Sei sulla pagina 1di 204

UNIVERSITY OF CALIFORNIA,

IRVINE
Numerical Simulation of Taylor Cone-Jet
DISSERTATION
submitted in partial satisfaction of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
in Mechanical and Aerospace Engineering
by
Ronne Toledo

Dissertation Committee:
Professor Roger H. Rangel, Chair
Professor Manuel Gamero-Castao
Professor Dimitri Papamoschou

2013

UMI Number: 3597852

All rights reserved


INFORMATION TO ALL USERS
The quality of this reproduction is dependent upon the quality of the copy submitted.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.

UMI 3597852
Published by ProQuest LLC (2013). Copyright in the Dissertation held by the Author.
Microform Edition ProQuest LLC.
All rights reserved. This work is protected against
unauthorized copying under Title 17, United States Code

ProQuest LLC.
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106 - 1346

2013Ronne Toledo

DEDICATION
To Emily and Eduardo, whose lives have inspired me to be a better person.

ii

Contents
Page
LIST OF FIGURES

vi

LIST OF TABLES

ix

ACKNOWLEDGMENTS

CURRICULUM VITAE

xi

ABSTRACT OF THE DISSERTATION

xii

1 Introduction
1.1 The Electrospray Experimental Observations . .
1.2 The Electrospray Applications . . . . . . . . . .
1.2.1 Aerosol . . . . . . . . . . . . . . . . . . .
1.2.2 Beads-Particle Fabrication . . . . . . . .
1.2.3 Coating and Painting . . . . . . . . . . .
1.2.4 Combustion . . . . . . . . . . . . . . . .
1.2.5 Drug Encapsulation . . . . . . . . . . . .
1.2.6 Biocomposite and Nanofiber Fabric . . .
1.2.7 Emulsification . . . . . . . . . . . . . . .
1.2.8 Field Injection Atomization (FIA) . . . .
1.2.9 Ionization . . . . . . . . . . . . . . . . .
1.2.10 Ion Propulsion . . . . . . . . . . . . . .
1.2.11 Electrospray Mass Spectrometry (ESMS)
1.2.12 MEMS Fabrication . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.

1
2
19
19
20
22
24
25
26
28
30
31
34
36
37

.
.
.
.
.
.
.
.

38
40
43
44
46
47
49
51
52

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

2 Mathematical Description of The Problem


2.1 Electric Field . . . . . . . . . . . . . . . . . . . . . . .
2.2 Far Field Electric Potential . . . . . . . . . . . . . . .
2.3 Charge Transport Equation . . . . . . . . . . . . . . .
2.4 Electric Forces in The Liquid . . . . . . . . . . . . . .
2.5 Mass and Momentum Balance for the Isothermal Flow
2.6 Liquid/Vacuum Interface Dynamics . . . . . . . . . . .
2.7 Mechanical Energy Equation . . . . . . . . . . . . . . .
2.8 Scaling Laws . . . . . . . . . . . . . . . . . . . . . . . .
iii

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

2.9

2.8.1 De la Moras Current Scaling Law . . . . . . . . . . . . . . . . . . . .


2.8.2 Gan-Calvos Current Scaling Law . . . . . . . . . . . . . . . . . . .
2.8.3 Higueras Current Scaling Law . . . . . . . . . . . . . . . . . . . . . .
2.8.4 Gamero-Castao Geometric Scaling Law . . . . . . . . . . . . . . . .
Dimensional Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.9.1 Higueras Dimensionless System . . . . . . . . . . . . . . . . . . . . .
2.9.2 Dimensionless System For The Invariant Transition Region . . . . . .
2.9.3 Alternative Dimensionless System For The Invariant Transition Region
2.9.4 Rescale of the Dimensionless Systems . . . . . . . . . . . . . . . . . .
2.9.5 General Considerations Regarding the Differences Among the Dimensionless Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3 The Boundary Element Method


3.0.6 The Greens Identities . . . . . . . . .
3.0.7 The Boundary Element Discretization
3.0.8 The Fundamental Solutions . . . . . .
3.0.9 The Composite Regions . . . . . . . .
3.1 The Boundary Element Mesh . . . . . . . . .

53
57
65
67
75
78
81
88
91
92

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

99
99
102
104
108
113

4 The Finite Volume Method


4.1 Mathematical Formulation . . . . . . . . . . . . . . . .
4.2 The Finite Volume Discretization . . . . . . . . . . . .
4.2.1 Interpolation Schemes . . . . . . . . . . . . . .
4.3 The Treatment of the Mesh Non-orthogonality . . . . .
4.3.1 The Domain Integration Method . . . . . . . .
4.4 The Pressure-Velocity Coupling . . . . . . . . . . . . .
4.4.1 The Semi Implicit Linked Equations (SIMPLE)
4.5 The Boundary Conditions . . . . . . . . . . . . . . . .
4.6 The Linear System Solver . . . . . . . . . . . . . . . .
4.6.1 The Strong Implicit Procedure (SIP) Solver . .
4.7 The Code Accuracy Check . . . . . . . . . . . . . . . .
4.8 The Taylor Cone Jet Profile . . . . . . . . . . . . . . .
4.9 The Updating Profile Technique . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

121
121
124
125
130
131
132
134
135
136
137
139
140
144

5 The
5.1
5.2
5.3

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

Results and Analysis


148
Numerical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
Interface Updating Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . 160
Taylor Cone-jet Emitted Current Scaling Law, Revisited . . . . . . . . . . . 160

6 Conclusions

166

Bibliography

170

iv

Appendix A Mathematical Expressions and Numerical Coefficients


A.1 Normal and Tangential Directions in the Curvilinear System of Coordinates .
A.2 Summary of The Governing Equations and Boundary Conditions . . . . . .
A.3 Collocated Mesh Finite Volume Coefficients . . . . . . . . . . . . . . . . . .
A.4 SIMPLE Method Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . .

187
187
188
189
189

List of Figures
Page
1.1

Reproduction of an electrospray photographed during Zeleny (1917) experiments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.1
2.2

Taylor Cone jet sketch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .


Electric current emitted by electrospray in Taylor cone-jet mode of highly
conducting liquids. Data originally presented in the work of De La Mora &
Loscertales (1994), figure 8. . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3 Dimensionless current functional dependence on dimensionless flow rate for
electrospray of different liquids. Data plot based on the measurements originally obtained by De La Mora & Loscertales (1994), figure 8. Figure 9 of
De La Mora & Loscertales (1994) paper shows a similar plot. . . . . . . . . .
2.4 Dimensionless current normalized by the slope d/d versus dimensionless
flow rate, . The slope of the curves were obtained from linear regression
for individual set of data measured for each liquid. Data plot based on the
measurements originally obtained by De La Mora & Loscertales (1994), figure
8. Figure 12 of De La Mora & Loscertales (1994) paper shows a similar graph.
2.5 Slope of the dimensionless current as function of dielectric constant. Data
based on the measurements originally obtained by De La Mora & Loscertales
(1994), figure 8. Figure 11 of De La Mora & Loscertales (1994) paper shows
a similar graph. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.6 The current as function of flow rate for the experimental data given by GanCalvo et al. (1997). Solid lines represent the I Q1/2 and I Q1/4 current
laws. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.7 Experimental data of Gan-Calvo et al. (1997) normalized by Q0 . . . . . .
2.8 Higueras numerical results of dimensionless flow rate versus dimensionless
current. The solid lines were scanned from figure 5 (b) of Higuera (2003).
The points on the graph were numerical data collected from within the text.
The I/Io = 2.6 (Q/Qo )1/2 is the Gan-Calvo (1999) experimental correlation,
and I Q1/4 is the scaling law introduced by Gan-Calvo et al. (1997). . .
2.9 Dimensionless current emitted by a Taylor cone as function of flow rate. The
data plotted are collected from Gamero-Castao (2010, pg 497, firuge 2b). .
2.10 Dimensionless voltage deficit as function of the inverse of Reynolds number.
The voltage deficit was normalized by 0 , as defined in eq. 2.80 and Reynolds
number defined in eq. 2.81. . . . . . . . . . . . . . . . . . . . . . . . . . . .
vi

3
39
57

58

59

60
64
65

68
69
72

2.11 Higueras interface profile renormalized by Gamero-Castao (2010) geometric


scale rt . The inset plot shows the original profiles of Higuera (2003, figure 2,
pg 313) obtained from numerical calculation for dielectric constant = 5 and
two different flow rates, QH = 0.27(I = 1.1) and QH = 4.8(I = 5.0). . . . . .
3.1
3.2
3.3
3.4
3.5
3.6
3.7
3.8
3.9
4.1
4.2
4.3
4.4
4.5
4.6
4.7
4.8

4.9

(a) Control Volume region delimited by the boundary interface . (b) BEM
discretization of the interface . . . . . . . . . . . . . . . . . . . . . . . . . .
Three dimensional Taylor Cone jet schetch. The ring source located at x0 is
integrated over direction to obtain the axisymmetric fundamental solution
(axisymmetric Greens function). . . . . . . . . . . . . . . . . . . . . . . . .
Schematic drawing of two subregions separated by a common interface. . . .
BEM mesh I. The arrows represent the surface normal direction. . . . . . . .
New mesh topology with boundary refinement. . . . . . . . . . . . . . . . . .
Inlet and Outlet zoom in for mesh topology I and II. . . . . . . . . . . . . .
Unstructured triangular mesh sample used to create the iso-potential plots. .
Fluid electric potential and its normal gradient along the interface boundary
for both mesh topologies. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Electric charge along the interface boundary for both mesh topologies. . . .
Body fitted non-orthogonal grid of CV. . . . . . . . . . . . . . . . . . . . .
Non-orthogonal mesh diagram. . . . . . . . . . . . . . . . . . . . . . . . . . .
Five-diagonal sparse matrix of coefficients for two dimensional and axisymmetric structured mesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Lid driven cavity flow problem sketch. . . . . . . . . . . . . . . . . . . . . .
u and v velocity profiles along the cavity center line in x and y directions,
respectively. Calculations were performed for Re = 100. Upwind, hybrid,
exponential and power law interpolation scheme were tested. . . . . . . . . .
u and v velocity profiles along the cavity center line in x and y directions,
respectively. Calculations were performed for Re = 400. Upwind, hybrid,
exponential and power law interpolation scheme were tested. . . . . . . . . .
u and v velocity profiles along the cavity center line in x and y directions,
respectively. Calculations were performed for Re = 1000. Upwind, hybrid,
exponential and power law interpolation scheme were tested. . . . . . . . . .
Taylor interface profile scanned data from Higuera (2003, pg 313, figure 2a)
paper. This interface profile is the Higueras numerical solution for flow rate
QH = 0.27. The inset plot is the spatial derivative of the profile superimposed
with the fitted arctan and the noise reduced curves. . . . . . . . . . . . . . .
Taylor interface profile scanned data from Higuera (2003, pg 313, figure 2b)
paper. This interface profile is the Higueras numerical solution for flow rate
QH = 4.8. The inset plot is the spatial derivative of the profile superimposed
with the fitted arctan and the noise reduced curves. . . . . . . . . . . . . . .

vii

75
101
106
108
114
116
116
118
119
120
125
131
138
140
141
142
143

145

146

5.1

The counter plot of the electric potential superimposed with the fluid stream
lines. The numerical results for this plot were obtained with QH = 0.27 and
= 5.0. The isopotential lines colored with black are the result of the present
calculations with Higuera (2003) isopotential lines superimposed with symbols
for comparison purposes. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 The counter plot of the electric potential superimposed with the fluid stream
lines. The numerical results for this plot were obtained with QH = 4.8 and
= 5.0. The isopotential lines colored with black are result of the present
calculations with Higuera (2003) isopotential lines superimposed with symbols
for comparison purposes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3 The counter plot of the electric potential for QH = 0.27 superimposed with
results obtained for QH = 4.8. In the figure, length and electric potential
data from figures 5.1 and 5.2 were rescaled using the invariant dimensionless
system where characteristic parameters are presented in table 2.1. . . . . . .
5.4 The electrostatic properties of the interface along the zI axis ordinate for
two different flow rates. The black lines, marked with symbols, represent
the solution for flow rate QH = 0.27 and the red lines, without symbols, for
QH = 4.8. Chart (a) is the electric potential (left abscissa) and surface charge
(right abscissa). Chart (b) is the tangential and normal electric field at the
surface for the liquid (superscript (1) ) and vacuum (superscript (2) ). Chart (c)
is the difference between liquid and vacuum normal and tangential Maxwell
stress tensor. Chart (d) is the bulk and surface current. . . . . . . . . . . . .
5.5 The velocity profile for two different flow rate conditions along the rz cross
section planes located at five different axial positions. Chart (a) represents
the velocity in the axial direction for QH = 0.27 located at zI axial position
indicated in the legend. Chart (b) is the velocity in the axial direction for
QH = 4.8. Chart (c) is for velocity in the radial direction for QH = 0.27.
Chart (c) is for velocity on radial direction for QH = 4.8. . . . . . . . . . . .
5.6 The viscous normal and tangential forces together with the Maxwell tangential
force at the interface. The inset plot is the difference normalized by capillary
force. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.7 The normal viscous and Maxwell stress forces at the interface. Pressure and
capillary is also shown in the plot. Inset graph shows the balance error normalized by capillary forces. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.8 The normal stress balance error profile and average error for two iterative
procedures. Graph (a) and (c) are for = 1e 3 and (b) and (d) are for
= 1e 6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.9 The dimensionless current versus Capillary number for a number of experimental data collected from literature. . . . . . . . . . . . . . . . . . . . . . .
5.10 The y intercept (y0 ) of the curves shown on figure 5.9 as a function of Higueras
definition of Reynolds number. . . . . . . . . . . . . . . . . . . . . . . . . . .
5.11 The general dimensionless current dependence on the Capillary number. . . .

viii

149

150

153

154

156
157
159
161
163
164
165

List of Tables
Page
2.1
2.2
4.1

Dimensionless system rescaling parameter between Higeras and the Invariant


system. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Dimensionless system rescaling parameter between Higeras and Alternative
Invariant system. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

93
94

Finite Volume Discretization Coefficient . . . . . . . . . . . . . . . . . . . . 128

A.1 Governing Equations Summary . . . . . . . . . . . . . . . . . . . . . . . . . 188


A.2 Boundary Conditions Summary . . . . . . . . . . . . . . . . . . . . . . . . . 188

ix

ACKNOWLEDGMENTS
As a man of faith, I am profoundly grateful to the Living God for the gift of science and faith
allowing those who accept it to appreciate His love and care for His creatures. For His laws
of nature, that in their original state, has provided a perfect environment for humankind
to live and prosper. For the death of His son, Jesus Christ, allowing our redemption and
restoration of nature laws.
Thank you to Prof. Rangel for giving me the opportunity to come to UCI and pursuit my
Ph.D., all the advice offered to me during the program, and the financial support obtained
from the TA and GSR opportunities offered to me. Thank you to Prof. Gamero-Castao for
introducing me to the exciting problem of the Taylor Cone Jet and for the financial support
throughout the projects that I have been involved in, and the discussions that allowed me
to evolve and mature in academic terms.
Thank you, my parents, Alaor Jose Toledo and Maria de Lourdes Toledo for their love, care,
emotional and financial support. For their teachings that have inspired me to keep me going
until the end, and to never give up. Also thank you to my beautifully wife, Eliana Gisele
Rodrigues Toledo, whose tireless effort in caring for our children has allowed me to progress
in the in development of this work.
Thank you to my friend Riju Lavanya whose friendship has offered me precious emotional
support.
Thank you to the Brazilian citizens and its government, whose taxes has financially supported
me through CAPES scholarship. Thank you to Miguel Velez scholarship for the financial
support. And last but not least, to UCI MAE department, its faculty and staff members for
hosting me, and offering me a wonderful physical and academic environment through classes,
seminars, TA and Lecturer opportunities.

CURRICULUM VITAE
Ronne Toledo
1994-1999

B.A. in Mechanical Engineering, University of Brasilia,


Brazil

2000-2003

M.A. in Mechanical Engineering, Technical University


of Lisbon, Lisbon, Portugal

2003-2004

Research Assistant, University of Brasilia, Brazil

2006-2008

Teaching Assistant, Mechanical and Aerospace Engineering, University of California, Irvine

2009-2010

Lecturer, Mechanical and Aerospace Engineering, University of California, Irvine

2004-2013

Ph.D. in Mechanical and Aerospace Engineering, University of California, Irvine

xi

ABSTRACT OF THE DISSERTATION


Numerical Simulation of Taylor Cone-Jet
By
Ronne Toledo
Doctor of Philosophy in Mechanical and Aerospace Engineering
University of California, Irvine, 2013
Professor Roger H. Rangel, Chair

The Taylor cone-jet is a particular type of electrohydrodynamic phenomenon where electrostatic stresses and surface tension effects shape the interface of the jet in a peculiar conical
shape. A thin jet is issued from the cone apex that further breaks up into a fine aerosol.
Due to its monodispersive properties, this fine aerosol has found a number of applications,
ranging from mass spectrometry, colloidal space propulsion, combustion, nano-fabrication,
coating/painting, and many others. In this study, a general non-dimensional analysis is
performed to derive the governing equations and boundary conditions. In accordance with
the observations of Gamero-Castao (2010), noting that droplet electric potential is insensitive to the flow rate conditions, a particular set of characteristic parameters is proposed,
based on the terminal jet diameter. In order to solve the non-dimensional set of governing
equations and boundary conditions, a numerical method combining the Boundary Element
Method and the Finite Volume Method is developed. Results of electric current have shown
good agreement with numerical and experimental data available in the literature. The main
feature of the algorithm developed is related to the decoupling of the electrostatic from the
hydrodynamic problem, allowing us to accurately prescribe the far field electric potential
boundary conditions away from the hydrodynamic computational domain used to solve the
hydrodynamics of the transition region near the cone apex.

xii

Chapter 1
Introduction
This dissertation investigates the problem of electrospray in the cone jet regime. The phenomenon of electrospray has been widely researched in the past few decades, and general
interest in this area has grown especially with the advent of fast computers that can provide
support and validation for various numerical and mathematical techniques.
The electrospray phenomenon can be classified as a multi-physics, multi-scale problem. It
is considered a multi-physics problem because of its electrostatic and fluid dynamic considerations. In particular, the electrospray problem is also a multi-scale problem because the
liquid jet issued at the apex of the menisci is much smaller than the size of the capillary that
feeds the liquid into the electrostatic field.
This phenomenon occurs in the meniscus of a electric conducting liquid subject to an electric
field. This electric field induces charge on the surface of the liquid, altering the surface stress
balance at the interface and resulting in a deformed meniscus. With an appropriate electric
field, the meniscus deforms to a cone like shape. Different operational modes are possible as
well (Cloupeau & Prunet-Foch, 1989). In particular, the operational mode where a thin jet
is issued by the cone apex is the main object of investigation of this dissertation.
1

This thesis proposes a technique to solve the Taylor cone jet governing equations and boundary conditions using a combination of the Boundary Element Method and the Finite Volume
Method. This novel technique provide the means to accurately place the far field electric
potential boundary condition without incurring the tremendous increase in computational
costs, in stark opposition to traditional techniques.
The following sections of this chapter present a comprehensive bibliography of this field and
its applications. This review describes research performed in the areas of the experimental
observations and the diverse range of applications of this phenomenon. Chapter two reviews
in detail the mathematical modeling used to describe the problem and its scaling laws resultant for dimensional analysis and experimental data found in the literature. We then propose
a novel generalized non-dimensionalization procedure for this problem to obtain the set of
governing equations and boundary conditions with the use of the general set of dimensionless
parameters. Chapter three is dedicated to detailing the numerical method used to solve the
electrostatic equations, the Boundary Element Method (BEM). Chapter four contains the
descriptions of the different numerical techniques used to solve the dynamics of the flow, and
the associated Computational Fluid Dynamics (CFD) techniques. Chapter five presents the
results of the simulations performed with the different tools developed. Chapter six presents
conclusions drawn from our study.

1.1

The Electrospray Experimental Observations

In 1600, William Gilbert, Queen Elizabeths personal physician, observed that electric fields
deform fluids interfaces (Gilbert, 1991). Subsequently, Zeleny (1914) extended his previous
work on electric discharge from pointed conductors to liquid points (Zeleny, 1907). Zeleny
(1917) experimentally investigated the disintegration of drops under strong electric field in
an effort to study the formation of thunderstorms. He was among the first scientists to
2

Figure 1.1: Reproduction of an electrospray photographed during Zeleny (1917) experiments.


observe and photograph the narrow jet formation from a liquid meniscus formed in the tip
of a capillary tube.
Zeleny (1917) showed that instability is the main cause of the meniscus disintegration with
a instability criteria of
2
=C
ro

(1.1)

Where , ro , are the potential, drop radii, surface tension respectively and C an arbitrary
constant. Figure 1.1 reproduce a earlier image experimentally acquired by Zeleny (1917).
Zeleny. (1920) summarizes his observations in the investigation of this branch of the EHD
problems.
Macky (1928) was interested in the field of electrostatic, termed by him as frictional electricity. Macky (1931) extended his previous work to investigate the behavior of charged falling
drops in the presence of a strong electric field, where he observed the formation of a cone
like shape at the rear tip of the drop and a thin long jet issued by the apex of the cone.
Wilson & Taylor (1925) conducted experimental work to investigate the behavior of soapbubbles in the presence of a electric field. Vonnegut & Neubauer (1952) introduced the term
3

electrical atomization in his experimental work and qualitatively observed the production
of monodisperse aerosols having particle radius of a micron or less, using low electrical
conductivity liquids.
Taylor (1964) developed a theoretical framework and established the first simple model for
the problem. In his theory and based on the observations, he assumed that the shape of
the deformed drop under the electric field could be approximated as a cone like shape. This
phenomenon is also known as the Taylor Cone-Jet after Taylors analysis.

Since the curvature of a conical surface is inversely proportional to the distance from the vertex the stress normal to it which can balance the surface
tension must also be inversely proportional to distance from the vertex. Since
the fluid will be assumed to be conducting the conical surface must be an equipotential, and to balance the surface tension the potential gradient there must be
1

proportional to R 2 where R is the radial co-ordinate.

Under this assumption, the electric field that satisfies the stress conditions can be expressed
as
1

T = 0 + AR 2 P 1 (cos ())
2

(1.2)

where the cone axis is given by = 0 and = , P1/2 is the Legendre function of order 1/2
(Abramowitz & Stegun, 1965), R the distance from the cone apex and A a given constant.
For = o , the conical equipotential surface is given by T = 0 , where 0 is a arbitrary
constant, and in general chosen to be equal to zero. Therefore, 0 can be calculated from

P 1 (cos (0 )) = 0
2

(1.3)

Since the function P1/2 has only one zero for 0 < < , this results in 0 = 130.7099o . From
this result, the Taylor-Cone jet semi-vertical angle is defined as T = 0 = 49.29o . The
curvature of such a cone can be expressed as 1/tan(T )R, therefore, from the stress balance,

1
= 0 En2
tan () R
2

(1.4)

where E = , En = E n and is defined from equation 1.2 with

A=

2
02
0 P1/2 (cos (T )) sin2 (T ) tan (T )

!1/2
(1.5)

and 0 = 0.
Drozin (1955) conducted experimental investigation with a glass capillary and different liquids in the presence of a strong electric potential. He experimented with polar and nonpolar
organic liquids and described his observations regarding the dispersibility of the aerosol formation. He also calculated the electrostatic pressure in the meniscus of the liquid using a
theoretical analysis of Onsagers equation (Onsager, 1936) and experimental correlation of
Henriquez equation (Henriquez, 1935). Taylor & McEwan (1965) further extended the work
of Taylor (1964) to investigate the instability of a conduction liquid under the influence of
an electrostatic field force.
Kim & Turnbull (1976) succeeded in electrospraying insulating liquids by placing a sharp
electrode (needle) inside the liquid and raising it to a high potential. This needle injects
charges into the insulating liquid, producing large enough surface charges to overcome the
inadequate low electrical conductivity. Bailey & Balachandran (1981) measured the droplet
dispersion of electrospray hydrocarbon using a particle size analyzer to quantify the light
energy distribution from a laser source. He concluded that since surface charge density is
mainly controlled by liquid resistivity, directly affecting the liquid electrical relaxation time,
and consequently the electric field, liquid resistivity is one of the most important parameters
5

controling the liquid breakup. Viscosity has a minor role in further controlling the size of
the droplet, since viscous effects increase wavelength resulting in larger drops (Bellman &
Pennington, 1954).
Masayuki (1984) experimented with a.c. potential in order to increase the monodispersity
of the aerosol produced from electrospray of water. Smith (1986) reported his experimental
observations in a parametric study of electrospray of different liquids to pin point the effect
of conductivity, surface tension, viscosity in the onset voltage and current operation in the
Taylor cone jet mode. In his observations, he concluded that: a) liquids with high conductivity would produce smaller droplets; b) Increasing viscosity would produce long wavelength
for the varicose instability, generating large droplets; c) The onset voltage has a quadratic
variation with increasing surface tension; and d) Liquid hydrostatic pressure has low effect
on the onset voltage in the Taylor cone jet mode. Hayati et al. (1986, 1987b,a) in a series of
papers, produced streak photography with licopodium particles to investigate the circulation
pattern inside the cone. He investigated the electric field surrounding the Taylor cone by
employing finite element calculations and observed the relationship between voltage, current
and flow rate in the stability of the Taylor cone mode. His work provides strong evidence
that the surface of the Taylor cone jet is not equipotential, therefore, the tangential electric
forces are not negligible, especially near the cone apex.
Cloupeau & Prunet-Foch (1989) reported experimental observation with the goal of producing droplets with uniform diameter smaller than 20m. He introduced the term cone-jet
to identify the specific jet mode of operation where the Taylor cone and the thin jet from
the apex is present in a stable/steady condition. Experimenting with liquids in the conductivity range of 101 109 S/m, and flow rate ranging from 103 10mm3 /s, he was able
to establish a lower bound limit of steady cone-jet operation, where the upper bound was
not observed. Further observations, Cloupeau & Prunet-Foch (1990) propose a classification
of different modes of electrospray operations, such as: dripping, microdripping, cone-jet,
6

spindle, simple jet and ramified jet. This classification was further discussed in Cloupeau
(1994) and Cloupeau & Prunet-Foch (1994).
Meesters et al. (1990, 1991) describe the development of aerosol generator design, the Delft
Aerosol Generator (DAG). On this design, a shielding electrode between the needle and
the ground electrode. The shielding electrode voltage was select to produce corona discharge
between the droplets in order to neutralize its electric charge. Meesters et al. (1992) reported
operation conditions in the presence of a corona discharge, that further he defined as gas
discharge mode (Meesters et al., 1993). Due to the corona discharge characteristics, he could
not find the current/flow rate relation observed by others researches. Instead, he reported
similarities in the electric behavior between the Taylor cone jet and metallic needle corona
discharge. He was further criticized by Fernndez de la Mora and Gomez (Meesters et al.,
1993) and clarified some of the previously observed and reported results.
De la Mora et al. (1990) successfully produced a highly monodisperse aerosol of droplets in
the size range of submicron. He experimented with glycerol seeded with N aI and collected
droplet dimension and charge data by using a variable-speed impactor and differential mobility analyzer (DMA). From his observations and previous data available in the literature,
he proposed the first scaling law for the jet diameter, written as,
a I 2 / 2 2

 31

where a is the jet diameter, liquid density, surface tension and I electric current.
De La Mora (1992) investigated the effect of charge emission on the Taylor cone jet. Base
on observations, he assumed that the charged droplets distribution generated from the cone
apex would modify the electric field produced by the liquid menisci initially proposed by
Taylor (1964). Rosell-Llompart & Fernndez de la Mora (1994) reported a parametric study
of highly conducting liquids electrosprayed into droplets ranging from 0.3m < d < 4m.

The goal of this study was to characterize the droplet size spectra of the produced aerosol
using a aerodynamic size spectrometer. The results show that at the minimum flow rate,
only primary and satellite droplets were formed. At higher than the minimum flow rate,
the number of distributions increased. The diameter of the largest drops in the distribution
1

scales with r = (Q ) 3 , where is the electric relaxation time. The others diameters in the
distribution were independent of the flow rate, with the minimum droplet size scaling with
1

dmin = ( 2 /) 3 . They also found that the droplet diameter for all distributions was depen1
( 2 ) 3
dent upon the viscous parameter =
. Further, De La Mora & Loscertales (1994)

conducted experimental work of electrospray in cone jet regime where the diameter of the jet
was much smaller than the diameter of the capillary. He used five different types of solvent:
ethylene glycol; triethylene glycol, formamide, water, 1-octanol and bensyl alcohol where
he seeded all solvents with different concentrations of LiCL, except the 1-octanol, where
H2 SO4 was used. Different concentration of the solute produced solutions with different
electric conductivity ranging from 5.4e 5 to 0.27 siemens per meter (Sm1 ). De La Mora
& Loscertales (1994) observed that in the cone jet regime the structure of the jet became
independent of the electrostatic parameters, voltage and electrode configuration and that
the measured current was mostly function of the liquid properties, and is given as
1

I = f () (QK/) 2

(1.6)

This scaling law was proposed considering that a sink flow convect the surface charges towards the cone apex disturbing the equilibrium charge distribution given by Taylor solution.
This disturbance becomes relevant in a distance r from the apex, large compared to a certain
charge relaxation length, . This distance is in the order of r = (Q0 /K)1/2 with the corresponding surface current convected towards the apex scales as Is = (QL/)1/2 . This sink
flow simplified analysis shows good agreement for sufficiently small jet Reynolds numbers.
From photographs, De La Mora & Loscertales (1994) also finds that the jet diameter scales

with dj 0.4r showing that the jet diameter forms soon the charge relaxation effects is
over. In the limit of large dielectric constant (  1), an upper bound is found for the
convected surface current which accounts for only one-quarter of the total measured spray
current, indicating that the renaming current is transported by conduction throughout the
bulk. Aguirre-de Carcer & de la Mora (1995) further investigate the effect of the surrounding
gases in the current-flow rate characteristics of the electrospray. Using air, CO2 and SF6
with pressure varying from 130 1500T orr, they conclude that although the mobility of the
particle and the width of the spray highly depends on the background gas and pressure, the
current-flow rate characteristics and the minimum flow rate was unaffected, leading to the
conclusion that space charges have a negligible effect in the electrospray phenomenon.
Rulison & Flagan (1994) also reported electrospray experimental data of electrolytic solution
with conductivity varying from 0.00684 0.356S/m. The only aspect that differentiates this
study from previous work was the use of a aluminum extension gadget coupled to the capillary
needle that somehow modified the electric field, thus modifying the results.
Gomez & Tang (1994) report electrospray of heptane in a charge-to-volume ratio investigation. They have observed that large the droplets have its ratio charge-to-volume closer to
the Rayleigh limit, given by
q 2 = 8 2 o d3
and that droplets about 7080% of the limit end up going to a rupture process, termed by the
authors as Coulomb Fission. They used phase Doppler anemometry and flash shadowgraph
to investigate the rupture phenomenon and further satellite formation. One interesting
observations during the droplet rupture process is that part of the droplet surface distorted
to a conical like structure with a thin jet been ejected from its apex. This small Taylor
cone-jet like structure is responsible for producing equally sized smaller offspring droplets
satellites with large charge-to-mass ratio. This result might implicate the biochemical mass
spectrometry, since these satellites droplets are the primary source of ions via ion desorption.
9

Further extension of his previews work, Tang & Gomez (1995, 1996) reported electrospray
results of water and liquids with low conductivity. Chen et al. (1995) reported experimental
results with conducting liquids to produce droplets in the range of 4nm to 1.8m.
Gan-Calvo et al. (1993) undertook a analytical study using small perturbation (Dyke
(1975); Kevorkian & Cole (1996)) for the small parameter

o
1
Q

(1.7)

This small parameter was obtained from the ratio of hydrodynamic residence time, th = Q/
with charge relaxation time, tr = i / assuming that = i /o  1, where is the liquid
electric conductivity. The reported results of the analysis were in good agreement with
experimental data. The scaling law for the relation between current and flow rate was given
as,
(1.8)

p
I = 2.47 Q

In order to identify the minimum flow rate, current and jet diameter under which polar
liquids can be electrosprayed, Gan-Calvo (1994) developed a theory based on the ratio of
charge and hydrodynamic relaxation time. By assuming that the minimum cone-jet mode is
attained when these two time scales are of same order, the minimum flow rate, current and
jet diameter can be respectively given by
1

Qmin
Imin
dmin

i 2

k

1
o 2 2

 2  13
1
o
3
k 2

where the droplet charge can be obtained by q =

10

d3 I
,
6Q

scaling to the Rayleighs limit.

