1,376 232 5MB
Pages 414 Page size 433.44 x 695.28 pts Year 2009
SURFACE IMPEDANCE BOUNDARY CONDITIONS A Comprehensive Approach
SURFACE IMPEDANCE BOUNDARY CONDITIONS A Comprehensive Approach
Sergey V. Yuferev Nathan Ida
Boca Raton London New York
CRC Press is an imprint of the Taylor & Francis Group, an informa business
CRC Press Taylor & Francis Group 6000 Broken Sound Parkway NW, Suite 300 Boca Raton, FL 33487-2742 © 2010 by Taylor and Francis Group, LLC CRC Press is an imprint of Taylor & Francis Group, an Informa business No claim to original U.S. Government works Printed in the United States of America on acid-free paper 10 9 8 7 6 5 4 3 2 1 International Standard Book Number: 978-1-4200-4489-8 (Hardback) This book contains information obtained from authentic and highly regarded sources. Reasonable efforts have been made to publish reliable data and information, but the author and publisher cannot assume responsibility for the validity of all materials or the consequences of their use. The authors and publishers have attempted to trace the copyright holders of all material reproduced in this publication and apologize to copyright holders if permission to publish in this form has not been obtained. If any copyright material has not been acknowledged please write and let us know so we may rectify in any future reprint. Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying, microfilming, and recording, or in any information storage or retrieval system, without written permission from the publishers. For permission to photocopy or use material electronically from this work, please access www.copyright. com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users. For organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged. Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. Library of Congress Cataloging-in-Publication Data Yuferev, Sergey V. Surface impedance boundary conditions : a comprehensive approach / authors, Sergey V. Yuferev, Nathan Ida. p. cm. “A CRC title.” Includes bibliographical references and index. ISBN 978-1-4200-4489-8 (hardcover : alk. paper) 1. Electromagnetic waves--Scattering--Mathematical models. 2. Radio wave propagation--Mathematical models. 3. Electromagnetic surface waves--Mathematical models. 4. Impedance (Electricity)--Mathematical models. 5. Boundary value problems. I. Ida, Nathan. II. Title. QC665.S3Y84 2010 537.6’2--dc22 Visit the Taylor & Francis Web site at http://www.taylorandfrancis.com and the CRC Press Web site at http://www.crcpress.com
2009026274
To our parents
Authors Sergey Yuferev was born in St. Petersburg, Russia, in 1964. He received his MSc in computational fluid mechanics from St. Petersburg Technical University, St. Petersburg, in 1987, and his PhD in computational electromagnetics from the A.F. Ioffe Institute, St. Petersburg, in 1992. From 1987 to 1998, he worked at with the Dense Plasma Dynamics Laboratory, A.F. Ioffe Institute. From 1999 to 2000, he was a visiting associate professor at the University of Akron, Akron, Ohio. Since 2000, he has been with the Nokia Corporation, Tampere, Finland. His current research interests include numerical and analytical methods of computational electromagnetics and their application to electromagnetic compatibility and electromagnetic interference problems of mobile phones. Nathan Ida is currently a distinguished professor of electrical and computer engineering at the University of Akron, Akron, Ohio. He teaches electromagnetics, antenna theory, electromagnetic compatibility, sensing and actuation, and computational methods and algorithms. His current research interests include numerical modeling of electromagnetic fields, electromagnetic wave propagation, theoretical issues in computation, and nondestructive testing of materials at low and microwave frequencies as well as in communications, especially in low-power remote control and wireless sensing. He has published extensively on electromagnetic field computation, parallel and vector algorithms and computation, nondestructive testing of materials, surface impedance boundary conditions, and other topics. He is the author of three books and co-author of a fourth. Dr. Ida is a fellow of the IEEE and the American Society of Nondestructive Testing.
vii
Preface The idea for this book grew out of work on surface impedance boundary conditions (SIBCs) over a period of nearly 12 years. We felt that the concept of surface impedance, especially at a low frequency, had been neglected to a large extent. When it was used, it was often viewed as an esoteric issue and incorporated in computation as an ad hoc means of simplifying the treatment for specific problems. In addition, the use of surface impedance was limited almost exclusively to the Leontovich concept—that of a first-order SIBC. No attempt has been made to formalize the concept of SIBC or to extend it to higher orders. This book intends to remedy this state of affairs by providing a comprehensive, consistent, and thorough approach to the subject. The book outlines not only the need for SIBCs, but also offers a simple, systematic method of construction of SIBCs of any order based on a perturbation approach. As one of the first tasks, we develop a socalled surface impedance toolbox, which allows the reader to develop SIBCs for any application. The formulation of the SIBC within common numerical techniques, such as the boundary integral equations method, the finite element method, and the finite difference method, is discussed in detail and is elaborated with specific examples. We felt that one particular reason to shun SIBCs was the fact that their implementation usually required extensive modification of existing software. To mitigate this problem, we developed our own SIBCs so that they could be incorporated within existing software without having to modify the software’s structure. The variousorder SIBCs are built by adding terms in an expansion so that increasing the order of approximation only requires the addition of terms without modification of the existing terms. Although the development of the various SIBCs and their formulations is rigorous and detailed, the emphasis in this book is on the practical aspects. Therefore, we include examples for all formulations, and discuss in detail the conditions of applicability and the errors to be expected from the inclusion of SIBCs in formulations. A practical set of guidelines based on the physical dimensions and properties of the problem is outlined to allow the reader to first decide if an SIBC can benefit his or her calculation and then to evaluate, a priori, the maximum errors he or she can expect by incorporating the SIBC. This book has been in the making for over 4 years, but its roots can be traced back to its modest beginnings nearly 12 years ago when the authors first began their association. Over the years, others, including graduate students, have contributed to this book, but we wish to acknowledge the
ix
x
Preface
contribution of Professor Luca Di Rienzo from the Polytecnica di Milano, in particular. Many of the results reported here were derived from his work on implementation of the various formulations. The value of this book would have been greatly diminished without his contributions. Sergey V. Yuferev Nathan Ida
Contents Introduction .......................................................................................................... xv 1. Classical Surface Impedance Boundary Conditions ................................. 1 1.1 Introduction ............................................................................................ 1 1.2 Skin Effect Approximation ................................................................... 2 1.3 SIBCs of the Order of Leontovich’s Approximation......................... 8 1.4 High-Order SIBCs ................................................................................ 10 1.4.1 Mitzner’s Approach................................................................ 10 1.5 Rytov’s Approach ................................................................................ 17 1.5.1 General ..................................................................................... 17 1.5.2 Calculation of the Field inside the Conductor ................... 18 1.5.3 Boundary Conditions at the Conductor Surface................ 21 1.5.4 Particular Case of a Planar Interface.................................... 22 1.5.5 Notes on Applicability of the Method................................. 23 References......................................................................................................... 25 2. General Perturbation Approach to Derivation of Surface Impedance Boundary Conditions ............................................................... 27 2.1 Introduction .......................................................................................... 27 2.2 Local Coordinates ................................................................................ 28 2.3 Perturbation Technique ....................................................................... 35 2.4 Tangential Components ...................................................................... 41 2.5 Normal Components ........................................................................... 45 2.6 Normal Derivatives.............................................................................. 49 2.7 Components of the Curl Operator ..................................................... 52 2.8 Surface Impedance ‘‘Toolbox’’ Concept ........................................... 57 2.9 Numerical Example ............................................................................. 65 ~ ~ f ~g ........................ 73 ~ g , , and Appendix 2.A.1: Calculation of ~ 3 2 2 h jk h References......................................................................................................... 76 3. SIBCs in Terms of Various Formalisms .................................................... 79 3.1 Introduction .......................................................................................... 79 3.2 Basic Equations..................................................................................... 79 3.3 Electric Field–Magnetic Field Formalism ......................................... 80 3.4 Magnetic Scalar Potential Formalism................................................ 89 3.5 Magnetic Vector Potential Formalism............................................... 95 3.6 Common Representation of Various SIBCs Using a Surface Impedance Function.............................................. 102 3.7 Surface Impedance near Corners and Edges ................................. 105 References....................................................................................................... 113 xi
xii
Contents
4. Calculation of the Electromagnetic Field Characteristics in the Conductor’s Skin Layer................................................................... 115 4.1 Introduction ........................................................................................ 115 4.2 Distributions across the Skin Layer................................................. 116 4.3 Resistance and Internal Inductance ................................................. 127 4.4 Forces Acting on the Conductor ...................................................... 131 5. Derivation of SIBCs for Nonlinear and Nonhomogeneous Problems ........................................................................................................ 135 5.1 Introduction ........................................................................................ 135 5.2 Coupled Electromagnetic–Thermal Problems................................ 136 5.3 Magnetic Materials............................................................................. 143 5.4 Nonhomogeneous Conductors ........................................................ 156 5.4.1 PEC-Backed Lossy Dielectric Layer ................................... 156 5.4.2 Two-Layer Conducting Structure....................................... 159 References....................................................................................................... 165 6. Implementation of SIBCs for the Boundary Integral Equation Method: Low-Frequency Problems .......................................................... 167 6.1 Introduction ........................................................................................ 167 6.2 Two-Dimensional Problems ............................................................. 169 6.2.1 E–H Formalism...................................................................... 169 6.2.2 A–K Formalism...................................................................... 176 6.2.3 Common Representation ..................................................... 182 6.3 Three-Dimensional Problems ........................................................... 187 6.4 Properties of the Surface Impedance Function .............................. 193 6.5 Boundary Element Formulations for Two- and Three-Dimensional Problems in Invariant Form ........ 195 6.6 Numerical Examples.......................................................................... 204 6.7 Quasi-Three-Dimensional Integro-Differential Formulation for Symmetric Systems of Conductors..................... 209 References....................................................................................................... 218 7. Implementation of SIBCs for the Boundary Integral Equation Method: High-Frequency Problems ......................................................... 221 7.1 Introduction ........................................................................................ 221 7.2 Integral Representations of High-Frequency Electromagnetic Fields....................................................................... 222 7.3 SIBCs for Lossy Dielectrics ............................................................... 227 7.4 Direct Implementation of SIBCs into the Surface Integral Equations .............................................................................. 232 7.5 Implementation Using the Perturbation Technique...................... 237 7.6 Numerical Example ........................................................................... 248
Contents
xiii
Appendix 7.A.1: Efficient Evaluation of Time Convolution Integrals ................................................................................... 255 References....................................................................................................... 260 8. Implementation of SIBCs for Volume Discretization Methods......... 263 8.1 Introduction ........................................................................................ 263 8.2 Statement of the Problem.................................................................. 264 8.3 Finite-Difference Time-Domain Method......................................... 265 8.4 Finite Integration Technique............................................................. 280 8.5 Finite-Element Method ...................................................................... 290 Appendix 8.A.1: Basics of Contour-Path FDTD Method ........................ 294 References....................................................................................................... 296 9. Application and Experimental Validation of the SIBC Concept ....... 299 9.1 Introduction ........................................................................................ 299 9.2 Selection of the Surface Impedance Boundary Conditions for a Given Problem ...................................................... 299 9.2.1 Characteristic Values of the Problem................................. 300 9.2.2 Asymptotic Expansions ....................................................... 301 9.2.3 Methodology ......................................................................... 302 9.3 Experimental Validation of SIBCs ................................................... 315 9.3.1 Physical Configuration......................................................... 316 9.3.2 Example: Reconstruction of Currents from Measured Magnetic Fields................................................... 322 9.3.3 Example: Calculation of p.u.l. Parameters in Multiconductor Transmission Lines.............................. 328 References....................................................................................................... 335 Appendix A: Review of Numerical Methods ............................................. 337 A.1 Introduction ....................................................................................... 337 A.2 Finite-Element Method..................................................................... 338 A.2.1 Physical Equations.............................................................. 339 A.2.2 Discretization ...................................................................... 341 A.2.3 Approximation.................................................................... 342 A.2.4 Minimization ....................................................................... 345 A.2.5 Solution ................................................................................ 352 A.2.6 Postprocessing..................................................................... 353 A.3 Finite-Difference Time-Domain Method........................................ 353 A.4 Boundary-Element Method ............................................................. 356 References....................................................................................................... 362 Index .................................................................................................................... 367
Introduction
I.1 Concept of Surface Impedance In most electromagnetic problems, the domain under consideration consists of several different media. The electromagnetic field governing equations written for each region are linked by boundary conditions that involve values on both sides of the interface. Thus one has to solve the problem for all media simultaneously, even if the main interest is focused only on one of them. However, under certain conditions, the number of regions involved in the solution procedure may be reduced. A classical example is the elimination of a body of infinite conductivity—a perfect electrical conductor (PEC)—from the computational space by enforcing the tangential electric field or normal magnetic flux to be equal to zero at the boundary (the so-called PEC boundary condition): ~ n~ Bjinterface ¼ 0 n ~ Ejinterface ¼ 0; ~
(I:1)
In practice, any real material has finite conductivity so that the PEC is merely a model of a good conductor in which the skin depth is assumed to be zero. Although the PEC condition is very attractive for implementation, the diffusion of electromagnetic fields into the conductor may be neglected only in a limited number of practical cases. This means that the application of the PEC limit in practical designs is very limited. For example, the electromagnetic penetration depth d in copper at an incident frequency of 1 MHz is approximately 2 105 m. Is this skin layer thin or thick? Obviously, the question is meaningful only in relation to the dimensions of the media so that they may be compared. The characteristic size D of the conductor may be used for this comparison. In our example, the penetration depth in the conductor is definitely not small if the conductor’s thickness is equal to 5 105 m (typical thicknesses of conductors in printed circuit boards) and the PEC condition may not be applied in this case. One may expect the use of the PEC condition to lead to errors proportional to d=D. Since the PEC is the limiting case of a real conductor, it is only natural to expect that the PEC condition is also a particular case of a more general approximate boundary condition relating electromagnetic quantities at conductor=dielectric interfaces. The existence of approximate boundary conditions of this genre follows directly from Snell’s law of refraction; if the electromagnetic wave propagates from a low-conductivity medium to a high-conductivity medium, the refraction angle is about 908 and, in practical terms, does not depend on the incidence angle. Suppose the conducting xv
xvi
Introduction
region is so large that the wave attenuates completely inside the region. Then, the electromagnetic field distribution in the conductor’s skin layer can be described as a damped plane wave propagating in the bulk of the conductor normal to its surface. In other words, the behavior of the electromagnetic field in the conducting region may be assumed to be known a priori, as in the case of the PEC. The electromagnetic field is continuous across the conductor’s surface so that the intrinsic impedance of the wave remains the same at the interface. Therefore, the ratio Ex at the xy-plane of a dielectric=conductor interface is assumed to be equal to the intrinsic impedance of the plane wave propagating in the conductor, in the positive z-direction: sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Ex jvsource mcond 1þj ¼ vsource mcond d Hy interface scond þ jvsource econd 2 sve sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 d¼ vsource scond mcond
(I:2)
The relation in Equation I.2 does not depend on coordinates and, therefore, the surface impedance is assumed to be constant over the conductor’s surface. The ‘‘surface’’ relation in Equation I.2, taking into account parameters of the conductor’s material and the source, contains all the necessary information about the field distribution in the conductor’s ‘‘volume.’’ Thus, it may be used as a boundary condition to the governing equations for the dielectric space and, by doing so, excludes the conductor from the solution region. In contrast to Equation I.1, which is a Dirichlet boundary condition, Equation I.2 can be viewed as an additional equation relating different unknowns at the interface. This is the basic idea of the surface impedance concept.
I.2 Historical Perspective The roots of the surface impedance concept can be traced to the end of the nineteenth century, a time when the main computational tool of electrical engineers was circuit theory. The lumped circuit representation is based on the approximation that it is acceptable to assign electromagnetic processes such as energy supply, dissipation, and storage to individual components, concentrated virtually at a point (lumped) in space. A lumped circuit is considered to be a group of components (resistors, capacitors, and inductors) interconnected in a certain topology occupying no physical space; a signal propagates from one point to another without time delay. To denote the ratio of amplitudes of an electromotive force V and the alternating current I
xvii
Introduction
produced by it in the circuit comprised of a resistance R and an inductance L, Oliver Lodge introduced the term ‘‘impedance’’ in 1889 [18]. The instantaneous voltage–current relation in such a circuit is described by the following differential equation: L
dI þ RI ¼ V dT
(I:3)
In the steady-state case, Equation I.3 takes the form (R þ jvL)I ¼ V
(I:4)
where v is an angular frequency of the source. Thus, the impedance Z of the electric circuit is represented in the form: Zcirc ¼ V=I ¼ R þ jvL
(I:5)
In real physical systems, energy storage and dissipation are mixed together and distributed over relatively large areas. The lumped component representation is applicable if the physical size of the problem is much smaller than the wavelength (electrically small system) and the propagation time of disturbances is negligible compared to their period. However, it is also possible to use a ‘‘circuit’’ approach in devices that do not satisfy this condition, such as transmission lines. The space variation of the complex voltage V across and the electric current I in a transmission line (let it be directed along the x-axis) is described by the following well-known equations: dV=dx ¼ ZI
(I:6)
dI=dx ¼ YV
(I:7)
where the coefficients of proportionality Z and Y are known as the distributed ‘‘series impedance’’ and the distributed ‘‘shunt admittance’’ of the line. In a generalized transmission line, Z and Y may be functions of x and may depend on frequency. If Z and Y are independent of x, they depend upon the distributed series resistance R, the shunt conductance G, the series inductance L, and the shunt capacitance C in the following manner: Z ¼ R þ jvL;
Y ¼ G þ jvC
(I:8)
Equations I.6 through I.8 can be combined to solve for V(x) and I(x): d2 V ¼ g2 V dx2
(I:9)
xviii
Introduction
d2 I ¼ g2 I dx2
(I:10)
where g ¼ a þ jb ¼
pffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ZY ¼ (R þ jvL)(G þ jvC)
(I:11)
is the propagation constant whose real and imaginary parts, a and b, are the attenuation and phase constants of the line, respectively. Solutions of Equations I.9 and I.10 represent waves propagating in the þx- and x-directions with a certain finite velocity maintaining a fixed phase between V and I: V(x) ¼ V þ (x) þ V (x) ¼ V0þ exp (gx) þ V0 exp (gx)
(I:12)
I(x) ¼ I þ (x) þ I (x) ¼ I0þ exp (gx) þ I0 exp (gx)
(I:13)
From Equation I.12, one obtains dV ¼ gV dx
(I:14)
Substituting Equation I.14 in Equations I.6 and I.7, the characteristic impedance of the line, that is, the ratio of voltage and current, is obtained: R þ jvL g ¼ ¼ Z0 ¼ g G þ jvC
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi R þ jvL G þ jvC
(I:15)
Transmission line theory describes wave propagation phenomena. For example, Equations I.6, I.7, I.9, and I.10 are of the same form as the equations governing the propagation of plane electromagnetic waves with ~ E ¼~ ay E and ~ ¼~ H az H in a lossy medium in the x-direction: dE ¼ jvmH dx
(I:16)
dH ¼ (s þ jve)H dx
(I:17)
d2 E ¼ g2 E dx2
(I:18)
d2 H ¼ g2 H dx2
(I:19)
xix
Introduction
where the propagation constant g takes the form: pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi g ¼ jvm(s þ jve) ¼ jv
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi s me 1 þ jve
(I:20)
Obviously, the solutions of Equations I.18 and I.19 have the same form as Equations I.12 and I.13: E(x) ¼ Eþ 0 exp (gx) þ E0 exp (gx)
(I:21)
H(x) ¼ H0þ exp (gx) þ H0 exp (gx)
(I:22)
Since even the physical meanings of E and H are closely related to those of V and I (E is V per unit length and H is I per unit length), the terms of impedance are naturally extended to the electromagnetic wave theory and applications. By analogy with Equation I.15, the quantity E Z0 ¼ ¼ H
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi jvm jvm g ¼ ¼ s þ jve g s þ jve
(I:23)
has been called the ‘‘intrinsic impedance’’ of the medium. In the general case Z0 is complex, but in the particular case of a lossless medium it is real. The intrinsic impedance will frequently occur as a multiplier in expressions for the impedances of various types of waves. Practical applications of impedance in power engineering, transmission lines, shielding, etc., have been analyzed by Schelkunoff, who introduced the term ‘‘impedance concept’’ in 1938 to generalize the idea of impedance [6]. Schelkunoff also recognized analogies between circuit impedance and similar ratios of major quantities in other engineering disciplines where the term ‘‘impedance’’ is also common (such as mechanics) or not so common (such as hydrodynamics and thermodynamics). For example, the well-known onedimensional equation describing linear oscillations of a mass in a damping medium is written in the form: m
d(n exp (iwt)) þ r(n exp (iwt)) ¼ F exp (iwt) dt
where n is the velocity F is the applied force m is the mass r is the resistance coefficient
(I:24)
xx
Introduction
This equation is similar to Equation I.3, so that mechanical impedance, that is, the ratio ‘‘force=velocity,’’ can be represented in the form: Zmech ¼ r þ iwm
(I:25)
Another example is the propagation of a heat wave that is governed in the one-dimensional case by the following equations: @T v ¼ @x K
(I:26)
@v @T ¼ cr @x @t
(I:27)
where T is the temperature v is the rate of heat flow K is the thermal conductivity r is the density c is the specific heat For simple periodic waves, Equations I.26 and I.27 are reduced to dT v ¼ dx K
(I:28)
dv ¼ jvcrT dx
(I:29)
Thus, the characteristic impedance and the propagation constant of heat waves are Zheat
1 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi ; jvcrK
rffiffiffiffiffiffiffiffiffi jvcr g¼ K
(I:30)
In the 1930s, the newly evolving radio technology required the development of the theory of propagation of electromagnetic waves of an antenna over the Earth’s surface. An analytical solution for the particular case of a vertical dipole radiating over a conducting half-space has been obtained by Sommerfeld [4], but the more general problem involving layered media separated by curved interfaces has not been solved so far. Leontovich [7,8] and Schukin [9] (independently of one another) proposed a different approach, namely, to restrict the general problem by considering only the air region using the surface impedance boundary condition (SIBC) at the air=Earth interface. Of course, the Earth is not a good conductor in the common sense. On the other
Introduction
xxi
hand, the characteristic dimensions in this class of electromagnetic problems are sufficiently large for attenuation of the wave in the Earth, so that the surface impedance method may be applied. The considerations of both Leontovich and Schukin are based on the assumption that the variation of the field along the surface is small compared to the variation inside the Earth. Thus the field derivatives in the directions tangential to the surface may be neglected compared to the normal derivative, and the original twodimensional or even three-dimensional equation of the field diffusion into the conductor is reduced to a one-dimensional problem that yields the relationship in Equation I.2. This is frequently referred to as the skin effect approximation [10]. Note that the same approximation is at the root of the theory of boundary layers in fluid mechanics. Schukin published this derivation in 1940 [9]. Leontovich published his results without derivation later, in 1944 [7]. However, there is evidence that Leontovich actually came to the idea of approximate boundary conditions at the end of the 1930s, at the same time as Schukin. The first rigorous mathematical analysis was done by Rytov, who wrote the following in his memoirs [2]: In 1939 he (Leontovich) posed the following problem to me: calculate the electromagnetic field inside a good conductor near the surface, i.e. develop an approximate, but general theory for strong skin effect. This problem had the following history. Probably in 1938, Leontovich proposed his now well-known boundary conditions for the electromagnetic field on the surface of good conductors. The meaning of these conditions is that they allow solutions to problems of diffraction of electromagnetic waves outside conducting regions without consideration of the field inside the conductors, greatly simplifying the solution of the problem. Although the correctness of the boundary conditions was beyond all doubt, Leontovich was not completely satisfied because the questions concerning accuracy and applicability of the conditions as well as that of acceptable curvature of the conductor’s surface were unresolved. In other words, he wanted to have validation of his boundary conditions and estimation of their accuracy. However, when he raised the problem of the skin effect theory, he told me nothing about the approximate boundary conditions (he probably wanted to test my capacity at guesswork). I have developed the theory for the general case of an arbitrarily directed incident wave and double curved surface of the conductor. The theory certainly gave the approximate boundary conditions as the skin depth is reduced, but I missed this fact and so I failed the exam on guesswork.
Rytov’s fundamental contribution is the development of the perturbation approach to the problem of the field inside and outside the conductor [12]. He sought a solution in the form of a power series in the small parameter proportional to the ratio d=D. The first-order terms of the expansions
xxii
Introduction
(actually, first nonzero terms) gave Leontovich’s condition. Thus, an improvement of the Leontovich condition can be obtained by the inclusion of the next higher-order terms of the expansions. Rytov also stated the problems of calculation of the surface impedance at curved interfaces and nonhomogeneous conductors. Unfortunately, even though [12] has been translated into French [13], Rytov’s contribution is not as well known as Leontovich’s work. One may think that the time for approximate analytical approaches passed with the advent of robust numerical methods and fast computers. However, computational electromagnetics is called upon to solve and analyze problems in structures so vast and complex that the cliche ‘‘one basic truth to life is that demand will fill up all available resources!’’ [14] has never sounded truer. Because of this, the surface impedance concept is gaining acceptance as a technique that allows significant savings in computer resources and improved solutions. The Leontovich SIBC is efficiently used in combination with such numerical methods as the finite-element method (FEM) [26–38,79,87–90,108,112,117–182], the boundary-integral equation method [39–56], and the finite-difference time-domain method (FDTD) [57–77]. Among practical applications of the surface impedance concept are transformers [30,33,79–81,166,180], waveguides [82–87], inductive heating devices [88–92], microstrip and transmission lines [44,93–98], electric machines [99–101], electromagnetic scattering [49,65,102–107], electromagnetic compatibility [108–111,160], nondestructive testing analysis [112–115], plasma and magnetic levitation devices [32,116,117], electromagnetic casting [111,112], and many others, including applications outside the sphere of electromagnetics.
I.3 New Developments The growing number of applications was the main impetus to the further development of the concept in the following basic directions: 1. Extensions of the planar SIBCs to curved surfaces and edges. Indeed, the Leontovich SIBC can be applied as long as the smallest radius of curvature of the surface is much larger than the wavelength inside the conductor. Practical important steps have been taken by Mitzner [15], who developed SIBCs, known by his name, for any smooth surface conductors by introducing terms of the order of d2 that allow for the conductor’s curvature. Although Mitzner has derived the SIBCs in his own way, calculation of the second-order terms in Rytov’s expansions leads to the same result. A significantly different situation occurs in the vicinity of a conducting edge where the
Introduction
xxiii
magnetic field distribution is singular and the diffusion may not be described by a one-dimensional equation. Semiempirical attempts have been made to modify Leontovich’s condition near corners and edges [142–152,194], but the only rigorous solution was obtained using the perturbation approach [153,154]. First, the field distribution in free space surrounding the edge is calculated with the PEC condition (zero-order solution). Then this distribution along the interface is used as a boundary condition to the two- or threedimensional problem for the field inside the real conductor and its asymptotic solution is used to derive the first-order SIBC near the edge. 2. Time-domain SIBCs. The concept of surface impedance can be applied to transient problems, when, for instance, the current pulse duration is so short that the field has no time to diffuse deep into the conductor and remains concentrated near the surface. There are two basic approaches to solve transient problems: (a) by obtaining the solution in the frequency domain for a time-harmonic exciting source and then using the inverse Fourier transform to calculate the required transient data, and (b) by formulating the problem directly in the time domain. The time-domain form of the Leontovich condition can be obtained from Equation I.1 using the Laplace transform. High-order timedomain SIBCs have been obtained using the perturbation approach [21,23–25,55,56,76,77,174]. The choice between time- and frequencydomain forms of the SIBC is often governed by the method used for solution of the problem (for example, time-domain SIBCs are naturally suited for FDTD methods). Due to the frequency-dependent nature of the SIBC, it contains a convolution integral in the time domain. Direct computation of this convolution integral requires large computer resources and, therefore, approximate algorithms have been developed for efficient implementations of SIBCs in the time domain [57–75,155–164]. 3. SIBCs for nonlinear media. Often, the magnetic or thermal properties of the conductor’s material may not be approximated by constants. Traditionally the B(H) or s(T) curves are approximated by an analytical function exploiting features of the problem under consideration and then inserted into the frequency-domain Leontovich condition [90,165–172]. However, in the general case, neither magnetic nor electric fields in the conductor with nonlinear properties are actually time harmonic and the use of a single-frequency SIBC is not feasible. Thus, formulating the problem in the time domain seems to be a more general approach for both linear and nonlinear problems [173,174]. In this case, the SIBC is not obtained as the ratio of the tangential fields at the interface. Instead, it is formulated in the form of a onedimensional nonlinear equation that must be solved numerically.
xxiv
Introduction
On the other hand, time-domain nonlinear SIBCs are usually applicable to any isotropic homogeneous material since they are not restricted to any special approximation of the material property curve. 4. SIBCs for nonhomogeneous (composite) structures. The most important case in practice is a conductor coated by a layer of dielectric material. If the nonconducting layer is sufficiently thin compared to the wavelength, the skin layer approximation will still be valid. SIBCs of different orders of approximation for this class of problems have been developed and given the name generalized impedance boundary conditions (GIBCs) [198–207]. As a result of these activities, a wide variety of SIBCs have been developed and are commonly used in computational electromagnetics.
I.4 Purpose and Method of Presentation The concept of this book grew out of a need to present to the practicing engineer as well as to students a comprehensive, coherent exposition of the ideas, applications, and the general issues involved with the use of SIBCs. In particular, the authors felt that SIBCs are viewed by the computational electromagnetics community as somewhat esoteric and as an ad hoc solution to specific problems. Furthermore, most users of the method unnecessarily restrict themselves to the Leontovich condition. Yet, the concept is much more general and simpler than most potential users suspect, and the authors felt it necessary to expose these aspects of the method. The purpose of this book is to help the practitioner in the understanding, selection, use, and development of impedance boundary conditions for his=her own applications. In the process, the theoretical arguments and mathematical relations needed for the proper definition of SIBCs are developed. However, the main emphasis of the book is on the practical. It provides the engineer with a practical and versatile tool for low-penetration problems in computational electromagnetics. The book provides the following specific benefits: 1. Decision. The book provides guidance on how to decide if the surface impedance concept is applicable to a particular problem and if the problem can benefit from the use of the surface impedance concept. 2. Selection. The choice of the most appropriate SIBC (if one exists) is important both for accuracy and implementation. This issue is discussed based on performance and accuracy, and simple guidelines to the selection process are provided.
Introduction
xxv
3. Development. When an SIBC appropriate for a specific application does not exist, one can be developed based on the methods discussed in this book. 4. Implementation. For an SIBC to be truly useful, its implementation in existing code or in new code development should be simple and efficient. A number of formulations of SIBCs in conjunction with well-known numerical techniques are discussed. Structure of the book The discussion of SIBCs starts with the Leontovich condition in Chapter 1. The Leontovich condition is derived starting with the electric and magnetic fields, following the classical approach. This serves as an introduction to the more general methods of deriving SIBCs to be taken up in subsequent chapters. It also emphasizes the steps and ideas needed to develop SIBCs for specific applications. The assumptions made in the development of SIBCs and the conditions under which these SIBCs may be used are critical to their successful use. These assumptions and conditions are introduced in Chapter 1, which also discusses the Mitzner and Rytov methods as an introduction to the elaborations and developments to follow. Chapter 2 introduces the reader to the general idea of building an SIBC as an asymptotic expansion based on a perturbation approach. This method, due to Rytov, is a general tool for the development of SIBCs and is particularly useful for the development of SIBCs of high order of approximation. Following the introduction and discussion of Rytov’s formula, a unified time- and frequency-domain approach to the development of SIBCs is introduced. The chapter outlines a general surface impedance ‘‘toolbox,’’ that is applicable universally and is later used repeatedly in following chapters. SIBCs in terms of various formalisms are introduced in Chapter 3. In addition to the electric field–magnetic field formalisms, we also discuss the magnetic scalar potential and the magnetic vector potential formalisms as well as introduce common representations in terms of surface impedance functions. Finally, we discuss the issue of surface impedance near corners and edges. An important issue in the use of SIBCs is the surface layer itself. Chapter 4 discusses these issues and shows how distributions across the skin layer may be calculated in the context of the formulations of Chapter 3. Chapter 5 tackles the implementation of SIBCs for nonlinear and nonhomogeneous applications. We treat coupled electromagnetic–thermal formulations as well as magnetic materials. In Chapter 6, we treat the implementation of SIBCs for the low-frequency boundary equation method. Starting with two-dimensional implementations, we treat in particular the electric field–magnetic field formulation and then introduce the surface current concept followed by the magnetic vector potential–surface current formulation in the context of boundary
xxvi
Introduction
integral formulations. The concept of common representations discussed first in Chapter 3 is again invoked to extend the implementation to other formulations. Three-dimensional problems are then discussed, followed by a discussion of the surface impedance function and the development of invariant formulations in two and three dimensions. Chapter 7 continues the subject of implementation of SIBCs for boundary integral equations, but this time we focus on high-frequency problems. Because of the assumption of high frequencies, SIBCs for lossy dielectrics can now be used. Implementations using direct methods and perturbation techniques are discussed. Although SIBCs are most readily used with boundary integral methods, they can be successfully implemented in volume discretization methods. We take up this subject in Chapter 8 and discuss implementation within the finite-difference time-domain (FDTD), finite-integration technique (FIT), and the finite-element method (FEM). Chapter 9 presents two important issues. The first is the selection of appropriate SIBCs for a given problem. Starting from the characteristic values of the physical problem, we develop a simple methodology that tells the user the appropriate order of the SIBC and estimates the maximum errors involved. Then we look at experimental validation of SIBCs in two distinct types of applications: one is a simple comparison of computed and measured data; the other is the problem of reconstruction of currents from measured field data. These emphasize not only the accuracy expected from SIBCs, but also the range of applications of the method. Additional applications dealing with multiconductor transmission lines look into the issues of applicability and accuracy of SIBCs. The book concludes with an appendix discussing representative numerical methods including the FEM, the boundary-element method, and the FDTD method. This appendix only touches on the methods, but it does outline the basic steps so that the user may get a general idea of the issues involved. A concise list of references given may also help in guiding the user to further study.
References History 1. C. Pelosi and P. Ya. Ufimtsev, The impedance-boundary condition, IEEE Antennas and Propagation Magazine, 38(1), 1996, 31–35. 2. S.M. Rytov, In the oscillations laboratory, in Memoires to Academician Leontovich, Moscow, Nauka, 1996, pp. 40–66 (in Russian). 3. J.R. Wait, The ancient and modern history of EM ground-wave propagation, IEEE Antennas and Propagation Magazine, 40(5), 1998, 7–24.
Introduction
xxvii
4. A. Ishimaru, J.D. Rockway, Y. Kuga, and S-W. Lee, Sommerfeld and Zenneck wave propagation for a finitely conducting one-dimensional rough surface, IEEE Transactions on Antennas and Propagations, 48(9), 2000, 1475–1484 (also rough surfaces).
General Theory 5. Lord Rayleigh, On the self inductance and resistance of straight conductors, Philosophical Magazine, 21, 1886, 381–394. 6. S.A. Shelkunoff, The impedance concept and its application to problems of reflection, radiation, shielding and power absorption, Bell System Technical Journal, 17, 1938, 17–48. 7. M.A. Leontovich, On one approach to a problem of the wave propagation along the Earth’s surface, Academy of Science USSR, Series Physics, 8, 1944, 16–22. 8. M.A. Leontovich, On the approximate boundary conditions for the electromagnetic field on the surface of well conducting bodies, in B.A. Vvedensky, Ed., Investigations of Radio Waves, Academy of Sciences of USSR, Moscow, 1948. 9. A.N. Shchukin, Propagation of Radio Waves, Svyazizdat, Moscow, 1940. 10. G.S. Smith, On the skin effect approximation, American Journal of Physics, 58(10), 1990, 996–1002. 11. H.G. Booker, The elements of wave propagation using the impedance concept, Journal of the Institute of Electrical Engineers, 3(94), 1947, 171–184. 12. S.M. Rytov, Calculation of skin effect by perturbation method, Journal Experimental’noi i Teoreticheskoi Fiziki, 10(2), 1940, 180–189 (in Russian). 13. S.M. Rytov, Calcul du skin-effet par la méthode des perturbations, Journal of Physics, 2(3), 1940, 233–242. 14. C. Brench, TC-9 computational EMC, IEEE EMC Society Newsletter, 186, 2000, 13. 15. K.M. Mitzner, An integral equation approach to scattering from a body of finite conductivity, Radio Science, 2(12), 1967, 1459–1470. 16. C.A. Balanis, Advanced Engineering Electromagnetics, John Wiley & Sons, New York, pp. 210–213. 17. J. van Bladel, Electromagnetic Fields, Appendix 2, McGraw-Hill Book Co., Inc., New York, 1964. 18. O. Lodge, On lightning, lightning conductors, and lightning protectors, Electrical Review, May 1889, 518. 19. J.R. Wait, The scope of impedance boundary conditions in radio propagation, IEEE Transactions on Geoscience Remote Sensing, 28(4), 1990, 721–723. 20. Z. Godzinski, The surface impedance concept and the theory of radio waves over real earth, Proceedings of the IEE, 108C, March 1961, 362–373. 21. S. Yuferev and N. Ida, Time domain surface impedance boundary conditions of high order of approximation, IEEE Transactions on Magnetics, 34(5), 1998, 2605–2608. 22. S. Yuferev and N. Ida, Selection of the surface impedance boundary conditions for a given problem, IEEE Transactions on Magnetics, 35(3), 1999, 1486–1489. 23. S. Yuferev, N. Ida, and L. Kettunen, Invariant BEM-SIBC formulations for timeand frequency-domain eddy current problems, IEEE Transactions on Magnetics, 36(4), 2000, 852–855 (also use with numerical methods: integral equations). 24. S. Yuferev and N. Ida, Time domain surface impedance concept for low frequency electromagnetic problems-Part I: Derivation of high order surface impedance
xxviii
Introduction
boundary conditions in the time domain, IEE Proceedings on Science, Measurement and Technology, 152(4), 2005, 175–185. 25. S. Barmada, L. Di Rienzo, N. Ida, and S. Yuferev, Time domain surface impedance concept for low frequency electromagnetic problems. II. Application to transient skin and proximity effect problems in cylindrical conductors, IEE Proceedings on Science, Measurement and Technology, 152(5), 2005, 207–216.
Use of SIBCs with Numerical Methods FEM 26. E.M. Deeley and J. Xiang, Improved surface impedance methods for 2-D and 3-D problems, IEEE Transactions on Magnetics, 24(1), 1988, 209–211. 27. J. Sakellaris, G. Meunier, A. Raizer, and A. Darcherif, The impedance boundary condition applied to the finite element method using the magnetic vector potential as state variable: A rigorous solution for high frequency axisymmetric problems, IEEE Transactions on Magnetics, 28(2), 1992, 1643–1646. 28. A.M. El-Sawy Mohamed, Finite-element variational formulation of the impedance boundary condition for solving eddy current problems, IEE Proceedings on Science, Measurement and Technology, 142(4), 1995, 293–298. 29. M. Gyimesi, J.D. Lavers, T. Pawlak, and D. Ostergaard, Impedance boundary condition for multiply connected domains with exterior circuit conditions, IEEE Transactions on Magnetics, 30(5), 1994, 3056–3059. 30. J. Sakellaris, G. Meunier, X. Brunotte, C. Guerin, and J.C. Sabonnadiere, Application of the impedance boundary condition in a finite element environment using the reduced potential formulation, IEEE Transactions on Magnetics, 27(6), 1991, 5022–5024 (also applications: transformers). 31. S. Subramaniam, M. Feliziani, and S.R. Hoole, Open boundary eddy-current problems using edge elements, IEEE Transactions on Magnetics, 2(2), 1992, 1499–1503. 32. M. Enokizono and T. Todaka, Approximate boundary element formulation for high-frequency eddy current problems, IEEE Transactions on Magnetics, 29(2), 1993, 1504–1507 (also applications: magnetic levitation devices). 33. S.A. Holland, G.P. O’Connel, and L. Haydock, Calculating stray losses in power transformers using surface impedance with finite elements, IEEE Transactions on Magnetics, 28(2), 1992, 1355–1358 (also applications: transformers). 34. D. Rodger, P.J. Leonard, H.C. Lai, and P.C. Coles, Finite element modeling of thin skin depth problems using magnetic vector potential, IEEE Transactions on Magnetics, 33(2), 1997, 1299–1301. 35. G. Pan and J. Tan, General edge element approach to lossy and dispersive structures in anisotropic media, IEE Proceedings—Microwave, Antennas, Propagation, 144(2), 1997, 81–90. 36. I. Mayergoyz and C. Bedrosian, On finite element implementation of impedance boundary conditions, Journal of Applied Physics, 75, 1994, 6027–6029. 37. J. Gyselinck, P. Dular, C. Geuzaine, and R.V. Sabariego, Surface-impedance boundary conditions in time-domain finite-element calculations using the magneticvector-potential formulation, IEEE Transactions on Magnetics, 45(3), 2009, 1280–1283. 38. S. Yuferev and L. Kettunen, Implementation of high order surface impedance boundary conditions using vector approximating functions, IEEE Transactions on Magnetics, 36(4), 2000, 1606–1609.
Introduction
xxix
Integral Equations 39. S. Subramaniam and S.R. Hoole, The impedance boundary condition in the boundary element vector potential formulation, IEEE Transactions on Magnetics, 4, November 1988, 2503–2505. 40. A. Nicolas, 3D eddy current solution by BIE techniques, IEEE Transactions on Magnetics, 24(1), 1988, 130–133. 41. M.T. Ahmed, J.D. Lavers, and P.E. Burke, On the use of the impedance boundary condition with an indirect boundary formulation, IEEE Transactions on Magnetics, 24(6), 1988, 2512–2514. 42. M.T. Ahmed, J.D. Lavers, and P.E. Burke, Direct BIE formulation for skin and proximity effect calculation with and without the use of the surface impedance approximation, IEEE Transactions on Magnetics, 25, September 1989, 3937–3939. 43. K. Ishibashi, Eddy current analysis by boundary element method utilizing impedance boundary condition, IEEE Transactions on Magnetics, 31(3), 1995, 1500–1503. 44. T.E. van Deventer, P. Katehi, and A. Cangellaris, An integral equation method for the evaluation of conductor and dielectric losses in high-frequency interconnects, IEEE Transactions on Microwave Theory and Techniques, 37(12), 1989, 1964–1972—also applications: microstrip interconnects (Ohmic losses due to the finite conductivity of the strips constitute the prelevant loss effect at microwave and millimeter-wave frequencies). 45. H. Tsuboi, K. Sue, and K. Kunisue, Surface impedance method using boundary elements for exterior regions, IEEE Transactions on Magnetics, 27(5), 1991, 4118–4121. 46. A.A. Kishk and R.K. Gordon, Electromagnetic scattering from conducting bodies of revolution coated with thin magnetic materials, IEEE Transactions on Magnetics, 30(2), 1994, 3152–3155. 47. A.H. Chang, K.S. Yee, and J. Prodan, Comparison of different integral equation formulations for bodies of revolution with anisotropic surface impedance boundary conditions, IEEE Transactions on Antennas and Propagation, 40(8), 1992, 989–991 (also non-homogeneous structures). 48. A. Bendali, M’B. Fares, and J. Gay, A boundary-element solution of the Leontovich problem, IEEE Transactions on Antennas and Propagation, 47(10), 1597–1605, October 1999. 49. P.L. Huddleston, Scattering by finite, open cylinders using approximate boundary conditions, IEEE Transactions on Antennas and Propagation, 37(2), 253–257, 1989 (also applications: scattering). 50. O. Sterz and C. Schwab, A scalar boundary integrodifferential equation for eddy current problems using an impedance boundary condition, Computing and Visualization in Science, 3, 209–217, 2001. 51. H.C. Jayatilaka and I.R. Ciric, Performance of surface impedance integral equations for quasi-stationary field analysis in axisymmetric systems, IEEE Transactions on Magnetics, 40(2), 2004, 1350–1353. 52. G.W. Hanson, On the applicability of the surface impedance integral equation for optical and near infrared copper dipole antennas, IEEE Transactions on Antennas and Propagation, 54(12), 2006, 3677–3685. 53. S. Yuferev and V. Yuferev, Calculation of the electromagnetic field diffusion in a system of parallel conductors, Soviet Physics Technical Physics Journal, 36, 1991, 1208–1211.
xxx
Introduction
54. S. Yuferev, Generalized BIE-BLA formulation of skin and proximity effect problems for an arbitrary regime of the magnetic field generation, Second International IEE Conference on Computation in Electromagnetics, IEE Conference Publication No. 384, 1994, pp. 307–310. 55. S. Yuferev and N. Ida, Application of approximate boundary conditions to electromagnetic transient scattering problems, Third International IEE Conference on Computation in Electromagnetics, IEE Conference Publication No. 420, 1996, pp. 51–56. 56. S. Yuferev and N. Ida, Efficient implementation of the time domain surface impedance boundary conditions for the boundary element method, IEEE Transactions on Magnetics, 34(2), 1998, 2763–2766.
FDTD 57. J.A. Roden and S.D. Gedney, The efficient implementation of the surface impedance boundary condition in general curvilinear coordinates, IEEE Transactions on Microwave Theory and Techniques, 47(10), 1999, 1954–1963. 58. J.G. Maloney and G.S. Smith, The use of surface impedance concepts in the finitedifference time-domain method, IEEE Transactions on Antennas and Propagation, 40(1), 1992, 38–48. 59. K.S. Yee, K. Shlager, and A.H. Chang, An algorithm to implement a surface impedance boundary condition for FDTD, IEEE Transactions on Antennas and Propagation, 40(7), 1992, 833–837. 60. C.F. Lee, R.T. Shin, and J.A. Kong, Time domain modeling of impedance boundary condition, IEEE Transactions on Microwave Theory and Techniques, 40(9), 1992, 1847–1851. 61. S. Kellahi and B. Jecko, Implementation of a surface impedance formalism at oblique incidence in FDTD method, IEEE Transactions on Electromagnetic Compatibility, 35(3), 1993, 347–356. 62. K.S. Oh and J.E. Schutt-Aine An efficient implementation of surface impedance boundary conditions for the finite-difference time-domain method, IEEE Transactions on Antennas and Propagation, 43(7), 1995, 660–666. 63. C.F. Lee, R.T. Shin, and J.A. Kong, Time domain modeling of impedance boundary condition, IEEE Transactions on Microwave Theory and Techniques, 40(9), 1992, 1847–1851. 64. J.H. Beggs, R.J. Lubbers, K.S. Yee, and K.S. Kunz, Finite-difference time-domain implementation of surface impedance boundary conditions, IEEE Transactions on Antennas and Propagation, 40(1), 1992, 49–56. 65. C.W. Penney, R.L. Luebbers, and J.W. Schuster, Scattering from coated targets using a frequency-dependent, surface impedance boundary condition in FDTD, IEEE Transactions on Antennas and Propagation, 44(4), 1996, 434–443 (also applications: scattering). 66. K.S. Yee and J.S. Chen, Impedance boundary condition simulation in the FDTD=FVTD hybrid, IEEE Transactions on Antennas and Propagation, 45(6), 1997, 921–925. 67. J.H. Beggs, A FDTD surface impedance boundary condition using Z-transform, Applied Computational Electromagnetics Society Journal, 13(3), 1998, 14–24.
Introduction
xxxi
68. M.K. Karkkainen and S.A. Tretiakov, Finite-difference time domain model of interfaces with metals and semiconductors based on a higher order surface impedance boundary conditions, IEEE Transactions on Antennas and Propagation, 51(9), 2003, 2448–2455. 69. M.K. Karkkainen, FDTD surface impedance model for coated conductors, IEEE Transactions on Electromagnetic Compatibility, 46(2), 2004, 222–233 (also non-homogeneous layered structures). 70. P.G. Petropoulos, Approximating the surface impedance of a homogeneous lossy half-space: An example of ‘‘dialable’’ accuracy, IEEE Transactions on Antennas and Propagation, 50(7), 2002, 941–943. 71. R. Makinen, An efficient surface-impedance boundary condition for thin wires of finite conductivity, IEEE Transactions on Antennas and Propagation, 52(12), 2004, 3364–3372. 72. Q. Zeng, A method for efficient computation of time domain surface impedances of a lossy half space, Proceedings 20th Zurich International Symposium on Electromagnetic Compatibility, 2009, 29–32. 73. T. Kashiwa, O. Chiba, and I. Fukai, A formulation for surface impedance boundary conditions using the finite-difference time-domain method, Microwave Optical Technology Letter, 5(10), 1992, 486–490. 74. T. Ohtani, K. Taguchi, T. Kashiwa, Y. Kanai, and J.B. Cole, Scattering analysis of large-scale coated cavity using the complex nonstandard FDTD method with surface impedance boundary condition, IEEE Transactions on Magnetics, 45(3), 2009, 1296–1299. 75. M. Feliziani, F. Maradei, and G. Tribellini, Field analysis of penetrable conductive shields by the finite-difference time-domain method with impedance network boundary conditions (INBC’s), IEEE Transactions on Electromagnetic Compatibility, 41(4), 1999, 307–319. 76. S. Yuferev, N. Farahat, and N. Ida, Use of the perturbation technique for implementation of surface impedance boundary conditions for the FDTD method, IEEE Transactions on Magnetics, 36(4), 2000, 942–945. 77. N. Farahat, S. Yuferev, and N. Ida, High order surface impedance boundary conditions for the FDTD method, IEEE Transactions on Magnetics, 37(5), 2001, 3242–3245.
Hybrid Formulations 78. H. Wang, M. Xu, C. Wang, and T. Hubing, Impedance boundary conditions in a hybrid FEM=MOM formulation, IEEE Transactions on Electromagnetic Compatibility, 45(2), 198–205, 2003.
Applications Transformers 79. C. Guerin, G. Tanneau, and G. Meunier, 3D eddy current losses calculation in transformer tanks using the finite element method, IEEE Transactions on Magnetics, 29(2), 1993, 1419–1422 (also FEM).
xxxii
Introduction
80. L. Lin and C. Xiang, Losses calculation in transformer tie plate using the finite element method, IEEE Transactions on Magnetics, 34(5), 1998, 3644–3647. 81. Y. Higuchi and M. Koizumi, Integral equation method with surface impedance model for 3D eddy current analysis in transformers, IEEE Transactions on Magnetics, 36(4), 2000, 774–779 (also Integral Equations).
Waveguides 82. J.D. Wade and R.H. Macphie, Conservation of complex power technique for waveguide junctions with finite wall conductivity, IEEE Transactions on Microwave Theory and Techniques, 38(4), 1990, 373–378. 83. A. Weisshaar, Impedance boundary method of moments for accurate and efficient analysis of planar graded-index optical waveguides, Journal of Lightwave Technology, 12(11), 1994, 1943–1951. 84. M.S. Islam, E. Tuncer, and D.P. Neikirk, Calculation of conductor loss in coplanar waveguide using conformal mapping, Electronic Letters, 29(13), 24 June 1993, 1189–1191. 85. K. Yoshitomi and H.R. Sharobim, Radiation from a rectangular waveguide with a lossy flange, IEEE Transactions on Antennas and Propagation, 42(10), 1994, 1398–1403. 86. K.A. Remley and A. Weisshaar, Impedance boundary method of moments with extended boundary conditions, Journal of Lightwave Technology, 13(12), 1995, 2372–2377 (digital-type pulse is considered; looks very interesting). 87. J. Tan and G. Pan, A new edge element analysis of dispersive waveguide structures, IEEE Transactions on Microwave Theory and Techniques, 43(11), 1995, 2600–2607 (also FEM).
Induction Heating Systems 88. S.M. Mimoune, J. Fouladgar, A. Chentouf, and G. Develey, A 3D impedance calculation for an induction heating system for materials with poor conductivity, IEEE Transactions on Magnetics, 32(3), 1996, 1605–1608 (also coupled electrical= thermal problems and FEM). 89. W. Mai and G. Henneberger, Field and temperature calculations in transverse flux inductive heating devices heating non-parametric materials using surface impedance formulations for non-linear eddy-current problems, IEEE Transactions on Magnetics, 35(3), 1999, 1590–1593 (also coupled electrical=thermal problems and FEM). 90. J. Nerg and J. Partanen, A simplified FEM based calculation model for 3-D induction heating problems using surface impedance formulations, IEEE Transactions on Magnetics, 37(5), 2001, 3719–3722 (also non-linear materials). 91. F. Bioul and F. Dupret, Application of asymptotic expansions to model twodimensional induction heating systems. Part I: Calculation of electromagnetic field distribution, IEEE Transactions on Magnetics, 41(9), 2005, 2496–2505. 92. F. Bioul and F. Dupret, Application of asymptotic expansions to model twodimensional induction heating systems. Part II: Calculation of equivalent surface stresses and heat flux, IEEE Transactions on Magnetics, 41(9), 2005, 2506–2514.
