next to the electrodes and an electroneutral bulk, while in the saturation

regime the heterocharge layers overlap and there is no electroneutral bulk. The working regime

of a pump is characterized by the dimensionless numbers β and C0. The number β characterizes

a given pump, independently of the applied electric potential. For β of order 1 or greater the

pumps always work in the ohmic regime. For β small they operate in or near the saturation

114
regime. In the ohmic regime the dimensionless electric force acting upon the liquid is

proportional to C0−1, while it is proportional to C0 in the saturation regime. This is true even if

the motion of the liquid is taken into account in a closed cavity.

Figure 5.15 Experimental [19] vs. predicted pressure generation for R-134a, with C0 = 0.02 and β = 0.06.
Solidlines correspond to a = 8A ̇

The usual way to compute the value of C0 requires using the Walden Rule to get an

approximate value of the ionic mobility. However, this procedure does not produce accurate

values and, in particular, does not allow to characterize accurately the change of C0 with

temperature. We have shown that it is possible to evaluate the change of the density of ions in

the liquid, and from it the number C0, without using the value of the ionic mobility. We have

shown that the value of the density of ions decreases with the working temperature, and as a

consequence the value of C0 decreases. Along with the value of the number β, this allows to

determine the operational regime of the pumps and explain the observed behavior of the

generated pressure with temperature in our experiments and in the others works discussed in the

paper.

115
 Another conclusion is that it was possible to induce a change in C0 and β in such a fashion

that the behavior of the EHD conduction pump will change from one regime to another, causing

pressure performance to change its trend of behavior. This can be done by changing the working

fluid’s properties, varying the distances between the electrodes, and more. Such a change in

regime can even occur at different applied voltages if the pump is designed with C0 ∼ 1. If this

effect is not desired, future designs for EHD conduction pumps should take into account the full

range of C0 and β values for their pumping device, accounting for all potential changes in

operational parameters and environmental conditions, such as temperature.

This work can be extended furthermore. An obvious improvement would be to solve the

energy equation coupled with the Navier-Stokes and EHD equations in order to compute the

temperature. We could then take into account the dependence on temperature of the properties

of the hydrodynamic and electric properties of the liquid in the physical model. In this way we

could deal with pump configurations with heat sources or sinks present. Also, the exact value

for the minimum distance of approach between oppositely charged ions depends not only on the

properties of the fluid and the ionic species, but also on the specifics environment conditions

that affect the kinetics of the dissociated ions. Obtaining the true values of this parameter for all

ionic species in all EHD conduction working fluids would therefore require extensive further

research that is not a part of the current study. Such an effort would be worthwhile for obtaining

a more accurate predictive tool for the change of EHD conduction pressure generation with

temperature for any given EHD pump and a given working fluid.

