University of Kentucky University of Kentucky
UKnowledge
UKnowledge
Theses and Dissertations--Electrical andComputer Engineering Electrical and Computer Engineering
2021
Boundary Integral Equation Method for Electrostatic Field
Boundary Integral Equation Method for Electrostatic Field
Prediction in Piecewise-Homogeneous Electrolytes
Prediction in Piecewise-Homogeneous Electrolytes
Christopher Keith Pratt
University of Kentucky, chrispratt400@gmail.com
Digital Object Identifier: https://doi.org/10.13023/etd.2021.136
Right click to open a feedback form in a new tab to let us know how this document benefits you. Right click to open a feedback form in a new tab to let us know how this document benefits you.
Recommended Citation Recommended Citation
Pratt, Christopher Keith, "Boundary Integral Equation Method for Electrostatic Field Prediction in Piecewise-Homogeneous Electrolytes" (2021). Theses and Dissertations--Electrical and Computer Engineering. 164.
https://uknowledge.uky.edu/ece_etds/164
This Master's Thesis is brought to you for free and open access by the Electrical and Computer Engineering at UKnowledge. It has been accepted for inclusion in Theses and Dissertations--Electrical and Computer Engineering by an authorized administrator of UKnowledge. For more information, please contact UKnowledge@lsv.uky.edu.
STUDENT AGREEMENT: STUDENT AGREEMENT:
I represent that my thesis or dissertation and abstract are my original work. Proper attribution has been given to all outside sources. I understand that I am solely responsible for obtaining any needed copyright permissions. I have obtained needed written permission statement(s) from the owner(s) of each third-party copyrighted matter to be included in my work, allowing electronic distribution (if such use is not permitted by the fair use doctrine) which will be submitted to UKnowledge as Additional File.
I hereby grant to The University of Kentucky and its agents the irrevocable, non-exclusive, and royalty-free license to archive and make accessible my work in whole or in part in all forms of media, now or hereafter known. I agree that the document mentioned above may be made available immediately for worldwide access unless an embargo applies.
I retain all other ownership rights to the copyright of my work. I also retain the right to use in future works (such as articles or books) all or part of my work. I understand that I am free to register the copyright to my work.
REVIEW, APPROVAL AND ACCEPTANCE REVIEW, APPROVAL AND ACCEPTANCE
The document mentioned above has been reviewed and accepted by the student’s advisor, on behalf of the advisory committee, and by the Director of Graduate Studies (DGS), on behalf of the program; we verify that this is the final, approved version of the student’s thesis including all changes required by the advisory committee. The undersigned agree to abide by the statements above.
Christopher Keith Pratt, Student Dr. John C. Young, Major Professor Dr. Daniel Lau, Director of Graduate Studies
BOUNDARY INTEGRAL EQUATION METHOD FOR ELECTROSTATIC FIELD PREDICTION IN PIECEWISE-HOMOGENEOUS ELECTROLYTES
________________________________________ THESIS
________________________________________ A thesis submitted in partial fulfillment of the
requirements for the degree of Master of Science in Electrical Engineering
in the College of Engineering at the University of Kentucky
By
Christopher Pratt Lexington, Kentucky
Director: Dr. John Young, Professor of Electrical Engineering Lexington, Kentucky
2021
ABSTRACT OF THESIS
BOUNDARY INTEGRAL EQUATION METHOD FOR ELECTROSTATIC FIELD PREDICTION IN PIECEWISE-HOMOGENEOUS ELECTROLYTES
This thesis presents a method to predict electrostatic fields, potentials, and currents in regions containing piecewise-homogeneous electrolytes. Additionally, an efficient electric field calculation is presented. A boundary integral equation is formulated for the boundary potentials and currents and is discretized using the Locally Corrected Nyström method. Solution convergence with respect to the mesh discretization and basis order is investigated. The techniques are validated through analysis of problems with either analytic solutions, with published data, or with other solution methods.
KEYWORDS: Corrosion-related field, Boundary integral equation, Locally Corrected Nyström method
Christopher K. Pratt May 14, 2021 Date
BOUNDARY INTEGRAL EQUATION METHOD FOR ELECTROSTATIC FIELD PREDICTION IN PIECEWISE-HOMOGENEOUS ELECTROLYTES
By
Christopher Pratt
Dr. John Young Director of Thesis
Dr. Daniel Lau Director of Graduate Studies
May 14, 2021 Date
iii
ACKNOWLEDGMENTS
I am very grateful to have been given this opportunity to work with my advisor Dr. John C. Young. The advice and instruction provided by Dr. Young during my research gave me a great learning experience in this complex and interesting field. The courses in theoretical and computational electromagnetics instructed by Dr. Young inspired me to further my education and purse graduate research in this field. I am thankful for Dr. Young’s time and expertise throughout my studies.
I would also like to thank my committee members Dr. Robert J. Adams and Dr. Cai-Cheng Lu for reviewing my thesis and for their excellent courses that assisted in my understanding in electromagnetics and computational methods. I am grateful for Dr. Johné Parker’s inspiration to seek higher education during my time assisting a part of her research as an undergraduate.
I am appreciative of the opportunities and the support of this research provided by the Office of Naval Research under grant N00014-16-1-3066.
Lastly, I would like to thank my family for supporting me during my education. I would not have been able to have this achievement without their love and support.