Introduction
xxxiii
Transmission Lines 93. W. Thiel, A surface impedance approach for modeling transmission line losses in FDTD, IEEE Microwave and Guided Wave Letters, 10(3), 2000, 89–91. 94. S. Kim and D.P. Neikirk, Time domain multiconductor transmission line analysis using effective internal impedance, IEEE 6th Topical Meeting on Electronic Packaging, IEEE Cat. Number 97TH8318, San Jose, CA, October 27–29, 255–258 (EPE97_reprint.pdf).
Microstrip Structures 95. E.M. Deeley, Microstrip calculations using an improved minimum-order boundary element method, IEEE Transactions on Magnetics, 30(5), 1994, 3753–3756 (also BEM). 96. I.P. Theron and J.H. Cloete, On the surface impedance used to model the conductor losses of microstrip structures, IEE Proceedings on Microwave Antennas and Propagation, 142(1), 1995, 35–40. 97. L.L. Lee, W.G. Lyons, T.P. Orlando, S.M. Ali, and R.S. Withers, Full-wave analysis of superconducting microstrip lines on anisotropic substrates using equivalent surface impedance approach, IEEE Transactions on Microwave Theory and Techniques, 41(12), 1993, 2359–2367. 98. D. De Zutter and L. Knockaert, Skin effect modeling based on a differential surface admittance operator, IEEE Transactions on Microwave Theory and Techniques, 53(8), 2005, 2526–2538.
Electrical Machines 99. K. Adamiak, G.E. Dawson, and A.R. Eastham, Application of impedance boundary conditions in finite element analysis of linear motors, IEEE Transactions on Magnetics, 27(6), 1991, 5193–5195 (also FEM). 100. V.C. Silva, Y. Marechal, and A. Foggia, Surface impedance method applied to the prediction of eddy currents in hydrogenerator stator end regions, IEEE Transactions on Magnetics, 31(3), 1995, 2072–2075 (also FEM). 101. J.F. Eastham, D. Rodger, H.C. Lai, and H. Nouri, Finite element calculation of fields around the end region of a turbine generator test rig, IEEE Transactions on Magnetics, 29(2), 1993, 1415–1418.
Scattering 102. A. Sebak and L. Shafai, Scattering from arbitrarily-shaped objects with impedance boundary conditions, IEE Proceedings-H, 136(5), 1989, 371–376. 103. M. Osuda and A. Sebak, Scattering from lossy dielectric cylinders using a multifilament current model with impedance boundary conditions, IEE Proceedings-H, 139(5), 1992, 429–434. 104. T.B.A. Senior and J.L. Volakis, Generalized impedance boundary conditions in scattering, Proceedings of the IEEE, 79(10), 1991, 1413–1420. 105. A.A. Kishk, Electromagnetic scattering from composite objects using mixture of exact and impedance boundary conditions, IEEE Transactions on Atennas and Propagation, 39(6), 826–833, 1991 (also Integral Equations).
xxxiv
Introduction
106. R.D. Graglia, P.L.E. Uslenghi, R. Vitiello, and U. D’Elia, Electromagnetic scattering for oblique incidence on impedance bodies of revolution, IEEE Transactions on Antennas and Propagation, 43(1), 1995, 11–26. 107. L.N. Medgyesi-Mitschang and J.M. Putnam, Integral equation formulations for imperfectly conducting scatters, IEEE Transaction on Antennas and Propagation, 33(2), 1985, 206–214 (also integral equations).
EMC 108. A. Darcherif, A. Raizer, J. Sakellaris, and G. Meunier, On the use of the surface impedance concept in shielded and multiconductor cable characterization by the finite element method, IEEE Transactions on Magnetics, 2(2), 1992, 1446–1449 (also FEM). 109. M. D’Amore and M.S. Sarto, Theoretical and experimental characterization of the EMP-interaction with composite-metallic enclosures, IEEE Transactions on Electromagnetic Compatibility, 42(1), 2000, 152–163—also Time-domain, FDTD (FDTDSIBC formulation for EMC problems). 110. C. Yang and V. Jandhyala, A time-domain surface integral technique for mixed electromagnetic and circuit simulation, IEEE Transactions on Advanced Packaging, 28(4), 2005, 745–753. 111. F. Maradei and M. Raugi, Analysis of upsets and failures due to ESD by the FDTD-INBCs method, IEEE Transactions on Industry Applications, 38(4), 2002, 1009–1017.
Inverse Problems and NDT 112. Z. Badics, Y. Matsumoto, K. Aoki, and F. Nakayasu, Effective probe response calculation using impedance boundary condition in eddy current NDE problems with massive conducting regions present, IEEE Transactions on Magnetics, 32(3), 1996, 737–740 (also FEM). 113. H. Sahinturk, On the reconstruction of inhomogeneous surface impedance of cylindrical bodies, IEEE Transactions on Magnetics, 40(2), 2004, 1152–1155 (also surface impedance measurement and modeling). 114. R. Grimberg, H. Mansir, J.L. Muller, and A. Nicolas, A surface impedance boundary condition for 3-D non destructive testing modeling, IEEE Transactions on Magnetics, MAG-22(5), September 1986, 1272–1274. 115. T.H. Fawzi, M.T. Ahmed, and P.E. Burke, On the use of the surface impedance boundary conditions in eddy current problems, IEEE Transactions on Magnetics, 21(5), 1985, 1835–1840.
Plasma Devices 116. F.-Z. Louai, D. Benzerga, and M. Feliachi, A 3D finite element analysis coupled to the impedance boundary condition for the magnetodynamic problem in radiofrequency plasma devices, IEEE Transactions on Magnetics, 32(3), 1996, 812–815. 117. A. Helaly, E.A. Soliman, and A.A. Megahed, Electromagnetic waves scattering by non-uniform plasma cylinder, IEE Proceedings of the Microwave Antennas and Propagation, 144(2), 1997, 61–66.
Introduction
xxxv
Electromagnetic Casting 118. M.R. Ahmed and J.D. Lavers, Boundary element analysis of the electromagnet casting mould, IEEE Transactions on Magnetics, 25(4), 1989, 2843–2845. 119. B. Dumont and A. Gagnoud, 3D finite element method with impedance boundary condition for the modeling of molten metal shape in electromagnetic casting, IEEE Transactions on Magnetics, 36(4), 2000, 1329–1332.
New Developments Nonhomogeneous Layered Structures 120. S.N. Karp and F.C. Karal, Generalized impedance boundary conditions with applications to surface wave structures, in J. Brown, Ed., Electromagnetic Wave Theory, Pt. 1, Pergamon, New York, 1965, pp. 479–483. 121. J.R. Wait, Electromagnetic Waves in Stratified Media, Pergamon, New York, 1970. 122. D.J. Hoppe and Y. Rahmat-Samii, Higher order impedance boundary conditions applied to scattering by coated bodies of revolution, IEEE Transactions on Antennas and Propagation, 42(12), 1994, 1600–1610. 123. T.B.A. Senior, Approximate boundary conditions, IEEE Transactions on Antennas and Propagation, 29, September 1981, 826–829. 124. T.B.A. Senior and J. LO. Volakis, Derivation and application of a class of generalized impedance boundary conditions, IEEE Transactions on Antennas and Propagation, 37, December 1989, 1566–1572. 125. D.J. Hoppe and Y. Rahmat-Samii, Impedance Boundary Conditions in Electromagnetics, Taylor & Francis, Washington, DC, 1994. 126. T.B.A. Senior and J. LO. Volakis, Approximate Boundary Conditions in Electromagnetics (IEE Electromagnetic Waves Series), IEE, London, 1995. 127. R. Cicchetti, A class of exact and higher order surface boundary conditions for layered structures, IEEE Transactions on Antennas and Propagation, 44(2), 1996, 249–259. 128. R. Cicchetti and A. Faraone, Exact impedance=admittance boundary conditions for complex geometries: Theory and applications, IEEE Transactions on Antennas and Propagation, 48(2), 2000, 223–230. 129. O. Marceaux and B. Stupfel, High-order impedance boundary conditions for multiplayer coated 3-D objects, IEEE Transactions on Antennas and Propagation, 48(3), 2000, 429–436. 130. S.M. Apollonskii and V.T. Erofeenko, Impedance boundary conditions for anisotropic bodies, IEE Proceedings on Microwave Antennas and Propagation, 142(2), 1995, 89–92. 131. T.B.A. Senior, J.L. Volakis, and S.R. Lagault, High order impedance and absorbing boundary conditions, IEEE Transactions on Antennas and Propagation, 45(1), 1997, 107–114 (links between SIBCs and AbsoBC; also on application of Rytov’s approach to inhomogeneous bodies).
Nonplanar Interfaces Smooth Curved Boundaries 132. K.M. Mitzner, An integral equation approach to scattering from a body of finite conductivity, Radio Science, 2(12), 1967, 1459–1470.
xxxvi
Introduction
133. H. Ammari and S. He, Effective impedance boundary conditions for an inhomogeneous thin layer on a curved metallic surface, IEEE Transactions on Antennas and Propagation, 46(5), 1998, 710–715 (also non-homogeneous structures). 134. W.T. Shaw and A.J. Dougan, Curvature corrected impedance boundary conditions in an arbitrary basis, IEEE Transactions on Antennas and Propagation, 53(5), 2005, 1699–1705. 135. B.E. Fridman, Skin effect in massive conductors used in pulsed electrical devices: I. Electromagnetic field of massive conductors, Technical Physics, 47(9), 2002, 1112–1119.
Rough Surfaces 136. T.B.A. Senior, Impedance boundary conditions for imperfectly conducting surfaces, Applied Science Research B, 8, 1960, 418–436. 137. C.L. Holloway and E.F. Kuester, Impedance-type boundary conditions for a periodic interface between a dielectric and a highly conducting medium, IEEE Transactions on Antennas and Propagation, 48(10), 2000, 1660–1672 (application of SI concept to 2-D periodic highly conducting rough surface with small scale roughness; new EGIBC improving Leontovich BC for this kind of problems is developed (R)). 138. C.L. Holloway and E.F. Kuester, Power loss associated with conducting and superconducting rough interfaces, IEEE Transactions on Microwave Theory and Techniques, 48(10), 2000, 1601–1610 (also applications: power losses in superconductors). 139. A.A. Maradudin, The impedance boundary condition for a one-dimensional, curved, metal surface, Optics Communications, 103(3=4), 1993, 227–234 (in this paper the SIBC for a rough conducting surface is derived as function of position; for comments see previous paper by Holloway, etc.). 140. T.T. Ong, V. Celli, and A.A. Maradudin, The impedance of a curved surface, Optics Communications, 95(1), 1993, 1–4. 141. J.R. Poirier, A. Bengali, and P. Borderies, Impedance boundary conditions for the scattering of time-harmonic waves by rapidly varying surfaces, IEEE Transactions on Antennas Propagation, 54(3), 2006, 995–1005. - rough surfaces.
Corners and Edges 142. S.R.H. Hoole and J. Carpenter, Surface impedance models for corners and slots, IEEE Transactions on Magnetics, 21(5), 1985, 1841–1843. 143. Y.-G. Park, T.-K. Chung, H.-K. Jung, and S.-Y. Hahn, Three dimensional eddy current computation using the surface impedance method considering geometric singularity, IEEE Transactions on Magnetics, 31(3), 1995, 1400–1403. 144. E.M. Deeley, Surface impedance near edges and corners in 3D media, IEEE Transactions on Magnetics., 26(2), March 1990, 712–714. 145. E.M. Deeley, The use of modified surface impedance in a minimum-order boundary element method, IEEE Transactions on Magnetics, 26(5), 1990, 2762–2764. 146. J. Wang, J.D. Lavers, and P. Zhou, Modified surface impedance boundary conditions for 3D eddy current problems, IEEE Transactions on Magnetics, 29(2), March 1993, 1826–1829. 147. E.M. Deeley, Avoiding surface impedance modification in BE methods by singularity-free representations, IEEE Transactions on Magnetics, 28(5), 1992, 2814–2816.
Introduction
xxxvii
148. E. Tuncer and D.P. Neikirk, Efficient calculation of surface impedance for rectangular conductors, Electronic Letters, 29(4), 25 November 1993, 2127–2128. 149. W.E. Boyse and K.D. Paulsen, Accurate solutions of Maxwell’s equations around PEC corners and highly curved surfaces using nodal finite elements, IEEE Transactions on Antennas and Propagation, 45(12), 1997, 1758–1767. 150. E.M. Deeley, Improved impedance boundary condition for finite elements by cross-coupling at corners, IEEE Transactions on Magnetics, 30(5), 1994, 2881–2884. 151. K. Ishibashi, Eddy current analysis by boundary integral equation utilizing edge boundary condition, IEEE Transactions on Magnetics, 32, 1996, 832–835. 152. H.T. Anastassiu, D.I. Kaklamani, D.P. Economou, and O. Breinbjerg, Electromagnetic scattering analysis of coated conductors with edges using the method of auxiliary sources (MAS) in conjunction with the standard impedance boundary condition (SIBC), IEEE Transactions on Antennas and Propagation, 50(1), 2002, 59–66. 153. S. Yuferev, L. Proekt, and N. Ida, Surface impedance boundary conditions near corners and edges: Rigorous consideration, IEEE Transactions on Magnetics, 37(5), 2001, 3465–3468. 154. L. Proekt, S. Yuferev, I. Tsukerman, and N. Ida, Method of overlapping patches for electromagnetic computation near imperfectly conducting cusps and edges, IEEE Transactions on Magnetics, 38(2), 2002, 649–652.
Transients 155. F.M. Tesche, On the inclusion of loss in time-domain solutions of electromagnetic interaction problems, IEEE Transactions on Electromagnetic Compatibility, 32(1), 1990, 1–4 (also Integral equations). 156. S. Celozzi and M. Feliziani, Time domain finite element formulation of conductive regions, IEEE Transactions on Magnetics, 28(2), 1992, 1705–1710 (also FEM). 157. K.R. Davey and L. Turner, Prediction of transient eddy current fields using surface impedance methods, IEEE Transactions on Magnetics, 25(5), 1989, 4156–4158. 158. K.R. Davey and L. Turner, Transient eddy current analysis for generalized structures using surface impedances and the Fast Fourier Transform, IEEE Transactions on Magnetics, 26(3), 1990, 1164–1170. 159. S. Celozzi and M. Feliziani, Transient scattering problems solution by surface equivalent sources, IEEE Transactions on Magnetics, 30(5), 1994, 3148–3151. 160. M.S. Sarto, A new model for the FDTD analysis of the shielding performances of thin composite structures, IEEE Transactions on Electromagnetic Compatability, 41(4), 1999, 298–306 (also FDTD, applications: EMC). 161. D. Rodger and H.C. Lai, A surface impedance method for 3D time transient problems, IEEE Transactions on Magnetics, 35(3), 1999, 1369–1371 (also FEM). 162. S. Celozzi and M. Feliziani, Analysis of fast transient electromagnetic fields: A frequency dependent 2-D procedure, IEEE Transactions on Magnetics, 28(2), 1992, 1146–1149. 163. L. Di Rienzo, N. Ida, and S. Yuferev, Application of surface impedance concept to inverse problems of reconstructing transient currents, IEEE Transactions on Magnetics, 39(3), 2003, 1626–1629. 164. Y. O’Keefe, D. Thiel, and K. O’Keefe, Time-dependent surface impedance from sferics, IEEE Geoscience Remote Sensing Letters, 2(2), 2005, 104–107.
xxxviii
Introduction
Nonlinear Materials 165. L. Krahenbuhl, O. Fabregue, S. Wanser, M. De Sousa Dias, and A. Nicolas, Surface impedances, BIEM and FEM coupled with 1D non linear solutions to solve 3D high frequency eddy current problems, IEEE Transactions on Magnetics, 33(2), 1997, 1167–1172. 166. C. Guerin, G. Meunier, and G. Tanneau, Surface impedance for 3D non-linear eddy current problems—Application to loss computation in transformers, IEEE Transactions on Magnetics, 32(3), 1996, 808–811 (also applications: transformers). 167. F. Azzouz and M. Feliachi, Non-linear surface impedance taking account of thermal effect, IEEE Transactions on Magnetics, 37(5), 2001, 3175–3177 (also coupled problems). 168. F. Bertoncini, E. Cardelli, S. Di Fraia, and B. Tellini, Evaluation of surface impedance of hysteretic materials, IEEE Transactions on Magnetics, 39(3), 2003, 1369–1372. 169. I. Mayergoyz, Nonliner Diffusion of Electromagnetic Fields, Academic Press, New York, 1998. 170. A. Kost, J.P. Bastos, K. Miethner, and L. Janicke, Improvement of nonlinear impedance boundary conditions, IEEE Transactions on Magnetics, 38(2), 2002, 573–576. 171. B. Paya and C. Guerin, Magnetic field dependence of nonlinear surface impedance: Which field to choose? IEEE Transactions on Magnetics, 38(2), 2002, 585–588. 172. P. Dular, R.V. Sabariego, and L. Krähenbuhl, Subdomain perturbation finite element method for skin and proximity, IEEE Transactions on Magnetics, 44(6), 2008, 738–741. 173. S. Yuferev and L. Kettunen, A new boundary element technique to transient nonlinear low penetration problems of multiconductor systems, IEEE Transactions on Magnetics, 34(5), 1998, 2613–2616. 174. S.Yuferev and L. Kettunen, A unified time domain surface impedance concept for both linear and non-linear skin effect problems, IEEE Transactions on Magnetics, 35(3), 1999, 1454–1457 (also Coupled EM=thermal problems).
Coupled EM=Thermal Problems 175. J.P. Sturgess and T.W. Preston, An economic solution for 3-D electromagnetic and thermal eddy current problems, IEEE Transactions on Magnetics, 28(2), 1992, 1267–1270. 176. S. Yuferev and J. Yufereva, A new boundary element formulation of coupled electromagnetic and thermal problems for non-harmonic regimes of current passage, IEEE Transactions on Magnetics, 32(3), 1996, 1038–1041.
Thin-Wall Conductors and Shell Elements 177. K.R. Davey, Shell impedance conditions for steady state and transient eddy current problems, IEEE Transactions on Magnetics, 25(4), 1989, 3007–3009 (also transients; skin effect approximation). 178. D. Zheng and K.R. Davey, A boundary element formulation for thin shell problems, IEEE Transactions on Magnetics, 32(3), 1996, 675–677.
Introduction
xxxix
179. D. Rodger, P.J. Leonard, H.C. Lai, and R.J. Hill-Cottingham, Surface elements for modeling eddy currents in high permeability materials, IEEE Transactions on Magnetics, 27(6), 1991, 4995–4997 (also FEM). 180. C. Guerin, G. Tanneau, G. Meunier, P. Labie, T. Ngnegueu, and M. Sacotte, A shell element for computing 3D eddy currents—Application to transformers, IEEE Transactions on Magnetics, 31(3), 1995, 1360–1363 (also FEM; applications: transformers). 181. M. Feliziani and F. Maradei, Finite-difference time-domain modeling of thin shields, IEEE Transactions on Magnetics, 36(4), 2000, 848–851. 182. L. Krahenbuhl and D. Muller, Thin layers in electrical engineering. Example of shell models in analyzing eddy-currents by boundary and finite element methods, IEEE Transactions on Magnetics, 29(2), 1993, 1450–1455.
Motion 183. H.-Y. Pao and J.R. Wait, Electromagnetic induction and surface impedance in a half-space from overhead moving current system, IEEE Transactions on Antennas and Propagation, 48(9), 2000, 1301–1305. 184. D. Rodger, E. Melgoza, and H.C. Lai, A surface impedance method for moving conductors, IEEE Transactions on Magnetics, 36(4), 2000, 736–740.
Surface Impedance Measurements and Modeling 185. C. Wang, J.L. Drewniak, M. Li, J. Fan, J.L. Knighten, N.W. Smith, R. Alexander, and J. Huang, FDTD modeling of skin effect, Proceedings of 2002 3rd International Symposium On Electromagnetic Compatibility, 2002, 246–249. 186. I. Akduman and A. Yapar, Surface impedance determination of a planar boundary by the use of scattering data, IEEE Transactions on Antennas and Propagation, 49(2), 2001, 304–307. 187. D.V. Thiel, On measuring electromagnetic surface impedance—Discussions with professor James R. Wait, IEEE Transactions on Antennas and Propagation, 48(10), 2000, 1517–1520. 188. S.R.H. Hoole, T.H. Walsh, and G.H. Stevens, The slot impedance: Experimental verification, IEEE Transactions on Magnetics, 24(6), 1988, 3156–3158. 189. D.A. James, S.G. O’Keefe, and D.V. Thiel, Eddy current modeling using the impedance method for surface impedance profiling, IEEE Transactions on Magnetics, 35(3), 1999, 1107–1110. 190. D.V. Thiel and R. Mittra, Surface impedance modeling using the finite difference time domain method, IEEE Transactions on Geoscience Remote Sensing, 35(5), 1997, 1350–1356.
Limits and Applicability 191. B. Wagner, W. Renhart, and C. Magele, Error evaluation of surface impedance boundary conditions with magnetic vector potential formulation on a cylindrical test problem, IEEE Transactions on Magnetics, 44(6), 2008, 734–737.
xl
Introduction
192. S.R.H. Hoole, Experimental validation of the impedance boundary condition and a review of its limitations, IEEE Transactions on Magnetics, 25(4), 1989, 3028–3030 (also experimental validation). 193. D.-S. Wang, Limits and validity of the impedance boundary condition on penetrable surfaces, IEEE Transactions on Antennas and Propagation, 35(4), 1987, 453–457. 194. N.G. Alexopoules and G.A. Tadler, Accuracy of the Leontovich boundary condition for continuous and discontinuous surface impedance, Journal of Applied Physics, 46, 1975, 3326–3332. 195. N. Aymard, M. Feliachi, and B. Paya, An improved modified surface impedance for transverse electric problems, IEEE Transactions on Magnetics, 33(2), 1997, 1267–1270 (also corners and edges). 196. J.R. Wait, Electromagnetic surface impedance of an anisotropic cylinder, IEE Proceedings A, 138(3), 1991, 184–186. 197. I. Hanninen and K. Nikoskinen, Implementation of method of moments for numerical analysis of corrugated surfaces with impedance boundary condition, IEEE Transactions Antennas Propagation, 56(1), 2008, 278–281.
GBIC 198. T.B.A. Senior and J.L.O. Volakis, Approximate Boundary Conditions in Electromagnetics (IEE Electromagnetic Waves Series), IEE, London, 1995. 199. H.H. Syed and J.L. Volakis, High-frequency scattering by a smooth coated cylinder simulated with generalized impedance boundary conditions, Radio Science, 26, 1991, 1305–1314. 200. T.E. van Deventer and L.P.B. Katehi, Generalized boundary conditions with applications to submillimeter and optical waveguides, Radio Science, 31(6), 1996, 1407–1416. 201. T.B.A. Senior and J.L. Volakis, Derivation of a class of generalized boundary conditions, IEEE Transactions on Antennas and Propagation, AP-37, 1989, 1566–1572. 202. V. Galdi and I.M. Pinto, SDRA approach for higher-order impedance boundary conditions for complex multi-layer coatings on curved conducting bodies, Progress in Electromagnetics Research, PIER 24, 1999, pp. 311–335. 203. S.A. Tretyakov, Generalized impedance boundary conditions from isotropic multilayers, Microwave Optics and Technology Letters, 17(4), 1998, 262–265. 204. X. Antoine and H. Barucq, Approximation by generalized impedance boundary conditions of a transmission problem in acoustic scattering, Mathematical Modeling and Numerical Analysis, 39(5), 2005, 1041–1059. 205. H. Ammari and S. He, Effective impedance boundary conditions for an inhomogeneous thin layer on a curved metallic surface, IEEE Transactions on Antennas and Propagation, 46(5), 1998, 710–715. 206. T.B.A. Senior, Generalized boundary conditions for scalar fields, Journal of the Acoustic Society of America, 97, 1995, 3473–3477. 207. M.A. Ricoy and J.L Volakis, Generalized impedance boundary conditions for isotropic multilayers, Radio Science, 25(4), 1990, 391–405.
1 Classical Surface Impedance Boundary Conditions
1.1 Introduction Although in recent years, the surface impedance boundary condition (SIBC) concept has taken increasing importance both at low and at high frequencies, its origins are rather modest and are deeply rooted in the concept of skin depth and skin effect. Schelkunoff [1] is credited with coining the term and introducing the concept in the early 1930s, by analogy with circuit theory, as the ratio of the electric and magnetic field intensities. With the rapid advance of the theory of radio waves and propagation, the concept has been further developed for the analysis of wave propagation over the earth’s surface. The need for solutions for propagation of waves generated by antennas in the presence of layered, curved surfaces has led to what we refer here as the classical SIBCs. The troubled times before and during World War II seem to have been a catalyst in this development, as the need for these important solutions became more acute. The first systematic attempt to treat SIBCs was undertaken by Rytov [2,3] in 1939. His perturbation method approach allowed calculation of fields inside and outside conductors based on power series expansions, and encompassed the Leontovich condition as the first-order term in the expansion. Inclusion of higher-order terms allows treatment of curved boundaries as well as variations of the field on the surface of conductors. The method is not as well known as others, primarily because Rytov’s work was published in Russian and, with the exception of one paper [4] which was translated and published in French [3], none of his work has ever been translated. The work of Leontovich and the SIBC bearing his name are much better known and have found wide acceptance in the past as well as in contemporary work. The first-order SIBC developed by Leontovich [5,6] was published in 1940 but was in existence at least since 1938 [2]. Later, a modified form, which applies to curved surfaces, has been developed [6] and has been discussed (and corrected) by Mitzner [7]. Mitzner developed a second-order SIBC using an integral equations approach solving for the scattered field due to a general scatterer. Although Mitzner’s SIBC is equivalent to Rytov’s second-order SIBC and can in fact be developed based on the expansion method of Rytov, Mitzner’s contribution includes discussion of errors for his own and 1
2
Surface Impedance Boundary Conditions: A Comprehensive Approach
Leontovich’s conditions as well as limits of applicability and modifications for large skin depths. The SIBCs due to Leontovich, Mitzner, and Rytov, together with the limiting case of the perfect electric conductor (PEC), form what we call the classical SIBCs and are discussed in this chapter. The purpose is to show how these SIBCs were developed, discuss the limits on their applicability and in so doing to pave the way for additional, new developments to be taken up in subsequent chapters. We start with a short discussion of the skin effect and the skin effect approximation, followed by the Leontovich SIBC. The Mitzner SIBC is introduced as an example of higher-order SIBCs and also to point out to the method by which he arrives at the SIBC. The general approach of Rytov is discussed and used to analyze the conditions of applicability for low- and high-order SIBCs. In the following chapter, we will also show how Rytov’s approach can be used to obtain the Leontovich and Mitzner’s SIBCs as special cases of the more general higher-order SIBCs.
1.2 Skin Effect Approximation The basis of the surface impedance concept is intertwined with the concept of skin effect. The idea of skin effect is usually introduced in elementary electromagnetics through the concept of propagation of plane waves from free space into a good conductor, and dates back to 1886 when Lord Rayleigh introduced it [8]. This approach has the advantage of being both simple and, at the same time, quite general even if the approach to derivation hides more than it reveals as to the physical processes and approximations involved. For example, by assuming plane wave propagation, we limit ourselves to planar surfaces and it is not immediately clear how the skin depth applies to curved surfaces or to cylindrical bodies. Similarly, the skin effect is normally understood in the context of time-harmonic fields and its applicability and use in time-domain applications is not obvious. Other questions, including those related to the sources of the fields, will have to be clarified before we can use the idea of skin depth in any general sense. Nevertheless, we shall start with the classical skin depth derivation and then see how the results may be modified and what are the approximations and conditions needed for its universal applicability to derivation of SIBCs. At this point, we limit ourselves to time-harmonic fields. Treatment in the time domain will be taken up later in this book. Consider a source-free electromagnetic wave propagating in free space, perpendicular to the surface of a conducting half-space as shown in Figure 1.1. The source-free Maxwell’s equations inside the conductor may be written as
Classical Surface Impedance Boundary Conditions
Free space
3
Conductor
E
x
pˆ
z
y
H
FIGURE 1.1 Conducting half-space used to define skin depth. ^p shows the direction of propagation of the wave.
nˆ (ε0, μ0, σ = 0)
(ε, μ, σ)
~ r~ E ¼ jvmH
(1:1)
~ ¼ jvec~ rH E
(1:2)
~¼0 rD
(1:3)
~¼0 rH
(1:4)
where ec is the complex permittivity given by s ec ¼ e 1 j ve
(1:5)
where e is the dielectric constant of the medium (i.e., the ‘‘static’’ value of permittivity). By direct substitution, we obtain E jvm(s þ jve)~ E¼0 r2~
(1:6)
~ jvm(s þ jve)H ~¼0 r2 H
(1:7)
and
These equations are general and apply equally well to nonconducting media by setting s ¼ 0. However, there are two related assumptions that have been used so far. 1. Equations are source free. 2. Plane wave assumption implies the source to be at infinity. Now, assuming the conductor to be a good conductor, the loss tangent is high and we can write s ve
(1:8)
4
Surface Impedance Boundary Conditions: A Comprehensive Approach
With this, we have E jvms~ E¼0 r2~
(1:9)
~ jvmsH ~¼0 r2 H
(1:10)
Under the specific constraints assumed here, the wave is a transverse electromagnetic (TEM) wave whereby both the electric and magnetic field intensities are perpendicular to the direction of propagation and to each other (Figure 1.1). Assuming, arbitrarily that ~ E is in the x-direction, and propagation is in the z-direction, we can write @ 2 Ex jvmsEx ¼ 0 @z2
(1:11)
@ 2 Hy jvmsHy ¼ 0 @z2
(1:12)
The solution to these equations, assuming only forward propagation, is Ex ¼ E0 egz
(1:13)
where E0 is the electric field in the conductor at z ¼ 0 (i.e., at the surface) g is the propagation constant pffiffiffiffiffiffiffiffiffiffiffi g ¼ a þ jb ¼ jvms ¼ (1 þ j)
rffiffiffiffiffiffiffiffiffiffi vms 2
(1:14)
where a is the attenuation constant b is the phase constant We now define a quantity called the skin depth as 1 d¼ ¼ a
sffiffiffiffiffiffiffiffiffiffi 2 [m] vms
(1:15)
The electric field intensity inside the conductor then may be written as 1þj
Ex ¼ E0 e d z
(1:16)
We can, of course, write the solution for the magnetic field intensity as 1þj
H y ¼ H 0 e d z
(1:17)
5
Classical Surface Impedance Boundary Conditions
where H0 is the magnetic field intensity at z ¼ 0, since Equations 1.11 and 1.12 are identical in form. However, we can also write from Equation 1.1: @Ex ¼ jvmHy @z
!
Hy ¼
1þj j @Ex (1 þ j) j (1 j) 1þjz ¼ E 0 e d z ¼ e d E0 vm @z d vm vmd (1:18)
Now, from Equation 1.15, we note that v ¼ 2=msd2 and we have @Ex ¼ jvmHy @z
!
Hy ¼
(1 j)sd 1þjz e d E0 2
(1:19)
This, together with Equation 1.16, defines a wave impedance in the conductor as 1þj
Ex E0 e d z 2 1þj ¼ ¼ (1j)sd 1þj ¼ hc ¼ z Hy sd e d E0 (1 j)sd
[V]
(1:20)
2
Also, the wave number in the conductor is kc ¼
1þj d
(1:21)
We shall see shortly that Equation 1.20 can be, and often is, taken as a surface impedance. The foregoing development is simple but it also seems very restrictive in that we assumed plane waves, planar surfaces, and semi-infinite thickness. To relax these constraints and therefore to obtain a more universally applicable skin depth approximation, we proceed by assuming a curved surface, finite thickness, and a general field. Finite thickness: Taking now a conductor of thickness t, we can argue as follows: since the analysis above assumed only a forward propagating wave in the z-direction, the equivalent requirement is that within the conductor there is no reflection from the far surface. For a semi-infinite conductor, this is always satisfied. For practical purposes, t can be finite and not necessarily large as long as it is a few skin depths thick since the field decays to 1=e in one skin depth and because any reflection from the opposite surface travels the same distance t to arrive back at the interface between conductor and free space. Thus t must be larger than the skin depth and some have taken this as t > 2d [9]. A thickness of 2d would ensure that the amplitude of the reflected wave at the interface is about 1.8% of the launched wave. A thickness of 3d will reduce this to about 0.2%. Thus, the minimum thickness required
6
Surface Impedance Boundary Conditions: A Comprehensive Approach
Free space Conductor x nˆ
R min z
(ε, μ, σ)
y
(ε0, μ0, σ = 0)
FIGURE 1.2 Local system of coordinates to allow treatment of curved surfaces.
depends on the accuracy one needs, but it is only a very few skin depths for all but the most demanding applications. We can write this as follows: tmin > 2 3d
(1:22)
Curved surfaces: If the smallest radius of curvature on a surface is much larger than the skin depth, then we can assume that the surface is locally flat. Assuming now a local system of coordinates as in Figure 1.2, the considerations above apply unchanged. Thus the requirement is Rmin d
(1:23)
General field source: Finally, we now assume that a general field impinges on the surface in Figure 1.2, i.e., on a locally flat surface. This brings us back to Equations 1.9 and 1.10. In the local coordinates, we write Equation 1.9 as x~ Ex þ ~ y~ Ey þ ~ z~ Ez ) jvms(~ x~ Ex þ ~ y~ Ey þ ~ z~ Ez ) ¼ 0 r2 (~
(1:24)
First, it is important to note that the wave propagation inside the conductor is essentially perpendicular to the interface for any angle of incidence. This is best seen from Figure 1.3, where a wave impinges on the interface at angle ui and refracts inside the conductor at an angle ut. Assuming propagation from free space into a good conductor, we can write for the refraction angle [10]:
7
Classical Surface Impedance Boundary Conditions
Free space (ε0, μ0, σ = 0)
Conductor (ε, μ, σ)
pˆ t
θt θi
pˆ i y
Ei
sin ut ¼
where n1 is n2 is g1 is g2 is
the the the the
Et
x z
FIGURE 1.3 Incidence and refraction angles at a conducting interface.
pffiffiffiffiffiffiffiffiffiffi jv e0 m0 n1 g qffiffiffiffiffiffiffiffi sin ui 0 sin ui ¼ 1 sin ui ¼ n2 g2 (1 þ j) vm20 s
(1:25)
index of refraction in free space index of refraction in the conductor propagation constant in free space propagation constant in the conductor
Clearly, Equation 1.25 applies for any good conductor at any frequency for which the condition in Equation 1.8 is satisfied. Inside the conductor, the variation of the components of the electric and magnetic fields parallel to the surface is assumed to be small compared to the variation of the components perpendicular to the surface. That is, while the wave is quickly attenuated in directions perpendicular to the surface, there is no lateral propagation because of the result in Equation 1.25. This can be written as follows: 2 2 @ Ex @ E x ; @x2 @z2
2 2 @ Ey @ Ey ; @y2 @z2
2 2 @ Ex @ Ex ; @z2 @x@y
2 2 @ Ey @ Ey @z2 @x@y (1:26)
The same considerations apply to the components of the magnetic field intensity: 2 2 @ H x @ Hx ; @x2 @z2
2 2 @ H y @ H y ; @y2 @z2
2 2 @ Hx @ H x ; @z2 @x@y
2 2 @ H y @ Hy @z2 @x@y (1:27)
8
Surface Impedance Boundary Conditions: A Comprehensive Approach
Applying these conditions to the general equations in Equations 1.9 and 1.10 results in equations identical to Equations 1.11 and 1.12, which then lead again to the solution in Equation 1.13 and to the skin depth. Although we have used here the ideas of waves propagating in the conductor, Equations 1.11 and 1.12 are diffusion equations and therefore the dominant physical behavior of the fields in the conductor is one of diffusion perpendicular to the surface with high attenuation, characterized by a small value of the skin depth. We can now summarize the conditions under which a skin layer may be assumed: 1. Loss tangent is high—conduction effects dominate (Equation 1.8). 2. Conductor is thick with respect to the skin depth (Equation 1.22). 3. Minimum radius of curvature of the conductor is much larger than the skin depth (Equation 1.23).
1.3 SIBCs of the Order of Leontovich’s Approximation Leontovich, while not the first to derive or to use the SIBC that bears his name (it is most often associated with Rytov [3] and Shchukin [11]), was the first to publish its derivation and to discuss its limits of applicability and, since these results were available early on [5,6,11], it was only natural that the SIBC he espoused should be associated with his name. The basic assumptions needed to derive the SIBC are those we assumed in Section 1.2. If the loss tangent of a medium is high (i.e., a high loss dielectric or a conductor), then the wavelength inside the medium is small so that the conditions we discussed in Section 1.2 apply. Specifically, the skin depth is small and Snell’s law can be applied at the surface. In addition, the fields are required to vary slowly from point to point on the surface. Under these conditions, the field can be assumed to vary only in the direction normal to the surface of the medium. Although this implies a plane wave in the conducting medium, the wave is only approximately a plane wave in the sense that its radius of curvature is large compared to the skin depth. To summarize, the conditions under which the SIBC is derived are as follows: 1. Loss tangent of the material must be high. 2. As a consequence, the wavelength and skin depth inside the material are small. 3. Fields do not vary or vary slowly on the surface of the medium.
9
Classical Surface Impedance Boundary Conditions
4. As a consequence, the surface of the material must be flat or, equivalently, its smallest radius of curvature must be large compared to the skin depth. 5. Material is sufficiently thick so that no reflected fields across the interface from within the medium exist. 6. There are no sources within the medium. Although the original SIBC assumes a homogeneous medium, Leontovich also discuses corrections due to inhomogeneities as well as corrections due to finite curvatures. The latter will be discussed in the following section as well as in the more general derivation due to Rytov in Section 1.5. With these conditions, the fields inside the medium are related to each other as in any plane wave (as a first approximation) and both the electric field intensity and magnetic flux density are parallel to the surface of the conductor. Assuming a system of coordinates as in Figure 1.2, the fields inside the medium are related as rffiffiffiffiffi mc Hy ; Ex ¼ ec
rffiffiffiffiffi mc Ex Ey ¼ ec
(1:28)
where ec is the complex permittivity mc is the complex permeability of the medium Since the electric and magnetic fields are tangential and tangential fields are continuous across the interface, the electric and magnetic fields also exist in the space outside the lossy medium. Thus, Equation I.2 constitutes the SIBC known as the Leontovich SIBC. The condition in Equation 1.28 is often written in terms of the electric Km , as surface current density, ~ Ke , and the magnetic surface current density, ~ ~ n~ Ke Km ¼ Zc~
(1:29)
~ ~ ~ Ke ¼ ~ n H; Km ¼ ~ n~ E
(1:30)
where
and the system of coordinates is the same as in Figure 1.2. Leontovich [6] also analyzed the errors and conditions of applicability involved in the use of this SIBC. In Section 1.5, we discuss the limits of applicability of the method and, in the next section, also discuss the errors in the Leontovich approximation.
10
Surface Impedance Boundary Conditions: A Comprehensive Approach
1.4 High-Order SIBCs 1.4.1 Mitzner’s Approach Following the introduction of the Leontovich SIBC and its successful use for a variety of applications, as well as the more general work by Rytov, it became apparent that higher-order boundary conditions are both possible and desirable. Whereas the Leontovich SIBC assumes that the minimum radius of curvature is large compared to the skin depth (Equation 1.23), a correction for smaller radii of curvatures has been implicitly introduced by Rytov [3] and later given by Leontovich [6] and reported, with corrections, by Mitzner [7]. Furthermore, the method by Rytov extends to higher-order SIBCs, as we shall see in Section 1.5. Mitzner’s approach is a different way of extending the SIBC to smooth surfaces on which the curvature is not large compared to the skin depth. Unlike Rytov’s method, which uses asymptotic expansions, Mitzner [7] derives the SIBC from the magnetic field integral equation (MFIE) and the electric field integral equation (EFIE) on the smooth surface of a homogeneous conductor. We shall show in Chapter 2 that Mitzner’s SIBC is in fact identical to Rytov’s method in the sense that the same SIBC is obtained from Rytov’s method when the order of the expansion is 2. Nevertheless, Mitzner’s method is important in that it uses the physical process of scattering by a homogeneous object to derive the SIBC as well as to analyze its accuracy. This method of derivation also indicates the close relationship between SIBCs and integral equation methods and that the use of SIBCs is most natural in boundary integral methods of computation. The latter idea will be used in Chapters 6 and 7 to implement SIBCs for low- and highfrequency applications. On the other hand, Mitzner’s method is ad hoc and does not lend itself to generalization to higher-order SIBCs. Mitzner started with the general idea of integral equations for scattering by a conducting body but without assuming perfect conductivity. By writing the fields inside and outside the scatterer, he was able to derive impedance boundary conditions (first and second order) as well as to give error estimates to the use of these boundary conditions. The following is a summary of Mitzner’s development of his SIBC using, for the most part, his own notation. By first considering a general homogeneous region V (Figure 1.4), the expressions relating the fields in V and the tangential fields on the surface of the region are given by the EFIE and the MFIE: ~ ~ ¼~ E L*(~ n~ E) þ ZM*(~ n H) Einc
(1:31)
~ L*(~ ~ þ (1=Z)M*(~ ~ inc H n H) n~ E) ¼ H
(1:32)
where ~ n is the unit normal vector out of V and the incident fields, ~ Einc and inc ~ , are solely due to sources in V. The operator L* and M* are defined as H
Classical Surface Impedance Boundary Conditions
E inc
H inc
11
nˆ V0
V
E
(ε0, μ0, σ = 0)
H
Scatterer FIGURE 1.4 Scatterer in free space and the fields inside and outside the scatterer. The normal unit vector points out of V.
(ε, μ, σ)
ð L*~ F ¼ r0 G ~ F0 dS0 j M*~ F¼ k
ð
(1:33)
S
(k2 GI rr0 G) ~ F0 dS0
(1:34)
S
In these relations ~ ~ F represents either ~ E or H I is the unit dyadic k is the wave number in V the symbol ‘‘ 0 ’’ denotes the variables of integration G is the Green function 0
G¼
pffiffiffiffiffiffi ejkj~r~r j , k ¼ v me 4pj~ r ~ r0 j
(1:35)
which may be written either inside or outside V by using the appropriate wave number (kc or k0). Equations 1.33 and 1.34 are valid everywhere inside V but not on the Km are defined: boundary s. To obtain the SIBCs, the surface currents ~ Ke and ~ ~ ~ ~ Ke ¼ ~ n H; Km ¼ ~ n~ E
(1:36)
In Equation 1.36, the normal is assumed to point out of the scatterer V. The surface currents are related through the Leontovich SIBC [6]: ~ n~ Ke Km ¼ Zc~
(1:37)
Zc is the wave impedance in the scatterer: Zc ¼
pffiffiffiffiffiffiffiffiffiffiffiffi mc =ec ;
ec ¼ e þ js=v;
mc ¼ m þ jt=v
(1:38)
12
Surface Impedance Boundary Conditions: A Comprehensive Approach
The SIBC due to Mitzner is developed assuming e and t to be negligible (nonferrous, metallic conductors) although the SIBC obtained can be extended to more general media. To derive the condition, Equation 1.32 is first written for the free space, V0, outside the scatterer, V, by replacing Zc by n(~ rs ) and Equation 1.32 (where ~ rs is on Z0. Next, the cross product between ~ the surface S) is written ~ ~ ~ (1=Z0 )~ ~ inc (1:39) ~ n(~ rs ) H n(~ rs ) L*(~ n H) n(~ rs ) M*(~ n~ E) ¼ ~ n(~ rs ) H Now the following identity is added to the left-hand side of the equation ð S
j 0 (Km (~ rs ) rs )(~ n(~ rs ) r0 G) k0 Z 0 þ
j (~ n0 Km (~ rs ))(k0u þ k0v )(~ n(~ rs ) r0 G) ds0 ¼ 0 k0 Z 0
(1:40)
rs ) and ~ n(~ rs ) are independent of ~ r, and k0u and k0v are the where both ~ Km (~ principal curvatures on the surface in the u- and v-directions [12]. Taking the limit as ~ r !~ rs , the MFIE outside the scatterer becomes 1~ ~ inc nH Ke L0 ~ Ke (1=Z0 )M0 ~ Km ¼ ~ 2
(1:41)
Similarly, starting with Equation 1.31, he assumes that the incident electric field intensity inside the scatterer is zero (a conducting scatterer) and, following identical steps, writes the EFIE inside the scatterer as 1~ Km Zc Mc ~ Ke ¼ 0 Km þ Lc ~ 2
(1:42)
The operators L and M are defined through the following: ð n (r0 G ~ L~ K¼ ~ K0 )ds0 ¼ 0 ð
(1:43)
S
h
i 1 0 ~ n k2 G~ K0 (k0u þ k0v )(r0 G)(~ n0 ~ n) ~ K þ (~ K0 ~ K) rs r0 G ds0 M~ K¼ jk S
(1:44) In these relations, S denotes the surface of the scatterer excluding the singularity (punctured surface) and the Green’s function is either the free space Green’s function or the Green’s function in the scatterer as indicated by the
13
Classical Surface Impedance Boundary Conditions
t
ρ
2t
L S r –t
FIGURE 1.5 Local system of coordinates used to derive the SIBC. 0
use of k0 or kc for k in Equation 1.35. rs denotes the tangential gradient with respect to ~ r 0. To derive the SIBC, it is necessary here to digress and derive some auxiliary relations on the surface S of the subdomain V, which take into consideration the local geometry of S. First, a local system of coordinates centered ev , ~ n) so that (~ eu ~ ev ¼ ~ n) and ~ n points into V. at ~ r is defined (Figure 1.5) as (~ eu ,~ A second, cylindrical system (r, u, z), defines a cylinder t z t, r t, and t is sufficiently small so that the z-coordinate on S, within the cylinder, is single valued. Taking L to be the tangent plane to S at ~ r, and St a disklike subdomain cut by the right cylinder of radius t, then the projection of St on L n ~ n0 )dS0 . Now, writing a function g with odd periodicity, is D and dD0 ¼ (~ g(r, u, p) ¼ g(r, u) and a function h with even periodicity, h(r, u, p) ¼ h(r, u), Mitzner defines the following integral formulas ð D
ð D
~ K0 g=rn dD0 ¼ lim
ðt
D!0
ðp h i dr (~ Kþ ~ K )g=rn1 du 0
D
(~ K0 ~ K)h=rn dD0 ¼ lim
ðt
D!0 D
(1:45)
ðp h i dr (~ Kþ 2~ Kþ~ K )g=rn1 du
(1:46)
0
where ~ Kþ ¼ ~ K½ r, u, f (r, u) ; ~ K ¼ ~ K½r, u, p, f (r, u, p)
(1:47)
and the integral on D indicates the integral over the punctured surface which excludes the singularity at ~ r.
14
Surface Impedance Boundary Conditions: A Comprehensive Approach
The surface St is required to be sufficiently smooth everywhere so that it can be represented in terms of the principal curvatures at ~ r (excluding error terms) as z ¼ f (r, u) ¼
r2 ku cos2 u þ kv sin2 u 2
(1:48)
With these results and the assumptions on St, 0
r Gc ¼
s* e(j1)R=d (r=R)b1~ 2 4pR
(1:49)
For any vector ~ C: e ~ n (~ C r0 )r0 Gc ¼
(j1)R=d
4pR3
h
i b1 ~ C* ~ s*)~ s* ~ n C* ~ n b2 (~
(1:50)
In these relations, ~ C* is the component of ~ C normal to ~ n0 : ~ C* ¼ ~ n0 (~ C ~ n0 ) and
b1 ¼ 1 (j 1)R=d; b2 ¼ 3 3(j 1)R=d 2j(R=d)2 (r=R)2 h i ~ r ~ r0 r ~ s* ¼ ¼ (cos u~ ex þ sin u~ ey ) þ (ku cos2 u þ kv sin2 u)~ n r 2
(1:51) (1:52)
Further approximations are possible based on the fact that the skin depth is small, again, excluding error terms. Specifically r 1 R
!
e(j1)R=d ¼ e(j1)r=d
(~ n ~ n0 )1 1; ~ n ~ n0 ¼ r(ku cos u~ ex þ kv sin u~ ey )
(1:53) (1:54)
Now, in a conducting medium, the Green’s function is approximated as G¼
ejkc R e(j1)R=d ¼ 4pR2 4pR2
!
jGj ¼
eR=d 4pR2
With this approximation, the operators Lc and Mc in Equation 1.42 are approximated by the operators LtC and MtC , in which the integration is over a subarea St on S as defined above. St has dimensions of the order of d and, in fact, will be taken to be a smooth curved disk of radius d. Once these new operators are evaluated, the two approximations are written as Km LtC ~ Km ; LC ~
Ke MtC ~ Ke MC ~
(1:55)
Classical Surface Impedance Boundary Conditions
15
Km is approximated by LtC ~ Km and MC~ Ke by MtC ~ Ke . These approxiThus, LC~ mations facilitate the evaluation of the various terms needed to obtain the SIBC. These operators are evaluated using the previous relations. Starting with LtC : ð e(j1)R=d LtC ~ n ~ n0 )1 b1 (r=R)3 Km ¼ (~ 4pr2 D n h i 0 h 0 i o 0 s * þ (~ n ~ s *) ~ Km ~ n ~ n) ~ s * dD0 (~ n ~ s *)~ Km (~ n ~ n0 )~ Km ~ Km (~ Km ~ Km ~
(1:56)
To evaluate the integral in Equation 1.56, the various terms are substituted from Equations 1.51 through 1.54 after removing the error terms but, in addition, instead of integrating on D, the integration is performed on the tangent plane L. With these, Equation 1.56 is approximated by LC ~ Km ¼
ð
e(j1)r=d [1 (j 1)r=d] 4pr2
L
h i 1 Km ku cos2 u(~ Km ~ eu )~ eu þ kv sin2 u(~ Km ~ ev )~ ev dD0 (ku cos2 u þ kv sin2 u)~ 2
(1:57) and, from this, on the tangent plane to S Km ¼ LC ~
i ph ~ (Km ~ eu )~ eu (~ Km ~ ev )~ ev 2
(1:58)
where 1 p ¼ (1 þ j)d(kv ku ) 4
(1:59)
A similar discussion leads to approximation of MC~ Ke , which is Ke ¼ (1 j) MC ~
ð
~ n~ Ke e(j1)r=d (r=d)~ n~ Ke dD0 ¼ 2 4pr 2
(1:60)
L
Finally, by substituting Equation 1.59 into Equation 1.43, an expression for the SIBC is obtained as eu þ (1 p)Kmv~ ev ¼ 2Zc Mc Ke (1 þ p)Kmu~
(1:61)
16
Surface Impedance Boundary Conditions: A Comprehensive Approach
where Kmu ~ eu Kmu ¼ ~ Kmv ¼ ~ Kmv ~ ev Zc may be written explicitly in terms of the skin depth as Zc ¼
(1 þ j) vmd 2
(1:62)
Ke , Equation 1.61 can be With this and using Equation 1.60 to evaluate MC~ ~ in the u- and written explicitly in terms of the components of ~ E and H v-directions as (1 þ j) (1 j) vmd 1 þ d[kv ku ] Hv 2 2 (1 þ j) (1 j) Ev ¼ vmd 1 d[kv ku ] Hu 2 2
Eu ¼
(1:63) (1:64)
The expression in Equation 1.61 is the SIBC known as the Mitzner boundary condition as published in [7]. Equations 1.63 and 1.64 are in explicit form, useful for implementation. Also, this form is identical to the curvaturedependent SIBC given by Leontovich [6]. It is also interesting to note that the curvature-dependent (first-order) interface condition due to Leontovich as well as the classic (zeroth-order)Leontovich condition can be obtained from the second-order SIBC in Equation 1.61 by writing eu þ (1 p)Kmv~ ev ¼ Zc~ n Ke (Keu~ ev Kev~ eu ) (1 þ p)Kmu~
(1:65)
Separating the components (1 þ p)Kmu ¼ Zc Kev ;
(1 p)Kmv ¼ Zc Keu
(1:66)
Since the correction term, p, is due to the curvature (see Equation 1.59), simply omitting the term p gives the first-order SIBC due to Leontovich: Kmu ¼ Zc Kev ;
Kmv ¼ Zc Keu
(1:67)
Mitzner also analyzed the errors involved in the use of the second-order SIBCs as well as of the first-order SIBC. We will discuss errors of the method in following chapters, including comparison with analytic, numerical, and experimental data, but we note here that the second-order SIBCs are of the
Classical Surface Impedance Boundary Conditions
17
order of O d2 k02 , O d2 =h2 , and O(d2 k2 ), where h is the distance to the nearest source and k is the larger of jkuj and jkvj. k0 is the free space wave number. From the same analysis, he concludes that the curvature-dependent Leontovich SIBC in Equation 1.65 has errors of the order O d2 k02 , The classical Leontovich O(d2 =h2 ), and O(d2 k2 ), just like the Mitzner SIBC. SIBC in Equation 1.66 has errors of the order O d2 k02 , O(d2 =h2 ), and O(dk).