(1.9)
(1.10)
(1.11)

Gan-Calvo et al. (1997) observation also supports the I Q 2 law for the dependence
of the current with flow rate, initially proposed by De La Mora & Loscertales (1994) and
1

expressed by equation 1.6. He also discussed the existence of a I Q 4 law for liquids with
low viscosity and conductivity. Further extending his analysis, Gan-Calvo (1997a) used
small perturbation techniques to predict the behavior of the electrostatic spray in the cone jet
regime for liquid in the quasielectrostatic/perfectly conducting (QEPC) limit. The QEPC
limit is suitable for perfect conducting liquids, or leaky dielectric (not perfectly conducting)
liquids operating in a low speed compared with the electric relaxation time (te = i /K).
This theory departs from Taylor solution for the electric potential and perturb it along the
liquid interface ( = T + (R)),
= T + G

(1.12)

where the perturbation function is written as a series of the form

G (R, ) =

D (Q () + C P ()) R

(1.13)

where Q and P are the Legendre functions of order . To determine the coefficients of
the above series, Gan-Calvo (1997a) linearized the perturbation correction (eq. 1.13)
assuming  T and balancing out the liquid momentum with surface shear stress along
the jet issued by the cone apex, written as,


1 Q2
2s
d
P+ 2 4 =
dR
2

where = ( ) R is the jet local radius, assuming  R. Gan-Calvo (1997a) also neglected the influence of pressure in the jet, assuming that p / is much smaller than kinetic
energy, Q2 / (2 2 4 ), in the limit of very thin jets. From the parameters {, K, , 0 }, liquid
density, conductivity, surface tension and vacuum permittivity, respectively, Gan-Calvo
(1997a) defined several characteristic quantities: flow rate, Q0 = K 1 1
0 , electric current

11

1/2
1 2
Io = 1
, length d0 = 2 2
0
0 K

1/3

1
and electric field E0 = 21
0 d0

1/2

Gan-Calvo (1997a) indicated that these characteristic values for the flow rate, electric
current and length, are the minimum values that can be obtained by electrospraying (Chen
et al., 1995; De La Mora & Loscertales, 1994; Gan-Calvo et al., 1997). Based on the non
dimensional flow rate, Q = Q/Q0 , and the number, L0 = ln Q1/2 , Gan-Calvo (1997a)
defines typical values for jet Radius, R0 = d0 Q1/2 , axial distance, Lo = d0 QL0 , normal and
1/2

axial electric field, En0 = E0 L0

and Es0 = E0 (QL0 )1/2 respectively, and potential varia-

tion along the jet, 0 = E0 L0 . It is noteworthy that for the limit of Q  1, the characteristic
axial length of the jet, L0 , is very large compared with the jet diameter, d0 , and the number
L0 is also very large. Using these typical values, the nondimensional universal governing
equation can be written as

es = es0 a

(1.14)

f = 2af 3 es

(1.15)

es f 2 a
+ 2 =I
2
f

(1.16)

where the nondimensional quantities are given as, z = R/L0 , f = /R0 , a = A/ (En0 R0 ) =

A/ Es0 L0 L1
, eS = ES /Es0 and I = I/ (8I02 Q/L0 ). Asymptotic solution of the universal
0
governing equation for the electric spray were obtained for the limiting case of z .
Solution of the region around the Taylor cone apex, z O(1), is consistently found in terms
of power series of the form
k = k 1 +

!
n z 3n/4

n=1

where k is a symbolic label for f , a or es .


Chen & Pui (1997) compared the two competing scaling laws of De La Mora & Loscertales
(1994) and Gan-Calvo (1994) by experimental observations of electrospray of dielectric
liquids with a dielectric constant ranging from 12.5 to 182. He measured the electric current,
12

the flow and the droplet size of the monodisperse aerosol by measuring the size of the particle
produced after the liquid of the droplet been evaporated. Measurements of the particle size
spectrum were taken using a TSI scanning mobility particle sizer (SMPS). Results show
that for solvents with high dielectric constants the droplet size are in agreement with scaling
law proposed by Gan-Calvo (1994). For this scale law, the low dielectric constant liquids
do not show good agreement, in part due to the long jet characteristic length assumed by
Gan-Calvo (1994) and not experimentally observed in Chen & Pui (1997). For most of
the observed emitted currents, Chen & Pui (1997) results deviate from the scaling law of
those proposed and observed by De La Mora & Loscertales (1994), except for the solution of
benzyl alcohol. This could be explained by the fact that the solute used to vary the electric
conductivity in the work of De La Mora & Loscertales (1994) was LiCL, different from the
HN O3 used by Chen & Pui (1997). The authors justify the differences by criticizing the
assumption of negligibility of the ions mobility proposed by De La Mora & Loscertales (1994)
in the electrospraying process. Based on the results obtained, Chen & Pui (1997) proposed
modification in the minimum liquid feed rate, previously proposed by the authors and an
expression to determine the maximum feed rate.
Saville (1997) presented a seminal review work in the leaky dielectric liquid model initially
developed by Taylor (1964) and extensively used by Melcher (Melcher & Taylor, 1969). The
remarkable aspect of this work is the derivation and simplification of the balance equations,
including charge for the bulk and interface.
In order to unify the two distinct branches of research investigating the Taylor cone-jet phenomenon, the electrospray of liquid metal (LMIS) in vacuum and the electrospray of electrolytes in a pressurized environment, Gamero-Castao et al. (1998) conducted experiments
where electrolytes were electrosprayed in a environment with controllable pressure, varying
from 0.02 771 T orr. One important conclusion from this study is that the surrounding
pressure has negligible effect on the current emitted from the Taylor cone, its shape, charge
13

distribution, or its droplet diameter. Another conclusion is the fact that space charge might
have a influence on the mechanism in the current transmitted by the gas, but has negligible
effect on the current emitted by the cone-jet itself.
Barrero et al. (1999) investigated the effect of electrical conductivity and viscosity on the
motion inside Taylor cone-jet. From experimental observations using optical devices, they
concluded that the flow pattern from liquids of low viscosity and conductivity have different
characteristics than those with high range of viscosity and conductivity. In a attempt to
model the flow field inside the cone jet, they superimposed a sink flow with a self-similar
circulation conical solution, balancing out the tangential shear stress at the interface with the
tangential electric stress. With mathematical modeling valid for the low Reynolds number
spectrum, they show good qualitative agreement with experimental observations.
Hartman et al. (1999) proposed a numerical model to investigate the cone-jet phenomenon.
Their model is based upon the balance of the viscous and electrical stresses at the interface.
To calculate the velocity gradient at the interface, they assume a flat velocity profile. By
a means of a iterative process, they correct the interface profile by reducing the error in
the stresses balance at the interface. They also compared his results with data obtained
experimentally and available from the literature. Although some agreement was obtained,
they indicate the need to improve the velocity profile calculation procedure in the proposed
model. From their analysis, they proposed a modified scaling law that could be thought as
a intermediate scaling law between the one proposed by De La Mora & Loscertales (1994)
and Gan-Calvo et al. (1994).
Loscertales et al. (2001) proposed a novel technique of electrospray in cone-jet mode of
two co-flowing immiscible liquids. This technique generates a outer jet in Taylor cone-jet
mode, while the inner liquid is injected. This mode of operation could be used to create tiny
capsules filled with the inner liquids. Lpez-Herrera et al. (2003) extended the previous work
of Loscertales et al. (2001), characterizing the different modes of operation and deriving the
14

scaling law for coaxial jets in Taylor cone-jet mode.


Higuera (2003) employed a numerical algorithm to accurately calculate the velocity profile of
the transition region of the Taylor cone-jet. This algorithm uses finite difference techniques
to solve the stream-function and vorticity equation with the proper boundary condition at
the interface. The computational domain of Higuera (2003) solution was confined to the
transition region from the cone to jet. Charge balance equation was also solved at the interface together with Laplaces equation for the electric potential inside and outside the cone
region. The boundary condition for the electric potential faraway was the Taylor potential.
From numerical solutions, he concluded that the current carried by the jet increases monotonically with flow rate, as expected. The asymptotic structure of the solution for large
non-dimensional flow rate show two distinct regions, the cone region, where conduction is
the main mechanism of charge transport, converging to a slender region where the current
is dominated by surface convection. In the slender region, the electric shear stress acting
on the charged surface accelerates the whole cross-section of the fluid by viscous diffusion
mechanism. The current of this regime is proportional to Q1/2 as experimentally observed
by others authors.
Gan-Calvo (2004) suggested the existence of five asymptotic regions with different scaling law for each region. Based on the regions dominant phenomenon he classified regions as following: inertia and electrostatic; inertia and polarization; viscous and electrostatic; viscous and polarization; and surface tension. Comparison with experimental
data partially supports his suggestion, but since most of the data available are in the inertia/electrostatic/polarization region, he was not able to fully support his proposed scaling
laws for the others regions. Barrero et al. (2004) present experimental data of a electrospray
of a conducting liquid in a insulating bath of a immiscible liquid. Current measurements
show the same I Q1/2 trend as electrospray in air or vacuum. He also found that if the
insulating bath have negligible viscosity compare with the liquid been electrosprayed, the
15

droplet size have the same scaling law as one observed for electrospray in air/vacuum. One
difference is that the cone in the insulating bath have semiangle smaller than the one obtained in air/vacuum. One possible explanation may lie in the fact that the drag present
in the insulated bath does change the spatial charge distribution associated with a narrow
cloud of droplets, consequently changing the electric potential around the cone.
Marginean et al. (2004) studied the relationship between the current oscillation and the
pulsating Taylor cone-jet mode. He investigated the formation process of the Taylor cone-jet
by measuring the apex velocity in pulsating mode and correlated this phenomenon with the
current measurements. Good quality time-lapse images of the Taylor cone and jet formation
were seen for the first time in scientific literature. He concluded that measurements of electric
current emitted by the jet could be used as a feedback parameter to stabilize the jet.
Higuera & Barrero (2005) investigated the existence of a linear scaling law for very polar
liquids electrosprayed in Taylor cone-jet mode at low flow rate regime. Gan-Calvo (2004)
was the first to proposed the existence of a linear scaling law, different from the I Q1/2
scaling law well known for other regime. Different from Gan-Calvo (2004), Higuera &
Barrero (2005) used a order-of-magnitude analysis to derive such a law, but emphasized the
need of experimental data to support this proposed linear scaling law.
Smith et al. (2006b,a) revisited a much controversial subject in the electrospray field, which
is the influence of the applied voltage, fluid property and capillary geometry in the flow rate
of electrospray on liquid in Taylor cone-jet mode. A majority of the scientific community
agrees with observations provided by De La Mora & Loscertales (1994) and Gan-Calvo
et al. (1997) that voltage has negligible effect on the current and/or flow rate emitted by
the cone-jet, assuming that the jet is in the range of Taylor cone-jet operation. However,
earlier observations from Yahiku et al. (1970) has exhibited a linear dependency between
mass flow rate and voltage (Smith et al., 2006b). One of the arguments that motivates
Smith et al. (2006b) to collect a new set of experimental electrospray data was associated
16

with the imprecision of the flow rate measurements in the available systems reported in the
literature, specially those conducted in vacuum. In his novel system of measurements, flow
rate was accurate measured by a on-line automated system. He concluded that based on
the "experimental data collected, the volumetric flow rate is indeed sensitive to the applied
voltage". In fact, not only the data but also images collected from the Taylor cone show
sensitive changes to the voltage applied.
Paine et al. (2007) reported experimental investigation of unforced nanoelectrospray in pulsation regime. From the experiment, they observed that increasing the conductivity of the
three liquids tested, ethylene glycol, triethylene glycol and water, the frequency of the pulse
increases, reaching a maximum frequency of 635kHz for water. They also observed that the
dimension of the needle affects the frequency of the pulse; smaller the diameter of the needle,
higher is the frequency. From energy analysis, this authors shown that bulk conduction has
a role in the charge transport process and that the pulsation energy is dependent on fluid
conductivity and viscosity.
Collins et al. (2007) reported a numerical investigation of the breakup mechanism of a Newtonian perfectly conducting liquid jet in the presence of a electric field. They compare two
type of algorithms, the axisymmetric two dimensional algorithm and the one-dimensional
slender jet algorithm. Collins et al. (2008) has investigated the formation of the Taylor cone
and the thin jet issued by its apex using extension of his previously developed numerical
algorithm. They also conducted experimental observations to collect images to qualitatively
compare their numerical results. From the calculations, Collins et al. (2008) found that the
q
1/3
o is the Ohnesorge number,
drop size has a scaling law of rd kOh , where Oh = / L
k = /o is the permittivity ratio and = te /tc , where te = /, is the charge relaxation
q
3o /, is the capillary timescale. Collins et al. (2008)

time, is the conductivity and tc = L


 1/3
tc
compared this result with the scaling of De La Mora & Loscertales (1994), rD Q
,

1/2 tc / 1/6 , and show that when the viscous effect is
and Gan-Calvo (1997a), rG Q
17

= tc (/), both scale rD rG 2/3 , differ from the rdrop 1/3


neglected, defining Q
= l3 /t , where l is the vissuggested. However, when viscous effect is included (defining Q
cous length and t is the viscous time), the drop scaling law becomes, rd Oh1/3 , agreeing
with the scale reported. Further extension of his work, Collins et al. (2013) investigates the
breakup of charged liquid drop in the presence of the electric field. In order to achieved that,
they numerically simulated the transient evolution of a spherical droplet to the ellipsoidal
and further formation of the Taylor cone-jet streaming and breakup. From the numerical
results, they were able to report a universal scaling law for the droplet charge of q = 0.44
qR ,
where q is the amount of charge that the droplet carries, and qR is the Rayleigh limit of
stability.
Krpoun & Shea (2008) used the finite element method to calculate the onset voltage of a
single and an array of a micro-fabricated needles. He used a predefined cone like shape
with a 49.3 Taylors half-angle where the apex of the cone was described with a smooth
Bernstein-Bzier curve. From the finite element calculations the electric field was calculated
for several conic shapes with different apex radii. From the known critical electric field,
which is the field to yield ionic emission, the shape and the onset voltage was obtained.
Ryan et al. (2009, 2012) investigated the influence of the emitter geometry and the flow rate
in the onset voltage of the electrospray in cone-jet mode of operation.
Gamero-Castao (2010) reported experimental data of time-of-flight and retarding potential
for a wide range of liquids properties electrosprayed in a low pressure chamber filled with
hydrogen. He showed that approximately 17% of the electric energy used to generate the
aerosol is in deficit. From mechanical energy balance equation, he further demonstrated that
this energy is dissipated by ohmic heating (Joule effect) and viscous dissipation.
Bhattacharjee et al. (2010) reported experimental data of homopolymers solutions in electrospining regime. The scaling reported shows good agreement with the I Q1/2 law, where
the electric potential E and conductivity K were used in the constant to group the results
18

in the same line, I/Io = E/Eo (K/Ko )0.4 (Q/Qo )1/2 where the characteristic values follow
Gan-Calvo (1998) suggestions. Gundabala et al. (2010) reported a voltage dependence in
the current of an electrospray process in the presence of a coflowing liquid in a microfluidic
device, contrary to the observed in the classical electrospray experiments.

1.2
1.2.1

The Electrospray Applications


Aerosol

Aerosol has a large variety of applications in calibration instruments, testing procedures, applicators and manufacturing processes. Many researchers have successfully used electrospray
to produce aerosol of fine particles.
Meesters et al. (1990, 1991, 1992), in a series of papers describe the configuration designed
and operation of the Delft Aerosol Generator (DAG). They successfully show that the electrospray in the Taylor Cone jet regime can produce aerosol in the frequency ranging from
108 1010 droplets per second with droplets sizes in the order of 1m for liquid with modest
conductivity. They also show that the operation of volatile liquid with non volatile solute
(DOP in ethylene glycol), droplets of smaller sizes ( 0.08m) can be further achieved with
solvent evaporation during the droplet flight. They conclude that an shielding ring between
the needle and ground is necessary to stabilize the cone jet regime and consequently produce
a monodisperse aerosols with standard deviation in the order of 0.2m (for droplets in the
1m range). They also identify the influence of the corona discharge in the total electric
current for different values of voltage. He conclude that the corona discharge is a necessary
mechanism to produce droplets with neutral electric charge.
More recently, Ude & de la Mora (2003) has shown the use of hypersonic impactors and

19

differential mobility analyzer (DMA) having the capability to measure the mass of particles
smaller than 2nm. In order to calibrate these measurements, the monodisperse of aerosol
plume of particles in the range of 2nm has to be generated. Electrospraying of commercial
Polyethylene glycol (PEG) with molecules ranging from 2.1 9.8nm has been successfully
tested. To further lower the size of particles that can have their mass measured, Ude &
de la Mora (2005) has used the mass spectrometry combined with DMA. To obtain aerosol
standards for these techniques, Ude et al. (2006) electrospray polystyrene (pSty) dissolved
in 1-methyl-2-pyrrolidone (NMP) seeded with dimethyl ammonium formate. By controlling
the electric conductivity of the solution, Ude et al. (2006) was able to obtain monodisperse
aerosol ranging from 3 11nm. These techniques and solutions, allowed them to produces
4 mass standards ranging from 9200 g/mol to 96,000g/mol.

1.2.2

Beads-Particle Fabrication

One standard application in which electrospray has been a valuable asset is in the beadsparticle fabrication. The food processing industry and drug fabrication make use of beadsparticles or droplets to add chemical compounds to their products in different stages of the
fabrication. Masayuki (1984) has shown that the use of alternating current (AC) can produce
monodisperse droplets of distiled water in the range of 2.4mm and 28m. Each droplet was
formed synchronously with each cycle of the applied AC and its size controlled by the liquid
flow rate and AC frequency. Martn-Banderas et al. (2005) has demonstrated the use of
flow focusing (FF) technique to produce monodisperse particle with diameter lower than
25m. The FF technique has shown compliance with the requirements of uniform shape and
surface and homogeneous size distribution particles necessary to the field of analyte research
and diagnosis for parallel multiplex analysis (Martn-Banderas et al., 2006). It is evident
that the production of monodisperse particle in the size range required by this application
can be substituted by the use of an electrospray technique. Indeed, Gan-Calvo (2007)
20

has shown that the use of FF and electrospray (ES) techniques can enhance the parametric
research of microjetting beyond the range obtained by the two techniques separately. This
combine technique was termed by Gan-Calvo (2007) as Electro-flow focusing. GanCalvo & Montanero (2009) provides an in depth analysis of both techniques, ES and FF,
and shows that if electrospray data are expressed in terms of an effective pressure drop, both
phenomena satisfy the same scaling law for the droplet size, exhibiting similarities in the
particle size distribution. The effective pressure drop is a function of the liquid parameters
only, not been dependent upon operational parameters as voltage applied and flow rate.
Fundamental coincidence in the stability limit regimes in the two phenomena was obtained
by stability analysis, resulting in useful general description and predictive scaling laws for
micro to nanospraying from the cone jet.
Kim et al. (2006) was able to produced microsized poly methylmetacrylate (PMMA) polymer
droplets. He characterized the droplet size distribution by spraying the droplets in a PET film
and further characterizing in a optical microscope. The size of the droplets was controlled
by a electrospray design using different auxiliary electrodes and an AC-controllable target
electrode. Xie et al. (2006) has successfully shown that the electrohydrodynamic atomization
can be used in the production of biodegradable polymeric micro- and nanoparticles. In
his study, poly-lactic-co-glycolic acid solution was atomized and particles in the size range
of 250nm were obtained. Organic salt was used to enhance conductivity of the spraying
solution. He demonstrated the ability to produce solid and porous particles by controlling
several parameters of operation and liquid properties. A seminal overview of the FF and
ES techniques can be found in Barrero & Loscertales (2007). There, the authors indicate
the widely accepted scaling laws of both phenomena and a brief explanation of the physics
underlying the two phenomena. He also detailed some of the applications of these techniques.
Larriba-Andaluz and Fernndez de la Mora (2010; 2011) show that monodisperse droplet
can be also produced from insulating liquids by injection of charged nanodrop of ionic liquids
21

in the inner core of the insulating liquids. These charged nanodrops drifts to the surface
of the insulating liquids, deforming in a cone like shape that further atomizes, producing
monodisperse micron size droplets of the insulating liquids. This work extends and confirms
the previous results obtained by Kim & Turnbull (1976). Due to its features this technique
also can be employed, as example, in the drug encapsulation processing, where a given
chemical compound has to be encapsulated by a different insulating compound, as will be
detailed later.
Sun et al. (2011) has successfully reported the use of electrospray in the preparation of quantum dots encoded microspheres to be used in the detection of biomolecules. He reported a
novel method by electrospraying polymer solution and quantum dots solution to obtain a
series of composite microspheres with functional nanoparticle. He further tested the performance of this microspheres in the immunofluorescence analysis of antigens immobilized in
the particles.

1.2.3

Coating and Painting

Perhaps one of the oldest applications for the electrospray is in the industrial painting and
coating process, specially in the automotive industries. US Patents using electrospraying
for painting and coating are found to be as old as 1934 (Darrah, 1934) and many more
have been filled after that, Ransburg (1959); Probst & Probst (1973); Heyningen (1983);
Kocot et al. (1983); Seaver & Eckhardt (1988); Roberts et al. (1992); Kim & Ryu (1994);
Leonard (1996) just to cite a few. Hines (1966), from the Ford Motor Company, located in
Michigan, has estimated that 30% - 50% of the paint does not land on the intended surface
when compressed air sprayers are used. He conducted a parametric analysis of a single jet,
where he identified the effect of the corona discharge as a limiting effect in the atomization
rate. He also reported an apparatus developed to produce a multiplex of cone jet, the knife

22

edge atomizer. This apparatus was able to produce the large quantity of jets along the
edge, where paint was fed by centrifugal force. This technique provided the atomization
rate necessary to industrial scale painting using electrospray. Siefert (1984) enhanced the
deposition efficiency in the coating process called spray pyrolysis by means of electrostatic
field with corona discharge. This technique has improved control in the time of flight of the
charged droplets resulting in a higher quality of the coating process.
Uematsu et al. (2004) uses electrospray to produce protein thin films from aqueous lactalbumin. The surface morphology was further characterized by use of scanning electron
microscopy and atomic force microscopy. Further reaction to ions of calcium has demonstrated that the film produced has its biological activity preserved. This result shows the
potential use of electrospray coating process in the production of protein-based biomaterials. Matsumoto et al. (2005) has further demonstrated that organic and inorganic hybrid
nano-microstructured coating can be electrospray deposited on a insulated plate as well.
Rietveld et al. (2006) has characterized the coating process of poly(vinylidene fluoride) in
the production of thin films by electrospraying coating. His observations are in agreement
with previous work showing that solute concentration (controls the solution conductivity)
and flow rate are the best parameters to control the film morphology, since these parameters
control the droplet sizes and dispersion. Rietveld et al. (2009) further extends his parametric analysis of thin film production of poly(vinylidene fluoride) using electrospray coating
technique. He was able to obtain films in the range of 60 120nm within 25 45s. The
parameters controlling the dimension and quality of the film was identified to be the droplet
size, polymer concentration, shear rate at impact and volume of flux. Kim et al. (2011) has
also characterized microstructures of silica layers produced with electrospray. In his parametric research, he observed that by controlling the viscosity of the precursor solution he
was able to change the microstructure of the silica layer and consequently, the wettability of
the coated surface. This has direct impact on the production of surface-related applications
such as self-cleaning, anti-fingerprint, anti-contamination and anti-fogging.
23

1.2.4

Combustion

One of the earliest studies in the field of electrostatic dispersion of liquid fuels was given
by Weinberg (1968). He discussed the application of electrospray in dispersion of the liquid
fuel and the morphological changes of the flame under different electric fields. Lawton &
Weinberg (1969) further extended this discussion on the application of such technology in the
combustion field (Jones & Thong, 1971). The work of Jones & Thong (1971) characterized
the electrical dispersion of jet kerosene. He observed that within a finite small range of
applied voltages, droplet dimensions is independent of voltage applied, and strong dependent
upon flow rate. He also developed a theory where the controlling parameter of the current
is given by the electric field strength at the surface of the jet.
Thong & Weinberg (1971) systematically studied the electrostatics dispersion of liquids (electrospray) and solids, its trajectories path and resulting charged particles to manipulate their
burning process. He employed experimental and theoretical tools to investigate powdered
solids in distiled water and paraffin with antistatic additive as fuel. To measure and quantify
his observations, he used optics and MgO-coated microscope slides, from where he acquired
information regard droplet size and charge. From his findings, he designed burners operated
entirely by electric field which draw in and accelerate air from the surroundings using corona
discharges in multi-stage ion pumps.
One of the great challenges to efficiently burn liquid fuels is in how fast can liquid fuel be
converted to the gas phase. The underlying phenomenon associated with this process is
liquid evaporation. To enhance the evaporation rate of the liquid, two parameters should
be accounted for, the surrounding temperature and the surface area of the liquid. There is
no doubt that smaller the size of the fuel droplets, faster will be the evaporation rate. With
this goal, electrospray techniques have the features required for the task. The technical issue
still remaining in how to multiplex the atomization process to satisfy the flow rate required

24

by the combustion process.


Chen & Gomez (1992) identified the structure of counterflow diffusion flames of heptane.
Measurements of droplet size distribution, velocity and gas-phase temperature was characterized for two types of flames. The first flame behaved like a gaseous diffusion flame, since
all the liquid has been converted to gas before flame front. For the second flame experiment, some droplets cross over the flame front and burn in an oxygen rich environment.
Consequently, soot formation changes the characteristics of the flame, like color and peak
temperature.
Kyritsis et al. (2002) has successfully reported the development of a mesoscale catalytic
combustor operating in liquid JP8 jet fuel. The fuel was atomized by means of electrospray
technique. From gas chromatographic analysis, the authors reported a estimated combustion efficiency of 97%. They also reported that no fouling, nor soot, nor N OX was detected
from the exhaust gases. This remarkable result shows great potential in the technique of
electrospray hydrocarbon liquid fuel to produce efficient and low emission combustor. The
challenge still exists in how to multiplex the electrospray to obtain the necessary flow rates.
This issue was partially addressed in the work of Deng et al. (2007) where microfabrication
techniques were used to develop a microcombustor with a multiplexed electrospray sources.
This device had volumetric heat release reported to be as large as 270M W/m3 . The combustion efficiency was accessed by comparison of CO/CO2 ratio, found to be under 1% and
after 10h of operation, the inner walls of the combustor did not reveal any traces of deposits.

1.2.5

Drug Encapsulation

The use of microsphere as potential drug delivery system has been discussed by Mathiowitz
et al. (1997). Although the authors of this paper employed phase inversion nanoencapsulation method to produce the microspheres, electrospray could be a reasonable substitute. In
25

fact, Loscertales et al. (2002) successfully demonstrate applicability of electro-hydrodynamic


(EHD) technique in the production of a monodisperse aerosol where the diameter of the
particle varying from 10 0.15m. The droplets of these aerosols have an outer liquid
encapsulating a inner one. The authors used electrified coaxial liquid jets to produce the
droplets. Xie & Wang (2007) report to successful encapsulate living cells using electrospray in
dripping mode. Xie et al. (2008) extended his previous work by developing a coaxial electrospray system to encapsulate protein-based drugs in biodegradable polymeric microparticles.
To test the design, the bovine serum albumin and lysozyme were used as the protein based
drugs. He characterized the morphology of the particles produced by means of Scanning
electron microscopy. After biologically testing the particles, they concluded that co-axial
electrospray could represent a revolutionary approach to encapsulate biomacromolecules.
Almera et al. (2010) reported a extensive morphologic characterization of poly(lactide-coglycolide) aliphatic polyester justified by the fact that morphology of the particle is a key
aspect of the drug delivery profile. Almera et al. (2011) go one step further and demonstrate
that a multiplex system could present a serious advantages compared to traditional methods
of drug encapsulation. In his work he uses poly(DL-lactic-co-glycolic acid) to encapsulate
amphiphilic agents such as duxorebucin and rhodamine. One of the major advantages reported by using this method of encapsulation is in the simplicity of the method. Only one
single step (electrospray) is necessary to produce the final drug delivery system, compared
with the multi-steps necessary to the traditional method (emulsification).

1.2.6

Biocomposite and Nanofiber Fabric

Electrospinning is a electrospray operational mode in which the atomization does not take
place in the apex of the cone jet. Instead, a fine and long thread is issued from the apex.
This thread, or solid fiber, is produced when the polymeric fluid is used as the working

26

solution in a electrospray process. Hohman et al. (2001a) develops a instability analysis of


the electrospray process, finding asymptotic approximation for the equations. He reported
the existence of three different instabilities, the Rayleigh, electric field induced axisymmetric
and whipping instabilities. The dominant instability was further identified based on the
surface charge density and radius of the jet. Hohman et al. (2001b) and Shin et al. (2001)
further identify the operational parameters when electrospinning will occur and identify the
changes in the electrospinning process according with fluid conductivity and viscosity. A
good review of the electrospinning technique is presented by Yu & Rutledge (2002). More
recently, Bhattacharjee et al. (2010) has conducted electrospinning experiments to measure
the current generated by the process.
Apart from its academic interest that had led electrospinning studies of unorthodox substances like chocolate (Luo et al., 2011), the most valuable use of this technique are in the
field of biocomposite and nanofiber fabric fabrication. Doshi & Reneker (1995) report a
parametric and morphological study in the production of fibers from polymer solution by
electrospinning process. He lists some of the commercial application for these types of fibers:
reinforcing fibers of composite materials, non-wetting surface layer on textiles, support for
thin polymeric membranes, insecticide application on plants, production of non-woven fabrics and wound dressing material. Reneker et al. (2000) studied the bending instability of
the nanofibers produced by electrospinning of polimeric solution. A theory was developed
to describe the bending mechanism using the viscoelastic dumbbell model and comparison
was made with experimental data. Li & Xia (2004) review the process of production of
nanofibers from electrospinning by focusing on the mechanism and the theoretical models
developed to describe the process. Few applications are also mentioned where the nanofiber
was used in the process to fabricated reinforced composites, membranes and smart cloths,
biomedical material, supports for enzymes and catalysis, sensors, electrodes, electronic and
optical devices, sacrificial templates, etc.

27

A similar technique to electrospinning, electrospray deposition has been proven to be effective


in the preparation of ion-exchange of fiber fabric, as reported by Matsumoto et al. (2006).
Wakamatsu et al. (2006) further tested the performance of ion-exchange nanofiber fabrics in
water splitting in bipolar membrane, and Matsumoto et al. (2007) test the electrokinetic and
absorption behavior of nanofiber fabric produced from chitosan/poly(ethylene oxide) blend
solution.
Reneker et al. (2007) discuss the interaction of surface tension and electrical forces on the
change of the shape of viscoelastic fluids used in the production of nanofibers. Some of
key aspects of his work is to identify the parameters that modified the morphological characteristics of the fibers. Greiner & Wendorff (2007) review the subject of electrospinning
for preparation of ultrathin fibers. In application section of the paper, he mentioned few
samples that already uses this technique. Among the mentioned applications are: filter
and textile fabrication, catalysis, tissue engineering, wound healing, transport and release of
drugs, tumor therapy, inhalation therapy, etc.

1.2.7

Emulsification

The production of emulsion has various uses in the pharmaceutical and food industries.
Characterization and control of the emulsion behavior have always attracted attention from
the research community. The traditional techniques of emulsification always rely upon the
stirring mechanism of two immiscible liquids, where the monodisperse characteristics is not
always achieved. Due to its features of producing monodisperse droplets, some authors have
been suggested electrospray as a viable alternative to the emulsification process.
Sato et al. (1993) report successfully electrostatic emulsification of viscous liquids in an insulating bath where traditional methods have failed to produce emulsion. He also characterizes
the size of the droplets in the emulsion as function of the voltage applied to the electrodes.
28

Ramos & Castellanos (1994) has extended Taylor (1964) analysis to liquid-liquid electrospray
interfaces in the limit of perfectly insulated liquids. Characterization of the cone angle has
been presented as a function of the permittivity ratio of the two liquids. He concluded that
the classical Taylor cone solution is the limiting case for infinite conductivity of inner liquids,
but solutions with different angles can also be achieved.
Tsouris et al. (1997) compared the emulsions formed by a more traditional method of emulsification, the stirred tank, and electrostatic spraying emulsification. From energy calculations,
he showed that electrostatic emulsification was able to produce emulsification more efficiently
than by mechanical agitation. Further, Tsouris et al. (1998) experimented with electrostatic
spraying of air and organic solvents in relatively conductive fluids, as water. He measured
and related the pressure inside the capillary with the voltage applied to the electrodes. He
observed that electric field can cause simultaneous pumping, spraying and mixing of fluids.
Barrero et al. (2004) experimented with electrospray of liquid in insulator liquid baths,
and highlighted the applicability of this technique to produce emulsions of uniformly sized
droplets in the nanometric range. From current measurements, he also found that the scale
law of electrospray in air is preserved and applicable also to liquid-liquid electrospray. Mejia
et al. (2009) reported the production of uniform discotic wax particle using electrospray
emulsification in the process. Gundabala et al. (2010) found that from electrospray of liquidliquid in a coflowing microfluidic device, current show strong voltage dependence and weak
flow rate dependence. This result is in apparent contradiction with experimental results
from electrospray of liquid in gas/vacuum. Two factors can account for the unexpected
behavior, a) The force balance at the interface is fundamentally altered by the dynamics of
the coflowing design; and b) The scale of the experiment might affect the observation, when
compared with more traditional electrospray device dimensions.

29

1.2.8

Field Injection Atomization (FIA)

Field injection is a technique first reported by Henson (1964) where ions were injected into
liquid argon and liquid nitrogen through an immersed tungsten point electrode. This phenomenon was further characterized by Halpern & Gomer (1969a) and extended to liquid
H2 , D2 , He, Ar, N2 , O2 , and C6 H6 . Halpern & Gomer (1969b) extended his previous work
to Field ionization of liquid H2 , Ar, N2 , and O2 . Field ionization is a similar technique to
field injection, but due to sufficient ion acceleration in the high-field region near the emitter,
charges are multiplied.
Kim & Turnbull (1976) used field injection technique to introduce charges into a insulating
liquid to atomize it in a monodisperse spray using electrospray method. In his study, he uses
liquid Freon 113, but he discussed the possibility of using the same technique to others cryogenic liquids as liquid nitrogen and liquid hydrogen. This method has important application
in the production of cryogenic droplets for refueling fusion reactor and laser fusion targets.
Field Injection Atomization provides a simple technique to electrospray insulated liquid
without the need to be seeded with salt, resulting in liquid droplets of pure solvent, without
any "contaminant", except for the ion injected. Larriba-Andaluz & Fernandez de la Mora
(2010) discussed the charge mechanism in the FIA technique. The basic underlying physics
is the same as in the electrospray of leaky dielectric or conduction liquids: the charge drifts
to the surface of the insulator liquid and destabilizing it, leading to the formation of the jet
and subsequent atomization. The novel design, proposed by Larriba-Andaluz & Fernandez
de la Mora (2010), is to use a liquid charge-injector that continuously renews and sharpens
itself, instead of a solid tungsten sharp tip, previously reported in Henson (1964). This
design provides a long-term atomization operation. Larriba-Andaluz & Fernandez de la
Mora (2010) reported a monodisperse spray ranging from 4 20m of liquid heptane as
insulator liquid and 1 ethyl 3 methulimidazolium BF4 as the ionic liquid. He also

30

observed that the scaling law of Kim & Turnbull (1976) is confirmed, implying that drops
are on the average charged to 50%-60% of the Rayleigh limit.