iv
TABLE OF CONTENTS
ACKNOWLEDGMENTS ... iii
LIST OF FIGURES ... v
CHAPTER 1. Introduction ... 1
CHAPTER 2. Piecewise-Homogeneous Electrolyte Method ... 3
2.1 Boundary Integral Equation Method ... 3
2.2 Locally Corrected Nyström Discretization ... 8
CHAPTER 3. Electric Field Calculation ... 12
CHAPTER 4. Results and Validation... 15
4.1 Insulated box filled with two electrolytes. ... 15
4.2 Insulated Box with Split Electrolyte Interface ... 21
4.3 Insulated Box with Multiple Electrolyte Interfaces ... 25
4.4 Insulated Box containing Sphere Electrolyte Interface ... 31
4.5 Box with Linearly Varying Conductivity ... 35
4.6 Box with Aperture ... 43
4.7 Insulated cylindrical tank filled with two electrolytes. ... 48
4.8 Sphere encapsulating an electrode ... 51
4.9 Sphere encapsulating a Brick... 54
CHAPTER 5. Conclusion ... 57
References ... 58
v
LISTOFFIGURES
Figure 2.1: Conductivity models with I homogeneous regions ...3 Figure 4.1: A two-meter brick with two homogeneous electrolytes. ...16 Figure 4.2: (a) Electric potential vs. position and (b) electric field vs. position for various
basis orders p containing homogeneous electrolyte with σ1=σ2 =4 S/m and meshed
using quadrilateral cells of edge length 0.25 m. ...17 Figure 4.3: (a) Relative error vs. maximum edge length of the electric potential and (b) electric field for various basis orders p containing homogeneous electrolyte with
1 2 4 S/m
σ =σ = . ...18 Figure 4.4: (a) Electric potential vs. position and (b) electric field vs. position for various
basis orders p containing piecewise-homogeneous electrolytes with σ1 =6 S/m and
2 3 S/m
σ = and meshed using quadrilateral cells of edge length 0.25 m. ...19
Figure 4.5: (a) Relative error vs. maximum edge length of the electric potential and (b) electric field for various basis orders p containing piecewise-homogeneous electrolytes with σ1=6 S/m and σ2 =3 S/m. ...20
Figure 4.6: A two-meter brick with two homogeneous electrolytes and a split electrolyte interface. The arrows indicate the direction of the surface normals. ...22 Figure 4.7: (a) Electric potential vs. position and (b) electric field vs. position for various
basis orders p containing homogeneous electrolytes with σ1=σ2 =4 S/m and meshed
using quadrilateral cells of edge length 0.25 m. ...23 Figure 4.8 (a) Electric potential vs. position and (b) electric field vs. position for various
basis orders p containing piecewise-homogeneous electrolytes with σ1 =3 S/m and
2 5 S/m
σ = and meshed using quadrilateral cells of edge length 0.25 m. ...24
Figure 4.9: A three-meter brick with three homogeneous electrolytes. ...26 Figure 4.10 (a) Electric potential vs. position and (b) electric field vs. position for various
basis orders p containing homogeneous electrolytes with σ1=σ2 =4 S/m and meshed
vi
Figure 4.11: (a) Relative error vs. maximum edge length of the electric potential and (b) electric field for various basis orders p containing homogeneous electrolyte with
1 2 4 S/m
σ =σ = . ...28 Figure 4.12 (a) Electric potential vs. position and (b) electric field vs. position for various
basis orders p containing piecewise-homogeneous electrolytes with σ1 =6 S/m,
2 4 S/m
σ = , and σ3 =2 S/m and meshed using quadrilateral cells of edge length 0.25 m.
...29 Figure 4.13: (a) Relative error vs. maximum edge length of the electric potential and (b) electric field for various basis orders p containing piecewise-homogeneous electrolytes with σ1=6 S/m, σ2 =4 S/m, and σ3 =2 S/m. ...30
Figure 4.14: A two-meter long brick with two homogeneous electrolytes and a spherical electrolyte interface. ...32 Figure 4.15: (a) Electric potential vs. position and (b) electric field vs. position for various
basis orders p containing homogeneous electrolytes with σ1=σ2 =4 S/m and meshed
using quadrilateral cells of edge length 0.1 m. ...33 Figure 4.16: (a) Electric potential vs. position and (b) electric field vs. position for various
basis orders p and piecewise-homogeneous electrolytes with σ1 =2 S/m, and σ2 =3 S/m
and meshed using quadrilateral cells of edge length 0.1 m. ...34
Figure 4.17: (a-e) One-meter box with n =1,...,5 homogeneous electrolyte(s). ...37
Figure 4.18: Electric potential vs. position and (b) electric field vs. position for various
basis orders p containing electrolyte varying linearly from z =0 0 m with σ0 =1 S/m and
1 1 m
z = with σ1=2 S/m with n = regions and meshed using quadrilateral cells of edge 1
length 0.25 m. ...38 Figure 4.19: Electric potential vs. position and (b) electric field vs. position for various
basis orders p containing electrolyte varying linearly from z =0 0 m with σ0 =1 S/m and
2 1 m
z = with σ1 =2 S/m with n = regions and meshed using quadrilateral cells of edge 2
vii
Figure 4.20: Electric potential vs. position and (b) electric field vs. position for various
basis orders p containing electrolyte varying linearly from z =0 0 m with σ0 =1 S/m and
3 1 m
z = with σ1 =2 S/m with n = regions and meshed using quadrilateral cells of 3
maximum edge length 0.25 m. ...40 Figure 4.21: Electric potential vs. position and (b) electric field vs. position for various
basis orders p containing electrolyte varying linearly from z =0 0 m with σ0 =1 S/m and
4 1 m
z = with σ1 =2 S/m with n = regions and meshed using quadrilateral cells of 4
maximum edge length 0.25 m. ...41 Figure 4.22: Electric potential vs. position and (b) electric field vs. position for various
basis orders p containing electrolyte varying linearly from z =0 0 m with σ0 =1 S/m and
5 1 m
z = with σ1 =2 S/m with n = regions and meshed using quadrilateral cells of 5
maximum edge length 0.25 m. ...42 Figure 4.23: A box with aperture inside a larger insulating box. ...44 Figure 4.24: Electrical potential vs. position within the interior box (a) without using zoning and (b) using zoning for various basis orders p and both boxes are meshed using quadrilateral cells of maximum edge length 5 cm. ...45 Figure 4.25: Electric potential vs. position outside the interior box (a) without using zoning and (b) using zoning for various basis orders p and both boxes are meshed using quadrilateral cells of maximum edge length 5 cm. ...46 Figure 4.26: (a) Electrical potential vs. position within the interior box and (b) electric potential vs. position outside interior box (b) for various basis orders p calculated with zoning without the exterior box and the interior box is meshed using quadrilateral cells of maximum edge length 5 cm. ...47 Figure 4.27: A cylindrical tank with a false surface placed along the y-axis. ...48
Figure 4.28: Ex vs. position (a) without using zoning and (b) using zoning for various basis
orders p and the tank is meshed using quadrilateral cells of maximum edge length 20 cm with a depth of 2.625 m. ...49
Figure 4.29: e vs. position (a) without using zoning and (b) using zoning for various basis z
orders p and the tank is meshed using quadrilateral cells of maximum edge length 20 cm with a depth of 2.625 m. ...50
viii
Figure 4.30: A pair of electrodes with a fictious sphere encompassing an electrode. ...51 Figure 4.31: Electric potential vs. position (a) without using zoning and (b) using zoning for various basis orders p and meshed using quadrilateral cells of maximum edge length 0.25 m. ...52 Figure 4.32: Electric field vs. position (a) without using zoning and (b) using zoning for various basis orders p and meshed using quadrilateral cells of maximum edge length 0.25 m. ...53 Figure 4.33: A brick encapsulated by a fictious sphere. ...54 Figure 4.34: Electric potential vs. position (a) without using zoning and (b) using zoning for various basis orders p and meshed using quadrilateral cells of maximum edge length 0.25 m. ...55 Figure 4.35: Electric field vs. position (a) without using zoning and (b) using zoning for various basis orders p and meshed using quadrilateral cells of maximum edge length 0.25 m. ...56
1
CHAPTER 1. INTRODUCTION
Modern engineering involving conductors surrounded by or immersed in electrolyte, such as the hull of a ship in seawater, is faced with the challenge of mitigating corrosion [1-4]. These conducting structures may be difficult to protect from corrosion over time due to the conditions the structure is exposed to. While the use of insulating paint can be used to prevent corrosion of conducting surfaces, this coating may be damaged during regular operation of the system. In other cases, insulating paint may not be an option for some surfaces.