1.5 Rytov’s Approach 1.5.1 General Rytov’s approach to the idea of SIBCs was much more radical and general than his contemporaries while also preceding most of them. It all started in an attempt to verify and quantify the Leontovich SIBC at the request of Leontovich [2,3]. In the process, Rytov developed a general approach to low- and high-order SIBCs based on the perturbation method [2,3]. This approach also allowed the systematic analysis of the accuracy and applicability of the method and has pointed to the relative value of higher-order SIBCs. Rytov started with the source-free Maxwell’s equations and proceeded to calculate the electromagnetic fields inside and outside the conductor. The electric and magnetic fields were assumed to vary exponentially inside the conductor and were written as a power series expansion in the skin depth d. The expansion coefficients were evaluated for a system of equations that is obtained by equating terms with equal powers of d. Equating the external and internal approximations leads to the Rytov boundary conditions. The number of terms retained in the expansion defines the order of the SIBC obtained. Thus, higher-order SIBCs are as easily obtained as lower-order SIBCs. Further, since a higher-order SIBC is obtained by adding terms to the lower-order SIBC, the method is ideally suited for numerical computation since the form of the expansion allows a modular approach whereby addition of terms does not require rewriting software. In the section that follows, we describe the method and derive the first four terms of the expansion. The zeroth-order term (first term in the expansion) will be shown to be the Leontovich SIBC while the first-order term corresponds to the Mitzner SIBC. The second- and third-order terms lead to higherorder SIBCs, which, in addition to curvature, allow for variations of fields on the surface of the conductor. Since one of the goals of Rytov’s work was the applicability and accuracy of SIBCs, the last part of this section discusses in detail the applicability of the method. The discussion focuses on skin depth as a small parameter and the errors expected are in terms of the skin depth.
18
Surface Impedance Boundary Conditions: A Comprehensive Approach
1.5.2 Calculation of the Field inside the Conductor In a conducting medium, under time-harmonic excitation and no source currents, the electric and magnetic field distribution inside the conductor can be described by the following: ~ r~ E ¼ jvmH;
~ ¼ s~ rH E
(1:68)
For simplicity, we assume that only one conductor exists and that its shape admits introduction of a system of curvilinear coordinates (j1, j2, h) in which the coordinates j1 and j2 are directed along the surface and coordinate h is normal to it pointing into the conductor. We seek the solution of Equation 1.68 in the following form ~ 1 , j2 , h) exp (c(h0 )); ~ E ¼ A(j
~ ¼~ H B(j1 , j2 , h) exp (c(h0 ))
(1:69)
where h0 ¼ h=d
(1:70)
where d is the skin depth characterizing the electromagnetic penetration and is written in the form sffiffiffiffiffiffiffiffiffiffi 2 d¼ vsm
(1:71)
Equation 1.71 can be rewritten in the form: s¼
2 vmd2
(1:72)
Substituting Equation 1.69 into Equation 1.68 and taking into account Equation 1.71, we obtain dc hh Aj2 ¼ idBj1 dh0
(1:73)
~ þ dc hh Aj ¼ idBj d(r A) j2 1 2 dh0
(1:74)
~ ¼ idBh d(r A) h
(1:75)
~ d(r A) j1
B)j1 d d2 (r ~
dc hh Bj2 ¼ 2Aj1 dh0
(1:76)
Classical Surface Impedance Boundary Conditions
d2 (r ~ B)j2 þ d
dc hh Bj1 ¼ 2Aj2 dh0
B)h ¼ 2Ah d2 (r ~
19
(1:77) (1:78)
where the Lame coefficient, hh, corresponding to coordinate h was obtained from the following well-known formula: ds ¼ 2
dj1 h j1
2 2 2 dj2 dh þ þ h j2 hh
(1:79)
~ and ~ B in the form of power series in d gives Representing A ~ 1 þ d2 ~ ~¼~ B2 þ B¼~ B0 þ d~ B1 þ d2~ A A0 þ dA A2 þ ; ~
(1:80)
The system of equations for the expansion coefficients is obtained by comparing the terms of equal powers of d. The zero-order approximation takes the form (c0 ¼ (dc=dh0 )): c0 hh A0j2 ¼ 0; 2A0j1 ¼ 0;
c0 hh A0j1 ¼ 0
2A0j2 ¼ 0;
2A0jh ¼ 0
Therefore, A0j1 ¼ A0j2 ¼ A0h
(1:81)
The first-order approximation is as follows: 8 0 > < c hh A0j2 ¼ iB0j1 c0 hh A0j1 ¼ iB0j2 ; > : 0 ¼ iB0h
8 0 > < 2A1j1 ¼ c hh B0j2 0 2A1j1 ¼ c hh B0j1 > : 2A1h ¼ 0
(1:82)
Therefore, B0h ¼ A1h ¼ 0
(1:83)
The function c0 can be obtained from the condition of compatibility between the equations for A1jm and B0jm, m ¼ 1, 2, and is written in the form: qffiffiffiffiffiffiffiffiffiffiffi c0 ¼ 2i=hh
(1:84)
In the last expression, the negative sign was chosen to reflect the fact that the field decays with increase in the distance from the surface. Performing the integration over the skin layer in Equation 1.84 yields
20
Surface Impedance Boundary Conditions: A Comprehensive Approach
0
h ðh pffiffiffiffi ð dh0 1 þ i dh c ¼ 2i ¼ hh d hh
(1:85)
0
0
Using Equation 1.84, the relations between the tangential components of ~ A1 and ~ B0 can be represented in the form: A1j1
rffiffiffi i B0j ; ¼ 2 2
A1j2
rffiffiffi i B0j ¼ 2 1
(1:86)
The second-order approximation looks as follows: c0 hh A2j2 ¼ iB1j1 hj2 hh
@(A1j2 =hj2 ) @h @(A1j1 =hj1 ) @h
c0 hh A2j1 ¼ iB1j2 þ hj1 hh 0 ¼ iB1h þ hj1 hj2
2A2j2 ¼ c0 hh B0j1 þ hj1 hh 2A2h ¼ hj1 hj2
(1:88)
@(A1j2 =hj2 ) @(A1j1 =hj1 ) @j1 @j2
2A2j1 ¼ c0 hh B1j2 hj2 hh
(1:87)
@(B0j2 =hj2 Þ @h
@(B0j1 =hj1 ) @h
@(B0j2 =hj2 ) @(B0j1 =hj1 ) @j1 @j2
(1:89) (1:90) (1:91)
(1:92)
Combining Equations 1.86 and 1.92, we obtain B1h
hj1 hj2 @(B0j2 =hj2 ) @(B0j1 =hj1 ) ¼ pffiffiffiffi @j1 @j2 2i
(1:93)
The system of equations in Equations 1.90 through 1.93 can be solved if the following condition is satisfied: ! ! B20j1 B20j2 @ @ ¼ ¼0 @h hj1 hj2 @h hj1 hj2
(1:94)
The pairs of Equations 1.87 and 1.91, and Equations 1.88 and 1.90 are compatible only if Equation 1.94 is satisfied and then the following solutions are obtained:
Classical Surface Impedance Boundary Conditions
A2j1 A2j2
rffiffiffi hh @ ln (hj1 =hj2 ) i B0j2 ; B1j2 ¼ @h 4 2 rffiffiffi hh @ ln (hj2 =hj1 ) i B0j1 ¼ B1j þ @h 4 2 1
21
(1:95)
Substituting Equations 1.83, 1.86, and 1.95 into Equations 1.80 and then into Equation 1.69, we obtain rffiffiffi
hh @ ln (hj1 =hj2 ) i B0j2 þ Ej1 ¼ mvd exp (c) (1:96) B0j2 þ d B1j2 pffiffiffiffi @h 2 2 2i rffiffiffi
hh @ ln (hj2 =hj1 ) i B0j1 þ (1:97) B0j1 þ d B1j1 pffiffiffiffi Ej2 ¼ mvd exp (c) @h 2 2 2i Eh ¼
@(B0j2 =hj2 ) @(B0j1 =hj1 ) mvd2 þ exp (c) hj1 hj2 @j1 @j2 2
(1:98)
Hj1 ¼ exp (c){B0j1 þ dB1j1 þ }
(1:99)
Hj2 ¼ exp (c){B0j2 þ dB1j2 þ }
(1:100)
hj1 hj2 @(B0j1 =hj2 ) @(B0j2 =hj1 ) þ Hh ¼ d exp (c) pffiffiffiffi þ @j1 @j2 2i
(1:101)
Equations 1.96 through 1.101 lead to the following conclusions about the behavior of the electromagnetic field in the skin layer. Due to the term, exp(c), all components of the electric and magnetic fields inside the conductor decay with increase in distance from the surface. The tangential components of the magnetic field are quantities of the zero-order magnitude and they tend to be discontinuous at the conductor surface in the limiting case of d ! 0. The normal magnetic and tangential electric fields decay as quantities of the first order of magnitude, and the normal electric field behaves as a quantity of second-order magnitude. 1.5.3 Boundary Conditions at the Conductor Surface The expressions in Equations 1.96 through 1.101 giving the electric and magnetic fields inside the conductor must be linked to the external fields Ee and He via the following boundary conditions at h ¼ 0: Eejm ¼ Ejm ;
m1 Hjem ¼ mHjm ;
e1 Eeh eEh ¼ x; Hhe ¼ Hh
(1:102)
22
Surface Impedance Boundary Conditions: A Comprehensive Approach
where x is the surface charge density e1 and m1 are the electric permittivity and magnetic permeability of the dielectric space surrounding the conductor We now expand the external fields and the surface charge density as power series in the small parameter d: ~e ¼ H ~ e þ dH ~ e þ d2 H ~e þ ; ~ Ee0 þ d~ Ee1 þ d2~ Ee2 þ ; H Ee ¼ ~ 0 1 2 2 ~e e e e ~ ~ ~ H ¼ H 0 þ dH 1 þ d H 2 þ
(1:103)
Substituting the expansions (Equation 1.103 into Equation 1.102) and equating the coefficients of equal powers of d, the following relations at the conductor surface (h ¼ 0) are obtained: Zero-order approximation: Ee0jm ¼ 0;
e H0j ¼ 0; m
e1 Ee0h ¼ x0 ;
e H0j ¼ B0jm m
(1:104)
These are conditions on the surface of an ideal conductor (PEC condition). First-order approximation: Ee1jm
rffiffiffi rffiffiffi i i e 3m B0j ¼ (1) mv H ¼ (1) mv , m ¼ 1, 2 2 3m 2 0j3m hj hj @(B0j1 =hj2 ) @(B0j2 =hj1 ) e ¼ m p1 ffiffiffiffi2 þ m1 H1h @j1 @j2 2i 3m
e1 Eeh1 ¼ x1 ;
e H1j ¼ B1jm m
(1:105) (1:106) (1:107)
The conditions in Equations 1.106 and 1.107 provide corrections to the field distribution around the ideal conductor due to dissipation of energy by the real conductor. 1.5.4 Particular Case of a Planar Interface Consider the particular case of a planar conductor–dielectric interface. Let the x and y Cartesian coordinates be directed along the surface, and the z-coordinate normal to it inside the conductor. Thus hx ¼ hy ¼ hz ¼ 1 and derivations are significantly reduced so that the distribution of the electric and magnetic field components inside the conductor can be represented as functions of the external field at the interface and the z-coordinate:
Classical Surface Impedance Boundary Conditions
23
" !# rffiffiffi( e e @ 2 H0y @ 2 H0y 1þi i z e e Ex ¼ mvd exp þ H0y þ d H1y þ pffiffiffiffi @y2 d 2 2 2i @x2 " !# ) e e e @ 2 H0y @ 2 H0y 1 @ 2 H0x 2 2 þ d B2y þ þ þ (1:108) 4i @x@y @x2 @y2 2 e rffiffiffi( e @ H0x @ 2 H0x 1þi i z e e H0x þ d H1x þ pffiffiffiffi þ Ey ¼ mvd exp @y2 d 2 2 2i @x2 " !# ) e e e @ 2 H0y @ 2 H0y @ 2 H0x 1 2 2 þ þ (1:109) þ d B2y þ @x@y @x2 @y2 4i " ( e e e e @H0y @H0x @H1y @H1x mvd2 1þi exp þ þd Ez ¼ 2 @x @y @x @y d !# ) e e e @ 3 H0y @ 2 H0y z @ 3 He @ 3 H0x p ffiffiffiffi þ þ þ 2 0x 3 2 3 @x@y @x @y @y 2 2i @x
(1:110)
2 e
e @ H0x @ 2 H0x 1þi z 2 e e Hx ¼ exp þ d H0x þ d H1x þ pffiffiffiffi þ B þ 2x @y2 d 2 2i @x2 (1:111) " ! # ( ) e e @ 2 H0y @ 2 H0y 1þi z 2 e e þ d B2y þ þ H0y þ d H1y þ pffiffiffiffi Hy ¼ exp @y2 d 2 2i @x2 (1:112) " ( e e e e @H1y d 1þi @H0x @H0y @H1x þ þd þ z Hz ¼ pffiffiffiffi exp @x @y @x @y d 2i !# ) 3 e 3 e e e @ H0y @ H0y z @ 3 H0x @ 2 H0x þ pffiffiffiffi þ þ 2 þ þ 3 2 @x@y @x @y @y3 2 2i @x
(1:113)
1.5.5 Notes on Applicability of the Method For validity of the asymptotic behavior of the solution, the first-order terms must be much smaller than the zero-order terms as follows: 2 e 2 e e H dH e þ pz ffiffiffiffi @ H0x þ @ H0x 0x 1x 2 2i @x2 2 @y
(1:114)
24
Surface Impedance Boundary Conditions: A Comprehensive Approach
Equation 1.114 can be written in the following form without consideration e of H1x for now: e e e zd @ 2 H0x @ 2 H0x H p ffiffiffiffi þ 0x @y2 2 2i @x2
(1:115)
As z increases, the condition in Equation 1.115 will eventually become invalid, but the multiplier exp(z=d) keeps the solution valid for z d. Therefore, Equation 1.115 can be represented in the form:
2 e 2 e e H d2 @ H0x þ @ H0x 0x @x2 @y2
(1:116)
The condition in Equation 1.115 means that the field should vary slowly along the interface. In the case of illumination of the conductor by a spherical wave, the distance, R, between the wave source and the interface must be restricted as follows
d2 R 2 d 2 l2
if if
Rl Rl
(1:117)
where l is the incident wavelength. The second condition in Equation 1.117 can be represented in another form f ms
(1:118)
where f ¼ v=2p is the source frequency. The condition in Equation 1.118 is always met for metals, but it should be checked when poor conductors are analyzed. Consider now the first term on the right-hand side of Equation 1.114. The condition e H dHe 0x 1x
(1:119)
must be satisfied, if we wish to restrict ourselves to zero-order approximation. For spherical waves, Equation 1.119 is met while R is small. In other words, for any R it is possible to estimate a value for d so that Ee and He will be approximated by Ee0 and H0e with an accuracy specified a priori. In the general case of a nonplanar surface, an additional condition allowing for the curvature of the interface must be taken into account. Comparing terms of the order of magnitude of d and d2 in Equations 1.96 and 1.97, it is readily found that the following condition must be met in order for terms of the order d2 to be neglected:
Classical Surface Impedance Boundary Conditions dhh @ ln (hj1 =hj2 ) 1 pffiffiffi @h 2 2
25
(1:120)
The right-hand side of Equation 1.120 vanishes in the case of planar interfaces (hj1 ¼ hx ¼ 1; hj2 ¼ hy ¼ 1), but in the case of cylindrical surfaces (hj1 ¼ hz ¼ 1; hj2 ¼ hf ¼ 1=r; hh ¼ hr ¼ 1; h ¼ r), Equation 1.120 takes the form: d r pffiffiffi 2 2
(1:121)
Application of Equation 1.120 to a sphere (hj1 ¼ hu ¼ 1=R; hj2 ¼ hf ¼ 1=(R sin u); hh ¼ hR ¼ 1; h ¼ R) yields @ ln (hj1 =hj2 ) @ ln ( sin u) ¼ ¼0 @h @R
(1:122)
Clearly, this does not mean that the skin depth should not be small as compared to the radius in the case of a sphere. This is a common requirement for all curved surfaces, but it is formulated for each surface in a different way as follows from Equations 1.96 through 1.101.
References 1. S.A. Schelkunoff, The impedance concept and its application to problems of reflection, radiation, shielding and power absorption, Bell System Technical Journal, 17, 1938, 17–48. 2. S.M. Rytov, In the oscillations laboratory, in Memoires to Academician Leontovich, Nauka, Moscow, 1996, pp. 40–66 (in Russian). 3. S.M. Rytov, Calcul du skin-effet par la méthode des perturbations, Journal of Physics, 2(3), 1940, 233–242. 4. S.M. Rytov, Calculation of skin effect by perturbation method, Journal Experimental’noi i Teoreticheskoi Fiziki, 10(2), 1940, 180–189 (in Russian). 5. M.A. Leontovich, On one approach to a problem of the wave propagation along the Earth’s surface, Academy of Sciences USSR, Series Physics, 8, 1944, 16–22. 6. M.A. Leontovich, On the approximate boundary conditions for the electromagnetic field on the surface of well conducting bodies, in B.A. Vvedensky (ed.), Investigations of Radio Waves, Academy of Sciences of USSR, Moscow, 1948. 7. K.M. Mitzner, An integral equation approach to scattering from a body of finite conductivity, Radio Science, 2(12), 1967, 1459–1470. 8. Lord Rayleigh, On the self inductance and resistance of straight conductors, Philosophical Magazine, 21, 1886, 381–394.
26
Surface Impedance Boundary Conditions: A Comprehensive Approach
9. G.S. Smith, On the skin effect approximation, American Journal of Physics, 58(10), 1990, 996–1002. 10. C.A. Balanis, Advanced Engineering Electromagnetics, John Wiley & Sons, New York, 1989, pp. 210–213. 11. N. Shchukin, Propagation of Radio Waves, Svyazizdat, Moscow, 1940. 12. J. van Bladel, Electromagnetic Fields, Appendix 2, McGraw-Hill Book Co., Inc., New York, 1964.
2 General Perturbation Approach to Derivation of Surface Impedance Boundary Conditions
2.1 Introduction In Chapter 1, we have considered three different approaches (Leontovich’s, Mitzner’s, and Rytov’s) to the development of SIBCs and discussed the historical background. Although Rytov’s approach [1] seems to be more robust than those of Mitzner [2] and Leontovich [3,4] and allows one to take into account such high-order effects as variation of the field along the conductor’s surface, its application area is restricted to time-harmonic problems with singlefrequency excitation. However, Rytov’s conditions have been derived in Cartesian, cylindrical, and spherical coordinate systems and therefore, unlike Mitzner’s SIBCs, they cannot be directly applied to arbitrarily curved interfaces with variable radii of curvature. Therefore, none of these approaches can be considered a ‘‘universal tool’’ for the development of approximate boundary conditions for skin effect problems. A general approach should have only those limitations that are common to a wide range of skin effect problems. One such basic limitation is the requirement that the skin depth, d, be much smaller than the characteristic size, D, of the conductor’s surface d¼
pffiffiffiffiffiffiffiffiffiffiffi t=sm D
(2:1)
where t is the ratio 2=v in the case of time-harmonic fields or the incident pulse duration in the case of transient sources. The condition in Equation 2.1 is related to the well-known fact that the normal magnetic field and tangential electric field in the skin layer of a conductor is much smaller than the tangential magnetic field. In other words, Maxwell’s equations for the conducting domain combine terms of different orders of magnitude. In the limiting case of a perfect electrical conductor the skin depth is zero and the normal magnetic and tangential electric fields vanish at the conductor=dielectric interface. It is natural to assume that the equations governing the field distributions in the skin layer ‘‘implicitly’’ include a small parameter proportional to the skin depth. Then the approximate relation between the tangential electric and magnetic fields at the interface can be sought in the form of asymptotic expansions in this small parameter. The question is how to choose the small parameter? Rytov 27
28
Surface Impedance Boundary Conditions: A Comprehensive Approach
demonstrated that the choice does not look difficult in the particular case of time-harmonic excitation. Consideration of the general case requires a definition suitable for any time dependence of the incident source, even in nonlinear cases, when the one-dimensional problem of the field diffusion into the conducting half-space does not have an analytical solution. It is perhaps better to reformulate the question as follows: Do we need to define the parameter ourselves, in some way, or could it be ‘‘extracted’’ from the governing equations? That is, can the equations be transformed to reveal this parameter? Perturbation theory leads us to favor this possibility. Following this theory, we first introduce the scale factors for the basic variables in the governing equations. Selection of the scale factors is based on knowledge of the characteristic variation of such input data as total current and dimensions of the conductor. Then the equations are rewritten in terms of nondimensional variables so that each is a function of the corresponding dimensional variable and its scale factor. As a result, combination(s) of the scale factors will appear in the governing equations as parameters. Unlike nondimensional variables that are of the same order of magnitude under the definition, the scale factors (as well as the original dimensional variables) are of different orders of magnitude so that one of the combinations of the scale factors is expected to be the desired small parameter of the problem. The distribution of various electromagnetic quantities in the conducting domain can be described by one specific equation—the equation of diffusion. @~ f r (r ~ f ) þ sm ¼ 0 @t
(2:2)
where the vector function ~ f may denote various electromagnetic fields and potentials. The purpose of this chapter is transformation of this equation with the use of perturbation techniques and derivation of a set of approximate relationships between components and derivatives of ~ f at the interface. Those results can be applied not only to electromagnetics, but also to any other area where functions under consideration are governed by the diffusion equation.
2.2 Local Coordinates For simplicity, we first consider the planar case of long parallel conductors. Let us direct the x-coordinate along the conductors and neglect all derivatives with respect to x: @ ¼0 @x
General Perturbation Approach to Derivation of SIBCs
29
dη
η = const
dξ
ξ = const d FIGURE 2.1 Local orthogonal curvilinear coordinate system related to the surface. (Adapted from Mestel, A.J., IMA J. Appl. Math., 45, 33, 1990.)
Local to the surface, we define inward normal and tangential orthogonal coordinates, h and j, as in Figure 2.1. From Figure 2.1 it is clear that, when the curvature is positive, the scale factor corresponding to j in an orthogonal system hj is a decreasing function of h. In fact, when j represents the arc length and the contour of the conductor’s surface, the Lame coefficients can be written in the form hj ¼ 1 h=d;
hh ¼ 1; hx ¼ 1
(2:3)
where d ¼ d(j) is the local radius of curvature. With these new coordinates, the curl, div, and curl curl operators applied to an arbitrary vector function ~ f take the form 1 @(hj fj ) @(hh fh ) 1 @((1 h=d)fj ) @fh ~ ¼ (r f )x ¼ @j @j hj h h @h 1 h=d @h 1 @fj fj @fh @fj fj d @fh ¼ (1 h=d) ¼ @h d @j @h d h d h @j 1 h=d (2:4) 1 @(hx fx ) @(hh fh ) @fx ¼ (2:5) (r ~ f )j ¼ @x @h h x hh @h 1 @(hj fj ) @(hx fx ) d @fx (r ~ f )h ¼ (2:6) ¼ h x hj @x @j d h @j @(hj hh fx ) @(hx hh fj ) @(hx hj fh ) 1 ~ þ þ rf ¼ @x @j @h hx hj hh @fh fh fh 1 @fj d @fj @fh ¼ þ (1 h=d) ¼ þ (2:7) @h d 1 h=d @j d h @j @h d h
30 h
Surface Impedance Boundary Conditions: A Comprehensive Approach
r (r ~ f)
i
1 @j hh @(hj fj ) @(hx fx ) @ hj @(hx fx ) @(hh fh ) x @x hj hh @ hx hj @x @j @h hx hh @h 1 @ 1 @fx @ @fx hj ¼ @h hj @j hj @j @h d @ d @fx d @ 2 fx @ d h @fx d h @ 2 fx ¼ @h d h @j d h @j d h @j2 @h d d @h2 d hd0 @fx d @ 2 fx 1 @fx d h @ 2 fx þ ¼ dh d @h2 (d h)2 @j d h @j2 d @h ¼
¼ h
hdd0 @fx d2 @ 2 fx 1 @fx @ 2 fx þ 3 @j 2 2 d h @h @h2 (d h) (d h) @j
(2:8)
i 1 @ hh @(hj fj ) @(hx fx ) @ hx @(hh fh ) @(hj fj ) ~ r (r f ) ¼ j @j hx hh @x hx hj @x @j @h hj hh @h @ 1 @fh @fj @hj @ 1 @fh @fj fj @hj hj fj ¼ ¼ @h @h @h hj @j @h hj @j @h hj @h @ d @fh @fj fj þ ¼ @h d h @j @h d h ¼
@fh d d @ 2 fh @ 2 fj fj 1 @fj þ 2þ þ 2 @j 2 @h@j @h d h d h @h (d h) (d h)
(2:9)
h i 1 @ hj @(hx fx ) @(hh fh ) @ hx @(hh fh ) @(hj fj ) r (r ~ f) ¼ h @x @j hx hj @x hx hh @h @j hj hh @h 1 @ 1 @fh @hj @fj 1 @ 1 @fh fj @hj @fj fj hj ¼ ¼ @h @h hj @j hj @j hj @j hj @j hj @h @h d @ d @fh fj @fj ¼ þ d h @j d h @j d h @h ¼
hdd0 @fh d2 @ 2 fh dd0 d @fj d @ 2 fj þ f þ 3 @j 2 2 3 j 2 @j d h @j@h (d h) (d h) (d h) @j (d h)
(2:10)
where d0 denotes the derivative of d with respect to j. The characteristic lengths associated with the coordinates j and h are the characteristic size D of the conductor’s surface and the skin depth d, respectively. These lengths may be of different orders of magnitude due to the condition in Equation 2.1. Thus, it is natural to introduce nondimensional variables, ~j and h ~ , that have variation ranges of the same order of magnitude and are related to j and h, respectively, as follows: ~j ¼ j=D;
h ~ ¼ h=d
(2:11)
General Perturbation Approach to Derivation of SIBCs
31
Here and below, the sign ‘‘’’ denotes nondimensional quantities. Switching to variables ~j and h ~ in Equations 2.4 through 2.10, we obtain fj d 1 @fj 1 @fh ~ (2:12) (r f )x ¼ d D @h ~ d d~ h d d~ h @~j @fx (r ~ f )j ¼ d1 @h ~ (r ~ f )h ¼ r ~ f ¼
@fh fh d @fj D1 þ d1 ~ d d~ h @h ~ d d~ h @j
h i r (r ~ f ) ¼ d x
þ h
i r (r ~ f) ¼ j
h i r (r ~ f) ¼ d h
h ~ dd0 (d d~ h)
3
D1
(d d~ h)
2
D1
(d d~ h)
3
D1
(2:15)
(2:16)
@fh @ 2 fh d @ 2 fj þ d2 2 d1 D1 @h ~ h @~j d d~ @h ~ @~j
fj 1 @fj d1 þ 2 @h ~ d d~ h (d d~ h) h ~ dd0
(2:14)
@fx d2 @ 2 fx D2 2 2 @~j (d d~ @~j h)
1 @fx @ 2 fx d2 2 d1 @h ~ @h ~ d d~ h
d þ
d @fx D1 d d~ h @~j
(2:13)
(2:17)
@fh @ 2 fh d2 dd0 D2 2 þ fj 2 @~j (d d~ @~j h) (d d~ h) 3
d @fj d @ 2 fj þ D1 d1 D1 2 @~j d h @~j@ h ~ (d d~ h)
(2:18)
Using Equation 2.1, let us introduce a small parameter p~ proportional to the ratio of the penetration depth to the characteristic size of the conductor’s surface: pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ~ p ¼ d=D ¼ t=(smD2 ) 1 (2:19) The second relation in Equation 2.11 can be represented in another form using Equation 2.19: h ~ ¼ h=(~ pD)
(2:20)
The local radius of curvature, d, is directly related to the variation of the function, ~ f , in the direction tangential to the surface of the conductor. This leads to the following representation: ~ d ¼ d=D
(2:21)
32
Surface Impedance Boundary Conditions: A Comprehensive Approach
Substituting Equations 2.20 and 2.21 into Equations 2.12 through 2.18, we get " # ~d @fh @f f j j 1 (r ~ f )x ¼ d ~p (2:22) ~ p ~d ~ph ~ @h ~ ~ @~j d~ ph ~ @fx (r ~ f )j ¼ d1 @h ~ p (r ~ f )h ¼ d1 ~ p r ~ f ¼ d1 ~ h
i p3 r (r ~ f ) ¼ d2 ~ x
(2:23)
~ d
@fx ~ d~ ph ~ @~j
~ d
fh @fj @fh ~p þ ~d ~ph ~ ~ ~ d~ ph ~ @ ~j @ h
(2:24) ! (2:25)
~ ~p @fx @ 2 fx h ~~ d~ d0 @fx @ 2 fx d2 2 ~ 2 p þ 2 ~d ~ph ~ @h ~ ~h ~ @h (~ dp ~ )3 @~j (~ d ~ph ~ )2 @~j
!
(2:26) h
i r (r ~ f ) ¼ d2 ~ p2 j
~ ~ ~ @fh @ 2 fh @ 2 fj d d fj p @fj þ~ p 2þ~ p2 þ 2 2 ~ ~ ~ ~ ~ ~ @ h ~ ~ ~ ~ ~ @ j @ h ~ @ j (d ph ~) d ph ~ (d ph ~) d~ ph ~ @h
!
(2:27) h i r (r ~ f ) ¼ d2 ~p3 h
~ ~ @ 2 fh h ~ ~d~d0 @fh d2 d~ d0 ~p2 þ~ p2 fj 2 3 2 (~d p~h ~ ) @~j (~d ~ ph ~ ) @~j (~ d~ ph ~ )3 ! ~d ~d @fj @ 2 fj 2 þ ~p ~p ~d p~h ~ (~d p~h ~ )2 @~j ~ @~j@ h
(2:28)
In the general three-dimensional case, both tangential coordinates are curved. We denote them as j1 and j2. Their radii of curvature are d1 and d2, respectively, and the Lame coefficients can be written in the form: hjk ¼ 1 h=dk ;
hh ¼ 1,
k ¼ 1, 2
(2:29)
The curl, div, and curl curl operators with coordinates (j1, j2, h) are written in the following forms: @(hj3k fj3k ) @(hh fh ) 1 k ~ (r f )jk ¼ (1) @j3k @h hj3k hh
@ (1 h=d3k )fj3k @fh 1 ¼ @h @j3k 1 h=d3k @f fj3k @fh 1 j k 3k ¼ (1) (1 h=d3k ) @h d @j3k 1 h=d3k @fj3k fj3k @fh d3k , k ¼ 1, 2 (2:30) ¼ (1)k @h d3k h d3k h @j3k
33
General Perturbation Approach to Derivation of SIBCs
X 2 @(hj2 fj2 ) @(hj1 fj1 ) @fji 1 d3i ¼ (1)i (2:31) @j1 @j2 d3i h @j3i h j1 h j2 i¼1 " # 2 2 2 X X X @(h h f ) @(h h f ) 1 di @fji @fh 1 j h j j j h i 3i 1 2 r ~ f¼ þ þ fh ¼ d h @j @h h hj1 hj2 hh i¼1 d @h @j3i i i¼1 i l¼1 l (r ~ f )h ¼
h
i r(r~ f)
jk
@(hj3k fj3k ) @(hjk fjk ) hh @ @ hj3k @(hjk fjk ) @(hh fh ) @j3k @h @jk @jk hj3k hh @j3k hjk hj3k @h hjk hh hj @hjk hj3k @fh @fjk 1 @ 1 @fj3k 1 @fjk @ þfjk 3k ¼ hj hjk @h hjk @jk hj3k @j3k hjk @jk hj3k @j3k @h 3k @h (" #) @fjk 1 1 @ 2 fj3k @ 1 1 @ 2 f jk ¼ hj3k hjk @jk @j3k @j3k hj3k @j3k hj3k @j23k 1
¼
(2:32)
@hj3k @fjk @ 2 f jk @ hj3k @hjk hj3k @fjk @hjk @ hj3k @fh hj3k @ 2 fh þhj3k þ þf jk @h @h @h2 @h hjk @h @h @h hjk @jk hjk @h@jk @h hjk ( ) 0 hd d3k @fjk d23k @ 2 fj3k @ 2 f jk @fjk dk d3k 1 þ 3k þ ¼ (dk h)(d3k h) @jk @j3k (d3k h)3 @j3k (d3k h)2 @j23k d3k h @h
@ 2 fjk dk d3k 1 @fjk dk (dk d3k ) @fh dk @ 2 fh þ fjk þ 2 2 2 @h (dk h) (d3k h) dk h @h (dk h) (d3k h) @jk dk h @h@jk @ 2 fj @fjk 1 1 dk @ 2 fh dk d3k þ þ ¼ 2k þ fjk @h d3k h dk h @h dk h @h@jk (dk h)2 (d3k h)
2 hd d3k @fjk @ 2 fji dk (dk d3k ) @fh d3k X d3i (1)kþi þ 3k , k ¼ 1, 2 2 3 @j d h d h @j @j (dk h) (d3k h) k 3i 3k 3i 3k (d3k h) @j3k i¼1 0
(2:33) h
i r (r ~ f)
h
¼ ¼
hj2 @(hj1 fj1 ) @(hh fh ) hj1 @(hh fh ) @(hj2 fj2 ) 1 @ @ @h @j1 @j2 @h hj1 hj2 @j1 hj1 hh @j2 hj2 hh 2 X i¼1
X 2 1 @ hj3i @(hji fji ) @fh 1 @ 1 @hji @fji 1 @fh þ fji ¼ @h @ji hji hj3i @ji hji h @ji hji @h @h hji @ji i¼1 ji
" # 2 X @ 2 f ji 1 @fji 1 @hji @ 1 @hji fji @ @hji @ 1 @fh 1 @ 2 fh ¼ þ f ji þ þ @ji @h @ji hji @ji hji @j2i h @ji hji @h @ji hji @h hji @ji @h i¼1 ji " # 0 0 0 2 X @fji @ 2 fh di hdi di di @ 2 fji hdi di @fh d2i fj þ þ f ji þ ¼ di (di h) i di h @ji @h (di h)3 @ji (di h)2 @j2i (di h)2 @ji (di h)3 i¼1
(2:34)
34
Surface Impedance Boundary Conditions: A Comprehensive Approach
The scale factor, D, for tangential coordinates is naturally defined as D ¼ min (d1 , d2 ) so that ~jk ¼ jk =D;
~ dk ¼ dk =D, k ¼ 1, 2
(2:35)
Substituting Equation 2.35 into Equations 2.30 through 2.34, we obtain " (r ~ f )jk ¼ (1) d
k 1
# ~d3k @fj3k fj3k @fh ~ p ~ p , ~ ~d3k ~ph @h ~ ph ~ ~ @~j3k d3k ~
k ¼ 1, 2 (2:36)
p (r ~ f )h ¼ d1 ~
i¼1
" 2 X 1 ~ ~ p rf ¼d
" ¼d
2
~ p2 ~ p2
(1)i
~ d3i
@fji ~ ~ @ ~ ~j3i d3i ph
~ di
2 X @fji @fh 1 ~pfh þ ~ ~ ~ ~ ~h ~ @ ji @ h ph ~ di p l¼1 dl ~
i¼1
h i r (r ~ f)
2 X
(2:37) # (2:38)
jk
! ~dk @ 2 f jk @fjk @ 2~fh 1 1 þ~ p þ 2 þ~ p ~ ~ @h ~ ~ ~ @~jk ph ~ ~ ph ~ @h ~ @h dk ~ph d3k ~ dk ~
~k ~ ~ ~ @fh d3k d3k ) dk ~ dk ( d f jk ~ p2 2 2 ~h (~ dk p ~ ) (d3k ~ ph ~) ph ~ ) (d3k ~ph ~ ) @~jk (~ dk ~ ~ d3k
2 X
~ ph ~ d3k ~
i¼1
(1)kþi
0 h~d3k d~3k @fjk @ 2 f ji þ ~p3 , k ¼ 1,2 ~ ph ~ @~j3i @~j3k (~d3k ~ph ~ )3 @~j3k d3i ~
~ d3i
(2:39) " 2 h i X 2 ~ p ~ p r (r f ) ¼ d ~ h
0 ~d0 ~ @fji h ~ ~di di 2 2 i ~ ~ p f þ p f ji j 2 ~ 3 i ~ ~di ~ph ~di (~di ~ph ~ @ j ( d p h ~ ) ( ~ ) ~) i i i¼1 # 0 ~ ~d2 @ 2 fji @ 2 fh h ~~ di ~ di di @fh 2 i þ þ~ p ~p (2:40) ~i p ~ ~h ~ ph ~ @~ji @ h (d ~ )3 @~ji (d~i p~h ~ )2 @ ~j2 di ~
i
Obviously, these derivations are valid only for smooth surfaces (the cases of corners and edges will be considered separately in Section 3.7). In most practically important problems, the radius of curvature varies so slowly
General Perturbation Approach to Derivation of SIBCs
35
along the surface that the terms containing ~d0i can be neglected in further derivations, and Equations 2.39 through 2.40 are reduced to the following form: " ! h i ~dk @ 2 f jk @fjk @ 2~fh 1 1 2 ~ þ ~p þ 2 þ~ p r (r f ) ¼ d ~dk ~ph ~ jk @h ~ ~ ~ @ ~jk ph ~ ~ ~ @h ~ @h d3k ~ dk ~ph ~ ~dk (~dk ~d3k ) @fh d3k dk ~ fjk ~p2 2 ~h (~ dk p ~ ) (d3k ~ ph ~) ~ )2 (d3k ~ph ~ ) @~jk (~dk ~ph # 2 2 ~ ~ X @ f d d j 3i 3k i ~ p2 (1)kþi (2:41) ~ ~ ~h ~ i¼1 ~ @~j3i @~j3k d3k p d3i p~h
~ p2
" 2 h i X 2 ~ p ~ p r (r f ) ¼ d ~ h
i¼1
~ ~di @fji @ 2 fji @ 2 fh di d~2i ~ þ p ~h ~ (~ di p ~ )2 @~ji ~ ~ @~ji @ h (~di ~ph ~ )2 @~j2i di ~ph
#
(2:42) The condition in Equation 2.1 allowed the representation of the vector operators in the form of asymptotic expansions in the small parameter, ~p, that is proportional to the ratio of the depth of penetration to the characteristic size of the conductor’s surface. It is a well-known fact that the tangential derivatives in the conductor’s skin layer are much smaller than the normal derivative [5]: @ @ ; @h @jk
@2 @2 @h2 @j2k
(2:43)
Equations 2.36 through 2.38 and Equations 2.41 and 2.42 not only illustrate Equation 2.43, but also facilitate estimation of the tangential and normal derivatives for a given thickness of the skin layer. Comparison of Equations 2.41 and 2.42 leads us to the following important conclusion: the normal component of the curl of the function ~ f in the conductor’s skin layer is much smaller than its tangential component. In the limiting case ~ p ¼ 0 (d ¼ 0), the normal component of curl, ~ f , vanishes.
2.3 Perturbation Technique Since the parameter ~ p is small, the following functions can be represented as expansions in the small parameter ~ p:
36
Surface Impedance Boundary Conditions: A Comprehensive Approach
1 ~ ph ~ dk ~
¼
~ dk
~ ph ~ dk ~
1 h ~ h ~2 þ~ p þ~ p2 þ ; ~ ~ ~ d3 dk d2
k ¼ 1, 2
(2:44)
h ~ h ~2 þ~ p2 þ ; k ¼ 1, 2 ~ ~ d2 dk
(2:45)
k
k
¼1þ~ p
k
Substituting the expansions in Equations 2.44 and 2.45 into Equations 2.36 through 2.38 and Equations 2.41 and 2.42, we obtain " ! ! # fj3k fj3k @fh 1 @fh k 1 @fj3k 2 ~ ~p h þ , ~p (r f )jk ¼ (1) d þ ~ þ ~d3k @~j ~ @h ~ d3k @~j3k d23k ~ 3k
p (r ~ f )h ¼ d1 ~
2 X
(1)i
i¼1
@fji @~j3i
1þ~ p
h ~
~ d3i
(2:46)
!
þ ~p2
k ¼ 1, 2
h ~2 þ ~d2
(2:47)
3i
" ! ! 2 2 2 2 X X X X @fji 1 1 @fji 1 @fh 2 2 ~ ~ þ~ p rf ¼d fh ~ fh dl þ~ p h ~ ~ ~ ~ @h ~ i¼1 @ ji i¼1 di @ ji l¼1 dl l¼1 !# 2 2 X X 1 @fji 3 2 3 ~ þ~ p h ~ fh d (2:48) l ~2 ~ i¼1 di @ ji l¼1 h
r (r ~ f)
h
i
" # " 2 2 @ 2 fj k @fjk X @fj X @ 2 fh @ 2 fh 1 2 ~ ~ þ~ p h di þ d2 ¼d 2 þ ~p ~ k ~~ d1 i þh k ~ jk @h ~ @h ~ i¼1 @h ~ i¼1 @h ~ @ jk @h ~ @~jk # ) 2 ~dk ~d3k ~dk ~d3k @fh X @ 2 fji þ , k ¼ 1, 2 fjk (1)kþi ~d2 ~d3k ~dk ~d3k @~jk @~j3i @~j3k i¼1 k
r (r ~ f)
(
2
i h
2 X
"
@ 2 f ji h ~ @ 2 f ji 1 @fji @ 2 fh þ~ p ~ ~ ~ ~ ~di @~ji @~j2i di @ ~ji @ h i¼1 @ ji @ h ! # 2 2 @ f @f @ f h ~ 2 2 j j h i i þ~ p2 h ~ þ ~ ~ ~ d2i @~ji @ h d2i @~ji ~di @~j2i
p ¼ d2 ~
!
(2:49)
(2:50)
Consider the practically important case in which the distribution of the function, ~ f , in a conductor with constant material properties can be described by the diffusion equation (Equation 2.2). Note that the displacement current was neglected in Equation 2.2, therefore its application area is restricted to good conductors. The more general case of lossy dielectrics should be considered using the so-called telegraph equation, which can be represented in the form: r (r ~ f ) þ sm
@~ f @ 2~ f þ em 2 ¼ 0 @t @t
(2:51)
37
General Perturbation Approach to Derivation of SIBCs
Transformations of Equations 2.2 and 2.51 using the perturbation technique are carried out in a similar fashion. To simplify the description, in this chapter we will consider Equation 2.2 only. The diffusion equation (Equation 2.2) must be supplemented with the boundary and initial conditions in the form: h ¼ 0 (conductor surface): f ¼ f b ;
h ! 1 (depth of conductor): f ¼ 0;
t ¼ 0: f ¼ 0
(2:52)
Here and below, the superscript ‘‘b’’ denotes values at the interface (boundary). A natural scale factor for time t is the quantity t defined as the ratio 2=v in the case of time-harmonic fields or the incident pulse duration in the case of transient sources. Therefore, we can introduce a ‘‘nondimensional time’’ ~t as follows: ~t ¼ t=t
(2:53)
Substituting Equation 2.53 into Equation 2.2 and taking into account Equation 2.1, we obtain sm @~ f @~ f r (r ~ f) ¼ ¼ d2 t @~t @~t
(2:54)
Replacing the operator r (r ~ f ) in Equation 2.54 with Equations 2.49 and 2.50, we get " # " 2 2 ~dk ~d3k @ 2 f jk @fjk X @ 2 fh @fj X @ 2 fh 1 2 ~ ~ 2 þ~ p ~ k þh ~ ~d1 f jk þ~ p h di þ d2 i k ~d2 ~d3k @h ~ @h ~ i¼1 @h ~ i¼1 @h ~ @~jk @h ~ @~jk k # 2 ~ @ 2 f ji @fj d3k @fh X dk ~ kþi (1) þ O(~ p3 ) ¼ k , k ¼ 1, 2 ~ ~ ~ ~ ~ @~t @ j3i @ j3k dk d3k @ jk
(2:55)
i¼1
~ p
2 X i¼1
"
@ 2 f ji h ~ @ 2 fji 1 @fji @ 2 fh þ~ p ~ @~ji @ h ~ ~ ~ di @~ji @ h di @~ji @~j2i
h ~ @ 2 f ji 2 @fji 2 @ 2 fh ~h ~ þp ~ ~ ~ d2i @~ji @ h d2i @~ji ~ di @~j2i 2
!
#
!
þ O(~p ) ¼ 3
@fh @~t
(2:56)
38
Surface Impedance Boundary Conditions: A Comprehensive Approach
Since the parameter ~ p is small, we represent the tangential and normal components of ~ f in the form of the asymptotic expansions in the parameter ~p: fjk ¼
1 X
~ pm ( ~ fm )jk ,
k ¼ 1, 2
(2:57)
m¼0
fh ¼
1 X
~ pm ( ~ fm )h
(2:58)
m¼0
where (~ f m)jk and (~ f m)h are unknown coefficients. Substituting Equations 2.57 and 2.58 into Equations 2.55 and 2.56, and equating coefficients of equal powers of ~ p, the following equations for the expansion coefficients are obtained m¼0 @ 2 (~ f0 )jk @(~ f 0 ) jk ¼0 2 @h ~ @~t
(2:59)
@(~ f0 )h ¼0 @~t
(2:60)
2 2 ~ @ 2 (~ f1 )jk @(~ f1 )jk @(~ f0 )jk X ~1 þ @ ( f0 )h ¼ d @h ~2 @h ~ l¼1 l @~t @~jk @ h ~
(2:61)
2 X @ 2 (~ f0 )ji @(~ f 1 )h ¼ ~ @~t ~ i¼1 @ ji @ h
(2:62)
m¼1
m¼2 2 2 ~ @(~ @ 2 (~ f2 )jk @(~ f2 )jk @(~ f1 )jk X f0 )h ~1 1 ~1 1 ~d1 þ @ ( f1 )h þ (~ ~d2 ~ þ d d ) d d f ¼ 0 j l k k 3k k 3k k @h ~2 @h ~ l¼1 @~t @ ~jk @ h ~ @~jk
2 X i¼1
(1)kþi
2 @ 2 (~ f0 )ji @(~ f0 )jk X @ 2 (~ f 0 )h ~ d2 þh ~ þh ~~ d1 i k @h ~ @~j3k @~j3i @~jk @ h ~
(2:63)
i¼1
" # 2 2 ~ 2 ~ ~ X @ 2 (~ f1 )ji @ ( f ) @( f ) @(~ f 2 )h @ ( f ) 0 0 0 j j h 1 1 i i þh ~~ di ~ di ¼ @~t @~ji @ h ~ @~ji @ h ~ @~ji @~j2i i¼1
(2:64)
39
General Perturbation Approach to Derivation of SIBCs
m ¼ 3 (for (~ f3 )h only) " 2 X @ 2 (~ f 2 ) ji @ 2 (~ f1 )ji ~1 @(~ f1 )ji @ 2 (~ @(~ f3 )h f 1 )h þh ~~ d1 d ¼ i i @~t @~ji @ h ~ @~ji @ h ~ @~ji @~j2i i¼1 !# f 0 ) ji f 0 ) ji 2 @ 2 ( ~ f0 )h ~d0i @(~ f0 )h h ~ @ 2 (~ 2 @(~ þ þh ~ ~d2 @~ji ~ ~ ~ ~ d2 @~ji @ h d2 @~ji di @~j2 i
i
i
(2:65)
i
Boundary and initial conditions for Equations 2.32, 2.34, and 2.36 are obtained by substituting Equations 2.30 and 2.31 into Equations 2.23 and 2.24: fmb )jk ; h ¼ 0: (~ f m ) jk ¼ ( ~ t ¼ 0: (~ fm )jk ¼ 0;
(~ f m )h ¼ ( ~ fmb )h ;
h ! 1: (~ fm )jk ¼ 0;
(~ fm )h ¼ 0
(~ fm )h ¼ 0; (2:66)
The formulations (Equations 2.59 through 2.65) with the boundary conditions (Equation 2.66) state that the field at any point inside the conductor is determined solely by the field at the geometrically closest point on the surface. Viewed in this way, it is not surprising that the expansions breakdown at h ~ dk when fields due to equidistant points on the surface interfere with each other. Taking into account that higher-order terms of the expansions reduce the approximation error, their derivation is not an easy task. Therefore, we restrict ourselves by keeping the first three nonzero terms in the expansions. Considering Equation 2.60 with the boundary condition h ! 1: (~ fm )h ¼ 0 leads us to the following very important conclusion: ‘‘the coefficient (~ f0 )h is equal to zero everywhere inside the conductor including its interface,’’ i.e. (~ f0 )h ¼ 0
(2:67)
Equations 2.60 and 2.67 reduce Equations 2.61 through 2.65 to the following form: m¼1 2 @ 2 (~ f1 )jk @(~ f1 )jk @(~ f 0 ) jk X ~d1 ¼ @h ~2 @h ~ l¼1 l @~t
(2:68)
2 X @(~ f 1 )h @ 2 (~ f0 )ji ¼ ~ ~ @t ~ i¼1 @ ji @ h
(2:69)
40
Surface Impedance Boundary Conditions: A Comprehensive Approach
m¼2 2 @ 2 (~ f2 )jk @(~ f2 )jk @(~ f 1 ) jk X @ 2 (~ f 1 )h 1 ~ ~ ~d2 ~d1 ~d1 þ ( f ¼ þ ) d 0 j l k k 3k k @h ~2 @h ~ l¼1 @~t @~jk @ h ~
2 X
(1)kþi
i¼1
2 @ 2 (~ f 0 ) ji @(~ f 0 ) jk X ~d2 þh ~ i @h ~ @~j3k @~j3i
(2:70)
i¼1
" # 2 2 ~ ~ X @(~ f 2 )h @ 2 (~ f1 )ji @ ( f ) @( f ) 0 0 j j 1 1 i i þh ~~ di ~di ¼ @~t @~ji @ h ~ @~ji @ h ~ @~ji i¼1
(2:71)
m ¼ 3 (for (~ f3 )h only) " 2 X @ 2 (~ f 2 ) ji h f1 )ji 1 @(~ f 1 ) ji @ 2 ( ~ @(~ f 3 )h f1 )h ~ @ 2 (~ þ ¼ ~ ~ ~ ~ ~ ~ ~ @t @ ji @ h ~ ~ @ j2i di @ j i @ h di @ j i i¼1 !# f0 )ji f0 )ji h ~ @ 2 (~ 2 @(~ þh ~ 2 2 ~ ~ ~ ~ di @ j i @ h di @~ji
(2:72)
Substituting Equation 2.69 into Equation 2.70 gives ðt X 2 2 @ 2 (~ f2 )jk @(~ f2 )jk @(~ f 1 ) jk X @ 4 (~ f0 )ji 1 ~ ~ ~d2 ~d1 ~d1 ¼ dt þ ( f ) d 0 j l k k 3k k ~ ~ ~2 @h ~2 @h ~ l¼1 @~t i¼1 @ jk @ ji @ h 0
2 X i¼1
(1)kþi
2 @ (~ f 0 ) ji @(~ f 0 ) jk X ~d2 þh ~ @h ~ i¼1 i @~j3k @~j3i 2
(2:73)
Simplification of Equation 2.73 gives ~ f 2 2 2 2 ~ ~ ~ X 0 @ 2 (~ f2 )jk @(~ f2 )jk @(~ f1 )jk X @ ( f ) @(~ f 0 )j k X 0 j jk d3k dk 1 k ~ ~2 d d þ þh ~ ¼ l i 2 2 ~ ~ ~ ~j @h ~ @h ~ l¼1 @ h ~ @~t @ d d d k k 3k i i¼1 i¼1 (2:74)
Note that the diffusion equations (Equations 2.59, 2.68, and 2.74) contain only the tangential components (~ f m ) jk . Calculation of the tangential components (~ fm )jk is a task of primary importance in the surface impedance concept because all other quantities (such as the normal component and tangential and normal derivatives) can be then derived as ‘‘postprocessing’’ operations using the (~ fm1 )jk obtained at previous step.