1.2.9

Ionization

Ions (atom or molecule in which the total number of electrons is not equal to the total
number of protons, Wikipedia 2013) can be produced from chemical and physical processes.
Ions sources have application in different scientific areas. Field emissions guns have increased
by order of four the performance of electron microscopes(Denizart et al., 1981a,b; Pinna
et al., 1983). Point ion sources consist of fieldionizing a material near the apex of a metallic
tip. Mller & Tsong (1969) ionize gas in his field ion microscope apparatus. Others authors
have employed electrohydrodynamic (EHD) techniques to produce ions from the Taylor cone
apex (Hendricks & Swatik, 1968). This techniques consist of either wetting the metalic tip
(Prewett & Mair, 1991) or feeding the liquid-metal through tiny capillary. For a critical
voltage range, the cone like shape is formed at the tip of the metallic/capillary tip and
intense and stable ion emission appears. This liquid-metal ion sources (LMIS) have found
many interesting applications in different fields, as microlithography (Seliger et al., 1979;
Komuro et al., 1983; Marche et al., 1983), local implantation, micromachining, scanning ion
microscopy, high resolution secondary ion mass spectrometry, etc.
Mahoney et al. (1969b) reported atomic ions, molecular ions and charge droplets from liquid
cesium, N aK/cesium alloy and sodium. Clampitt et al. (1975) reported generation of
intense beams of metal ions of Li, Cs, Sn, Ga and Hg by field-ion emission of molten
metal films and solid wire needles. Krohn & Ringo (1975) reported the production of ions
using liquid gallium, cesium and mercury, where gallium has produced higher brightness
than cesium and mercury. Swanson & Schwind (1978) reported electron emission from
liquid Ga/12% In. He observed that the liquid metal apex exhibits an incandescent glow

31

corresponding to a temperature rise of 1000 C. Swanson et al. (1980) has conducted


measurements of energy distribution of gallium liquid metal ion source, and concluded that
the deficit in the energy supports a field evaporation mechanism of ion formation. This
mechanism was further investigated by Bell et al. (1982) who study the composition and
energy distribution of a beam of ions obtained from liquid aluminum.
Due to its limitation, the classical field ionization theory did not provide sufficient explanation for the phenomenon.Gomer (1979) extend Taylor (1964) and similarly assume a rounded
cone like shape for the apex of the jet. This theory did not provided satisfactory agreement
with experimental data, and it was argued by Ward & Seliger (1981) that the presence of
space charges is not compatible with the sphere-on-cone shape (Benassayag et al., 1985, pg
2).
Benassayag et al. (1985) addressed this problem by investigating the electrospray of gallium
in the Taylor cone regime performed in the scope of a High Voltage Electron Microscope
(HVEM) using similar principles of a Transmission Electron Microscope (TEM). Images
from the apex of the cone were collected, thereby enhancing knowledge of its shape. Gallium
was selected due to its low temperature melting point, low vapor pressure and low electrohydrodynamic emission onset voltage, reducing field distortion. From his observations,
Benassayag et al. (1985) concluded that precise examination of the apex of the cone, reveals
a cusp like shape and the existence of a jet-like protrusion of short distance, different from
theoretical predictions available at the time the paper was published.
Although electrohydrodynamic ionization was primarily made with liquid-metals, Fenn et al.
(1989) has started discussing the impact of EHD ionization principles and techniques in the
production of intact ions for mass spectrometry analysis.
Most of the liquid metal ion source (LMIS) experiments are preferably done in a vacuum
chamber, with few exceptions. Gamero-Castao et al. (1998) experimented with electrolytes
32

electrosprayed in a chamber where the surrounding pressure was varied from 0.02 Torr to 737
Torr. This series of experiments had the goal of identifying the difference and resemblance of
the scaling law and its pertinent physics observed in LMIS and electrolytes in gas electrospray.
They concluded that in both cases, the liquid properties and flow rate controls the current
passing through the meniscus. The charge effects controls the current transmitted through
the gas/vacuum. Gamero-Castao & Fernndez de la Mora (2000) further investigated the
ion evaporation kinetics from liquid surface of electrolytes. The goal was to estimate the
electric field E present in the liquid surface. To achieve this objective, they proposed a
novel technique of combining stopping potential measurements techniques with preferential
deceleration of the ions by a background gas. Gamero-Castao (2002) also was able to proof
the ion field evaporation from dielectric liquid by combining measurements techniques of
time-of-flight and retarding potential.
Not only liquid-metal and electrolytes but also ionic liquid (molten salt) can be used as
ion source. One of the problems associated with ionic-liquid ion source (ILIS) is the electrochemical decomposition of the ions, when single polarity ions is generated. Lozano &
Martnez-Snchez (2004) proposed a novel technique by alternating the voltage of the electrode in a specific frequency associated with the electrochemical window to minimize any
electronic reaction with the metalic emitter. In his observations liquid EM I BF4 (1ethul-3-methyl imidazolium tetrafuoroborate) was electrosprayed in vacuum using similar
technique of the LMIS, where the ionic liquid externally wetted the sharp tungsten emitter
(Lozano & Martnez-Snchez, 2005). Garoz et al. (2007) experimented with vacuum capillary electrospray of different ionic liquids to characterize the purely ionic regime at room
temperature.

33

1.2.10

Ion Propulsion

Ion propulsion is a branch of propulsion systems founded on the physics of charged particle acceleration techniques. Also known as Electric Propulsion, this type of propulsion
has the characteristics of low acceleration and high specific impulse. These characteristics
have promising applicability in satellite thrusters designed for space missions of long term
or continuous orbital correction. Different designs of thrusters have been developed, i.e.
hall thrusters, ion engines, pulsed inductive thrusters, contact ionization thrusters, colloidal
thrusters, field-effect electrostatic propulsion, etc (Khayms, 2000).
Electric spray technique is the core mechanism for charged particle generation for systems
like the colloidal thrusters and field-effect electrostatic propulsion. In colloidal thrusters, a
plume of charged droplets is generated from electrospraying the liquid propellants. These
droplets are electrostatically accelerated to high speeds and are neutralized by an external
cathode, producing thrust by reaction. Field-Effect electrostatic propulsion works in similar
principles as the colloidal thrusters, except that the propellants are liquid-metal, and the
electrospray operates in the pure ions generation regime, therefore no droplets are formed,
highly increasing the specific impulse characteristics.
The possibility of using charged particle to generate thrust is a fundamental idea proposed
since the beginning of the 1900 century (Choueiri, 2004). The use of electrospray to generate the charged particle in the Electrostatic propulsion has began almost simultaneously
with the development of colloidal thrusters (Gignoux et al., 1964). Work of Hendricks Jr.
(1962); Carson et al. (1964b,a) had the specific goal of understanding and characterizing the
electrospray phenomenon to be applied to the development of colloidal thruster technology.
While Hunter (1966); Mahoney et al. (1967, 1969a); Mahoney & Perel (1969, 1976) research
had the objective of developing the colloidal thruster technology itself, applying the recently
developed electrospraying of the propellant technology. Jahn & Mickelsen (1969) provide a

34

comprehensive state of development of electric propulsion by the end of the 1960 decade.
Bartoli et al. (1984) addressed some of the problems associated with single injection point of
ion source for space propulsion by developing a novel array of injection from a micrometersized slit injector. Khayms (2000) has discussed the use of electric thrusters for small scale
satellites. He also developed a small scale hall thruster and investigated the electrohydrodynamic of Taylor cone jet using numerical techniques. Romero-Sanz et al. (2003) characterized
electrospray from Taylor cone of ionic liquid under different surrounding temperature. In
his experiment, he was able to reach a pure ionic regime of EMI-Beti electrospray above a
temperature threshold of 199 C. At that condition, he obtained a thrust of 1N at specific impulse of 2000s. This reported result was higher than previously obtained from single
Taylor cone of organic electrolytes. Carretero et al. (2004); Carretero (2005) addressed
the emission mechanics from colloid thrusters using a numerical modeling. He employed a
quasi-one-dimensional numerical technique to describe the cone jet profile and a 3D particlein-cell methodology to model the plume generated from the jet. The mixed ion-droplet
regime is modeled with additional ion evaporation model incorporated into the simulations.
Martinez-Sanchez & Akinwande (2005) summarized the colloidal thrusters technology development realized by his research group, including the previously mentioned numerical work
of Carretero (2005). Blandino et al. (2007) reported a novel particle counting algorithm to
measure the droplet energy distribution without the need to to collect retarding potential
curves for the entire beam or use dedicated charged particle counter. Quang Tran Si et al.
(2007) investigated the influence of emitter spacing in an array of emitters of a microthurster.
He had developed a theory and experimentally tested the array of emitters.

35

1.2.11

Electrospray Mass Spectrometry (ESMS)

Mass spectrometry is the technique used to identify the chemical composition of a sample
material by measuring the masses of single chemical elements in that material. The masses
are obtained by measuring the mass-to-charge ratio of a charged particle (ion) by injecting its
trajectory in a electrostatic or electromagnetic field. Electrospray techniques play important
role in the ion generation of liquid electrolytes.
Although interest in ion generation for propulsion purposes is attributed to Krohn (1962), the
first study to report detection of ions from electrospray was (Cook, 1986). Soon Hendricks
& Swatik (1968) realized that by generating ions from electrospray and measuring the timeof-flight of the charged particles, the mass to charge ratio spectra could be obtained. Dole
et al. (1968) also reported detection of macroions, that was was further interpreted in the
light of Iribarne & Thomson (1976) model by Yamashita & Fenn (1984) that soon realized
that this technique could revolutionize the field of mass spectrometry of large molecules,
frequently present in biological materials. Fenn et al. (1989) describe the full potential of
such technique in the field of biology, and Fenn et al. (1990) provide a historic review and
specific details of the mass spectrometry techniques with electrospray. Smith et al. (1990)
also reviews the subject and describe the collision-induced dissociation methods to investigate
the structure of large molecules. Winger et al. (1993) report resolution accuracy increase
for mass measurements of biomolecules using a novel electrospray ionization technique, the
ion cyclotron resonance mass spectrometer. Wilm & Mann (1994, 1996) report a analytical
description of the electrospray phenomenon to predict the size of the electrospray jet. He
further designed a micro electrospray capable of producing droplets in the 200nm range or
below. Due to the low level of flow rate, this new design is capable of mass spectrometry
of sample in the range of 0.5l. Gaskell (1997) review the underlying principles of the
electrospray process. Loo (1997); Chang et al. (2002) reported successfully use of electrospray
mass spectrometry to study complex biomolecules and proteins.
36

1.2.12

MEMS Fabrication

One of the most recent applications beginning to benefit from electrospray techniques is
the micromachining of microelectromechanical systems (MEMS). As demonstrated by Suh
et al. (2005), the nanoxerography can benefit from electrospray techniques due to its quality
of producing charged monodisperse aerosol in the nanoscale range. Gamero-Castao &
Mahadevan (2009) successfully produced sputtering yields in a Si, SiC and B4 C substrate
by electrospraying silicon, silicon carbide and boron carbide. This technique can be used in
surface treatment in some fabrication process of micromachining MEMS devices.
Deng et al. (2010) discuss the benefit of using electrospray in a controlled deposition maskless bottom-up fabrication technique instead of the traditional inkjet printing technique. In
his proof-of-concept design he had a system consisting of 61-nozle array microfabricated in
silicon with the extractor made of flexible polyimide substrate. Each nozzle can be controlled independently with a microfabrication acurracy of 675m in a frequency of 100Hz,
representing a considerable resolution improvement compared to available systems on the
market. Some of the benefits reported with the use of electrospray is the possibility of using
high viscous liquids, avoidance of high shear rates and low risk of clogging.
Kim et al. (2013) report the use of electrospray techniques in the preparation of superhydrophobic SiO2 layers. The hierarchical surface structure was produce from Au nanoparticles
depositions from the electrospray and use of ultraviolet-enhanced chemical reduction. The
resulted surface showed a water-repellency of 170 .

37

Chapter 2
Mathematical Description of The
Problem
The electrohydrodynamic problem could be classified as a multi-physics, multi-domain problem. The dynamic and electric properties of each domain is intrinsically related to the interface motion and location. In the liquid domain, the dynamics is governed by momentum
balance equation (Navier-Stokes). Electric potential in the region is prescribed by Laplaces
equation. In the outer gas/vacuum side the dynamics is negligible and the electric potential
is also prescribed with Laplaces equation.
To correctly account for the motion of the interface between the two domains, forces should
be balanced out among the two regions. Figure 2.1 presents a sketch of the problem. To
organize the description of the problem, 1 is labeled after the volume occupied by the liquid,
and 2 by the gas/vacuum. In theory, the vacuum1 domain can be extended to infinity. In
practice, and from a computational perspective, the domain is limited to a finite dimension.
1

For simplicity, from here on, the words gas/vacuum will be substituted by vacuum only. But the reader
should be aware that in certain conditions gas might be present instead of vacuum only. Nevertheless, the
presence of gas does not significantly change the dynamics of the problem described.

38

To avoid any spurious results in the computational calculations, the dimension of the vacuum
region is defined to produce negligible influence in the liquid/vacuum interface.

Taylor Cone Jet


te

rfa

ce

ow inlet

in

outlet

Figure 2.1: Taylor Cone jet sketch


To facilitate the mathematical description of this problem it is convenient to define a cylindrical system of coordinate. Additionally, a local curvilinear system is often used to describe
the interface. In the cylindrical system of coordinates, position is defined as x = {r, , z} and
velocity as u = {ur , u , uz }, where r, and z are the radial, azimuthal and axial directions
respectively (see figure 2.1 for more details). From the assumed axisymmetric properties
of the problem2 , the azimuthal gradients, / = 0, can be neglected. To simplify the
description of this problem it will be assumed that the swirl velocity3 is also negligible.
2
3

Under certain unstable conditions, the problem is not axisymmetric


A swirl velocity has been experimentally observed by Barrero et al. (1999) even in stable conditions.

39

2.1

Electric Field

The governing equations to model the electric phenomena are the classic Maxwells equations.
For a medium with electric permittivity and permeability 4 , these equations are known
as: The Gausss law (Feynman, 1970, Vol. 2),
e

(2.1)

B =0

(2.2)

E =

Gausss law for magnetism,

Faradays law of induction,


E =

B
t

(2.3)

and Amperes circuit law,


B = J +

E
t

(2.4)

where E is the electric vector field, B, the magnetic vector field, J, the total current density,
and e the charge density. The permittivity of the medium can be written as a factor, r , of
the free space property (vacuum)
(2.5)

= r 0
Same can be applied to the permeability, where r is the medium factor,

= r 0

(2.6)

The permittivity of the free space in a SI units is given by Farads per meter, 0 = 8.85...
1012

F/m, and the permeability as Henries per meter, 0 4107

H/m. The relation

The symbol is ambiguously defined as permeability of the medium and fluid viscosity. For this section,
in the context of the derivation of the electrostatic equations, is the permeability of the medium. For all
other sections, in the hydrodynamic context, is the absolute fluid viscosity.

40

between permittivity and permeability can be defined with the speed of light in a free space,
c0 = 299, 792, 458 m/s, as
0 =

1
c20 0

(2.7)

From the non-dimensional Faradays law of induction it can be shown that the two fields
become decoupled when E  M ,
E =

M B
E t

(2.8)

where M B0 l/E0 = l2 is the magnetic time scale, and E /, the electrostatics time
scale. Rearranging the time scale inequality, E  M , it can be shown that

l
co 0

2

(2.9)

In classic electrospray experiments, i.e. Gamero-Castao (2010), the diameter of the inner
feeding line tube is 100 106

m, therefore it is expected that the apex of the cone jet is

smaller then this value, implying the jet length scale is of order of l 106 m. The electric
conductivity of the different fluids used in the experiment ranges from 1 101 S/m to
4 106 S/m. Within these experimental ranges of values, the electric field can be regarded
as decoupled from the magnetic field if /  5 109 . Assuming that the fluids used in the
experiments can be regarded as approximately magnetic inert, 1, for a relative dielectric
constant much larger than  5 109 and in an absence of an external magnetic field, the
Maxwell set of equations can be simplified to the electrostatic equations, given by Gausss
law,
e

(2.10)

E =0

(2.11)

E =
and Faradays law of induction,

41

Boundary conditions at the interface, I , can be derived from equations 2.10 and 2.11 using
a pill-box system spanning the interface. This derivation process shows that tangential components of the electric field are continuous across the interface, and the normal components
increase by an amount proportional to the interface charge density, q, given by Coulombs
per unit area, and written as,
(2.12)

(2) 0 E (2) n (1) 0 E (1) n =


(2)

To simplify the notation, and since the outer subregion is a vacuum (r

= 1), the liquid

(1)

relative dielectric constant will be written as r = r , which removes the need to explicitly
(2)

write the vacuum relative dielectric constant, since r = 1.


The electrostatic governing equations are valid for the two subregions, liquid and vacuum. Since the application of Faradays law results in null curl of the electric field for
both subregions, it is convenient to define a potential variable represented by the scalar
( () = 0), such that
E =

(2.13)

where the negative sign is used due to historical convention. For the vacuum region, the
charge density is identically equal to zero, e 0, therefore is a harmonic function and
Laplaces equation holds true,
2 (2) = 0

(2.14)

For the liquid region, Poissons equation for the electric potential holds,
2 (1) =

42

e(1)

(2.15)

2.2

Far Field Electric Potential

Computationally speaking, the vacuum is a domain of finite dimensions. Therefore, the


outer boundary should have explicit conditions, known as the far field electric potential
boundary conditions. The assumptions made in the current work follows Higuera (2003),
and assumes that far away from the conical apex the electric potential is dominated by the
charge distribution along the conical meniscus, known as the Taylor potential. The others
contributions, assumed to be negligible, are the charges distributed along the interface of a
cusp like shape near the apex, and the charges of the droplets distributed along the conical
spray. The charges along the interface near the apex of the cone, the transition regions, are
assumed to be a local phenomenon of negligible dimensions compared with the dimensions of
the conical shape of the meniscus. Although De La Mora (1992) has shown that the electric
potential of the conical spray of charged drops decays with the same order of magnitude
as the Taylor potential, it can be considered a far downstream phenomena. Therefore, its
effect can be considered negligible in the region far upstream where the vacuum boundary
conditions is placed, assuming that the vacuum domain is large enough.
De La Mora (1992) has also shown that in certain conditions, the combination of the Taylor
potential and the conical spray potential can have a considerable effect on the meniscus
region, generating a conical shape with angle different from the one obtained by Taylor
(1964). These effects will not be considered in the scope of this work, and are cited here for
future reference. Therefore, the only charges that will be considered to the far field vacuum
electric potential boundary conditions are the charges present in the meniscus region of
conical shape. The upstream part of the conical shape of the meniscus will be considered to
have defined dimensions with a Taylor angle, = 49.29o .
In the Taylor cone region (region upstream to the transition region), the electric field in
the liquid decays as I/R2 to keep a constant electric current, consequently, the surface
43


velocity induced by the shear stress also decays rapidly, u O R3/2 (Barrero et al.,
1999). Assuming that the normal viscous stress is also smaller compared with the surface
tension, the Taylor solution is recovered where En = O(R1/2 ) (eq. 1.4). Consequently,
the interface electric field is much larger than the liquid electric field. Moreover, since the
electric charge on the jet far downstream in the transition region decays faster than R1/2 ,
the leading-order solution for the vacuum potential when R is the Taylors solution
(eq 1.2).

2.3

Charge Transport Equation

The Gausss and Faradays equations describe the electric field resulting from a spatial
distribution of an electric charge, but do not give any information regarding the spatial
distribution of these charges. In order to do that, it is necessary to establish a charge
transport mechanism. This mechanism was formally introduced by Melcher & Taylor (1969)
for electrolytes, which is now known as the Taylor-Melcher Leaky Dielectric Model, reviewed
by Saville (1997). This model describes the transport of the electric charge been carried
out by the ions present in the liquid phase (electrolytes). Saville (1997) uses a species
conservation equation to account for molecules and ions carried by the flow and movement
due to response to gradients in the electrochemical potential, written as


nk
+ u nk = k ez k nk E + k kB T nk + rk ,
t

k = 1, ..., N

(2.16)

where e is the charge of a proton, z k is the valence of the k th species whose concentration is
nk and mobility k , kB the Boltzmanns constant, T , the absolute temperature.
The LHS of the equation is the material transport of a chemical species. The first term
on the RHS is the ion migration in the electric field. The second term is the transport by

44

diffusion. And the third term is the production due to chemical reactions, since the neutral
species might act as a source of ions in the bulk of the liquid. For a single species, N = 1
and r1 = 0 and ions are produced by reactions at the electrodes, characterizing a unipolar
injection. Charge density can be related to ion concentration by

N
X

ez k nk

k=1

Saville (1997) has extended this analysis for liquids with charge from a single 11 electrolytes
(N = 3) where forward and backward reactions were included to account for ions productions
in the bulk of the liquid, beside the electrodes.
To describe the charge transport mechanism at the interface, species balance equation, eq
2.16, is integrated across the interface assuming that no reaction takes place at the surface,
and diffusion time is much larger than the characteristic convection time, therefore ignored5 .
The interface charge transport can be written as,

+ u s = n (n )u + k KEk n
t

(2.17)

In the RHS of above equation, the first term accounts for change in the charge concentration due to the dilatation and compression of the interface, the second term is the electromigration due to conduction. Since the fluid is surrounded by vacuum, there will be no
electro-migration outside the liquid and for a steady state condition, the transport of charge
at the interface can be written as
s (u) = KE (1) n
where s ( ) is the surface divergence.
5

Saville (1997, pg 35)

45

(2.18)

Equation 2.18 shows that charge at the interface is being carried out (convected) by the
surface velocity at the same time, electro-migration occurs, transferring free charges from
the bulk of the surface of the liquid by conduction. Under these conditions, the total current
can be split into two components, the bulk current, IB , and the surface current, IS , and can
be written as
I = IB + IS

(2.19)

where the bulk current is given by

rs

Ex(1) rdr

IB = 2

(2.20)

and the surface current by


IS = 2rs us

(2.21)

(1)

where Ex is the liquid electric field in the x direction, rs = rs (x) is the interface profile, us ,
the surface velocity and the surface charge density.
The scope of this work does not include reactions in the bulk of the liquid, therefore, charge
will not be created within the bulk. Therefore, the total current is assumed to be constant
along the axis of the jet.

2.4

Electric Forces in The Liquid

The electric forces of the free charge in the presence of a electric potential field, can be
described by
1
FE = e E E 2
2

46

(2.22)

The above equation (eq 2.22) can be written in divergence form as

FE = M
where
1
M ij = Ei Ej ij Ek Ek
2

(2.23)

is the Maxwell stress tensor. This form of the tensor ignores electrostriction forces, since
these forces are only important for dilatational fluid motions Melcher & Taylor (1969).
Combining the isotropic part of the Maxwell stress tensor (the electrically induced pressure)
with the hydrodynamic pressure, the Maxwell stress tensor can be alternatively written as



 
1
e
= o EE o 1
E E
2
e T

(2.24)

The pressure resulting from the combination of the electrically induced pressure and the
hydrodynamic pressure, mentioned above, is the total pressure.

2.5

Mass and Momentum Balance for the Isothermal Flow

The Navier-Stokes equation represents the momentum balance for the liquid jet. To simplify
this balance, it is convenient to list the set of fundamental assumptions used:

1. Viscous, Newtonian, Incompressible liquid.


2. Isothermal flow.
3. Low charge density within the liquid bulk, e(1) v 0, since most of the free charge is
conducted to the interface. The remaining charge in the bulk is spread throughout the
47

volume, resulting in a low volumetric charge density.


4. Constant liquid electric permittivity value, , throughout the volume. This assumption
is based on the fact that electric permittivity (dielectric constant) is only temperature
dependable, = (T ), and the flow is isothermal.
5. The Maxwell stress tensor previously presented, neglects the electrostriction forces
(Gamero-Castao, 2010). This is a reasonable assumption for electric conductive liquids with no dilatational fluid motion (Melcher & Taylor, 1969).
For the incompressible flow assumption, mass conservation can be enforced through the
classical incompressible continuity equation written as,

u=0

(2.25)

where for the cylindrical coordinate is given by,


ur ur uz
+
+
=0
r
r
z

The low bulk charge density assumption, e(1) v 0, simplifies Poissons equation (eq 2.15)
for the electric potential within the liquid domain, resulting in the Laplaces equation,
2 (1) = 0

(2.26)

For momentum balance, the forces due to the electric field are added into the balance equation. Therefore, the classic Navier-Stokes equation for an incompressible fluid needs to be
modified by the introduction of the electric term, written as,

u
+ u u = p + + M
t
48

(2.27)

where M is the Maxwell stress tensor (eq 2.24) and pressure is the total pressure which, as
previously mentioned, already accounts for the electrically induced pressure.
The term M in the equation represents the volumetric forces due to the electric field.
Using the Maxwell Stress Tensor (eq 2.24), these forces can be expressed as,

 



1
1
e
E

o E 2 + e(1) E
= o
2
e T
2

(2.28)

Since the low bulk charge density, e(1) v 0, and constant electric permittivity, the bulk
forces due to electric potential is negligible, M 0, the bulk balance equation recovers
the classic Navier-Stokes equation for an incompressible flow. Since charge density is not
negligible at the liquid/vacuum interface, electric force will be induced there and should be
balanced out.

2.6

Liquid/Vacuum Interface Dynamics

At the interface, besides the electric forces, capillary forces have significance as well. This
happens since both curvature radii of the jet interface eventually become small resulting in
large capillarity forces. To accomplish that, the Young-Laplace equation, first introduced by
Young (1805) and later formalized by Laplace (1829), provides the necessary pressure jump
condition, written as,

p =

1
1
+
RC1 RC2


= ( n
) =

1 + R02 RR00
R(1 + R02 )3/2

(2.29)

where RC1 and RC2 are the two principal radii of curvature. RC1 is the jet cross section radii
of curvature in the r plane, and is always negative, pulling inward and contributing to
the jet collapses. RC2 is the radii of curvature in the r z plane, which may be positive or
49

negative depending on the second derivative of the interface profile.


Accounting all the surface forces, the balance equation can be written as,


(1)

n p + + M

(2)

n = M n

(2.30)

where is the surface tension coefficient, p is pressure, is the viscous stress tensor and
M (1) the Maxwell stress tensor for the liquid and M (2) for the vacuum. Force balance can
be decomposed in the two curvilinear principal directions,


(2)
(1)
t n = t M M n


(2)
(1)
n ( n p + ) n = n M M n

T angential :
N ormal :

(2.31)
(2.32)

In both cases, the right-hand side of the equation represents the total electric forces acting
on the surface, and the left-hand side is the fluid reaction, including the capillarity, pressure,
viscous shear and normal stresses.
The liquid strain rate in cylindrical coordinates is expressed as(Kundu & Cohen, 2004),

e=

1
2

ur
r

ur
r

uz
z

ur
z

uz
r

1
2

ur
x

uz
r

(2.33)

and the viscous stress tensor for a Newtonian fluid is given by

= 2e

50

(2.34)

2.7

Mechanical Energy Equation

The mechanical energy balance is derived from the momentum balance equation with the
inner product of the velocity vector, given by

(vEk ) = (pv) + ( v) : v + v M

(2.35)

where,
Ek =

vv
2

(2.36)

is the kinetic energy.


As previously discussed, the low charge density within the bulk results in negligible electrostatic forces in the bulk, M 0. Integrating the above equation over a general control
volume (CV) will result in the integral form of the mechanical energy balance, written as,

(vEk ) d =

(pv) d +

( v) d

: vd

(2.37)

From the divergence theorem,

( F ) d =

(2.38)

(F n) d

the divergence terms of equation 2.37 can be converted from volumetric to surface integral,
reducing the mechanical energy balance to,

(vEk ) nd =

(pv) nd +

( v) nd

51

: vd

(2.39)

The LHS of equation 2.39 accounts for the kinetic energy balance within the CV. On the
RHS, the first term is the energy associated with pressure. The second term is the balance
of energy associated with the work done by the shear forces on the surface of the CV. Finally, the last term is energy dissipated by shear friction (viscous dissipation) within the CV.
Further development of this equation requires the knowledge of the velocity profile along a
given section of the jet. This velocity profile could be obtained from experimental measurements, which might present a extremely difficult challenge due to the small dimensions of
the jet. Alternatively, it could be obtained from direct numerical simulation (DNS) using
computational fluid dynamic algorithms. One last and albeit less accurate alternative is the
use of simplified velocity profiles. This last alternative might, indeed, help in estimating the
initial conditions to be used in further DNS calculations, as will be shown.

2.8

Scaling Laws

Since Galileos observations, scaling laws have been an important tool for the physical science.
In the context of electrospray in the Taylor cone-jet mode, scaling laws proposals represents
a substantial part of the research work associated with experimental data collection. Indeed,
this is a consequence of the number of independent parameters of the problem, that can be
broadly classified as the material properties of the medium and the operation characteristics
of the spray. As described in previous sections, the material properties controlling the
problem are density, viscosity, permittivity and conductivity for the liquid, interfacial tension
coefficient, and permittivity of the vacuum. The operation characteristics can be further split
in two, the controlling parameters, flow rate, applied electrode voltage and dimensions, and
the outcome parameters, electric current, jet and droplets diameter.
As might be expected, some of these properties and parameters play a minor role in the
electrospray mechanism. Cloupeau & Prunet-Foch (1989) reported that applied electrode
52

voltage and dimensions have negligible effect on emitted current and jet/droplets diameter of high conducting liquids. In fact, this behavior has been already observed by Taylor
(1964), where the hydrostatic solution does not depend upon the aforementioned parameters.
De La Mora & Loscertales (1994) further confirmed this behavior by experimental observation and scaling arguments. Consequently, the applied electrode voltage and dimensions will
not be part of the dimensional analysis, and will not appear in any of the discussed scaling
laws.

2.8.1

De la Moras Current Scaling Law

De la Mora et al. (1990) was among the first authors to propose a scaling law for the jet
diameter (dj ). He found that the number
x = 2 2 d3j / Q2

1/3

(2.40)

where x3 is the ratio of surface tension overpressure, /dj , and Bernouillis dynamic pressure,

U 2 /2. U is the terminal velocity of the jet, U = Q/ d2j . From microscopic measurements,
he observed that, x = O(1), leading to the first scaling law of for the jet diameter
dj Q2 / 2 2

1/3

(2.41)

with good agreement with experiments, valid for dj > 2.8m.


De La Mora & Loscertales (1994) realized that since the jet diameter is much smaller than
the capillary diameter, there must exist a transition region between the Taylor cone, in
the order of the capillary, and the jet, infinitesimally small compared with the capillary.
This limiting behavior is more evident for liquids with high conductivity, since higher the
conductivity, smaller is the jet diameter. Within this region, the liquid interface departs from
53

its equipotential equilibrium state (conical shape) to a cusp like shape, where electric charge
monotonically increases, altering the electric potential of the region. In fact, the physics of
this region is considered to be the only controlling mechanism that will determine the jet
diameter and the current emitted by the Taylor cone-jet phenomena.
From their analysis, the dimensionless flow rate was defined as,
= (KQ/(0 ))1/2

(2.42)

where the characteristic flow rate, Q0 = 0 /K, was used. From experimental data, they
also defined a jet diameter scaling law,
dj (Q0 /K)1/2

(2.43)

different from the one shown in eq. 2.41. Based on simple reasoning, this authors concluded
that the beginning of the transitions region, electric potential and charges should asymptotically match the hydrostatic Taylor (1964) solution of the problem. The mechanism
proposed by De La Mora & Loscertales (1994), assumes that Taylor equilibrium (equipotential) solution is perturbed by electrohydrodynamics effects, where charges on the surface are
convected due to the surface hydrodynamic velocity, and bulk conduction is not sufficiently
fast to restore the equilibrium. To model this mechanism, they used charge balance equation,

s js = jb n

(2.44)

where, js = us q is the surface current, jb = qE (1) is the bulk current and s is the surface
gradient. This balance equation is identical to equation 2.18 previously discussed. Casting
the charge balance equation into an ordinary differential equation of a linear relaxation
prototype,
te

dIs
= (Is Ie )
dt
54

(1)

where dt = ds/dt, te = 0 /K the electrical relaxation time, and Ie = 2Rus En

the

equilibrium current (obtained from Taylor (1964) hydrostatic solution). The relaxation linear
equation allows two limiting solutions. In the range of te dlnIe /dt  1, the solution is Is = Ie .
On the other extreme, in the range where te dlnIe /dt  1, Is depends on its own temporal
history, and the solution is obtained from the time integration. The later limiting case was
termed by De La Mora & Loscertales (1994) as the "charge relaxation phenomena". The
importance of this phenomena can be inferred by the ratio of the electric relaxation time, te

and the hydrodynamic residence time, th = ds/us .