One preventative method for slowing corrosion is an impressed-current corrosion protection system. This system slows the corrosion process by injecting electric current into the electrolyte to adjust the structure’s surface potential. Impressed-current corrosion protection systems are complex to design and are often dependent on numerical modeling of the system to predict the corrosion-related currents and electromagnetic fields. Numerical modelling techniques include finite-difference methods [5], finite-element methods [6, 7], and boundary element methods (BEM) [8]. The BEM is a convenient method since the electric currents and fields can be solved using a surface integral equation without the need to mesh the surrounding electrolytic medium.
In the analysis of a marine structure and buried pipes, the prediction of corrosion currents and fields in an electrolyte may assume the surrounding electrolyte to be homogeneous. In reality, the conductivity of electrolytes, e.g., seawater or soil, typically varies with location. This implies that the assumption of a constant conductivity may be a poor approximation in some cases. For a better approximation of inhomogeneous conductivity, a piecewise-constant conductivity model, referred to as zoning in this work, is discussed. Herein, each region with a homogeneous electrolyte is called a zone. The application of zoning will be applied with marine structures in mind, but these methods may be applicable to other systems with a surrounding inhomogeneous conductivity.
In this thesis, a boundary integral equation is formulated for the analysis of piecewise-homogeneous electrolytes. The integral equation is discretized using the Locally Corrected Nyström method [9-11], and the solution convergence with respect to mesh discretization and basis order is investigated. An efficient electric field calculation
2
is presented as well and is verified through comparison to literature data [12, 13], analytic solutions, and data for single electrolyte problems.
3
CHAPTER 2. PIECEWISE-HOMOGENEOUS ELECTROLYTE METHOD
2.1 Boundary Integral Equation Method
Consider a piecewise-homogeneous electrolyte model for I homogeneous regions
i
V where region Vi has homogeneous conductivity σi for i = 1… I. Let the region V0 be
the region outside the problem domain. Figure 2.1 illustrates an electrolyte with I
homogeneous regions. The shared surface between regions Vi and Vj is denoted Γ = Γij ji
with the unit surface normal ˆn pointing from region ij Vi into region Vj. The exterior
boundary of the total region is Γ0, which may recede to infinity for unbounded problems,
and the total boundary of Viis Γi.
Figure 2.1: I regions with piecewise-homogeneous conductivity σi.
The potential Φiwithin Vi is given by [12]
( )
( )
ˆ(
,)
(
,)
ˆ( )
i i i i G d G i d Γ Γ ′ ′ ′ ′ ′ ′ ′ ′ ′ ′ Φ r + Φ
∫
r n ⋅∇ r r Γ =
∫
r r n ⋅ Φ∇ r Γ (2.1)4
where ˆ′n points out of region Vi. The electrostatic Green’s function
(
,)
1/ 4G r r′ = π r r− ′ is assumed unless noted otherwise. For simplicity, it will be
assumed all surfaces Γi are bounded except possibly for Γ0 .
To obtain a set of coupled integral equations with a unique solution, the shared surfaces between each region must have both the electric potential be continuous
( )
( )
i j
Φ r = Φ r (2.2)
and the normal electric current density be continuous
( )
( )
ˆ i ˆ j ij⋅ = ij⋅ n J r n J r (2.3) for r∈Γijwhere( )
( )
, i i i i V σ = − ∇Φ ∈ J r r r . (2.4)The boundary condition on insulating surfaces Γi,insulating ⊂ Γi is that the normal current
density is zero
( )
ˆ⋅ i =0
n J r (2.5)
for r∈Γi,insulating. The excitations in the model are the boundary conditions over the electrodes Γi,electrode⊂ Γi and are either a fixed potential
( )
,electrode( )
i i
Φ r = Φ r (2.6)
or a fixed normal current density
( )
,electrode( )
ˆ i i
i J
− ⋅n J r = r (2.7)
for r∈Γi,electrode based on whether the boundary condition is Dirichlet or Neumann,
respectively.
For interior, linear Neumann problems, a one-dimensional null space corresponding to a constant potential offset is addressed by enforcing the average potential over some
surface Γavgto a constant C
( )
avg . d C Γ ′ ′ Φ Γ =∫
r (2.8)5 ,electrode ˆ 0 i i d Γ ′ ⋅ Γ =
∫
n J (2.9)is satisfied over all electrode surfaces Γi,electrode. Additionally, (2.9) must be a priori satisfied
for linear Neumann problems but is not required to be enforced explicitly for linear, Dirichlet or linear, mixed boundary condition problems since these problems already have the potential fixed somewhere in the problem. Enforcing the conditions (2.2) - (2.9) produces a uniquely solvable system.
The reference direction of the unit surface normal ˆn on a shared interface ij Γij is
always directed from i into j. On a shared interface Γij, let Φij
( )
r = Φi( )
r = Φj( )
r and( )
( )
( )
ˆ ij ˆ i ˆ j
ij⋅ = ij⋅ = ij⋅
n J r n J r n J r where i < j to fix the reference direction for the
normal current density. In the following, i j< is assumed unless otherwise noted.