General Perturbation Approach to Derivation of SIBCs
41
2.4 Tangential Components We now apply the Laplace transform following the rule: fj (~s) ¼ k
1 ð
fjk (~t) exp (~s~t)d~t
(2:75)
0
In the Laplace domain (LD), Equations 2.59, 2.68, and 2.74 are written in the form: m¼0 f @2 ~ 0 @h ~2
jk
f ~s ~ ¼0 0 jk
(2:76)
m¼1 f @2 ~ 1 @h ~2
jk
f @ ~ 2 X 0 jk ~ ~d1 ~s f 1 ¼ l jk @h ~ l¼1
(2:77)
m¼2 f @2 ~ 2
~ 2 ~ f f f f @ ~ @ ~ 2 2 @ 2 ~ ~ X X X 1 0 0 0 d d jk jk jk jk 3k jk k ~ ~d1 ~ f d2 ~ s ¼ þ þ h ~ 2 ~ ~ jk @h ~2 @h ~ l¼1 l @h ~ i¼1 i @~j2i dk dk ~ d3k i¼1
(2:78)
f Derivation of ~ 0
jk
The solution of Equation 2.76 with boundary conditions (Equation 2.66) can be written in the form: pffiffi ~ ~ f b exp (~ f ¼ h ~s) 0 0 jk
jk
(2:79)
Therefore, f @ ~ 0 @h ~
jk
pffiffi b pffiffi f ¼ ~s ~ exp (~ h ~s) 0 jk
(2:80)
42
Surface Impedance Boundary Conditions: A Comprehensive Approach
f Derivation of ~ 1
jk
Equation 2.77 can be represented in the following form using Equation 2.80: f @2 ~ 2 1 pffiffi~b X pffiffi jk ~ f f ~d1 exp (~ ~ ~ s ¼ s h ~s) (2:81) l 0 1 2 jk jk @h ~ l¼1 The solution of Equation 2.81 is 0 1 ~ f b 2 X 0 pffiffi jk B b ~ ~1 C f ¼@ ~ ~ h ~s) d f1 þ h l A exp (~ 1 jk jk 2 l¼1 f Calculation of the normal derivative of ~ yields 1 jk 2 3 ~ f f b @ ~ 2 X p pffiffi p ffiffi 0 ffiffi 1 jk jk 6 f b þ (1 h ~d1 7 ¼ 4 ~s ~ h ~s) ~ ~s) l 5 exp (~ 1 jk @h ~ 2 l¼1 f Derivation of ~ 2
(2:82)
(2:83)
jk
Substituting Equations 2.79 and 2.82 into Equation 2.78, we obtain 8 !2 ~ f b > < pffiffi X 2 2 X 0 p ffiffi jk jk ~ ~ 1 ~d1 þ (1 h ~ f f b ~ ~ ~ s ¼ s ~ s ) d l l 2 1 > jk jk @h ~2 2 : l¼1 l¼1 9 ~ 2 ~ f b f b > = 2 @ 2 ~ ~ X X 0 0 pffiffi pffiffi ~ jk jk d3k dk 2 b ~ ~ exp (~ h ~s) þ h ~ s f0 dn 2 ~j ~dk ~dk ~d3k > jk @ ; i n¼1 i¼1
f @2 ~ 2
(2:84)
The right-hand side of Equation 2.84 is transformed as follows: 8 2 3 !2 2 ~ f b > @ 2 2 2 < pffiffi X 1 X ~dk ~ X 0 d jk 3k ~ ~ b 1 b 1 ~ ~ 5 þ dl þ f 0 4 d ~s f 1 l 2 2~ ~ ~ > j j 2 @ j k k d d : i i¼1 l¼1 l¼1 k 3k 9 2 3 !2 > = 2 2 X pffiffi pffiffi~b 1 X 1 2 5 ~ ~ 4 exp (~ h ~s) h ~ ~s f 0 þ dl dn > jk 2 ; n¼1 l¼1 pffiffi ~ ~ 1 þ~ 2 ¼ W exp (~ h ~s) (2:85) h W jk
jk
43
General Perturbation Approach to Derivation of SIBCs
where
2 3 !2 2 ~ f b @ 2 2 2 ~d3k ~dk X 0 pffiffi ~b X 1 1 X ~1 jk ~ ~ b ~ 4 5 þ dl þ f 0 dl W 1 ¼ ~s f 1 2 2 ~ ~ ~ jk jk jk 2 @ j d d i i¼1 l¼1 l¼1 k 3k 2 ~ f b 2 @ 2 2 ~ ~ ~ X 0 pffiffi~b ~ þ 3 d d þ d d jk k 3k ~ b k 3k ¼ ~s f 1 þ f 0 (2:86) 2 2 2 ~ ~ ~ ~ jk ~ j @ j k 2dk d3k dk d3k i i¼1 2 3 !2 2 2 1 X X ffiffi p pffiffi~ 3~ d3k þ 3~ d2k þ 2~ dk ~ d23k ~ 1 2 5 ~ ~ f b 4 f b 2 ¼ ~s ~ ~ þ ¼ d d W s l n 0 0 jk jk 2 jk 2~ d2k ~ d23k n¼1 l¼1
(2:87)
The solution of Equation 2.84 can be represented in the form: 8 1 2 0 39 ~ 2 > W < = > pffiffi 1 j B C 6 ~ ~ ~ ~ f f b pffiffi 4h 1 þ pffiffi k A þ h 2 7 ~2 W h ~s) ¼ ~ @2 W 5 exp (~ 2 2 > jk jk jk jk > ~s 4 ~s : ; 8 2 > < h ~ 6 pffiffi b ~ b f pffiffi 4 ~s ~ ¼ f1 2 > j jk k 2 ~s :
3 2 ~ f b 2 @ ~dk þ ~d3k ~d2 þ 3~d2 X 0 jk 7 k 3k f b þ ~ 5 0 2 2 ~dk ~d3k ~ ~ jk 2~ @ j2i dk d3k i¼1
9 > = 3d~2 þ 2~d ~d þ 3~d2 2 2 2 3~ ~ ~ ~ pffiffi d þ 2 d þ 3 d d h ~ ~ h ~ k 3k k 3k ~ k 3k k 3k f b þ pffiffi f 0b exp (~ h ~s) þ 0 > jk jk 4 4 ~s 2~d2k ~d23k 2~d2k ~d23k ; 8 ! > < ~d þ ~d 2~d2k þ 6~d23k 3~d2k þ 2~dk ~d3k þ 3~d23k h ~ ~ k 3k ~ ~ f b f b f b þ h p ffiffi ~ þ þ ¼ 2 1 0 > jk jk 2~ jk 2 ~s dk ~d3k 4~d2k ~d23k 4~d2k ~d23k : 9 2 ~ f b > = 2 @ 3~d2 þ 2~d ~d þ 3~d2 X pffiffi 0 h ~ 2 ~ h ~ k 3k jk k 3k f b p ffiffi exp (~ h ~s) þ þ 0 2 2 2 ~ ~ ~ jk 4 @ ji > 2 ~s i¼1 2dk d3k ; 8 > < d~ þ ~d 3~d2 þ 2~d ~d þ 3~d2 k 3k k 3k ~ k 3k f b þ h f b f b ¼ ~ ~ þh ~2 ~ 2 1 0 ~ ~ > j j jk k k 2 d 8~ d2k ~d23k d : k 3k 39 2 2 ~ f b > 2 @ = ~d2 þ 2~d ~d 3~d2 X 0 pffiffi h ~ 6~ k 3k jk 7 k 3k exp (~ h ~s) þ þ pffiffi 4 f 0b 5 2 ~j > jk @ 2 ~s 4~d2k ~d23k ; i i¼1
(2:88)
Equations 2.79, 2.82, and 2.88 give the distributions of the coefficients f m, m ¼ 0, 1, 2, in the normal direction inside the conductor in the LD. Substitution of Equations 2.79, 2.82, and 2.88 into Equation 2.57 results in the LD distributions of the tangential components of the function f of different orders of approximation:
44
Surface Impedance Boundary Conditions: A Comprehensive Approach
First-order approximation (only the first nonzero term of the expansions is taken into account) pffiffi fj ¼ f b exp(~ h ~s) jk k
(2:89)
Second-order approximation (only the first and second nonzero terms of the expansions are taken into account) fj ¼ k
f b jk
þ
~ ph ~fjbk
~ d3k dk þ ~ ~ ~ 2dk d3k
!
pffiffi exp(~ h ~s)
(2:90)
Third-order approximation (three nonzero terms of the expansions are taken into account) " ~ 3~ d2 þ 2 ~ dk ~d3k þ 3~d23k d3k dk þ ~ fj ¼ f b þ ~ ph ~fjbk þ~ p2 h ~ 2fjbk k jk k 2~ dk ~ 8~ d2k ~d23k d3k !# 2 @ 2 2 2 ~ ~ ~ ~ X pffiffi fjbk þ 2 d 3 d d d ~ k 3k 2 h b k 3k þ~ p pffiffi fjk þ exp(~ h ~s) 2 ~ 2 ~s 2~ d2k ~ d23k i¼1 @ ji
(2:91)
In the case of time-harmonic fields, Equations 2.89 through 2.91 can be represented in the frequency domain (FD) replacing the transform variable ~s by 2j as follows: First-order approximation h f_jk ¼ f_jbk exp (~
pffiffiffiffi 2j) ¼ f_jbk exp((1 þ j)~ h)
(2:92)
Second-order approximation f_jk ¼
f_jbk
þ
~ ph ~ f_jbk
~ d3k dk þ ~ ~ ~ 2dk d3k
! exp((1 þ j)~ h)
(2:93)
Third-order approximation "
~ 3~ d2 þ 2 ~ dk ~ d3k þ 3~d23k d3k dk þ ~ ph ~ f_jbk þ~ p2 h ~ 2 f_jbk k f_jk ¼ f_jbk þ ~ 2~ dk ~ 8~ d2k ~ d3k d23k !# 2 @ 2 f_ b ~~ ~2 ~2 1X jk 21j b dk þ 2dk d3k 3d3k _ h ~ fjk þ þ~ p exp((1 þ j)~ h) (2:94) 2 2 i¼1 @~j2i 8~ d2k ~ d23k
45
General Perturbation Approach to Derivation of SIBCs
The transfer back to the time domain (TD) is performed using the inverse Laplace transform, changing the representation in Equations 2.92 through 2.94 to the following TD form: First-order approximation ~1 fjk ¼ fjbk * T
(2:95)
Second-order approximation ! ~ d3k b ~ dk þ ~ 1þ~ ph ~ fjk * T1 2~ dk ~ d3k
f jk ¼
(2:96)
Third-order approximation fjk ¼
! ~2 ~~ ~2 ~ d3k dk þ ~ 2 2 3dk þ 2dk d3k þ 3d3k ~1 1þ~ ph ~ þ~ p h ~ fjbk * T 2 2 ~ ~ ~ ~ 2dk d3k 8d d k 3k
þ~ ph ~ 2
fjbk
2 @2f b ~ dk ~ d23k 1 X d2k þ 2~ d3k 3~ jk þ 2 2 ~ ~ ~ 2 i¼1 @ j2i 8dk d3k
!
~ * T2
(2:97)
where the sign ‘‘*’’ denotes the nondimensional time convolution product defined as follows: ð~t ~ ~ w(t) * f(t) ¼ w(~t0 )f(~t ~t0 )d~t0 0
~ 1 and T ~ 2 , are obtained using the and the nondimensional TD functions, T following Laplace pairs [6]: pffiffi 1 ~3=2 h ~2 ~ 1 (~ ~ pffiffiffiffi h ~t ¼T exp h, ~t) (2:98) exp (~ h s) , 2 p 4~t pffiffi 1 pffiffi exp (~ h ~s) ~s
,
1 ~1=2 h ~2 ~ 2 (~ pffiffiffiffi t ¼T exp h, ~t) p 4~t
(2:99)
2.5 Normal Components The LD Equations 2.69, 2.71, and 2.72 can be represented in the form: 2 ~ f @ 2 X 0 ji ~ 1 f ¼ ~s (2:100) 1 ~ h @ j @ h ~ i i¼1
46
Surface Impedance Boundary Conditions: A Comprehensive Approach 3 2 ~ ~ f f f @ ~ @ 2 X 1 0 0 @ 6 h ~ ji ji ji 7 ~ 1 f þ ¼ ~s 4 5 2 ~ ~ ~ h @h ~ ~ di di @ h i¼1 @ ji
(2:101)
2 0 13 ~ f f f f f @ ~ @ ~ @ ~ @ ~ 2 X 2 1 1 0 1 @ h ~ h ~ 2 j j j h j 6 B C7 ~ ~ i i i i f f ¼ ~s1 þh ~@ þ 3 ~di ~ji ~ji 4 @ h ~di @ h ~d2 @ h ~d2 0 ji A5 h ~ ~ ~ @ @ i¼1 i i
(2:102)
From Equation 2.100: ~ f b @ 2 X 0 pffiffi ji ~ 1=2 f ¼ ~s exp (~ h ~s) 1 ~ h @ ji i¼1
(2:103)
From Equation 2.101: 3 2 ~ ~ f f 2 ~ ~ X 0 0 pffiffi di þ d3i ~ @ 6 1=2 ~ ji ji 7 ~ 1 1=2 1=2 f f f ¼ ~ s ~ s þ (1 ~ s h ~ ) ~ s h ~ h ~s) 4 5 exp (~ 2 1 0 ~ ~ ~ ~ ~ h ji ji di di 2di d3i i¼1 @ ji ¼ ~s1=2
" # 2 ~di þ 3~d3i ~ ~ X pffiffi @ ~ ~ 1=2 di d3i ~ f f f exp (~ h ~s) þ h ~ ~ s 1 0 0 ~ ~ ~ ~ ~ j j j i i i 2di d3i 2di d3i i¼1 @ ji
(2:104)
From Equation 2.102 (see details in Appendix 2.A.1): 8 " # > 2 ~d þ 3~d 3~ X d2i þ 6~ di ~ d23i d3i þ 11~ @ < 1=2 ~ i 3i ~ ~ 2 ~ f b f f b þ h f b ~ ¼ ~ þ h ~ s 3 2 1 0 ~> h ji ji 2~ ji di ~d3i 8~ d2i ~ d23i i¼1 @ ji : 13 2 0 2 ~ f b 2 @ d~ ~d 5~d2 2~d ~d 7~ 2 X 0 d 1 i 3i ji C7 B b 6 b i 3i i 3i þh ~@ ~ f0 ~s1 4 ~ f1 A5 2 2 2 ~ ~ ~ ~ ji 2~ j 2 @ j i di d3i 8di d3i l l¼1 39 2 f b f b f b > @2 ~ @2 ~ @2 ~ = ~d2 þ 2~d ~d 3~d2 0 0 0 pffiffi 1 1 j3i 7 i 3i ji ji 6 b i 3i exp (~ h ~s) ~s3=2 4 ~ þ f0 5 2 2 2 2 ~ ~ ~ ~ ~ ~ ji 2 @ ji 2 @ j3i @ ji @ j3i > 8di d3i ;
(2:105)
Substituting Equations 2.103 through 2.105 into Equation 2.58, we obtain First-order approximation fh ¼ ~ p
2 X i¼1
~s1=2
pffiffi @fjbi exp (~ h ~s) ~ @ ji
(2:106)
47
General Perturbation Approach to Derivation of SIBCs
Second-order approximation " # 2 ~ ~ ~di ~d3i X pffiffi @ þ 3 d d i 3i 1=2b f b fh ¼ ~s1=2 ~ f b þ ~ ~s) ~ ~ p p h ~ p s f exp (~ h j j j i i i ~ 2~ di ~ 2~di ~d3i d3i i¼1 @ ji (2:107) Third-order approximation ( 2 ~2 ~~ ~2 ~ ~ X @ 2 2b 3di þ 6di d3i þ 11d3i fh ¼ ~s1=2 ~ f b di þ 3d3i þ ~ f b þ ~ p p h ~ p h ~ f j j j i i i ~ 2~ di ~ 8~d2i ~d23i d3i i¼1 @ ji " !# b 2 @ 2 2 ~ ~ ~d2 ~ ~ ~ X f 5 d þ 2 d 7 d d 1 d i 3i j i 3i 3i i þ~ ph ~ fjbi i ~ p~s1=2 fjbi 2 l¼1 @~j2l 2~ di ~ 8~ d2i ~ d3i d23i " #) 2 b 2 b ~~ ~2 ~2 pffiffi @ 2fjb3i 1 @ fji 1 @ fji 2 1 b di þ 2di d3i 3d3i exp (~ h ~s) þ ~ p ~s fji 2 2 2 @~ji 2 @~j3i @~ji @~j3i 8~ d2i ~ d23i (2:108) The FD and TD versions of Equations 2.106 through 2.108 can be represented in the form: First-order approximation 1j f_h ¼ ~ p 2
2 @ f_ b X ji expð(1 þ j)~ hÞ ~ @ ji
(2:109)
i¼1
fh ¼ ~ p
2 @f b X ji ~ * T2 @~ji
(2:110)
i¼1
Second-order approximation 1j p f_h ¼ ~ 2
p fh ¼ ~
2 X @ ~ i¼1 @ ji
2 X @ ~ji @ i¼1
~ d3i 1 j _ b ~di d~3i di þ 3 ~ ph ~ f_jbi ~ p f_jbi þ ~ f 2 ji 2~di ~d3i 2~ di ~ d3i
"
!
1þ~ ph ~
! expð(1 þ j)~ hÞ
~ ~di ~d3i d3i b ~ di þ 3 ~ ~3 fjb * T fji * T2 ~p ~i d ~3i 2d 2~di ~d3i i
#
(2:111) (2:112)
48
Surface Impedance Boundary Conditions: A Comprehensive Approach
Third-order approximation ( 2 ~ X 3~d2 þ 6~di ~d3i þ 11~d23i 1 j @ d3i di þ 3~ p ph ~ f_jbi þ~ p2 h ~ 2 f_jbi i f_jbi þ ~ f_h ¼ ~ 2 i¼1 @~ji 2~ di ~ 8~d2i ~d23i d3i " !# 2 @ 2 f_ b 2 ~ ~ ~d2 ~ X 5 d þ 2 d 7 d 1 j _b ~ d3i 1 di ~ i 3i j 3i i þ~ ph ~ f_jbi i ~ p fji 2 2 l¼1 @~j2l 2~ di ~ 8~ d2i ~ d3i d23i " #) 2 b 2 b ~~ ~2 ~2 @ 2 f_jb3i 1 @ f_ji 1 @ f_ji 2 j _ b di þ 2di d3i 3d3i þ~ p exp ((1 þ j)~ h) þ f 2 ji 2 @~j2i 2 @~j23i @~ji @~j3i 8~ d2i ~ d23i (2:113) ( ! 2 ~2 ~~ ~2 ~ X @ d3i di þ 3~ 2 2 3di þ 6di d3i þ 11d3i ~2 fh ¼ ~ fjbi * T p 1þ~ ph ~ þ~ p h ~ ~ ~d2 ~d2 ~ ~ji @ 2 d 8 d i 3i i¼1 i 3i " !# 2 @2f b 2 2 ~ ~~ ~ ~ ~ 1X ji b di d3i b 5di þ 2di d3i 7d3i ~ ~ p fji þ~ ph ~ fji * T3 2 l¼1 @~j2l 2~ di ~ 8~ d2i ~ d3i d23i ! ) 2 b 2 b ~~ ~2 ~2 @ 2 fjb3i 1 @ fji 1 @ fji 2 b di þ 2di d3i 3d3i ~4 ~ p fji þ T 2 @~j2i 2 @~j23i @~ji @~j3i * 8~ d2i ~ d23i (2:114) ~ 4 have been obtained using the inverse Laplace ~ 3 and T Here the functions T transform: 1
~s
h ~ 1 ~ 3 (~ ~ , (pt) erfc pffiffiffiffi ¼ T h, ~t) (2:115) 2~t 2 h ~ h ~ ~ 4 (~ h, ~t) (2:116) h ~ erfc pffiffiffiffi ¼ T 2~t1=2 p1=2 exp 4~t 2~t
pffiffi exp (~ h ~s)
pffiffi ~s3=2 exp (~ h ~s)
,
and 2 erfc(x) ¼ 1 erf(x) ¼ pffiffiffiffi p
1 ð
exp (u2 )du
(2:117)
x
As follows from Equations 2.106 through 2.108, the normal component of a function governed by the equation of diffusion in the conductor’s skin layer (including surface) is much smaller than the tangential components and can be neglected in the zero-order approximation.
General Perturbation Approach to Derivation of SIBCs
49
2.6 Normal Derivatives The normal derivatives of the tangential components in the LD can be obtained directly from Equations 2.79, 2.82, and 2.88 and are written in the following forms: f @ ~ 0 pffiffi jk f b exp (~ (2:118) ¼ ~s1=2 ~ h ~s) 0 jk @h ~ " # f @ ~ ~ ~ pffiffi 1 þ d d jk k 3k ~ ~ ¼ ~s1=2 f 1b þ (1 h ~~s1=2 ) f 0b exp (~ h ~s) jk jk 2 ~ @h ~ dk ~ d3k " # ~ ~d þ ~d pffiffi d3k dk þ ~ k 3k ~ ~ 1=2 1=2 ~ b b b ¼ ~s ~ f0 ~s exp (~ h ~s) f1 þh f0 ~ ~ ~ ~ jk jk 2dk d3k jk 2dk d3k f @ ~ 2 @h ~
(2:119)
jk
8 > < pffiffi pffiffi ~dk þ ~d3k pffiffi 3~d2 þ 2~dk ~d3k þ 3~d2 k 3k f b þ (1 h f b f b ¼ ~s ~ ~ ~s) ~ þ (2~ hh ~ 2 ~s) ~ 2 1 0 > jk jk 2~ jk dk ~d3k 8~d2k ~d23k : 39 2 2 ~ f b > ~2 2 @ = 2 ~ ~ ~ X pffiffi 0 dk þ 2dk d3k 3d3k 1 h ~ 6 ~ jk 7 þ pffiffi exp (~ h ~s) þ 4 f 0b 5 2 2 2 ~ ~ ~ > jk @ ji 2 ~s 2 4dk d3k ; i¼1 8 > < ~d þ ~d 3~d2 þ 2~d ~d þ 3~d2 k 3k k 3k k 3k f b þ h f b f b ¼ ~s1=2 ~ ~ ~ þh ~2 ~ 2 1 0 > jk jk 2~ jk dk ~d3k 8~d2k ~d23k : 1 0 2 ~ f b 2 @ ~d þ ~d 5~d2 þ 2~d ~d þ 9~d2 X 0 h ~ k 3k jk C B b k 3k k 3k f b ~s1=2 @ ~ f1 þh ~ ~ A 0 2 2 ~ ~ ~ ~ jk 2~ j 2 @ j2i k dk d3k 8dk d3k i¼1 19 0 2 ~ f b > = 2 @ ~d2 þ 2~d ~d 3~d2 X 0 pffiffi 1 k 3k jk C B b k 3k exp (~ h ~s) f0 þ (2:120) ~s1 @ ~ A 2 2 2 > jk 2 @~j 8d~ d~ ; k 3k
i¼1
i
Similarly, we obtain the normal derivatives of the normal component in the LD from Equations 2.103 through 2.105: ~ b f @ ~ @ 2 X f0 j pffiffi 1 h i (2:121) ¼ exp ( h ~ ~s) ~ @h ~ @ ji i¼1 f @ ~ 2 @h ~
h
" # 2 ~ ~ X pffiffi @ ~ d3i 1=2 ~ d3i di þ 3~ di ~ 1=2 ~ f 1 þ (~ f0 ¼ h ~s ) f0 ~s exp (~ h ~s) ~ ~ ~ ~ji ji ji 2~ j @ i di d3i 2di d3i i¼1 " # 2 ~ ~ ~ ~ X pffiffi @ ~ þ 3 d þ d d d i 3i i 3i f b f b þ h f ~ ~ ~s1=2 ~ (2:122) exp (~ h ~s) ¼ 1 0 0 ~ ~ ~ ~ ~ j j j i i 2di d3i i di d3i i¼1 @ ji
50
Surface Impedance Boundary Conditions: A Comprehensive Approach
8 " > < 2 ~d þ 3~d X @ h i 3i ~ f b þ(~ f b ¼ h ~s1=2 ) ~ 2 1 ~ > j ji 2~ @h ~ i di ~d3i i¼1 @ ji :
f @ ~ 3
# 3~d2 þ 6~d ~d þ 11~d2 i 3i ~ i 3i b þ (~ h 2~s h ~) f 0 ji 8~d2i ~d23i 13 2 0 2 ~ f b 2 @ ~d ~d 5~d2 þ 2~d ~d 7~d2 X 0 1 i 3i ji C7 i 3i 6 b B b i 3i ~s1=2 4 ~ f1 þ (~ h ~s1=2 )@ ~ f0 A5 2 2 ~ ~ ~ ~ ~ ji 2di d3i ji 2 l¼1 @ j2l 8di d3i 39 2 f b f b f b > @2 ~ @2 ~ @2 ~ = d~2 þ 2d~ d~ 3d~2 0 0 0 pffiffi 1 1 j3i 7 i 3i ji ji 6 b i 3i exp(~ h ~s) ~s1 4 ~ f0 þ 5 2 2 2 2 ~ ~ ~ ~ ~ ~ > ji 2 2 @ ji @ j3i @ ji @ j3i ; 8di d3i 8 # >" 2 d~ þ 3d~ 3d~2 þ 6d~ d~ þ 11d~2 X @ < ~ ;b i 3i i 3i ~ 2 ~ i 3i b b ¼ f 2 þh ~ f1 þh ~ f0 ~> ji ji 2~ ji di ~d3i 8~d2i ~d23i i¼1 @ ji : 2
1=2
13 2 0 2 ~ f b 2 @ ~d þ ~d 11~d2 þ 14~d ~ 2 ~ X 0 d þ 15 d 1 i 3i ji C7 B ~ i 3i 1=2 6 ~ i 3i b b ~s þh ~@ f 0 4 f1 A5 ji ~ ji 2 l¼1 @~j2l di ~d3i d23i 8~d2i ~ 39 2 f b f b > @2 ~ @2 ~ = ~d2 þ ~d2 0 0 pffiffi j j3i 7 6 i 3i i f b exp(~ h ~s) ~s1 4 ~ þ (2:123) 5 0 2 ~2 d~2 j i 2d @~j @~ji @~j3i > ; i
3i
3i
Substituting Equations 2.121 through 2.123 into Equation 2.57 and representing the results in the FD and TD, we obtain First-order approximation @ f_jk ¼ (1 þ j)f_jkb exp ((1 þ j)~ h) (FD) @h ~
(2:124)
@ f_jbi expð(1 þ j)~ hÞ (FD) @~ji
(2:125)
2 X @ f_h ¼ ~ p @h ~ i¼1
@fjk d b ~ ¼ f T2 @h ~ d~t jk * 2 @f b X @fh ji ~ ¼ ~ p * T1 ~ @h ~ @ j i i¼1
(TD)
(2:126)
(TD)
(2:127)
Second-order approximation ~ @ f_jk d3k 1 j _ b ~dk þ ~d3k dk þ ~ ¼ (1 þ j) f_jbk þ ~ ph ~ f_jbk ~ p f @h ~ 2 jk 2~dk ~d3k 2~ dk ~ d3k
! exp ((1 þ j)~ h) (2:128)
51
General Perturbation Approach to Derivation of SIBCs
2 X @ f_h @ ¼ ~ p ~ji @h ~ @ i¼1
~ d3i 1 j _ b ~di þ ~d3i di þ 3~ ph ~ f_jbi ~ p f_jbi þ ~ f 2 ji ~di ~d3i 2~ di ~ d3i
! exp ((1 þ j)~ h) (2:129)
@fjk d ¼ @h ~ d~t
"
2 X @fh @ ¼ ~ p ~ji @h ~ @ i¼1
! # ~ ~dk þ ~d3k d3k b ~ dk þ ~ ~1 fjk * T2 þ ~p fjb * T 1þ~ ph ~ ~ ~3k 2 dk d 2~dk ~d3k k "
(2:130)
! # ~ ~di þ ~d3i d3i b ~ di þ 3~ b ~ 1þ~ ph ~ f T fji * T1 ~p ~di ~d3i ji * 2 2~ di ~ d3i
(2:131)
Third-order approximation " ~ 3~ d2 þ 2~dk ~d3k þ 3~d23k @ f_jk d3k dk þ ~ ¼ (1 þ j) f_jbk þ ~ ph ~ f_jbk þ~ p2 h ~ 2 f_jbk k @h ~ 2~ dk ~ 8~d2k ~d23k d3k
! 2 @ 2 f_ b 2 ~ ~ ~d2 ~ X 5 d þ 2 d þ 9 d 1 j _b ~ d3k h ~ dk þ ~ k 3k j 3k k ~ p þ~ ph ~ f_jbk k ~p fjk 2 2 i¼1 @~j2i 2~ dk ~ 8~ d2k ~ d3k d23k !# 2 @ 2 f_ b 2 2 ~ ~ ~ ~ X þ 2 d 3 d d d 1 k 3k j 2 j b k 3k k þ~ p þ expð(1 þ j)~ hÞ (2:132) f_ 2 jk 2 i¼1 @ ~j2i 8~ d2k ~ d23k
" 2 ~ X 3~d2 þ 6~di ~d3i þ 11~d23i @ f_h @ _b d3i di þ 3 ~ ¼ ~ p ph ~ f_jbi þ~ p2 h ~ 2 f_jbi i fji þ ~ ~ @h ~ 2~ di ~ 8~d2i ~d23i d3i i¼1 @ ji
! 2 @ 2 f_ b 2 ~ ~ ~d2 ~ X 11 d þ 14 d þ 15 d 1 j _b ~ d3i 1 di þ ~ i 3i j i 3i i ~ p þ~ ph ~ f_jbi ~ph ~ fji ~ 2 2 l¼1 @~j2l 8~ d2i ~ di ~ d3i d23i !# 2 2 ~ ~ @ 2 f_jb3i @ 2 f_jbi d þ d 2 j b i 3i þ 2 expð(1 þ j)~ hÞ (2:133) f_ þ~ p 2 ji 2~ @~j3i @ ~ji @~j3i d2i ~ d23i
@fjk d ¼ @h ~ d~t þ~ p
þ~ p2
"
! # 2 ~ ~ ~d2 ~ ~ 3 d þ 2 d þ 3 d d3k dk þ ~ k 3k 3k ~2 1þ~ ph ~ þ~ p2 h ~2 k fjbk * T 2~ dk ~ 8~ d2 ~d2 d3k k 3k
2 @2f b ~ 5~ d2 þ 2 ~ dk ~ d3k þ 9~d23k d3k 1X dk þ ~ jk þ~ ph ~ fjbk k ~ph ~ 2 2 2 ~ ~ ~ ~ ~ 2 @ j 2dk d3k 8dk d3k i i¼1 ! # 2 @2f b ~ dk ~ d23k 1 X d 2 þ 2~ d3k 3~ jk ~2 fjbk k þ T 2 i¼1 @~j2i * 8~ d2k ~ d23k
fjbk
! ~ * T1
(2:134)
52
Surface Impedance Boundary Conditions: A Comprehensive Approach
2 X @fh @ ¼ ~ p ~ji @h ~ @ i¼1
"
! ~2 ~~ ~2 ~ d3i di þ 3 ~ 2 2 3di þ 6di d3i þ 11d3i ~1 1þ~ ph ~ þ~ p h ~ fjbi * T ~i d ~3i 2d 8~d2 ~d2 i
3i
! 2 @2f b 2 ~ ~ ~d2 ~ ~ ~ X 11 d þ 14 d þ 15 d þ d 1 d i 3i j i 3i i 3i i ~2 ~ fjbi þ~ ph ~ fjbi ~ph ~ fjbi p T ~ 2 l¼1 @~j2l * 8~ d2i ~ di ~ d3i d23i ! # ~ 2 ~2 @ 2 fjb3i @ 2 fjbi 2 b di þ d3i ~ ~ p fji þ 2 (2:135) * T3 @~j3i @~ji @~j3i 2~ d2i ~ d23i
~ 2 , and T ~ 3 , are given ~1, T Here the nondimensional time-dependent functions, T in Equations 2.99, 2.98, and 2.115, respectively.
2.7 Components of the Curl Operator To simplify notations in the following derivations we denote curl ~ f as ~ g: r ~ f ¼~ g
(2:136)
~, Rewriting Equation 2.136 in local nondimensional coordinates, ~j1, ~j2, and h using Equations 2.46 and 2.47, we obtain " gj3k ¼ (1)
3k 1
d
! # @fjk fjk @fh f jk 1 @fh 2 ~ p ~ p h þ ~ þ þ , ~ ~ @h ~ dk @~jk d2k ~dk @~jk
k ¼ 1, 2 (2:137)
2 X @fji h ~ ~ pþ~ p2 (1)i þ gh ¼ d1 ~d3i @~j3i i¼1
(2:138)
The tangential and normal components of the function ~ g are now represented as expansions in the small parameter ~ p: g jk ¼
1 X m¼0
~ pm ( ~ g m ) jk ;
gh ¼
1 X
~ pm ( ~ gm )h , k ¼ 1, 2
(2:139)
m¼0
Substituting Equations 2.139 into Equations 2.137 and 2.138 and equating coefficients of equal powers of ~ p, the following equations are obtained (we restrict ourselves to the first three nonzero terms of the expansions):
53
General Perturbation Approach to Derivation of SIBCs
First-order approximation
(~ g0 )j3k ¼ (1)3k (~ g1 ) h ¼
2 X
@(~ f 0 ) jk , @h ~ (1)i
i¼1
k ¼ 1, 2
(2:140)
@(~ f0 )ji @~j3i
(2:141)
Second-order approximation
(~ g1 )j3k ¼ (1)
@(~ f1 )jk (~ f0 )jk @(~ f0 )h ~ @h ~ @~jk dk
3k
! ¼ (1)
3k
! @(~ f1 )jk (~ f 0 )j k , k ¼ 1, 2 ~dk @h ~ (2:142)
(~ g2 )h ¼
2 X i¼1
(1)
i
@(~ f 1 ) ji f0 )ji h ~ @(~ þ ~ ~ ~ @ j3i d3i @ j3i
! (2:143)
Third-order approximation "
(~ g2 )j3k
@(~ f 2 ) jk ( ~ f1 )jk @(~ (~ f 0 ) jk f1 )h f 0 )h 1 @(~ ¼ (1) h ~ þ 2 ~ ~ ~ ~ ~jk @h ~ @ j @ dk dk dk k " # ~ (~ f1 )jk @(~ (~ f0 )jk f1 )h 3k @( f2 )jk h ~ , k ¼ 1, 2 ¼ (1) ~ ~d2 @h ~ @~jk dk k
!#
3k
(~ g3 ) h ¼
2 X
(1)
i¼1
i
@(~ f2 )ji f 1 ) ji f0 )ji h ~ @(~ h ~ 2 @(~ þ þ 2 ~ ~ ~ ~ ~ @ j3i d3i @ j3i d3i @ j3i
(2:144)
! (2:145)
Representing Equation 2.140 through 2.145 in the LD and substituting Equations 2.79, 2.82, 2.88, and 2.103, we obtain the following expressions:
~ g0
j3k
pffiffi f b exp (~ ¼ (1)3k~s1=2 ~ h ~s) 0 jk pffiffi f b exp (~ ¼ (1)k~s1=2 ~ h ~s), k ¼ 1, 2 0 jk
(2:146)
54
Surface Impedance Boundary Conditions: A Comprehensive Approach
3 ~ f b d~ þ ~d pffiffi 0 jk 7 6 k 3k ~ 3k ~ 1=2 1=2 b b g1 )j3k ¼ (1) 4~s h ~s) (~ ~) f 0 f 1 þ(1 ~s h 5 exp (~ ~ ~ jk j k 2~ dk d3k dk 2
"
# d~ ~d ~d þ ~d pffiffi k 3k k 3k ~ ~ ~ 1=2 b b b ¼ (1) f1 þ f0 exp(~ h ~s) ~s ~s h ~ f0 ~ ~ ~ jk jk 2~ j k 2dk d3k dk d3k ! ~d þ ~d ~d ~d pffiffi k 3k k 3k ~ k 1=2 ~ 1=2 ~ b b b ~ ¼ (1) s ~ f0 ~s f 1 þh f0 exp (~ h ~s), k ¼ 1,2 jk jk 2~ jk 2~ dk ~d3k dk0 ~d3k 3k
1=2
8 > < ~d þ ~d 3~ d3k þ 3~ d2k þ 2~ dk ~ d23k k 3k f b f b þ h f b (~ ~ ~ þh ~2 ~ g2 )j3k ¼ (1)k~s1=2 ~ 2 1 0 > jk jk 2~ jk d23k dk ~d3k 8~ d2k ~ :
(2:147)
13 0 2 ~ b 2 @ f0 5~d2 þ 2~ 2 ~ ~ ~ ~ X d d þ 3 d d d 1 k 3k ji C7 B ~ 3k ~ 1=2 6 k k 3k b b ~@ f 0 þ f 1 h ~s 4 ~~ A5 jk jk 2 @~j2 d2 2dk d3k 2~ d2 ~ 2
k 3k
l¼1
l
19 0 f b f b f b > @2 ~ @2 ~ @2 ~ = ~d2 þ 2~d ~d 3~d2 0 0 0 1 1 k 3k j3k C jk jk B b k 3k ~s1 @ ~ þ f0 A 2 2 2 2 ~ ~ ~ ~ ~ ~ jk 2 @ jk 2 @ j3k @ jk @ j3k > 8dk d3k ; pffiffi exp(~ h ~s),
k ¼ 1, 2
(~ g1 ) h ¼
2 X
(1)i
f b @ ~ 0
ji
(2:148) pffiffi exp (~ h ~s)
(2:149) @~j3i 1 0 ~ ~ ~ f b f b f @ @ @ 2 ~ ~ X 1 0 0 pffiffi þ d h ~ d j j ji C B i 3i i i i exp (~ h ~s) (~ g2 ) h ¼ (1) @ þh ~ þ A ~i d ~3i @~j3i ~d3i @~j3i @~j3i 2d i¼1 i¼1
¼
2 X i¼1
1 0 f b f b @ ~ @ ~ ~ ~ 1 0 pffiffi 3di þ d3i ji ji C iB (1) @ þh ~ exp (~ h ~s) A @~j3i @~j3i 2~ di ~ d3i
(2:150)
2 3~ 11~d2 þ 6~d ~d þ 3~d2 @ 6~ di þ ~ d3i i 3i i 3i f b f b (~ g3 )h ¼ (1)i ~ ~ þh ~2 ~ 4 f 2b þ h 1 0 2~ 2 ~ ~ ~ ~ j j j @ j i i i 2 d 8 d d d i 3i 3i i¼1 i 3i 13 0 2 ~ b 2 @ f0 ~ pffiffi di ~ d23i 1 X d2i þ 2~ d3i 3~ ji C7 1=2 B ~ b þh ~~s h ~s) (2:151) þ @ f0 A5 exp(~ 2 2 2 ~ ~ ~ ji 2 @ jl 8di d3i l¼1 2 X
Details of derivation of Equations 2.148 and 2.151 are given in Appendix 2.A.1. ) given in Equations 2.146 through 2.148, ~ g Substituting the coefficients ( ~ m jk
~ gm )h given in Equations 2.149 through 2.151 into the expansions in and ( ~
55
General Perturbation Approach to Derivation of SIBCs
~g of Equation 2.139, we obtain the tangential and normal components of ~ various orders of approximation: First-order approximation pffiffi h ~s), gj3k ¼ (1)k~s1=2fjbk exp (~ gh ¼ ~ p
2 X i¼1
(1)i
k ¼ 1, 2
(2:152)
pffiffi @fjbi exp (~ h ~s) ~ @ j3i
(2:153)
Second-order approximation gj3k ¼ (1) ~s
k 1=2
gh ¼ ~ p
2 X i¼1
~ ~dk ~d3k d3k dk þ ~ f b þ ~ ph ~ fjbk ~ p~s1=2fjbk jk 2~ dk ~ 2~dk ~d3k d3k
(1)i
@ 3~ di þ ~ d3i b f b þ ~ p h ~ fj j i ~ ~ ~ @ j3i 2di d3i i
!
!
pffiffi exp (~ h ~s ) (2:154)
pffiffi exp (~ h ~s)
(2:155)
Third-order approximation ( ~ d3k þ 3~ 3~ d2 þ 2~ dk ~ d23k dk þ ~ d3k f b þ ~ gj3k ¼ (1) ~s ph ~fjbk þ~ p2 h ~ 2fjbk k jk 2 2 ~ ~ ~ ~ 2dk d3k 8dk d3k " !# 2 @ 2 ~2 ~~ ~2 ~ ~ fjbk 1X 1=2 dk d3k b b 5dk þ 2dk d3k þ 3d3k ~ f jk þ ~p~s fj h 2 l¼1 @~j2l d3k k d23k 2~dk ~ 2~ d2k ~ !) 2 b 2 b pffiffi ~2 ~~ ~2 @ 2fjb3k 1 @ fjk 1 @ fjk 2 1 b dk þ 2dk d3k 3d3k fjk þ exp ~ h ~s , k ¼ 1,2 ~p ~s 2 2 2 @~jk 2 @~j3k @~jk @~j3k 8~ d2k ~ d23k k 1=2
(2:156) " 2 ~2 ~~ ~2 ~ ~ X @ 2 2b 11di þ 6di d3i þ 3d3i f b 3di þ d3i þ ~ f b þ ~ gh ¼ ~ p (1)i p h ~ p h ~ f j ji ji @~j3i i 2~ di ~ 8~d2i ~d23i d3i i¼1 !# 2 @ 2 pffiffi ~~ ~2 ~2 fjbi 1X 2 1=2 b di þ 2di d3i 3d3i þ~ ph ~~s þ exp ~ h ~s (2:157) fji 2 l¼1 @~j2l 8~ d2i ~ d23i
Equations 2.152 through 2.157 give the tangential and normal components of ~ g in the LD as functions of the tangential components, fjbk , at the conductor’s surface. Finally, let us represent Equations 2.152 through 2.157 in the FD and TD: First-order approximation g_ j3k ¼ (1)k (1 þ j)f_jbk expð(1 þ j)~ hÞ
(2:158)
56
Surface Impedance Boundary Conditions: A Comprehensive Approach
g_ h ¼ ~ p
2 X
(1)i
i¼1
@ f_jbi exp ((1 þ j)~ h) @~j3i
gj3k ¼ (1)k p gh ¼ ~
2 X
(2:159)
d b ~ f T2 d~t jk *
(2:160)
@fjbi ~ * T1 @~j3i
(2:161)
(1)i
i¼1
Second-order approximation "
g_ j3k
# ~ d3k 1 j d~k d~3k _ b dk þ ~ ¼ (1) (1 þ j) 1 þ ~ ph ~ ~ p h) f exp ((1 þ j)~ 2 2~dk ~d3k jk 2~ dk ~ d3k k
g_ h ¼ ~ p
2 X
(1)i 1 þ ~ ph ~
i¼1
"
3~ di þ ~ d3i ~ ~ 2di d3i !
!
(2:162) @ f_jbi exp ((1 þ j)~ h) @~j3i
~dk ~d3k d b ~ ~1 fjk * T2 ~p fjb * T gjk ¼ (1) d~t 2~dk ~d3k k ! 2 X 3~ di þ ~d3i b ~ i @ gh ¼ ~ p (1) 1þ~ ph ~ fji * T1 @~j3i 2~di ~d3i i¼1 k
~ d3k dk þ ~ 1þ~ ph ~ ~ ~ 2dk d3k
(2:163) # (2:164)
(2:165)
Third-order approximation "
g_ j3k
~ 3~d2 þ 2~dk ~d3k þ 3~d23k d3k dk þ ~ ¼ (1) (1 þ j) f_jbk 1 þ ~ ph ~ þ~ p2 h ~2 k 2~ dk ~ 8~d2k ~d23k d3k
!
k
2 @ 2 f_ b 5~ d2 2 ~ dk ~ d3k 3~d23k 1 j _b ~ d3k 1X dk ~ jk þ~ ph ~ f_jbk k ~ph ~ ~ p fjk 2 2 l¼1 @~j2l 2~ dk ~ 2~ d2k ~d23k d3k !# 2 _b 2 _b 2 _b 2 2 ~ ~ ~ ~ @ f @ @ f f þ 2 d 3 d d d 1 j k 3k j j 2 j b 3k 3k k þ~ p þ 2k f_ k 2 jk 2 @~j2k @ ~j3k @~jk @~j3k 8~ d2k ~ d23k
exp ((1 þ j)~ h), k ¼ 1, 2
!
(2:166)
" 11~d2i þ 6~di ~d3i þ 3~d23i @ 3~ di þ ~ d3i g_ h ¼ ~ p (1) ph ~ f_jbi þ~ p2 h ~ 2 f_jbi f_jbi þ ~ @~j3i 2~ di ~ 8~d2i ~d23i d3i i¼1 !# 2 @ 2 f_ b 2 2 ~ ~ ~ ~ X þ 2 d 3 d d d 1 i 3i j 21j b i 3i i þ þ~ p h ~ f_ji exp ((1 þ j)~ h) (2:167) 2 2 l¼1 @~j2l 8~ d2i ~ d23i 2 X
i
General Perturbation Approach to Derivation of SIBCs
57
! ~2 ~~ ~2 ~ d3k d b ~ dk þ ~ 2 2 3dk þ 2dk d3k þ 3d3k ¼ (1) 1þ~ ph ~ þ~ ph ~ f T2 ~k d ~3k d~t jk * 2d 8~ d2k ~d23k ! 2 @2f b ~ 5~ d2k 2~ dk ~ d23k b d3k 3~ d3k b 1X dk ~ jk ~1 fj þ ~ ph ~ fjk ~ph ~ T ~ p 2 l¼1 @~j2l * 2~ dk ~ 2~ d2k ~ d3k k d23k ! 2 b 2 b ~~ ~2 ~2 @ 2 fjb3k 1 @ fjk @ fjk 2 b dk þ 2dk d3k 3d3k ~ 2 , k ¼ 1, 2 ~ p fjk þ T 2 @~j2k @ ~j23k @~jk @~j3k * 8~ d2k ~ d23k "
gj3k
k
(2:168) ! 2 X 11~d2i þ 6~di ~d3i þ 3~d23i b ~ @ 3~ di þ ~ d3i p (1)i 1þ~ ph ~ þ~ p2 h ~2 fji * T1 gh ¼ ~ @~j3i 2~ di ~ 8~d2i ~d23i d3i i¼1 ! # 2 @2f b ~~ ~2 ~2 1X ji 2 b di þ 2di d3i 3d3i ~ ~ fji þ T2 (2:169) þ~ ph 2 l¼1 @~j2l * 8~ d2i ~ d23i "
2.8 Surface Impedance ‘‘Toolbox’’ Concept The results obtained in the previous sections can be summarized into a set of formulae that can be then used for derivation of appropriate SIBCs for a given problem. Suppose the function ~ f obeys the equation of diffusion (Equation 2.2). Then its normal component, derivatives of tangential and normal components in direction normal to the surface, and tangential and normal components of curl ~ f at the interface (h ¼ 0) can be written in a form depending on the order of approximation in the LD, FD, and TD: Zero-order approximation (diffusion into the body is neglected) fh ¼ 0
@fh
b ¼0 @h ~ (r ~ f )bh ¼ 0
(2:170) (2:171) (2:172)
First-order approximation (only diffusion in the normal direction is considered; the surface of the body is assumed to be planar) 8 2 @ X fjbi > > 1=2 > ~ ~ p s LD > > ~ > > i¼1 @ ji > > > > < 2 @ f_ b 1j X ji b (2:173) fh ¼ ~ p FD ~ > 2 @ j > i i¼1 > > > > > 2 @f b X > > ji >~ ~ b TD > p T : ~ji * 2 @ i¼1
58
Surface Impedance Boundary Conditions: A Comprehensive Approach 8 1=2b
b > > ~s fjk LD @fjk
< (1 þ j)f_ b FD, k ¼ 1, 2 ¼ jk @h ~ > > : d fb T ~b TD d~t jk * 2 8 2 @ X fjbi > > > ~ p LD > > ~ > > i¼1 @ ji > >
> 2 @ f_ b X @fh
b < ji ~ ¼ FD p ~ji > @h ~ > @ i¼1 > > > > 2 X > @fjbi > > ~ b TD ~ > T e p : ~ji * 1 @ i¼1 8 > (1)k~s1=2fjbk LD > > < k b (r~ ~ f )bj3k ¼ (1) (1 þ j)f_jk FD, k ¼ 1, 2 > > > (1)k d f b T ~b : TD ~ j 2 * dt k 8 2 X @fjbi > > i > ~ p (1) LD > > > @~j3i > i¼1 > > > < X 2 @ f_jbi b ~ ~ (r f ) h ¼ ~ p (1)i FD > @~j3i > i¼1 > > > > 2 X > @fjbi > i > ~ b TD ~ > T (1) p : ~j3i * 1 @ i¼1
(2:174)
(2:175)
(2:176)
(2:177)
Second-order approximation (only diffusion in the normal direction is considered, the curvature of the surface of the body is taken into account) ! 8 2 ~d3i ~ X > @ d i > 1=2 b 1=2 b f ~ f > ~ p~s p~s LD > ji > ~ ~d3i ~ji ji > @ 2 d > i i¼1 > > ! > 2 < 1j X ~d3i ~ @ 1 j d i fhb ¼ ~ (2:178) p p f_ b ~ f_ b FD > 2 i¼1 @~ji ji 2 ji 2~di ~d3i > > > ! > > 2 ~ X > @ d3i b ~ b di ~ > b ~b > > p f T ~ p fj * T3 TD :~ ~ ji * 2 2~ di ~ d3i i i¼1 @ ji ! 8 ~ d3k dk þ ~ > 1=2 b 1=2 b > > ~s p~s fjk ~ fjk LD > > > 2~ dk ~ d3k > > !
> ~ @fjk
b < d3k dk þ ~ 1j b b _ _ ¼ (1 þ j) fj ~ p 2 fjk FD, k @h ~ > > 2~ dk ~ d3k > > > > ~ > d3k b ~ b dk þ ~ > > ~b þ ~ : d~ fjb * T f T TD p 2 dt k ~k d ~3k jk * 1 2d
k ¼ 1, 2
(2:179)
59
General Perturbation Approach to Derivation of SIBCs 8 2 X > > > ~ p > > > > i¼1 >
b > > < 2 X @fh
¼ ~ p @h ~ > > i¼1 > > > > 2 > X > > > ~ p : i¼1
~ ~ (r f )bj3k
~ @ b d3i di þ ~ p~s1=2fjbi fji ~ ~ @~ji di ~ d3i
!
@ @~ji
1 j _b ~ di þ ~d3i p f_jbi ~ fji ~ 2 di ~ d3i
@ @~ji
fjbi
LD !