116
References
[1] A. Castellanos, ed., Electohydrodynamics (Springer-Verlag, 1998).
[2] S.-I. Jeong and J. Seyed-Yagoobi, IEEE Transactions on Industry Applications 40, 900 (2004).
[3] M. A. W. Siddiqui and J. Seyed-Yagoobi, IEEE Transactions on Industry Applications 45, 3
(2009).
[4] M. R. Pearson and J. Seyed-Yagoobi, Journal of Electrostatics 69, 479 (2011).
[5] V. K. Patel and J. Seyed-Yagoobi, Journal of Heat Transfer 137, 41504 (2015).
[6] I. A. Ashikhmin and Y. K. Stishkov, Surface Engineering and Applied Electrochemistry 50,
246 (2014).
[7] V. A. Chirkov, Y. Stishkov, and S. A. Vasilkov, IEEE Transactions on Dielectrics and
Electrical Insulation 22, 2709 (2015).
[8] D. Testi, F. D’Ettorre, D. Della Vista, and W. Grassi, International Journal of Thermal
Sciences 117, 1 (2017).
[9] N. Nourdanesh, S. Hossainpour, and E. Esmaeilzadeh, Applied Thermal Engineering 157, 113711
(2019).
[10] M. Mirzaei and M. Saffar-Avval, Experimental Thermal and Fluid Science 93, 108 (2018).
[11] N. Nourdanesh, S. Hossainpour, and K. Adamiak, Applied Thermal Engineering 180, 115825
(2020).
[12] Ryoichi Hanaoka, Yoji Fujita, Takuma Kajiura, and Hidenobu Anzai, Journal of Energy and
Power Engineering 13, 229 (2019).
[13] W. Grassi, D. Testi, and D. Della Vista, Journal of Physics: Conference Series 501 (2014),
10.1088/1742-6596/501/1/012006.
[14] L. Yang, M. Talmor, B. C. Shaw, K. S. Minchev, C. Jiang, and J. Seyed-Yagoobi, IEEE
Transactions on Industry Applications 53, 1431 (2017).
[15] Z. Sun, D. Sun, J. Hu, P. Traoré, H. L. Yi, and J. Wu, Journal of Electrostatics 106 (2020),
10.1016/j.elstat.2020.103454.
[16] T. Sato, Y. Yamanishi, V. Cacucciolo, Y. Kuwajima, H. Shigemune, M. Cianchetti, C. Laschi,
and S. Maeda, Chemistry Letters 46, 950 (2017).
[17] A. Sobhani, S. Nasirivatan, R. Gharraei, and E. Esmaeilzadeh, Journal of Electrostatics 73, 71
(2015).
[18] P. Atten and J. Seyed-Yagoobi, IEEE Transactions on Dielectrics and Electrical Insulation 10,
27 (2003).
[19] S.-I. Jeong and J. Didion, Journal of Thermophysics and Heat Transfer 22, 90 (2008).
[20] N. Nourdanesh and E. Esmaeilzadeh, Applied Thermal Engineering 50, 327 (2013).
[21] P. A. Vázquez, M. Talmor, J. Seyed-Yagoobi, P. Traoré, and M. Yazdani, Physics of Fluids 31,
113601 (2019).
[22] D. a. Saville, Annual Review of Fluid Mechanics 29, 27 (1997).
[23] L. Onsager, Journal of Chemical Physics 2, 599 (1934).
[24] S. A. Vasilkov, V. A. Chirkov, and Y. K. Stishkov, Physics of Fluids 29, 063601 (2017).
[25] L. Yang, M. Talmor, and J. Seyed-Yagoobi, in 2014 IEEE 18th International Conference on
Dielectric Liquids (ICDL) (IEEE, 2014) pp. 1–4.
[26] M. R. Pearson and J. Seyed-Yagoobi, IEEE Transactions on Dielectrics and Electrical
Insulation 16, 424 (2009).
[27] J. E. Bryan, Fundamental Study of Electro-hydrodynamically Enhanced Pool and Convective
Boiling Heat Transfer, Ph.D. thesis, Texas A& M University, College Station, TX (1998).
[28] DuPont, “HCFC-123 Properties, Uses, Storage, and Handling,” (1998).
[29] N. J. Felici, Revue General d’Electrostatique 78, 717 (1969).
[30] A. Jawichian, L. Davoust, and S. Siedel, Physics of Fluids 32 (2020), 10.1063/5.0010358.
[31] J. Lyklema, Fundamentals of Interface and Colloid Science: Solid-Liquid Interfaces (Academic
Press, 1991).

117
[32] P. H. Rieger, Electrochemistry, 1st ed. (Champan & Hall, 1994).
[33] J. P. Gosse, NATO ASI Series 193, 503 (1988).
[34] E. Lemmon, M. Huber, and M. McLinden, “NIST Standard Reference Database 23: Reference
Fluid Thermodynamic and Transport Properties-REFPROP,” (2013).
[35] A. Castellanos and A. T. Perez, in Springer Handbook of Experimental Fluid Mechanics
(Springer, 2007).
[36] P. Langevin, Annales de chimie et de physique 28, 223 (1903).

118
 Appendix
List of Publications

Journal Publications
Talmor, M. and Seyed-Yagoobi, J., “Numerical Study of Micro-scale EHD Conduction Pumping: The
Effect of Pump Orientation and Flow Inertia on Heterocharge Layer Morphology and Flow Distribution
Control”, Journal of Electrostatics, Vol. 111, pp. 1-13, May 2021
(https://doi.org/10.1016/j.elstat.2020.103548).
O’Connor, N. J., Castaneda, A. J., Christidis, P. N., Tobar, N. V., Talmor, M., and Yagoobi, J. S.,
“Experimental Study of Flexible Electrohydrodynamic Conduction Pumping for Electronics Cooling”,
ASME Journal of Electronic Packaging, Vol. 142, pp. 041105-1 to 6, December 2020.
Vázquez, P. A.., Talmor, M., Yagoobi, J.S., Traoré, P., and Yazdani, M., “In-depth Description of
Electrohydrodynamic Conduction Pumping of Dielectric Liquids: Physical Model and Regime analysis”,
Physics of Fluids, Vol. 31, 113601, 2019 (https://doi.org/10.1063/1.5121164 ).