At r∈Γij, if we approach along a path from outside the respective region, then
( )
lim( )
0 and( )
lim( )
0,j i
i i j j
A r = −r→r+Φ r = A r = −r r→ +Φ r = (2.10)
where r+ indicates the limit is taken along a path outside the respective region. The
limiting operation for a small surface∆S [14] is
( )
(
)
( )
(
)
( )
0 0 1 ˆ ˆ lim , P.V. , 2 S S G ds G ds ± → ∆∫
Φ ′ ′ ′ ⋅ ′ ′= ∆∫
Φ ′ ′ ′ ⋅ ′ ′± Φ r r r ∇ r r n r ∇ r r n r (2.11) where(
( )
)
0 ˆ lim h + h ± → = ±r r n r and the surface normal is ˆn . P.V. denotes a principal value
integral and is suppressed throughout.
To enforce (2.2) and (2.3) on a shared interface Γij, a linear combination ofAiand
j
A
( )
( )
0i iA jAj
α r +α r = (2.12)
is taken for r∈Γij. Splitting the integrals into ones over Γij and not over Γij and applying
6
( )
( )
(
)
( )
(
)
(
)
( )
(
)
( )
in, in, 1 0 ( ) 2 ˆ , ˆ , 1 , ˆ 1 , ˆ i ij ij i ij ij i ij i ij i i ij i i ij i ij i i A G d G d G d G d σ σ Γ −Γ Γ Γ −Γ Γ = = Φ ′ ′ ′ ′ ′ ′ ′ ′ ′ ′ − Φ ⋅ Γ + Φ ⋅ Γ ′ ′ ′ ′ ′ ′ ′ ′ − ⋅ Γ + ⋅ Γ∫
∫
∫
∫
r r r n r r r n r r r r n J r r r n J r ∇ ∇ (2.13) and( )
( )
(
)
( )
(
)
(
)
( )
(
)
( )
in, in, 1 0 ( ) 2 ˆ , ˆ , 1 , ˆ 1 , ˆ j ij ij j ij ij j ij j ij j ij j ij j ij j j A G d G d G d G d σ σ Γ −Γ Γ Γ −Γ Γ = = Φ ′ ′ ′ ′ ′ ′ ′ ′ ′ ′ − Φ ⋅ Γ − Φ ⋅ Γ ′ ′ ′ ′ ′ ′ ′ ′ − ⋅ Γ − ⋅ Γ∫
∫
∫
∫
r r r n r r r n r r r r n J r r r n J r ∇ ∇ (2.14)where nˆin,i = −n is into the electrolyte of region ˆi Vi and nˆin,j = −n is into the electrolyte ˆj
of region Vj. Here αi =αj =1 is chosen to eliminate the strong singularity in the integral
over Φ . Applying these constants gives
( )
( )
( )
( )
(
)
( )
(
)
(
)
( )
(
)
( )
(
)
( )
in, in, in, in, ˆ , ˆ , 1 , ˆ 1 , ˆ 1 1 , ˆ 0 i ij j ij i ij j ij ij i i j j ij i i j j i i i j j j ij ij i j A A G d G d G d G d G d α α σ σ σ σ Γ −Γ Γ −Γ Γ −Γ Γ −Γ Γ ′ ′ ′ ′ ′ + = Φ − Φ ⋅ Γ ′ ′ ′ ′ ′ − Φ ⋅ Γ ′ ′ ′ ′ − ⋅ Γ ′ ′ ′ ′ − ⋅ Γ ′ ′ ′ ′ + − ⋅ Γ =∫
∫
∫
∫
∫
r r r r n r r r n r r r r n J r r r n J r r r n J r ∇ ∇ (2.15) for r∈Γij.To enforce (2.2) and (2.3), a linear combination
( )
( )
0 i iB jBj β r +β r = (2.16) is taken where( )
( )
in, in, ˆ ˆlim 0 and lim 0
j i
i i j j
i j
7
for r∈Γij and where the limits are taken along paths outside the respective regions.
Separating these integrals gives
( )
( )
(
)
( )
(
)
( )
(
)
( )
(
)
( )
in, in, in, in, 0 in, in, in, in, ˆ ˆ 0 , ˆ ˆ lim , 2 ˆ , ˆ ˆ , ˆ i ij ij a i ij i i i i i ij i ij i i i a S i i i i i i ij i i i B G d G d a G d G d σ σ σ σ σ σ σ Γ −Γ → Γ − Γ −Γ ′ ′ ′ ′ = = Φ ⋅ ⋅ Γ Φ ′ ′ ′ ′ + Φ ⋅ ⋅ Γ + ′ ′ ′ ′ ′ + ⋅ ⋅ Γ ′ ′ ′ ′ ′ + ⋅ ⋅∫
∫
∫
r r n n r r r r n n r r n r r n J r n r r n J r ∇ ∇ ∇ ∇ ∇ ∇ ˆin,( )
2 ij ij i i i σ σ Γ ⋅ Γ −∫
n J r (2.18) and( )
( )
(
)
( )
(
)
( )
(
)
( )
(
)
( )
in, in, in, in, 0 in, in, in, in, ˆ ˆ 0 , ˆ ˆ lim , 2 ˆ , ˆ ˆ , ˆ j ij ij a j ij j j j j j ij j ij j j j a S j j j j j j ij j j j B G d G d a G d G d σ σ σ σ σ σ σ Γ −Γ → Γ − Γ −Γ ′ ′ ′ ′ = = Φ ⋅ ⋅ Γ Φ ′ ′ ′ ′ + Φ ⋅ ⋅ Γ + ′ ′ ′ ′ ′ + ⋅ ⋅ Γ ′ ′ ′ ′ ′ + ⋅ ⋅∫
∫
∫
r r n n r r r r n n r r n r r n J r n r r n J r ∇ ∇ ∇ ∇ ∇ ∇ ˆin,( )
. 2 ij ij j j j σ σ Γ ⋅ Γ −∫
n J r (2.19)The procedure for dealing with the ∇∇ term is taken from [15] where S is a a
circular disk of radius a centered on the field point r∈Γij. A linear combination is taken
using i 1 i β σ = and j 1 j β σ
= − , which eliminates the terms over Γij and the circular disk
a
S in (2.18) and (2.19) while maintaining the non-integrated J rij
( )
terms. Applying this8
( )
( )
( )
(
)
( )
(
)
(
)
( )
(
)
( )
in, in, in, in, in, in, in, in, 1 1 ˆ ˆ , ˆ ˆ , 1 ˆ , ˆ 1 ˆ , ˆ 1 1 i ij j ij i ij j ij i j i i i i i j i j j i j j i i i i j i j j i j B B G d G d G d G d σ σ σ σ σ σ σ σ σ σ Γ −Γ Γ −Γ Γ −Γ Γ −Γ ′ ′ ′ ′ − = Φ ⋅ ⋅ Γ ′ ′ ′ ′ + Φ ⋅ ⋅ Γ ′ ′ ′ ′ ′ + ⋅ ⋅ Γ ′ ′ ′ ′ ′ + ⋅ ⋅ Γ + −∫
∫
∫
∫
r r r n n r r r n n r r n r r n J r n r r n J r ∇ ∇ ∇ ∇ ∇ ∇(
)
( )
( )
in, in, in, ˆ , ˆ ˆ 1 1 2 0 ij ij i i ij i i j G d σ σ Γ ⋅ ′ ′ ′ ⋅ ′ Γ′ ′ ⋅ − + =∫
n r r n J r n J r ∇ (2.20) for r∈Γij.For r∈Γi, such that r is not on a surface between electrolytes, it is only necessary