~ ~ ~ b p di þ d3i f b T ~b j * 2 * T1 ~ ~ di ~ d3i i
(2:180)
FD ! TD
8 ! ~ ~ > d d > k 3k k > p~s1=2fjbk LD > (1) ~s1=2 fjbk ~ > > 2~ dk ~ d3k > > > > ! > < ~ ~ d d k 3k 1j k b b ¼ (1) (1 þ j) f_jk ~ p 2 f_jk FD, k ¼ 1, 2 > 2~ dk ~ d3k > > > " # > > > ~ > d3k b ~ b dk ~ > k d b ~b > fj * T 1 p TD > : (1) d~t fjk * T2 ~ 2~ dk ~ d3k k 8 2 X @fjbi > i > > ~ p (1) LD > > > @~j3i > i¼1 > > > < X 2 @ f_jbi b ~ ~ (r f ) h ¼ ~ p (1)i FD > @~j3i > i¼1 > > > > 2 > X @fjbi > i > ~ b TD > ~ T p (1) : ~j3i * 1 @ i¼1
(2:181)
(2:182)
Third-order approximation (diffusion in the normal and tangential directions is considered) 8 " 2 ~ X > @ b d3 i di ~ > 1=2 > ~ ~ p~s1=2fjbi p s fji ~ > > ~ ~ ~ji > @ 2 d d > i 3i i¼1 > > > !# > 2 b 2 b > 2 2 ~ ~ ~ ~ > @ 2fjb3i d d þ 2 d 3 d 1 @ fji 1 @ fji > i 3i 3i 2~1 b i > fji ~p s þ LD > > > 2 @~j2i 2 @~j23i @~ji @~j3i 8~ d2i ~ d23i > > > " > > 2 > X > @ _b 1 j _b ~ d3i di ~ > > ~p 1 j p fji ~ fji > > ~ < 2 i¼1 @ ji 2 2~ di ~ d3i fhb ¼ !# 2 b 2 b > ~2 ~~ ~2 @ 2 f_jb3i > 1 @ f_ji 1 @ f_ji > b di þ 2di d3i 3d3i 2j _ > þ f FD > þ~p > > 2 ji 2 @~j2i 2 @~j23i @~ji @~j3i 8~ d2i ~ d23i > > > " > > 2 > ~ X > @ b ~b d3i b ~ b di ~ > > p fji * T3 fji * T2 ~ > > ~p ~ ~ ~ > @ j 2 d d > i 3i i i¼1 > > ! # > > 2 b 2 b > ~2 ~~ ~2 @ 2 fjb3i > 1 @ fji 1 @ fji > 2 b di þ 2di d3i 3d3i b ~ > ~ p f þ T TD > ji : 2 @~j2i 2 @~j23i @~ji @~j3i * 4 8~ d2i ~ d23i
(2:183)
60
Surface Impedance Boundary Conditions: A Comprehensive Approach
8 " ~ > d3k dk þ ~ > 1=2 > f b ~ > ~s p~s1=2 fjbk > jk > ~ ~ > 2dk d3k > > !# > > 2 @ 2 2 > ~ ~ fjbk > d d3k 3~ þ 2 dk ~ d23k 1 X > 2~1 b k > f jk ~p s þ LD > > 2 i¼1 @~j2i > 8~ d2k ~ d23k > > " > > > > 1 j _b ~ d3k dk þ ~ > p f
b > > (1 þ j) f_jbk ~ ~ 2 jk 2 ~ @fjk
< dk d3k (2:184) !# @h ~ > 2 @ 2 f_ b ~k d ~2 ~2 þ 2d ~3k 3d > d j 1X > jk b k 3k 2 > _ > FD, k ¼ 1, 2 þ~p þ f > > 4 jk 2 i¼1 @~j2i 8~ d2k ~ d23k > > > > > ~ > d b ~b d3k b ~ b dk þ ~ > > p fjk * T1 fjk * T2 þ ~ > > ~ ~ ~t > d 2 d d > k 3k > ! > > 2 @2f b > ~ > dk ~ d23k 1 X d2k þ 2~ d3k 3~ jk > b ~ b TD > T þ f jk þ > : 2 i¼1 @~j2i * 2 d23k 8~ d2k ~ " 8 ~ 2 @ > d3i di þ ~ > ~p P f b ~ > p~s1=2 fjbi > j > i ~ ~ ~ > @ j d d i¼1 > i 3i i > > !# > > 2 ~ ~ ~2 ~ > @ 2fjb3i @ 2fjbi 3 d þ 2 d d > i 3i þ d3i i 2~1 b > ~ p þ LD s f > ji > ~2 ~2 d > @~j23i @~ji @~j3i 8d > i 3i > > " > > 2 > 1 j _b ~ d3i di þ ~ > > ~p P @ f_ b ~ > p fji
b > j i < ~ ~ ~ 2
@ j d d i¼1 @fh i 3i i !# ¼ (2:185) 2 ~ ~ @h ~ > ~ @ 2 f_jb3i > 3 d þ 2 d þ~ d23i @ 2 f_jbi d j i 3i > b i 2 _ > FD þ~p þ 2 f > > 2 ji > @~j3i @~ji @~j3i 8~ d2i ~ d23i > > " > > > ~ 2 @ > P di þ ~ d3i b ~ b > ~b ~ > fjbi * T ~p p f T > 1 > ~i d ~3i ji * 2 ~ji > @ d i¼1 > > > ! # > > ~2 ~~ ~2 > @ 2 fjbi @ 2 fjb3i > 2 b 3di þ 2di d3i þ d3i > ~b > T TD ~ p f þ : ji * 3 @~j23i @~ji @~j3i 8~ d2i ~ d23i " 8 ~ > d3k dk ~ > (1)k~s1=2 f b ~ > p~s1=2 fjbk > jk > ~ ~ > 2dk d3k > > > !# > 2b 2 b > 2 ~ ~ ~ ~2 > @ 2fjb3k d d þ2 d 1 @ fjk > k 3k 3d3k 1 @ fjk 2~1 b k > ~p s þ LD fjk > > > 2 @~j2k 2 @~j23k @~jk @~j3k d23k 8~ d2k ~ > > > " > > > 1 j _ b ~ d3k dk ~ > > > (1)k (1 þj) f_jbk ~ p fjk > > ~ ~ 2 > 2dk d3k < b !# ~ ~ 2 b 2 b ¼ r f ~k d ~2 ~2 þ 2d ~3k 3d @ 2 f_jb3k d j 1 @ f_jk 1 @ f_jk b j3k > k 3k 2 _ > FD, þ > > þ~p 2 fjk 2 @~j2k 2 @~j23k @~jk @~j3k > 8~ d2k ~ d23k > > " > > > ~k d ~3k > d > k d b ~b ~b > T f (1) fjb * T ~ p > j 2 1 * > k ~ ~ ~ > d t 2dk d3k k > > > ! # > > ~~ ~2 ~2 > @ 2 fjbk 1 @ 2 fjbk @ 2 fjb3k > 2 b dk þ 2dk d3k 3d3k 1 b > ~ > T ~p fjk þ TD > > 2 @~j2k 2 @~j23k @~jk @~j3k * 2 8~ d2k ~ d23k > > : k ¼ 1,2
(2:186)
61
General Perturbation Approach to Derivation of SIBCs 8 2 X @fjbi > i > > ~ p (1) LD > > > @~j3i > i¼1 > > > < X 2 @ f_jbi b ~ ~ (r f ) h ¼ ~ p (1)i FD > @~j3i > i¼1 > > > > 2 > X @fjbi > > ~ b TD >~ T p (1)i : ~j3i * 1 @ i¼1
(2:187)
Here we used the following:
~ b (~t) ¼ jT ~ 1 (~ ~ 0 (~t) T h, ~t) h~ ¼0 ¼ U 1
(2:188)
~ 2 (~ ~ b (~t) ¼ jT h, ~t) h~ ¼0 ¼ (p~t)1=2 T 2
(2:189)
~ b (~t) ¼ jT ~ 3 (~ ~ ~t) T h, ~t) h~ ¼0 ¼ U( 3
(2:190)
~ b (~t) ¼ jT ~ 4 (~ T h, ~t) h~ ¼0 ¼ 2~t1=2 p1=2 4
(2:191)
where U(t) and U0 (t) denote the Heaviside function and its derivative (the Dirac function), respectively. The return to dimensional variables can be performed by substituting Equations 2.19, 2.20, 2.35, and 2.53 into Equations 2.173 through 2.191. For example, the FD and TD versions of Equation 2.173 are 2 2 @ f_ b X @ f_jbi d X ji 1=2 ¼ (2j) d f_hb ¼ (2j)1=2 @j D i¼1 @(ji D1 ) i i¼1
fhb ¼
(2:192)
ðt ðt 2 2 X d X @ t t0 1=2 0 1 @ b 0 1=2 f (t ) p d(t t ) ¼ dt fjbi (t0 )(p(t t0 ))1=2 dt0 t D i¼1 @(ji D1 ) ji @j i i¼1 0
¼ dt1=2
2 @f b X j i
i¼1
0
b
T ; @ji * 2
T2b (t) ¼ (pt)1=2
(2:193)
Taking into account the definition of skin depth in the FD and TD 8 sffiffiffiffiffiffiffiffiffiffi > > 2 > > < vsm d ¼ rffiffiffiffiffiffiffi > > t > > : sm
FD (2:194) TD
62
Surface Impedance Boundary Conditions: A Comprehensive Approach
casts Equations 2.193 and 2.194 into the following form: f_hb ¼ ( jvsm)1=2
2 @ f_ b X ji i¼1
fhb ¼ (sm)1=2
2 @f b X ji i¼1
(2:195)
@ji
b
T @ji * 2
(2:196)
Transforming Equations 2.174 through 2.191 in a similar way, and taking into account that ðt f (t
0
)T1b (t
0
0
0
ðt
t )dt ¼ f (t0 )U 0 (t t0 )dt0 ¼ f (t)
(2:197)
0
and
dTb (t) d b ^ b (t) f (~ r, t) þ T2b (0)f (~ r, t) ¼ 2 * f (~ r, t) ¼ T T2 (t) * f (~ 2 * r, t) (Duhamels theorem) dt dt (2:198)
the following approximate boundary conditions are finally obtained: First-order approximation
fhb
¼
8 2 @ f_ b X > > ji > > (jvsm)1=2 > < @j i i¼1 2 @f b > X > ji > 1=2 b > (sm) > * T2 : @j i i¼1
FD (2:199) TD
8
b < (jvsm)1=2 f_jbk FD
@fjk ¼ , k ¼ 1, 2 @h : (sm)1=2 f b T ^ b TD jk * 2 8 2 @ f_ b X > > ji > FD
b > > < @j i @fh
i¼1 ¼ 2 @f b @h > X > ji > > TD > : @j i i¼1
(2:200)
(2:201)
63
General Perturbation Approach to Derivation of SIBCs
(r ~ f )bj3k ¼
8 < (1)k (jvsm)1=2 f_jb
FD
k
: (1)
(r ~ f )bh ¼
k
(sm)1=2 fjbk
^b * T2
8 2 X > @ f_jbi > > > (1)i > < @j3i
,
TD
k ¼ 1, 2
FD
i¼1
2 > X @fjbi > > i > (1) > : @j i¼1
(2:202)
(2:203) TD
3i
Second-order approximation 8 2 X > @ _b > 1=2 1=2 di d3i _ b > > FD (jvsm) (jvsm) f f > < 2di d3i ji @ji ji i¼1 b fh ¼ 2 > X > @ > 1=2 1=2 di d3i b b b b > TD f T (sm) f T > : (sm) 2di d3i ji * 3 @ji ji * 2 i¼1 8 d þ d3k _ b
b > > ( jvsm)1=2 f_jbk þ k FD f 2dk d3k jk @fjk
< , k ¼ 1, 2 ¼ @h > > ^ b þ dk þ d3k f b TD : (sm)1=2 f b T jk * 2 2dk d3k jk 8 2 X > @ _b > 1=2 di þ d3i _ b > > FD (jvsm) f f
> ji di d3i ji @fh
b < i¼1 @~ji ¼ 2 @h > > > X @ 1=2 di þ d3i b b b > TD f (sm) f T > : ~ ji di d3i ji * 2 i¼1 @ ji
(r ~ f )bj3k
8 dk d3k _ b > k 1=2 _ b > ( jvsm) f f (1) FD > jk < 2dk d3k jk , ¼ > > k 1=2 ^ b dk d3k f b > TD : (1) (sm) fjbk * T jk 2 2d d
(2:204)
(2:205)
(2:206)
k ¼ 1, 2 (2:207)
k 3k
(r
~ f )bh
¼
8 2 X > @ f_jbi > i > > (1) > < @j3i
FD
i¼1
2 > X @fjbi > > i > (1) > : @j i¼1
3i
(2:208) TD
64
Surface Impedance Boundary Conditions: A Comprehensive Approach
Third-order approximation " 8 2 X > @ _b di d3i _ b 1=2 > > fji (jvsm)1=2 f (jvsm) > > @ji 2di d3i ji > > i¼1 > !# > > 2 _b 2 _b 2 _b 2 2 > > > (jvsm)1 f_ b di þ 2di d3i 3d3i 1 @ fji þ 1 @ fji @ fj3i > FD > ji < 2 @j2i 2 @j23i @ji @j3i 8d2i d23i b " fh ¼ 2 > X > @ b b di d3i b b 1=2 > > (sm) fji * T2 (sm)1=2 f T > > @j 2di d3i ji * 3 > i > i¼1 > ! # > > 2 b 2 b 2 b 2 2 > @ f @ f @ f d þ 2d d 3d 1 1 > j i 3i j j 1 3i 3i i i > > fjbi i þ T b TD : (sm) 2 ~j2i 2 @j23i @ji @j3i * 4 8d2i d23i (2:209) 8 dk þ d3k _ b 1=2 _ b > > fjk > (jvsm) fjk þ 2d d > > k 3k > > !# > > 2 @ 2 f_ b 2 2 > 1X > jk 1=2 _ b dk þ 2dk d3k 3d3k > > FD þ(jvsm) þ fjk > < 2 i¼1 @j2i 8d2k d23k
@fjk
b ¼ , @h > dk þ d3k b > 1=2 b ^ b > (sm) f þ f T > jk * 2 > > 2dk d3k jk > > ! > > 2 @2f b 2 2 > > 1X jk 1=2 > b dk þ 2dk d3k 3d3k > T b TD þ(sm) f þ : jk 2 i¼1 @j2i * 2 8d2k d23k
k ¼ 1, 2
8 dk þ d3k _ b 1=2 _ b > > fj k > (jvsm) fjk þ 2d d > > k 3k > > !# > > 2 @ 2 f_ b 2 2 X > d þ 2d d 3d 1 > j k 3k 1=2 k 3k k b > > FD f_jk þ(jvsm) þ > < 2 8d2k d23k @j2i
@fjk
b i¼1 ¼ @h > d þ d k 3k > 1=2 b ^b þ > (sm) fjbk * T f > 2 > > 2dk d3k jk > > ! > > 2 @2f b > d2k þ 2dk d3k 3d23k 1 X > jk 1=2 > b > þ(sm) fjk þ T b TD : 2 i¼1 @j2i * 2 8d2k d23k
(2:210)
(2:211)
" 8 dk d3k _ b > k > > f (1) (jvsm)1=2 f_jkb > > 2dk d3k jk > > > > > !# > > 2 b 2 b 2 2 > @ 2 f_jb3k > 1 @ f_jk 1 @ f_jk 1=2 _ b dk þ 2dk d3k 3d3k > > f FD (jvsm) þ > jk > 2 @j2k 2 @j23k @jk @j3k 8d2k d23k < b ~ , k ¼ 1,2 (r f )j3k ¼ " > > > > (1)k (sm)1=2 f b T b dk d3k b ^ > f > jk * 2 > 2dk d3k jk > > > > > ! # > > > @ 2 fjb3k > d2 þ 2dk d3k 3~ d23k 1 @ 2 fjkb 1 @ 2 fjkb 1=2 > b > TD fjkb k þ T : (sm) 2 @j2k 2 @j23k @jk @j3k * 2 8d2k d23k
(2:212)
General Perturbation Approach to Derivation of SIBCs
(r
~ f )bh
¼
8 2 X @ f_jbi > i > > (1) > < @j3i i¼1 2 >X
> > > :
(1)
i¼1
i
@fjbi
@j3i
65
FD (2:213) TD
Here we used the following: T2b (t) ¼ (pt)1=2
(2:214)
T3b (t) ¼ U(t)
(2:215)
T4b (t) ¼ 2t1=2 p1=2
(2:216)
b
^ b ¼ dT2 ¼ (4pt3 )1=2 T 2 dt
(2:217)
and the nondimensional convolution product was replaced with ðt t1 w(t0 )f(t t0 )dt0
(2:218)
0
The formulae we derived are applicable to such functions as the magnetic field, the electric field, and the magnetic vector potential (with Coulomb’s gauge) since they obey the diffusion equation (Equation 2.2). We emphasize that the application of ‘‘toolbox’’ formulae requires that the equations be represented in local nondimensional coordinates.
2.9 Numerical Example Consider an infinitely long straight conducting cylinder of circular cross section. In this case the coordinate j1 is directed along the cylinder as shown in Figure 2.2 so that d1 ¼ 1. The radius d ¼ d2 of the cylinder cross section is constant and is used as the characteristic scale D. Let the distribution of the time-harmonic function ~ f in the cylinder be described by the diffusion equation and assume no variation of ~ f in the j1-direction, i.e. @ ¼0 @j1
(2:219)
The problem is therefore considered one-dimensional in cylindrical coordinates.
66
Surface Impedance Boundary Conditions: A Comprehensive Approach
ξ1 ξ2 η
FIGURE 2.2 The test problem analyzed: infinitely long straight conducting cylinder of circular cross section.
d
n
Suppose that the function ~ f has only a circumferential component (analogous to the magnetic field in our current carrying infinitely long straight cylinder): ~ ej2 ¼ f~ ej2 f ¼ fj2~
(2:220)
Then the governing equation of diffusion is
@ 2 f_ 1 @ f_ f_ 2_ _
¼ f_ b þ þ q f ¼ 0; f r¼d @r2 r @r r2
(2:221)
where q2 ¼ jvsm ¼ 2j=d2 ;
q¼
pffiffiffiffiffiffiffiffi 2j=d ¼ (1 j)=d
(2:222)
and r is the radial coordinate in the cylindrical coordinate system. It varies from 0 (center of the cylinder) to d (radius of the cylinder). The exact solution of Equation 2.221 can be written in the form [7] J1 (qr) f_ (r) ¼ f_ b J1 (qd)
(2:223)
where J1 is the Bessel function of the first kind and first order. Note that d 1j qd ¼ (1 j) ¼ ~ p d
(2:224)
Turning to the dimensionless variable ~r ¼ r=d
(2:225)
General Perturbation Approach to Derivation of SIBCs
67
we represent Equation 2.223 in the form: h f_ (d~r) ¼ f_ b
J1
i
1j ~ ~p r
h
J1
i
1j ~p
(2:226)
where ~r varies from 0 (center of the cylinder) to 1 (surface of the cylinder). An approximate solution under the conditions of skin effect is given by Equations 2.92 through 2.94 which, in this case, reduce to the following forms: First-order approximation f_ (h) ¼ f_ b exp (h
pffiffiffiffiffiffiffiffiffiffiffi 1þj b _ jvsm) ¼ f exp h d
(2:227)
Second-order approximation h 1þj h f_ (h) ¼ f_ b 1 þ exp 2d d
(2:228)
Third-order approximation ! 2 _f (h) ¼ f_ b 1 þ h þ 3h þ pdffiffiffiffi h exp 1 þ j h 2d 8d2 d 2j 8d2
(2:229)
Introducing d as a scale factor for h h ~ ¼ h=d
(2:230)
we represent Equations 2.227 through 2.229 in the form: First-order approximation 1þj h ~ f_ (~ h) ¼ f_ b exp ~ p
(2:231)
h ~ 1þj b _f (~ _ h ~ exp h) ¼ f 1 þ ~p 2
(2:232)
Second-order approximation
Third-order approximation ! ~ p 1 3 2 1þj b _f (~ _ h ~ h) ¼ f 1 þ h ~þ h ~ þ pffiffiffiffi h ~ exp ~p 2 8 8 2j
(2:233)
68
Surface Impedance Boundary Conditions: A Comprehensive Approach
Note that the variables h ~ and ~r have the same scale factor, d, and are related as follows: ~r ¼ 1 h ~
(2:234)
Using Equation 2.234, we recast Equation 2.226 into the form: h f_ (~ h) ¼ f_ b
J1
1j ~p (1
h ~) h i
J1
i (2:235)
1j ~p
Figures 2.3 through 2.11 show distributions of the real part, imaginary part, and amplitude of the function f_ given by the approximate formulae (Equations 2.231 through 2.233) and the exact solution (Equation 2.235) for the parameter ~ p equal to 0.1, 0.3, and 0.5. The nondimensional variable, h ~ , varies from 0 (surface) to 0.5 (half the distance from the surface to the center of the cylinder). The boundary value, f_ b, is equal to 1. It is clearly seen that
Analytical Leontovich Mitzner Rytov
~ p = 0.1
1
. Re ( f )
0.8 0.6 0.4 0.2 0 –0.2
0
0.1
0.2
0.3
0.4
0.5
η/d FIGURE 2.3 Real part of f_ as a function of the distance from the cylinder surface, for the case ~p ¼ 0.1 and f_ b ¼ 1.
69
General Perturbation Approach to Derivation of SIBCs
0
~ p = 0.1
. Im ( f )
–0.1
Analytical Leontovich Mitzner Rytov
–0.2
–0.3 0
0.1
0.2
0.3
0.4
0.5
η/d FIGURE. 2.4 Imaginary part of f_ as a function of the distance from the cylinder surface, for the case ~p ¼ 0.1 and f_ b ¼ 1.
1
~ p = 0.1
Analytical Leontovich Mitzner Rytov
0.8
. |f|
0.6
0.4
0.2
0
0
0.1
0.2
0.3
0.4
0.5
η/d FIGURE 2.5 Amplitude of f_ as a function of the distance from the cylinder surface, for the case ~p ¼ 0.1 and f_ b ¼ 1.
70
Surface Impedance Boundary Conditions: A Comprehensive Approach
Analytical Leontovich Mitzner Rytov
~ p = 0.3
1
. Re ( f )
0.8 0.6 0.4 0.2 0 –0.2
0
0.1
0.2
0.3
0.4
0.5
η/d FIGURE 2.6 Real part of f_ as a function of the distance from the cylinder surface, for the case ~p ¼ 0.3 and f_ b ¼ 1.
0 Analytical Leontovich Mitzner Rytov
~ p = 0.3
. Im ( f )
–0.1
–0.2
–0.3
–0.4
0
0.1
0.2
0.3
0.4
0.5
η/d FIGURE 2.7 Imaginary part of f_ as a function of the distance from the cylinder surface, for the case ~p ¼ 0.3 and f_ b ¼ 1.
71
General Perturbation Approach to Derivation of SIBCs
1 Analytical Leontovich Mitzner Rytov
~ p = 0.3 0.8
. |f|
0.6
0.4
0.2 0
0.1
0.2
0.3
0.4
0.5
η/d FIGURE 2.8 Amplitude of f_ as a function of the distance from the cylinder surface, for the case ~p ¼ 0.3 and f_ b ¼ 1.
1 Analytical Leontovich Mitzner Rytov
~ p = 0.5
. Re ( f )
0.8
0.6
0.4
0.2 0
0.1
0.2
0.3
0.4
0.5
η/d FIGURE 2.9 Real part of f_ as a function of the distance from the cylinder surface, for the case ~p ¼ 0.5 and f_ b ¼ 1.
72
Surface Impedance Boundary Conditions: A Comprehensive Approach
0 Analytical Leontovich Mitzner Rytov
~ p = 0.5
. Im ( f )
–0.1
–0.2
–0.3
–0.4 0
0.1
0.2
0.3
0.4
0.5
η/d FIGURE 2.10 Imaginary part of f_ as a function of the distance from the cylinder surface, for the case ~p ¼ 0.5 and f_ b ¼ 1.
1
~ p = 0.5
Analytical Leontovich Mitzner Rytov
. |f|
0.8
0.6
0.4
0
0.1
0.2
0.3
0.4
0.5
η/d FIGURE 2.11 Amplitude of f_ as a function of the distance from the cylinder surface, for the case ~p ¼ 0.5 and f_ b ¼ 1.
73
General Perturbation Approach to Derivation of SIBCs
~ ¼ 0.1, the Leontovich’s SIBC provides suitable accuracy; in the case of p however with the increase in thickness of the skin layer (~p ¼ 0.3), higherorder SIBCs (Miztner’s and Rytov’s approximations) should be applied. When the skin depth equals half the cylinder radius (~p ¼ 0.5), the surface impedance concept yields significant error.
~ ~ f ~ Appendix 2.A.1: Calculation of ~ , g , and g~2 2 3 h
jk
h
The following notations are introduced to simplify further derivations: pffiffi ~ f ¼ a20 þ h ~ a21 þ h ~ 2 a22 þ h ~~s1=2 a23 exp ~ h ~s 2 ji
(2:A:1)
pffiffi ~ f ¼ (a10 þ h ~ a11 ) exp (~ h ~s) 1
(2:A:2)
pffiffi ~ f ¼ a00 exp (~ h ~s) 0
(2:A:3)
ji
ji
2 i X pffiffi @ h ~ 1=2 1=2 f ~ ~ ¼ s a þ h ~ a s a exp (~ h ~s) 10 24 25 2 ~ h i¼1 @ ji 2 X pffiffi @a00 ~ 1=2 f ~ ¼ s exp (~ h ~s) 1 ~ h i¼1 @ ji
(2:A:4)
(2:A:5)
where f b a20 ¼ ~ 2
a23
ji
(2:A:6)
~ ~ f b di þ d3i a21 ¼ ~ 1 j i 2~ di ~ d3i
(2:A:7)
3~ d2i þ 2~ di ~ d3i þ 3~d23i f b a22 ¼ ~ 0 ji 8~ d2i ~ d23i
(2:A:8)
2 ~ f b @ 2 ~ 2 2 ~ ~ ~ 0 di þ 2di d3i 3d3i 1 X ji ~ b ¼ f0 þ 2 2 ~ ~ ~ ji 2 l¼1 @ j2l 8di d3i ~ d3i di þ 3~ f a24 ¼ ~ 0 ~ ~ ji 2di d3i
(2:A:9)
(2:A:10)
74
Surface Impedance Boundary Conditions: A Comprehensive Approach d ~i ~ d3i f a25 ¼ ~ 0 ~ ~ ji 2di d3i f b a10 ¼ ~ 1
(2:A:11) (2:A:12)
ji
~ ~ f b di þ d3i a11 ¼ ~ 0 j i 2~ di ~ d3i f b a00 ¼ ~ 0
(2:A:13) (2:A:14)
ji
Equations 2.102, 2.144, and 2.145 can now be represented in the forms: 2 0 13 ~ f f f f f @ ~ @ ~ @ ~ @ ~ 2 X 2 1 1 1 0 @ h ~ h ~ 2 ji ji ji ji 6 B C7 h ~ ~ f f þ ¼ ~s1 þh ~@ A5 3 0 ~ ~ ~ ~ ~ 4 @h h ji ~ ~ ~ @~ji di @ h d2i @ h d2i di i¼1 @ ji ¼ ~s1
( 2 i X @ h 1=2 a20 þ h ~ a21 þ h ~ 2 a22 þ h ~~s1=2 a23 þ a21 þ 2~ ha22 þ~s1=2 a23 ~s ~ i¼1 @ ji þ
2 i 1 X h ~ h 1=2 @ 2 a00 ~s (a10 þ h ~ a11 ) þ a11 [a10 þ h ~ a11 ] ~s1=2 ~ji @~jl ~ ~ @ di di
" þh ~ a00 ~s1=2
1
¼ ~s
h ~ 2 ~ d2i ~ d2i
#)
l¼1
pffiffi exp ~ h ~s
( " # 2 2 X X @ @ 2 a00 ~s1=2 a23 ~ ~ ~ i¼1 @ ji l¼1 @ ji @ jl "
a11 a10 a11 2a00 þ ~ ha23 þ a21 þ 2~ ha22 þ h ~ h ~ h ~ ~ ~di ~ ~ di di d2 i
" ~ a21 h ~ 2 a22 h ~ þ ~s1=2 a20 h
a10 a11 a00 h ~2 h ~2 ~i ~ ~ d di d2
#
#)
pffiffi exp h ~ ~s
i
(
¼ ~s1
"
#
2 2 X X @ @ 2 a00 ~s1=2 a23 ~ ~ ~ i¼1 @ ji l¼1 @ ji @ jl " !# a10 2a00 þ a21 þ h ~ a23 þ 2a22 ~i ~d2 d i " !#) pffiffi a a11 a00 10 1=2 2 þ~s a20 h ~ a21 þ exp h ~ ~s h ~ a22 þ þ ~ ~ di di ~ d2i
(2:A:15)
75
General Perturbation Approach to Derivation of SIBCs 3 2 ~ ~ ~ ~ f f f f @ @ 2 1 1 0
jk jk h jk 7 3k 6 ~ g2 j ¼ (1) 4 h ~ 5 2 3k ~ ~dk ~ @h ~ @ jk dk " 3k
¼ (1)
~ a21 þ h ~ 2 a22 þ h ~~s1=2 a23 þ a21 þ 2~ ha22 ~s1=2 a20 þ h
# 2 X pffiffi 1 @ 2 a00 a00 1=2 þ~s a23 (a10 þ h ~ a11 ) ~s h ~ exp (~ h ~s) 2 ~ ~ ~ ~dk dk l¼1 @ jk @ jl " ! a a a 10 11 00 ~ a21 þ h ~ 2 a22 ) þ a21 h ~ a23 þ 2a22 þ þ ¼ (1)3k ~s1=2 (a20 þ h ~ ~ dk dk ~ d2k !# 2 X pffiffi @ 2 a00 exp (~ h ~s) (2:A:16) þ ~s1=2 a23 ~ ~ l¼1 @ jk @ jl 1=2
1 0 ~ ~ f f f @ ~ @ @ 2 2 1 0 h ~ h ~ ji ji ji C iB ~ ( g3 )h ¼ (1) @ þ þ A 2 ~ ~ ~ ~ ~ @ j3i d3i @ j3i d3i @ j3i i¼1 2 X
" # @ h ~ h ~2 2 1=2 ¼ (1) a20 þ h ~ a21 þ h ~ a22 þ h ~~s a23 þ (a10 þ h ~ a11 ) þ a00 ~ ~ @~j3i d23i d3i i¼1 " ! # 2 X a10 a11 a00 i @ 2 1=2 þh ~~s (1) a20 þ h ~ a21 þ þ a23 þh ~ a22 þ ¼ ~ ~ @~j3i d3i d3i ~ d23i i¼1 2 X
i
(2:A:17)
Substitution of Equations 2.A.6 through 2.A.14 into Equations 2.A.15 through 2.A.17 leads to the following results: 8 " # > < 2 ~d þ 3~d 3~d2 þ 6~d ~d þ 11~d2 X @ i 3i i 3i ~ ~ ~ 1=2 2 ~ i 3i b b b f f2 þh ¼ ~s ~ f1 þh ~ f0 3 ~> h ji ji 2~ ji di ~d3i 8~d2i ~d23i i¼1 @ ji : 13 2 0 2 ~ f b 2 @ ~d ~d 5~d2 þ 2~d ~d 7~d2 X 0 1 i 3i ji C7 i 3i 6 B ~ ~ 1 i 3i b b ~ þ s 4 f1 þh ~@ f 0 A5 ji 2~ ji 2 l¼1 @~j2l di ~d3i 8~d2i ~d23i 39 f b f b f b > @2 ~ @2 ~ @2 ~ = ~d2 þ 2~d ~d 3~d2 0 pffiffi 0 0 1 1 j3i 7 i 3i ji ji 6 b i 3i exp (~ h ~s) þ f0 þ ~s3=2 4 ~ 5 2 2 2 2 ~ ~ ~ ~ ~ ~ > ji 2 @ ji 2 @ j3i @ ji @ j3i ; 8di d3i 2
(2:A:18)
76
Surface Impedance Boundary Conditions: A Comprehensive Approach 2
~ g2 j
3k
6 ¼ (1)3k 4 ~s1=2
d~ þ ~d 3~d2 þ 2~d ~d þ 3~d2 k 3k k 3k ~ k 3k f b þ h f b f b ~ ~ þh ~2 ~ 2 1 0 jk jk 2~ jk dk ~d3k 8~d2k ~d23k
!
1 0 2 ~ f b 2 @ 5~d2 þ 2~d ~d þ 3~d2 ~ X 0 d3k ~ 1 dk ~ k 3k ji C B ~ k 3k f b h f b þ ~ þ @ A 1 0 jk jk 2 @~j2 2~ dk ~ 2~d2 ~d2 d3k l¼1
k 3k
l
13 0 f b f b f b @2 ~ @2 ~ @2 ~ ~d2 þ 2~d ~d 3~d2 0 0 0 pffiffi 1 1 j3k C7 k 3k jk jk B b k 3k h ~s) f0 þ~s1=2 @ ~ þ A5 exp(~ 2 2 2 2 ~ ~ ~ ~ ~ ~ jk 2 2 @ j @ j @ j @ j 8dk d3k k 3k k 3k
(2:A:19)
2
2 3~ 11~d2 þ 6~d ~d þ 3~d2 X
@ 6~ di þ ~ d3i i 3i i 3i f b f b ~ g3 h ¼ (1)i ~ ~ þh ~2 ~ 4 f 2b þ h 1 0 2~ 2 ~ ~ ~ ~ ji ji 2di d3i ji @ j 8 d d 3i i¼1 i 3i
13 2 ~ f b 2 @ ~ 2 2 ~ ~ ~ X 0 pffiffi di þ 2di d3i 3d3i 1 ji C7 1=2 B ~ b ~ þ þh ~s exp (~ h ~s) @ f0 A 5 2 ji 2 l¼1 @~jl 8~ d2i ~ d23i 0
(2:A:20)
At the conductor’s surface (~ h ¼ 0), Equations 2.A.18 through 2.A.20 reduce to the following formulae: ~ f 3
h
2 2 ~d ~d X @ 6 1=2 ~ i 3i f b f b þ ~ ¼ ~s1 ~s 4 2 1 ~ ji j i 2~ di ~d3i h ~ ¼0 i¼1 @ ji 13 0 f b f b f b @2 ~ @2 ~ @2 ~ d~2 þ 2~d ~d 3~d2 0 0 0 1 1 j3i C7 i 3i ji ji 1=2 B ~ i 3i b þ~s þ A5 @ f0 2 2 2 2 ~ ~ ~ ~ ~ ji 2 2 @ ji @ j3i @ j@~j3i 8di d3i
~ g2 j 3k
(2:A:21)
2
~ ~ 6 f b þ dk d3k ~ f b ¼ (1)3k 4 ~s1=2 ~ 2 1 jk jk h ~ ¼0 2~ dk ~ d3k 13 0 f b f b f b @2 ~ @2 ~ @2 ~ ~ 2 2 ~ ~ ~ 0 0 0 1 jk jk j3k C7 B b dk þ 2dk d3k 3d3k 1 þ~s1=2 @ ~ f0 þ A5 2 2 2 2 jk 2 @~jk 2 @~j3k @~jk @~j3k d3k 8~ dk ~
~ g3 h
h ~ ¼0
¼
2 X i¼1
(1)i
f b @ ~ 2 @~j3i
(2:A:22) ji
(2:A:23)
General Perturbation Approach to Derivation of SIBCs
77
References 1. S.M. Rytov, Calcul du skin-effet par la méthode des perturbations, Journal of Physics, 2(3), 1940, 233–242. 2. K.M. Mitzner, An integral equation approach to scattering from a body of finite, Radio Science, 2(12), 1967, 1459–1470. 3. M.A. Leontovich, On one approach to a problem of the wave propagation along the Earth’s surface, Academy of Science USSR, Series Physics, 8, 1944, 16–22. 4. M.A. Leontovich, On the approximate boundary conditions for the electromagnetic field on the surface of well conducting bodies, in B.A. Vvedensky (ed.), Investigations of Radio Waves, Academy of Sciences of the USSR, Moscow, 1948. 5. G.S. Smith, On the skin effect approximation, American Journal of Physics, 58(10), 1990, 996–1002. 6. M. Abramowitz and I.A. Stegun (Eds.), Handbook of Mathematical Functions with Formulas, Graphs and Mathematical Tables, Applied Math Series 55, National Bureau of Standards, Washington, DC, 1964. 7. K. Simonyi, Theoretische Elektrotechnik, Ver Deutscher Verlag, Berlin, 1956. 8. A.J. Mestel, More accurate skin-depth approximations, IMA Journal of Applied Mathematics, 45, 1990, 33–48.
3 SIBCs in Terms of Various Formalisms
3.1 Introduction Surface impedance boundary conditions are usually used in combination with a numerical method. Therefore, both the formulation of the numerical method and the SIBC should be represented in terms of the same variables. It is rather natural that boundary conditions should not dictate the selection of the numerical technique. Instead, the numerical method is selected first and boundary conditions must then be derived in the state variables of the numerical method. Numerical methods used in practice frequently operate in terms other than the electric and=or magnetic fields (for example, the magnetic vector or scalar potential formalisms are commonly employed). The method may require quantities such as normal and tangential components of a function or the function and its normal derivative being related at the interface. These types of relationships are obtained in this chapter using the ‘‘surface impedance toolbox.’’ Since most electromagnetic quantities in conductors are governed by the diffusion equation, it is natural to suppose they are mutually connected. Thus, another goal of this chapter is to derive general relations suitable for representation of various SIBCs in terms of different formalisms. In this sense, the discussion in this chapter forms the cornerstone of implementations that follow in Chapters 5 through 8.
3.2 Basic Equations Consider a homogeneous body of finite conductivity (m1 ¼ m, s1 ¼ s), surrounded by a nonconductive medium (e2 ¼ e0, m2 ¼ m0, s2 ¼ 0). The parameters of the conducting material are assumed to be constant. In this chapter, we restrict ourselves to good conductors so that displacement currents inside the body as well as surface charges can be neglected. The more general case of a lossy dielectric materials will be considered in Chapter 8. The electric ~ inside the body can be described by Maxwell’s field, ~ E, and magnetic field, H, equations: 79
80
Surface Impedance Boundary Conditions: A Comprehensive Approach
~ @H r ~ E ¼ m @t
(3:1)
~ ¼ s~ rH E
(3:2)
The exact boundary conditions at the surface of the conductor are ~ ~ nH
cond
~ ¼~ nH
diel
~ ; m1~ nH
cond
~ ¼ m2~ nH
diel
; ~ n ~ E
cond
¼~ n ~ E
diel
; ~ n ~ E
cond
¼~ n ~ E
diel
(3:3)
Let the time variation of the incident field be such that the electromagnetic penetration depth, d, into the body remains small compared with the characteristic dimension, D, of the surface of the body so that the condition in Equation 2.1 is satisfied.
3.3 Electric Field–Magnetic Field Formalism Equations 3.1 and 3.2 with the local coordinates ~j1, ~j2, and h ~ , introduced in Section 2.2, are written in the following form using Equations 2.46 and 2.47: " ! # @Hj3k @Ejk Ejk @Eh Ejk 1 @Eh 2 þ ¼ (1)k m ~p , k ¼ 1, 2 ~p h d þ ~ þ 2 ~ ~ ~ ~dk ~ @h ~ @t @ jk dk dk @ jk 1
(3:4)
d1 " 1
d
2 X i¼1
@Eji @Hh h ~ ~ pþ~ p2 (1)i þ ¼ m ~ @t @~j3i d3i
! # @Hj3k Hj3k @Hh Hj3k 1 @Hh 2 þ ¼ (1)k sEjk , ~p h þ ~ þ ~p ~d3k @~j3k ~d2 ~d3k @~j3k @h ~ 3k
(3:5) k ¼ 1, 2
(3:6)
2 X ~ 3i @Hj3i 1 2 h ~ pþ~ p d (1) þ ¼ sEh ~ @~ji di i¼1
(3:7)
Solution of Equations 3.4 through 3.7 using perturbation theory requires switching to nondimensional electric and magnetic fields. We denote the scale factors for the electric and magnetic fields as E and H, respectively, ~ by the nondimensional quantities ~ ~ and and replace the dimensional ~ E and H E ~ ~ multiplied by the appropriate scale factors: H
81
SIBCs in Terms of Various Formalisms
~ ~ E¼~ EE;
~ ~ ¼ HH ~ H
(3:8)
The relation between E and H follows directly from Equation 3.1: mD H t
E¼
(3:9)
The relation in Equation 3.9 can be represented in another form: E¼
ms D D D H ¼ d2 H ¼ d2 H ¼ ~p1 (sd)1 H t s s s
(3:10)
In many practical problems, the total current is used as input data. In such cases it is only natural to use the characteristic current I as one of the scale factors and express E and H in terms of I. The relation between I and H can be obtained using the Biot–Savart law: ð ~ ¼ (4p)1 ~ I ð~ R=R3 Þdl H
(3:11)
L
where L is any closed contour surrounding the conductor carrying the current ~ I. Transfer to nondimensional variables in Equation 3.11 yields ð ~ ~I ~ ~ R ~ 3 d~l ~ ¼ I(4pD)1 ~ R= HH
(3:12)
L
The latter equation can be rewritten in the form: ð ~ ~I ~ ~ R ~ 3 d~l ~ ¼ (4p)1 ~ R= H
(3:13)
L
if H is chosen as H ¼ ID1
(3:14)
Substitution of Equation 3.14 in Equation 3.9 yields the new definition of E: E¼
m I t
(3:15)
Scale factors for all quantities are collected in Table 3.1. Note that there are three ‘‘basic’’ scale factors: I, t, and D. The other scale factors are combinations of the basic scale factors.
82
Surface Impedance Boundary Conditions: A Comprehensive Approach
TABLE 3.1 Quantities and Appropriate Scale Factors Quantity
Scale Factor
Unit
j 2 , j2
D ¼ min (d1, d2)
m
@=@j1, @=@j2 h
D1 ~D d¼p
m
@=@h
d1 ¼ (~pD)1
m1
T
t
s
I
I
A
E H
I t1m I D1
V m1 A m1
A
I
A
‘‘*’’ (in time convolution products)
t
s
m1
Substituting Equation 3.8 into Equations 3.4 through 3.7 and using Equations 3.9 and 3.10, we obtain ! ! ~j ~j @E ~j ~j ~h ~h @H E E @E 1 @E 2 3k k k k ~ p , k ¼ 1, 2 (3:16) þ ~ þ ~ ph ¼ (1)k ~p 2 ~ ~ ~ ~ ~ @h ~ @~t dk @ j k dk dk @ jk 2 X i¼1
~j ~h @E @H ~ 2 h i ~ ~ ¼ ~p pþp (1) ~ @~t @~j3i d3i
~j ~j ~h H @H @H 3k 3k ~ p þ ~ ~ @h ~ d3k @ j3k
i
!
~j ~h H 1 @H 3k ~h ~ þ p 2 ~ ~ ~ d3k d3k @ j3k 2
(3:17)
!
~ j , k ¼ 1, 2 ¼ (1)k ~p1 E k (3:18)
2 X i¼1
(1)3i
~j @H h ~ 3i ~h ~þ~ ¼ ~p1 E p p2 ~ @~ji di
(3:19)
Equations 3.16 through 3.19 do not contain the scale factors for the electric and magnetic fields any longer. We now apply the Laplace transform following the rule: ~ ~ ~s) ¼ E(
1 ð
~ ~ ~t) exp (~s~t)d~t, E(
0
~ ~ ~s) ¼ H(
1 ð
~ ~ ~t) exp (~s~t)d~t H(
(3:20)
0
In the Laplace domain, Equations 3.16 through 3.19 are written in the form: ! ! ~j @E ~j ~j ~h ~h E E @E 1 @E ~ , k ¼ 1, 2 (3:21) 2 k k k ~ p þ ~ þ ~ ph ¼ (1)k ~p~sH j3k 2 ~ ~ ~ ~ ~ @h ~ @ j @ j dk k dk d k k
83
SIBCs in Terms of Various Formalisms
2 X
(1)i
i¼1
~j @H 3k ~ p @h ~
~j ~ H @H h 3k þ ~ ~ d3k @ j3k
!
~j @E h ~ ~ i ~þ~ p p2 ¼ ~p~sH h ~ @~j3i d3i
~ ~j H 1 @H h 3k ~ þ ~ p2 h 2 ~ ~ ~ d3k d3k @ j3k
!
(3:22)
~ , k ¼ 1, 2 ¼ (1)k ~p1 E jk (3:23)
2 X ~ ~ ~ 3i @ H j3i 2 h ~ ¼ ~p1 E pþ~ p (1) h ~ji ~ @ d i i¼1
(3:24)
~ is small, we represent the functions for which Since the parameter p the solutions are sought in the form of asymptotic expansions in the parameter ~ p: X ~ ~ ~m ~¼ ~ pm E E 1
(3:25)
m¼0
X ~ ~ ~m ~ ¼ ~ pm H H 1
(3:26)
m¼0
Substituting the expansions (Equations 3.25 and 3.26) into Equations 3.21 through 3.24 and equating coefficients of equal powers of ~p, we obtain the following equations for the expansion coefficients: m ¼ 0: ~ ~ ~ ~ ~ ~ ~ ~ ¼0 E0 ¼ E0 ¼ E0 ¼ H 0 j1
j2
h
(3:27)
h
m ¼ 1: ~ ~1 @ E @h ~
jk
~ ~0 ¼ (1)k~s H ~ ~1 @ E
2 X
(1)i
i¼1
~ ~0 @ H @h ~
j3k
@~j3i
ji
,
k ¼ 1, 2
(3:28)
j3k
~ ~ ¼ ~s H 1
(3:29)
h
~ ~1 , ¼ (1) E k
jk
k ¼ 1, 2
(3:30)
84
Surface Impedance Boundary Conditions: A Comprehensive Approach ~ ~1 ¼ 0 E
(3:31)
h
m ¼ 2: ~ ~2 @ E
~ ~1 E
jk
~ ~1 ¼ (1)k~s H
jk
, k ¼ 1, 2 ~ dk j3k 1 0 ~ ~ ~2 ~1 @ E @ E 2 B C X h ~ ~ ji ji C ~ iB (1) B þ ¼ ~s H C 2 ~ ~ ~ @ A @ j @ j d h 3i 3i 3i i¼1 @h ~
~ ~1 @ H
j3k
@h ~
~ ~0 H
2 X
~ d3k
~ ~ ¼ (1)k E , 2
3i
@~ji
i¼1
k ¼ 1, 2
(3:34)
jk
~ ~0 @ H
(1)
(3:33)
j3k
(3:32)
j3i
~ ~ ¼ E 2
(3:35)
h
m ¼ 3: ~ ~ @ E 3 @h ~
jk
~ ~ E 2
~dk
~ ~ E 1
~ ~ @ E 2 jk
h
@~jk
h ~
~d2 k
jk
~ ~2 ¼ (1)k~s H
,
k ¼ 1, 2
3 2 ~ ~ ~ ~ ~ ~ 1 @ E2 @ E 2 6@ E3 7 2 X h ~ h ~ ~ j j ji 7 6 ~ i i i (1) 6 þ 7 ¼ ~s H 3 2 ~ ~ ~ ~ ~ 4 5 @ j @ j @ j d 2 d h 3i 3i 3i 3i i¼1 3i ~ ~ @ H 2 @h ~
j3k
~ ~1 H
~ ~0 H
~ ~1 @ H j3k
~ d3k
@~j3k
h
h ~
(3:36)
j3k
j3k
~ d23k
(3:37)
~ ~ ¼ (1) E , k ¼ 1, 2 3 k
3 2 ~ ~ ~1 ~0 @ H @ H 2 6 7 X h ~ j j3i 7 6 ~ 3i ~3 (1)3i 6 þ 7¼ E ~ ~ji ~ji 4 5 h @ @ d i i¼1
jk
(3:38)
(3:39)
85
SIBCs in Terms of Various Formalisms
From Equation 3.27 it follows that only the magnetic field at the interface does not vanish at d ! 0. The tangential electric field and the normal magnetic field tend to zero as ~ p whereas the normal electric field tends to zero as ~ p 2. The representation in Equations 3.25 and 3.26 has a clear physical meaning, namely 1. Zero-order terms of the Equations 3.25 and 3.26 give the solution of the problem in the perfect electrical conductor (PEC) limit, in which the electromagnetic field diffusion into the body is neglected. 2. First-order terms describe the diffusion in the well-known Leontovich approximation, in which the surface of the body is considered as a plane and the field is assumed to be penetrating into the body only in the direction normal to the surface of the body. 3. Second-order terms yield a correction by taking into account the curvature of the body’s surface, but the diffusion is assumed to be only in the direction normal to the surface as in the Leontovich approximation. This is Mitzner’s approximation. 4. Third-order terms and higher allow for electromagnetic field diffusion in directions tangential to the surface of the body. This approximation is referred to as Rytov’s approximation. Equations 3.28 through 3.39 can be solved sequentially to derive the first-, second-, and third-order terms of expansion of the tangential electric and magnetic fields in the conductor’s skin layer. However, the shortest way to obtain the mentioned quantities is application of the surface impedance ‘‘toolbox’’ since the magnetic field obeys the diffusion equation (Equation 2.2). As a first step, we represent Equation 3.2 with nondimensional variables in the following form: b ~ b ~ ~ ~ p rH E jk ¼ ~
(3:40)
jk
Equation 3.40 is similar to Equation 2.136. Therefore, treating the magnetic field as ~ f and applying Equations 2.176, 2.181, and 2.186 to the interface (~ h ¼ 0), we obtain relations between the coefficients of expansions directly in the Laplace domain: First-order (Leontovich) approximation: b b ~ ~ ~ ~1 p~s1=2 H ¼ (1)k ~ E 0 j3k
jk
(3:41)
86
Surface Impedance Boundary Conditions: A Comprehensive Approach
Second-order (Mitzner) approximation: ! b b b ~ ~d3k d ~ ~ ~ k k 1=2 1=2 ~2 ~ 1 ~ ~0 p~s E H H ¼ (1) ~ p~s ~~ j3k jk jk 2dk d3k
(3:42)
Third-order (Rytov) approximation:
"
b b b ~ ~ dk d3k ~ ~ ~ ~ ~ k 1=2 1=2 ~ ~ ~ ~ ¼ (1) ~ p s s p E H H 3 2 1 ~~ j3k jk jk 2dk d3k b 13 b b 0 ~ ~ ~ 2 ~ 2 ~ 2 ~ @ @ @ H H H 0 0 0 B ~ b ~d2 þ 2~d ~d 3~d2 C7 1 1 j3k C7 j j k 3k B ~ k 3k k k ~p2~s1 B H þ C7, 0 2 2 @ 2 2 @~j3k @~jk @~jk @~j3k A7 8~d2 ~d2 5 jk
k ¼ 1,2
k 3k
(3:43)
Substituting Equations 3.41 through 3.43 into the expansions in Equations 3.25 and 3.26 and using Equations 2.176, 2.181, and 2.186, we obtain Zero-order approximation (PEC limit): ~b ¼ 0 E j3k
(3:44)
First-order (Leontovich) approximation:
~b E j3k
8 ~ b LD > p~s1=2 H (1)k ~ > jk > > < k ~_ b FD, k ¼ 1, 2 p(1 þ j)H ¼ (1) ~ jk > > > d > (1)k ~ ~b * T ~b : p H TD j 2 k d~t
(3:45)
Second-order (Mitzner) approximation:
~b E j3k
! 8 ~ ~ d d > k 3k b b k > ~ ~ ~ > p H (1) ~s1=2 ~ p~s1=2 H LD > jk jk > > 2~ dk ~ d3k > > > ! > < 1 j ~_ b ~ d3k dk ~ _ b k ~ ¼ (1) (1 þ j)~ p H jk ~ p H jk FD, k ¼ 1, 2 > 2 2~ dk ~ d3k > > > " # > > > ~ > d b ~ b d3k b ~ b dk ~ k > > p f * T2 ~ f T TD p : (1) ~ ~k d ~3k jk * 1 d~t jk 2d
(3:46)
87
SIBCs in Terms of Various Formalisms
Third-order (Rytov) approximation:
~b E j3k
8 " ~ ~ > > k 1=2 > ~b ~ ~ b dk d3k > (1) ~s ~p H p~s1=2 H > jk jk ~ ~ > > 2dk d3k > > !# > > 2 ~b ~ b 1 @2 H ~b 2 ~ ~2 ~ ~ > @2H þ 2 d d d > 1@ H j3k jk jk k 3k 3d3k b k 2 1 > ~ > H jk ~p ~s LD þ > > 2 @~j2k 2 @~j23k @~jk @~j3k > d23k 8~ d2k ~ > > " > > > 1 j ~_ b ~ d3k dk ~ > > ~_ b ~ (1)k (1 þ j)~ p H p H jk > j > k ~k d ~3k < 2 2d 0 13 ¼ _b _b _b 2~ 2~ 2~ 2 2 ~ ~ ~ ~ > H H H @ @ @ þ 2 d 3 d d d j 1 1 > j j j k 3k _ 3k 3k A5 k k > ~b k > FD, k ¼ 1, 2 þ þ~p2 @H > jk > 2 2 @~j2k 2 @~j23k @~jk @~j3k > d23k 8~ d2k ~ > > " > > > ~ > d ~ b ~b d3k ~ b ~ b dk ~ > k > (1) ~p Hjk * T1 p Hjk * T2 ~ > > ~ ~ ~ > d t 2 d d > k 3k > ! # > > 2 ~b 2 ~b ~b ~~ ~2 ~2 > @2H > 1@ H 1@ H j3k jk jk > 2 ~ b dk þ 2dk d3k 3d3k b ~ > H þ T TD ~ p > jk : 2 @~j2k 2 @~j23k @~jk @~j3k * 2 d23k 8~ d2k ~
(3:47)
In Equations 3.45 through 3.47, LD, FD, and TD stands for the Laplace domain, frequency domain, and time domain respectively. Note that the electric field also obeys Equation 2.2. Thus, the technique can be applied to Equation 3.1, which is written with nondimensional variables in the Laplace domain as ~ b ¼ ~s1 ~ p1 H jk
b ~ ~ ~ rE
(3:48)
jk
Therefore, we get Leontovich’s approximation: pffiffi b ~ b , k ¼ 1, 2 ~ b ¼ ~s1 ~ ~ ¼ (1)3k ~ p1 (1)k ~sE p1~s1=2 E H j3k jk jk
(3:49)
Mitzner’s approximation: ~b H
j3k
¼ (1)
3k 1
~ p
~b ~s1=2 E jk
~ b ~ p~s1 E jk
~dk ~d3k 2d~k d~3k
! (3:50)
Rytov’s approximation: " ~ ~ b 3k 1 ~b ~ ~ b dk d3k ~ ~ p ~s1=2 E p~s1 E H j3k ¼ (1) jk jk 2~ dk ~ d3k 2 3=2
~ p ~s
~ b !# ~ b ~ b 2 ~ ~2 þ 2~ ~ @2E @2E @2E d 3 d d d 1 1 j3k k 3k j j b k 3k k k ~ E þ jk 2 @~j2k 2 @~j23k @~jk @~j3k 8~ d2k ~ d23k (3:51)
88
Surface Impedance Boundary Conditions: A Comprehensive Approach
Both sets of equations (Equations 3.45 through 3.47 and Equations 3.49 through 3.51) give relationships between the tangential electric and magnetic fields at the interface. However, the presence of ~p1 in Equations 3.49 through 3.51 makes their use less practical as ~ p ! 0 compared with Equations 3.45 through 3.47. The return to dimensional variables in Equations 3.45 through 3.47 can be performed by representing the nondimensional variables in the following form using the scale factors given in Table 3.1: ~t ¼ t=t; ~jk ¼ jk =D;
h ~ ¼ h=d;
~ ¼ Et=(mI); H ~ ¼ HD=I E
(3:52)
Substituting Equation 3.52 into Equation 3.45, we obtain the Leontovich approximation: _ b DI 1 E_ bj3k 2(mvI)1 ¼ (1)k dD1 (1 þ j)H jk 1=2 1 t 1 k b 1 1 @ b Hjk * pffiffiffiffiffi t Ej3k t(mI) ¼ (1) dD DI t @t pt
(3:53) (3:54)
Taking into account that t1=2 pffiffiffiffiffiffiffi ¼ sm d we get Equations 3.53 and 3.54 in the following form: Leontovich’s approximation: rffiffiffiffiffiffiffiffi _Eb ¼ (1)k (1 þ j) mvd H _ b ¼ (1)k jvmH _b j3k jk 2 s jk rffiffiffiffi m@ Hjbk * T2b Ebj3k ¼ (1)k s @t
(3:55) (3:56)
Here T2b is given by Equation 2.113. The transfer to dimensional variables in Equations 3.50 and 3.51 is done in a similar way. We give below the final results: Mitzner’s approximation: 1 E_ bj3k ¼ (1)k s
pffiffiffiffiffiffiffiffiffiffiffi b dk d3k b _ _ H jvsmH jk 2dk d3k jk
(3:57)
89
SIBCs in Terms of Various Formalisms
Ebj3k ¼ (1)k
d d 1 pffiffiffiffiffiffiffi @ b k 3k b H jk sm Hjk * T2b 2dk d3k s @t
(3:58)
Rytov’s approximation: E_ bj3k
" pffiffiffiffiffiffiffiffiffiffiffi b dk d3k b 1 _ _ H ¼ (1)k jvsmH jk 2dk d3k jk s 2 _b 2 _b _b @2H d2 þ 2dk d3k 3d23k _ b 1 1@ H 1@ H j3k jk jk H pffiffiffiffiffiffiffiffiffiffiffi k þ j3k 2 @j2k 2 @j23k @jk @j3k 8d2k d23k jvsm
!#
(3:59) "
Ebj3k ¼ (1)k
1 pffiffiffiffiffiffiffi @ b b dk d3k b H sm H *T 2dk d3k jk s @t jk 2
! # 2 b 2 b @ 2 Hjb3k d2k þ 2dk d3k 3d23k b 1 1 @ Hjk 1 @ Hjk b pffiffiffiffiffiffiffi Hj3k þ * T2 sm 2 @j2k 2 @j23k @jk @j3k 8d2k d23k
(3:60)
It is a simple matter to show that by neglecting the terms containing d2, Equations 3.59 and 3.60 reduce to the Mitzner condition in Equations 1.63 and 1.64. Similarly, by setting d1 ! 1 and d2 ! 1, the Rytov condition at the interface in Equations 1.108 and 1.109 is obtained. Note that the third term in Equations 3.59 and 3.60 is absent in both Mitzner’s and Rytov’s conditions since Equations 3.59 and 3.60 cannot, naturally, be obtained by adding Equations 1.63 and 1.64 and Equations 1.108 and 1.109. Neglecting all terms except for the very first leads to the original Leontovich SIBC as given in Section I.2.