De La Mora & Loscertales (1994) did not directly solved the hydrodynamic problem, instead
approximated it with an inviscid, spherically symmetric sink flow,

ur =

Q
2(1 cos)r2

which is, in fact, the average velocity field for a given flow rate in the spherical frame of
reference. With this approximation, th = r3 / (3a), and considering that for the electric
relaxation phenomena to be relevant, te th , the characteristic electrical relaxation length
can be defined as,

=

3Qte
2 (1 cos)

1/3

where remarkable agreement is found with the jet diameter length scale proposed from
experimental data, given by equation 2.43.
From this analysis, the surface current was approximated as
3

Is = I x 2

55

valid for x = r/  1, where

I =

2sin2 QK
3 (1 cos)

1/2

where for the Taylor angle = T = 49.29 , I = 2.44 (QK/)1/2 .


Based on his theoretical analysis, De La Mora & Loscertales (1994) proposed the characteristic current, I0 = (0 /)1/2 , with dimensionless current dependence of the form
I
= (, , Re )
I0

For limiting case of Re  1, this dependence could be reduced to


I
1

f (, Z)

(KQ/) 2
2

where Z = te3 /Q 3 , and Reynolds number defined as Re = Q 3 /te3 (De La Mora &
Loscertales (1994) experimental data was in the range of 0.506 Re 3.15). Figure 2.2
show the dimensional experimental data measured by De La Mora & Loscertales (1994)
1

where the I Q 2 relation is strongly suggested. This data was scanned from figure 8 of his
original work and dimensionless values were accordingly calculated to generate the plot of
figure 2.3. This figure shows that for different liquid properties, linear regression of the curves
= (, ) have distinct slope for each liquid. Figure 2.4 shows the normalized dimensionless
current where the normalization factor, f = d () /d, is the slope of the set of data for
each individual liquid, obtained from linear regression. The function f = d () /d is show
in figure 2.5.

56

The current emitted by highly conducting Taylor cones

I(nA)

1000

100

10
0.01

FM
W
EG
TEG
BA
1-8OL
0.1

10

100

1000

Q(nl/s)

Figure 2.2: Electric current emitted by electrospray in Taylor cone-jet mode of highly conducting liquids. Data originally presented in the work of De La Mora & Loscertales (1994),
figure 8.

2.8.2

Gan-Calvos Current Scaling Law

The theoretical analysis of Gan-Calvo et al. (1997) has the following main assumptions:

1. Electrical relaxation time, te 0 /K is small compared with the hydrodynamic time,


th L/U (residence time of a fluid particle with velocity U to travel the distance L).
2. All the free charges are confined in a thin layer, close to the surface, of the order of
Debyes length, and small compared with all others length scale of the problem.
3. Charges are relaxed at the surface, otherwise they would not be in steady state condition. This means that te  th 0 /K  LR2 /Q (here, the flat profile jet flow
approximation was used, U Q/R2 ).
57

Dimensionless Spray current vs.


200

150

100

50
FM
W
EG
TEG
BA
1-8OL

0
0

10

12

Figure 2.3: Dimensionless current functional dependence on dimensionless flow rate for electrospray of different liquids. Data plot based on the measurements originally obtained by
De La Mora & Loscertales (1994), figure 8. Figure 9 of De La Mora & Loscertales (1994)
paper shows a similar plot.
The third assumption leads us to conclude that the liquid normal electrical field at the
(1)

(2)

interface is negligible compared with the external one, En  En . From charge equation
(eq 2.18) one has,
U 0 En0 /L KEni
and consequently,
Eni /En0 (0 /K) (U/L) te /th  1

From

-theorem, Gan-Calvo et al. (1997) define the dimensionless functions

I/I0 = f (, , )

58

Dimensionless Spray current divided by f vs.


12

10

/(d/d)

FM
W
EG
TEG
BA
1-8OL

0
0

10

12

Figure 2.4: Dimensionless current normalized by the slope d/d versus dimensionless flow
rate, . The slope of the curves were obtained from linear regression for individual set of
data measured for each liquid. Data plot based on the measurements originally obtained by
De La Mora & Loscertales (1994), figure 8. Figure 12 of De La Mora & Loscertales (1994)
paper shows a similar graph.
and
dj /d0 = g (, , )
choosing , , K and 0 and the dimensionally independent parameters. Based on experimental observation, they ruled out the influence of the electrode voltage and geometry,
including its location relative to the ground. The characteristic current, length and flow rate
were defined as,


I0 =

0 2

1/2

d0 =

59

20
K 2

1/3

Q0 =

0
K

(2.45)

Slope of lines vs for different dielectric constant


20

18

f=d()/d

16

14

12

10

6
0

20

40

60

80

100

120

Figure 2.5: Slope of the dimensionless current as function of dielectric constant. Data
based on the measurements originally obtained by De La Mora & Loscertales (1994), figure
8. Figure 11 of De La Mora & Loscertales (1994) paper shows a similar graph.
respectively, with the dimensionless flow rate written as

1
Q
KQ
=
=

Q0
0

and viscosity dimensionless parameters given by,


= Q0 / (d0 ) = 0 2 / K3

60

1/3

Gan-Calvo et al. (1997) also proposed a set of simplified governing equations to describe
this problem; the momentum equation,
d
dx



1 2
2 (x)
P + v =
2
R (x)

(2.46)

the interface normal stress balance




0
= P Pa +
Eno2 + ( 1) Es2
R(x)
2

(2.47)

I = R2 (x) KEs (x) + 2R (x) v (x)

(2.48)

and the total current

with the outer electric field satisfying the Laplaces equation,


Eo = 0

(2.49)

and surface charge density given by = 0 Eno . These equations have been simplified from
its original counterparts assuming a flat velocity profile, v(x) Q/R(x)2 , relaxed charges at
the surface, te /th  1 (Eni  Eno ) and surface tension for slender jet, n R (x)1 .
Further non-dimensionalization of equations 2.46, 2.47 and 2.48 result in the system of three
dimensionless equations

d
f (z)4 = 4en (z) es (z) f (z)1
dz

(2.50)

f (z)1 = p (z) + en (z)2

(2.51)

es f (z)
+ en (z) f (z)1
2

61

(2.52)

where the following dimensionless variable were defined as,

x/Lo

(2.53)

f (z)

R (z) /Ro

(2.54)

p (z)

(2.55)

en (z)

(P (x) Pa ) Ro /

1/2
o Ro
Eno
2

1/2
o Ro
Es
2

es (z) = ( 1)1/2

I ( 1)1/4

(8QK)1/2

(2.56)
(2.57)
(2.58)

, dimensionless radius and jet slenderness given by,

R0 =
=

Q0 ( 1)1/2
2K

!1/3
(2.59)

R0
Q0
0
=
=
L0
Q
KQ

(2.60)

respectively. Is noteworthy that the slenderness of the jet was assumed to be always small,
and as long as ( 1)1/2  1, the resultant system of equations (equations 2.50-2.52) will
be independent of dimensionless parameters. Gan-Calvo et al. (1997) assumes a transition
region is small compared with any other dimensional length, concluding that the transitions
region is a local phenomenon. This implies that as long as the non-dimensional boundary
conditions for equations 2.50-2.52 be independent of any physical parameters, the eigenvalue
is a pure number ( = O(1)). They concluded than the scaling law for current and jet
diameter from equations, 2.58 and 2.59 given by,

dj

(QK)1/2

(2.61)

( 1)1/4
( 1)1/2 Qo
K
62

!1/3
(2.62)

or in dimensionless form as,


Q

I0
dj
d0 ( 1)1/3

!1/2
(2.63)

( 1)1/2 Q0
Q

( 1)1/2 Q0

!1/3
(2.64)

Gan-Calvo et al. (1997) limited the validity of his analysis for the cases having the characteristic axial length, L Q/, much shorter than the transition length, L  L0 . The
characteristic length L was defined based upon the length to have the variation of liquid
kinetic energy of the same order as kinetic energy itself, (vx2 ) vx2 . This length is the
necessary length to allowthe boundary layer to develop a radial dimension of the same order
of jet diameter. This is achieved only when
1/2  1

(2.65)

Gan-Calvo et al. (1997) also proposed the I Q1/4 scaling law for liquids with low viscosity
and conductivity, which is equivalent to the limiting case of 1/2  1. This scaling law
was based on the Rayleigh maximum charge limit, resulting in the given relations for the
current and jet diameter,
I

I0

Q
Q0

1/4

dj

d0

Q
Q0

1/2

(2.66)
(2.67)

respectively.
Figure 2.6 shows the dimensional measurements obtained by Gan-Calvo et al. (1997).
These data was scanned from Fig. 4 and Fig. 7 in Gan-Calvo et al. (1997) journal article.

63

The two limiting scaling laws were also plotted for reference. The dimensionless counterpart
of these data is plotted in figure 2.7 where the best fit for the two limiting cases is also
shown, and given respectively by,

I/Io = 6.2

!1/2

(2.68)

2.0

( 1)1/2 Qo

for the limiting case of 1/2  1 (the I Q1/2 law), and



I/Io = 11.0

Q
Qo

1/4

(2.69)

5.0

for the case of 1/2  1 (the I Q1/4 law).


Current Vs Flow Rate
I~Q1/2
DIOX66
DIOX50
DIOX25
OCTA
METH
PROP
SUNF
DODE
I~Q1/4
HEP10
HEP04
DIOX4
HEP01
DIOX2

I [A]

1e-07

1e-08
1e-12

1e-11

1e-10

1e-09

1e-08

3 -1

Q [m s ]

Figure 2.6: The current as function of flow rate for the experimental data given by GanCalvo et al. (1997). Solid lines represent the I Q1/2 and I Q1/4 current laws.

64

Dimensionless Current Vs Flow Rate

0.5

(Q/Qo)1/4
2

1.5

2.5

3.5

50

DIOX66
DIOX50
DIOX25
OCTA
METH
PROP
SUNF
DODE

40

30

10

DIOX4*
HEP01*
DIOX2*
BEST FIT

I/Io
20

BEST FIT
HEP10*
HEP04*

0
0

4
5
(Q/Qo(-1)1/2)1/2

Figure 2.7: Experimental data of Gan-Calvo et al. (1997) normalized by Q0 .

2.8.3

Higueras Current Scaling Law

Higuera (2003) approached the electrospray problem by developing a numerical algorithm to


solve the hydrodynamic and electrostatic governing equations simultaneously. The scope of
his work was limited to the Taylor cone-jet mode. The key characteristic of his approach was
to use a vorticity-stream function formulation to solve the hydrodynamics part of the problem, where the governing equations were transformed to a curvilinear system of coordinate
aligned with the interface of the jet, r = r/rs . The electrostatic equations were simultaneously solved in the same curvilinear system of coordinate. For the far field boundary
conditions, Higuera (2003) used Taylor (1964) potential solution for the Laplace s equation
and Barrero et al. (1999) stream function solution and zero vorticity for the stream function
and vorticity equations, respectively. This hydrodynamic boundary conditions were used
upstream only. It is not clear from Higueras paper what was the downstream boundary
65

condition used. In general, when not specified, the downstream CFD simulations use zero
gradient velocity and calculate pressure within the method.
To obtain the dimensionless counterpart of equations, Higuera (2003) scales the variables
accordingly with a set of characteristic parameters. This scale assumes that the transition
region length, Ro , is small compared with any other characteristic length of the problem.
The characteristic velocity of the liquid within this region is given by, vo Qo /Ro2 . The
characteristic flow rate, Qo , can be determined by assuming:

1. Inertia of the liquid in the transition region should scale with surface tension, vo2
1/3

/Ro Ro (Q2o /)

(this comes from the condition that inertia scales with

pressure variation, and consequently, surface tension).


2. Electric stress at the surface should also be included, and from Taylor hydrostatic
balance, Eo (/o Ro )1/2
3. Within the transition region, bulk current is converted in surface current, therefore,
both should be of the same order, Ro vo o KEo R02 .
4. The scales of electric potential are given by Laplaces equation, o = Eo Ro , and surface
charge density by Gausss/Faradays interface jump conditions (Eq. 2.12), o = o Eo .
In particular, these two scales are very convenient since they eliminate the characteristic
parameters from the electrostatic equations and its interface jump boundary condition.

From this assumptions, the characteristic flow rate and, consequently, all the others scales,
can be determined by,

66

Qo =

o
,
K

Ro =

o
K 2

1/3

vo =

K
o

1/3

,
(2.70)

1/3


Eo = 1/25/2K
,
o

1/2

Io =

o
,
1/2

o = Eo Ro ,

o = o E o

These scales are similar to the characteristics parameters presented in the set of equations
2.45, obtained by Gan-Calvo et al. (1997) from dimensional analysis (-theorem). Figure
2.8 show the numerical results obtained by Higuera (2003). The line data were scanned from
the figure 5 and figure 6 shown in his paper, where the points were numerical data available
within the figures caption. For comparison, figure 2.8 also plot the I/Io = 2.6 (Q/Qo )1/2
experimental correlation given by Gan-Calvo (1999), and an arbitrary I Q1/4 curve to
show the trend of the scaling law introduce by Gan-Calvo et al. (1997) for low viscosity
and conducting liquids. From the figure 2.8, is evident that the liquids simulated by Higuera
(2003) were not in the range of validity of the I Q1/4 scaling law.

2.8.4

Gamero-Castao Geometric Scaling Law

Gamero-Castao (2010) has contributed to the understanding of physics of the electrospray


in the Taylor cone-jet regime by experimental investigation of the stopping potential. Using a similar apparatus detailed in Gamero-Castao & Hruby (2001), instrumented with a
Differential Retarding Potential Analyzer (RPA), a second RPA in tandem operation with
a Induction Charge Detector (ICD) (Gamero-Castao, 2009) and a Time-of-Flight (TOF)
collector, he accurately measured the droplet mass, charge, stopping potential and current
emitted by the electrospray.
From the current emitted by the Taylor cone, Gamero-Castao (2010) has observed and
67

Higueras Numerical Results of Flow Rate Vs Current Emitted by a Taylor Cones

I/Io

10

0.1
0.01

=5, Re=0
=5, Re=1
=5, Re=1
=50, Re=0
=50, Re=1
=50, Re=1
=50, Re=10
=150, Re=1 1/2
I/Io=2.6*(Q/Qo)
I~Q1/4
0.1

10

100

1000

Q/Qo

Figure 2.8: Higueras numerical results of dimensionless flow rate versus dimensionless current. The solid lines were scanned from figure 5 (b) of Higuera (2003). The points on the
graph were numerical data collected from within the text. The I/Io = 2.6 (Q/Qo )1/2 is the
Gan-Calvo (1999) experimental correlation, and I Q1/4 is the scaling law introduced by
Gan-Calvo et al. (1997).
confirmed the I Q1/2 law for the current dependence on flow rate, as can be seen from figure
2.9. As previously discussed, this relationship has also been observed by other researchers.
The novelty of his work is that from the stopping potential measurements, which indirectly
accounts for the sum of kinetic and potential energy of the droplet,

s =

Q v (z)2
+ (z) = constant
I
2

(2.71)

an average deficit of 17% of the electrical power has been observed. The logical reasoning
to explain this deficit is that it is been degraded in the cone-jet, instead of being converted
into kinetic power. Another striking observation is that this deficit is independent of the
flow rate. The surface energy is responsible for 10 to 22% of the deficit, the remaining is
68

(/20)1/2

Dimensionless Voltage Deficit Vs Inverse of Reynolds

10
PC2 Re=0.40
PC3 Re=0.74
PC4 Re=0.98
FORM2 Re=0.34
TBP1 Re=0.29
TBP2 Re=0.50
TBP3 Re=0.64
TBP4 Re=1.12
TBP5 Re=2.34
~QH1/2
10

100
QH

Figure 2.9: Dimensionless current emitted by a Taylor cone as function of flow rate. The
data plotted are collected from Gamero-Castao (2010, pg 497, firuge 2b).
degraded between viscous and ohmic dissipation, and both contribute to the increase in the
internal energy of the liquid.
To estimate the voltage deficit from surface energy, viscous dissipation and Joule heating,
Gamero-Castao (2010) integrates the mechanical energy equation in a volume V , delimited
by a r plane at the beginning of the transition region and right before the jet breakup,
and the surface of revolution of the liquid interface. Assuming a flat velocity profile, he
demonstrated that the mechanical energy balance can be given by


v22 p2
+
2

v12 p1
+
2


=

I
P + P
(1 2 )
Q
Q

(2.72)

where subscription 1 and 2 stand for the planes at the beginning of the transition region,
and right before the jet breakup, respectively. Conservation of charge requires that charge69

to-mass ratio () of the droplet be equal to the liquid, d = J = I/Q, and conservation of
momentum requires a jump in velocity,
2R2 R22 p2

Q
Q

v2 v3 =

(2.73)

where subscript 3 stands for a r plane crossing the center of the droplet. Substituting
equations 2.73 and 2.72 into equation 2.71, the expression for the stopping potential can be
simplified assuming that kinetic energy and pressure are negligible upstream to the transition
region (due to the increasing cross-section of the Taylor cone); and that the periodic breakup
occurs within a small gap, so that the potential of the liquid jet is nearly identical to the
potential of the emitted droplets. After some algebra, the stopping potential can be rewritten
as,
P + P 2Q Q 2
s = 1

+
I
IR2
I v 2

p2

R2
2

2

(2.74)

or in terms of voltage deficit ,


(2.75)

= + +
where
= 1 s ,

P
,
I

P
,
I

2Q
IR2

(2.76)

The viscous power dissipated in the cone-jet is given by

P =
and ohmic dissipation by,

: vd V

(2.77)

E J d V

(2.78)

P =

Another striking observation of Gamero-Castao (2010) is that the voltage deficit, , is


70

independent of the flow rate, varying only with liquid properties. This indicates that the
physics of the cone-jet could be also regarded as being independent of the flow rate. From
equation 2.75 is evident that the voltage deficit is proportional to viscous dissipation, which
is proportional to viscosity. Therefore, Gamero-Castao (2010) proposed a linear dependence
of the voltage deficit with the inverse of Reynolds number,
1

= () + ()
0
Re

(2.79)

where the Voltage deficit was scaled with

0 =

2/3
1/6

K 1/3 0 1/6

(2.80)

and Reynolds number defined as


1/3

1/3 0 2/3
Re =
K 1/3

(2.81)

This linear scaling law for voltage deficit with respect to viscosity is supported with experimental observation, as can be seen from figure 2.10. Gamero-Castao (2010) also attempts
to obtain the surface voltage deficit dependence on the operational parameters. He combined
De La Mora & Loscertales (1994) current scale, I = a () (QK)1/2 , with Gan-Calvo (2004)
jet diameter scale,

rGC =

0 Q3
K

1/6

(2.82)

to obtain

= g ()
0

(2.83)

Unfortunately, this law has not yet received full validation from experimental data. This fact
is due to the difficulties associated with the measurements of the jet diameter, as indicated

71

by Gamero-Castao (2010), although, direct estimation of jet diameter of Gamero-Castao


& Hruby (2002) supports the Gan-Calvo (2004) scaling law, and consequently, equation
2.83.
Dimensionless Voltage Deficit Vs Inverse of Reynolds
50

PC
FORM
TBP
/0=5.94/Re+8.60
/0=1.92/Re+5.33

40

/0

30

20

10

0
0

3
1/Re

Figure 2.10: Dimensionless voltage deficit as function of the inverse of Reynolds number.
The voltage deficit was normalized by 0 , as defined in eq. 2.80 and Reynolds number defined
in eq. 2.81.
In order to estimate the dimensions of the transition region, Gamero-Castao (2010) uses
the viscous dissipation term from the mechanical energy equation,

: v = 2

vr
r

2
+

 v 2
r


+

vz
z

2


+

vz vr
+
r
z

2 !

(2.84)

arguing that dissipative process is concentrated on the transition region where the greatest
velocity gradients occur. Considering that the last term of viscous dissipation is negligible

72

and using continuity equation, he bounded the viscous dissipation term by



2

vz
z

2


< : v < 4

vz
z

2

(2.85)

and assuming a flat velocity profile, vz = Q/R (z)2 , integrates the upper bound, showing
that
16Q3/2 1

3a 1/2 K 1/2 rn3

a 1/2 Q1/2

K 1/2 rn



rn
1+
ln

(2.86)



ln
1+
4rn

(2.87)

With this estimation for the viscous and ohmic voltage deficit, the total voltage deficit can
be also estimated from equation 2.75,
1/6

8Q3/2 0 1/6 1
3a 7/6 K 1/6 rn3





1/6
rn
aQ1/2 0 1/6 1
ln
1+
+
1 +
ln
K 1/6 1/6 rn
4rn
= () + g () + ()

K 1/3
1/3

1/3 0 2/3

(2.88)

from where the transition geometric scaling law, rt , can be derived,



rt =

0 Q3
K

1/6

(2.89)

This geometric scaling law can be casted as a geometric average of two length scales, rt =

r rF M , the electric relaxation scale, rF M = (0 Q/K)1/3 , and the scale where the dynamic
1/3

pressure is of the order of capillary pressure, r = (Q2 /)

. Except for the 1/6 factor,

this scaling law is identical to the jet diameter scaling law of Gan-Calvo (2004) (equation
2.82). Gamero-Castao (2010) argued that the 1/6 factor should be included to: a) add

73

physical significance to the electrical relaxation length, rF M ; b) to account for the electrical
relaxation phenomenon in the transition region, as experimentally observed by the influence
of the dielectric constant in the electrospray current and the total voltage deficit; c) The
theory of Gan-Calvo (2004) that resulted in the rGC scaling law, produced a axial length
given by L = (2 KQ3 /0 )

1/3

, in which, combined with radial scaling law, rGC , would result

in a total voltage deficit dependent on flow rate, inconsistent with experimental observations
of Gamero-Castao (2010).
Gamero-Castao (2010) claims that interface profile of the transition region, normalized
by rt , should be independent of the flow rate, surface tension, density and viscosity. To
independently verify those claims, he rescales Higueras interface profile results obtained
from numerical calculations for two different flow rate conditions (Higuera, 2003, figure 2, pg
313). Figure 2.11 show a inset plot with the original interface profile of the transition region
normalized by Higueras length scale, and the main plot shows the same profiles renormalized
with Gamero-Castanos geometric scale, rt .
Inspite of the expectations for a better match of the interface profiles normalized by the rt ,
the results seems to strongly indicate that if not independent of the flow rate, surface tension,
density and viscosity, those factors exert minor influence on it. Alternative justification for
the small discrepancy in the renormalized profiles could be attributed to innacuracies resulting from scanning the profile data from Higueras article. Nevertheless, Gamero-Castao
(2010) suggest that a slight shifting of the axial coordinate could improve this match, where
this shifting should be proportional to the error (in the order of rt /1/6 Q1/2 ) associated with
the asymptotic relation used by Higuera (2003) to fix his system of coordinate.

74

Higueras Interface Profile Normalized With rt


8
20
7

r/rH

6
10

r/rt

4
0
-20

-10

0
z/rH

10

20

QH=0.27, =5
QH=4.80, =5
Taylor Cone

0
-8

-6

-4

-2

0
z/rt

Figure 2.11: Higueras interface profile renormalized by Gamero-Castao (2010) geometric


scale rt . The inset plot shows the original profiles of Higuera (2003, figure 2, pg 313)
obtained from numerical calculation for dielectric constant = 5 and two different flow
rates, QH = 0.27(I = 1.1) and QH = 4.8(I = 5.0).

2.9

Dimensional Analysis

In order to derive the non-dimensional form of the governing equation and its respective
boundary conditions, it is convenient to use generic characteristics values that later will be
defined. Therefore, the non-dimensional variable could be generically defined as,
z = z R0 ,

r = r R0 ,

(2.90)

u = u U0 ,

v = v U0 ,

(2.91)

= 0 ,

E = E E0 ,

(2.92)

= 0

(2.93)

p = p P0

75

Substituting the non-dimensional variables into the governing equation would result in the
following set of governing equations:
Momentum Equation (eq. 2.27):

1
u
+ u u = p +
u + uT
t
Re

(2.94)

where pressure is normalized by dynamic pressure, P0 = U02 , the gradient is in the nondimensional form, = /xi and Reynolds number is defined in its standard form Re =
U0 R0 / (ratio of inertia over viscosity).
Laplaces equation (eq. 2.26):
(2.95)

2 (i) = 0
where superscript (i) is i = 1 for liquid and i = 2 for vacuum/gas.
Surface Current Equation (eq.2.18 ):

d
KE0 R0 (1) 
0 2 1/2

(r v ) =
r E
1 + rs
dx s s
U0 0 s n

(2.96)

These equations are subject to the following boundary conditions:


The far field boundary condition (Taylor solution, eq. 1.2):
T
where6

R0
o 20

1/2

(2.97)

A R 2 P 1 (cos ())
2

1/2

A =

P (cos (T )) sin (T )2 tan (T )


0
1
2

(2.98)

Taylor (1964) has the value of 0.974 but uses P 1 (cos (0 )) with 0 = 130.7099 , where T = 0 .
2

76

with


0

P 1 (cos (T )) =

3
2


cos(T )P 1 ( cos(T )) + 32 P 3 ( cos(T )) sin(T )
2

cos2 (T ) 1

(2.99)

= 0.9746610605954051

and, consequently,
(2.100)

A = 1.7755839907209399

The Interface force balance condition (eq. 2.30):




0 E02 (1)
0 E02 (2)

n
=

n
U02 R0
U0 R0
U02 M
U02 M

(2.101)

where
(2.102)

= 2e
with

T
e = u + u
=


1
2

ur
z

ur
r

ur
r

uz
z

ur
z

uz
r

1
2

uz
r

(2.103)

and
(i)
M

(i)

= E

(i)

(i)



 
1 (i)
(i) (i)

1 (i)
E (i) E (i) ,
2

(i) T

i = 1, 2

(2.104)

The electrostatic interface jump condition (eq. 2.12):


=


0 E 0
E (2) n E (1) n
0

77

(2.105)

Additional to these governing equations and boundary conditions, the flow definitions is also
used, the electric field (eq. 2.13):

(i)

(i)

and flow rate:


U0 R02
Q =
2
Q0

2.9.1

rs

0
E0 R0

un r dr

(2.106)

(2.107)

Higueras Dimensionless System

This set of equations exemplifies the convenience of the dimensional system proposed by
Higuera (2003). In his system, both

=1
U02 R0

(2.108)

0 E02
=1
U02

(2.109)

and

thus, eliminating the need to have any dimensionless coefficient in the interface force balance
equation. Additionally, he also set

0 E0
=1
0

(2.110)

0
=1
E0 R0

(2.111)

from interface charge balance equation, and electric field definition, respectively. With that,
these two equations will also be free from any dimensionless parameter. He finally also

78

removes any dimensionless parameters from the interface charge balance equation by setting
KE0 R0
=1
U0 0

(2.112)

From combining equations 2.108 and 2.112 using 2.110 to simplify, Higuera (2003) obtained
the length scale used in his calculations, given as,

R0 =

 13

20
K 2

(2.113)

From equation 2.108, the velocity scale,



U0 =

K
o

1/3

(2.114)

and consequently the flow rate scale


Q0 = U0 R02 =

o
K

(2.115)

From equation 2.109 the electrostatic field characteristic parameter,



E0 =

1/2 K

1/3

(2.116)

5/2

and from equations 2.110 and 2.111 the characteristic electric charge density and the characteristic electric potential, given respectively by,
1/2
0 = 0 E0 = 1/2
K
o

and
0 = E0 R0 =

2
1/2

1/2 0 K

79

1/3

(2.117)

!1/3
(2.118)

With those characteristics values, the dimensionless set of governing equations for steady
state conditions is given by
u u = p +


1
u + uT
Re

(2.119)

2 (i) = 0

(2.120)



d
0 2 1/2

(1)
(r v ) = rs En
1 + rs
dx s s

(2.121)

with the following boundary conditions


1/2

T =

P (cos (T )) sin (T )2 tan (T )


0
1
2

R 2 P 1 (cos ())
2

(2.122)



1
M (1)

+
n p +
n = M (2) n
Re

(2.123)

= En(2) En(1)

(2.124)

Notice that the dimensionless parameter in the electric potential far field boundary condition,
the Taylor solution, has also conveniently canceled out, R0 /o 20 = 1. Higuera (2003)
provides physical arguments to justify his choice of parameters, among them are the fact
that in the transition region, dynamic pressure should be of the order of electrostatic pressure
and capillary pressure, and also bulk current should be in the order of surface current for
the bulk charge been conducted from the liquid to the surface.
Indeed the wise choice of parameter provided by Higuera (2003) resulted in a convenient set
of equations free from any dimensionless parameter, except for Reynolds number. What is
80

not so convenient is the fact that each Taylor cone jet will have different non-dimensional
length and terminal diameter according to the flow rate and fluid properties selected for the
simulation.
In the dimensional space, each Taylor cone jet should have different dimensional length
and terminal diameters, and those dimensions should be proportional to the operational
conditions and fluid properties. Nevertheless, from the theoretical point of view, to compare
the different Taylor cone jets would be highly convenient if a dimensional system could be
designed such that the interface profile length and terminal diameter can be similar for all
operation conditions and fluid properties.

2.9.2

Dimensionless System For The Invariant Transition Region

With the goal of obtaining a system of characteristics parameters that would facilitate the
comparison and classification of different Taylor cone jets, we follow the footsteps of GameroCastao (2010) in the sense that the terminal jet diameter should be the length scale of
choice. Supported with experimental data of Gamero-Castao & Hruby (2002), the terminal
jet diameter scale, proposed by Gan-Calvo (2004) and expressed in equation 2.82 is an
excellent candidate, except for the factor 1/6 included in the scale as previously discussed
by Gamero-Castao (2010)7 . Therefore, the characteristic dimensional length scale used for
this system is

Ro =

0 Q3
K

1/6

(2.125)

The author of this work is not complete convinced that: ...the geometry of the transition
region normalized with rt should be insensitive to changes of the liquid flow rate, surface
tension, density and viscosity. as stated by Gamero-Castao (2010, pg 509, 1st paragraph).
The reason for that, as will be demonstrated later, is that the governing equations do not sup7

see section 2.8.4 for a brief summary of this discussion.

81

port such strong statement, and apparently, neither the normalized Higueras8 profile shown
on figure 2.11, although, Gamero-Castao (2010) has justified the differences by blaming the
asymptotic relation used by Higueras to fix his system of coordinates. Instead, here we propose a less stringent statement saying that the geometry of the transition region normalized
with R0 should have low sensitivity to liquid flow rate, surface tension density and viscosity,
therefore, not detectable by the voltage deficit measurements techniques. Nevertheless, the
proposed dimensionless system here, will be based on this quasi-invariant interface profile
length scale, since, in spite of any differences in the interface profile due to the changes in
the flow rate and fluid properties, it is assumed that the length of the transition region will
remain invariant.
By assuming a dimensionless invariant transition region, the dimensionless far field electric
potential should remain invariant. To accomplish that, it would be necessary to have
R0
=1
o 20

(2.126)

It is also convenient to have the definition of the electric field and the electrostatic interface
boundary condition free from any dimensionless parameter, therefore from equations 2.105
and 2.106

0 = 0 E0

(2.127)

0 = E0 R0

(2.128)

Assuming that the far field electric potential is invariant with flow rate, it is equivalent to
8

Higuera (2003)

82

say that electrostatic forces is of the same order of the capillary forces,
0 E02 =

R0

as can be easily demonstrated by substituting definitions 2.127 and 2.128 into equation 2.126.
Another feature we wish to have in this proposed dimensionless system of equations, is a
inlet flow rate boundary condition not depending on the real flow rate,
u =

Q
R2

(2.129)

which means Q = 1 and consequently, Q0 = Q, therefore,


Q
R02

U0 =

(2.130)

With this set of equations all the characteristics parameters can be defined in the following
order,

0 =

R0
0

1/2

5/2 1/2 1/2 Q3/2

0 R0


0 = 0 E0 =

2K 2
2 2 20

1/2

0
R0

(2.131)

5/2

0 K 1/2

Q
U0 = 2 =
R0

0
E0 =
=
R0

!1/6

1/2
=

1/6

(2.132)

!1/6

K 1/2 7/2
7/2

Q3/2 1/2 0 1/2

5/2
K 1/2 7/2 0
Q3/2 1/2 1/2

(2.133)

!1/6
(2.134)

Using the characteristics values, the dimensionless governing equation is given by: The
83

momentum balance equation for steady state,

u u = p +


1
u + uT
Re

(2.135)

Laplaces equation,

(2.136)

2 (i) = 0
and the interface charge balance equation,


d
KE0 R0 (1) 
0 2 1/2

(r v ) =
r E
1 + rs
dx s s
U0 0 s n

(2.137)

where the following dimensionless parameters naturally appears in the equations: The
Reynolds number,
U0 R0
=
Re =

KQ3 5
6 0

1/6

(2.138)

measuring the ratio of inertia over viscous forces, and,

KE0 R0
Le =
=
U0 0

KQ

1/2

(2.139)

quantifying the ratio of bulk current over surface current.