to enforce
( )
0 lim i 0 + → Φ = r r r for 0 0∈Vr , giving the equation
( )
( )
in,(
)
(
)
in,( )
1 ˆ 1 ˆ 0 , , . 2 i i i i i i G i d σi G i i d Γ Γ ′ ′ ′ ′ ′ ′ ′ ′ ′ = Φ r − Φ∫
r n ⋅∇ r r Γ −∫
r r n ⋅J r Γ (2.21)Finally, note that in (2.15), (2.20), and (2.21), whenever the source point r′ is on a
shared surface Γik, i.e., r′∈Γik, then Φi
( )
r = Φik( )
r and( )
( )
ˆ i ˆ ik , i⋅ = + ik⋅ i k< n J r n J r (2.22) and( )
( )
ˆ i ˆ ki , i⋅ = − ki⋅ i k> n J r n J r . (2.23)Similarly, for terms in region Vj.
2.2 Locally Corrected Nyström Discretization
The integral equations are discretized using the Locally-Corrected Nyström method as described in [9-11]. To illustrate the method, consider the integral equation
9
( )
(
,) ( )
, , h G f ds Ω ′ ′ ′ =∫
∈Ω r r r r r (2.24)where h is the excitation, f is the unknown, Ω is the domain, and G is the kernel. The
domain Ω can be divided into N subdomains s Ω such that n
1 s N n n= Ω =
Ω , giving the equation( )
(
) ( )
1 , s n N n h G f ds = Ω ′ ′ ′ =∑ ∫
r r r r . (2.25)In the Nyström method, the integral in each subdomain is approximated using Q -point n
quadrature rules
{
rqn,ωqn}
where r are the quadrature points and qn ωqn are the quadratureweights. Then, (2.25) is enforced at the field points
{ }
r to obtain qm(
) ( )
1 1 ( )m s n m, n n n ( )n n N Q q q q q q q n q h G f ω g = = ≈∑∑
r r r r r , (2.26)where g( )rqn is the cell Jacobian at the qth quadrature point on the cell n. The matrix
equation formed from (2.26) is
[ ][ ] [ ]
Z f = h . (2.27)The interaction between the quadrature point r of the field cell qm Ω and the quadrature m
point r on the source cell qn Ω is n ωq qm, nsuch that
[ ]
Zmn q qm,n =ωq qm,n , (2.28)where
[ ]
Zmn is the sub-matrix of[ ]
Z for the field cell Ω and source cell m Ω and n{
ωq qm n,}
is the corresponding matrix element value.
For a field cell Ω and source cell m Ω separated such that the integrand in (2.26) is n
sufficiently smooth, the matrix entry ωq qm n, is
(
)
, , ( ).
m n m n n n
q q G q q q g q
10
For field and source cell interactions close enough that standard integration rules begin to lose accuracy, a set of locally corrected quadrature weights may be utilized. The matrix entries
{
ωq qm n,}
are locally corrected weights calculated by solving( )
,(
)
( )
1 , n n m n m n n Q k q q q q k q f ω G f ds = Ω ′ ′ ′ =∑
r ∫
r r r (2.30)for k =1,...,Kn, where
{
fk( )
r}
is an appropriate set of K basis functions defined on the nsource cell Ω . The weights n
{
ωq qm n,}
can be solved from[ ]
m,n mn n q q q
L ω = d , (2.31)
where the local correction matrix
[ ]
,( )
nn n k q k q L = f r and m
(
m,)
( )
n n q k q k d G f ds Ω = ′ ′ ′ ∫
r r r .The basis functions
{
fk( )
r are the set}
{
( ) ( )
1 2}
m n
P u P u where
m
P and P are Legendre polynomials of degree m and n, respectively, such that n
0<m n p, ≤ for quadrilateral cells with local coordinates
(
u u1, 2)
and a representation oforder p. Furthermore, for an order p representation on a quadrilateral cell, Kn =
(
p+1)
2.In discretizing (2.15), (2.20), and (2.21) using the Locally-Corrected Nyström
method, the Nyström unknowns fqn are the sampled potential i
( )
n nq fqΦ
Φ r = and the
sampled normal current ˆ
( )
n n
i
q fq
⋅ = J
n J r at the quadrature point r subject to and qn
continuity constraints and the sign convention given in (2.22) and (2.23) on shared
surfaces. The excitation vector
[ ]
h is determined from the known boundary conditionsover the electrodes and the insulating surfaces. If the normal current ˆ ⋅n J is known
(Neumann boundary type) over an electrode, then the integrals over n Jˆ ⋅ from (2.15),
(2.20), and (2.21) are part of the excitation vector
[ ]
h while the integrals over Φ are partof the system matrix [ ]Z . If the potential Φ is known (Dirichlet boundary type) over the
electrode, then the integrals over Φ from (2.15), (2.20), and (2.21) are part of
[ ]
h while11
( )
ˆ⋅ =0
n J r , all integrals over ˆ ⋅n J are 0 a priori satisfied. On the shared surfaces between
the electrolytes, both Φ and ˆ ⋅n J are unknown.