3.4 Magnetic Scalar Potential Formalism We first restrict ourselves to a source-free region where any closed path, c, does not enclose any current so that the following condition is satisfied: þ ~ d~l ¼ 0 H (3:61) c
Then the magnetic scalar potential, f, can be introduced as follows: ~ ¼ rf H
(3:62)
90
Surface Impedance Boundary Conditions: A Comprehensive Approach
Substituting Equation 3.62 into Equation 3.3, we obtain the Laplace equation governing the scalar potential distribution in free space: r2 f ¼ 0
(3:63)
The exact boundary conditions at the dielectric=conductor interface follow directly from Equation 3.62: ~ ~ n rfjdiel ¼ ~ nH
cond
@f ~ ¼~ nH cond @~ n diel
(3:64) (3:65)
The use of the conditions in Equations 3.64 and 3.65 requires that the problem inside the conductor in terms of the magnetic field be considered together with Equation 3.63. To eliminate the conducting region from the numerical procedure, another relation between the magnetic scalar potential and its normal derivative at the conductor surface, and taking into account material properties of the conductor, must be derived and used instead of Equations 3.64 and 3.65. Obviously, this is equivalent to derivation of an equation relating the tangential and normal components of the magnetic field at the interface. Since the magnetic field in the conductor obeys the diffusion equation, Equations 2.170, 2.173, 2.178, and 2.183 of the surface impedance ‘‘toolbox’’ developed in Chapter 2 can be applied and the results written in the form: PEC limit: Hh ¼ 0
(3:66)
Leontovich’s approximation: 8 2 @H b X > ji > 1=2 > ~ ~ p s LD > > ~ > @ j i > i¼1 > > > < 1j X 2 @H _b ji b Hh ¼ ~ p FD ~ > 2 @ j > i i¼1 > > > > X 2 @H b > > ji > ~ b TD > p *T :~ 2 ~ i¼1 @ ji
(3:67)
SIBCs in Terms of Various Formalisms
91
Mitzner’s approximation: ! 8 2 ~d3i ~ X @ b d > i 1=2 1=2 b > > ~ p~s H ji ~ H p~s LD > ji > > @~ji 2~di ~d3i > i¼1 > > ! > 2 < 1j X @ 1 j _ b ~di ~d3i b b _ p Hji ~ p H ji FD Hh ¼ ~ > 2 i¼1 @~ji 2 2~di ~d3i > > > ! > > 2 > ~ ~ X > @ d d i 3i > b b b b ~ ~ ~ > p Hji * T p H ji * T TD :~ 2 3 ~ 2~ di ~ d3i i¼1 @ ji
(3:68)
Rytov’s approximation: " 8 2 ~ ~ X > @ b > 1=2 b di d3i > ~ ~ Hji ~ p s p~s1=2 H > ji ~ ~ > ~ > 2di d3i > i¼1 @ ji > > !# > > b b 1 @2H b @2H 2 2 ~ ~ ~ ~ > @2H > j3i ji ji > ~p2~s1 H b di þ 2di d3i 3d3i 1 LD þ > ji > > 2 @~j2i 2 @~j23i @~ji @~j3i 8~ d2i ~ d23i > > > " > > 2 > @ _b 1j _ b ~ d3i di ~ > 1j X > > ~ Hji ~ p p Hj i > < 2 ~ ~ ~ 2 @ j 2 d d i 3i i i¼1 !# Hhb ¼ 2 _b 2 _b _b 2 2 > ~ ~ ~ ~ @2H > 1@ H 1@ H ji ji j3i 2j b di þ 2di d3i 3d3i > _ > þ H FD þ~ p > ji > 2 2 @~j2i 2 @~j23i @~ji @~j3i > 8~ d2i ~ d23i > > " > > 2 > ~ X > @ d3i b ~ b di ~ > ~b ~ > ~ p Hjbi * T p Hji * T3 > 2 > ~ ~ ~ > @ j 2 d d > i 3i i i¼1 > > ! # > > ~~ ~2 ~2 > @ 2 Hjbi 1 @ 2 Hjbi @ 2 Hjb3i > > ~p2 H b di þ 2di d3i 3d3i 1 ~b T þ TD > * ji 4 : 2 @~j2i 2 @~j23i @~ji @~j3i d23i 8~ d2i ~
(3:69)
The shortest way to write Equations 3.67 through 3.69 with dimensional variables is to use Equations 2.198, 2.203, and 2.208: Leontovich’s approximation: 8 2 @H _b X > ji 1=2 > > FD (jvsm) > < @ji i¼1 b Hh ¼ 2 @H b > X > ji > b > (sm)1=2 * T2 TD : @j i i¼1 Mitzner’s approximation: 8 2 X > @ 1=2 1=2 di d3i _ b b > _ > FD H H (jvsm) (jvsm) > ji < 2di d3i ji @ji i¼1 b Hh ¼ 2 > X > @ 1=2 1=2 di d3i b b b b > > T T TD (sm) H (sm) H * * : ji 2 2di d3i ji 3 @ji i¼1
(3:70)
(3:71)
92
Surface Impedance Boundary Conditions: A Comprehensive Approach
Rytov’s approximation: 8 " 2 X > @ _b di d3i _ b > 1=2 > Hji (jvsm)1=2 H (jvsm) > > > 2di d3i ji @ji > i¼1 > > !# > > 2 _b 2 _b _b 2 2 > @2H 1@ H 1@ H > ji ji j3i b di þ 2di d3i 3d3i > (jvsm)1 H _ > FD þ > ji < 2 @j2i 2 @j23i @ji @j3i 8d2i d23i " Hhb ¼ 2 > X > @ di d3i b b > > (sm)1=2 Hjbi * T2b (sm)1=2 H *T > > 2di d3i ji 3 @ji > > i¼1 > ! # > > 2 b 2 b 2 b > 2 2 > 1 @ Hji 1 @ Hji @ Hj3i > 1 b di þ 2di d3i 3d3i b > TD Hji þ > (sm) * T4 : 2 @j2i 2 @j23i @ji @j3i 8d2i d23i
(3:72)
From Equation 3.62, it directly follows that H ji ¼
@f ; @ji
Hh ¼
@f @h
(3:73)
So that @ 2 Hjbi @j23i
¼
@2 @f @3f ¼ @j23i @ji @j23i @ji
and
@ 2 Hjb3i
@ji @j3i
¼
@2 @f @3f ¼ 2 @ji @j3i @j3i @j3i @ji (3:74)
Substituting Equation 3.73 into Equations 3.70 through 3.72 and taking into account Equation 3.74 yields the boundary relations between the normal and tangential derivatives of the scalar potential in the following form: Leontovich’s approximation:
8 2_b 2 > > 1=2 P @ f > > 2 < (jvsm) i¼1 @j
FD @f b i ¼ 2 b > 2 @~ n > 1=2 P @ f > > (sm) T b TD : 2 * 2 i¼1 @ji
(3:75)
Mitzner’s approximation: 8 2 X > @2 _ b 1=2 1=2 di d3i _ b > > FD f (jvsm) f b > 2 < (jvsm) 2di d3i @f i¼1 @ji ¼ 2 > X @~ n > @2 di d3i b b 1=2 1=2 b b > > (sm) TD f * T2 (sm) f * T3 : 2 2di d3i i¼1 @ji (3:76)
SIBCs in Terms of Various Formalisms
93
Rytov’s approximation: 8 " 2 2 > di d3i _ b > 1=2 P @ > > f (jvsm) f_ b (jvsm)1=2 > 2 > 2di d3i > i¼1 @ji > > !# > > 2 2 > 1 @ 2 f_ b 1 @ 2 f_ b > 1 _ b di þ 2di d3i 3d3i > f (jvsm) FD > b > < 2 @j2i 2 @j23i 8d2i d23i @f " ¼ > @~ n 2 P > @2 b b di d3i b b 1=2 > > (sm) f * T2 (sm)1=2 f * T3 > 2 > 2di d3i > @j i¼1 > i > ! # > > > 2 2 2 b 2 b > d þ 2d d 3d 1 @ f 1 @ f i 3i > 1 b 3i > (sm) fb i > * T4 TD : 2 @j2i 2 @j23i 8d2i d23i (3:77) Since the tangential derivative of a function can be approximated using the function calculated at nodes located along the conductor’s surface, Equations 3.75 through 3.77 are SIBCs of various orders of approximation relating the scalar potential and its normal derivative at the surface of the conductor. Now let a source current ~ I(t) flow in the conductor so that Equation 3.61 takes the form: þ ~ d~l ¼ I H (3:78) c
To ensure uniqueness of the solution under the condition of Equation 3.78, the scalar magnetic potential in nonconducting regions can be introduced as follows ~¼H ~ s rf H
(3:79)
~ s is the ‘‘source’’ magnetic field satisfying only the following where H requirement: þ ~ s d~l ¼ I H (3:80) c
~ s will be discussed in Chapter 6. Applying Equation Details of selection of H 3.79 to the interface and using Equations 3.75 through 3.77, the following boundary conditions are obtained and written in the form: 2 2 X @f @ h si X @ ~ s H) ~ ¼ Hs Hh ¼ H s ¼~ n (H F H ji þ F½@f=@ji h h @h @j @j i i i¼1 i¼1 (3:81)
94
Surface Impedance Boundary Conditions: A Comprehensive Approach
where the operator-function F takes the following forms depending on the order of approximation and excitation applied: 8 < (jvsm)1=2 f_jb FD
k , k ¼ 1, 2 (3:82) FLeontovich fjk ¼ : (sm)1=2 f b T b TD * jk 2
FMitzner fjk ¼
8 dk d3k _ b 1=2 _ b > > fjk (jvsm)1 fjk FD < ( jvsm) 2d d k 3k
, k ¼ 1, 2 > > : (sm)1=2 f b * Tb (sm)1 dk d3k f b * T b TD jk 3 jk 2 2dk d3k
(3:83)
8 dk d3k _ b > > f ( jvsm)1=2 f_jbk (jvsm)1 > > > 2dk d3k jk > > > ! > 2 b 2 b > 2 2 > @ 2 f_jb3k 1 @ f_jk 1 @ f_jk > 3=2 _ b dk þ 2dk d3k 3d3k > f FD (jvsm) þ > jk > < 2 @j2k 2 @j23k @jk @j3k 8d2k d23k Rytov , k ¼ 1, 2 F [ fjk ] ¼ > > 1=2 b 1 dk d3k b b b > f (sm) f (sm) T T > * * jk 2 > > 2dk d3k jk 3 > > > ! > 2 b 2 b > 2 ~2 > @ 2 fjb3k 1 @ fjk 1 @ fjk > 3=2 b dk þ 2dk d3k 3d3k b > (sm) fjk þ > * T4 TD : 2 @j2k 2 @j23k @jk @j3k 8d2k d23k
(3:84)
The product F[f] has dimensions of fD where f is the scale factor of the function f. Let us also represent F in nondimensional form in the frequency, time, and Laplace domains as follows (we will need these representations later for implementation of SIBCs into numerical methods) ~ ~f ] F[~f ] ¼ DF[
(3:85)
~ ~f ] takes the following form depending where the nondimensional product F[ on the order of approximation: 8 > ~ p~s1=2~f bjk LD > > < ~Leontovich [ ~fj ] ¼ ~ ~_ b FD, k ¼ 1, 2 (3:86) F p 1j k 2 f jk > > > : ~s ~ b ~ pfjk * T2 TD 8 d3k dk ~ > ~ > > ~s1=2 ~ p~f bjk ~ p2~s1~f bjk LD > > ~ ~ > 2 d d k 3k > > > < j ~ d3k ~_ b dk ~ Mitzner ~ ~ ~f_ b þ ~ F [ fjk ] ¼ ~ p 1j f j FD, k ¼ 1, 2 p2 j 2 k > 2 2~ dk ~ d3k k > > > > > ~k ~ > d3k ~b ~ b > ~~b ~ b ~2 d > fj * T3 TD : pfjk * T2 p ~ ~ 2dk d3k k
(3:87)
95
SIBCs in Terms of Various Formalisms
8 > ~dk ~d3k > > ~p~s1=2~f bjk ~p2~s1~f bjk > > 2~dk ~d3k > > 0 1 > > > > ~d2 þ 2~dk ~d3k 3~d2 @ 2~f bjk 1 @ 2~f bjk @ 2~f bj3k > 1 b > k 3k 3 3=2 ~ @f A LD > ~p ~s þ > jk > 2 @~j2k 2 @~j23k @~jk @~j3k > 8~d2k ~d23k > > > > > > 1j _ b > d~ d~3k > > ~ ~ ~2 j ~_ b k > < p 2 f jk þ p 2 f jk 2~d ~d k 3k ~Rytov [~fj ] ¼ 0 1 F k _b _b _b 2~ 2~ 2~ > 2 ~ ~dk ~d3k 3~d2 f @ f f @ @ > d þ 2 1 1 j j j _ > 1þj b 3k k 3k k k A FD > ~p3 4 @~f jk þ > > > 2 @~j2k 2 @~j23k @~jk @~j3k 8d~2k d~23k > > > > > > ~ ~ > > ~ b ~p2 dk d3k ~f b * T ~b > ~p~fjbk *T > 2 > ~dk ~d3k jk 3 > 2 > > ! > > 2 b 2 b ~d2 þ 2~dk ~d3k 3~d2 > @ 2~fjb3k > 1 @ ~fjk 1 @ ~fjk b k 3k > 3 ~ ~ b TD > þ *T : ~p fjk 4 2 @~j2 2 @~j2 @~jk @~j3k 8~d2 ~d2 k
k 3k
3k
(3:88)
The physical meaning of the operator-function F will be explained in Section 3.6.
3.5 Magnetic Vector Potential Formalism ~ is defined as follows: The magnetic vector potential A ~ ~ ¼rA H
(3:89)
Substituting Equation 3.89 into Equations 3.1 and 3.2, the distribution of the vector potential in the conductor can be written in the form: r ~ E ¼ r
! ~ @A m @t
~ ¼ s~ rrA E
(3:90) (3:91)
From Equation 3.90 we obtain ~ @A ~ þ rV E ¼ m @t
(3:92)
96
Surface Impedance Boundary Conditions: A Comprehensive Approach
where V is the electric scalar potential. Substitution of Equation 3.92 into Equation 3.91 gives ~ ~ ¼ sm @ A þ srV rrA @t
(3:93)
Following the approach developed by Emson and Simkin [1], we split the vector potential into ‘‘source’’ and ‘‘eddy’’ components: ~e ~¼A ~s þ A A
(3:94)
~s can be represented in the form: where the source component A ~s ¼ 1 A m
ð rVdt
(3:95)
Substitution of Equation 3.95 into Equation 3.92 yields the relation between the electric field and the eddy component of the magnetic vector potential: ~e @A ~ E ¼ m @t
(3:96)
Substitution of Equation 3.94 into Equation 3.93 leads to elimination of the electric scalar potential for the equation of diffusion of the eddy component of the magnetic vector potential: ~e ¼ sm rrA
~e @A @t
(3:97)
Application of the Coulomb gauge, ~e ¼ 0 rA
(3:98)
to Equation 3.97 gives ~e ¼ sm r2 A
~e @A @t
(3:99)
The eddy component of the vector potential vanishes when the skin depth is zero. To illustrate this fact, we calculate the circulation of the electric field over any closed contour G using Equation 3.92 as follows: þ G
@ ~ E d~l ¼ m @t
þ G
þ ~ d~l þ rV d~l A G
(3:100)
97
SIBCs in Terms of Various Formalisms
Taking into account the fact that þ
rV d~l 0
(3:101)
ð ~ d~ ~ d~l ¼ r A s (Stokes0 s theorem) A
(3:102)
G
and þ G
S
Equation 3.100 reduces to the following form þ G
@ ~ E d~l ¼ m @t
ð S
~ d~ rA s ¼ m
@ @t
ð
~e d~ rA s ¼ m
S
@ @t
ð S
@F ~ d~ H s¼ @t (3:103)
where F is the magnetic flux crossing the surface S. Assuming the conductor to be stationary, the only physical process causing variation of the magnetic flux through the conductor is the diffusion of the electromagnetic field causing expansion of the penetration depth and, consequently, the surface S, which is a function of the skin layer thickness d. Thus, it is natural to suppose that in the case of skin effect (when the skin layer remains shallow), the eddy component of the vector potential is proportional to the skin depth and, consequently, much smaller than the source component. In the PEClimit (~ p ¼ 0), the field does not penetrate inside the conductor, the magnetic flux crossing the conductor does not vary in time, and the vector potential has only a source component. For further derivations in terms of nondimensional variables, we introduce the scale factors A and V for the magnetic vector potential and electric scalar potential, respectively. The first of these can be obtained directly from Equation 3.89 using Equation 3.14 as follows: ~ ~ ~ ¼ D1 r ~ AA HH
(3:104)
So that A ¼ H;
D¼I
(3:105)
and ~ ~ ~ ~ ¼rA H
(3:106)
98
Surface Impedance Boundary Conditions: A Comprehensive Approach
Substitution of Equation 3.105 into Equation 3.92 yields the scale factor for the electric scalar potential mIt1 ¼ mIt1
~ ~ @A ~ þ VD1 rV @~t
(3:107)
so that V ¼ mIDt1
(3:108)
It is natural to consider how the tangential components of the magnetic vector potential at the interface can be expressed in terms of the tangential magnetic field. The shortest way is to use Equation 3.96 relating the vector potential and the electric field, and then substitute the E–H boundary condition obtained in Section 3.2. As a first step, we represent Equation 3.96 in terms of nondimensional variables: ~ ~e ~ ~ ¼ @A E @t
(3:109)
Rewriting Equation 3.109 in the Laplace domain and applying the interface, we obtain
~ ~e A
b
b ~ ~ ¼ ~s1 E
(3:110)
Substituting Equations 3.45 and 3.47 in Equation 3.110, the A–H relations are obtained in the various approximations: Leontovich’s approximation: 8 ~ b LD > p~s1=2 H (1)3k ~ > jk > < b 1 j ~ e _ 3k ~ ~ b FD, ¼ (1) ~ A p H jk > j3k 2 > > : 3k ~ b ~ b pHjk * T2 TD (1) ~
k ¼ 1, 2
Mitzner’s approximation: ! 8 ~d3k ~ > d k > 3k b b 1=2 1 ~ ~ > (1) ~ p ~s H jk ~ p~s H jk LD > > > 2~ dk ~ d3k > > > > ! > < b 1 j ~_ b j ~_ b ~ d3k dk ~ ~ 3k e ~ A p ¼ (1) ~ FD, k ¼ 1, 2 H jk þ ~ pH > j3k 2 2 jk 2 ~ dk ~ d3k > > > > ! > > ~3k ~k d > d > 3k b ~b b ~b > > p fjk * T2 ~ p fj * T 3 TD : (1) ~ 2~ dk ~ d3k k
(3:111)
(3:112)
99
SIBCs in Terms of Various Formalisms
Rytov’s approximation: " 8 > ~ b ~p~s1 H ~ b ~dk ~d3k > 3k > ~ ~s1=2 H (1) p > jk jk ~ ~ > > 2dk d3k > > > > ~ b ~ b ~ b !# > 2 ~ ~2 ~ ~ > @2H þ 2 d d d > 1 @2H 1 @2H j3k jk jk k 3k 3d3k b > k 2 3=2 ~ > LD þ H jk ~ p ~s > > ~d2 ~d2 2 @~j2k 2 @~j23k @~jk @~j3k > 8 > k 3k > 2 > > > > ~ ~ > _ b þ j ~pH > 3k 4 1 j ~ ~_ b dk d3k > H p > jk > (1) ~ 2 2 jk 2~dk ~d3k < e b ~ ~ ¼ A 0 13 > j3k ~_ b ~_ b 1 @ 2 H ~_ b > ~~ ~2 ~2 @2H > 1 @2H jk jk j3k A5 _ > b dk þ 2dk d3k 3d3k 2 1þj @ ~ > ~ p 4 H jk þ FD, k ¼ 1, 2 > > 2 @~j2k 2 @~j23k @~jk @~j3k > 8~d2k ~d23k > > > " > > > ~ ~ > > ~ b ~p dk d3k H ~b ~b *T ~ b *T > p H (1)3k ~ > jk 2 jk 3 > > 2~dk ~d3k > > > ! # > > 2 ~b 2 ~b ~b ~~ ~2 ~2 > @2H 1@ H 1@ H > j3k jk jk 2 ~ b dk þ 2dk d3k 3d3k > ~ b TD > ~ p T H þ * : jk 4 2 @~j2 2 @~j2 @~jk @~j3k 8~d2 ~d2 k 3k
k
3k
(3:113)
Switching in Equations 3.111 through 3.113 back to dimensional variables, we represent the A–H relations in the frequency- and time-domain as follows: Leontovich’s approximation: 8 _ b FD < (1)3k (jvsm)1=2 H b jk ~e , ¼ A : (1)3k (sm)1=2 H b * T b TD j3k jk 2
k ¼ 1, 2
(3:114)
Mitzner’s approximation: 8 > 3k _ b (jvsm)1 dk d3k H _ b FD > (jvsm)1=2 H > jk jk < (1) b 2dk d3k ~e ¼ A , k ¼ 1, 2 > j3k d > k d3k b > Hjk * T3b TD : (1)3k (sm)1=2 Hjbk * T2b (sm)1 2dk d3k
(3:115)
Rytov’s approximation: 8 " > > 3k > _ b (jvsm)1=2 dk d3k H _b (1) (jvsm)1=2 H > jk > > 2dk d3k jk > > > > !# > 2 b 2 _b > 2 2 > @ 2 H_ jb3k 1 @ H_ jk 1 @ H > jk 3=2 b dk þ 2dk d3k 3d3k > _ (jvsm) þ FD Hjk > > < b 2 @j2k 2 @j23k @jk @j3k 8d2k d23k ~e , A ¼ " > j3k > > > (1)3k (sm)1=2 H b * T b (sm)1 dk d3k H b * T b > jk 2 > > 2dk d3k jk 3 > > > ! # > > 2 b 2 b 2 b > 2 ~2 > 1 @ Hjk 1 @ Hjk @ Hj3k > 3=2 b dk þ 2dk d3k 3d3k b > (sm) H þ T TD > * j 4 : k 2 @j2 2 @j2 @j @j 8d2 d2 k 3k
k
3k
k
k ¼ 1, 2
3k
(3:116)
100
Surface Impedance Boundary Conditions: A Comprehensive Approach
The SIBCs in Equations 3.111 through 3.113 and Equations 3.114 through 3.116 can be represented in much shorter form using the surface impedance defined in Equations 3.86 through 3.88 and Equations 3.82 through 3.84 as follows: Nondimensional form: b h i ~ ~e ~ H ~b ¼ (1)3k F A jk
(3:117)
b h i ~e ¼ (1)3k F Hjbk A
(3:118)
j3k
Dimensional form:
j3k
In some formulations [2–5], employing the vector potential formalism relationships of the following kind are required at the interface to eliminate the conducting domain from the numerical procedure: b ~ @ ~ nA ~b ¼ fb ~ n A @~ n
(3:119)
b ~ n~ Ab n r~ A ¼ fa ~
(3:120)
where fa and fb are functions to be derived. Since the distribution of the magnetic vector potential in the conductor is described by the diffusion equation (Equation 2.2), the surface impedance ‘‘toolbox’’ can again be used and application of Equations 2.199, 2.204, 2.209, 2.201, 2.206, and 2.211 directly, yields the appropriate approximations: Leontovich’s approximation: b b 8 1=2 ~_ e ~e > > FD (jvsm) A @ A < jk k j ¼ b @h > d e b ~ > : (sm)1=2 dt A * T2 jk
,
(3:121)
TD
8 k 1=2 ~_ e b > )j FD < (1) (jvsm) (A b e k ~ rA ¼ b k 1=2 d > e b j3k ~ : (1) (sm) dt A * T2 jk
k ¼ 1, 2
, TD
k ¼ 1, 2
(3:122)
101
SIBCs in Terms of Various Formalisms
Mitzner’s approximation: b d þ d b b 8 k 3k ~_ e 1=2 ~_ e ~e > > þ FD A A @ A < (jvsm) jk jk 2dk d3k jk ¼ b b d @h > ~e ~e * Tb þ dk þ d3k A > : (sm)1=2 A 2 jk jk 2dk d3k dt
, k ¼ 1, 2 (3:123) TD
8 b d d b k 3k ~_ e k 1=2 ~_ e > > A A FD < (1) (jvsm) b j jk 2dk d3k ~e k rA ¼ , k ¼ 1, 2 (3:124) b b > d j3k e > ~e * Tb dk d3k ~ : (1)k (sm)1=2 TD A A 2 jk jk 2dk d3k dt
Rytov’s approximation: 8 b d þ d b 3k ~_ e > ~_ e þ k > A (jvsm)1=2 A > > jk jk 2dk d3k > > 0 > b 1 > > 2 ~_ e > > A @ 2 2 2 > B _ e b dk þ 2dk d3k 3d3k 1 X > jk C > ~ C FD A þ(jvsm)1=2 B þ b > > A @ 2 2 2 > jk 2 8d d > @j e k 3k ~ i i¼1 @ A > < jk , ¼ b b > @h > 1=2 d ~e * Tb þ dk þ d3k A ~e > (sm) A > 2 dt > j jk 2dk d3k > > 0 k > b 1 > > > 2 ~e > A 2 @ b d2 þ 2d d 3d2 > X B > 1 k 3k jk C 1=2 B ~e > k 3k C * Tb TD > A þ(sm) þ > A 2 @ 2 2 2 > j 2 8d d > @j k : k 3k i i¼1
~e rA
b j3k
¼
8 " > > b d d b > > 3k ~_ e > ~_ e k > (1)k (jvsm)1=2 A A > > > j jk 2d d k k 3k > > > > > 0 > b > > > 2 ~_ e > @ A > 2 2 b B _ e dk þ 2dk d3k 3d3k 1 > jk > ~ > (jvsm)1=2 B A > @ > 2 2 2 > j 2 8d d @j k > k 3k k > > > > 1 # > > b b > > ~_ e ~_ e > @2 A @2 A > > 1 j3k C jk > C FD > þ > A 2 > @j @j 2 > @j k 3k > 3k
dk ~ d3k > k > ~ ~s1=2fjbk ~ (1) s p~s1fjbk > > > d3k 2~ dk ~ > > > !# > > 2 b 2 b ~2 ~~ ~2 > @ 2fjb3k > 1 @ fjk 1 @ fjk > 2~3=2 b dk þ 2dk d3k 3d3k > fjk LD þ ~p s > > 2 @~j2k 2 @~j23k @~jk @~j3k > d23k 8~ d2k ~ > > > " > > > dk ~ 1 j _b j ~ d3k > > > fjk þ ~ p f_jbk (1)k 2j > < b 2 2 2~ d3k dk ~ ¼ r~ ~ f !# 2 b 2 b > j3k ~2 ~~ ~2 @ 2 f_jb3k > 1 @ f_jk 1 @ f_jk > 2 1 þ j _ b dk þ 2dk d3k 3d3k > f FD, k ¼ 1, 2 ~ p þ > jk > > 4 2 @~j2k 2 @~j23k @~jk @~j3k d23k 8~ d2k ~ > > > " > > ~ > > dk ~ d b ~b d3k b ~ b > > (1)k f *T ~ p fjk * T3 > > ~ ~t jk 2 ~ d > d 2 d k 3k > > > ! # > 2 b 2 b > ~2 ~~ ~2 > @ 2 fjb3k 1 @ fjk 1 @ fjk > b dk þ 2dk d3k 3d3k b 2 > ~ T f þ ~ p > * 4 TD jk : ~2 ~2 d 2 @~j2 2 @~j2 @~jk @~j3k 8d k 3k
k
3k
(3:128)
103
SIBCs in Terms of Various Formalisms
" 8 > dk d3k _ b k > > (1) jvsm (jvsm)1=2 f_jbk (jvsm)1 f > > 2dk d3k jk > > > > !# > > 2 _b 2 _b 2 _b 2 2 > > > (jvsm)3=2 f_ b dk þ 2dk d3k 3d3k 1 @ fjk þ 1 @ fjk @ fj3k > FD > jk < b 2 @j2k 2 @j23k @jk @j3k 8d2k d23k , k ¼ 1, 2 r ~ f ¼ " > j3k > > (1)k sm d (sm)1=2 f b Tb (sm)1 dk d3k f b Tb > > * * j 2 j 3 > k > 2dk d3k k dt > > > ! # > > 2 b 2 b 2 2 ~ > @ 2 fjb3k 1 @ fjk 1 @ fjk > 3=2 b dk þ 2dk d3k 3d3k b > > T (sm) f þ TD * : jk 4 2 @j2k 2 @j23k @jk @j3k 8d2k d23k
(3:129)
Taking into account Equations 3.84 and 3.88, we obtain h i 8 3k 1 ~ Rytov b > ~ ~sF (1) p LD fjk > > > < h i b ~Rytov f_ b k ¼ 1, 2 p1 2jF ¼ (1)3k ~ FD, r~ ~ f jk > j3k > h i > > : (1)3k ~ ~Rytov f b p1 @@~t F TD jk 8 h i > < (1)3k jvsmFRytov f_jbk b FD h i r ~ f , k ¼ 1, 2 ¼ > j3k @ Rytov : (1)3k sm @t F fjbk TD
(3:130)
(3:131)
Clearly, Equations 3.130 and 3.131 hold for the lower-order Leontovich and Mitzner approximations. With the use of Equations 3.130 and 3.131, the SIBCs in terms of various formalism can be represented as follows: E–H formalism: Ebj3k ¼ (1)3k m
h i @ h bi ^ Hb , F Hjk ¼ (1)3k mF jk @t
k ¼ 1, 2
(3:132)
H–f formalism: 2 2 X @w @ h si X @ ¼ Hhs F Hji þ F½@w=@ji (this is Equation 3:81) @h @j @j i i i¼1 i¼1 Hhb ¼
2 X @ h bi F Hji @ji i¼1
(these are Equations 3:70 through 3:72)
A–V formalism:
~e A
b
h i ¼ (1)3k F Hjbk , k ¼ 1, 2 (this is Equation 3:118) j3k b b 3k @ e ~e ~ , k ¼ 1, 2 F A ¼ (1) rA j3k jk @t
(3:133)
104
Surface Impedance Boundary Conditions: A Comprehensive Approach
^ of the surface impedance function can be calculated as The time-derivative F follows: ( (jvsm)1=2 f_jbk FD @ Leontovich ^Leontovich , k ¼ 1, 2 (3:134) f jk ¼ F f jk ¼ F ^ b TD @t (sm)1=2 fjbk * T 2 8 dk d3k _ b > > f FD > (jvsm)1=2 f_jbk 2dk d3k jk @ Mitzner ^Mitzner < , F fjk ¼ F fjk ¼ > @t dk d3k b ^ b > 1=2 b ^ b > f * T TD : (sm) fjk * T2 2dk d3k jk 3
k ¼ 1, 2
(3:135) @ Rytov ^Rytov F fjk ¼ F fjk @t 8 dk d3k _ b > 1=2 _ b > > > (jvsm) fjk 2dk d3k fjk > > > ! > > 2 b 2 b > 2 2 > @ 2 f_jb3k 1 @ f_jk 1 @ f_jk > 1=2 _ b dk þ 2dk d3k 3d3k > fjk (jvsm) þ FD > > < 2 @j2k 2 @j23k @jk @j3k 8d2k d23k , ¼ > > > (sm)1=2 f b * T ^ b dk d3k f b * T ^b > jk 2 > > 2dk d3k jk 3 > > > ! > > 2 ~d2 > @ 2 fjb3k @ 2 fjbk 1 @ 2 fjbk d þ 2d d 3 1 > k 3k 1=2 k 3k b > (sm) ^ b TD fjk þ > *T 4 : 2 @j2k 2 @j23k @jk @j3k 8d2k d23k
k ¼ 1,2
(3:136)
^ m are obtained using the Duhamel theorem and written in The functions T the form: d dTm (t) dTm (t) ðTm (t) * f (~ r, t)Þ ¼ r, tÞ ¼ r, tÞ þ Tm (0)f ð~ r, tÞ * f ð~ * f ð~ dt dt dt ^ m (t) * f ð~ ¼T r, tÞ, m ¼ 2, 3, 4
(3:137)
b
^ b ¼ dT2 ¼ (4pt3 )1=2 T 2 dt
(this is Equation 2:217)
b
^ b ¼ dT3 ¼ U 0 ¼ Tb T 3 1 dt
(3:138)
b
^ b ¼ dT4 ¼ (pt)1=2 ¼ Tb T 4 2 dt
(3:139)
The operator-function F may be called a surface impedance function since it describes the perturbation of the external field surrounding the body due to the field diffusion into the body and dissipation of its energy by the body. The first term on the right-hand side of Equation 3.84 gives the contribution from the field diffusion in the direction normal to the planar surface. The second
105
SIBCs in Terms of Various Formalisms
and third terms describe the field diffusion in the direction normal to the curvilinear surface. The fourth term takes into account the field diffusion in ^ m demonthe directions tangential to the planar surface. The functions Tm or T strate the evolution of these processes in time.
3.7 Surface Impedance near Corners and Edges As was mentioned in the Introduction chapter, the surface impedance concept has been originally developed under the assumption that the electromagnetic field distribution in the conductor may be described by a one-dimensional equation in the direction normal to the conductor surface. Clearly, this approximation is only valid for smooth bodies. To modify the SIBC near a 908-conducting wedge, infinitely long in the z-direction (two-dimensional problem), Deeley [6] postulated that the transverse magnetic field can be described in terms of the superposition of two modes: Hx ¼ Hxo exp (zy);
Hy ¼ Hyo exp (zx),
z2 ¼ jvsm
(3:140)
where Hxo and Hyo are assumed to be constant. Jingguo and Lavers [7] generalized Deeley’s approach for an arbitrary corner angle. However, rigorous consideration of this problem requires the field equations being solved in both the conducting and nonconducting regions near the corner. It is well known that in the classical limiting case of a perfect conductor, the distribution of the magnetic field along the surface of the edge is singular and can be described by the following function [8–11]: Hr ¼ Crg
(3:141)
where C is a known coefficient r and a are polar coordinates (see Figure 3.1) g ¼ 1=3 for a 908 edge y ρ
σ=0
Hx
α
δ
x
Hy σ>0
2D
E = Ezez
FIGURE 3.1 Derivation of the SIBCs near corners: geometry of the problem.
106
Surface Impedance Boundary Conditions: A Comprehensive Approach
It is natural to suppose that discontinuities also occur near imperfectly conducting wedges. Clearly then, the assumption in [6] is not correct because it ignores the singularity. In the present work, we apply the approach previously used to obtain SIBCs of high order of approximation, namely, the use of the perturbation technique in the small parameter equal to the ratio of the skin depth d and characteristic size D of the conducting domain. The solution of the problem for the perfect conductor is naturally considered as the zero-order approximation. In the first approximation, the distribution of the magnetic field along the perfectly conducting edge is used as a boundary condition for the field diffusion problem inside the conductor. Thus the solution of this problem gives the correction to the zero-order approximation by taking into account the finite conductivity of the body. In other words, we seek a generalization of the solution for the perfectly conducting edge. We consider an eddy current problem consisting of a rectangular conductor in the external electric field directed along the z-axis so that the magnetic field has only x- and y-components (Figure 3.1). We also assume that the dimension D of the conductor remains small compared to the wave length l. In this case, the distributions of the electric and magnetic fields in the conducting and nonconducting regions can be described by the Maxwell equations in the following form: Conductor: @Hy @Hx ¼ sE @x @y jvmHx ¼
(3:142)
@E @y
(3:143)
@E @x
(3:144)
@Hy @Hx ¼0 @x @y
(3:145)
jvmHy ¼ Dielectric:
jvmHx ¼
@E @y
jvmHy ¼
(3:146)
@E @x
(3:147)
Boundary conditions: ~ cond H
surf
~ diel ¼H
surf
; ~ Econd
surf
¼~ Ediel
surf
(3:148)
107
SIBCs in Terms of Various Formalisms
Assuming that the condition in Equation 2.1 is satisfied, we seek the surface impedance in the region near an edge, i.e., we need to calculate the following functions: Zx ¼
Ez Hx
;
Zy ¼
a¼2p
Ez Hy
for r D
(3:149)
a¼3p=2
where r is the distance from the edge. Following the perturbation theory, we now switch to dimensionless variables by choosing scale factors defined in Equations 3.13 and 3.14 for the magnetic and electric fields, respectively. The skin depth d is a natural choice as scale factor for the polar coordinate r (as well as both Cartesian coordinates x and y). With nondimensional variables, equations in Equations 3.142 through 3.144 take the form: ~ y @H ~x @H ~¼~ E p @~ x @~ y
!
~ ~ x ¼ @E ~~k2 H p @~ y ~ ~ y ¼ @E ~~k2 H p @~ x qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ~ p ¼ d=D ¼ 2=ðvsmD2 Þ 1
(3:150)
(3:151) (3:152) (3:153)
where ~k2 ¼ 2j. Substituting the scale factors into Equation 3.141 gives ~ r ID1 ¼ C~ pDÞg rg ð ~ H
,
~ r ¼ ~pg ~rg CI 1 Dgþ1 H
(3:154)
Introducing the following scale factor for C C ¼ IDg1
(3:155)
we represent Equation 3.141 in the form: ~ ~r ¼ ~ H p~ rg C
(3:156)
As a result of transfer to nondimensional variables, the small parameter ~p appears in the governing equations. We will seek the magnetic and electric fields in the conducting region in the form of power series in the small parameter ~ p: ~ ~ ~ ~0 þ ~ ~1 þ ; ~ ~ 0 þ ~pgþ1~ ~1 þ ~¼~ ~ ¼~ E E pgþ1 H E p g~ H pg H
(3:157)
108
Surface Impedance Boundary Conditions: A Comprehensive Approach
Since g þ 1 > 0 (recall that g > 0) the terms in Equation 3.157, beginning ~ 0 and E ~ 0 in from the second vanish when ~ p ¼ 0. Thus the first terms H Equation 3.157 describe the field distribution in the PEC limit whereas the others take into account the field diffusion into the conductor. Substituting Equation 3.157 into Equations 3.150 through 3.152, and 3.156 and equating the coefficients of equal powers of ~ p, we obtain the following equations for the coefficients in Equation 3.157: ~ pg: ~0 ¼ 0 E ~ ~ ~0 ¼~ rg C H r
(3:158) (3:159)
surface
~ pgþ1: ~1 ¼ E
~0 @ H y
~0 @ H x @~ y
(3:160)
@~ x ~ ~k H ~ 1 =@~ ~ 0 ¼ @E y
(3:161)
~ ~k H ~ 1 =@~x ~ 0 ¼ @ E
(3:162)
x
y
Equation 3.158 indicates that the electric field is zero everywhere inside the perfect conductor including on the surface. Therefore, the expansions in Equation 3.157 have the same physical meaning as in the classical surface impedance theory for smooth surfaces in spite of the fact that in the latter ~ 1 using Equation 3.160, we must case g ¼ 0 in the power series. To obtain E ~ ~ ~ 0 ) and (H ~ 0 ) along the know the normal derivatives of the functions (H x y conductor’s surface. For this purpose we consider the boundary value problem of distribution of the magnetic field inside the conductor. The governing equations are r2 Hx ¼ jvsmHx
(3:163)
r2 Hy ¼ jvsmHy
(3:164)
By switching in Equations 3.163 and 3.164 to nondimensional variables and substituting Equation 3.157, the following equations are obtained:
109
SIBCs in Terms of Various Formalisms ~ ~ ~0 ~ 0 ¼ ~k2 H D H x
x
y
y
~ ~ ~0 ~ 0 ¼ ~k2 H D H
(3:165) (3:166)
The problems in Equations 3.165 and 3.166 can be supplemented by the following boundary conditions based on Equation 3.159: 9 ~ ~ ~ a ¼ 3p=2: H ~0 ¼ 0> ~0 ¼ ~ = rg C; a ¼ 2p: H x x ~ ~ ~0 ! 0 > ~ 0 ! 1; r ; ~ ! 1: H r ¼ 0: H ~ x
(3:167)
x
9 ~ ~ ~> ~ 0 ¼ 0; a ¼ 3p=2: H ~ 0 ¼ ~rg C a ¼ 2p: H = y y ~ ~ ~0 ! 0 > ~ 0 ! 1; ~ ; r ! 1: H r ¼ 0: H ~ y
(3:168)
y
In Equations 3.167 and 3.168 we used the fact that the normal magnetic field vanishes on the surface of a perfect conductor. Note that the formulations in Equations 3.165 and 3.168 can provide only approximate solutions of the problem of the magnetic field diffusion inside a real conductor near the edge because the boundary conditions for the ideal conductor are used. ~ ~ 1 have been neglected, but it is a suitaThis means that terms containing H ~ 1 as follows from Equations 3.158 ble approximation for calculation of ~ E through 3.162. Equations 3.165 and=or 3.167 and Equations 3.166 and=or 3.168 are similar ~ ~ 0 )x . Let f(a) so we can restrict ourselves by consideration of the problem for (H be an integrable function. Then it can be proven that Equation 3.165 or Equation 3.167 is equivalent to the following boundary value problem in the domain ABcDE (Figure 3.2):
y ρ a
a
α B D
C
A
x
o
E
FIGURE 3.2 The boundary value problem under consideration.
110
Surface Impedance Boundary Conditions: A Comprehensive Approach ~ ~0 @2 H
~ 2 ~ ~0 ~0 H @ H @ 1 1 2 ~ x x x ~0 ¼ 0 þ H þ k x @~ r2 r r @~ ~ r2 @a2 ~ ~ ~ ~ rg ; ~ 0 ~ 0 H ¼ C~ H ¼0
x a¼2p
~ ~0 H
x r¼a
(3:169)
x a¼3=2p
¼ f (a);
~ ~ 0 H
x r¼1
¼0
provided that the radius a of the arc BcD tends to zero. The problem in Equation 3.169 has been solved using the Green function technique and the solution can be represented in the form:
~ ~ H 0
¼ F1 þ F2
(3:170)
1 X Kln (k~ x) Fn sin (2an=3) (k) K ln n¼1
(3:171)
x
where F1 ¼ k
ð 1 X Kln ðk~ xÞ ~0 ÞKln (k) Iln (k)Kln ðk~ ln sin (2an=3) x0 gðIln ðk x0 ÞÞdx0 (k) K l n n¼1 x
~ ip=2 F2 ¼ Cke
1
1 ð Iln (k) ~ þ Cki ln sin (2an=3) Iln ðk~ xÞ x ~0 gKln ðk~ x0 Þd~ x0 Kln (kx) Kln (k) n¼1 1 X
(3:172)
x
ln ¼ 2n=3;
~ ¼ r=a; k ¼ ka x
(3:173)
Fn is a Fourier transform of the function f(a) Im and Km are the m-order modified Bessel and Hankel functions, respectively It can be shown that G1 vanishes when a ! 0 and Equation 3.170 takes the form:
~ ~ 0 H
x a!0
¼
8 9 1 ðr ð < = ~ Kl (k~ eip=2 sin (2na=3)C r) ~ r0 gIln (k~ r0 )d~ r0 þ Iln (k~ r) ~ r0 gKln (k~ r0 )d~ r0 n : ; a
1 X pn n¼1
0
r
(3:174)
The problem described in Equation 3.166=3.168 has been solved in the same way and the result can be written in the form:
111
SIBCs in Terms of Various Formalisms
~ ~0 H
y
¼ a!0
1 X 2n n¼1
3
e
ip=2
9 8 1 ðr ð = < 0 0 0 0 0 0 ~ (1) sin (2na=3)C Kln (k~ r) ~ r gIln ðk~ r Þd~ r þ Iln (k~ r) ~ r gKln (k~ r )d~ r ; : n
0
r
(3:175)
Asymptotic analysis of Equations 3.174 and 3.175 yields distributions of the ~ ~ ~ 0 ) and (H ~ 0 ) in the conductor, near the corner: functions (H x
y
1=3 2 cos (a=3) ~ ¼C x k~ r cos (2p=3) 1=3 sin ((2p a)=3) ~ ~ 2 ~0 ¼ C H y k~ r cos (2p=3)
~ ~0 H
(3:176) (3:177)
Substituting Equations 3.176 and 3.177 into Equation 3.160, the electric field is obtained as 2 ~ 4=3 2 a 2 2p a 1 4a 1 2p 4a ~ sin a cos þ sin cos a þ sin þ sin E1 ¼ C~ r 3 3 3 3 3 3 3 3 3 (3:178) To illustrate the preceding theory, consider the problem of diffraction of plane waves by an imperfectly conducting edge as shown in Figure 3.3. Solution of the problem for a perfect conductor is known [8] and can be represented in the form: pffiffiffiffiffiffi 3pk inc ip=3 2 1=3 2cinc a cos H e sin 2 G(2=3) kr 3 pffiffiffiffiffiffi 3pk inc ip=3 2 1=3 2cinc a sin H e sin 2 G(2=3) kr 3
Hx jr1
(3:179)
Hy jr1
(3:180)
H inc
ψ inc
FIGURE 3.3 Diffraction of a plane wave by a conducting edge.
112
Surface Impedance Boundary Conditions: A Comprehensive Approach
Thus, in this case the coefficient C takes the form: C¼H e
inc ip=3
pffiffiffiffiffiffi 3pk sin (2cinc =3) G(2=3)
(3:181)
where G denotes the gamma-function. The distribution of the magnetic field intensity in the conducting and dielectric regions near the edge was calculated using Equations 3.176 and 3.177 and shown in Figures 3.4 and 3.5. To calculate the surface impedance, we switch to nondimensional variables and substitute the series (Equation 3.157), resulting in the following formula: 2 ~1 þ O ~ ~1 ~ ~E p p ~ pE E ~ ¼ þ O ~p2 , ¼ Zi ¼ ~ ~ ~ Hi ~ 0 þ~ ~0 ~ 1 þOð~ H H p H p2 Þ i
i
i ¼ x, y
(3:182)
i
Substitution of Equations 3.176 through 3.178 into Equation 3.182 leads to the following expressions: tan (p=6) ~ x ¼ ~ ; p Z r ~
tan (p=6) ~y ¼ ~ Z p ~r
(3:183)
6 5 4 3 2 1 0 0.5 0.5 y
0 −0.5 −0.5
FIGURE 3.4 The magnetic field intensity near the corner.
0 x
113
SIBCs in Terms of Various Formalisms
H 0.4 0.3 0.2 0.1 y
0 –0.1 –0.2 –0.3 –0.4 –0.5 –0.5
–0.4
–0.3
–0.2
–0.1
0
0.1
0.2
0.3
0.4
x FIGURE 3.5 The magnetic field intensity near the corner—contour plot.
Returning in Equation 3.183 to dimensional variables, we finally obtain Zx ¼
tan (p=6) ; sr
Zy ¼
tan (p=6) sr
(3:184)
References 1. C.R.I. Emson and J. Simkin, An optimal method for 3-D eddy currents, IEEE Transactions on Magnetics, MAG-21(6), November 1985, 2231–2234. 2. W.M. Rucker and K.R. Richter, A BEM code for 3-D eddy current calculations, IEEE Transactions on Magnetics, 26(2), March 1990, 462–465. 3. H. Igarashi, A. Kost, and T. Honma, A three dimensional analysis of magnetic field around thin magnetic conductive layer using vector potential, IEEE Transactions on Magnetics, 34(5), September 1998, 2539–2542. 4. M. Cao and P.P. Biringer, BIE formulations for skin and proximity effect problems of parallel conductors, IEEE Transactions on Magnetics, 26(5), September 1990, 2768–2770. 5. D. Rodger and H.C. Lai, A surface impedance method for 3D time transient problems, IEEE Transactions on Magnetics, 35(3), May 1999, 1369–1371.
114
Surface Impedance Boundary Conditions: A Comprehensive Approach
6. E.M. Deeley, Surface impedance near edges and corners in three dimensional media, IEEE Transactions on Magnetics, 26(2), March 1990, 712–714. 7. W. Jingguo and J.D. Lavers, Modified surface impedance boundary conditions for 3D eddy current problems, IEEE Transactions on Magnetics, 29(2), March 1993, 1826–1829. 8. G.A. Greenberg, On a method of solving the fundamental problem of electrostatics and allied problems, Russian Journal of Experimental and Theoretical Physics, Part I, 8(3), March 1938, 221–252; Part II, 9(6), June 1939, 725–728 (in Russian). 9. J. Meixner, The behavior of electromagnetic fields at edges, IEEE Transactions on Antennas and Propagation, 20(4), July 1972, 442–446. 10. J. Van Bladel, Electromagnetic Fields, Chapters 5 and 12, A SUMMA Book, Hemisphere, Washington, DC, 1985. 11. K. Nikoskinen and I. Lindell, Image solution for Poisson’s equation in wedge geometry, IEEE Transactions on Antennas and Propagation, 43(2), February 1995, 179–187.
4 Calculation of the Electromagnetic Field Characteristics in the Conductor’s Skin Layer
4.1 Introduction In Chapters 2 and 3, we derived the distributions of the magnetic field, electric field, and vector potential ‘‘inside’’ the conductor as a function of the tangential magnetic field at ‘‘the conductor’s surface.’’ This information is sufficient for calculation of such important practical quantities as current density, energy stored in the electromagnetic field, inductance, and power losses. In the case of constant material properties of the conductor, calculation of the quantities mentioned is possible without numerical consideration of the conducting domain. In other words, the surface impedance concept should provide not only local boundary relations between fields or potentials, but also analytical formulae for calculation of ‘‘integral’’ or ‘‘volume’’ quantities using only the distributions of the EM field at the surface of the conductor. Once the distribution of the electric and magnetic fields inside a domain V is obtained, energy-related quantities such as ohmic power loss (Joule loss) Pw and the magnetic field energy Wm associated with V can be calculated as follows: ð (4:1) PW ¼ s1 J 2 dv V
1 Wm ¼ 2
ð
~ ~ H B dv
(4:2)
V
where ~ J and ~ B are the current density and the magnetic flux density, respectively. In circuit theory, these quantities are defined as PW ¼ I 2 R
(4:3)
1 Wm ¼ LI 2 2
(4:4)
where R and L are the resistance and inductance of the conductors, respectively. 115
116
Surface Impedance Boundary Conditions: A Comprehensive Approach
From a computational point of view, it is very attractive to solve the problem of electromagnetic field distribution in the domain under consideration, use the solution to calculate ‘‘extract’’ lumped parameters and then apply them in circuit analysis. However, any combination of field and circuit representations must be done with great care because the procedure is applicable only in particular cases. For example, it is impossible to guarantee in the general case that resistance defined using Equations 4.3 and 4.1 can be then used to calculate voltage drop between two selected points. Here, we will restrict ourselves to the practically important case of long parallel conductors where description of the electromagnetic processes using lumped parameters is possible. In this chapter, we demonstrate how to calculate the following quantities based on known distributions of the tangential magnetic field along the conductor’s surface in the practically important case of time-harmonic fields: . .