We were unable to locate in the literature the name of the dimensionless number representing
the bulk/surface current ratio, therefore, we name it Leaky Dielectric number, after TaylorMelcher Leaky Dielectric Model. This model has its roots in Taylor (1966) but was formally
introduced by Melcher & Taylor (1969)9 . The Leaky Dielectric Model is used to derive the
surface charge balance equation, reviewed by Saville (1997), and used in the present work.
9

Saville (1997)

84

The Leaky Dielectric number does not appear in the system of equations derived by Higuera
(2003) dimensionless system, since he conveniently chose dimensionless bulk current on the
same order as the dimensionless surface current, making it easy for comparison. Therefore,
KE0 R02 = U0 0 R0 .
We do agree that dimensional bulk and surface current should be of the same order, but not
necessarily the dimensionless one. Indeed, the assumption that the dimensionless bulk and
surface current have to be of the same order, combined with the given choice of dimensionless
flow rate, would produce a scale associated with the length necessary for the bulk charges
to balance out the charges depleted from the surface, as shown by Higueras dimensionless
system. In the dimensionless system proposed here, the length scale is fixed to the scale of
the transition region, which, combined with a invariant inlet dimensionless flow rate, will
result in a bulk/surface ratio (the Leaky Dielectric number) necessary to accommodate the
charge relaxation process, where bulk charge is transfer to the depleted surface.
The Leaky Dielectric number is a dimensionless number that accounts for the physical capability of the liquid to transfer, by conduction, charges from the bulk to the surface, where
charge is been depleted at the ratio proportional to the flow rate conditions. Having said
that, one might expect that the flow rate information should be embedded in the Leaky
Dielectric number, as, indeed, it is. It is also reasonable to think that this number might be
strongly associated with the upper and lower bound operation conditions of a stable Taylor
cone jets, since charge relaxation is the main physical mechanism regulating this type of jets.
It is interesting to notice that Reynolds number is also dependent upon flow rate. Intuitively,
that might be against the initial thoughts that in this quasi-independent dimensional system,
both interface geometry and flow pattern should match for different conditions of flow rate.
To resolve this conflict, we borrow ideas from another fluid problem, the HagenPoiseuille
problem. In the Poiseuille laminar case, two cylindrical pipes having different diameters and
operating in a different flow rate conditions would have different Reynolds number, based on
85

pipe diameter and flow rate. Nevertheless, the normalized velocity profile in both cases is
the same, the parabolic profile. That occurs because although the two pipes have different
Reynolds number, the shear stress at the pipe wall is also different and proportional to the
Reynolds number. Naturally, in the pipe flow problem, we dont need to impose the stress
boundary condition, since substitution with the no-slip velocity condition is sufficient to
solve the problem. The key point to successfully solve the Poiseuille problem is to assume
zero velocity gradient and constant pressure gradient in the pipe axial direction (constant
pressure drop). Obviously, these assumptions does not apply to the Taylor cone problem,
but it is reasonable to think that pressure drop and velocity gradient in the Taylor cone
problem, somehow, combined with viscous forces would produce a similar (or quasi-similar)
interface geometry when normalized by a specific R0 . With the appropriate choice of U0
it is expected that flow pattern should also be similar, regardless of the Reynolds number.
In fact, appearance of the flow rate in the Reynolds number dimensionless parameters is
also strong evidence that the interface profile could be quasi-invariant with flow rate, but
definitely, not independent. Although, transition region length should be invariant.
This governing equations are subjected to the following boundary conditions: The far field
Taylor potential condition,
1/2

T =

P (cos (T )) sin (T )2 tan (T )


0
1
2

R 2 P 1 (cos ())
2

(2.140)

the electrostatic interface jump condition

= En(2) En(1)
and, the force balance interface condition,

86

(2.141)


1
1 M (2)
1
1 M (1)

n=
n p +
+

n
We
Re
Ge
Ge

(2.142)

In the force balance equation, additional to the Reynolds number, Re, two dimensionless
numbers appears, the ratio of capillary forces to the inertia, known as Webber number,

U02 R0
=
We =

KQ
0

1/2

(2.143)

and the ratio of inertia to electrostatic forces,


U02
Ge =
=
0 E02

KQ
0

1/2

(2.144)

named in the scope of this thesis as Geoffrey number10 , after Sir Geoffrey Ingram Taylor,
G.I. Taylor for short. Sir Taylor was a British physicist and mathematician whose work has
brought great contribution to the modern views of the electrohydrodynamics field (Taylor,
1964), among others. As previously demonstrated, due to the choices of the far field electric
potential boundary conditions, the Geoffrey number is identical to the Webber number. This
is not so surprising after we realize that by fixing the far field electric potential boundary
condition to the Taylor potential, the electric forces will be in the order of the capillary
forces, since Taylor solution was based on that equality, and indeed representing the correct
physics upstream to the transition region. Therefore, W e = Ge. Further simplification of
the force balance interface equation lead to,

n p W e + Ca + M (1) n = M (2) n

(2.145)

reducing the number of dimensionless parameters to Webber number and the Capillary
10

It is possible that this dimensionless parameter has already been named after another scientist. We
havent found a reference in literature, and propose Sir G.I. Taylors name in recognition of his great contributions to the field of electrohydrodynamics.

87

number,
U0
Ca =
=

K3
2 0

1/3

(2.146)

quantifying the ratio of viscous forces over capillary forces.


The dimensional analysis proposed for the quasi-independent system has not posed any
constraint to the definition of the characteristic current. Following the work of De La Mora
& Loscertales (1994) we define the characteristic current as,
(2.147)

I0 = (KQ)1/2
resulting in the following dimensionless total current equation,

I =

R0 U0 0
I0

2rs us


+

KE0 R02
I0

rs

Ex(1) r dr

(2.148)

with
R0 U0 0
I0
=
=
I0
KE0 R02

2.9.3

0
KQ

1/4

(2.149)

Alternative Dimensionless System For The Invariant Transition Region

While developing the dimensionless system for the invariant transition region, we realize
that, although it is very effective to compare the velocity field among different flow rate
conditions, the previous proposed set of characteristics parameters is not so convenient to
compare bulk and interface current. That happens because we constrain length scale, R0
and velocity scale, U0 , therefore, convenient to impose a invariant inlet velocity condition.
Consequently, Le 6= 1.
Alternatively, we would like to have a dimensionless system to compare bulk and surface
88

current, inspite of possible loose the invariant inlet velocity boundary condition constraint
(dont understand the last part of the sentence....). To accomplish that we alternative constrain Le = 1, similarly as Higuera (2003) did for his dimensionless system. That will result
in a characteristic velocity parameter as
KE0 R0
0

U0 =

and consequently a characteristic flow rate parameter as


KE0 R03
0

Q0 =

Using the same previously defined length scale, electric potential and electric charge (equations 2.125, 2.126, 2.127 and 2.128), the alternative set of characteristic parameters can be
derived, as

0 =

R0
0

1/2

5/2 1/2 1/2 Q3/2

5/2

0 K 1/2

KR0
U0 =
=
0


Q0 =

E0 =

K 5 Q3
50

KQ3
0

1/2

7/2

Q3/2 1/2 0 1/2

K 1/2 7/2 0
Q3/2 1/2 1/2

89

(2.150)

(2.151)

(2.152)

!1/6

K 1/2 7/2

5/2

0 =

1/6

!1/6

(2.153)

!1/6
(2.154)

U0 R0
Re =
=

K 2 Q3 4
3 20

U02 R0
W e = Ge =
=

U0
Ca =
=

1/3

KQ1/3
0

K 5 Q3 6
7 50

3/2

1/6

(2.155)

(2.156)

(2.157)

The dimensionless set of governing equation then becomes,

u u = p +


1
u + uT
Re

(2.158)

2 (i) = 0

(2.159)



d
0 2 1/2

(1)
(r v ) = rs En
1 + rs
dx s s

(2.160)

subject to the following boundary conditions,

T = A R 2 P 1 (cos ())

(2.161)

= En(2) En(1)

(2.162)

90


n p W e + Ca + M (1) n = M (2) n

(2.163)

As initially intended, the resultant total current equation is free from any dimensionless
parameter and given by,

I =

2rs us

rs

+ 2

Ex(1) r dr

(2.164)

where the following characteristic current,


I0 =

K 3 Q3
0

1/4

(2.165)

was used to derived the non-dimensional equation.

2.9.4

Rescale of the Dimensionless Systems

Higuera (2003) has provided substantial amount of information regarding the Taylor cone
jets, including two jets profiles operating on different flow rate conditions. In order to compare with Higueras work, the numerical model that was developed within the scope of this
thesis uses the dimensionless system proposed by him. Therefore, it is necessary to translate
data from Higueras dimensionless system to the invariant systems proposed here. To designate the Higueras characteristic parameters the subscript H will be used. Subscript I and
AI will label the characteristics parameters of the Invariant and Alternative Invariant
system, respectively.
Any dimensionless variable can be rescaled to another system by multiplying it with the
ratio of the characteristic parameter of that particular variable. For example, any length can

91

be rescale as,
rI =

RH
r
RI H

where rI = r/RI is the dimensionless length in the radial direction of the quasi-independent
dimensionless system. Table 2.1 have a summary of characteristics parameters and its corresponding constant of proportionality for the Invariant and the Alternative Invariant
dimensionless system. Table 2.2 provides the constant for the alternative quasi-independent
system.

2.9.5

General Considerations Regarding the Differences Among the


Dimensionless Systems

While developing the different dimensionless system of equations, we were able to generalize
a mathematical procedure to generate any set of characteristic parameters and dimensionless
equations. This procedure solves the problem of defining seven characteristic parameters,
R0 , 0 , 0 , E0 , Q0 , U0 and I0 , respectively the characteristic length, electric potential,
electric charge density, electric field, flow rate, velocity and electric current. Regardless of
the dimensionless system, it is desirable that three dimensionless definitions should be free
from any dimensionless parameter, the velocity flow rate relation (eq. 2.107), electrostatic
jump condition (eq. 2.105) and electrostatic field definition (eq. 2.106), Therefore, the
following invariant relations should be satisfied in any dimensionless system,

Q0 = U0 R02

0 = E0 R0

0 = 0 E0

(2.166)

Additional to this invariant relations, we describe four dimensionless parameters in the nondi-

92

93
2

2 0
K3




1/3

CaH = Re1
H

1/3

1/2

1/2 K
5/2
o

0 K

W eH = 1

ReH =

IH =

EH =

o
QH = K
1/3

1/2
H = o 1/2 K

1/3
2
H = 1/2 1/2

Leaky Dielectric Number


Le
LeH = 1
Geoffrey Number
Ge
GeH = 1
KQ

QH = 0 is Higuera (2003) dimensionless flow rate.

We
Re

We

Webber Number
Ca =

Re

Reynolds Number

Capillary Number

I0

Electric Current

Electric Potential
E0

Surface Charge Density

Electric Field

Q0

Flow Rate
5/2

K 1/2 7/2 0
Q3/2 1/2 1/2

1/6

LeI = (QH )1/2


GeI = 1

II = (KQ)1/2
 3 5 1/6

ReI = KQ
0
 1/2
Q
W eI = H

 13
K3
CaI = 2 0

I =

1/6
5/2 1/2 1/2 Q3/2
I =
5/2
0 K 1/2

1/6
K 1/2 7/2
EI =
3/2 1/2 7/2 1/2

QI = Q

1/2

LeH /LeI = (QH )1/2


GeH /GeI = 1

ReH CaI = 1/3

W eH /W eI = (/QH )1/2

ReH /ReI = 1/3 /QH

IH /II = (QH )1/2

1/4

1/4
EH /EI = QH 1/3

H /I = QH 1/3

QH /QI = Q1
H
1/4
H /I = QH 1/3

Table 2.1: Dimensionless system rescaling parameter between Higeras and the Invariant system.
Description
Symbol Higeras Parameters Invariant Parameters Higueras / Invariant
 2  13

1/6

0
0 Q3
1/3 1/2
Length
R0
RH = K 2
RI =
R
/R
=
Q

H
I
H
K
1/3
 1/3

K
Velocity
U0
UH = K
U
=
UH /UI = 1/3
I
o
0

94

Leaky Dielectric Number


Le
LeAI = 1
Geoffrey Number
Ge
GeAI = 1
QH = KQ
is Higuera (2003) dimensionless flow rate.
0

LeH /LeAI = 1
GeH /GeAI = 1

Table 2.2: Dimensionless system rescaling parameter between Higeras and Alternative Invariant system.
Description
Symbol Alternative Invariant Parameters Higueras / Alternatives

1/6

0 Q3
1/3 1/2
Length
R0
RAI =
R
/R
=
Q

H
AI
H
K
 5 3 1/6
Velocity
U0
UAI = K Q5
UH /UAI = RH /RAG
 30 1/2
3/2

Flow Rate
Q0
QAI = KQ
QH /QAI = QH 1/3
0

1/6
5/2
1/4
K 1/2 7/2 0
Surface Charge Density
0
AI = Q3/2 1/2 1/2
H /AI = QH 1/3

1/6
1/4
5/2 1/2 1/2 Q3/2
Electric Potential
0
AI =
H /AI = QH 1/3
5/2 1/2
0 K

1/6
1/4
K 1/2 7/2
Electric Field
E0
EAI =
EH /EAI = QH 1/3
7/2 1/2
3/2
1/2
Q 0
 3 3 1/4
3/4
Electric Current
I0
IAI = K Q0
IH /IAI = QH 1/3
 2 4 1/3
1
Reynolds Number
Re
ReAI = KQ
ReH /ReAI = QH 1/3
3 2
 3 3 0 3 1/2
1/2
Webber Number
We
W eAI = k Q3
W eH /W eAI = (Q3
3
H )
0
 5 3 6  16
1/2
K Q
e
Capillary Number
Ca = W
Ca
=
ReH CaAI = QH 1/3
AI
Re
7 5

mensionalization process, the Webber number,

We =

U02 R0

(2.167)

Re =

U0 R0

(2.168)

Le =

KE0 R0
U0 0

(2.169)

the Reynolds number,

the Leaky Dielectric number,

and the Geoffrey number


U02
Ge =
0 E02

(2.170)

Regardless of the dimensionless system used, assuming that far field electric potential should
satisfy Taylor solution is equivalent to saying that capillary forces are of same order of
magnitude as electrostatic forces, therefore, W e = Ge and consequently,

= 0 E02
R0

(2.171)

With these four invariant properties, four parameters can be obtained, the characteristic
velocity,
U0 =

R02
Q0

(2.172)

the characteristic electric field,




0 R0

R0
0

E0 =

1/2

(2.173)

the characteristic electric potential,

0 =

95

1/2

(2.174)

and the characteristic charge density,



0 =

0
R0

(2.175)

showing that only three characteristic parameters needed to be arbitrarily defined, Q0 , R0


and I0 .
Up to this point, there is no difference among all the dimensionless system of equation
and characteristic parameters. For Higueras11 system, by assuming that, in the transition
region the dimensionless surface electric charge is of the same order as bulk charge, Le = 1,
additional constraint is obtained,
Q0 K = R0 0

(2.176)

Finally, Higuera (2003) also assumed that dynamic pressure should be on the same order as
capillary pressure, W e = Ge = 1, leading to another constraint
Q20 = R05

(2.177)

allowing him to obtain his characteristic length scale,



R0 =

20
K 2

1/3

(2.178)

and consequently all the other characteristic parameters except for I0 , which is conveniently
selected to be
I0 = KE0 R0 = U0 0

(2.179)

elimination any dimensionless parameter from the total current equation.


It is worth noting that the two additional constraint obtained by Higuera (2003) to defined
11

Higuera (2003)

96

his system of dimensionless parameters, is equivalent to the relation Le = W e, leading to


the expression to the invariant length scale
U03 0
=1
K

R0 =

Q30 0
K

1/6

(2.180)

Additionally, he also constrained Le = 1, resulting in his characteristics parameters. Similarly to Higueras, the proposed invariant dimensionless system also assumes Le = W e 6= 1,
but arbitrarily defines Q0 = Q, resulting in the system with a invariant inlet dimensionless
flow rate, Q = 1, suitable to velocity comparison, as extensively discussed. The inclusion
of the parameter 1/6 in to the length scale, can be simply justified by a arbitrary choice
of the definition of both, the electrostatic jump condition (eq. 2.105) and electrostatic field
definition (eq. 2.106).
Different from both previously discussed systems, the alternative invariant system does not
constrain Le = W e, instead, assumes Le = 1 and constrains the length scale to

R0 =

Q3 0
K

1/6

resulting in the system conveniently designed to compare surface charge, as previously discussed.
Regardless of the dimensional system used, the set of governing equations can be expressed
as following,
u u = p +


1
u + uT
Re

2 (i) = 0

97

(2.181)

(2.182)



d
0 2 1/2

(1)
(r
v

)
=
Ler
E
1
+
r
s n
s
dx s s

(2.183)

subject to the following boundary conditions,

T = A R 2 P 1 (cos ())

(2.184)

= En(2) En(1)

(2.185)


0
n p + Ca + M (1) n = M (2) n

(2.186)

where p0 = W ep . Accordingly with table 2.1, for the invariant dimensionless system proposed, Ca and the inlet dimensionless flow rate condition are invariant with flow rate. Dimensionless shear stress is proportional 0 E0 Q1/8 , therefore, of small influence from flow
rate in the order of unit or higher. The only remaining factor is Re Q1/2 , controlling the
diffusion and, consequently, the viscous dissipation. The diffusion term in the momentum
balance equation is not the dominant term, since inertia and pressure gradient is of the order of unit. This analysis shows that Taylor cone jet interface profile should be very similar
regardless of the flow rate condition, but it can not be regarded invariantly identical, since
flow rate does have a small contribution to momentum diffusion, proportional to Q1/2 .

98

Chapter 3
The Boundary Element Method
The Boundary Element Method is a numerical technique based on Greens theorem, created
in order to solve a specific class of partial differential equations. With this method, diffusion
problems can be solved without the need for a volumetric discretization of the computational
domain.

3.0.6

The Greens Identities

The Boundary Element method to solve Laplaces equation is derived from Greens first
identity. Greens identity states that any two functions, twice-differentiable, f (x) and g (x),
satisfy the equation:
g2 f = (gf ) g f

(3.1)

and its converse relation (interchanging f and g).


Subtracting Greens first identity from its converse counterpart, Greens second identity is
defined as:
g2 f f 2 g = (gf f g)
99

(3.2)

When, these two functions, f (x) and g (x) are either singular or non-singular harmonic functions (satisfy the Laplaces equation, or singularly forced Laplaces equation), the reciprocal
relation for harmonic functions is also defined as:

(gf f g) = 0

(3.3)

Greens function is a special kind of function with the required properties needed to develop
the BEM. The first characteristic of Greens function is the ability to satisfy the singularly
forced Laplaces equation, given as:
2 G (x, x0 ) + (x x0 ) = 0

(3.4)

where x is an arbitrary variable location and x0 is the fixed location of the "singular point"
or "pole" and is the Diracs delta function.
Substituting g (x) by Greens singular harmonic function, g (x) G (x, x0 ), Greens second
identity (eq 3.2) becomes

f (x x0 ) = (G (x, x0 ) f f G (x, x0 ))

(3.5)

To further derive the method, figure 3.1 defines an arbitrary region in space delimited by
the boundary , where n is the unit normal vector pointing out of region and perpendicular
to its boundary, . For three dimensional problems, is a volumetric region in R, and ,
its surface boundary in R. For two dimensional and axisymmetric problems, is a area
region in R, and its line contour boundary in R.

100

Figure 3.1: (a) Control Volume region delimited by the boundary interface . (b) BEM
discretization of the interface .
Integration of Greens second identity (eq 3.5) over the region ,

(G (x, x0 ) f f G (x, x0 )) d

f (x x0 ) d =

(3.6)

including its boundary , might result in three possibilities: If the singular point x0 is placed
outside the region, resulting in

(G (x, x0 ) f f G (x, x0 )) nd = 0

(3.7)

If x0 is located anywhere inside the region ,

f (x) G (x, x0 ) nd =

f (x0 ) +

G(x, x0 )f (x) nd

(3.8)

and if x0 is located anywhere at the boundary ,


1
f (x0 ) +
2

f (x) G (x, x0 ) nd =

G(x, x0 )f (x) nd

101

(3.9)

Some of the steps to obtain these results were omitted (see Pozrikidis (2002); Brebbia et al.
(1984) for more details) from the text, but is worth mentioning that the derivation steps
included the use of the integral property of the delta function, given by,

(x x0 ) d = 1

x0

f (x) (x x0 ) d = f (x0 )

x0

f (x) (x x0 ) d = 0

x0 6

(3.10)

the Gauss divergence theorem (eq 2.38) and integration by parts.

3.0.7

The Boundary Element Discretization

A important result demonstrated by the Greens boundary-integral equation 3.9, is the fact
that a integral of Greens second identity over a region , can be expressed as a integral
over the boundary . The Boundary Element Method takes advantage of this result and
subdivides the interface of region in a finite number of cells, or elements (see figure 3.1),
therefore the boundary-integral representation for a point located at the interface can be
expressed in a discretized form as:
N

X
1
f (xi ) +
2
j=1

f (xj ) G (xj , xi ) ndj =


Ej

N
X
j=1

G (xj , xi ) f (xj ) ndj

(3.11)

Ej

Assuming that all elements are sufficient small, the simplest approximation is to assume that
f (x) and f (x) n is constant along the element, which simplifies the elemental integral

102

to

f (xj ) G (xj , xi ) ndj f (xj )

Ej

G (xj , xi ) ndj

G (xj , xi ) f (xj ) ndj f (xj ) n


G (xj , xi ) dj

(3.12)

Ej

Ej

(3.13)

Ej

resulting in a discrete boundary-integral equation


N

X
X
dfj
1
ij fj =
fi +
H
Gij
2
dn
j=1
j=1

(3.14)

where

ij
H

G (xj , xi ) ndj

(3.15)

G (xj , xi ) dj

(3.16)

Ej

Gij

Ej

was written in short notation (fi f (xi ) and

dfj
dn

f (xj ) n) for simplification. It is

possible to consider other approximations allowing longer elements to be used, but since the
computational cost associated with the calculation of higher order approximations rises significantly, it is often more convenient to increase the number of elements, reducing the length,
to achieve the same accuracy. Therefore, throughout this work, the simplest approximation
will be used.
When a finite number of singular points xi are placed at the center of each boundary element,
an algebraic linear system of equations can be assembled as

Hu = Gq

where u [f1 , f2 , . . . , fn ]T , q

, df2 , . . . , dfdnn
dn dn

 df1

103

T

+ 1 I.
and H H
2

(3.17)

Then BEM linear system of equations (eq 3.17), equates the boundary values of a harmonic
function, f (x), with its gradient normal to the interface. Therefore, if the value of a harmonic
function is defined at the boundary (u is known), solution of the system 3.17 will provide
the gradient normal to the interface (q is the unknown), and vice-versa.
For general problems, the two types of boundary conditions coexist, and reordering the
system will be necessary to separate the unknown vector to the left-hand side (LHS) and
the known to the right-hand side (RHS), resulting in a standard system of equations,

Ay = F

3.0.8

(3.18)

The Fundamental Solutions

The fundamental solution is the solution for the singularly forced Laplaces equation
2 G (x, x0 ) + (x x0 ) = 0

(3.19)

The result of this equation defines Greens function G (x, x0 ), where x = {x, y, z} is the
variable "field point", x0 = {x0 , y0 , z0 } is the fixed location, a singular point, the pole.
There are different kinds of Greens function, according with the boundary conditions used.
Greens function of first kind satisfies equation 3.19 and zero value at the boundary of the
domain ,
:

G (x, x0 ) = 0

(3.20)

Greens function of second kind, the Neumann function, satisfies equation 3.19 and zero

104

normal gradient along the boundary of the domain,

(3.21)

G (x, x0 ) n (x) = 0

Free-space Greens function is the fundamental solution of equation 3.19 in a infinite domain
(no boundary conditions). The three dimensional free-space Greens function is given by:

G (x, x0 ) =

1
4x

(3.22)

where x = |xx0 | is the distance of the field point to the pole. It can be demonstrated that
the two dimensional free-space Greens functions can be derived from the three-dimensional
one by integrating over a line source point located at x0 perpendicular to the two-dimensional
plane. The two dimensional space, the free-space Greens function can be written as:
1
G (x, x0 ) =
ln
2

1
x

(3.23)

The axisymmetric fundamental solution can also be derived from the three-dimensional freespace Greens function by integrating over a ring source located at x0 = {r0 , z0 }, as shown
in figure 3.2. The three-dimensional free-space Greens function in cylindrical coordinate,
{r, , z}, is written as

Gc (x, x0 ) =

1
4 ro2 + r2 2r0 rcos (0 ) + (z0 z)2

 21

(3.24)

where the integration along the ring source result in the axisymmetric Greens function,

105

ring source

Figure 3.2: Three dimensional Taylor Cone jet schetch. The ring source located at x0 is
integrated over direction to obtain the axisymmetric fundamental solution (axisymmetric
Greens function).
given by:

GAx (x, x0 ) =

4 ro2 + r2 2r0 rcos (0 ) + (z0 z)2


4K (m)
0

(a + b) 2

 21 d =
(3.25)

where K (m) is the complete elliptic integral of the first kind, and,

m=

2b
a+b

(3.26)

a = r02 + r2 + (z0 z)2

(3.27)

b = 2r0 r

(3.28)

106

The gradient normal to the boundary can be written as

GAx (x, x0 ) n =

1
1

(a + b) 2

!
r02 r2 + (z0 + z)2
E (m) K (m) nr +

ab

z0 z
+
E (m) nz
ab
1
2r

(3.29)

where nr and nz are the components of the surface normal vector, n = {nr , nz }; m, a and
b were previously defined on equations 3.26-3.28, K (m) and E (m) are the complete elliptic
integral of first and second kind, respectively.
Substituting the definition of the Greens function into the element integral equation (eq 3.16
and eq 3.15), the coefficients of the BEM method can be calculated. For two dimensional
elements, with constant approximation of the harmonic function f (x) over the element, it
is possible to obtain analytical solution of the elemental integral, but due to its complexity,
three dimensional and axisymmetric cases rely on numerical integration methods in order to
obtain these coefficients.
It is possible to identify two type of elements, singular and non-singular ones. In case of
singular elements, the singular point is located within the contour, therefore, the integral
path. For non-singular elements, the singular point is located within another element, therefore, outside of the integral path. For both singular and non-singular elements, the Gauss
quadrature integration method can be used where the coefficients of the method are different
for singular and non-singular integration.

107

3.0.9

The Composite Regions

Often, physical problems present the challenge to solve Laplaces equation over composite
regions, where two subregions have different physical properties. To exemplify an application
of such a case, figure 3.3 is a schematic of the composite geometry for the Taylor cone jet
problem, studied under the current work. There, two distinct subregions in space, 1 and
S S S
S
2 , are delimited by two distinct boundaries, l = {1 2 3 I } and o = {4 I }
T
and a common boundary interface, I = {l o }. At each subregion, Laplaces equation
should be satisfied with its respective boundaries conditions and interface condition.

Figure 3.3: Schematic drawing of two subregions separated by a common interface.


A common practice is to reorder the algebraic system of equations (eq 3.17) for each subregion
in order to identify and group the interface function and its normal gradient as:

1 :

2 :

h
i u(1)
(1)
H (1) HI
(1)
uI

h
i u(2)
(2)
(2)
H HI
(2)
uI

h
i q (1)

(1) (1)
= G GI
(1)
qI

h
i q (2)

(2) (2)
= G GI
(2)
qI

108

(3.30)

(3.31)

where the superscript denotes the subregion and subscript, I, the interface variables. These
two algebraic systems can now be combined by use of the interface conditions. For the case
where the value of the functions and its gradient should be continuous, these conditions can
be expressed as:
(1)

(1)

(3.32)

uI = uI = uI
(1)

(2)

(3.33)

qI = qI = qI

where the negative sign was used in the normal gradient condition to satisfy the geometric
definition for the interface
(3.34)

n1I = n2I

Using the interface conditions, eq 3.32 and eq 3.33, a unique global banded algebraic linear
system of equations can be generated as:

(1)

u
(1)
(1)
HI
0
H

uI
(2)
(2)

0
H
HI
u(2)

(1)

(1)
G(1) GI
0
=

qI

(2)
(2)

0
G
GI
q (2)

Since for the interface, both, the harmonic function uI and its gradient qI are unknown, the
global system of equation for the composite problem can be reordered and written as:

(1)
HI

(1)
H

0
H (2)

(1)
GI

G(2)

H (2)

(1)

uI

q
I

u(2)

(1)
(1)

0 q
G
=

0
G(2)
q (2)

The general BEM for composite regions previously shown can be custom built to address
109

specific needs depending on the problem under investigation. For the problem under analysis
in this work, a jump condition for the flux across the interface will be embedded into the
equations before assembling the global system.
Consider the two subregions drawn in figure 3.3, where Laplaces equation is satisfied by the
harmonic function i ,

1 :

2 (1) = 0

2 :

2 (2) = 0

with the following boundary conditions:

1 :
2 :
3 :
4 :
I :

(1) = (2)

(1) = T (xinlet )
(1) n = 0
(1) = T (xout )
(2) = T (xouter )

and k2 (2) n(2) k1 (1) n(1) =

where ki is the material property of the subregion (a), i.e. the coefficient of diffusivity
for a heat transfer problems, or, electric permittivity of the medium, for electric potential
problems; T is a known boundary value for (a) (i.e. Taylor potential) and , the jump in
the normal flux condition (i.e. electric charge density).
It is convenient to define the boundary normal flux variable q (a) ki (a) n(a) and rewrite
the equations in terms of it. This procedure simplifies the algorithm to build the algebraic
system of equations and the application of the interface boundary conditions. Using the
discretized integral-representation of the harmonic function in the BEM equation (eq 3.14)
for the above problem setup, and placing the unknowns variables on the LHS and the known

110

on the RHS, the discrete BEM equation for the two subregions can be written as:

1 :

Gij

X
1 (1) X
1 (1)
(1)
qj +
Hij j
Gij qj +
k1
k1
2
3
X
X
1 (1)
(1)
+
Hij j
Gij qj =
k1
I
I
X
X
1 (1) X
(1)
(1)
=
Hij j +
Gij qj
Hij j
k1

(3.35)

X
X
1 (2) X
1 (2)
(2)
(2)
qj +
Hij j
Gij qj =
Hij j
k2
k
2
I
I

(3.36)

2 :

Gij

where

Hij

ij
H

ij +
H

i 6= j
1
2

i=j

ij and Gij being the elemental Greens function integration, defined at eq 3.15 and
with H
3.16 respectively.
(1)

(2)

(1)

(2)

(2)

(1)

Using the boundary conditions at the interface (nI = nj , j = j and qj = qj + ),


equation 3.36 is rewritten as

2 :

X
4

Gij

X
X
1 (2) X
1 (1) X
1
(1)
(2)
qj
Hij j +
Gij qj +
Gij j =
Hij j
(3.37)
k2
k2
k2
I
I
I

With the introduction of jump condition for the normal flux at the interface, an additional
variable is introduced into the system. If is prescribed as a boundary condition, this term
is a known variable and placed on the RHS of the equation. In the case of a unknown jump
conditions, the term is kept on the LHS and to solve the system, a additional equation
needs to be provided in order to close it. i.e. for the present development, this equation is
111

the conservation equation for the free surface charge, written as


dIs
02 1
= 2Rs KEni (1 + Rs ) 2
dx

(3.38)

where IS = 2Rs Us .
Using Finite Difference method it is possible to rewrite the charge conservation equation (eq
3.38) into a discrete form as
(1)

(3.39)

Mij j + Nij qj = 0

The global algebraic system of linear equation can be assembled based on equations 3.35,
3.37 and 3.39, written as:
(3.40)

Ay = F
where,

k11 G1

H2

k11 G3

(1)
HI

(1)
k11 GI

(2)
+ k12 GI

k12 GI

(2)
+ k12 GI

(2)
HI

T


y

H1

F =
0

1
G
k1 2

H3

H4

112

q 2

(3.41)

(3.42)

q1 2 q3 I qI q4 I

(3.43)

3.1

The Boundary Element Mesh

The computational domain of the BEM is defined in spatial boundary surface bounding the
3D space under analysis. Further simplification is achieved with the use of the axisymmetry
properties of the problem. Consequently, the 3D surface is simplified to a contour line in the
axisymmetric plane.
In order to test the robustness, effectiveness, accuracy and mesh Independence of the BEM
algorithm, two types of boundary meshes were designed and compared. The first boundary
mesh topology, labeled as mesh I, was designed based on the schematic presented in figure
3.3, where the inlet path, 1 , is a straight line from the inlet r zero coordinatexIN = {x1 , 0}
to the inlet interface coordinate, x1 = {x1 , Rs (x1 )}; the outer path, 2 , a circular arch centered at origin, xc = {0, 0}, starting from the inlet interface coordinate, x1 , ending at the
outlet interface coordinate, x2 = {x2 , Rs (x2 )} and the outlet path, a straight line starting
from the outlet interface coordinate, x2 , to the outlet r zero coordinate, xOU T = {x2 , 0}.
Figure 3.4 shows the electric potential iso-contour obtained from the BEM algorithm with
this mesh. Is expected that the discontinuity (corner) of the surface normal direction at
coordinate x1 present in this mesh reduces the accuracy of the results. This boundary error
associated with mesh corners is a well known issue in the field of numerical methods. This
type of error is associated with the limitation of the approximations used in t hemethod to
handle discontinuity. The standard procedure to remedy this issue and improve accuracy
near the corner is either to refine the mesh near the discontinuity and/or use higher order
approximations in the derivation of the numerical method. In general, higher order approximations type of solution is only necessary if the mesh refinement does not solve the problem.
This generally accepted procedure is justified with the fact that higher order approximation
increases the computational cost at a rate higher than the mesh refinement technique.
In order to assure that this type of discontinuity isnt affecting the accuracy of the results, a
113

Figure 3.4: BEM mesh I. The arrows represent the surface normal direction.
second type of mesh topology was employed, here labeled as mesh II. Comparison has shown
that the mesh II topology has a smoother boundary, since substituting the inlet and outlet
straight lines and the outer circular arch by a single circular arch extending from 0 to
centered at xc = {xc , rc } with radii Rc , where

xc =

x21 x22 + Rs (x1 )2 Rs (x2 )2


2 (x1 x2 )
rc = 0

Rc =

(x1 xc )2 + (Rs (x1 ) rc )2

(3.44)
(3.45)
(3.46)

Figure 3.4 shows this new boundary mesh topology represented by the circular dashed black
line on top of mesh I. Note that mesh I has two types of normal surface direction discontinuities in both ends: the discontinuity from the liquid inlet/outlet to the vacuum outer
boundary, and from the interface to the inlet/outlet/outer boundary. The mesh II topology addresses the problem of the corner at the interface liquid in both ends. Additionally,
114

improvment of the accuracy of the results is obtained by refinement near the interface inlet
and outlet tip.
The mesh II topology is refined near the region where the interface profile meets the inlet/outer (x1 ) and outlet/outer (x2 ) boundaries. This refinement is obtained from equation

= i +

(3.47)

where is the clustering function suggested by Anderson (1995, pg 190), and is given by

0
(sinh (( 0 ) ) + A)
A

(3.48)

where

A = sinh (0 )


1
1 + (exp () 1) 0
log
0 =
2
(exp () 1) 0

(3.49)
(3.50)

In the clustering function 3.48, is the constant that controls the degree of the clustering and
0 is the location on the dimensionless mapping (0 < < 1) where the maximum clustering
is to occur.