Interior, linear Neumann problems admit a one-dimensional null space which corresponds to a constant potential offset [16] that may be removed by enforcing the
average potential over some surface Γavgto a constant value C, as in (2.8). To enforce the
12
CHAPTER 3. ELECTRIC FIELD CALCULATION
Consider a region Vi of homogenous conductivity σi. The potential Φ within this
region is
( )
1(
,) ( )
( )
( )
ˆ , i i i i i i V V G ρ dv G G ds V ε ′ ′ ′ Γ ′ ′ ′ ′ ′ ′ Φ r =∫
r r r −
∫
Φ r ∇ − ∇Φ r ⋅n r∈ , (3.1)where ˆn is the unit normal to Γi out of Vi, ε is the permittivity of the material, and
(
,)
1/ 4G r r′ = π r r− ′ is electrostatic Green’s function. If there is no volume charge
density ρV in Vi, then
( )
( )
( )
ˆ , . i i i G G i ds Vi Γ ′ ′ ′ ′ ′ ′ Φ r = −
∫
Φ r ∇ − ∇Φ r ⋅n r∈ (3.2)The electric field E in the region Vi is
( )
( )
( )
(
,)
ˆ( )
ˆ(
,)
i i i i i G d i G d Γ Γ = − Φ ′ ′ ′ ′ ′ ′ ′ ′ ′ Φ ⋅ Γ − ⋅ Γ∫
∫
E r r r r r n J r n r r
∇ = ∇ ∇ ∇ (3.3) where(
)
(
)
3 2 ˆ 1 1 , , 4 4 G ′ = − ′G ′ = − π − ′ = − π ′ − r r u r r r r r r u ∇ ∇ (3.4)with u=
(
r r− ′)
. From the identity [17](
A B⋅)
= × × + × × +A B B A B(
⋅)
A A+(
⋅)
B, ∇ ∇ ∇ ∇ ∇ (3.5) (3.3) becomes( )
( )(
ˆ)
(
,)
( )
ˆ(
,)
. i i i i G d i G d Γ Γ ′ ′ ′ ′ ′ ′ ′ ′ ′ = Φ∫
⋅ Γ −∫
⋅ Γ E r
r n ∇ ∇ r r
J r n ∇ r r (3.6) Applying (3.4) to (3.6) and expanding the equation gives( )
( )
[
]
( )
3 2 2 1 ˆ 3 ˆ 4 ˆ 1 ˆ . 4 i i i i i d d π π Γ Γ ′ ′ ′ ′ = Φ − ⋅ Γ ′ ′ ′ − ⋅ − Γ ∫
∫
E r r n n u u u u u J r n u
(3.7)The integral over Γi is divided into Ni cells n
13
( )
i i n n i n n IΦ I Ω ∈Γ Ω ∈Γ =∑
−∑
J E r , (3.8) where( )
( )
3 2[
]
1 ˆ 3 ˆ 4 n i n I d π Φ Ω ′ ′ ′ = Φ − ⋅ Γ ∫
r r n n u u u u (3.9) and( )
( )
2 ˆ 1 ˆ 4 n i n I π d Ω ′ ′ = ⋅ − Γ ∫
J r J r n u u . (3.10)For field points sufficiently distant from the source cell Ω , the integrals are approximated n
using Q -point quadrature rules n
{
rqn,ωqn}
to obtain( )
( )
3 2 1 1 ˆ 3 ˆ ( ) 4 n n n n n n n n n n n Q i n q q q q q q q q q q I ω g π Φ = ′ ′ ≈ Φ − ⋅ ∑
r r n n u u r u u (3.11) and InJ becomes( )
( )
2 1 ˆ 1 ˆ ( ) 4 n n n n n n n n Q q i n q q q q q q I ω g π = ′ ≈ ⋅ − ∑
J r J r n u r u (3.12)where g( )rqn is the Jacobian of the source cell Ω , n nˆqn =n r , and ˆ
( )
qn uqn =(
r r− qn)
.When the field point is close to the source cell Ω , the quadrature rule approximation n
is not sufficiently accurate. Then, the original integrals are used with the potential approximated as
( )
( )
1 n K i k k k f γ = Φ r ≈∑
r (3.13)and the normal current approximated as
( )
( )
1 ˆ i Kn k k k f γ = ⋅ ≈∑
n J r r , (3.14)14
where
{ }
γk is a set of coefficients for the K set of basis functions n{
fk( )
rqn}
. Forquadrilateral cells, the basis functions
{
fk( )
r are the set of Legendre polynomials}
products defined in the previous section. The coefficients
{ }
γk are solutions to( )
( )
1 n n n K i q k k q k f γ = Φ r ≈∑
r (3.15) and( )
( )
1 ˆ n n n n K i q q k k q k f γ = ⋅ ≈∑
n J r r (3.16) for qn =1,...,Qn. Here,( )
n n i q fqΦ Φ r = and ˆ n i( )
n n q ⋅ q = fqJn J r are from the Locally-Corrected Nyström solution to (2.27) and subject to the sign convention given in (2.22) and (2.23) on shared surfaces.
15
CHAPTER 4. RESULTS AND VALIDATION
The problems presented in this section are for structures filled with piecewise-homogeneous electrolytes to verify the accuracy and correctness of the method.
4.1 Insulated box filled with two electrolytes.
Consider a rectangular box structure centered vertically along the z-axis with a length of 2 m and side lengths of 1 m with a fictitious surface at the center of the structure, as
shown in Figure 4.1. Electrodes are placed at the top of the structure z =2 1 m and bottom
0= 1 m
z − , sourcing and sinking 1 A, respectively. The surfaces on the sides of the structure
are insulating material. The average potential over the bottom electrode is forced to be
zero. The structure was analyzed using basis orders p =0,1,2.
The analytic electric potential and electric field for I regions of homogeneous
conductivity σi with the average potential on the bottom electrode enforce to be zero are
( )
1(
1)
1 1 1 ( ) , T i i i k k i k i k I z z z z V S σ σ − − − = − Φ = − + ∈ ∑
r r (4.1) and( )
T ˆ, i i i I V Sσ = − ∈ E r z r , (4.2)respectively, where IT is the electrode current, S is the cross-sectional area of the box,
and zi is the axial locations of the electrodes and surfaces in between regions. This analytic
solution is relevant for the box problems presented in Sections 4.1 – 4.4.