.
Distribution of the electric and magnetic fields across the skin layer Power losses, energy in the magnetic field, and force acting on the conductor Lumped parameters such as resistance and internal inductance per unit length of the conductor
4.2 Distributions across the Skin Layer In the previous chapters, we focused on derivation of relations between electromagnetic quantities at the interface between the dielectric and the conductor. Since the goals of this chapter require performing integrations over the conductor’s volume, we start from the distributions of magnetic and electric fields, and the magnetic vector potential throughout the skin layer of a good conductor (where displacement current may be neglected). Consider a straight long current-carrying conductor of constant crosssection as shown in Figure 4.1. In this case, the electric current flowing in
I ξ1
FIGURE 4.1 The two-dimensional problem of electromagnetic field diffusion into a long straight cylindrical conductor of constant cross-section.
η
ξ2
117
Calculations in the Conductor’s Skin Layer
the conductor has only a longitudinal component and the problem of the electromagnetic field distribution inside the conductor can be considered as two dimensional in its cross section. Let coordinate j1 of our surface-related coordinate system (j1 , j2 , h) be directed along the conductor so that @ ¼ 0; d1 ! 1 @j1
(4:5)
Then, the electric and magnetic field vectors can be represented in the form: ~ E ¼~ ej1 E;
~ ¼~ H e j2 H j2 þ ~ eh H h
(4:6)
Substituting Equation 4.5 into the ‘‘toolbox’’ formulae (Equations 2.92 through 2.94, 2.158, 2.162, and 2.166), we reduce them to the following form: First-order approximation: f_jk ¼ f_jbk expð(1 þ j )~ hÞ,
k ¼ 1, 2
(4:7)
g_ j3k ¼ (1)k (1 þ j )f_jbk expð(1 þ j )~ hÞ, k ¼ 1, 2
(4:8)
Second-order approximation: 1 expð(1 þ j )~ hÞ, k ¼ 1, 2 f_jbk þ ~ ph ~ f_jbk 2~ d2 1 1 j 1 _b g_ j1 ¼ (1 þ j ) 1 þ ~ fj expð(1 þ j )~ ph ~ þ~ p hÞ 2 2~ 2~ d2 d2 2 1 1 j 1 _b g_ j2 ¼ (1 þ j ) 1 þ ~ ph ~ ~ p hÞ fj expð(1 þ j )~ 2 2~ 2~ d2 d2 1 f_jk ¼
(4:9) (4:10) (4:11)
Third-order approximation: " !# 2 b 1 @ f_j1 b 1 2 2 _b 3 21j b 1 _ _ _fj ¼ f_ b þ ~ ~ ~ expð(1 þ j )~ hÞ h ~ fj1 ph ~ fj1 þp h ~ fj1 þp þ j1 1 2 2~ d2 8~ d22 8~d22 2 @~j22
(4:12) "
2 b 1 3 1j 3 1 @ f_j2 ph ~ f_jb2 þ~ p2 h ~ 2 f_jb2 þ~ p2 þ f_j2 ¼ f_jb2 þ ~ h ~ f_jb2 2 2~ d2 8~ d22 8~d22 2 @ ~j22
!# expð(1 þ j )~ hÞ
(4:13)
118
Surface Impedance Boundary Conditions: A Comprehensive Approach "
! ! 2 b 1 1 j _b 1 1 @ f_j2 b 2 2 3 b 3 _ _ þ~ p ph ~ þ~ ph ~ þ~ ph ~ fj2 þ~ ph ~ g_ j1 ¼ (1 þ j ) fj2 1 þ ~ fj2 2 2 @~j22 2~ 8~ 2~ 2~ d2 d22 d2 d22 !# @ 2 f_jb j _b 3 ~p2 expð(1 þ j )~ hÞ þ 22 (4:14) fj2 2 @~j2 4~d22 " ! ! @ 2 f_jb1 1 3 1 j 1 5 1 b 2 2 b b g_ j2 ¼ (1 þ j ) f_j1 1 þ ~ ~ p f_j1 ph ~ þ~ p h ~ þ ~ph ~ f_j1 ~ph ~ 2 2 @~j22 2~ d2 8~ d22 2~d2 2~d22 !# @ 2 f_jb j _b 1 expð(1 þ j )~ hÞ fj1 þ 21 (4:15) þ~ p2 2 4~ d22 @~j2
_ j ¼ f_j in Equations 4.7, 4.9, 4.12, and 4.13 and Letting E_ ¼ f_j1 and H 2 2 switching to dimensional variables, we obtain distributions of the tangential electric and magnetic fields as a function of h (across the skin layer) in the various approximations: Leontovich’s approximation: E_ ¼ E_ b exp (bh) _j ¼ H
_b H j
exp (bh)
(4:16) (4:17)
Mitzner’s approximation: h E_ ¼ E_ b 1 þ exp (bh) 2d _ b 1 þ h exp (bh) _j ¼H H j 2d
(4:18) (4:19)
Rytov’s approximation: " !# 2 2 _b _E ¼ E_ b þ E_ b h þ E_ b 3h þ d 1 j h E_ b 1 þ 1 @ E exp (bh) (4:20) 8d2 2d 2 8d2 2 @j2 " !# _b 2 @2H h 3h 1 j 3 1 j b b b b _j ¼ H _ þH _ _ _ H þd þ þH h H exp (bh) j j j j 8d2 2d 2 8d2 2 @j2 (4:21) where b¼
pffiffiffiffiffiffiffiffiffiffiffi 1 þ j jvsm ¼ d
(4:22)
Calculations in the Conductor’s Skin Layer
119
and the following notations have been introduced: j ¼ j2 ;
d ¼ d2
(4:23)
_ j ¼ f_j in Equations 4.8, 4.10, 4.11, 4.14, and 4.15, By denoting E_ ¼ g_ j1 and H 2 2 we obtain the tangential electric field distribution in terms of the tangential magnetic field at the interface: Leontovich’s approximation: 1þj _b H exp (bh) E_ ¼ sd j
(4:24)
_E ¼ 1 þ j 1 þ h þ 1 j d H _ b exp (bh) sd 2d 4 d j
(4:25)
Mitzner’s approximation:
Rytov’s approximation: " ! _b 2 @2H j _E ¼ 1 þ j H _ b 1 þ h þ 3h þ d 1 j H _b1þH _ b 3h þ h j j j 2 sd 2d 8d2 4 d d @j2 !# _b @2H j _b 3 j Hj 2 þ exp (bh) d2 2 2 4d @j
(4:26)
Improvement in accuracy of results by taking into account high-order terms at large values of ~ p ¼ d=d can be demonstrated through comparison of Equations 4.16 through 4.21 and Equations 4.24 through 4.26 with the exact solution, which is not limited by the skin layer approximation. An exact solution is available for the single current-carrying round conductor, shown in Figure 4.2, where the fields vary only radially and there is no variation in
I d
FIGURE 4.2 A solid, round, current-carrying conductor.
120
Surface Impedance Boundary Conditions: A Comprehensive Approach
the tangential direction (second derivatives in Equations 4.16 through 4.21 and Equations 4.24 through 4.26 vanish). Due to symmetry, the magnetic field has only one component: ~ ¼~ ej H H ej H j ¼ ~
(4:27)
The diffusion equations describing the electric and magnetic fields in the common cylindrical coordinates are the following: @ 2 E_ 1 @ E_ þ ¼ b2 E @r2 r @r _ 1 @H _ @2H 1 _ 2 H þ þ ¼ b @r2 r @r r2
(4:28) (4:29)
The radial coordinate r is directed from the center of the cylinder to its surface. The coordinates r and h are related as r¼dh
(4:30)
where d is the radius of the cylinder. Taking into account the fact that the electromagnetic field at the center of the cylinder (r ¼ 0) has finite value, the solutions of these equations can be represented in the form: _ ¼ E_ b I0 (br) E(r) I0 (bd)
(4:31)
_ _ b I1 (br) H(r) ¼H I1 (bd)
(4:32)
Here In is the modified Bessel function of the first kind and order, n. The distributions of the functions In are shown in Figure 4.3. The function I1 can be calculated as the derivative of I0 with respect to br: I1 (br) ¼ I00 (br)
(4:33)
Note that the solution in Equation 4.32 is equivalent to Equation 2.223. Indeed, comparing Equations 2.222 and 4.22, we get b ¼ jq; I1 (br) ¼ I1 (jqr) ¼ jJ1 (qr)
(4:34)
121
Calculations in the Conductor’s Skin Layer
3.5 I0 (x) I1 (x) I2 (x) I3 (x)
3.0 2.5 2.0 1.5 1.0 0.5 0.0 0
1
2 X
3
4
FIGURE 4.3 Distributions of modified Bessel functions of the first kind In, n ¼ 0, 1, 2, and 3.
where J1 is the Bessel function of the first kind and first order. With the use of Equation 4.34, we can represent Equation 4.32 in the form: _ _ b J1 (qr) H(r) ¼H J1 (qd)
(4:35)
In the particular case of a single circular cylinder, Equation 3.1 takes the form _ ¼ jvmH
dE_ dr
(4:36)
Substituting Equation 4.31 into Equation 4.36, we obtain _b _ ¼ (jvm)1 E dI0 (br) ¼ b E_ b I1 (br) H I0 (bd) dr jvm I0 (bd)
(4:37)
Comparison of Equations 4.32 and 4.37 gives _ b ¼ b E_ b I1 (bd) H jvm I0 (bd)
or
jvm _ b I0 (bd) E_ b ¼ H b I1 (bd)
(4:38)
122
Surface Impedance Boundary Conditions: A Comprehensive Approach
Therefore _ b I0 (br) ¼ b H _ b I0 (br) ¼ 1 þ j H _ b I0 (br) _ ¼ jvm H E(r) b I1 (bd) s I1 (bd) sd I1 (bd)
(4:39)
Figures 4.4 through 4.9 show distributions parts, of the
real parts, imaginary
_ b, h ~ and E_ ¼ g H ~ for three and amplitudes of the functions E_ ¼ f E_ b , h different values of the small parameter ~ p ¼ d=d, namely 0.1, 0.3, and 0.5. Eb 1 _ b and d H are assumed to be equal to 1 and the nondimensional variable h ~ varies from 0 (surface) to 0.5 (at half the distance from the surface to the center of the cylinder). Each figure gives four curves: approximate solutions in the Leontovich, Mitzner, and Rytov approximation and the exact solution. The disagreement between the curves gives the approximation error in our example. From Figures 4.4 and 4.7, it can be seen that for ~p ¼ 0:1 (very thin skin layer) all curves are very close to each other. This means that the formula (Equation 4.16) of the order of the Leontovich approximation is sufficiently accurate and there is no need to take into account even the radius of curvature of the conductor’s surface. The situation naturally changes as the parameter ~ p increases. According to the results shown in Figures 4.5 and 4.8, application of Equation 4.16 in the case ~p ¼ 0:3 leads to significant errors whereas Equation 4.18 obtained in the Mitzner approximation works
1 Analytical Leontovich Mitzner Rytov
~ p = 0.1 0.8
|f |
0.6
0.4
0.2
0
0.1
0.2
0.3 η/d
FIGURE 4.4
~ for ~p ¼ 0:1 and E_ b ¼ 1. Amplitude of E_ ¼ f E_ b , h
0.4
0.5
123
Calculations in the Conductor’s Skin Layer
1
Analytical Leontovich Mitzner Rytov
~ p = 0.3 0.8
|f|
0.6
0.4
0.2 0
0.1
0.2
0.3
0.4
0.5
η/d FIGURE 4.5
~ for ~p ¼ 0:3 and E_ b ¼ 1. Amplitude of E_ ¼ f E_ b , h
1
Analytical Leontovich Mitzner Rytov
~ p = 0.5
|f|
0.8
0.6
0.4
0
0.1
0.2
η/d
FIGURE 4.6
~ for ~p ¼ 0:5 and E_ b ¼ 1. Amplitude of E_ ¼ f E_ b , h
0.3
0.4
0.5
124
Surface Impedance Boundary Conditions: A Comprehensive Approach
1500
Analytical Leontovich Mitzner Rytov
p~ = 0.1
|g|
1000
500
0
0
0.1
0.2
η/d
0.3
0.4
0.5
FIGURE 4.7 b
_ b d1 ¼ 1. _ ,h ~ for ~p ¼ 0:1 and H Amplitude of E_ ¼ g H
Analytical Leontovich Mitzner Rytov
p~ = 0.3
500
|g|
400
300
200
100 0
0.1
0.2
0.3 η/d
FIGURE 4.8 b
_ b d1 ¼ 1. _ ,h ~ for ~p ¼ 0:3 and H Amplitude of E_ ¼ g H
0.4
0.5
125
Calculations in the Conductor’s Skin Layer
350
Analytical Leontovich Mitzner Rytov
p~ = 0.5
300
|g|
250
200
150
100
0
0.1
0.2
η/d
0.3
0.4
0.5
FIGURE 4.9 b
_ b d1 ¼ 1. _ ,h ~ for ~p ¼ 0:3 and H Amplitude of E_ ¼ g H
satisfactorily, and Equation 4.20 (Rytov’s approximation) provides high accuracy. When ~ p ¼ 0:5 even the Rytov approximation cannot provide agreement with the exact solution (Figures 4.6 and 4.9). Note that in the latter case one can expect more significant errors because the skin depth is so large that the surface impedance concept should not be generally applied (see Chapter 9 for range of applicability of the surface impedance concept). However, the existence of symmetry in the
circular conductor b reduces the _ b, h _ ,h ~ and E_ ¼ g H ~ leads to the error. Analysis of the functions H ¼ f2 H same conclusions. It is readily demonstrated that under the conditions of skin effect ðjbrj 1Þ Equations 4.31, 4.32, and 4.39 lead to Equations 4.16, 4.17, and 4.24, respectively. Taking into account the asymptotic behavior of the Bessel functions at large arguments: exp (z) In (z) pffiffiffiffiffiffiffiffi , 2pz
when jzj ! 1
(4:40)
we represent Equation 4.31 in the form: sffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffi d d b _ ¼E expðb(r d)Þ ¼ E_ exp (bh) E_ b exp (bh) E(r) r dh _b
(4:41)
126
Surface Impedance Boundary Conditions: A Comprehensive Approach
Equations 4.32 and 4.39 are transformed in the same way. Note that application of Equation 4.39 for jbrj 1 yields the classical Leontovich SIBC: 1 þ j _ b I0 (br) 1 þ j _ b H H E_ b ¼ sd I1 (bd) sd
(4:42)
Since under the conditions of skin effect problems the current is concentrated near the surface, it is natural to introduce the ‘‘surface current’’ density ~ K as 1 1 ð ð ~ ~ 1 , j2 , h)dh ~ K(j1 , j2 ) ¼ J(j1 , j2 , h)dh ¼ r H(j 0
(4:43)
0
In the two-dimensional case under consideration here, the surface current density is scalar. Substituting Equations 4.16 through 4.21 into Equation 4.43, we obtain Leontovich’s approximation: 1þj _b _b H ¼H K_ ¼ j db j
(4:44)
_b ¼ 1þ1j d H _b _K ¼ b 1 þ 1 þ 1 j d H j j b b2 2d 4 bd 2 d
(4:45)
Mitzner’s approximation:
Rytov’s approximation: " ! !# _b _b @2H 1j _ b 1 1 @2H d2 j _ b 3 j j b 3 _K ¼ b H _b 1þ 1 þ 3 _ þd Hj þ Hj 2 2 þ 2 Hj 2 þ j b 2 b b2 2d 2b3 8d2 4 bd 4d b d b @j2 @j2 ! ! _b @2H 1j _ b1 _ b 3 1 @ 2 H_ jb j _ b 3 þ1 _ b 1þ 1 þ 3 þd jd2 H ¼H Hj þ Hj 2 þ j j 2 2 2 2 b2d 2b 8d 4 d bd b @j 8d 2 @j2 ! _ b 3 @2H _b H j j _ b þ 1 j dH _ b jd2 þ (4:46) H j d2 2 @j2 2 d j
In terms of the electric field at the interface: Leontovich’s approximation: K_ ¼ E_ b s
1 ð
0
sd sd ¼ E_ b (1 j ) exp (bh)dh ¼ E_ b sb1 ¼ E_ b 1þj 2
(4:47)
127
Calculations in the Conductor’s Skin Layer
Mitzner’s approximation: K_ ¼ E_ b s
1 ð
0
h sd 1 sd d 1þ ¼ E_ b 1j 1þ 1þ exp ( bh)dh ¼ E_ b 2d 1þj b2d 2 2d (4:48)
Rytov’s approximation: 1 ð"
h 3h2 1j h E_ b þ E_ b þ E_ b 2 þ d 8d 2d 2 0 " sd _ b _ b 1 3 1 þ E þE þ E_ b 2 ¼ 1þj b2d b 16d2 b2 ( sd _ b d d2 d2 j E 1j 1þ þ 2 4 2 2d 8d
K_ ¼ s
1 1 @ 2 E_ b E_ b 2 þ 8d 2 @j2
!#
1 1 @ 2 E_ b E þ 2 8d 2 @j2 ) @ 2 E_ b @j2
exp (bh) !#
_b
(4:49)
The surface current density is an important quantity in the surface impedance concept and is frequently used in practical calculations.
4.3 Resistance and Internal Inductance We first consider an idealized problem in which a current-carrying conductor fills the region x 0. The conductor is assumed to be of finite length ‘ in the j-direction as shown in Figure 4.10.
x η Jx 1
l
ξ
FIGURE 4.10 A current-carrying conducting slab with the current density distribution indicated.
128
Surface Impedance Boundary Conditions: A Comprehensive Approach
The ratio of the electric field E_ b at the conductor’s surface to the total current I_‘ flowing through the conductor gives the internal impedance. The latter can be represented in the form: E_ b ¼ R0‘ þ jvL0i‘ I_‘
(4:50)
In this case, the total current is ð‘ _ _I‘ ¼ Kdj ¼ ‘K_
(4:51)
o
In the Leontovich approximation, Equation 4.51 takes the form: sd I_‘ ¼ ‘E_ b 1þj
(4:52)
Substituting Equation 4.52 into Equation 4.50, we obtain the resistance and internal inductance per unit length and per unit width in the Leontovich approximation: R0‘ ¼
1 ‘sd
(4:53)
L0i‘ ¼
1 ‘vsd
(4:54)
In the case of a single circular cylindrical conductor (see Figure 4.2) ‘ ¼ 2pd and the resistance and internal inductance per unit length of the conductor can be calculated as R0 ¼
1 2pdsd
(4:55)
L0i ¼
1 2pdvsd
(4:56)
More accurate results can be obtained by substituting Equation 4.48 into Equation 4.51: sd d 1j 1þ I_ ¼ 2pdE_ b 2 2d
(4:57)
Calculations in the Conductor’s Skin Layer
129
Thus 2 1 E_ b 1 d 1 d d 2 1j 1þ 1 þjþO 2 ¼ ¼ _I‘ d 2pdsd 2d 2pdsd 2d
(4:58)
and R0 ¼
1 d 1 2pdsd 2d
(4:59)
1 2pdvsd
(4:60)
L0i ¼
These formulae have been obtained assuming no tangential variation of the field. In the case of homogeneous distribution of the field along the interface, one can introduce the following quantities defined at each point of the cross section’s contour: E_ b ¼ R00 þ jvL00i K_ 1 d 1 R00 ¼ sd 2d L00i ¼
1 vsd
(4:61) (4:62) (4:63)
where the quantities R00 ¼ R00 (j) and L00i ¼ L00i (j) are functions of the contour coordinate j. Resistance and inductance can also be calculated using the Poynting vector:
þ _ ~_ ~ _ 2 ðR þ jvLi Þ E H* ~ nds ¼ jIj
(4:64)
S
~_ represents the conjugate of the magnetic field phasor and the normal where H* unit vector ~ n is directed outward to volume V bounded by surface S. In our two-dimensional case, Equation 4.64 can be represented in the following form using surface-related coordinates (x, j, h): þ L
*
_ 2 R0 þ jvL0 _ b dj ¼ jIj E_ b H j i
(4:65)
130
Surface Impedance Boundary Conditions: A Comprehensive Approach
where L is the contour of the cross section of the conductor. Using Equation 3.55, we can calculate the Poynting vector in the Leontovich approximation as follows: þ
þ 2 * _Eb H _ b dj ¼ 1 þ j H _ b dj j j sd
L
(4:66)
L
Substituting this result in Equation 4.64, we obtain R0 ¼ L0i
¼
1 _2 sdjIj
þ 2 _ b Hj dj
(4:67)
L
1 _2 vsdjIj
þ 2 _ b Hj dj
(4:68)
L
or, taking into account Equation 4.44, R0 ¼ L0i ¼
1 _2 sdjIj 1
_2 vsdjIj
þ
_ 2 dj jKj
(4:69)
L
þ
_ 2 dj jKj
(4:70)
L
It is easily demonstrated that Equations 4.69 and 4.70 reduce to Equations 4.55 and 4.56 in the case of the single circular conductor considered above. Indeed, in that case, I_ ¼ const K_ ¼ 2pd
(4:71)
and substitution of Equation 4.71 into Equations 4.69 and 4.70 results in Equations 4.55 and 4.56. In the next, higher approximation þ L
þ 2 * _Eb H _ b dj ¼ 1 þ d=(2d) þ j H _ b dj j j sd L
(4:72)
131
Calculations in the Conductor’s Skin Layer
and the improved R0 is
R0 ¼
1 þ d=(2d) sd
þ 2 _ b Hj dj L
2 I_
(4:73)
Another way to obtain R0 is to use the formula for ohmic power losses per unit length given by the Poynting theorem:
P0W
¼R I ¼ 0 _ 2
ð þ1
_ sE_ J*dj
(4:74)
L 0
Substituting Equations 4.24 or 4.25 into Equation 4.74, performing the required transformations and neglecting the terms of the order d2 d2 and higher, we obtain Equations 4.67 or 4.73, respectively. Note that Equations 4.67 and 4.68 allowing for nonhomogeneous distribution of the field along the surface have broader applicability than Equations 4.53 and 4.54 or Equations 4.55 and 4.56.
4.4 Forces Acting on the Conductor The force on a charged particle moving through a steady magnetic field may be written as the differential force d~ F exerted on a differential element of charge dQ d~ F ¼ dQ~ v~ B
(4:75)
where ~ u is the velocity of the charge. The differential element of charge may also be expressed in terms of the volume charge density rv and, consequently, current density ~ J: dQ ¼ rv dv
(4:76)
Substituting Equation 4.76 into Equation 4.75, we obtain d~ F ¼~ u~ Brv dv
(4:77)
132
Surface Impedance Boundary Conditions: A Comprehensive Approach
Taking into account that ~ u J ¼ rv~
(4:78)
we bring Equation 4.77 to the form: d~ F ¼~ J ~ Bdv
(4:79)
In long parallel conductors, the force is directed normal to the surface. Representing Equation 4.79 using surface-related coordinates, we obtain the differential force per unit length for time-harmonic fields: dF_ 0h ¼ J_ B_ j*dv
(4:80)
Integrating Equation 4.80 over the cross section of the conductor results in the formula for force per unit length: F_ 0h
þ
1 ð
¼ dj L
þ
J_ B_ j*dh ¼ dj L
0
1 ð
_ j*dh smE_ H
(4:81)
0
Substituting Equations 4.17 and 4.24 into Equation 4.81 and performing the integration, we obtain the force per unit length in the Leontovich approximation: ð þ 2 1 þ 2 _ b dj exp (2h=d) ¼ (1 þ j ) m H _ b dj _F0 ¼ 1 þ j m H h j j d 2 L
(4:82)
L
0
Using Equation 4.44, we can represent this formula in terms of the surface current density: þ þ m _b_ m _ 2 Hj K*dj ¼ (1 þ j ) K dj (4:83) F_ 0h ¼ (1 þ j ) 2 2 L
L
Correction of the order of the Mitzner approximation can be performed by substituting Equations 4.19 and 4.25 into Equation 4.81: ð þ 2 1 _ b dj 1 þ j þ d þ h 1 þ j þ d þ h2 1 þ j exp(2h=d) _F0 ¼ 1 m H h j d 2d d 4d2 4d2 L 0 2 þ þ 1 _ b 2 d 1þj d d d m _ b 2 d þ O 2 dj 1 þ þ j 1 þ ¼ m H j 1 þ j þ þ Hj dj d 2 2d 2d d 2d 2 L
L
(4:84)
133
Calculations in the Conductor’s Skin Layer
Finally, substituting Equation 4.45 into Equation 4.84 and neglecting the terms of the order d2 d2 and higher, we express the force in terms of the surface current density: þ _F0 1 þ d þ j 1 þ d m K_ 2 dj h 2d d 2
(4:85)
L
In this chapter, we demonstrated that for calculation of resistance, internal inductance, energy, power losses, and forces in the conductor under the condition of skin effect it is sufficient to know the distribution of the tangential component of the magnetic field along the surface of the conductor.
5 Derivation of SIBCs for Nonlinear and Nonhomogeneous Problems
5.1 Introduction In previous chapters, we assumed the properties of the conducting medium were constant. In practice, however, there are problems in which these assumptions are not valid and the governing Maxwell’s equations for the electromagnetic field distribution in conductors must be supplemented by additional equation(s) describing material properties. For example, the magnetic flux and magnetic induction in ferromagnetic materials are related through a nonlinear function. When currents flowing through the conductors are high enough, Joule heating starts to play a role and the dependence between conductivity and temperature cannot be neglected. In the latter case the heat transfer equation must be considered together with Maxwell’s equations. Objects consisting of several conductive layers or conductors coated with dielectrics belong to yet another class of problems where assumption of constant material properties is invalid. Intuitively it is clear that the SIBCs considered in Chapters 2 through 4 cannot be used directly in modeling nonlinear problems. However, the general application area of the surface impedance concept is not restricted to homogeneous conductors with constant material properties. The perturbation approach developed in Chapter 2 for derivation of SIBCs consists of two major stages: 1. Approximation of the original three-dimensional equations of the electromagnetic field distribution inside the conductor by onedimensional equations in the direction normal to the conductor’s surface 2. Derivation of the boundary conditions from analytical solution(s) of the one-dimensional equation(s) of diffusion of the magnetic field inside the conductor Stage 1 can be performed if the electromagnetic field in the skin layer varies in the normal direction much faster than in the tangential directions so that tangential derivatives are much smaller than normal derivatives. But this is exactly the main condition of applicability of the surface impedance concept 135
136
Surface Impedance Boundary Conditions: A Comprehensive Approach
to the problem under consideration. This means that the conditions of applicability of the surface impedance concept do not restrict the transformations performed at stage 2. In other words, even if the equations considered at stage 2 do not provide SIBCs in the classical single frequency form or do not allow exact analytical solution at all, it does not mean that the surface impedance concept cannot be applied. The objective of this chapter is to demonstrate how the perturbation approach developed in the previous chapters can be applied to derive boundary conditions for nonlinear and nonhomogeneous problems. It is expected that these SIBCs will be more complex than the classical conditions for homogeneous conductors. Thus we will restrict ourselves to low-order approximation to reduce the number of transformations. High-order terms can be derived following the same approach.
5.2 Coupled Electromagnetic–Thermal Problems When dependencies between the characteristics of the conductor’s material and temperature, T cannot be neglected, the equations governing electromagnetic field distribution in the conductor must be supplemented by the heat transfer equation in the form: cr
@T J2 ¼ r (lrT) þ s @t
(5:1)
where J, c, r, and l are the current density, specific heat, mass density, and thermal conductivity, respectively. We assume s and l to be functions of temperature and denote the electrical and thermal conductivities at T ¼ 293 K as s1 and l1, respectively. With the use of the following Maxwell equation: ~ ~ J ¼rH
(5:2)
we rewrite Equation 5.1 in the form: cr
@T ~ 2 ¼ r (lrT) þ s1 r H @t
(5:3)
Following the perturbation approach, we rewrite Equation 5.3 in nondimensional variables using the scale factors for basic quantities given in Table 3.1: T
~ @T l*t ~ ~ ~ ~ ~ ~ t I2 ~ 2 r l r T þ r H ¼T @~t crd2 cr~ ss*d2 D2
(5:4)
Derivation of SIBCs for Nonlinear and Nonhomogeneous Problems
137
where s ; s* ¼ s(T) s* ~ ¼ l ; l* ¼ l(T) l l* s ~¼
(5:5) (5:6)
Here s* and l* are the electrical conductivity and the thermal conductivity respectively, at the characteristic temperature T used as a scale factor. Representing the characteristic skin depth d and small parameter ~p in the forms: rffiffiffiffiffiffiffiffiffi t d¼ (5:7) s*m rffiffiffiffiffiffiffiffiffiffiffiffiffiffi d t ~ p¼ ¼ 1 (5:8) D s*mD2 transforms Equation 5.4 into the following form: T
2 2 ~ @T ~ ~ þs ~ m I ~ l ~T ~ H ~r ~ 1 r ¼ T~e2 r cr D2 @~t
(5:9)
where ~e ¼ ðl*s*m=(cr)Þ1=2
(5:10)
Introducing a scale factor for temperature as T¼
m I2 cr D2
(5:11)
and substituting Equation 5.11 into Equation 5.9, we finally write the heat transfer equation in nondimensional form: 2 ~ @T ~ ~ þs ~ ~ l ~T ~ H ~r ~ 1 r ¼ ~e2 r @~t
(5:12)
Equation 5.12 must be considered together with Equations 3.16 and 3.18 in describing the distribution of the tangential components of the electromagnetic field in the conductor. Substituting Equations 2.12 through 2.15, ~ , into Equation which give the vector operators in local coordinates j~1, j~2, h 5.12 and neglecting terms containing p2 (Leontovich’s approximation) in Equations 5.12, 3.16, and 3.18, we obtain ~ ~ @T @ ~ @T ¼ ~e2 l @h ~ @h ~ @~t
! 1
þs ~
2 X ~j @H i @ h ~ i¼1
!2 (5:13)
138
Surface Impedance Boundary Conditions: A Comprehensive Approach
~j @H k ~ j , k ¼ 1, 2 ¼ (1)3k s ~E 3k @h ~ ~j ~j @E @H 3k k p ¼ (1)3k ~ , k ¼ 1, 2 @h ~ @~t
~ p
(5:14) (5:15)
As follows from Equation 5.13, the distribution of temperature in the Leontovich approximation is described by a one-dimensional equation in the direction normal to the conductor’s surface. The tangential electric field at the interface can be calculated from Equation 5.14: ~b E j3k
@ p ¼ (1) ~ @~t
1 ð
k
~ j d~ H h k
(5:16)
0
The use of Equation 5.16 is impossible without solving the equation of diffusion of the magnetic field. The latter can be obtained, from Equations 5.14 and 5.15, and written in the form: ! ~ ~j @H @ 1 @ Hjk k s ~ , k ¼ 1, 2 (5:17) ¼ @h ~ @h ~ @~t Equations 5.13 and 5.17 must be solved together with the following boundary and initial conditions: ~ ! 0; ~t ¼ 0: H ~ ¼0 ~ ¼H ~ b; h ~ ! 1: H h ~ ¼ 0: H h ~ ¼ 0:
~ @T ~ T1 =T ; ¼g T @h ~
~ ! T1 =T; ~t ¼ 0: T
h ! 1:
(5:18) ~ ¼ T1 T T (5:19)
Solving the one-dimensional coupled problem (Equations 5.13 and 5.17 through 5.19), and substituting the results in Equation 5.16, we obtain the relationships between the electric field and the magnetic field at the conductor’s surface. Thus Equations 5.13 and 5.16 through Equation 5.19 can be treated as SIBCs for coupled electromagnetic–thermal problems in implicit form. Returning to dimensional variables in Equations 5.13 and 5.16 through Equation 5.19 gives @Hjk @ 1 @Hjk ¼ , k ¼ 1, 2 (5:20) (sm) @h @t @h 2 X @Hji 2 @T @ @T 1 ¼ l þs , k ¼ 1, 2 (5:21) cr @h @t @h @h ~ i¼1 Ebj3k ¼ (1)k m
@ @t
1 ð
Hjk dh 0
(5:22)
Derivation of SIBCs for Nonlinear and Nonhomogeneous Problems
139
As an example, consider the following temperature dependencies of conductivity and thermal conductivity: s¼
s1 1 þ aT
l ¼ xsT
(5:23) (5:24)
where the coefficients a and x can be assumed to be constant for pure materials (for example, in copper x ¼ 2.23 108). Substitution of Equations 5.23 and 5.24 into Equations 5.20 and 5.21 leads to the following equations: ! ~j ~j @H @H s* @ k k ~ 1 þ aTT ¼ @h ~ s1 @ h ~ @~t !2 ! 2 > X ~j ~ ~ ~ >@ T @H T l1 @ @T > 2 1 i > þs ~ : ~ ¼ ~e ~ @h l* @ h @h ~ ~ 1 þ aT T ~ @t i¼1 8 > > > >
@H @ > k k ~ > ¼ s1 m 1 þ aT T > < @h ~ @h ~ @~t (5:27) 2 X > @Hji 2 @T @ T @T > 1 > >cr ¼ xs1 þs : @h @t @h 1 þ aT @h i¼1 Under conditions of skin effect, the temperature is maximum at the surface of current-carrying conductors and decays monotonically inside the conductor with the increase in distance from the surface. The distributions of the magnetic field inside the conductor in the linear and nonlinear cases are similar, but the behavior of the current density may change significantly due to the decrease in conductivity at high temperatures (as follows from Equation 5.23). This results in a shift of the maximum in the current density distribution from the surface to the areas where Joule heating is not as significant (and consequently, the conductivity is constant). This phenomenon can be visualized by solving the one-dimensional problem of electromagnetic field diffusion (Equation 5.27) into the conducting half-space shown in Figure 5.1. Typical distributions of the current density inside copper at high currents (1 MA) are shown in Figure 5.2. To demonstrate the importance of taking into account the temperaturedependent properties in Equations 5.23 and 5.24 for practical calculation of the surface current density along the conductor’s surface, consider a pair of
140
Surface Impedance Boundary Conditions: A Comprehensive Approach
ξ2
H = eξ2 H E = eξ1 E
FIGURE 5.1 One-dimensional problem of electromagnetic field diffusion into a conducting half-space.
ξ1
η
Dielectric
Conductor
J
Linear problem (σ = const) Nonlinear problem (σ = σ (T ))
η FIGURE 5.2 Distribution of the current density inside the conducting half-space at high currents (1 MA).
l
b
ξ
l
a FIGURE 5.3 Statement of problem: Electromagnetic field diffusion due to passage of a current pulse in a system of two parallel conductors.
τ
t
l
infinitely long identical parallel conductors with circular cross sections with equal and oppositely directed current pulses flowing from an external source as shown in Figure 5.3. Under these conditions the current density has only one component directed along the conductors. The radius of each conductor
Derivation of SIBCs for Nonlinear and Nonhomogeneous Problems
141
5 I * = 1.2 MA
τ = 10–3 s
Material is copper
K (107 A/m)
4 A PEC-limit B SIBC, nonlinear C SIBC, linear
3 2 Power losses: Pc /Pb = 0.91
1 a 0
1
2
3
b
ξ (mm) FIGURE 5.4 Distributions of the surface current density along half the contour around the cross section of one conductor.
and the distance between the conductors are equal to 0.01 m (this is the characteristic scale D). The magnitude and duration of the current pulse are equal to 1 MA and 0.001 s, respectively. Figure 5.4 shows distributions of the surface current density K, obtained using different approximations of the electromagnetic field diffusion into the conductors, as a function of the surface coordinate j. The origin is at point a and proceeds around the contour, ending at point b (due to double symmetry only half the contour around the cross section of one conductor is shown). Curve A was obtained using the perfect electrical conductor (PEC) limit (the electromagnetic field diffusion is neglected). Curve C was obtained using the classical Leontovich SIBC in which the conductors’ properties are assumed to be constant and curve B was obtained using the nonlinear SIBC in Equation 5.27. Obviously, the solution of the problem obtained in the PEC-limit does not depend on the properties of the conductor because the field diffusion into the body is not taken into account. The diffusion induces a redistribution of the current along the conductor surface with time. The effect of this process is to smooth out nonuniformities in the surface current density distribution. Increase of the resistivity with temperature leads to deeper penetration of the field into the conductor compared to penetration in the linear material. Thus the maximum surface current density is significantly lower in the case of the nonlinear SIBC than the PEC-limit or the SIBC in which thermal properties of the conductors are neglected. Figure 5.5 shows the increase in surface temperature with time at point a of the conductor cross section as a result of Joule heating. It is easy to see that the s(T) dependence becomes a very important factor limiting the maximum surface temperature of the conductor (the difference in the surface
142
Surface Impedance Boundary Conditions: A Comprehensive Approach
8 Linear SIBC
T (102 K)
6 Nonlinear SIBC
4 2 0
0
2
4
6 t (10–4 s)
8
10
FIGURE 5.5 Increase in temperature of the conductor’s surface due to passage of the current pulse.
temperature between the curves obtained using the linear and nonlinear SIBCs is about 15%). As a conclusion of this section we now demonstrate how SIBCs for coupled problems can be derived in terms of the magnetic vector potential formalism. Substituting Equation 3.91, (which relates the magnetic vector potential and current density), into Equation 5.1, we obtain cr
@T ~2 ¼ r (lrT) þ s1 r2 A @t
(5:28)
As we have seen in Chapter 3, the distribution of the vector potential in the conductor is described in Equation 3.99. The scale factor for the magnetic vector potential is given in Equation 3.105. Turning to nondimensional variables in Equations 3.99 and 5.28 and neglecting terms of the order O(p2) yields the following coupled one-dimensional equations: ~e @A jk , k ¼ 1, 2 @h ~2 @~t ! ! 2 ~e 2 X ~ ~ @2A @T @ T j 2 @ 1 k ~ l ¼ ~e þs ~ @h ~ @h ~ @h ~2 @~t i¼1 s ~ 1
~e @2A jk
¼
(5:29)
(5:30)
These must be supplemented by the boundary and initial conditions: e b ~ ~ ~ e ! 0; ~t ¼ 0: A ~e ¼ 0 ~e ¼ A ; h ~ ! 1: A h ~ ¼ 0: A jk jk jk jk
(5:31)
Derivation of SIBCs for Nonlinear and Nonhomogeneous Problems
143
Returning to dimensional variables in Equations 5.29 and 5.30 gives the required SIBC: 8 @ 2 Aejk @Aejk > > > ¼ sm > < @h2 @t !2 2 e 2 X @ A > @T @ @T j > 1 k > > :cr @t ¼ @h l @h þ s @h2 i¼1
(5:32)
5.3 Magnetic Materials The presence of magnetic materials complicates the calculation of the surface impedance significantly, due to the following main reasons: .
The relationship between the magnetic field intensity (H) and magnetic flux density (B) in magnetic materials is nonlinear. If the relationship between the two is plotted for increasing levels of field intensity, it will follow a curve up to a point where further increases in magnetic field strength will result in no further change in flux density. This condition is called magnetic saturation.
.
The hysteresis loop due to the ‘‘history-dependent’’ nature of magnetization of the ferromagnetic material. Once the material has been driven to saturation, the magnetizing filed can be dropped to zero and the material will retain most of its magnetization.
The equation of the magnetic field diffusion taking into account both factors does not admit analytical solution in the general case. Therefore, the classical approach in practical electrical engineering calculations was replacement of the real BH-dependence by idealized functions such as equivalent rectangular or ellipsoidal functions. The first of these was introduced in the so-called limit theory where B(H) ¼ B0 ¼ const
(5:33)
where B0 is the saturation magnetic flux density corresponding to the amplitude of the magnetic field Hm at the conductor’s surface. Consider now the one-dimensional problem of the electromagnetic field diffusion into a ferromagnetic half-space and assume time-harmonic variation of the magnetic field along the conductor’s surface: H ¼ Hm sin (vt þ cc )
(5:34)
144
Surface Impedance Boundary Conditions: A Comprehensive Approach
B0
FIGURE 5.6 Idealized magnetization curves in the limit theory: with hysteresis (left) and without hysteresis (right).
–Hc
B0
Hc –B0
H
0 –B0
where cc is given by the initial condition (Figure 5.6, left): Hjt¼0 ¼ Hm sin cc ¼ Hc
(5:35)
If the hysteresis loop can be neglected then cc is assumed to be zero and the BH-curve takes the form shown in Figure 5.6 (right). According to Equation 5.33, the magnetic flux density B is assumed to have a constant value from the surface to a certain depth. Beyond this depth, the flux density has an opposite value. The reversal of the flux defines the wave front which sweeps into the conductor with velocity v ¼ dh=dt. When this wave front reaches its maximum depth, a new wave front starts from the surface. The wave front moving inside the conductor induces an electric field that has constant value from the surface up to the wave front: E ¼ 2B0 v
(5:36)
where the multiplier 2 was introduced because the magnetic flux changes values from B0 to þB0. With these conditions, Equation 3.2 reduces to the form: dH ¼ sE ¼ 2sB0 v dh
(5:37)
Solution of Equation 5.37, which shows a linear decrease of the magnetic field inside the conductor, can be written in the form: H ¼ 2sB0 vh þ const, 0 < h < hm
(5:38)
where hm is the distance that the wave front travels during time tm. The integration constant in Equation 5.37 is obtained from the condition in Equation 5.35: const ¼ Hjh¼0; t¼0 ¼ Hc
(5:39)
Derivation of SIBCs for Nonlinear and Nonhomogeneous Problems
145
B0 H E η
–B0
FIGURE 5.7 Distribution of electromagnetic field quantities inside a ferromagnetic half-space with a rectangular BH-curve.
The distributions of the functions B, H, and E inside the conductor are shown in Figure 5.7. Substituting Equation 5.39 into Equation 5.38 and performing simple transformations, we obtain H Hc dh d(h2 ) ¼ ¼ 2v ¼ 2h sB0 dt dt
(5:40)
Dividing both sides of Equation 5.40 by v: H Hc d(h2 ) ¼ vsB0 d(vt)
(5:41)
Substituting Equation 5.34 into Equation 5.41 and integrating with respect to the variable vt, we obtain
h2m
Hm ¼ vsB0
v ðt
½sin (vt þ cc ) sin cc d(vt)
(5:42)
0
Simplifying Equation 5.42 gives the depth the wave front reaches at the end of each half-cycle. The wave front penetration as a function of time is therefore: dnl hm ¼ pffiffiffi ½cos cc cos (vt þ cc ) vt sin cc 1=2 2
(5:43)
where sffiffiffiffiffiffiffiffiffiffiffi 2Hm dnl ¼ vsB0
(5:44)
146
Surface Impedance Boundary Conditions: A Comprehensive Approach
The subscript ‘‘nl’’ in Equations 5.43 and 5.44 means nonlinear. By neglecting hysteresis (cc ¼ 0) and setting vt ¼ p, Equation 5.43 reduces to the following form: dnl ¼ dnl hm jvt¼p ¼ pffiffiffi [1 cos vt]1=2 2 vt¼p
(5:45)
Therefore, dnl is the maximum electromagnetic penetration depth in ferromagnetic materials and occurs when hysteresis is neglected. According to the limit theory in [1] and [2] the surface impedance for rectangular BH-curves and sinusoidal magnetic fields can be calculated using the following formula: Zsnl ¼
8 (2 þ j) 3psdnl
(5:46)
The sinusoidal magnetic field formula is valid only for a small number of eddy current problems where the magnetic field reaches the surface of the ferromagnetic (steel) body mostly in a tangential direction, since the tangential magnetic field remains unchanged through the interface. In all other cases the magnetic field reaches the surface of the conductor in a mostly normal direction due to its high permeability. In such cases the electric field, instead of the magnetic field, is assumed to be sinusoidal and another formula of surface impedance has been proposed in [3]: Zsnl
27p3 4 j ¼ pffiffiffi 1þ 3p 2 5sdnl
(5:47)
The main weakness of the limit theory is that it is only accurate for high fields with strongly saturated materials but not for low fields where the magnetic characteristics of the material can be assumed to be linear. Deeley [4] and Guerin et al. [5] presented a balanced method in which the linear and nonlinear models due to Agarwal [1] are combined and weighted by a function f(Hs) taking into account the degree of saturation: Zsmagnetic ¼ f (Hm )Zs þ ð1 f (Hm )ÞZsnl
(5:48)
where Zs is the Leontovich surface impedance for linear problems. The choice of function f is empirical. In [4] and [5], the following function was used: f (Hm ) ¼
1 m 1 þ k HHknee
(5:49)
Derivation of SIBCs for Nonlinear and Nonhomogeneous Problems
147
where Hknee corresponds to the value of the magnetic field at the knee of the BH-curve k is a coefficient to be chosen Another weighting function has been proposed in [6] in the form: Ð Hm
B(H)dH f (Hm ) ¼ 2 1 B(Hm )Hm
!
0
(5:50)
In the models considered above the surface impedance in the case of nonlinear BH-dependence was represented in a form similar to that of single frequency SIBCs for linear materials. However, in the general case neither magnetic nor electric fields in the conductor with nonlinear properties are actually sinusoidal. To overcome this difficulty we consider below the derivation of SIBCs for magnetic materials in the time domain. In the general case of nonlinear BH-dependence one should use Maxwell’s equations in the following form instead of Equations 3.1 and 3.2: ~ ¼ s~ rH E
(5:51)
@~ B r ~ E¼ @t
(5:52)
r ~ B¼0
(5:53)
~ B ¼ n H ~
(5:54)
We assume s to be constant and the BH-curve to be known. Equations 5.51 through 5.53 in the local coordinates j1, j2, h take the following forms: @Hjk H jk dk @Hh ¼ (1)3k sEj3k , @h dk h @jk dk h 2 X i¼1
(1)3i
k ¼ 1, 2
di @Hj3i ¼ sEh di h @ji
@Bj3k @Ejk Ejk dk @Eh ¼ (1)k , k ¼ 1, 2 @h dk h @jk dk h @t
(5:55)
(5:56)
(5:57)
148
Surface Impedance Boundary Conditions: A Comprehensive Approach
2 X
(1)i
i¼1
@Eji @Bh di ¼ di h @j3i @t
(5:58)
2 2 X @Bh X di @Bji 1 þ ¼ Bh @h d h @j h d i i¼1 i i¼1 i
(5:59)
Following the perturbation approach, we now switch to dimensionless variables by introducing basic scale factors. The basic scale factors I, D, and t are introduced in the same way as in Chapter 3. Equation 3.14 giving H also remains valid. The reluctivity corresponding to the characteristic magnetic field H is called characteristic reluctivity and denoted as n*. With these, the scale factors B and E for the magnetic flux and electric field, respectively, can be introduced as H I ¼ n* n*D DH I ¼ E¼ tn* n*t B¼
(5:60) (5:61)
The scale factor d (characteristic penetration depth) for the normal coordinate h can be written in the form: pffiffiffiffiffiffiffiffiffiffiffiffi d ¼ tn*=s (5:62) With dimensionless variables, Equations 5.55 through 5.59 are written in the form: ~ p
~ ~j ~j ~h H @H @H dk k k ~j , ¼ (1)3k E ~ p2 ~ p2 3k ~ ~ ~ @h ~ ~ ~ ~ @ jk ~ dk ph dk ph p2
2 X
(1)3i
i¼1
~ di
~j @H 3i ~h ¼E ~ h @~ji di p~
~ ~j ~j ~j ~h @B E @E @E dk 3k k k ¼ (1)k ~p ~ p ~ p , ~ ~k ~ @h ~ @~t ph ~ @~jk ph ~ dk ~ d 2 X i¼1
(1)i
k ¼ 1, 2
(5:63)
(5:64)
k ¼ 1, 2
~i ~h ~i @B @E d ¼ ~ ~ @~t ph ~ @ j3i di ~
(5:65)
(5:66)
2 2 ~ X X ~h ~ ji @B @B 1 di ~h þ~ p ¼~ pB ~ ph ~ @h ~ h @~ji ~ i¼1 di p~ i¼1 di ~
(5:67)
~ B ~ ¼ n=n* ¼ ~ H= n
(5:68)
Derivation of SIBCs for Nonlinear and Nonhomogeneous Problems
149
Neglecting terms containing ~ p2 transforms (Equations 5.63 through 5.67) into the following form: ~ p
~j @H k ~ j , k ¼ 1, 2 ¼ (1)3k E 3k @h ~
(5:69)
~h ¼ 0 E
(5:70)
~j ~j @E @B 3k k p ¼ (1)k ~ , k ¼ 1, 2 @h ~ @~t
(5:71)
2 X
(1)i
i¼1
~h ~i @B @E ¼ ~ @~t @ j3i
(5:72)
2 X ~h ~ ji @B @B ¼ ~ p ~ @h ~ i¼1 @ ji
(5:73)
The tangential electric field can be expressed in terms of the magnetic field or magnetic flux using Equations 5.69, 5.71, and 5.68 as follows: ~ j b @H 3k ~ b ¼ (1)k ~ p E jk @h ~ @ ~ b ¼ (1)3k ~ E p jk @~t
1 ð
0
@ ~ j d~ B h ¼ (1)3k ~p 3k @~t
(5:74) 1 ð
0
~j H 3k d~ h ~n
b ~ j b ~ nB) k @ Hj3k k @(~ b 3k ~ ~ ~ Ejk ¼ (1) p ¼ (1) p @h ~ @h ~ @ ~ b ¼ (1)3k ~ p E jk @~t
1 ð
~ j d~ B h 3k
(5:75)
(5:76)
(5:77)
0
As was shown in Chapter 2, by neglecting terms of the order of magnitude of ~ p2 and higher, the distribution of the electromagnetic field inside the conductor is approximated by a one-dimensional equations of diffusion in the direction normal to the conductor’s surface, which is assumed to be planar (the Leontovich approximation). Substituting Equation 5.69 into Equation 5.71, we obtain ~j ~j @2H @B k k ¼ 0, @h ~2 @~t
k ¼ 1, 2
(5:78)
150
Surface Impedance Boundary Conditions: A Comprehensive Approach
Boundary and initial conditions for Equation 5.78 can be formulated as follows: ~¼B ~b; h ~ ¼ 0:B
~ ! 0; ~t ¼ 0 : B ~¼0 h ~ ! 1:B
(5:79)
~ ¼H ~ b; h ~ ¼ 0:H
~ ! 0; ~t ¼ 0 : H ~ ¼0 h ~ ! 1:H
(5:80)
or
The question of which function (H or B) should be used as a state variable to solve the problem in Equation 5.78 is defined by the specific BH-curve used in a given problem. In most practical cases the magnetization curve is obtained by approximation of experimental data. Appropriate approximating functions should represent the whole range from the origin to saturation with a minimum error over the whole range. This subject has been widely considered in the literature using power series, transcendental functions, Fourier transforms, or exponential functions. Exponentials have been used in [7] to generate the following curve fit: B(H) ¼ a0 þ
4 X
ai exp (ai H)
(5:81)
i¼1
Table 5.1 gives the coefficients ai and ai for several magnetic materials. Figures 5.8 through 5.11 show very good agreement between measured and computed magnetization curves for the materials given in Table 5.1. Another approximation for the BH-curve for various materials has been proposed in [8]: h i (5:82) H ¼ n(B2 )B ¼ ~k1 exp ~k2 B2 þ ~k3 B TABLE 5.1 Values for the Coefficients ai and ai (Equation 5.81) for Different Magnetic Materials Cold Rolled Steel a0
4.128
a1
1.191
a2 a3
Cast Steel
Cast Iron
4.295
2.338
0.797
0.358
0.266
2.638 1.24
2.173 1.045
2.887 3.809
1.724 1.177
a4
0.852
0.325
4.859
1.525
a1
8.89 103
3.81 103
1.78 103
1.07 103
5
5
3.61 105
a2 a3 a4
1.4 10
3.69
Annealed Sheet Steel
5 3
1.91 10 15.2 10
3
1.27 10
3
1.01 10
3
12.7 10
1.73 10
3
4.57 103
3
3.3 103
22.9 10 17.9 10
Source: El-Sherbiny, M., IEEE Trans. Mag, 9(1), 60, March 1973. With permission.
Derivation of SIBCs for Nonlinear and Nonhomogeneous Problems
151
1.5
B (Wb/m2)
1.2
0.9
0.6 Cold rolled steel 0.3 0
1.6
0
3.2
4.8
H (A/m) FIGURE 5.8 Comparison between measured BH-curve for cold rolled steel (solid points) and simulation using Equation 5.81 (solid curve).
1.5
B (Wb/m2)
1.2 0.9 0.6 Cast steel 0.3 0
0
1.6
3.2 H (A/m)
4.8
FIGURE 5.9 Comparison between measured BH-curve for cast steel (solid points) and simulation using Equation 5.81 (solid curve).