115

Figure 3.5: New mesh topology with boundary refinement.

Figure 3.6: Inlet and Outlet zoom in for mesh topology I and II.

116

The iso-contour field plot of the electric potential for the new mesh topology is shown in
figure 3.5. Qualitative inspection of the isocontours from the two plots does not present
considerable differences, except for the squeeze effect of the liquid isopotential lines near
the inlet boundary. This is expected, since the inlet boundary condition for both meshes
are the same, 1 : (1) = 0, but have different shapes. Note that, although different in
the beginning, after some axial distance, the equipotential lines relax and acquire similar
shapes in both results. This squeeze effect of the isocontour lines near the inlet region is a
consequence of the use of the Taylor Potential boundary condition at that particular interface.
This unwanted behavior can not be avoided, unless, some other boundary condition other
than Taylor Potential boundary condition is to be used. To reduce the effect of the errors
from the boundary on region of interest, the inlet boundary is placed far away from the
transition region. The isopotential lines of solution are in agreement with the assumption
that far upstream, the leading order of the electric potential is the Taylor potential, where
the liquid region is characterized by a sink of charges on the cone apex, resulting in a circular
isopotential (Higuera (2003), pg 311). Under this criterion, mesh II has a better performance
with the far upstream expected behavior.
To inspect the effect on the accuracy of the results, a zoom in plot of the inlet and outlet
regions is presented in figure 3.6.
From this qualitative comparison it is evident that both results are similar. It is convenient
to notice that some spurious behavior of the equipotential lines near the corner is associated
with the computational domain mesh used to plot the results. As detailed before, the BEM
only uses boundary values to assemble the BEM algebraic system of equations, and, its solution will provide only boundary information. After the solution procedure, a unstructured
triangular mesh (UTM) is used to plot the field electric potential based on the boundary
values using the integral representation of the boundary expressed in equation 3.8. Figure
3.7 shows samples of two UTMs used in the ISO-contour plots. This UTM is often used in
117

Finite Element (FE) simulations, and the first order shape function represents its scalar values within the triangle as a linear variation. This characteristic of the linear shape function
is what causes some spurious behavior on the isocontour plots near the corner. The UTM
shape and characteristics does not exert any influence on the BEM results, only on its field
isocontour plot.

Figure 3.7: Unstructured triangular mesh sample used to create the iso-potential plots.

Since inlet, outer and outlet boundaries geometry differ in the two mesh topologies, direct
comparison of its boundary values resulting from the BEM solution cannot be possible.
In order to quantify the difference in the results obtained from the two meshes topology,
interface data are the only data available, since the interface in both topologies was kept the
same. Figure 3.8 compares the interface electric potential and the liquid normal gradient
of the electric potential for both mesh topologies. One conclusion that can be drawn from
this result is that the modification in the mesh topology, including its refinement, does not
significantly change the outcome result of the BEM calculations. This shows the robustness
118

quality of the BEM method, its effectiveness to handle the boundary corners. On the other
hand, the oscillation observed in the gradient of the electric potential and the surface charge
(see figure 3.9) near the inlet boundary, indicates that refinement of the interface boundary
mesh might be necessary to improve the accuracy of the results. It is worth noticing that
this oscillatory behavior resembles that obtained from stiff problems when an inappropriate
numerical method is used in the solution process. Indeed, these stiffness characteristics at the
boundary might occur due to the electric potential boundary condition being incompatible
with charge boundary condition. In other words, the computational domain selected is too
short to have Taylor potential solution at its boundaries, since Taylor potential is a far field
boundary condition. This problem can be reduced by extending the computational domain
far up and down stream to the transition region.
Interface Boundary Mesh I and II Data Comparizon
0.15

mesh I
mesh II
d/dnmesh I
d/dnmesh I

-0.5

0.1
0.05

-1

0
Electric Potential

-1.5
-0.05
-2
-0.1
-2.5
-0.15
-3
-0.2
-3.5

-0.25

-4

Normal Gradient of Electric Potential d/dn

-0.3

-4.5

-0.35
-6

-4

-2

10

Figure 3.8: Fluid electric potential and its normal gradient along the interface boundary for
both mesh topologies.

119

Interface Boundary Mesh I and II Data Comparizon


0.8

qmesh I
qmesh II

0.7

Surface Charge (q)

0.6

0.5

0.4

0.3

0.2
-6

-4

-2

10

Figure 3.9: Electric charge along the interface boundary for both mesh topologies.

120

Chapter 4
The Finite Volume Method

4.1

Mathematical Formulation

The problem under the scope of this work regards fluid matter as an incompressible fluid.
The balance of linear momentum in a infinitesimal volume of fluid results in the well-known
Navier-Stokes (NS) equation,

1
(uR )
+ u u = p +
t

(4.1)

and mass balance in cylindrical coordinate (u = ur er + u e + ux ex ) is given by,

u=

1 (Rur ) 1 u ux
+
+
=0
r r
r
x

(4.2)

The stress tensor is defined as


ij = 2eij

121

(4.3)

with its components given by


1
eij = (u + uT )
2
ur
r

(4.5)

1 u ur
+
r
r

(4.6)

ux
x

(4.7)

err =

e =

exx =

er = er

1
=
2

ex = ex

 u  1 ur
r
+
r r
r

1
=
2

exr = erx

(4.4)

1
=
2

1 ux u
+
r
x

ur ux
+
x
r

(4.8)

(4.9)

(4.10)

Further axisymmetric simplification allows us to write the two components of momentum


balance equation as

ux
1 p
+ (u ) ux =
+ 2 ux
t
x

(4.11)


ur
1 p
ur 
+ (u ) ur =
+ 2 ur 2
t
r
r

(4.12)

122

where
1
=
r
2




1 2
2
r
+ 2 2+ 2
r
r
x

(4.13)

This two equations for momentum in each direction (eq. 4.12 and eq. 4.11), together with
continuity equation (eq. 4.2), result in a close system of three equations and three unknowns,
uR , ux and p.
For the development of the numerical scheme, it is convenient to write the momentum
equation in a conservative form as

() +
(ux ) +
(rur ) =
t
x
r r
x





1

+
r
+ P + S
x
r r
r

for = {ux , ur },
ux = ur =

P ux =

p
,
x

S ux = 0,

P ur =

S ur =

p
r

ur
R2

The conservative form of the momentum equation, cast in a general convection-diffusion


transport equation, will be discretized in the physical domain in a number of finite volumes,
culminating in an algebraic linear system of equations.

123

4.2

The Finite Volume Discretization

The Finite Volume (FV) method is based upon the volumetric integral of the mass and
balance equations, respectively given as,

(4.14)

v ndS = 0
S

ui v ndS =

ui d +

ij ij ndS

pii ndS

(4.15)

From Gauss theorem, the convective and diffusive terms can be converted from a volumetric
integral to a surface integral over control surfaces. The temporal and source terms can not
be converted to the surface integral, therefore, remain as volumetric integrals.
As previously mentioned, it is convenient to cast this balance equations in a form of a general
transport balance equation, written as,

nds +

v ndS =
s

q d V

(4.16)

The number of CVs necessary to obtain the solution of a given problem is determined by
the level of accuracy required for the solution, depending on the computational resources
available.
When solving irregular boundary problems it is convenient to use a body-fitted mesh. In
this type of mesh, the CV adjacent to the boundary is aligned with the boundary shape, as
presented in figure 4.1.
One of the distinguishing features of this type of mesh is the high level of accuracy in the
124

Figure 4.1: Body fitted non-orthogonal grid of CV.


results, since it has the capability to precisely fit the irregular boundary.
Each CV satisfies the conservation equation (eq. 4.16) and since each CV is interconnected,
a system of equations is obtained. The conservation equation also should be satisfied in
the entire computational domain, which provides the means to define some of the external
boundary conditions. Figure 4.1 presents a generic internal CV as P , where W , E, N and
S 1 are the neighbors CV positioned at West, East, North and South, respectively. The faces
of CV P are labeled as w, e, n and s2 for west, east, north and south faces respectively.

4.2.1

Interpolation Schemes

To evaluate the conservation equation (eq. 4.16), the surface integrals over the CV is conveniently divided into 6 faces integral for the three-dimensional case, and 4 faces integral in
the axisymmetric and two dimensional cases:

f ds =
s
1
2

X
k

fk ds

k = e, w, n, s, ...

Capital letters is used to label the center of the CV


Lower case letters label the center of the faces

125

(4.17)

where fk represents a given function of space, as the convective, v n, or diffusive,


n, normal flux with being the diffusion coefficient. Assuming that velocity component
and fluid property is known, the unknown variable needs to be define, and for that, an
approximation will be introduced.
Assuming that the CV is sufficient small to have a small variation of along each face, the
midpoint integral rule can be used. Therefore, the integral over the face is approximated as
a product of the integrand at the cell-face center (the mean value) and the cell-face area:

fk ds = f k Sk

(4.18)

In the case where the values of k is known along the face k, the mean value, f k , is defined
and the integral approximation is of second-order (Ferziger & Peric (1997) pg 74). When k
is the unknown, an interpolation scheme must be employed based on values of P and K
from the center of the CVs with coincident face.
Different interpolation schemes can be found in the literature. Out of these different schemes,
two have been widely used by different researchers, the upwind interpolation (upwind differencing scheme also known as UDS) and the linear interpolation (central differencing scheme
also known as CDS). The UDS approximate k with the value of at the center node
upstream to the flow direction, for instance, in the case of east face, e is defined as:

e =

P , if

(v n)e > 0

E , if

(v n)e < 0

(4.19)

On other hand, the CDS interpolation scheme is obtained by linear interpolation of the value
k at the center of the face, using the nearest center nodes values. For the east face, this is

126

translated into the following expression:

e = E e + P (1 e )

(4.20)

where e is the geometric factor for the linear interpolation, given by:

e =

|xe xP |
|xe xP | + |xe xE |

(4.21)

where xe is positioning vector of the center of the east face, and xE and xP the location of
the CV center of the East volume, E, and the center volume, P .
One characteristic of the UDS interpolation scheme is to be unconditionally stable when
solving the linear system of equation by using iterative methods. The unconditionally numerical stability characteristic of the UDS scheme is achieved by dumping the oscillations
with the introduction of numerical diffusion, consequently reducing the accuracy of the results. Therefore, the UDS interpolation accuracy is highly dependable on mesh size.
With the use of Taylor series expansion, it can be proved that the CDS interpolation scheme
is of second order accuracy. Consequently, oscillatory behavior might not be dumped under
certain conditions. Consequently, the CDS scheme is regarded as a conditionally stable
scheme.
The CDS interpolation is also used to evaluate the partial derivatives of the field function
at the faces of the CV. i.e. the gradient on x direction, needed to calculate the integral of
the diffusive flux, can be approximated as:


E P
xE xP

(4.22)

The Taylor series expansion also shows that this type of approximation for the gradient is
127

second-order accurate for uniform grid points where the east face is located midway between
P and E CV centers. For non-uniform grids, the accuracy of this approximation is formally
reduced to first-order accuracy, but Ferziger & Peric (1997) has observed that error reduction
behavior with grid refinement is similar to a second-order approximation (Ferziger & Peric
(1997) pg 78).
Using either UDS or CDS interpolation scheme to evaluate the convective flux at each face
of the CV, the surface integral can be written in its discrete counterpart as,

v ndS = AcE E + AcW W + AcN N + AcS S + AcP P

(4.23)

where the coefficients is defined on table 4.1, and m


k

(k = e, w, n, s) is the the mass flow

rate the crosses the respective face,

v nds

m
k=

(4.24)

and, Sk

(k = e, w, n, s), the surface area. Notice that for the two dimensional problems with

uniform grid spacing, the surface area is given by Se = Sw = yw and Sn = Ss = xw,


where w is the finite length in the direction perpendicular to the plane.
Table 4.1: Finite Volume Discretization Coefficient
U DS
CDS
AcE = min (m
e , 0)
AcE = m
e e
c
AW = min (m
w , 0) AcW = m
w w
AcN = min (m
n , 0) AcN = m
n n
c
c
AS = min (m
s , 0)
AS = m
s s
c
c
c
c
AP = (AE + AW + AN + AcS )
Using the CDS interpolation scheme, the diffusive flux can be written as:

nds = AdE E + AdW W + AdN N + AdS S + AdP P


s

128

(4.25)

where,
Se
;
xE xP
Se
AdE =
;
xE xP
AdE =

Sw
xP xW
Sw
=
xP xW

AdW =
AdW

AcP = (AcE + AcW + AcN + AcS )

The volume integrals terms of the conservation equation (eq. 4.16) can be approximated by
the product of the mean value of the integrand and the total volume of the CV,

q d = q P

QP =

(4.26)

where, q P is the mean value of the integrand q , geometric located at the center of the CV.
Since all the information for all variables is directly available at the center of the CV, no
interpolation is needed to calculate q P . Again, with the use of the Taylor series expansion,
it can be show that this type of approximation is of second-order accuracy. When q is
constant withing the CV, or have a linear variation, the volume integral is exact (Ferziger &
Peric (1997) pg 75).
In order to reduce the numerical oscillation often found in the convective terms, the first order
UDS scheme is blended with the second order CDS scheme. Although the UDS scheme is
less accurate, it always provides stable convergence towards the solution during the iterative
procedure. To improve accuracy, the UDS is gradually substituted by the CDS scheme in a
process named as deferred correction.

129

4.3

The Treatment of the Mesh Non-orthogonality

One important source of inaccuracy related with the use of body-fitted mesh employed in
this work is associated with the treatment of the non-orthogonality of the mesh, when it is
present. The non-orthogonality of the mesh can be measured by the angle between the two
lines of coordinates in the corners of each CV. A cartesian grid has /2 in all four corners of
the CV. Angles much lower or higher than /2 characterize a high level of non-orthogonality.
In most of the cases, it is possible to generate a body fitted mesh with orthogonal properties,
in which case, the accuracy of the method is not affected but the cost computational cost is
increased, due to the generation of the orthogonal mesh.
Ferziger & Peric (1997) has proposed a FVM where the physical CV is integrated in the
physical domain and the non-orthogonality of the mesh is treated by correcting the interpolated f at the face with the use of the gradient of, P , at the center of the volume
calculated from the previous iteration. This process allows the non-orthogonality to be implicitly corrected throughout the iterative process of the solution of the linear system of
equations.
Rhie & Chow (1983), Maliska (2004) and others authors have used a transformation of
coordinates from Cartesian to Curvilinear system to account for the non-orthogonality of the
mesh in the interpolation schemes. With these transformations, the non-orthogonal physical
domain (PHD) is transformed to a orthogonal Cartesian domain, aka computational domain
(CPD).
In the current work, we have implemented the Physical Domain Integration initially proposed
by Ferziger & Peric (1997).

130

4.3.1

The Domain Integration Method

Figure 4.2 shows a general schematic of a given internal CV and its neighbors in a collocated
arrangement. The staggered arrangement can also be used. The main difference between the
two grid arrangements is the interface velocities calculation. In the staggered arrangement,
the physical properties of the fluid at the face is readily available, whereas in the collocated
arrangement, interpolation between neighbors CV cell-center fluid properties is necessary to
obtain values at the faces. In the collocated arrangement, all velocities, pressure and any
scalar information is available at the center of the control volume. In this arrangement, the
CDS and UDS interpolation scheme can be used to obtain values at the face.

Figure 4.2: Non-orthogonal mesh diagram.


For the CV P , in the figure, the discrete form of the convective term can be written as,

ui v ndS m
s ui,s + m
n ui,n + m
w ui,w + m
e ui,e ;
S

131

i = x, y

(4.27)

where m
k is the mass flow rate that crosses the face k (eq 4.24) at a given time t. The
diffusive terms can be written as

ij ij ndS (ix Sx + iy Sy )s + (ix Sx + iy Sy )n + (ix Sx + iy Sy )w + (ix Sx + iy Sy )e (4.28)


S

where for collocated arrangement used, the viscous stress is given by,
uE uP
xE xP

(4.29)

une use
uE uP
+
xE xP
yne yse

(4.30)

pi ndS (pe Se pw Sw )

(4.31)

pj ndS (pn Sn ps Ss )

(4.32)

(xx )e 2
and
(yx )e
The pressure term is also discretized as

and

4.4

The Pressure-Velocity Coupling

The closed system of equations formed by the balance of momentum and mass over a CV
does not provide a evolution equation for p as it does for u and v. and since the fluid is
incompressible, there is no equation of state that could be used to evaluate p. This problem
was first addressed by Chorin (1967) that introduced the concept of artificial compressibility. With this method, a artificial equation of state is introduced in the iterative process

132

of solution for the linear system of equations. This compressibility property is gradually
removed the fluid as the pseudo temporal marching scheme advances in time. At the steadystate solution, the compressible effect is negligible and the result is accurate. This scheme is
only suitable for problems that only seek the final steady-state solution, where the transient
solution is not important, since it does not represent the accurate physical behavior.
Chorin (1968) also proposed a fractal step method (split method, projection method, etc)
in which the pressure field is decoupled from momentum equation in a two step iterative
process. i.e. For the unidimensional problem, in the solution procedure, the following partial
differential, given by
p

(u) +
= Fx u
t
x
is solved in two step,
1. For a given known velocity field, u, at time t, solve the equation

(u) = Fx u
t
2. Using the solution of the previous step, u , update the pressure field
pk+1 = p + D(u )

3. Update the velocity field u to


u = u t

pk+1
x

and return to step one until convergence.


D (u ) is the error associated with the divergence free velocity field (continuity equation) of
u . This is the reason this method is well known as Projection Method, since it projects
133

the non divergence-free intermediate velocity field into the divergence-free velocity field.
Notice that pressure in this method works as a Lagrange Multiplier and does not represent
the physical pressure. This is why, in the momentum equation, only pressure gradient is
physically relevant. Therefore, any pressure field p = pk+1 + pc is also a solution of the
problem, where pc is a arbitrary given constant.
Based on Chorins (Chorin (1968)) method, Patankar & Spalding (1972) proposed the widely
spread Semi Implicit Linked Equations (SIMPLE) procedure for the solution of incompressible flows.

4.4.1

The Semi Implicit Linked Equations (SIMPLE)

The solution of the algebraic system of equations previously derived from the momentum
equation produces updated velocity fields u and v that satisfy the momentum equation but
do not necessarily satisfy the continuity equation. In order to correct this field, pressure
is corrected to implicitly satisfy the continuity equation. This correction, also known as
divergence free projection, is then used to modify the velocity field. This procedure, repeated
iteratively, results in a velocity and pressure field that will satisfy momentum equation and
continuity simultaneously, within the defined accuracy range.
The algebraic equation for the pressure correction is derived from momentum and continuity
equations. Details of this derivation can be found in Ferziger & Peric (1997). The final form
of this algebraic equation is given by,
ApP p0P =

Apl p0l = m
p m
0p

134

(4.33)

where the coefficients are


ApE
ApN
ApP


=

S 2
Aup


S 2
=
Aup
 2 w
S
ApS =
Avp s

e
S 2
=
Avp n
X p
=
Al

ApW

,
,

(4.34)

l = E, W, N, S

With the use of the pressure correction p0 , the velocity correction u0i can be obtained. For
instance, for ue vertical velocity at face e, the correction is given by
u0e = uen1 ue

Si 0
0
ui (pE pP )
Ap

(4.35)

The final velocity field is updated with ui = ui + u0i and pressure by pn+1 = pn + p0 .

4.5

The Boundary Conditions

To accurately pose the numerical problem, boundary conditions need to be defined for all
control surfaces of the volumetric computational domain. In general, we can classify the
boundary conditions into two types, velocity and pressure boundary conditions. When velocity is specified, pressure is obtained by the solution procedure. When pressure is specified,
velocity is obtained from the numerical solutions. Both conditions can never be specified at
the same time, since it will result in an inconsistent problem.
From the velocity type of BC, a derived type is obtained for imposed mass flow rate BC. In
this type of boundary condition, the inside velocity is projected to the boundaries from the
previous iteration and a correction factor is calculated to assure that the imposed flow rate
is obtained. Using this method, the velocity profile at the boundary will be obtained from
135

the numerical solution.


Another type of boundary condition, also derived from velocity boundary conditions, is
designed to assure that a specific shear stress is imposed on a given surface. On those
boundaries, the tangential velocity is obtained from linear interpolation using the adjacent
inner volumes and satisfying the viscous shear stress condition. On these boundaries, velocity
normal to the surface is set to zero, since mass does not cross the these surfaces.
The simulations performed in this work use the imposed flow rate BC to calculate velocity
profile and pressure at the inlet and outlet boundary surfaces. The interface surface has zero
normal velocity, since mass does not cross the interface, and the shear stress is specified from
the BEM solution. Since velocity is imposed at the surface, pressure will be obtained from
the calculations. The normal forces balance equation is used to assure that the interface
profile is corrected, and if not, an iterative relaxation procedure is used.

4.6

The Linear System Solver

For a single CV, the discretized for of the conservation equation (eq 4.16) can be written as:

AE E + AW W + AN N + AS S + AP P = Qp

(4.36)

where the coefficient Ak (k = e, w, n, s), is the sum of the convective and diffusive coefficients,
Ak = Ack + Adk . The set of all CV equations assembles the system of linear equations for the
given problem, and it can be represented in a matrix for as:

A = F

136

(4.37)

where

A11 A12 A13

A21 A22 A23

A= .
..
..
..
.
.

An1 An2 An3

. . . A1n

. . . A2n

..
...
.

. . . Ann

(4.38)

is the coefficient matrix,


T


=

1 2 . . . n

(4.39)

the unknown vector, and


T


F =

Q1 Q2 . . . Qn

(4.40)

the forcing term.


The volumetric integral term is assumed to be a function of a known variable, therefore
treated as a source term and labeled as the forcing term F . Notice that for two dimensional
and axisymmetric problems, the discretized conservation equation (eq 4.36) have five coefficients only, therefore, the coefficient matrix for a structured mesh will be sparse with five
diagonals. Using a lexicographic ordering of CVs numbering, this system could be schematically represented by:

4.6.1

The Strong Implicit Procedure (SIP) Solver

To solve this sparse algebraic linear system of equations, direct method of solution such as
Gauss elimination is prohibitive in term of computational cost. The reason behind this is the
number of zero terms in the coefficient matrix that become non-zero during the elimination
process, demanding a storage capacity equivalent to a full matrix (N xN ) and its associated
computational calculations. To take advantage of the sparse property of the coefficient
matrix, a iterative method should be employed. Among the different methods suitable for
this kind of problem, the Strong Implicit Procedure (SIP) introduced by Stone (1968) is one
137

Figure 4.3: Five-diagonal sparse matrix of coefficients for two dimensional and axisymmetric
structured mesh
of the most efficient currently available.
The key idea of an iterative process is to find a approximate solution, n+1 , of the linear
system from a previous solution, n , and iterate until convergence is reached. This scheme
can be expressed as:
M n+1 = N n + B
where to satisfy the initial algebraic system of equations (eq 4.37) at the converged solution
(n+1 = n = ) the iteration matrix must be A = M N and B = F .
The SIP is part of the incomplete LU decomposition iterative family of methods, since it
approximates the iteration matrix M by a incomplete LU factorization (ILU), expressed
as:
M = LU = A + N
138

where the incomplete factorization scheme defines a lower diagonal matrix L to be sparse
and have a non-zero elements at the same location as the non-zero elements of the lower
diagonals of A. The same rule is applied to the upper diagonal U . Since M is the product
of L and U matrices, two additional diagonals will be generated. Defining N to store these
two extra diagonals, plus a combination of values of n such that N n 0. Details of the
derivation of this method for two dimensional and axisymmetric problems can be found at
Stone (1968) and Ferziger & Peric (1997), where the expressions to calculate the elements
of L and U can also be seen.
Once the ILU lower and upper diagonal matrices are defined, the iterative process successively
finds new approximation, n+1 , from one obtained from the previous iteration, n , using the
following steps.

1. Calculate the lower and upper diagonal matrices, L and U .


2. Calculate the residual of the initial guess, n = F A.
3. Solve the lower diagonal part of the ILU method, Rn = L1 n .
4. Solve the upper diagonal part of the ILU method, n+1 = U 1 Rn .
5. Obtain the new approximation, n+1 = n + n+1 .
6. If convergence is not achieved, set n = n+1 and return to step 2, otherwise, exit.

4.7

The Code Accuracy Check

In order to verify that the developed algorithm is accurate and bug free, we perform a
series of simulations for the lid driven cavity flow problem sketched in figure 4.4. This
problem has a number of documented numerical solutions available in the literature, making
139

Figure 4.4: Lid driven cavity flow problem sketch.


it suitable for our purposes. In our simulations, we test different interpolation schemes
and compare with solutions from Ghia et al. (1982). Figures 4.5,4.6 and 4.7 show u and v
velocity profiles in both x and y directions for Reynolds number Re = 100, Re = 400 and
Re = 1000 respectively. From the plots it is evident that better agreement with Ghia et al.
(1982) solutions is achieved with hybrid (UDS/CDS) and exponential interpolation schemes.
Further discussion about the properties of each interpolation scheme can be seen in Patankar
(1980).

4.8

The Taylor Cone Jet Profile

In order to define the geometric properties of the computational domain, a initial Taylor
cone jet interface should be known. We obtain this profile from numerical results of Higuera
(2003). Since we didnt have access to the original raw data of Higuera (2003), we scanned
the profiles from his paper (Higuera, 2003, pg 313). To scan and digitalize the curves from

140

Lid Driven Cavity Flow Re 100

-0.4
0.2

-0.2

u velocity
0.2
0.4

0.6

0.8

1
1

0.15
0.1

0.8

0.6

-0.05
-0.1

0.4

y coordinate

v velocity

0.05

-0.15
-0.2

0.2

-0.25
-0.3

0
0

0.2

0.4

0.6

0.8

x coordinate
Ghia et all 1982
Upwind

Hybrid
Power Law

Exponential

Figure 4.5: u and v velocity profiles along the cavity center line in x and y directions,
respectively. Calculations were performed for Re = 100. Upwind, hybrid, exponential and
power law interpolation scheme were tested.

141

Lid Driven Cavity Flow Re 400

-0.6
0.5

-0.4

-0.2

u velocity
0.2

0.4

0.6

0.8

1
1

0.4
0.3

0.8

v velocity

0.1

0.6

0
-0.1
0.4

-0.2

y coordinate

0.2

-0.3
0.2

-0.4
-0.5
-0.6

0
0

0.2

0.4

0.6

0.8

x coordinate
Ghia et all 1982
Upwind

Hybrid
Power Law

Exponential

Figure 4.6: u and v velocity profiles along the cavity center line in x and y directions,
respectively. Calculations were performed for Re = 400. Upwind, hybrid, exponential and
power law interpolation scheme were tested.

142

Lid Driven Cavity Flow Re 1000

-0.6
0.5

-0.4

-0.2

u velocity
0.2

0.4

0.6

0.8

1
1

0.4
0.3

0.8

v velocity

0.1

0.6

0
-0.1
0.4

-0.2

y coordinate

0.2

-0.3
0.2

-0.4
-0.5
-0.6

0
0

0.2

0.4

0.6

0.8

x coordinate
Ghia et all 1982
Upwind

Hybrid
Power Law

Exponential

Figure 4.7: u and v velocity profiles along the cavity center line in x and y directions,
respectively. Calculations were performed for Re = 1000. Upwind, hybrid, exponential and
power law interpolation scheme were tested.

143

Higueras paper, we used the digitizing software known as Engauge Digitizer3 . As expected,
the scanning process introduces error to the data. To reduce this noise, two techniques
were used, a two points moving average and a fitting curve. The noise is evident when
we plot the derivative of the interface, as it can be seen from the inset of figures 4.8 and
4.9 for two different flow rate conditions, QH = 0.27 and QH = 4.8 respectively. Since
the noise is more evident on the derivatives of the curves, both fitting curve and noise
reduction treatment is performed on the derivative of the curves. The interface profile is then
reconstructed from the first point boundary condition. The advantage of the arc-tan fitted
curve is the smoothness of the generated profile. A smooth profile is a desirable properties
since any wrinkles might introduce perturbation into the CFD calculations, resulting in a
slow convergence rate. Although desirable for the computational perspective, the arc-tan
profile does not fit the physics of the Taylor cone jet well, specially at the ends of the profile.
That conclusion is also evident from the spatial derivative of the profiles, where the arc-tan
appears to deviate slightly from the expected trend of the curve.

4.9

The Updating Profile Technique

The Taylor interface profiles obtained from Higueras calculations might be regarded as a
good starting point for the calculations. Indeed, as will be shown in the results, without
any changes the scanned profiles do provide a good approximation for the total current
emitted by a Taylor cone-jet. Nevertheless, an error is expected in the normal interface
force balance equation. Unfortunately, the code developed does not have the capability of
solve the unsteady moving interface problem, since it does not include in its formulation the
temporal terms or the spatial conservation law. Ferziger & Peric (1997, pg 379) has discussed
that if only the steady state solution is desirable, and implicit formulation is employed, the
spatial conservation law can be ignored, since the mesh will not be moving at the final
3

http://sourceforge.net/projects/digitizer/

144

Scanned data
Noise Reduction Profile
ArcTan Fitted Profile

4.5
0

-0.2
3.5
-0.4
RS(z)

RS(z)

-0.6

2.5

-0.8

-1
-1.2

1.5

-5

-4

-3

-2

-1

1
0.5
0
-5

-4

-3

-2

-1

0
z*H

Figure 4.8: Taylor interface profile scanned data from Higuera (2003, pg 313, figure 2a)
paper. This interface profile is the Higueras numerical solution for flow rate QH = 0.27.
The inset plot is the spatial derivative of the profile superimposed with the fitted arctan and
the noise reduced curves.

145

20

Scanned data
Noise Reduction Profile
ArcTan Fitted Profile

18
0

16

-0.2
14
-0.4
RS(z)

RS(z)

12

-0.6

10

-0.8

-1
-1.2
-20 -15 -10

-5

10

15

20

4
2
0
-20

-15

-10

-5

0
z*H

10

15

20

Figure 4.9: Taylor interface profile scanned data from Higuera (2003, pg 313, figure 2b)
paper. This interface profile is the Higueras numerical solution for flow rate QH = 4.8. The
inset plot is the spatial derivative of the profile superimposed with the fitted arctan and the
noise reduced curves.

146

iterative step. That is correct from theoretical considerations, but based on our experience,
the spatial conservation law does play a significant rule on the stability characteristics of the
iterative methods, even for the implicit formulations.
To overcome this limitation and still be able to update the interface with the method developed, the normal unbalanced error at the interface was successively reduced by changing
the height of the profile based on this errors. In this iterative procedure, an initial interface
profile guess is used in the CFD calculations. The pressure gradient resulting from the CFD
calculation is integrated along the interface to obtain pressure at the interface. This pressure
is introduced on the normal stress balance equation and a new profile is generated based on
the error of this equation.
(4.41)

RSn+1 (z) = RSn (z) + ERRn (z)


where
h

n
(2)
(1)
ERRn (z) = n ( n p + )nCF D M M

BEM

and the subscript

CF D

and

BEM

(4.42)

were used to indicate that calculation were performed with

CFD or BEM algorithm. is a relaxation parameter and restricted to be very small ( 106 )
in order to avoid unstable iterations.

147

Chapter 5
The Results and Analysis

5.1

Numerical Results

In order the check the accuracy of the combined BEM and CFD algorithms developed,
we replicate the simulations presented by Higuera (2003). Figures 5.1 and 5.2 show the
numerical results for QH = 0.27 and QH = 4.8 respectively. Both simulations are for = 5.0
and the interface profile obtained from Higueras paper fitted with the arc-tan procedure for
smoothness. For these calculations, we have not relaxed the profiles yet, since we want to
compare the solutions with Higueras data for the electric isopotential lines and streamlines.
The figures show the isopotential lines for the electrostatic field and the streamlines for the
hydrodynamics. The symbols in the figures represents the results obtained by Higueras
calculations, scanned from his paper (Higuera, 2003, pg 313, figures 2 (a) and (b)).
The consistent repeatability of this numerical experiment shows that both BEM and CFD
algorithm were properly coupled together. We assumed that Higueras profiles, obtained
from his paper, are the correct profiles, since they are the result of a converged numerical solution. But we also expected that due to the uncertainties introduced in the scanning/fitting
148

149

Figure 5.1: The counter plot of the electric potential superimposed with the fluid stream lines. The numerical results for
this plot were obtained with QH = 0.27 and = 5.0. The isopotential lines colored with black are the result of the present
calculations with Higuera (2003) isopotential lines superimposed with symbols for comparison purposes.