In Figure 4.2 (a) and (b) are plotted the computed and analytic electric potential and electric field, respectively, along the z-axis. The electrolytes filling the boxes have a
conductivity of σ1=σ2 =4 S/m. The gap seen in the data is a result of calculating the
potential and electric field within each region of the structure. The results are in good agreement with the analytic solution. In Figure 4.3 (a) and (b), respectively, are the relative error of the computed potential and electric field versus the maximum mesh edge length
for basis orders p =0,1,2,3. The potential and electric field for the relative error was
16
the potential and electric field begin the stagnate between 10−3 and 10−4, but the reason for
this has yet to be determined.
In Figure 4.4 (a) and (b) are plotted the computed and analytic electric potential and electric field calculated similarly but with different electrolytes in each box. The bottom
electrolyte has a conductivity of σ1=6 S/m and the top electrolyte has a conductivity of
2
σ = 3 S/m. The results are consistent with the analytic solution provided in (4.1) and (4.2)
. In Figure 4.5 (a) and (b), respectively, are the relative error of the computed potential and
electric field versus the maximum mesh edge length for basis orders p =0,1,2,3. The
potential and electric field for the relative error was calculated no closer than 5 cm to the surfaces between the regions. Both the potential and electric field begin the stagnate
between 10−3 and 10−4, but the reason for this has yet to be determined.
17 (a)
(b)
Figure 4.2: (a) Electric potential vs. position and (b) electric field vs. position for various
basis orders p and conductivity σ1 =σ2 =4 S/m and meshed using quadrilateral cells of
18 (a)
(b)
Figure 4.3: Relative error vs. maximum mesh edge length for (a) the electric potential and
19 (a)
(b)
Figure 4.4: (a) Electric potential vs. position and (b) electric field vs. position for various
basis orders p and conductivity σ1=6 S/m and σ2 =3 S/m and meshed using
20 (a)
(b)
Figure 4.5: Relative error vs. maximum edge length for (a) the electric potential and (b)
21 4.2 Insulated Box with Split Electrolyte Interface
Similar to the previous problem, the interior problem presented here is a rectangular box centered vertically along the z-axis with a length of 2 m and side lengths of 1 m with a fictitious surface at the center of the structure but this fictitious surface, shown in Figure 4.6, is split along the y-axis. The surfaces of the fictitious surface have oppositely directed surface normals. Although this breaks the convention that the reference direction for the
current Jij is defined as i j< , the equations may be modified to allow scenarios when the
surface normal are not pointing in the appropriate direction. Electrodes are placed at the
top of the structure z =2 1 m and bottom z = −0 1 m, sourcing and sinking 1 A,
respectively. The surfaces on the sides of the structure are insulating. The structure was
analyzed using basis orders p =0,1,2.
In Figure 4.7 (a) and (b) are plotted the computed and analytic electric potential and electric field, respectively, along the z-axis. The electrolytes filling the boxes have a
conductivity of σ1 =σ2 =4 S/m. The results are in good agreement with the analytic
solution. A comparison of the data shown in Figure 4.7 to the data in Figure 4.2 provides verification of the robustness of the method.
In Figure 4.8 (a) and (b), respectively, are plotted the potential and electric field along the z-axis as well but with different electrolytes filling the structure. The bottom electrolyte
has a conductivity σ1=3 S/m and the top electrolyte has a conductivity σ2 =5 S/m. The
22
Figure 4.6: A two-meter brick with two homogeneous electrolytes and a split electrolyte interface. The arrows indicate the direction of the surface normals.
23 (a)
(b)
Figure 4.7: (a) Electric potential vs. position and (b) electric field vs. position for various
basis orders p and conductivity σ1 =σ2 =4 S/m and meshed using quadrilateral cells of
24 (a)
(b)
Figure 4.8 (a) Electric potential vs. position and (b) electric field vs. position for various
basis orders p and conductivity σ1 =3 S/m and σ2 =5 S/m and meshed using
25
4.3 Insulated Box with Multiple Electrolyte Interfaces
Consider a rectangular box centered vertically along the z-axis with a length of 3 m and side lengths of 1 m with fictitious surfaces placed 0.5 m above and below the z-axis within the structure, as shown in Figure 4.9. Electrodes are placed at the top of the structure
3 1.5 m
z = and bottom of the structure z = −0 1.5 m, sourcing and sinking 1 A,
respectively. The surfaces on the sides of the structure are insulating material. The structure
was analyzed using basis orders p =0,1,2.
In Figure 4.10 (a) and (b) are plotted the computed and analytic electric potential and electric field, respectively, along the z-axis. The electrolytes filling the boxes have a
conductivity ofσ1=σ2 =σ3 =4 S/m. The results are in good agreement with the analytic
solution. In Figure 4.11 (a) and (b), respectively, are the relative error of the computed potential and electric field versus the maximum mesh edge length for basis orders
0,1,2,3
p = . The potential and electric field for the relative error was calculated no closer
than 5 cm to the surfaces between the regions. Both the potential and electric field relative
errors begin to stagnate around 10−4, but the reason for this has yet to be determined.
In Figure 4.12 (a) and (b) are plotted the computed and analytic electric potential and electric field, respectively, along the z-axis when the box is filled with different
electrolytes. The bottom electrolyte has a conductivity σ1 =6 S/m, the middle electrolyte
has a conductivity of σ2 =4 S/m, and the top electrolyte has a conductivity of σ3 =2 S/m.
The results show good agreement with the analytic solution provided in (4.1) and (4.2). In Figure 4.13 (a) and (b), respectively, are the relative error of the computed potential and
electric field versus the maximum mesh edge length for basis orders p =0,1,2,3. The
potential and electric field for the relative error was calculated no closer than 5 cm to the surfaces between the regions. Both the potential and electric field relative errors begin to
26
27 (a)
(b)
Figure 4.10 (a) Electric potential vs. position and (b) electric field vs. position for various
basis orders p and conductivity σ1 =σ2 =4 S/m and meshed using quadrilateral cells of
28 (a)
(b)
Figure 4.11: Relative error vs. maximum edge length for (a) the electric potential and (b)
29 (a)
(b)
Figure 4.12 (a) Electric potential vs. position and (b) electric field vs. position for various
basis orders p and conductivity σ1 =6 S/m, σ2 =4 S/m, and σ3 =2 S/m and meshed
30 (a)
(b)
Figure 4.13: Relative error vs. maximum edge length for (a) the electric potential and (b)
the electric field for various basis orders p and conductivity σ1 =6 S/m, σ2 =4 S/m,
31
4.4 Insulated Box containing Sphere Electrolyte Interface
Consider a rectangular box centered vertically along the z-axis with a length of 2 m and side lengths of 1 m with a fictitious surface at the center of the structure, as shown in Figure 4.14. A spherical fictitious surface is centered in the structure with a radius of 0.25
m. Electrodes are placed at the top z =2 1 m and bottom of the rectangular box z = −0 1 m
, sourcing and sinking 1 A, respectively. The surfaces on the sides of the structure are
insulating material. The structure was analyzed using basis orders p =0,1,2.