Yang, L., Talmor, M., Shaw, B.C., Minchev, K.S., Jiang, C., and Seyed-Yagoobi, J., “Flow Distribution
Control in Meso-Scale via Electrohydrodynamic Conduction Pumping”, IEEE Transactions on Industry
Applications, Vol. 53, No. 2, pp. 1431 – 1438, March 2017.
Yang, L., Talmor, M., Shaw, B.C., Minchev, K.S., Jiang, C., and Seyed-Yagoobi, J., “Flow Distribution
Control in Meso-Scale via Electrohydrodynamic Conduction Pumping”, IEEE Transactions on Industry
Applications, Vol. 53, No. 2, pp. 1431 – 1438, March 2017.

Talmor, M., Yagoobi, J.S., Vazquez, P.A., and Traoré, P., “Influence of Temperature on the Performance of
Electrohydrodynamic Conduction Pumps”, Physics of Fluids (to be submitted).
Talmor, M, Louste, C., and Yagoobi, J., “PIV Flow Field Measurements of Electrohydrodynamic
Conduction Pumping”, Journal of Electrostatics (to be submitted).

O’Connor, N.J., Talmor, M., Asar, M.E., and Yagoobi, J.S., “Electrohydrodynamic Flow Distribution
Control for a Microchannel Evaporator”, International Journal of Heat Mass Transfer (to be submitted).

Book Chapter
Talmor, M. and Seyed-Yagoobi, J., “Electrohydrodynamically Augmented Internal Forced Convection”,
a Chapter in Handbook of Thermal Science and Engineering, Edited by Francis A. Kulacki, Springer
Publishing Co., New York, New York, ISBN 978-3-319-28573-3, 2018 (The e-book version of the
Handbook of Thermal Science and Engineering ).

Conference Proceedings
Tobar, N.V., Christidis, P.N., O’Connor, N.J., Talmor, M., and Seyed-Yagoobi, J., “Experimental Study
of Flexible Electrohydrodynamic Conduction Pumping for Electronic Cooling”, ASME International
Technical Conference and Exhibition on Packaging and Integration of Electronic and Photonics
Microsystems – InterPACK2018, E-Proceedings, San Francisco, CA, August 2018.

119
Talmor, M, Louste, C., and Seyed-Yagoobi, J., “PIV Flow Field Measurements of Electrohydrodynamic
Conduction Pumping”, Electrostatics Society of America, E-Proceedings, Boston, Massachusetts, June
2018.
Talmor, M. and Seyed-Yagoobi, J., “Performance Characterization of an Innovative Micro-Scale
Electrohydrodynamic Conduction Pumping Device”, ASME International Technical Conference and
Exhibition on Packaging and Integration of Electronic and Photonics Microsystems – InterPACK2017,
CD ROM, San Francisco, California, August 2017.
Talmor, M., Lei, Y., and Seyed-Yagoobi, J., “Numerical Study of Directional EHD Conduction Driven
Flow Distribution Control in Small Scales in the Presence of External Pressure Loads”, Electrostatics
Society of America, CD ROM, Ottawa, Canada, June 2017.
Yang, L., Talmor, M., and Seyed-Yagoobi, J., “Flow Distribution Control between Two Parallel Meso-
Scale Evaporators with Electrohydrodynamic Conduction Pumping”, ASME International Mechanical
Engineering Congress and Exposition, CD ROM, Phoenix, Arizona, November 2016.
Talmor, M., Yang, L., Larkin, T. R., Kamat, O. K., Dancy, T. J., and Seyed-Yagoobi, J., “Flow
Distribution Control in Micro-Scale via Electrohydrodynamic Conduction Pumping”, Electrostatic Joint
Conference, CD ROM, West Lafayette, Indiana, June 2016.
Yang, L., Talmor, M., Shaw, B.C., Minchev, K.S., Jiang, C., and Seyed-Yagoobi, J., “Flow Distribution
Control in Meso-Scale via Electrohydrodynamic Conduction Pumping”, IEEE-IAS Annual Meeting, CD
ROM, Dallas, Texas, October 2015.
Yang, L., Talmor, M., and Seyed-Yagoobi, J., “Effect of Temperature on Electrohydrodynamic
Conduction Pumping Performance”, International Conference on Dielectric Liquids, CD ROM, Bled,
Slovenia, July 2014.

120