150

Figure 5.2: The counter plot of the electric potential superimposed with the fluid stream lines. The numerical results for this
plot were obtained with QH = 4.8 and = 5.0. The isopotential lines colored with black are result of the present calculations
with Higuera (2003) isopotential lines superimposed with symbols for comparison purposes.

Higueras data, it may result in a outstanding normal stress balance at the interface, despite
the agreement between our calculations and his data.
Higueras paper does not clarify the size of the computational domain. Based on our experience, the profile given in his paper might be small to avoid boundary errors propagate
into the region of interest. It is possible that he might have used a bigger domain but chose
only to plot the relevant part. The issue of a short computational domain is evident when
comparing the isopotential lines near the downstream boundary condition. When Higueras
profile is extended downstream, the agreement between his and our electric potential results
is further improved, as previously discussed in section 3.1 and shown in figure 3.5.
The downstream extended profiles were generated using the far downstream order of magnitude

Rs (z) = O

Q3/4
I 1/4 x1/8


,

provided from Gan-Calvo (1997a) analysis. For the upstream profile, the extended points
were obtained from a cubic polynomial using the left edge profile location and its derivative,
together with Taylor cone profile and its corresponding derivative far upstream. The cubic
curve was selected to produce a smooth profile from the transition region to the Taylor cone
region. The far upstream Taylor cone profile location and its derivative is also consistent
with the electric potential condition given by Taylor solution.
As previously discussed in the dimensional analysis, both profiles are similar when scaled
with a proper factor. For the purpose of comparing the two solutions for different flow rates,
the invariant system of dimensionless parameters is used to scale both solutions. Figure
5.3 compares the electric field obtained by the previously shown figures 5.1 and 5.2. The
characteristic parameters and conversion factors used are given in table 2.1. This figure
shows the similarity between outer/vacuum electric field of the different solutions, regardless
of the flow rate. That result is consistent with what has been observed by Gamero-Castao

151

(2010) from electric potential measurements using the RPA instrumentations. The small
difference can be also justified by the differences among the profiles, as previously discussed
in section 2.8.4. The inner liquid, electric field does have a shift proportional to the flow
rate, but regardless of the flow rate, it is negligible upstream to the transition region, due to
the shielding electric effect, as expected.
To further investigate the effects of increasing the flow rate in the problem, we plot a series
of electrostatic properties of the data obtained from the numerical calculations. Figure 5.4
(a) shows the electric potential on the left ordinate, and surface charge on the right ordinate.
Once again, the near invariance of the electric potential is in agreement with observations of
Gamero-Castao (2010), although minor changes might not be detectable by measurement
techniques. Surface charge profile begins from the same starting point, which is the charge
obtained from the Taylors solution. For the higher flow rate calculations, the charge profile
monotonically ascends up to a maximum value, from where it becomes apparently parallel
to the the lower flow rate charge profile. Since the overall surface charge increases with flow
rate, the liquid and vacuum normal electric field will also follow the same trend to satisfy the
electrostatic jump condition (equation 2.12), as shown in figure 5.4 (b). Interface tangent
electric field seems to have similar behavior as surface charge, showing that in fact, electric
potential might be regarded as near invariant, but not equal. Figure 5.4 (c) shows the normal
and tangential Maxwell stress tensor trend, behaving proportionally to the changes of the
electric charges and electric field. Figure 5.4 (d) shows the bulk and surface current charge
balance, where the bulk charges are conducted to the surface. This figure shows that the
charge balance mechanism model within the BEM is able to correctly capture the physics of
the problem. Higuera (2003, pg 313, figures 2 (a) and (b)) calculations have obtained a total

current of IH
= 1.1 for QH = 0.27 and IH
= 5.0 for QH = 4.8. When these values are scaled

to the dimensionless system used in the plot (Alternative Invariant), the Higueras current

reads IAI
= 1.96 and IAI
= 1.0 for flow rates of QH = 0.27 and QH = 4.8 respectively. This

shows an excellent agreement between our calculations and Higueras numerical solutions.
152

153

Figure 5.3: The counter plot of the electric potential for QH = 0.27 superimposed with results obtained for QH = 4.8. In the
figure, length and electric potential data from figures 5.1 and 5.2 were rescaled using the invariant dimensionless system where
characteristic parameters are presented in table 2.1.

-8

0.5

1.5

-4

||n.

-6

M.t||

||n.*M.n||

-6

-8

2.5

-3

-2

-1

-4

-2

-2

zI

(a)

0
z* I

(c)

0.2

0.4

0.6

0.8

*AI
8

E*I
I*
0

0.5

1.5

2.5

0.5

1.5

-8

-8

*(1)
n

I*b

I*s

-6

-6

E*(2)n

E*t

-4

-4

-2

-2

0
z* I

(d)

zI

(b)

Figure 5.4: The electrostatic properties of the interface along the zI axis ordinate for two different flow rates. The black lines,
marked with symbols, represent the solution for flow rate QH = 0.27 and the red lines, without symbols, for QH = 4.8. Chart
(a) is the electric potential (left abscissa) and surface charge (right abscissa). Chart (b) is the tangential and normal electric
field at the surface for the liquid (superscript (1) ) and vacuum (superscript (2) ). Chart (c) is the difference between liquid and
vacuum normal and tangential Maxwell stress tensor. Chart (d) is the bulk and surface current.

154

Radial velocity profile is shown in figure 5.5. From the plots it is clear that axial velocity
gradient in the radial direction is not zero in the upstream region. The radial velocity profile
in the radial direction shows a linear trend, except for the recirculation region that appears
at low flow rate. This provides us with good arguments to assume a constant radial velocity
gradient in the radial direction for moderate flow rate conditions (QH = O(1)). The axial
differences between the center velocity and the surface velocity is decreased as dimensionless
flow rate, QH , is increased. These observations might seem counterintuitive from a classical
hydrodynamics point of view, since one would expect low flow rate to be associated with a
higher velocity gradient. That happens because the velocity interface boundary condition is
controlled by Maxwell tangential stress, which increases with flow rate as well. Therefore,
increasing flow rate not only increases the overall inertia by increasing the core velocity, but
also increases the surface velocity by electrostatic acceleration, therefore, reducing the radial
gradient. Higueras dimensionless parameters contributes to this apparent confusion by not
having an invariant interface geometry, or a fixed dimensionless inlet condition, properties
attained in the Alternative Invariant system of parameters. In the AI system, the geometry
and inlet condition is fixed, resulting in a average velocity invariant with the flow rate.
In that case, the interface shear stress would be proportional to characteristic charge and
1/8

electrostatic field, and therefore proportional to QH

. This is consistent with the results,

since lower the flow rate, higher is the characteristic shear stress, resulting in the evident
recirculation zone.
The ultimate check to verify that Higueras profile used in the calculations is the correct
profile is given by the stress balance at the interface. Figure 5.6 shows the normal and tangential viscous forces together with Maxwell tangential forces. The tangential force balance
is also plotted in the inset graph, normalized by capillary forces. As expected, since the
velocity boundary condition is calculated from shear stress balance equation, the error is of
the order of the numerical precision of the computer, therefore, negligible.

155

r /R s

r /R s

0.5

1.5

v*/-u*

1
0.9
0.8
0.7
0.6
z*I=-5.20
0.5
z*I=-2.39
0.4
z*I=-0.21
0.3
z*I=2.91
0.2
z*I=7.13
0.1
0
-1.6 -1.4 -1.2 -1 -0.8 -0.6 -0.4 -0.2
0

1
0.9
0.8
0.7
0.6
z*I=-5.24
0.5
z*I=-2.40
0.4
z*I=-0.15
0.3
z*I=2.85
0.2
z*I=6.74
0.1
0
-1.6 -1.4 -1.2

-1

1.5

v*/-u*

-0.8 -0.6 -0.4 -0.2

(d)

0.5

(c)

-*
u /u

-0.5

z I=6.74

z I=2.85

z*I=-0.15

z I=-2.40

z I=-5.24

(b)

-*
u /u

0
0

0
-0.5

0.2

0.2

0.6

0.8

0.4

z I=7.13

z I=2.91

z I=-0.21

z I=-2.39

z I=-5.20

0.4

0.6

0.8

(a)

Figure 5.5: The velocity profile for two different flow rate conditions along the rz cross section planes located at five different
axial positions. Chart (a) represents the velocity in the axial direction for QH = 0.27 located at zI axial position indicated in
the legend. Chart (b) is the velocity in the axial direction for QH = 4.8. Chart (c) is for velocity in the radial direction for
QH = 0.27. Chart (c) is for velocity on radial direction for QH = 4.8.

156

157

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0.1

0.2

0.3

0.4

-5

-6e-15

-4e-15

-2e-15

2e-15

4e-15

6e-15

8e-15

1e-14

-5

-4

-4

-3

-2

-3

-1

-2

Capillary Normalized error

||t.*M.n||
*
||t. .n||
*
||n. .n||
balance error

-1

0
z*I

Interface Tangent Stress Balance

Figure 5.6: The viscous normal and tangential forces together with the Maxwell tangential force at the interface. The inset plot
is the difference normalized by capillary force.

Dimensionless Stress

Normal stress balance is shown in figure 5.7. From the plot, it is evident that the balance
error in the order of 20% is not a desirable solution, although, the electric problem is in
agreement with Higueras solution. That observation could lead to two possible conclusions,
either Higueras profile is not correct or the small changes in the profile necessary to satisfy
the normal stress balance would not change the outcome of the charge balance equation. In
either case, we acknowledge that at least some error could have been introduced from the
printing/scanning process of the paper. Another possibility to be considered is the fact that
Higueras numerical methods are different from ours.
Few differences between the method developed by Higuera (2003) and our model should be
highlighted. The first major difference to be pointed is the fact that we solve the problem with primitive variables, including pressure, and Higuera (2003) uses a stream-vorticity
formulation, eliminating pressure from being directly solved. Higuera (2003) uses finite difference discretization for both the hydrodynamic and electrostatic problems. We decouple
the two problems by solving the electrostatic field using the BEM. That allows us to properly
place the far field electric potential boundary condition (Taylor solution) without increasing the computational cost of the hydrodynamic problem. Higuera (2003) uses a curvilinear
transformation to solve the non-uniform geometric domain into a uniform grid computational
domain. We use a arbitrary volume integration method to discretize the hydrodynamic governing equation without any variable transformation. The non-orthogonality of the mesh is
taken into account by a iterative procedure as detailed in Ferziger & Peric (1997). Finally,
with our model, the flow and stress boundary conditions can be directly applied into the
computational algorithm, as opposed to Higuera (2003) method, where an ad hoc procedure
should be used to place the equivalent consistent vorticity and stream function boundary
conditions. We believe that this last difference between our method and Higuera (2003)
method contributing the most to the minor differences between the numerical results of
model and Higuera (2003) calculations.

158

159

-1

-5

-0.2

-0.1

0.1

0.2

-5

-4

-4

-3

-3

-2

-1

-2

Capillary Normalized Error

capillary
pressure
||n.*M.n||
*
||n. .n||
balance error

-1

0
z*I

Interface Normal Stress Balance

Figure 5.7: The normal viscous and Maxwell stress forces at the interface. Pressure and capillary is also shown in the plot.
Inset graph shows the balance error normalized by capillary forces.

Dimensionless Stress

5.2

Interface Updating Procedure

In order to reduce the normal stress balance error we have used the procedure discussed in
section 4.9. After an extensive trial and error attempt to define the relaxation coefficient ,
we found a good balance between stability and computational cost for 1e 6. Values
above that would result in a unstable system and the error wouldnt be further reduced.
Figure 5.8 shows the balance error profile and the average values at different iteration steps.

5.3

Taylor Cone-jet Emitted Current Scaling Law, Revisited

In the Taylor cone-jet classic literature we find opposing schools of thoughts regarding the
current scaling law. Apparently, under special circumstances, Gan-Calvo et al. (1997)
has proposed a I Q1/4 current law that contradicts the I Q1/2 previously proposed by
De La Mora & Loscertales (1994). These special circumstances might be related with the
rule of liquid viscosity, which has not been yet well understood. As a theoretical exercise,
we revisit this problem in a attempt to consolidate both theories. In order to do that, we
use the terminal jet diameter length scale proposed by Gamero-Castao (2010). Figure 5.9
shows dimensionless current versus Capillary number as defined in the Alternative Invariant
dimensionless system of parameters. The great feature of the AI Capillary number combined
with Higueras dimensionless number is the ability to fit all the data into an apparently unique
slope with different y intercept, including the I Q1/4 data given by Gan-Calvo et al.
(1997). At first, we speculated that the y intercept would be a function of the dielectric
constant, but after interpreting a few plots we realized that was not the case. Intuitively,
we thought that viscosity would play a role in the y intercept values of the curves. From
a linear regression of each single curve plotted on figure 5.9, we calculate the slope and y
160

161

Stress Balance Error

-5

-0.1

-0.05

0.05

0.1

-0.6

-0.4

-0.2

0.2

0.4

0.6

-4

-3

-2

(c)

0
z* I

5000
Iteration

-1

(a)

10000

0.1

0.15

0.2

0.25

-0.5

0.5

1.5

-5

-4

-3

-2

(d)

0
z* I

50000
Iteration

-1

(b)

100000

Figure 5.8: The normal stress balance error profile and average error for two iterative procedures. Graph (a) and (c) are for
= 1e 3 and (b) and (d) are for = 1e 6.

AVG Stress Balance error

intercept. The slope of the curves were very close to 1.0, as expected by a qualitative
inspection of the plot. In order to incorporate viscosity to our analysis we used Higueras
definition of Reynolds number. The reason for that is the fact that Higueras definition
of Reynolds number is only a function of the liquid properties, including viscosity. Figure
5.10 shows a singular dependence of the y intercept on the Reynolds number. The slope
of the curve can be approximated as 1.0, while the y0 1.532. From this y intercept (y0 ),
value calculated from the linear regression, it is possible to obtain a apparently universal
dependence of dimensionless current to the Capillary and Reynolds number, as shown in
figure 5.11.

162

163

IH

TEG2
BA

FM1
FM2
FM3
FM4

W3

W4

EG1

EG2

DOX50

DOX66

TEG1

EG4

1-8OL

EG3

0.1

W2

0.01

W1

0.1

10

DODCL

SUN

PROP

MTHL

DOX25

OCTL

1
CAAI

PC2

DOX2

HPT07

HPT19

DOX4

HPT45

10

TMP3

TMP2

TMP1

FORM2

PC4

PC3

100

TMP5

TMP4

1000

Figure 5.9: The dimensionless current versus Capillary number for a number of experimental data collected from literature.

164
0.01

0.1

De La Mora and Loscertales (1994)


GanCalvo (1997)

0.001
0.001

0.01

0.1

10

100

1000

1
ReH

100

GameroCastao (2010)
YINTCP=1.532*ReH;

10

1000

Figure 5.10: The y intercept (y0 ) of the curves shown on figure 5.9 as a function of Higueras definition of Reynolds number.

Y Intercept

165

I H/(1.532*ReH)

TEG2
BA

FM1
FM2
FM3
FM4

W3

W4

EG1

EG2

OCTL

DODCL

SUN

PROP

MTHL

DOX25

10
CAAI

PC2

DOX2

HPT07

HPT19

DOX4

HPT45

100

TMP3

TMP2

TMP1

FORM2

PC4

PC3

1000

TMP5

TMP4

Figure 5.11: The general dimensionless current dependence on the Capillary number.

DOX50

DOX66

TEG1

EG4

1-8OL

EG3

W2

0.1

W1

0.001
0.01

0.01

0.1

10

100

1000

10000

10000

Chapter 6
Conclusions
This dissertation has investigated the problem of the Taylor cone-jet using dimensional
analysis and numerical simulation tools. The numerical algorithm developed here has a
novel feature of combining different discretization techniques to obtain solutions for the
electrostatic and hydrodynamic set of governing equations and boundary conditions.
The Boundary Element Method was employed to solve the electrostatic equations inside and
outside the liquid. The two separated domains, the liquid and the vacuum, were coupled
together with the interface electric charge balance equation and the interface electric field
jump condition. A single set of algebraic linear system was generated from the two coupled
domains and the solution was obtained with the AMD version of LAPACK routines found
as part of the AMD Core Math Library (ACML).
The Finite Volume Method (FVM) was used to obtain the discrete counterpart of the hydrodynamic set of steady state governing equations. The shear viscous stress interface boundary conditions of the hydrodynamic problem were applied using electrostatic information
obtained from the BEM solution. The solution of the hydrodynamic problem was obtained
with the SIP Solver (Ferziger & Peric, 1997). The interface velocity solution from the FVM
166

was used to calculate the coefficient of the interface charge balance equation used by the BEM
to couple the liquid and vacuum domains. An iterative procedure was used to concurrently
solve the BEM and FVM algebraic systems in order to satisfy the interface charge balance
equation and viscous shear stress boundary condition simultaneously. The viscous normal
stress boundary conditions were not directly enforced in the FVM. Instead, the zero normal
interface velocity condition was imposed, which is associated with the steady state solution.
The FVM algorithm developed here did not include the transient terms of the momentum
balance equation or the Surface Conservation Law terms, therefore, the transient solution
of the moving interface problem could not be solved. For a given Taylor cone-jet interface
profile, the results of velocity, total current and electric potential from our calculations have
shown good agreement with data found in the literature for stream-vorticity formulation
(Higuera, 2003).
The normal stress interface balance error was used to update the interface geometry using
an iterative procedure. This procedure changes the rs location of the center of an arbitrary
interface element of area proportional to the normal stress balance error in that discrete
element. The procedure successfully resulted in reducing the error in most of the interface
cells, except for a few elements downstream to the transition region, near the exit boundary.
Different relaxation coefficients were tested in order to obtain a stable iterative solution.
In addition to the simulations, dimensional analysis was conducted to identify the key aspects
of the physics involved in the phenomenon. Four major forces can be identified in this
problem, capillary, inertia, viscous, and electrostatic. The Taylor cone-jet transition region
is defined as the region near the cone apex where the interface changes the profile from a
conical shape to a jet like shape. The length of the transition region is small compared with
the initial jet diameter, of the order of the injection needle inner diameter. Therefore, a fixed
length scale parameter could not be used in the process to obtain the non-dimensional set of
equations. This leaves room for different characteristic lengths to be proposed based on the
167

four types of forces existing in the jet. An extensive literature review of the application and
the proposed scaling laws was conducted. Two particular theories for the scaling laws were
investigated, the one proposed by De La Mora & Loscertales (1994) and the other proposed by
Gan-Calvo (1997b). Higuera (2003) proposed that the inertia, capillary and electrostatic
forces has to be of the same order in the transition region, recovering the characteristic
parameters previously proposed by Gan-Calvo (1997b). We propose a derivation of the
dimensionless set equation in a systematic procedure where no assumptions can be made
regarding the order of the different types of forces involved in the problem. We obtain a
set of equations with four dimensionless parameters controlling the problem, the Webber
number (W e), Reynolds Number (Re), Geoffrey number (Ge) and Leaky Dielectric (Le)
number. The Webber number is the ratio of inertia to capillary forces, Reynolds number
being the ratio of inertia to viscous forces, Geoffrey number being the ratio of inertia to
electrostatic forces, and the Leaky dielectric number is the ratio of bulk electric current to
surface current. From the hydrostatic solution provided by Taylor (1964), it can be shown
that W e = Ge is always true in the cone and the transition regions, which leaves us with only
three parameters to be defined, Ca = W e/Re, Re and Le. We were able to show that this
general non-dimensional derivation procedure allows us to obtain any set of characteristic
found in the literature. For example, Higueras characteristic parameters can be recovered
if we simply set W e = Re = Le = 1 in our model. Instead of using Higueras set of
parameters, we found it most convenient to define the transition length scale of the order of
the jet terminal diameter as proposed by Gamero-Castao (2010), since this scale is invariant
with respect to flow rate. Furthermore, we were able to show that this choice of invariant
length scale is equivalent to setting Le = W e.
Finally, we were able to show that with an appropriate choice between Higuera (2003) and
our set of dimensionless parameters, the I Q1/2 current law experimentally observed by
De La Mora & Loscertales (1994) and the I Q1/4 current law proposed by Gan-Calvo
(1997b) can be reconciled into a single I Q1/2 law where the viscous effects are already
168

embedded in the dimensionless current I and the dimensionless flow rate Q .

169

Bibliography
Abramowitz, Milton & Stegun, Irene A. 1965 Handbook of Mathematical Functions:
with Formulas, Graphs, and Mathematical Tables. Dover Publications.
Almera, Begoa, Deng, Weiwei, Fahmy, Tarek M. & Gomez, Alessandro 2010
Controlling the morphology of electrospray-generated PLGA microparticles for drug delivery. Journal of Colloid and Interface Science 343 (1), 125133.
Almera, Begoa, Fahmy, Tarek M. & Gomez, Alessandro 2011 A multiplexed
electrospray process for single-step synthesis of stabilized polymer particles for drug delivery. Journal of Controlled Release 154 (2), 203210.
Anderson, John David 1995 Computational fluid dynamics: the basics with applications.
McGraw-Hill.
Bailey, A.G. & Balachandran, W. 1981 The disruption of electrically charged jets of
viscous liquid. Journal of Electrostatics 10 (0), 99105.
Barrero, A., Gan-Calvo, A.M., Dvila, J., Palacios, A. & Gmez-Gonzlez,
E. 1999 The role of the electrical conductivity and viscosity on the motions inside taylor
cones. Journal of Electrostatics 47 (12), 1326.
Barrero, Antonio & Loscertales, Ignacio G. 2007 Micro- and nanoparticles via
capillary flows. Annual Review of Fluid Mechanics 39 (1), 89106.
Barrero, A., Lpez-Herrera, J.M., Boucard, A., Loscertales, I.G. & Mrquez,
M. 2004 Steady cone-jet electrosprays in liquid insulator baths. Journal of Colloid and
Interface Science 272 (1), 104108.
Bartoli, C., Rohden, H. von, Thompson, S. P. & Blommers, J. 1984 A liquid caesium field ion source for space propulsion. Journal of Physics D: Applied Physics 17 (12),
2473.
Bell, A. E., Schwind, G. A. & Swanson, L. W. 1982 The emission characteristics of
an aluminum liquid metal ion source. Journal of Applied Physics 53 (7), 46024605.
Bellman, R. & Pennington, R.H. 1954 Effects of surface tension and viscosity on taylor
instability. Quart. Appl. Math. 12, 151162.
170

Benassayag, G., Sudraud, P. & Jouffrey, B. 1985 In situ high voltage tem observation
of an electrohydrodynamic (EHD) ion source. Ultramicroscopy 16 (1), 18.
Bhattacharjee, P. K., Schneider, T. M., Brenner, M. P., McKinley, G. H. &
Rutledge, G. C. 2010 On the measured current in electrospinning. Journal of Applied
Physics 107 (4), 0443060443067.
Blandino, John, Shi, Xiaochuan, Rosenblad, Nathan, Flaherty, William &
Gamero-Castano, Manuel 2007 Charge detection mass spectrometer with integrated
retarding potential analyzer for study of colloid thruster plumes. American Institute of
Aeronautics and Astronautics.
Brebbia, C. A, Telles, J. C. F. (Jose Claudio Faria) & Wrobel, L. C. (Luiz C. )
1984 Boundary element techniques : theory and applications in engineering / C.A. Brebbia,
J.C.F. Telles, L.C. Wrobel . Berlin ; New York :: Springer-Verlag, includes bibliographical
references and index.
Aguirre-de Carcer, I. & de la Mora, J.Fernndez 1995 Effect of background gas on
the current emitted from taylor cones. Journal of Colloid and Interface Science 171 (2),
512517.
Carretero, Jorge A. 2005 Numerical simulation of a single emitter colloid thruster in
pure droplet cone-jet mode. Thesis, Massachusetts Institute of Technology, thesis (Ph.
D.)Massachusetts Institute of Technology, Dept. of Mechanical Engineering, 2005.
Carretero, Jorge A., Lozano, P. & Martinez-Sanchez, M. 2004 Numerical simulation of taylor cone-jets and colloid thruster plumes. , vol. 555, p. 106.
Carson, R. S., Hendricks, C. D., Hogan, J. & Schneider, J. M. 1964a Factors
influencing electrically sprayed liquids. AIAA Journal 2 (8), 14601461.
Carson, R. S., Hendricks, C. D., Hogan, J. & Schneider, J. M. 1964b Photomicrography of electrically sprayed heavy particles. AIAA Journal 2 (4), 733737.
Chang, Der-Yeou, Lee, Chia-Cheng & Shiea, Jentaie 2002 Detecting large
biomolecules from high-salt solutions by fused-droplet electrospray ionization mass spectrometry. Analytical Chemistry 74 (11), 24652469.
Chen, Da-Ren, Pui, David Y.H. & Kaufman, Stanley L. 1995 Electrospraying of
conducting liquids for monodisperse aerosol generation in the 4 nm to 1.8 m diameter
range. Journal of Aerosol Science 26 (6), 963977.
Chen, Da-Ren & Pui, David Y. H. 1997 Experimental investigation of scaling laws for
electrospraying: Dielectric constant effect. Aerosol Science and Technology 27 (3), 367
380.
Chen, Gung & Gomez, Alessandro 1992 Counterflow diffusion flames of quasimonodisperse electrostatic sprays. Symposium (International) on Combustion 24 (1),
15311539.
171

Chorin, Alexandre Joel 1967 A numerical method for solving incompressible viscous
flow problems. Journal of Computational Physics 2 (1), 1226.
Chorin, Alexandre Joel 1968 Numerical solution of the navier-stokes equations. Mathematics of Computation 22 (104), 745762.
Choueiri, E. Y. 2004 A critical history of electric propulsion: The first 50 years (19061956). Journal of Propulsion and Power 20 (2), 193203.
Clampitt, R., Aitken, K. L. & Jefferies, D. K. 1975 Abstract: Intense field-emission
ion source of liquid metals. Journal of Vacuum Science and Technology 12 (6), 12081208.
Cloupeau, M. 1994 Recipes for use of EHD spraying in cone-jet mode and notes on corona
discharge effects. Journal of Aerosol Science 25 (6), 11431157.
Cloupeau, M. & Prunet-Foch, B. 1989 Electrostatic spraying of liquids in cone-jet
mode. Journal of Electrostatics 22 (2), 135159.
Cloupeau, M. & Prunet-Foch, B. 1990 Electrostatic spraying of liquids: Main functioning modes. Journal of Electrostatics 25 (2), 165184.
Cloupeau, Michel & Prunet-Foch, Bernard 1994 Electrohydrodynamic spraying
functioning modes: a critical review. Journal of Aerosol Science 25 (6), 10211036.
Collins, Robert T., Harris, Michael T. & Basaran, Osman A. 2007 Breakup of
electrified jets. Journal of Fluid Mechanics 588 (-1), 75129.
Collins, Robert T., Jones, Jeremy J., Harris, Michael T. & Basaran, Osman A.
2008 Electrohydrodynamic tip streaming and emission of charged drops from liquid cones.
Nature Physics 4 (2), 149154.
Collins, Robert T., Sambath, Krishnaraj, Harris, Michael T. & Basaran, Osman A. 2013 Universal scaling laws for the disintegration of electrified drops. Proceedings
of the National Academy of Sciences 110 (13), 49054910, PMID: 23487744.
Cook, Kelsey D. 1986 Electrohydrodynamic mass spectrometry. Mass Spectrometry Reviews 5 (4), 467519.
Darrah, William A. 1934 Electrical spraying device. US Patent 1958406.
De La Mora, J. Fernndez 1992 The effect of charge emission from electrified liquid
cones. Journal of Fluid Mechanics 243, 561574.
De La Mora, J. Fernndez & Loscertales, I. G. 1994 The current emitted by highly
conducting taylor cones. Journal of Fluid Mechanics 260, 155184.
Deng, Weiwei, Klemic, James F., Li, Xiaohui, Reed, Mark A. & Gomez,
Alessandro 2007 Liquid fuel microcombustor using microfabricated multiplexed electrospray sources. Proceedings of the Combustion Institute 31 (2), 22392246.
172

Deng, Weiwei, Waits, C. Mike & Gomez, Alessandro 2010 Digital electrospray for
controlled deposition. Review of Scientific Instruments 81 (3), 035114.
Denizart, M., Roques, S., Sonier, F., Jouffrey, B. & Trinquier, J. 1981a Triode
and tetrode guns for field emission electron microscopes. Ultramicroscopy 7 (1), 6580.
Denizart, M., Roques, S., Sonier, F., Jouffrey, B. & Trinquier, J. 1981b Triode
and tetrode guns for field emission electron microscopes. Ultramicroscopy 7 (1), 107.
Dole, Malcolm, Mack, L. L., Hines, R. L., Mobley, R. C., Ferguson, L. D.
& Alice, M. B. 1968 Molecular beams of macroions. The Journal of Chemical Physics
49 (5), 22402249.
Doshi, Jayesh & Reneker, Darrell H. 1995 Electrospinning process and applications
of electrospun fibers. Journal of Electrostatics 35 (23), 151160.
Drozin, Vadim G 1955 The electrical dispersion of liquids as aerosols. Journal of Colloid
Science 10 (2), 158164.
Dyke, Milton D. Van 1975 Perturbation Methods in Fluid Mechanics, annotated ed edn.
Parabolic Pr.
Fenn, J. B., Mann, M., Meng, C. K., Wong, S. F. & Whitehouse, C. M. 1989
Electrospray ionization for mass spectrometry of large biomolecules. Science 246 (4926),
6471, PMID: 2675315.
Fenn, John B., Mann, Matthias, Meng, Chin Kai, Wong, Shek Fu & Whitehouse, Craige M. 1990 Electrospray ionizationprinciples and practice. Mass Spectrometry Reviews 9 (1), 3770.
Ferziger, Joel H. & Peric, Milovan 1997 Computational Methods for Fluid Dynamics,
corrected edn. Springer.
Feynman, Richard Phillips 1970 The Feynman Lectures on Physics (3 Volume Set).
Addison Wesley Longman.
Gamero-Castao, Manuel 2002 Electric-field-induced ion evaporation from dielectric
liquid. Physical Review Letters 89 (14), 147602.
Gamero-Castao, Manuel 2009 Retarding potential and induction charge detectors in
tandem for measuring the charge and mass of nanodroplets. Review of Scientific Instruments 80 (5), 0533010533014.
Gamero-Castao, M. 2010 Energy dissipation in electrosprays and the geometric scaling
of the transition region of cone-jets. Journal of Fluid Mechanics 662, 493513.
Gamero-Castao, M., Aguirre-de Carcer, I., de Juan, L. & Fernndez de la
Mora, J. 1998 On the current emitted by taylor cone-jets of electrolytes in vacuo: Implications for liquid metal ion sources. Journal of Applied Physics 83 (5), 24282434.
173

Gamero-Castao, M. & Hruby, V. 2001 Electrospray as a source of nanoparticles for


efficient colloid thrusters. Journal of Propulsion and Power 17 (5), 977987.
Gamero-Castao, Manuel & Hruby, Vladimir 2002 Electric measurements of charged
sprays emitted by cone-jets. Journal of Fluid Mechanics 459, 245276.
Gamero-Castao, Manuel & Mahadevan, Mahesh 2009 Sputtering yields of si, SiC,
and b[sub 4]C under nanodroplet bombardment at normal incidence. Journal of Applied
Physics 106 (5), 054305.
Gamero-Castao, M. & Fernndez de la Mora, J. 2000 Direct measurement of
ion evaporation kinetics from electrified liquid surfaces. The Journal of Chemical Physics
113 (2), 815832.
Garoz, D., Bueno, C., Larriba, C., Castro, S., Romero-Sanz, I., Fernandez
de la Mora, J., Yoshida, Y. & Saito, G. 2007 Taylor cones of ionic liquids from capillary tubes as sources of pure ions: The role of surface tension and electrical conductivity.
Journal of Applied Physics 102 (6), 06491306491310.
Gaskell, Simon J. 1997 Electrospray: Principles and practice. Journal of Mass Spectrometry 32 (7), 677688.
Gan-Calvo, A.M., Barrero, A. & Pantano-Rubio, C. 1993 04 o 01 the electrohydrodynamics of electrified conical menisci. Journal of Aerosol Science 24, Supplement
1, S19S20.
Gan-Calvo, A.M., Dvila, J. & Barrero, A. 1997 Current and droplet size in the
electrospraying of liquids. scaling laws. Journal of Aerosol Science 28 (2), 249275.
Gan-Calvo, A.M., Lasheras, J.C., Dvila, J. & Barrero, A. 1994 The electrostatic spray emitted from an electrified conical meniscus. Journal of Aerosol Science
25 (6), 11211142.
Gan-Calvo, Alfonso M. 1994 The size and charge of droplets in the electrospraying of
polar liquids in cone-jet mode, and the minimum droplet size. Journal of Aerosol Science
25, Supplement 1, 309310.
Gan-Calvo, Alfonso M. 1997a Cone-jet analytical extension of taylors electrostatic
solution and the asymptotic universal scaling laws in electrospraying. Physical Review
Letters 79 (2), 217220.
Gan-Calvo, Alfonso M. 1997b On the theory of electrohydrodynamically driven capillary jets. Journal of Fluid Mechanics 335, 165188.
Gan-Calvo, Alfonso M. 1998 The universal nature and scaling law of the surface
charge in electrospraying. Journal of Aerosol Science 29, Supplement 2, S975S976.
Gan-Calvo, Alfonso M 1999 THE SURFACE CHARGE IN ELECTROSPRAYING:
ITS NATURE AND ITS UNIVERSAL SCALING LAWS. Journal of Aerosol Science
30 (7), 863872.
174

Gan-Calvo, Alfonso M. 2004 On the general scaling theory for electrospraying. Journal of Fluid Mechanics 507, 203212a.
Gan-Calvo, Alfonso M. 2007 Electro-flow focusing: The high-conductivity lowviscosity limit. Physical Review Letters 98 (13), 134503.
Gan-Calvo, Alfonso M. & Montanero, Jos M. 2009 Revision of capillary cone-jet
physics: Electrospray and flow focusing. Physical Review E 79 (6), 066305.
Ghia, U, Ghia, K.N & Shin, C.T 1982 High-re solutions for incompressible flow using
the navier-stokes equations and a multigrid method. Journal of Computational Physics
48 (3), 387411.
Gignoux, D., Anton, H. F. & Shea, J. J. 1964 Development of a charged colloid source
for electrostatic propulsion. report no. 82. Tech. Rep. N65-15876; NASA-CR-54176. Cosmic, Inc., Washington, D.C.
Gilbert, William 1991 De Magnete. Dover Publications.
Gomer, R. 1979 On the mechanism of liquid metal electron and ion sources. Applied physics
19 (4), 365375.
Gomez, Alessandro & Tang, Keqi 1994 Charge and fission of droplets in electrostatic
sprays. Physics of Fluids 6 (1), 404414.
Greiner, Andreas & Wendorff, Joachim H. 2007 Electrospinning: A fascinating
method for the preparation of ultrathin fibers. Angewandte Chemie International Edition
46 (30), 56705703.
Gundabala, V. R., Vilanova, N. & Fernndez-Nieves, A. 2010 Current-voltage
characteristic of electrospray processes in microfluidics. Physical Review Letters 105 (15),
154503.
Halpern, Bret & Gomer, Robert 1969a Field emission in liquids. The Journal of
Chemical Physics 51 (3), 10311047.
Halpern, Bret & Gomer, Robert 1969b Field ionization in liquids. The Journal of
Chemical Physics 51 (3), 10481056.
Hartman, R.P.A., Brunner, D.J., Camelot, D.M.A., Marijnissen, J.C.M.
& Scarlett, B. 1999 ELECTROHYDRODYNAMIC ATOMIZATION IN THE
CONEJET MODE PHYSICAL MODELING OF THE LIQUID CONE AND JET. Journal of Aerosol Science 30 (7), 823849.
Hayati, I, Bailey, A & Tadros, Th.F 1987a Investigations into the mechanism of electrohydrodynamic spraying of liquids: II. mechanism of stable jet formation and electrical
forces acting on a liquid cone. Journal of Colloid and Interface Science 117 (1), 222230.