In Figure 4.15 (a) and (b) are plotted the computed and analytic electric potential and electric field, respectively, along the z-axis. The electrolyte filling the boxes has a
conductivity of σ1 =σ2 =4 S/m. The results match well with the analytic solution but have
some error in the electric field when observing the fields closer to the fictious sphere. This may be because the field on the curved surfaces of the sphere varies more rapidly which makes the approximation more inaccurate as the field point approaches a surface of the sphere. The error decreases significantly as the basis order is increased.
In Figure 4.16 (a) and (b) are plotted the computed and analytic electric potential and electric field, respectively, along the z-axis. In this case, the bottom box and the bottom
half of the sphere are filled with an electrolyte of conductivity σ1 =2 S/m and the top box
along with the top half of the sphere are filled with an electrolyte of conductivity σ2 =3
S/m. The results are in good agreement with the analytic solution in (4.1) and (4.2). The results also show the same type of error as discussed in Figure 4.15 (a) and (b) with the electric field approximation losing accuracy as it approaches the curved surfaces of the sphere at a low basis order.
32
Figure 4.14: A two-meter long brick with two homogeneous electrolytes and a spherical electrolyte interface.
33 (a)
(b)
Figure 4.15: (a) Electric potential vs. position and (b) electric field vs. position for
various basis orders p and conductivity σ1 =σ2 =4 S/m and meshed using quadrilateral
34 (a)
(b)
Figure 4.16: (a) Electric potential vs. position and (b) electric field vs. position for
various basis orders p and conductivity σ1=2 S/m, and σ2 =3 S/m and meshed using
35 4.5 Box with Linearly Varying Conductivity
Consider a cubic box aligned vertically with the z-axis and with sides of length 1
m. The bottom of the box is at z =0 m and the top of the box is at z =1 m. Electrodes
are placed at the top of the box and the bottom of the box, sourcing and sinking, respectively, 1 A. The surfaces on the sides are insulating. The electrolyte has a linear
conductivity profile σ
( )
r =σb+(
σ σt− b)
z where σb =1 S/m and σt =2 S/m. Theaverage potential over the bottom electrode is forced to be zero.
The analytic electric potential and electric field inside the box are
( )
(
tIT b)
ln ( )bz S σ σ σ σ Φ = − r (4.3) and( )
IT( )
ˆ Sσ z = − E r z , (4.4)respectively, where IT is the electrode current and S is the cross-sectional area of the
box.
The cubic box is approximated by being divided evenly along z into N boxes where
n n
z N
= for n= locates the surfaces bounding the regions in z. The conductivity 0, ,N
of the nth box is σn =σ
(
rcentern)
which is the average of the actual conductivity in the nthbox and where rcentern is the center of the nth box for n= . A set of 5 models, 1, ,N
illustrated in Figure 4.17, was analyzed for N = . 1, ,5
In Figure 4.18 (a) and (b) are plotted the computed and analytic electric potential and
electric field, respectively, along the z-axis with N = regions. The electrolyte filling the 1
box has a conductivity of σ1 =1.5 S/m. The results do not converge well to the analytic
solution since a single, homogeneous electrolyte is a poor approximation of the linear conductivity profile. The electric field can only be piecewise-constant, thus it does not represent the analytic solution well.
In Figure 4.19 (a) and (b) are plotted the computed and analytic electric potential and
36
conductivities of σ1=1.25 S/m and σ2 =1.75 S/m. The results are converging to the
analytic solution. The electric field can only be piecewise-constant, thus it does not represent the analytic solution well.
In Figure 4.20 (a) and (b) are plotted the computed and analytic electric potential and
electric field, respectively, along the z-axis with N = regions. The electrolytes filling the 3
boxes have conductivities of σ1=1.166 S/m, σ2 =1.5 S/m, and σ3 =1.833 S/m. The
results are converging nicely, but the electric field can only be piecewise-constant, thus it does not represent the analytic solution well.
In Figure 4.21 (a) and (b) are plotted the computed and analytic electric potential and
electric field, respectively, along the z-axis with N = regions. The electrolytes filling the 4
boxes have conductivities of σ1=1.125 S/m, σ2 =1.375 S/m, σ3 =1.625 S/m, and
4 1.875 S/m
σ = . The results are converging nicely, but the electric field can only be
piecewise-constant, thus it does not represent the analytic solution well.
Finally, in Figure 4.22 (a) and (b) are plotted the computed and analytic electric
potential and electric field, respectively, along the z-axis withN = regions. The 5
electrolytes filling the boxes with conductivities of σ1 =1.1 S/m, σ2 =1.3 S/m,
3 1.5 S/m
σ = , σ4 =1.7 S/m, and σ5 =1.9 S/m. The results are converging nicely, but the
electric field can only be piecewise-constant, thus it does not represent the analytic solution well.
37
(a) (b)
(c) (d)
(e)
38 (a)
(b)
Figure 4.18: (a) Electric potential vs. position and (b) electric field vs. position for N = 1
sub-boxes and various basis orders p and meshed using quadrilateral cells of edge length 0.25 m.
39 (a)
(b)
Figure 4.19: (a) Electric potential vs. position and (b) electric field vs. position for N = 2
sub-boxes and various basis orders p and meshed using quadrilateral cells of edge length 0.25 m.
40 (a)
(b)
Figure 4.20: (a) Electric potential vs. position and (b) electric field vs. position for N = 3
sub-boxes and various basis orders p and meshed using quadrilateral cells of maximum edge length 0.25 m.
41 (a)
(b)
Figure 4.21: (a) Electric potential vs. position and (b) electric field vs. position for N = 4
sub-boxes and various basis orders p and meshed using quadrilateral cells of maximum edge length 0.25 m.
42 (a)
(b)
Figure 4.22: (a) Electric potential vs. position and (b) electric field vs. position for N = 5
sub-boxes and various basis orders p and meshed using quadrilateral cells of maximum edge length 0.25 m.