175

Hayati, I, Bailey, A.I & Tadros, Th.F 1987b Investigations into the mechanisms of
electrohydrodynamic spraying of liquids: I. effect of electric field and the environment on
pendant drops and factors affecting the formation of stable jets and atomization. Journal
of Colloid and Interface Science 117 (1), 205221.
Hayati, I., Bailey, A. I. & Tadros, Th F. 1986 Mechanism of stable jet formation in
electrohydrodynamic atomization. Nature 319 (6048), 4143.
Hendricks, C. D. & Swatik, D. S. 1968 Production of ions by electrohydrodynamic
spraying techniques. AIAA Journal 6 (8), 15961597.
Hendricks Jr., Charles D 1962 Charged droplet experiments. Journal of Colloid Science
17 (3), 249259.
Henriquez, P.C. 1935 Rec. trav. chim. 54, 574.
Henson, Bob L. 1964 Mobility of positive ions in liquefied argon and nitrogen. Physical
Review 135 (4A), A1002A1008.
Heyningen, Van 1983 Liquid jet method for coating photographic recording media. US
Patent US4381342 A.
Higuera, F. J. 2003 Flow rate and electric current emitted by a taylor cone. Journal of
Fluid Mechanics 484, 303327.
Higuera, F. J. & Barrero, A. 2005 Electrosprays of very polar liquids at low flow rates.
Physics of Fluids 17, 8104.
Hines, R. L. 1966 Electrostatic atomization and spray painting. Journal of Applied Physics
37 (7), 27302736.
Hohman, Moses M., Shin, Michael, Rutledge, Gregory & Brenner, Michael P.
2001a Electrospinning and electrically forced jets. i. stability theory. Physics of Fluids
13 (8), 22012220.
Hohman, Moses M., Shin, Michael, Rutledge, Gregory & Brenner, Michael P.
2001b Electrospinning and electrically forced jets. II. applications. Physics of Fluids 13 (8),
22212236.
Hunter, R. 1966 Negatively charged colloid generation research. In 5th Electric Propulsion
Conference. American Institute of Aeronautics and Astronautics.
Iribarne, J. V. & Thomson, B. A. 1976 On the evaporation of small ions from charged
droplets. The Journal of Chemical Physics 64 (6), 22872294.
Jahn, R. & Mickelsen, W. 1969 Status of electric propulsion. In 5th Propulsion Joint
Specialist. American Institute of Aeronautics and Astronautics.
Jones, A. R. & Thong, K. C. 1971 The production of charged monodisperse fuel droplets
by electrical dispersion. Journal of Physics D: Applied Physics 4 (8), 1159.
176

Kevorkian, J. K. & Cole, J. D. 1996 Multiple Scale and Singular Perturbation Methods.
Springer.
Khayms, Vadim 2000 Advanced propulsion for microsatellites. Thesis, Massachusetts Institute of Technology, thesis (Ph.D.)Massachusetts Institute of Technology, Dept. of Aeronautics and Astronautics, 2000.
Kim, Eun-Kyeong, Hwang, Taejin & Kim, Sang Sub 2011 Change in microstructure
and surface properties of electrospray-synthesized silica layers. Journal of Colloid and
Interface Science 364 (2), 561565.
Kim, Geunhyung, Park, Jongha & Han, Houkseop 2006 Production of microsized
PMMA droplets using electrospraying with various auxiliary fields. Journal of Colloid and
Interface Science 299 (2), 593598.
Kim, Ji-Yeong, Kim, Eun-Kyeong & Kim, Sang Sub 2013 Micro-nano hierarchical
superhydrophobic electrospray-synthesized silica layers. Journal of Colloid and Interface
Science 392, 376381.
Kim, Kyekyoon & Ryu, Choon K. 1994 Method and apparatus for producing nanodrops
and nanoparticles and thin film deposits therefrom. US Patent US5344676 A.
Kim, Kyekyoon & Turnbull, Robert J. 1976 Generation of charged drops of insulating
liquids by electrostatic spraying. Journal of Applied Physics 47 (5), 19641969.
Kocot, Ray H., Rudd, Robert E. & Arnoldi, Douglas R. 1983 Electrostatic ink jet
system. US Patent US4404573 A.
Komuro, Masanori, Hiroshima, Hiroshi, Tanoue, Hisao & Kanayama, Toshihiko
1983 Maskless etching of a nanometer structure by focused ion beams. Journal of Vacuum
Science & Technology B: Microelectronics and Nanometer Structures 1 (4), 985989.
Krohn, V. E. 1962 Research on the Generation and Acceleration of Submicron-size Particles. Defense Technical Information Center.
Krohn, V. E. & Ringo, G. R. 1975 Ion source of high brightness using liquid metal.
Applied Physics Letters 27 (9), 479481.
Krpoun, Renato & Shea, Herbert R. 2008 A method to determine the onset voltage
of single and arrays of electrospray emitters. Journal of Applied Physics 104 (6), 064511
0645118.
Kundu, Pijush K. & Cohen, Ira M. 2004 Fluid Mechanics. ELSEVIER Academic Press.
Kyritsis, Dimitrios C., Guerrero-Arias, Ismael, Roychoudhury, Subir &
Gomez, Alessandro 2002 Mesoscale power generation by a catalytic combustor using electrosprayed liquid hydrocarbons. Proceedings of the Combustion Institute 29 (1),
965972.
177

Laplace 1829 Mcanique cleste : Laplace, Pierre Simon, marquis de, 1749-1827 : Free
Download & Streaming : Internet Archive.
Larriba, Carlos & Fernandez de la Mora, Juan 2011 Production of monodisperse
submicron drops of dielectric liquids by charge-injection from highly conducting liquids.
Physics of Fluids 23 (10), 10200310200311.
Larriba-Andaluz, Carlos & Fernandez de la Mora, Juan 2010 Electrospraying
insulating liquids via charged nanodrop injection from the taylor cone of an ionic liquid.
Physics of Fluids 22 (7), 0720020720027.
Lawton, James & Weinberg, Felix Jiri 1969 Electrical aspects of combustion. Clarendon P.
Leonard, William K. 1996 Multiple layer coating method. US Patent US5525376 A.
Li, D. & Xia, Y. 2004 Electrospinning of nanofibers: Reinventing the wheel? Advanced
Materials 16 (14), 11511170.
Loo, Joseph A. 1997 Studying noncovalent protein complexes by electrospray ionization
mass spectrometry. Mass Spectrometry Reviews 16 (1), 123.
Loscertales, I.G., Cortijo-bon, R., Barrero, A., Guerrero, I. & Gan-Calvo,
A.M. 2001 A novel technique to produce multicomponent micro/nano capillary jets and
micro/nano capsules by electrohydrodynamic forces. Journal of Aerosol Science 32, Supplement 1, 611622.
Loscertales, I. G., Barrero, A., Guerrero, I., Cortijo, R., Marquez, M. &
Gan-Calvo, A. M. 2002 Micro/Nano encapsulation via electrified coaxial liquid jets.
Science 295 (5560), 16951698, PMID: 11872835.
Lozano, Paulo & Martnez-Snchez, Manuel 2004 Ionic liquid ion sources: suppression of electrochemical reactions using voltage alternation. Journal of Colloid and Interface
Science 280 (1), 149154.
Lozano, Paulo & Martnez-Snchez, Manuel 2005 Ionic liquid ion sources: characterization of externally wetted emitters. Journal of Colloid and Interface Science 282 (2),
415421.
Luo, C. J., Loh, Shirin, Stride, Eleanor & Edirisinghe, Mohan 2011 Electrospraying and electrospinning of chocolate suspensions. Food and Bioprocess Technology
.
Lpez-Herrera, J.M, Barrero, A, Lpez, A, Loscertales, I.G & Mrquez, M
2003 Coaxial jets generated from electrified taylor cones. scaling laws. Journal of Aerosol
Science 34 (5), 535552.
Macky, Wallace A. 1928 On quantitative measurements in frictional electricity. Proceedings of the Royal Society of London. Series A 119 (781), 107132.
178

Macky, W. A. 1931 Some investigations on the deformation and breaking of water drops
in strong electric fields. Proceedings of the Royal Society of London. Series A 133 (822),
565587.
Mahoney, J., Moore, R., Perel, J. & Yahiku, A. 1967 Research on a charged particle
bipolar thrustor. In 6th Electric Propulsion and Plasmadynamics Conference. American
Institute of Aeronautics and Astronautics.
Mahoney, J. & Perel, J. 1969 Focusing and deflection of heavy charged particle beams. In
7th Electric Propulsion Conference. American Institute of Aeronautics and Astronautics.
Mahoney, J. & Perel, J. 1976 Particle interactions in colloid thruster exhaust plumes.
In 12th International Electric Propulsion Conference. American Institute of Aeronautics
and Astronautics.
Mahoney, J. F., Moore, R. D., Perel, J. & Yahiku, A. Y. 1969a Research and
development of a charged-particle bipolar thruster. AIAA Journal 7 (3), 507511.
Mahoney, John F., Yahiku, Arthur Y., Daley, Howard L., Moore, R. David
& Perel, Julius 1969b Electrohydrodynamic ion source. Journal of Applied Physics
40 (13), 51015106.
Maliska, Clovis R. 2004 Transferncia de calor e mecnica dos fluidos computacional .
Livros Tecnicos e Cientificos.
Marche, P. H. La, Levi-Setti, R. & Wang, Y. L. 1983 Focused ion beam microlithography using an etch-stop process in gallium-doped silicon. Journal of Vacuum Science &
Technology B: Microelectronics and Nanometer Structures 1 (4), 10561058.
Marginean, Ioan, Parvin, Lida, Heffernan, Linda & Vertes, Akos 2004 Flexing
the electrified meniscus: The birth of a jet in electrosprays. Analytical Chemistry 76 (14),
42024207.
Martinez-Sanchez, Manuel & Akinwande, Akintunde I. 2005 Colloid thrusters,
physics, fabrication and performance. Tech. Rep..
Martn-Banderas, Luca, Flores-Mosquera, Mara, Riesco-Chueca, Pascual,
Rodrguez-Gil, Alfonso, Cebolla, ngel, Chvez, Sebastin & GanCalvo, Alfonso M. 2005 Flow focusing: A versatile technology to produce sizecontrolled and specific-morphology microparticles. Small 1 (7), 688692.
Martn-Banderas, L., Rodrguez-Gil, A., Cebolla, ., Chvez, S., Berdnlvarez, T., Fernandez Garcia, J. M., Flores-Mosquera, M. & Gan-Calvo,
A. M. 2006 Towards high-throughput production of uniformly encoded microparticles.
Advanced Materials 18 (5), 559564.
Masayuki, Sato 1984 The production of essentially uniform-sized liquid droplets in gaseous
or immiscible liquid media under applied a.c. potential. Journal of Electrostatics 15 (2),
237247.
179

Mathiowitz, Edith, Jacob, Jules S., Jong, Yong S., Carino, Gerardo P., Chickering, Donald E., Chaturvedi, Pravin, Santos, Camilla A., Vijayaraghavan,
Kavita, Montgomery, Sean, Bassett, Michael & Morrell, Craig 1997 Biologically erodable microspheres as potential oral drug delivery systems. Nature 386 (6623),
410414.
Matsumoto, Hidetoshi, Mizukoshi, Tomoya, Nitta, Kazuya, Minagawa, Mie,
Tanioka, Akihiko & Yamagata, Yutaka 2005 Organic/inorganic hybrid nanomicrostructured coatings on insulated substrates by electrospray deposition. Journal of
Colloid and Interface Science 286 (1), 414416.
Matsumoto, Hidetoshi, Wakamatsu, Yuji, Minagawa, Mie & Tanioka, Akihiko
2006 Preparation of ion-exchange fiber fabrics by electrospray deposition. Journal of Colloid and Interface Science 293 (1), 143150.
Matsumoto, Hidetoshi, Yako, Hiroshi, Minagawa, Mie & Tanioka, Akihiko 2007
Characterization of chitosan nanofiber fabric by electrospray deposition: Electrokinetic
and adsorption behavior. Journal of Colloid and Interface Science 310 (2), 678681.
Meesters, G.M.H., Marijnissen, J.C.M. & Scarlett, B. 1993 Response to remarks
by fernndez de la mora and gomez on our paper Generation of micron-sized droplets
from the taylor cone. Journal of Aerosol Science 24 (5), 697702.
Meesters, G.M.H., Vercoulen, P.H.W., Marijnissen, J.C.M. & Scarlett, B.
1990 A monodisperse-aerosol-generator, using the taylor cone, for the production of 1 m
droplets. Journal of Aerosol Science 21, Supplement 1, S669S672.
Meesters, G.M.H., Vercoulen, P.H.W., Marijnissen, J.C.M & Scarlett, B. 1991
A (sub-)micron aerosol generator,using a high electric field. Journal of Aerosol Science 22,
Supplement 1, S11S14.
Meesters, G.M.H., Vercoulen, P.H.W., Marijnissen, J.C.M. & Scarlett, B.
1992 Generation of micron-sized droplets from the taylor cone. Journal of Aerosol Science
23 (1), 3749.
Mejia, Andres F., He, Peng, Luo, Dawei, Marquez, Manuel & Cheng, Zhengdong 2009 Uniform discotic wax particles via electrospray emulsification. Journal of Colloid and Interface Science 334 (1), 2228.
Melcher, J R & Taylor, G I 1969 Electrohydrodynamics: A review of the role of
interfacial shear stresses. Annual Review of Fluid Mechanics 1 (1), 111146.
De la Mora, J. Fernandez De La, Navascues, J., Fernandez, F. & RosellLlompart, J. 1990 Generation of submicron monodisperse aerosols in electrosprays.
Journal of Aerosol Science 21, Supplement 1, S673S676.
Mller, Erwin W. & Tsong, Tien Tzou 1969 Field ion microscopy: principles and
applications. American Elsevier Pub. Co.
180

Onsager, Lars 1936 Electric moments of molecules in liquids. Journal of the American
Chemical Society 58 (8), 14861493.
Paine, Mark D., Alexander, Matthew S. & Stark, John P.W. 2007 Nozzle and
liquid effects on the spray modes in nanoelectrospray. Journal of Colloid and Interface
Science 305 (1), 111123.
Patankar, Suhas 1980 Numerical Heat Transfer and Fluid Flow , 1st edn. Taylor & Francis.
Patankar, S.V & Spalding, D.B 1972 A calculation procedure for heat, mass and momentum transfer in three-dimensional parabolic flows. International Journal of Heat and
Mass Transfer 15 (10), 17871806.
Pinna, H., Liang, K., Denizart, M. & Jouffrey, B. 1983 Electronic environment for a
field emission gun in electron microscopy. Revue de Physique Applique 18 (10), 659665.
Pozrikidis, C. 2002 A Practical Guide to Boundary Element Methods with the Software
Library BEMLIB , 1st edn. CRC Press.
Prewett, P. D. & Mair, G. L. R. 1991 Focused Ion Beams from Liquid Metal Ion
Sources, 1st edn. John Wiley & Sons.
Probst, Richard O. & Probst, Richard O. 1973 Electrostatic deposition apparatus.
US Patent 3776187.
Quang Tran Si, Bui, Byun, Doyoung & Lee, Sukhan 2007 Experimental and theoretical study of a cone-jet for an electrospray microthruster considering the interference
effect in an array of nozzles. Journal of Aerosol Science 38 (9), 924934.
Ramos, A. & Castellanos, A. 1994 Conical points in liquid-liquid interfaces subjected
to electric fields. Physics Letters A 184 (3), 268272.
Ransburg, Edwin M. 1959 Method and apparatus for electrostatically coating. US Patent
2893894.
Reneker, D.H., Yarin, A.L., Zussman, E. & Xu, H. 2007 Electrospinning of nanofibers
from polymer solutions and melts. In Advances in Applied Mechanics (ed. Hassan Aref &
Erik van der Giessen), , vol. Volume 41, pp. 43346. Elsevier.
Reneker, Darrell H., Yarin, Alexander L., Fong, Hao & Koombhongse,
Sureeporn 2000 Bending instability of electrically charged liquid jets of polymer solutions in electrospinning. Journal of Applied Physics 87 (9), 45314547.
Rhie, C. M. & Chow, W. L. 1983 Numerical study of the turbulent flow past an airfoil
with trailing edge separation. AIAA Journal 21 (11), 15251532.
Rietveld, Ivo B., Kobayashi, K., Yamada, H. & Matsushige, K. 2006 Morphology
control of poly(vinylidene fluoride) thin film made with electrospray. Journal of Colloid
and Interface Science 298 (2), 639651.
181

Rietveld, Ivo B., Kobayashi, Kei, Yamada, Hirofumi & Matsushige, Kazumi
2009 Process parameters for fast production of ultra-thin polymer film with electrospray
deposition under ambient conditions. Journal of Colloid and Interface Science 339 (2),
481488.
Roberts, William P., Morgan, Charles R. & Havens, Marvin R. 1992 Antistatically conductive masking film for electrostatic spray painting.
Romero-Sanz, Ignacio, Fernandez de la Mora, Juan & Aguire-de Carcer,
Inigo 2003 Ionic propulsion based on heated taylor cones of ionic liquids. American Institute of Aeronautics and Astronautics.
Rosell-Llompart, J. & Fernndez de la Mora, J. 1994 Generation of monodisperse
droplets 0.3 to 4 m in diameter from electrified cone-jets of highly conducting and viscous
liquids. Journal of Aerosol Science 25 (6), 10931119.
Rulison, Aaron J. & Flagan, Richard C. 1994 Electrospray atomization of electrolytic
solutions. Journal of Colloid and Interface Science 167 (1), 135145.
Ryan, C. N., Smith, K. L., Alexander, M. S. & Stark, J. P. W. 2009 Effect of
emitter geometry on flow rate sensitivity to voltage in cone jet mode electrospray. Journal
of Physics D: Applied Physics 42 (15), 155504.
Ryan, C. N., Smith, K. L. & Stark, J. P. W. 2012 The influence of geometry on the flow
rate sensitivity to applied voltage within cone-jet mode electrospray. Journal of Applied
Physics 112 (11), 1145101145107.
Sato, Masayuki, Saito, Masahiro & Hatori, Toshihiro 1993 Emulsification and
size control of insulating and/or viscous liquids in liquid-liquid systems by electrostatic
dispersion. Journal of Colloid and Interface Science 156 (2), 504507.
Saville, D. A. 1997 ELECTROHYDRODYNAMICS:The taylor-melcher leaky dielectric
model. Annual Review of Fluid Mechanics 29 (1), 2764.
Seaver, Albert E. & Eckhardt, Carey J. 1988 Electrospray coating process. US
Patent US4748043 A.
Seliger, R. L., Ward, J. W., Wang, V. & Kubena, R. L. 1979 A high-intensity
scanning ion probe with submicrometer spot size. Applied Physics Letters 34 (5), 310
312.
Shin, Y.M., Hohman, M.M., Brenner, M.P. & Rutledge, G.C. 2001 Experimental
characterization of electrospinning: the electrically forced jet and instabilities. Polymer
42 (25), 0995509967.
Siefert, W. 1984 Corona spray pyrolysis: A new coating technique with an extremely
enhanced deposition efficiency. Thin Solid Films 120 (4), 267274.

182

Smith, David P H 1986 The electrohydrodynamic atomization of liquids. IEEE Transactions on Industry Applications IA-22 (3), 527535.
Smith, Katharine L., Alexander, Matthew S. & Stark, John P. W. 2006a The
sensitivity of volumetric flow rate to applied voltage in cone-jet mode electrospray and the
influence of solution properties and emitter geometry. Physics of Fluids 18 (9), 092104
0921047.
Smith, K. L., Alexander, M. S. & Stark, J. P. W. 2006b Voltage effects on the
volumetric flow rate in cone-jet mode electrospraying. Journal of Applied Physics 99 (6),
0649090649098.
Smith, Richard D., Loo, Joseph A., Edmonds, Charles G., Barinaga, Charles J.
& Udseth, Harold R. 1990 New developments in biochemical mass spectrometry: electrospray ionization. Analytical Chemistry 62 (9), 882899.
Stone, Herbert L. 1968 Iterative solution of implicit approximations of multidimensional
partial differential equations. SIAM Journal on Numerical Analysis 5 (3), 530.
Suh, Jeongsoo, Han, Bangwoo, Okuyama, Kikuo & Choi, Mansoo 2005 Highly
charging of nanoparticles through electrospray of nanoparticle suspension. Journal of Colloid and Interface Science 287 (1), 135140.
Sun, Lei, Yu, Xiaofang, Sun, Mingda, Wang, Hengguo, Xu, Shufei, Dixon,
John D., Wang, Y. Andrew, Li, Yaoxian, Yang, Qingbiao & Xu, Xiaoyi 2011
Preparation of quantum dots encoded microspheres by electrospray for the detection of
biomolecules. Journal of Colloid and Interface Science 358 (1), 7380.
Swanson, L. W. & Schwind, G. A. 1978 Electron emission from a liquid metal. Journal
of Applied Physics 49 (11), 56555662.
Swanson, L. W., Schwind, G. A. & Bell, A. E. 1980 Measurement of the energy
distribution of a gallium liquid metal ion source. Journal of Applied Physics 51 (7), 3453
3455.
Tang, Keqi & Gomez, Alessandro 1995 Generation of monodisperse water droplets
from electrosprays in a corona-assisted cone-jet mode. Journal of Colloid and Interface
Science 175 (2), 326332.
Tang, Keqi & Gomez, Alessandro 1996 Monodisperse electrosprays of low electric
conductivity liquids in the cone-jet mode. Journal of Colloid and Interface Science 184 (2),
500511.
Taylor, Geoffrey 1964 Disintegration of water drops in an electric field. Proceedings of
the Royal Society of London. Series A. Mathematical and Physical Sciences 280 (1382),
383397.

183

Taylor, Geoffrey 1966 Studies in electrohydrodynamics. i. the circulation produced in a


drop by electrical field. Proceedings of the Royal Society of London. Series A. Mathematical
and Physical Sciences 291 (1425), 159166.
Taylor, G. I. & McEwan, A. D. 1965 The stability of a horizontal fluid interface in a
vertical electric field. Journal of Fluid Mechanics Digital Archive 22 (01), 115.
Thong, K. C. & Weinberg, F. J. 1971 Electrical control of the combustion of solid and
liquid particulate suspensions. Proceedings of the Royal Society of London. A. Mathematical and Physical Sciences 324 (1557), 201215.
Tsouris, C., NEAL, S. H., SHAH, V. M., SPURRIER, M. A. & LEE, M. K. 1997
Comparison of liquid-liquid dispersions formed by a stirred tank and electrostatic spraying*. Chemical Engineering Communications 160 (1), 175197.
Tsouris, Costas, Shin, Won-Tae & Yiacoumi, Sotira 1998 Pumping, spraying, and
mixing of fluids by electric fields. The Canadian Journal of Chemical Engineering 76 (3),
589599.
Ude, S. & de la Mora, J.Fernndez 2003 Hypersonic impaction with molecular mass
standards. Journal of Aerosol Science 34 (9), 12451266.
Ude, Sven, Fernandez de la Mora, Juan, Alexander IV, James N. & Saucy,
Daniel A. 2006 Aerosol size standards in the nanometer size range: II. narrow size
distributions of polystyrene 311 nm in diameter. Journal of Colloid and Interface Science
293 (2), 384393.
Ude, S. & de la Mora, J. Fernndez 2005 Molecular monodisperse mobility and mass
standards from electrosprays of tetra-alkyl ammonium halides. Journal of Aerosol Science
36 (10), 12241237.
Uematsu, Ikuo, Matsumoto, Hidetoshi, Morota, Kenji, Minagawa, Mie, Tanioka, Akihiko, Yamagata, Yutaka & Inoue, Kozo 2004 Surface morphology and
biological activity of protein thin films produced by electrospray deposition. Journal of
Colloid and Interface Science 269 (2), 336340.
Vonnegut, Bernard & Neubauer, Raymond L. 1952 Production of monodisperse
liquid particles by electrical atomization. Journal of Colloid Science 7 (6), 616622.
Wakamatsu, Yuji, Matsumoto, Hidetoshi, Minagawa, Mie & Tanioka, Akihiko
2006 Effect of ion-exchange nanofiber fabrics on water splitting in bipolar membrane.
Journal of Colloid and Interface Science 300 (1), 442445.
Ward, J. W. & Seliger, R. L. 1981 Trajectory calculations of the extraction region of a
liquid-metal ion source. Journal of Vacuum Science and Technology 19 (4), 10821086.
Weinberg, F. J. 1968 Electrical aspects of aerosol formation and control. Proceedings of
the Royal Society of London. Series A. Mathematical and Physical Sciences 307 (1489),
195208.
184

Wikipedia 2013 Ion.


Wilm, Matthias & Mann, Matthias 1996 Analytical properties of the nanoelectrospray
ion source. Analytical Chemistry 68 (1), 18.
Wilm, Matthias S. & Mann, Matthias 1994 Electrospray and taylor-cone theory, doles
beam of macromolecules at last? International Journal of Mass Spectrometry and Ion
Processes 136 (23), 167180.
Wilson, C. T. R. & Taylor, G. I. 1925 The bursting of soap-bubbles in a uniform electric
field. Mathematical Proceedings of the Cambridge Philosophical Society 22 (05), 728730.
Winger, Brian E., Hofstadler, Steven A., Bruce, James E., Udseth, Harold R.
& Smith, Richard D. 1993 High-resolution accurate mass measurements of biomolecules
using a new electrospray ionization ion cyclotron resonance mass spectrometer. Journal of
the American Society for Mass Spectrometry 4 (7), 566577.
Xie, Jingwei, Lim, Liang Kuang, Phua, Yiyong, Hua, Jinsong & Wang, Chi-Hwa
2006 Electrohydrodynamic atomization for biodegradable polymeric particle production.
Journal of Colloid and Interface Science 302 (1), 103112.
Xie, Jingwei, Ng, Wei Jun, Lee, Lai Yeng & Wang, Chi-Hwa 2008 Encapsulation of
protein drugs in biodegradable microparticles by co-axial electrospray. Journal of Colloid
and Interface Science 317 (2), 469476.
Xie, Jingwei & Wang, Chi-Hwa 2007 Electrospray in the dripping mode for cell microencapsulation. Journal of Colloid and Interface Science 312 (2), 247255.
Yahiku, Arthur Y., Mahoney, J., Daley, H. & Perel, J. 1970 Experimental study
of colloid annular thrusters. In 8th Electric Propulsion Conference. American Institute of
Aeronautics and Astronautics.
Yamashita, Masamichi & Fenn, John B. 1984 Electrospray ion source. another variation
on the free-jet theme. The Journal of Physical Chemistry 88 (20), 44514459.
Young, Thomas 1805 An essay on the cohesion of fluids. Philosophical Transactions of the
Royal Society of London 95, 6587, ArticleType: research-article / Full publication date:
1805 / Copyright 1805 The Royal Society.
Yu, Jian H. & Rutledge, Gregory C. 2002 Electrospinning. In Encyclopedia of Polymer
Science and Technology. John Wiley & Sons, Inc.
Zeleny, John 1907 The discharge of electricity from pointed conductors differing in size.
Physical Review (Series I) 25 (5), 305.
Zeleny, John 1914 The electrical discharge from liquid points, and a hydrostatic method
of measuring the electric intensity at their surfaces. Physical Review 3 (2), 69.
Zeleny, John 1917 Instability of electrified liquid surfaces. Physical Review 10 (1), 1.
185

Zeleny., John 1920 Electrical discharges from pointed conductors. Physical Review 16 (2),
102125.

186

Appendix A
Mathematical Expressions and
Numerical Coefficients

A.1

Normal and Tangential Directions in the Curvilinear


System of Coordinates

In the curvilinear system of coordinates the two principal directions n and t, normal and
tangential, are defined as functions of the profile, Rs (z), expressed as,

Rs0 (z)

1
p
1 + Rs0 (z)2

p
, 0, p
1 + Rs0 (z)2
1 + Rs0 (z)2

n=

(
t=

R0 (z)
p s
, 0,
1 + Rs0 (z)2

187

(A.1)

(A.2)

A.2

Summary of The Governing Equations and Boundary Conditions


Table A.1: Governing Equations Summary
Domain
Description
1 : Liquid

2 : V acuum

E (1) = (1)

E (2) = (2)

Laplaces Equation

2 (1) = 0

2 (2) = 0

Continuity

u=0

N/A

u u = p +

N/A

Electric Potential

Momentum Balance

Table A.2: Boundary Conditions Summary


Description

Boundary Condition

Interface Charge Balance

I :

s (uq) = E (1) n

Normal Electric Jump

I :

0 E (2) n r 0 E (1) n = q

Tangential Electric Continuity

I :

E (1) t = E (2) t

I :

n p + n n =


(2)2
(1)2
(2)2
= 12 0 En r En
+ 21 (r 1) Et

Tangential Stress Balance

I :

t n = qEt

Far Field Electric Potential

0 :

(2) = o + AR 2 P 1 (cos())

Liquid Inlet Electric Potential

1 :

(1) = o + AR 2 P 1 (cos())

Normal Stress Balance

Liquid Outlet Electric Potential

3 :

1
2

(1) = o + AR P 1 (cos())
2

188

A.3

Collocated Mesh Finite Volume Coefficients

Substituting these approximations into the momentum equation, we obtain an algebraic


equation of the form:
AiP ui,P +

Aik ui,k = QiP ;

k = E, W, N, S

(A.3)

where the coefficients are given by


e Se
,
xE xP
n S n
min(m
in , 0)
,
xN xP
w Sw
,
min(m
iw , 0)
xP xW
s Ss
,
min(m
ip , 0)
xP xS
3 X u

Ak ,
t
k

AiE = min(m
ie , 0)
AiN =
AiW =
AiS =
AP =

A.4

(A.4)

k = E, W, N, S

SIMPLE Method Coefficients

The solution of the algebraic system of equation previously derived from the momentum
equation gives us updated velocity fields u and v that satisfy the momentum equation but
do not necessarily satisfy the continuity equation. In order to correct this field, the pressure
is corrected to implicitly satisfy the continuity equation. This correction, also known as
divergence free projection, is then used to modify the velocity field. This procedure, repeated
iteratively, results in a velocity and pressure field that will satisfy the momentum equation
189

and continuity simultaneously, within the defined accuracy range.


The algebraic equation for the pressure correction is derived from the momentum and continuity equations. Details of this derivation can be found in Ferziger & Peric (1997). The
final form of this algebraic equation is given by,
ApP p0P =

p m
0p
Apl p0l = m

(A.5)

where the coefficients are


ApE
ApN
ApP


=

S 2
Aup


,
e


S 2
=
Avp
X p n
Al
=


,
,


S 2
=
Aup w
 2
S
p
AS =
Avp s

ApW

(A.6)

l = E, W, N, S

With the use of the pressure correction p0 , the velocity correction u0i can be obtained. For
instance, for ue vertical velocity at face e, the correction is given by
u0e = uen1 ue

Si 0
(p p0P )
Aup i E

(A.7)

The final velocity field is updated with ui = ui + u0i and the pressure by pn+1 = pn + p0 .
The FV method uses a 3D set of control volumes with interconnected surface areas. In the
axisymmetric case, it becomes a planar grid of 2D interconnected finite area (see figure 4.1).

190

Potrebbero piacerti anche