152
Surface Impedance Boundary Conditions: A Comprehensive Approach
1.5
B (Wb/m2)
1.2 0.9 0.6 Annealed sheet steel 0.3 0
0
1.6
3.2
4.8
H (A/m) FIGURE 5.10 Comparison between measured BH-curve for annealed sheet steel (solid points) and simulation using Equation 5.81 (solid curve).
1.5
B (Wb/m2)
1.2
0.9
0.6
0.3 0
Cast iron
0
1.6
3.2
4.8
H (A/m) FIGURE 5.11 Comparison between measured BH-curve for cast iron (solid points) and simulation using Equation 5.81 (solid curve).
153
Derivation of SIBCs for Nonlinear and Nonhomogeneous Problems
1.8 1.5
B (T)
1.2 0.9 0.6 0.3 0
k1
k2
k3
+ ∙ Cast
49.4
1.46
520.6
x ∙ Annealed
2.6
2.72
154.4
∙ Cold rolled
3.8
2.17
396.2
2
4
6
8
H (kA/M) FIGURE 5.12 Comparison between measured and computed BH-curves using Equation 5.82 for three types of steel. (From Brauer, J., IEEE Trans. Magnet., 11(1), 81, January 1975. With permission.)
where kj are values determined from experimental data for any type of steel. In particular, for annealed steel k1 ¼ 2.6, k2 ¼ 2.72, k3 ¼ 154.4. Equation 5.82 also gives very good agreement between measured and computed data (see Figure 5.12). In dimensionless variables, Equation 5.82 is written in the following form by removing all terms higher than O(~ p2): ~ ¼~ ~ H nB;
h i ~ 0 )2 þ ~k3 n ¼ (n*)1 ~k1 exp ~k2 (B*)2 (B ~
(5:83)
With the aid of Equation 5.83, the problem in Equations 5.78 through 5.80 can be expressed in the form: 8 2 ~ ~ ~t ¼ 0; > nB)=@ h ~ 2 @ B=@ < @ (~ ~b ~b ¼ H h ~ ¼ 0: ~ nB > : ~ ! 0; ~t ¼ 0 : B ~¼0 h ~ ! 1:B
(5:84)
~ j on the surface of the conductor, the If we can specify the function H k ~ distributions of the function Bjk over the boundary layer can be obtained by solving the one-dimensional nonlinear problem in Equation 5.61. Substituting the results in Equation 5.76 or 5.77, we obtain the desired approximate boundary relation. Therefore, the relation in Equations 5.76 through 5.80 can be treated as the SIBC in implicit form. Note that the one-dimensional relation in Equation 5.78 is much more appealing for numerical implementation as compared with the initial equations in Equations 5.51 through 5.53.
154
Surface Impedance Boundary Conditions: A Comprehensive Approach
The results obtained by solving the one-dimensional problem in Equations 5.78 through 5.80 can also be used to calculate the normal components of the magnetic field or magnetic flux density at the conductor’ surface, using Equation 5.73: ~b ¼ ~ p B h
1 ð 2 X @ ~ j d~ B h i ~ji @ i¼1
(5:85)
1 ð ~ 2 X Bji @ d~ h ~ n ~ i¼1 @ ji
(5:86)
0
~b ¼ ~ H p~ n h
0
To illustrate these derivations consider again the passage of a current pulse in the system of conductors shown in Figure 5.3. The conductor is now cast steel (with the BH-curve given in [8]), I ¼ 100 A and t ¼ 0.1 s. The calculations are performed using four alternatives and the numerical results are shown in Figure 5.13: the PEC boundary conditions (curve A), the proposed nonlinear SIBC (Equations 5.76 through 5.80) (curve B), and two linear SIBCs (curves C and D). Curve D requires a preliminary nonlinear run. From this nonlinear run the point of largest field is used to determine the constant n. This maxfield method of choosing n leads to a 40% error in the calculated power losses. The accuracy of the power losses computed using a linear SIBC can be 4 A
K (kA/m)
3
I * = 100 A
τ = 0.1 s
Material is cast steel
C A B C D
B D
2
1
Comparison of power losses : PC /PB = 1.05
a 0 0
PEC-limit SIBC, nonlinear SIBC, linear SIBC, linear
PD /PB = 0.60 b 1
ξ (mm)
2
3
FIGURE 5.13 Distributions of the surface current density along half the contour around the cross section of one conductor shown in Figure 5.3.
Derivation of SIBCs for Nonlinear and Nonhomogeneous Problems
155
improved if the characteristic reluctivity n* is used (curve C) instead of the max-field n. However, the difference in the surface current density between curves B and C is larger than that between curves B and D. In some practical problems (for example, in induction heating) several nonlinear effects (s(T), l(T), and n (B)) play a combined role and have to be taken into account [9,10]. In such cases, the characteristic skin depth d and small parameter ~ p take the form: rffiffiffiffiffiffiffi tn* d¼ s* rffiffiffiffiffiffiffiffiffiffiffi d tn* ~ 1 p¼ ¼ D s*D2
(5:87) (5:88)
and the time-domain SIBC in implicit form can be represented as follows: ! 8 ~j ~j @H @B @ > 1 k k > > s ~ ¼ > < @h @h ~ ~ @~t ! !2 2 > X ~j ~ ~ > @H @T @ ~ @T > 2 1 i > l þs ~ : ~ ¼ ~e @h ~ @h ~ @h ~ @t i¼1
(5:89)
where the scale factor for temperature and the small parameter ~e have been used in the following form: T¼
1=2 1 I2 ; ~e ¼ l*s* ðn*crÞ crn* D2
(5:90)
The one-dimensional equations in Equation 5.89 must be solved with the boundary and initial conditions given in Equations 5.79 and 5.80, and Equations 5.19 and 5.20. The results are then be substituted into Equations 5.74 and 5.75, or Equations 5.76 and 5.77 to obtain the tangential electric field at the interface. Finally, we represent Equation 5.89 in dimensional variables as follows: ! 8 ~j > @ H @Bjk @ > 1 k > ¼ > < @h s @h @t 2 X > @Hji 2 @ @T > @T 1 > > cr ¼ l þ s : @h @t @h @h
(5:91)
i¼1
Implicit time-domain SIBCs require more effort in the numerical implementation than frequency domain conditions based on limit theory.
156
Surface Impedance Boundary Conditions: A Comprehensive Approach
However, time-domain SIBCs are not related to a single frequency and are therefore suitable for any BH-curve. In that sense, they have a wider application area.
5.4 Nonhomogeneous Conductors 5.4.1 PEC-Backed Lossy Dielectric Layer In this section, we consider the SIBC formulation for a thin lossy dielectric layer (Figure 5.14). The distribution of the electric and magnetic fields in the layer is described by the following equations: @~ E ~ ¼ s~ rH Eþe @t
(5:92)
~ @H r ~ E ¼ m @t
(5:93)
Let the thickness h of the layer be much thinner than the characteristic size D of the conductors surface so the following conditions are satisfied: hD
(5:94)
We introduce the scale factor h for coordinate h in the form h¼h
(5:95)
and represent the small parameter ~ p in the form: ~ p ¼ h=D 1
(5:96)
ξ h
η σμε FIGURE 5.14 A PEC-backed thin lossy dielectric layer.
PEC
Derivation of SIBCs for Nonlinear and Nonhomogeneous Problems
157
Scale factors for the electric and magnetic fields in both regions are obtained from Equations 3.14 and 3.15: H ¼ I=D;
E¼
m1 I t
(5:97)
Now, Equations 5.92 and 5.93 can be written with nondimensional variables neglecting the terms of order O(~ p2) (Leontovich’s approximation) as follows: ! ~j ~j @E @H 3k ~ k 3k ~ ¼ (1) Ej3k þ ~a , p @h ~ @~t
k ¼ 1, 2
~j ~j @E @H 3k k p ¼ (1)3k ~ , k ¼ 1, 2 @h ~ @~t
(5:98)
(5:99)
where ã ¼ e=(st). Equations 5.98 and 5.99 in the Laplace domain take the form: ~ p
~j @H k ~ j (1 þ ~a~s), k ¼ 1, 2 ¼ (1)3k E 3k @h ~
(5:100)
~j @E 3k ~j , p~sH ¼ (1)3k ~ k @h ~
(5:101)
k ¼ 1, 2
The electric field at the surface of the body (~ h ¼ 0) is ~b E j3k
b ~ j (1)3k @ H k ¼~ p , 1 þ ~s~a @ h ~
k ¼ 1, 2
(5:102)
At the opposite side of the layer (~ h¼~ h ¼ h=h ¼ 1) the tangential electric field is zero ~ ~ j ¼ @ H jk ¼ 0, h ~¼~ h: E 3k @h ~
k ¼ 1, 2
(5:103)
The distribution of the magnetic field is described by the following onedimensional equations of diffusion: ~j @2H k ~ j ¼ 0, ~s þ ~s2~a H k 2 @h ~
k ¼ 1, 2
(5:104)
158
Surface Impedance Boundary Conditions: A Comprehensive Approach
The boundary conditions for the equations in Equation 5.107 can be written in the form (k ¼ 1, 2): ~b ~j ¼ H h ~ ¼ 0: H jk k
(5:105)
~j @H k ¼0 h ~¼~ h: @h ~
(5:106)
The solution of the problem in Equations 5.104 through 5.106 can be written in the form: pffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ~h ~s þ ~s2~a ~ 2~ ~ ~ s þ s a þ exp h ~ exp h ~ h ~j ¼ H ~b pffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffi H jk k exp ~ h ~s þ ~s2~a þ exp ~h ~s þ ~s2~a
(5:107)
Taking the derivative with respect to h ~ , we obtain pffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ~h ~s þ ~s2~a ~ 2~ ~ ~ s þ s a þ exp h ~ exp h ~ h p ffiffiffiffiffiffiffiffiffiffiffiffiffiffi ~ @ H jk ~ b ~s þ ~s2~a pffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼H jk @h ~ exp ~ h ~s þ ~s2~a þ exp ~h ~s þ ~s2~a (5:108) Substituting Equation 5.108 into Equation 5.102, we obtain ~b E j3k
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ~h ~s þ ~s2~a ~ 2~ ~ ~ s þ s a exp exp h ~s ~b p pffiffiffiffiffiffiffiffiffiffiffiffiffiffi H ¼ (1)3k ~ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ~s þ ~s2~a jk exp ~ h ~s þ ~s2~a þ exp ~h ~s þ ~s2~a pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ~s ~ b tan h(~ ¼ (1)k ~ p pffiffiffiffiffiffiffiffiffiffiffiffiffiffi H h ~s þ ~s2~a) jk 2 ~s þ ~s ~a
(5:109)
In the case of a thin layer pffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffi h ~s þ ~s2~a tan h ~ h ~s þ ~s2~a ~
(5:110)
and Equation 5.109 becomes ~ b ¼ (1)k ~ ~b , E p~s~ hH j3k jk
k ¼ 1, 2
(5:111)
Derivation of SIBCs for Nonlinear and Nonhomogeneous Problems
159
Frequency- and time-domain analogs of Equation 5.111 can be written in the forms: ~_ b ¼ (1)k 2j~ ~_ b , E p~ hH j3k jk
k ¼ 1, 2
(5:112)
~b dH jk , d~t
k ¼ 1, 2
(5:113)
~ b ¼ (1)k ~ p~ h E j3k
These formulae are represented with dimensional variables in the following forms: _ b , k ¼ 1, 2 E_ bj3k ¼ (1)k jvmhH jk Ebj3k ¼ (1)k mh
dHjbk dt
, k ¼ 1, 2
(5:114) (5:115)
5.4.2 Two-Layer Conducting Structure We now consider derivation of the impedance boundary conditions at the surface of a body consisting of two layers of different materials shown in Figure 5.15. Material properties are assumed to be constant. The electromagnetic field distribution in each layer can be described by the following equations: ~ (i) ¼ si~ E(i) rH r ~ E(i) ¼ mi
~ (i) @H @t
ξ h
η σ1 μ1
σ2 μ2 FIGURE 5.15 A two-layer conducting half-space.
(5:116) (5:117)
160
Surface Impedance Boundary Conditions: A Comprehensive Approach
Here the index i denotes the region (1 or 2). The electromagnetic depth of penetration in each region is di ¼
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi t=(si mi )
(5:118)
Let the thickness h of the first layer and electromagnetic penetration depth d(2) in the second layer be much thinner than the characteristic size D of the conductors surface so that the following conditions are met: h D;
d1 D;
d2 D;
h þ d2 D
(5:119)
It is natural to suppose that h < d1
(5:120)
The scale factor h for coordinate h is now introduced in the form h ¼ h þ d2
(5:121)
and the small parameter ~ p in the form ~ p ¼ h=D
(5:122)
The scale factors for the electric and magnetic fields in both regions are H ¼ I=D;
E¼
m1 I t
(5:123)
Under these conditions Equations 5.116 and 5.117, which provide the distributions of the tangential electric and magnetic fields in the Leontovich approximation in each region can be written in the forms: ~ p
~ (i) @H jk @h ~
~ (i) @E j3k @h ~
~ (i) , ~2i E ¼ (1)3k g j3k p ¼ (1)3k ~
k ¼ 1, 2,
i ¼ 1, 2
(5:124)
~ (i) @H jk , k ¼ 1, 2, i ¼ 1, 2 @~t
(5:125)
where the nondimensional parameters g~i are ~1 ¼ h=d1 ; g
~2 ¼ hm2 =(d2 m1 ) g
(5:126)
161
Derivation of SIBCs for Nonlinear and Nonhomogeneous Problems
Equations 5.123 and 5.124 in the Laplace domain take the forms: ~ @H jk
(i)
~ p
~ , ~2i E ¼ (1)3k g j3k (i)
@h ~
~ @E j3k (i)
k ¼ 1, 2,
~ (i) , p~sH ¼ (1)3k ~ jk
@h ~
k ¼ 1, 2,
i ¼ 1, 2 i ¼ 1, 2
(5:127)
(5:128)
The electric field at the surface of the body (~ h ¼ 0) is ~ (1) @H (1) jk 3k 2 ~ ~1 Ej3k ¼ (1) ~ pg @h ~
(5:129)
The distribution of the magnetic field in both regions is described by the following one-dimensional equations of diffusion: ~ (i) @2H jk @h ~2
~2i g
~ (i) @H jk ¼ 0, @~t
k ¼ 1, 2,
i ¼ 1, 2
(5:130)
Equation 5.130 is represented in the Laplace domain as ~ @2H jk
(i)
@h ~2
~ ¼ 0, ~2i H ~sg jk (i)
k ¼ 1, 2,
i ¼ 1, 2
(5:131)
Boundary conditions for the equations in Equation 5.131 can be written in the form (k ¼ 1, 2): b (1) (1) ~ ~ h ~ ¼ 0: H jk ¼ H jk (1) ¼ H ~ ~ (2) ; h ~¼~ h: H jk jk
~ @H jk
(1)
@h ~
~ @H jk
(5:132)
(2)
¼
@h ~
;
~h ¼ h=h
~ (2) ¼ 0 h ~ ! 1: H jk
(5:133) (5:134)
Solutions of the problems in Equation 5.131 can be written in the forms: pffiffi (1) pffiffi (1) exp ~ ~ exp g ~ (1) ¼ X ~ ~ ~s ~ s þ Y H g h ~ h ~ 1 1 k jk k
pffiffi (2) pffiffi ~ exp g ~ (2) ¼ X ~ (2) exp ~ ~2 (~ H g2 (~ h~ h) ~s þ Y h ~h) ~s k jk k
(5:135) (5:136)
162
Surface Impedance Boundary Conditions: A Comprehensive Approach
Taking into account the boundary conditions in Equations 5.132 through ~ (i) can be obtained and written in the forms: ~ (i) and Y 5.134, the coefficients X k k
~ (1) X k
~ (1) Y k
¼
b ~ (1) H jk
pffiffi ~1 þ g ~2 Þ exp g ~1 ~h ~s ðg pffiffi pffiffi ~2 Þ exp g ~1 ~ ~2 Þ exp ~ ~1 þ g ~1 g ðg g1 ~h ~s h ~s þ ðg
b ~ (1) ¼H jk
pffiffi ~2 Þ exp g ~1 ~ ~1 þ g ðg h ~s pffiffi pffiffi ~1 ~ ~1 ~ g2 sh g h ~s þ 2~ h ~s 2~ g1 ch g
b ~ (1) ¼H jk
pffiffi ~2 Þ exp ~ ~1 g ðg g1 ~h ~s pffiffi pffiffi ~2 Þ exp g ~1 ~ ~2 Þ exp ~ ~1 g ~1 þ g ðg g1 ~h ~s h ~s þ ðg
b ~ (1) ¼H jk
pffiffi ~2 Þ exp ~ ~1 g ðg g1 ~ h ~s pffiffi pffiffi ~1 ~ ~1 ~ g2 sh g 2~ g1 ch g h ~s þ 2~ h ~s
(1) b ~ (2) ¼ H ~ X k jk b ~ (1) ¼H jk
(5:137)
(5:138)
2~ g1 pffiffi pffiffi ~1 þ g ~2 Þ exp g ~1 ~ ~2 Þ exp ~ ~1 g ðg g1 ~h ~s h ~s þ ðg
2~ g pffiffi 1 pffiffi ~ ~1 h ~s þ 2~ ~1 ~ 2~ g1 ch g h ~s g2 sh g
~ (2) ¼ 0 Y k
(5:139)
(5:140)
where ch(x) ¼
exp (x) þ exp (x) exp (x) exp (x) ; sh(x) ¼ 2 2
(5:141)
The normal derivative of the tangential magnetic field in the first region is obtained directly from Equation 5.135: ~ @H jk
(1)
@h
pffiffi (1) pffiffi pffiffi (1) ~ exp g ~ exp ~ ~1 ~s X ~1 h ¼g g1 h ~ ~s þ Y ~ ~s k k
(5:142)
Derivation of SIBCs for Nonlinear and Nonhomogeneous Problems
163
Application of Equation 5.142 on the surface of the conductor (~ h ¼ 0) gives pffiffi 2 b pffiffi ~ (1) b pffiffi ~ ~1 ~h ~s ~ ð þ g Þ exp g g @H 1 2 (1) jk (1) (1) ~ ~ ~ 4 pffiffi ~ ~ ~ ~ s g s g H ¼ X þ Y pffiffi 1 1 jk k k @h ~1 ~h ~s þ 2~ ~1 ~h ~s g2 sh g 2~ g1 ch g 3 pffiffi ~2 Þ exp ~ ~1 g ðg g1 ~ h ~s pffiffi pffiffi5 þ ~ ~1 h ~s þ 2~ ~1 ~ 2~ g1 ch g h ~s g2 sh g pffiffi pffiffi ~1 ~ ~1 ~ ~1 sh g ~2 ch g h ~s þ g h ~s pffiffi (1) b g ~ ~ 1 H jk (5:143) ¼ ~sg pffiffi pffiffi ~ ~ ~ ~ ~ sh g ~ ch g h ~s þ g h ~s g 1
1
2
1
Substituting Equation 5.143 into Equation 5.129, we obtain the desired SIBC between the tangential electric and magnetic fields at the conductor’s surface: pffiffi pffiffi ~ b ~ ~ ~ ~1 ~h ~s ~ s þ g sh g ch g h g p ffiffi 1 1 2 (1) ~ (1) b k 1 ~ ~1 ~s pg Ej3k ¼ (1) ~ pffiffi pffiffi H jk ~1 ch g ~1 ~ ~1 ~h ~s ~2 sh g g h ~s þ g pffiffi pffiffi ~1 ~ ~1 ~ sh g h ~s þ gg~~2 ch g h ~s p ffiffi 1 ~ (1) b pffiffi H p ~s pffiffi ¼ (1)k ~ jk ~1 ~ ~1 ~ h ~s þ gg~~2 sh g h ~s ch g
(5:144)
1
From Equation 5.126 it follows that ~2 m2 d1 s1 d1 g ¼ ¼ ~1 m1 d2 s2 d2 g
(5:145)
Substitution of Equation 5.145 into Equation 5.144 gives pffiffi s d pffiffi 1 1 ~1 ~ ~1 ~h ~s sh g ch g h ~s þ b p ffiffi (1) s ~ (1) b k 2 d2 ~ ¼ (1) ~ ~ s E p pffiffi s d pffiffi H j3k jk 1 1 ~1 ~h ~s ~1 ~ sh g h ~s þ ch g s2 d 2
(5:146)
Equation 5.146 is too complex for the inverse Laplace transform. We therefore restrict ourselves to its representation in the frequency domain: pffiffiffiffi s d pffiffiffiffi 1 1 ~1 ~ ~1 ~h 2j sh g ch g h 2j þ b b p ffiffiffiffi (1) s _E k 2 d2 ~ ¼ (1) ~ ~_ (1) p 2j pffiffiffiffi H j3k j pffiffiffiffi k s1 d 1 ~1 ~h 2j ~1 ~ sh g ch g h 2j þ s2 d 2
(5:147)
164
Surface Impedance Boundary Conditions: A Comprehensive Approach
Return to dimensional variables yields: s1 d1 shð(1 þ j)h=d1 Þ þ chð(1 þ j)h=d1 Þ b b 1 þ j s2 d2 k _ (1) H E_ (1) ¼ (1) j3k jk s1 d 1 d1 s1 shð(1 þ j)h=d1 Þ chð(1 þ j)h=d1 Þ þ s2 d 2
(5:148)
Therefore the surface impedance is 1þj Z¼ d1 s1
s1 d 1 chð(1 þ j)h=d1 Þ s2 d 2 s1 d 1 shð(1 þ j)h=d1 Þ chð(1 þ j)h=d1 Þ þ s2 d 2
shð(1 þ j)h=d1 Þ þ
(5:149)
Note that the surface impedance of layer ‘‘1’’ is Z(1) ¼
1þj d 1 s1
(5:150)
Equation 5.149 can be represented in another form: s 1 d1 chð(1 þ j)h=d1 Þ s 2 d2 Z ¼ Z(1) s1 d1 shð(1 þ j)h=d1 Þ chð(1 þ j)h=d1 Þ þ s2 d2 shð(1 þ j)h=d1 Þ þ
(5:151)
It is easy to see that Z ! Z(1)
when h=d1 ! 1
If layer ‘‘1’’ is so thin that sh ~ u(1) ~ h ~ u(1) ~ h;
ch ~ u(1) ~h 1
(5:152)
(5:153)
Equation 5.146 can be reduced to the following form: pffiffi s1 d1 ~1 ~ g h ~s þ b p pffiffi (1) b ffiffi b s2 d 2 ~ ~ (1) (1)k ~p ~sH ~ (1) ¼ (1)k ~ p ~s H E j3k j jk p ffiffi k s1 d1 ~ ~1 h ~s 1þ g s2 d2 pffiffi s1 d1 s1 d1 ~pffiffi ~1 h ~s ~1 ~ g 1 g h ~s þ s2 d 2 s2 d2 p ffiffi pffiffi b s1 d1 ~pffiffi ~ ~ (1) s1 d1 1 þ s2 d2 g ~ ~ ~ ~ p ~sH h s 1 h s g ¼ (1)k ~ jk s 2 d2 s1 d 1 1 s2 d 2 1 " # pffiffi (1) b s1 d1 pffiffi (s2 d2 )2 (s1 d1 )2 k ~ ~ ~1 h ~s p ~sH jk (1) ~ 1þg s 2 d2 s1 s2 d 1 d 2
(5:154)
Derivation of SIBCs for Nonlinear and Nonhomogeneous Problems
165
Representing the SIBC, in Equation 5.154, in the frequency and time domains, we obtain " # pffiffiffiffi _ (1) b s1 d1 pffiffiffiffi (s2 d2 )2 (s1 d1 )2 (1) b _E k ~ ~ (1) ~ ~1 ~ p 2jH 1þg h 2j (5:155) j3k jk s2 d 2 s1 s 2 d 1 d 2 " # b b b (s2 d2 )2 (s1 d1 )2 d ~ (1) b (1) k s1 d1 ~ (1) ~ ~ ~ ^ ~1 h (5:156) p Ej3k (1) ~ Hjk *T 2 þ g Hjk s2 d 2 s 1 s2 d 1 d 2 d~t Switching to dimensional variables in Equations 5.155 and 5.156, we obtain " # b b 2 2 _E(1) (1)k 1 þ j H ~_ (1) s1 d1 1 þ (1 þ j) h (s2 d2 ) (s1 d1 ) j3k s 1 s2 d 1 d 2 s1 d1 jk s2 d2 d1 " # h (s2 d2 )2 (s1 d1 )2 (1) b _ k1þj ~ H 1 þ (1 þ j) (5:157) ¼ (1) s1 s 2 d 1 d 2 s2 d2 jk d1 " # rffiffiffiffiffiffi b b b 2 2 m s d (s d ) (s d ) d 1 1 2 2 1 1 (1) (1) (1) 1 ^b þ h (5:158) Ej3k (1)k Hjk *T H jk 2 s 1 s2 d2 s1 s2 d 1 d 2 dt Naturally Equations 5.157 and 5.158 transform to Equations 3.55 and 3.56, respectively, by setting s1 ¼ s2 and m1 ¼ m2.
References 1. P.G. Agarwal, Eddy current losses in solid and laminated iron, AIEE Transactions, 78, Part 1, May 1959, 169–181. 2. W. MacLean, Theory of strong electromagnetic waves in massive iron, Journal of Applied Physics, 25(10), October 1954, 1267–1270. 3. D. Lowther and E. Wyatt, The computation of eddy current losses in solid iron under various surface conditions, COMPUMAG Conference, Oxford, S.W. 7, 1976. 4. E.M. Deeley, Flux penetration in two dimensions into saturating iron and the use of surface equations, Proceedings of the IEE, 126(2), February 1979, 204–208. 5. C. Guerin, G. Meunier, and G. Tanneau, Surface impedance for 3D nonlinear eddy current problems—Application to loss computation in transformers, IEEE Transactions Magnetics, 32(3), May 1996, 808–811. 6. W. Mai and C. Henneberger, Field and temperature calculations in transverse flux inductive heating devices heating nonparamagnetic materials using surface impedance formulations for nonlinear eddy-current problems, IEEE Transactions on Magnetics, 25(3), Part 1, May 1999, 1590–1593.
166
Surface Impedance Boundary Conditions: A Comprehensive Approach
7. M. El-Sherbiny, Representation of the magnetization characteristic by a sum of exponentials, IEEE Transactions on Magnetics, 9(1), March 1973, 60–61. 8. J. Brauer, Simple equations for the magnetization and reluctivity curves of steel, IEEE Transactions on Magnetics, 11(1), January 1975, 81. 9. F. Azzouz and M. Feliachi, Non-linear surface impedance taking account of thermal effect induction heating, IEEE Transactions on Magnetics, 37(5), Part 1, September 2001, 3175–3177. 10. J. Nerg and J. Partanen, A simplified FEM based calculation model for 3-D induction heating problems using surface impedance formulations, IEEE Transactions on Magnetics, 37(5), Part 1, September 2001, 3719–3722.
6 Implementation of SIBCs for the Boundary Integral Equation Method: Low-Frequency Problems
6.1 Introduction The numerical method best suited for use with SIBCs is the boundary element method (BEM) because in both the BEM and the SIBCs the functions are approximated at the same points on the interface between the media. When applied to an eddy current problem consisting of conducting and nonconducting regions, the BEM yields a system of two integral equations over the conductor’s surface with respect to two unknowns: the required function and its normal derivative at the conductor=dielectric interface [1]. Under the conditions of skin effect, the electromagnetic field behavior in the conductor is known, the surface impedance concept may be applied, and the formulation can be reduced to a single integral equation employing the fundamental solution of the Laplace equation. The extra unknown is eliminated using the SIBCs relating the function and its normal derivative at the conductor’s surface. This approach is almost ideal for solving timeharmonic problems. Because of this, BEM–SIBC formulations have been widely used for the analysis of skin and proximity effect problems of multiconductor systems [2–7]. In the general case of transient excitation of the surface integral equation, including the time-domain SIBCs in the form Equation 3.81, 3.118, or 3.132, must be solved at every time step due to the time convolution terms. This leads to a significant increase in the computational cost required for the solution and, for all practical purposes, renders its numerical implementation impractical. The best way to avoid this difficulty is the separation of variables in the formulation into space and time components. In this case, the integral equation for the space component needs to be solved only once for a given system of conductors, and the result multiplied by the time component to obtain the solution of the problem for any time dependence of the source. It is easy to see that the variables in the surface impedance function given in Equations 3.82 through 3.84 cannot be separated because the functions Tm are different. However, in the low-frequency case (the so-called quasistatic approximation, in which the displacement current can be neglected) each
167
168
Surface Impedance Boundary Conditions: A Comprehensive Approach
term in Equations 3.82 through 3.84, when considered independently, admits separations of variables. This circumstance and the fact that the SIBC in Equations 3.82 through 3.84 can be represented as power series in the small parameter are keys to the development of the formulation desired. Indeed, it is natural to suppose that use of the perturbation technique as described in Chapter 2 will lead to a set of integral equations so that every equation includes only one term of Equations 3.82 through 3.84 and, consequently, admits separation of variables. Such a formulation has a clear physical meaning: the zero-order integral equation gives the solution in the wellknown perfect electrical conductor limit and the other equations contribute corrections of the order of Leontovich, Mitzner, and Rytov approximations. Thus the total number of integral equations in the formulation will not exceed four. In fact, it may be even less depending on the problem as will be shown in Chapter 9. Derivation of this type of formulation is the aim of this chapter. Consider a system of N cylindrical conductors of arbitrary cross sections surrounded by a homogeneous nonconducting space as shown in Figure 6.1. The parameters of the conducting and nonconducting media are assumed to be constants. To simplify derivations and without loss of generality, let us assume that the magnetic permeabilities of the conducting and dielectric regions are the same. Let an external source produce quasisteady current pulses flowing through the conductors so that the condition of applicability of the surface impedance concept in Equation 2.1 is satisfied. In low-frequency problems, the electromagnetic field distribution in both the dielectric and conducting regions can be described by the following equations: Conducting domain: r ~ E ¼ m
~ @H @t
(6:1)
~ ¼ s~ rH E
(6:2)
y ξ=0 I
ξ I x
D FIGURE 6.1 Simulation setup.
L
ξ = πD
Implementation of SIBCs for the BEM: Low Frequency Problems
169
~¼0 rH
(6:3)
r ~ E¼0
(6:4)
~ @H r ~ E ¼ m @t
(6:5)
~¼0 rH
(6:6)
~¼0 rH
(6:7)
r ~ E¼0
(6:8)
Nonconducting domain:
6.2 Two-Dimensional Problems Consider the particular case of long parallel conductors of constant cross section as shown in Figure 6.1. We direct the tangential coordinate, j1, along the conductors; set d1 ! 1 and assume that no variation of the electromagnetic field takes place in this direction @f ¼0 @j1
(6:9)
where f denotes any function. Under these conditions, the problem of the electromagnetic field distribution can be considered as two-dimensional in ~1 ~ and A the plane of cross sections of the conductors, and the vectors ~ Ii , ~ E, H, can be represented in the following form: ~ Ii ¼ ðIi Þj1~ e1 ; ~ e1 ; E ¼ Ej1~
~ ¼ Aj ~ ~ ¼ Hj ~ e þ Hh~ e3 ; A e H 2 2 1 1
(6:10)
Further derivations for the two-dimensional case will be performed in terms of electric–magnetic fields and vector potential formalisms. 6.2.1 E–H Formalism In the two-dimensional case, Equations 6.1 through 6.8 are written in the following form with local variables (j2, h): Conducting domain: @Hj2 @E ¼ m @t @h
(6:11)
170
Surface Impedance Boundary Conditions: A Comprehensive Approach
~ ¼ s~ rH E
(6:12)
~¼0 rH
(6:13)
r ~ E¼0
(6:14)
r2 E ¼ 0
(6:15)
Nonconducting domain:
Since the distribution of the electric field in the dielectric region obeys the Laplace equation, the boundary integral equation method [1] yields the following surface integral equation ~cE þ
N þ X i¼1
Li
E
N X @G(2D) dj2 ¼ @~ n i¼1
þ G(2D) Li
@E dj @~ n 2
(6:16)
where Li is the contour of the cross section of the ith conductor ~c ¼ ~c(j2) is a nondimensional coefficient depending on the contour (for a smooth contour ~c ¼ 0.5) ~ n is the normal unit vector directed inside the conductor G(2D) is the fundamental solution of the two-dimensional Laplace equation in free space G(2D) j02 , j2 ¼ (2p)1 ln R j02 , j2
(6:17)
where R is the distance between the observation and integration points R j02 , j2 ¼ ~ r j02 ¼ j~ r ðj 2 Þ ~ r ~ r 0j
(6:18)
In the two-dimensional case, the SIBCs of the order of Rytov’s approximation given in Equations 3.59 and 3.60 reduce to the following form: Frequency domain: " !# 2 _b pffiffiffiffiffiffiffiffiffiffiffi b 1 _b 1 3 _b 1@ H j2 _Eb ¼ 1 _ H þ pffiffiffiffiffiffiffiffiffiffiffi Hj þ jvsmHj2 þ s 2d2 j2 jvsm 8d22 2 2 @j22 " !# _b _b H @2H jvm 1 1 3 1 j j b b 2 2 _ þ _ þ pffiffiffiffiffiffiffiffiffiffiffi þ ¼ pffiffiffiffiffiffiffiffiffiffiffi H H j2 2d2 jvsm jvsm 8d22 j2 2 @j22 jvsm
(6:19)
171
Implementation of SIBCs for the BEM: Low Frequency Problems
Time domain: " ! # 2 b 1 pffiffiffiffiffiffiffi @ b 1 b 1 3 b 1 @ H j2 b b E ¼ sm Hj2 * T2 þ H þ pffiffiffiffiffiffiffi H þ T s @t 2d2 j2 sm 8d22 j2 2 @j22 * 2 " ! # 2 b 1 @ pffiffiffiffiffiffiffi b 1 b 1 3 b 1 @ H j2 b b b ¼ H T þ pffiffiffiffiffiffiffi H þ smHj2 * T2 þ T s @t 2d2 j2 * 3 sm 8d22 j2 2 @j22 * 4 b
(6:20) where the functions Tib , i ¼ 2, 3, 4, are given in Equations 2.214 through 2.216, respectively. The index ‘‘b’’ will be omitted in this chapter since all quantities are related to the conductor’s surface. Let us represent the electric field as a combination of the source component, Es, and eddy component, Ee. Eðj2 , h, tÞ ¼ Es (t) þ Ee ðj2 , h, tÞ
(6:21)
Note that in the two-dimensional case the source field is a function of time only. Substituting Equation 6.21 in Equation 6.16, we obtain Es þ ~cEe þ
N þ X i¼1
Li
Ee
N X @G(2D) dj2 ¼ @~ n i¼1
þ G(2D) Li
@Ee dj @~ n 2
(6:22)
where the following identity was used: þ Es Li
@G(2D) dj2 ¼ Es @~ n
þ Li
@G(2D) dj2 ¼ (1 ~c)Es @~ n
(6:23)
The normal derivative of the electric field in Equation 6.22 can be replaced by the time derivative of the magnetic field using Equation 6.11 so that Equation 6.22 can be rewritten in the form: Es þ ~cEe þ
N þ X i¼1
Li
Ee
N X @G(2D) dj2 ¼ m @~ n i¼1
þ G(2D) Li
@Hj2 dj2 @t
Time domain (6:24)
E_ s þ ~cE_ e þ
N þ X i¼1
Li
@G dj2 ¼ jvm E_ e @~ n (2D)
N þ X i¼1
_ j dj2 G(2D) H 2
Frequency domain
Li
(6:25)
172
Surface Impedance Boundary Conditions: A Comprehensive Approach
The boundary integral equation (Equation 6.24) contains three unknowns, namely Es, Ee, and @Hj2=@t. Suppose the contours Li, i ¼ 1, 2, . . . , N, are discretized using M nodes alltogether, so that the total number of unknowns is (2M þ N) because Es has separate values for each conductor. Discretization of Equations 6.24 and 6.25 can provide only M algebraic equations. Another set of M equations can be obtained using the SIBCs in Equations 6.19 and 6.20 relating the functions Ee and @Hj2=@t at every node. And finally we have to add N equations by setting the total current flowing through each conductor: þ Hj2 dj2 ¼ Ii , i ¼ 1, 2, . . . , N (6:26) Li
In principle, the system of equations (Equations 6.24 through 6.26 and Equations 6.19 and 6.20) is now ready to be solved using an appropriate numerical technique. However, its direct implementation is impractical because substitution of Equations 6.19 and 6.20 into Equations 6.24 and 6.25 allocates terms containing time convolution products on the left-hand side of the integral equation. Obviously, numerical implementation would require considerably fewer computer resources if those terms moved to the righthand side. In the following, we demonstrate how this can be achieved using the perturbation approach developed in Chapter 2. As a first step, we represent the relations in Equations 6.24 through 6.26 and Equations 6.19 and 6.20 in the Laplace domain with nondimensional variables using the scale factors given in Table 3.1: X ~e þ ~ s þ ~cE E N
i¼1
þ Li
N þ X ~ (2D) ~ d~j ~ (2D) H ~e @G d~j2 ¼ ~s E G j2 2 @~ n i¼1 Li
2 ~ 1 3 ~ e ¼ ~s~ ~ j þ ~p2~s3=2 1 @ H j2 E H p ~s1=2 þ ~ p~s1 þ~ p2~s3=2 3k 2 ~ ~ ~ 2 @ j22 2d2 8 d2 þ ~ j dj2 ¼ ~I i , i ¼ 1, 2 . . . , N H 2
(6:27) ! (6:28) (6:29)
Li
Note that Equation 6.28 can be obtained by reducing Equation 3.47 to the two-dimensional case. We now expand the functions for which the solutions are sought in power series in the small parameter ~ p: X ~ em ; ~e ¼ ~ pE E 1
m¼1
X ~ ~ ~ ~ ¼ X ~pH ~ sm ; H ~s ¼ ~ pE E m 1
1
m¼0
m¼0
(6:30)
173
Implementation of SIBCs for the BEM: Low Frequency Problems
Substituting Equation 6.30 into Equations 6.27 through 6.29 and equating coefficients of equal powers of ~ p, the following equations for the coefficients of expansions are obtained: PEC limit (m ¼ 0): X ~ s ~s E 0 N
i¼1
þ
~ ~ (2D) H ~ 0 d~j2 ¼ 0 G
Li
þ
~ ~0 H
Li
(6:31)
j2
~I i d~j2 ¼
(6:32)
j2
Leontovich’s approximation (m ¼ 1): " # ~ e 1=2 ~0 ~ ¼ ~s ~s H E 1
(6:33)
j2
X ~ s ~s E 1 N
i¼1
þ Li
N þ ~ (2D) X ~ e @ G (2D) ~ e ~ ~ ~ ~ d~j2 E H 1 dj2 ¼ ~cE1 G 1 @~ n j2 i¼1 þ ~ ~ 1 d~j2 ¼ 0 H Li
(6:34)
Li
(6:35)
j2
Mitzner’s approximation (m ¼ 2): " # 1 ~ ~ ~ 1=2 1 e ~ ¼ ~s ~s ~ 1 þ ~s E H H 0 2 ~ 2 d j2 j2 2 X ~ s ~s E 2 N
i¼1
þ Li
(6:36)
N þ ~ (2D) X ~ e @ G (2D) ~ e ~ ~ ~ ~ d~j2 E H 2 dj2 ¼ ~cE2 G 2 @~ n j2 i¼1 þ ~ ~ 2 d~j2 ¼ 0 H Li
(6:37)
Li
(6:38)
j2
Rytov’s approximation (m ¼ 3):
2
~ ~ H 0
@ b 6 ~ ~ 6 e 1=2 ~ 1 1 3=2 3 3=2 1 ~ ~ ~ ~ H 2 þ ~s H 1 þ ~s H 0 þ ~s E3 ¼ ~s6~s 4 2 @~j22 2~ d2 8~ d22 j2 j2 j2 2
3 7 7 5
j2 7
(6:39)
174
Surface Impedance Boundary Conditions: A Comprehensive Approach
X ~ s ~s E 3 N
i¼1
þ Li
N þ X ~ (2D) ~ ~ e @ G ~ (2D) H ~ 3 d~j2 ¼ ~cE ~e d~j2 E G 3 3 @~ n j2 i¼1
(6:40)
Li
þ ~ ~ 3 d~j2 ¼ 0 H
(6:41)
j2
Li
Returning in Equations 6.31 through 6.41 to the frequency and time domains gives Frequency domain: X ~_ s 2j E 0 N
i¼1
þ Li
þ Li
~ ~ (2D) H ~_ 0 d~j2 ¼ 0 G
~ ~_ 0 H
(6:42)
j2
d~j2 ¼ ~I_ i
(6:43)
j2
~ _Ee ¼ pffiffiffiffi ~_ 0 ~ 2j H 1
(6:44)
j2
X ~_ s 2j E 1 N
i¼1
þ Li
~ ~ (2D) H ~_ 1 G
X ~_ e d~j2 ¼ ~cE 1 N
j2
i¼1
þ ~ ~_ 1 d~j2 ¼ 0 H Li
þ Li
~ ~_ e @ G d~j2 E 1 @~ n (2D)
(6:45)
(6:46)
j2
~ 1 ~ _ _Ee ¼ pffiffiffiffi ~ ~_ 0 ~ H 2j H 1 þ 2 ~ j2 2 d 2 j2 X ~_ s 2j E 2 N
i¼1
þ Li
(6:47)
N þ ~ (2D) X ~ ~ (2D) H ~_ 2 d~j2 ¼ ~cE ~_ e @ G ~_ e d~j2 E G 2 2 @~ n j2 i¼1 þ ~ ~_ 2 d~j2 ¼ 0 H Li
(6:48)
Li
(6:49)
j2
b ~ ~ 1 ~ 3 1 _ _ _Ee ¼ pffiffiffiffi ~ ~ ~ ~_ 0 þ p ffiffiffiffi H1 2j H 2 þ þ pffiffiffiffi H 3 2 ~ ~ 2 2j 8d2 2j j2 2 d 2 j2 j2
~ ~_ 0 @ H 2
@~j22
j2
(6:50)
Implementation of SIBCs for the BEM: Low Frequency Problems
X ~_ s 2j E 3 N
i¼1
þ Li
N þ X ~ (2D) ~ ~ (2D) H ~_ 3 d~j2 ¼ ~cE ~_ e @ G ~_ e d~j2 E G 3 3 @~ n j2 i¼1 þ ~ ~_ 3 d~j2 ¼ 0 H
175
(6:51)
Li
(6:52)
j2
Li
Time domain: N X ~s @ E 0 @~t i¼1
þ
~ ~ (2D) H ~ 0 d~j2 ¼ 0 G j2
(6:53)
Li
þ ~ ~ 0 d~j2 ¼ ~Ii H
(6:54)
j2
Li
~^ b ~ ~ ~0 * T ~e ¼ @ H ~0 * T ~b ¼ H E 1 2 2 j2 j2 @~t N þ N þ ~ (2D) X X ~ ~ (2D) H ~ 1 d~j2 ¼ ~cE ~e @G ~e ~s @ d~j2 E E G 1 1 1 j2 @~ n @~t i¼1 i¼1 Li
þ
(6:55) (6:56)
Li
~ ~1 H
j2
d~j2 ¼ 0
(6:57)
Li
1 ~ ~^ b ~^ b ~ ~ ~ ~0 * T ~0 * T ~1 * T ~1 * T ~e ¼ @ H ~b þ 1 H ~b ¼ H H E þ 2 2 3 2 2 (6:58) ~ ~ j2 j j j @~t 2 2 2 2 d2 2d2 N þ N þ ~ (2D) X X ~ ~ (2D) H ~s @ ~ 2 d~j2 ¼ ~cE ~e @G ~e E d~j2 E (6:59) G 2 2 2 j2 @~ n @~t i¼1 i¼1 Li
þ
Li
~ ~2 H
j2
d~j2 ¼ 0
(6:60)
Li
3 ~ ~0 @2 H j2 ~ ~ ~ ~e ¼ @ 6 ~1 * T ~0 * T ~2 *T ~b þ 1 H ~b þ 3 H ~b þ 1 ~ b7 E 4 H *T 3 2 3 4 45 j2 j2 j2 2 @~t @~j22 2~ d2 8~ d22 ~ ~0 @2 H 1 ~ 3 ~ j2 ~ ~ ~^ b 1 ~^ b ~ b b ~ ~ ^ ^ ~ H1 * T3 þ H0 * T4 þ ¼ H2 * T2 þ *T 4 2 2 ~ ~ ~ j2 j2 j2 2 @ j2 2 d2 8d2 2
(6:61)
176
Surface Impedance Boundary Conditions: A Comprehensive Approach
N X ~s @ E 3 @~t i¼1
þ Li
N þ X ~ (2D) ~ ~ (2D) H ~ 3 d~j2 ¼ ~cE ~e @G ~e d~j2 E G 3 3 j2 @~ n i¼1
þ
(6:62)
Li
~ ~3 H
j2
d~j2 ¼ 0
(6:63)
Li
~ b , i ¼ 2, 3, 4, are given in Equations 2.99, 2.115, and 2.116, The functions T i ~ ^ b , i ¼ 2, 3, 4, are obtained from Equations 2.217, respectively. The functions T i 3.138, and 3.139 and written in the form: ~ ^ b ¼ 4p~t3 1=2 T 2
(6:64)
~ ^b ¼ U ~0 T 3
(6:65)
~ ^ b ¼ (p~t)1=2 T 4
(6:66)
We emphasize the following basic advantages of the formulations in Equations 6.42 through 6.63: 1. The integral equations for the terms of the expansions differ only in the form of the right-hand side and can be solved by the same solution procedures. Therefore, new computational complications do not arise as compared with solving the problem using the well-known perfect conductor limit (m ¼ 0). 2. The convolution integrals in the integral equations are on the righthand sides only and can be calculated and tabulated before numerically solving the appropriate integral equation so that the computer resources required for the computations are greatly reduced as compared with the ‘‘original’’ equations in Equations 6.19, 6.20, 6.24 through 6.26. 6.2.2 A–K Formalism Derivations in terms of A–K are very similar to the E–H formulation we discussed in Section 6.2.1. In the low-frequency case, the distribution of the magnetic vector potential in nonconducting domains is described by the Laplace equation: r2 A ¼ 0
(6:67)
Therefore, the application of the boundary integral equation method gives the following integral equation over contours of the cross sections of the conductors:
Implementation of SIBCs for the BEM: Low Frequency Problems
~cA þ
N þ X i¼1
A
Li
N X @G(2D) dj2 ¼ @~ n i¼1
þ G(2D) Li
@A dj @~ n 2
177
(6:68)
Applying the representation in Equation 3.94 to the two-dimensional case, we obtain Aðj2 , h, tÞ ¼ As (t) þ Ae ðj2 , h, tÞ
(6:69)
Similar to Es, As is a function of time alone, but it has separate values for each conductor. Substitution of Equation 6.69 into Equation 6.68 yields As þ ~cAe þ
N þ X i¼1
Li
Ae
N X @G(2D) dj2 ¼ @~ n i¼1
þ G(2D) Li
@A dj @~ n 2
(6:70)
The additional relation between the magnetic vector potential and its normal derivative is given by Equation 3.125 which can be reduced to two dimensions and written in the following form: Frequency domain: ! 2_e pffiffiffiffiffiffiffiffiffiffiffi e @ A_ e 1 1 3 1 @ A e e ¼ jvsmA_ þ A_ þ pffiffiffiffiffiffiffiffiffiffiffi A_ þ @h 2d2 2 @j22 jvsm 8d22 " !# pffiffiffiffiffiffiffiffiffiffiffi 1 1 3 _ e 1 @ 2 A_ e A_ e e _ pffiffiffiffiffiffiffiffiffiffiffi þ ¼ jvsm A þ A þ 2d2 jvsm jvsm 8d22 2 @j22
(6:71)
Time domain: ! 2 b @Ae 1 b 1 3 b 1 @ H j2 pffiffiffiffiffiffiffi @ e b ¼ sm H þ pffiffiffiffiffiffiffi H þ Tb A * T2 þ @h @t 2d2 j2 sm 8d22 j2 2 @j22 * 2 " ! # @ 1 e 1 3 e 1 @ 2 Ae pffiffiffiffiffiffiffi e b b b smA * T2 þ ¼ A T þ pffiffiffiffiffiffiffi A þ T @t 2d2 * 3 sm 8d22 2 @j22 * 4 ! 1 e ^b 1 3 e 1 @ 2 Ae pffiffiffiffiffiffiffi e ^ b ^b ¼ smA * T2 þ A * T3 þ pffiffiffiffiffiffiffi A þ T 2d2 sm 8d22 2 @j22 * 4 (6:72) The remaining equation is the condition in Equation 6.29 imposed on the total current flowing in the conductor. To represent it in terms of the vector potential formalism, we first represent Equation 3.89 using Equation 2.5 in the form
178
Surface Impedance Boundary Conditions: A Comprehensive Approach
H j2 ¼
@Ae ¼K @h
(6:73)
where K is the surface current density. Thus, the condition in Equation 6.29 in terms of the magnetic vector potential for two-dimensional cases takes the form: þ Li
@Ae dj ¼ @h 2
þ Kdj2 ¼ Ii
(6:74)
Li
We emphasize that Equation 6.73 is valid for all approximations (Leontovich, Mitzner, and Rytov) unlike the three-dimensional case, in which the normal derivative of the vector potential is equal to the surface magnetic field=surface current density only in the Leontovich approximation. Equations 6.70 through 6.72 and Equation 6.74 can be solved to obtain the functions As(t), Ae(j2, t), and K(j2, t). To improve the efficiency of the numerical implementation, we transform Equations 6.70 and 6.74 to the Laplace domain in terms of nondimensional variables: X ~ s þ ~cA ~e þ A N
i¼1
þ Li
N þ ~ (2D) X ~ ~j ~ e @G ~ (2D) Kd d~j2 ¼ ~ p1 A G 2 @~ n i¼1
þ
(6:75)
Li
~ ~j2 ¼ ~I i Kd
(6:76)
Li
Representation of the functions As(t), Ae(j2, t), and K(j2, t) in asymptotic expansions in the small parameter ~ p is as follows: ~e ¼ A
1 X
~e ; ~ pm A m
~s ¼ A
m¼1
1 X
~s ; K ~¼ ~ pm A m
m¼0
1 X
~m ~pm K
(6:77)
m¼0
Taking into account that ~e ~ m ¼ @ Amþ1 , K @h ~
m ¼ 1, 2, 3
(6:78)
we obtain the following integral equations: PEC limit (m ¼ 0): X ~s A 0 N
i¼1
þ Li
~ (2D) K ~ 0 d~j2 ¼ 0 G
(6:79)
Implementation of SIBCs for the BEM: Low Frequency Problems þ
~ 0 d~j2 ¼ ~I i K
179
(6:80)
Li
Leontovich’s approximation (m ¼ 1): X ~s A 1 N
i¼1
þ
~e ~ (2D) K ~ 1 d~j2 ¼ ~cA G 1
Li
þ
N þ ~ (2D) X ~ e @ G d~j2 A 1 ~ @ n i¼1
(6:81)
Li
~ 1 d~j2 ¼ 0 K
(6:82)
Li
Mitzner’s approximation (m ¼ 2): X ~s A 2 N
i¼1
þ
X ~e ~ (2D) K ~ 2 d~j2 ¼ ~cA G 2 N
i¼1
Li
þ
þ Li
~ (2D) ~ e @ G d~j2 A 2 @~ n
~ 2 d~j2 ¼ 0 K
(6:83)
(6:84)
Li
Rytov’s approximation (m ¼ 4): ~s A 3
N þ X i¼1
Li
~e ~ (2D) K ~ 3 d~j2 ¼ ~cA G 3 þ
N þ ~ (2D) X ~ e @ G d~j2 A 3 ~ @ n i¼1
(6:85)
Li
~ 3 d~j2 ¼ 0 K
(6:86)
Li
~e ~ Additional equations relating A mþ1 and K m , m ¼ 1, 2, 3, can be obtained from the SIBCs (Equations 6.71 and 6.72). Switching to nondimensional variables in Equations 6.71 and 6.72, and substituting the expansions in Equation 6.77, we obtain ~e @A 1 ~e ¼ ~s1=2 A 1 @h ~
(6:87)
~e @A ~ e 2 ~e þ 1 A ¼ ~s1=2 A 2 1 ~ @h ~ 2d2
(6:88)
180
Surface Impedance Boundary Conditions: A Comprehensive Approach
2 ~e ~e @A ~ e 1 @ A 3 1 ~e þ 1 A ~ e þ ~s1=2 3 A ¼ ~s1=2 A þ 3 2 1 @h ~ 2 @ ~j22 2~ d2 8~d22
! (6:89)
~ e Since Equations 6.83 and 6.85 require the coefficients A mþ1 being expressed ~ in terms of Km , the relations in Equations 6.87 through 6.89 must be represented in the following form: ~e ~ ~ e ¼ ~s1=2 @ A1 ¼ ~s1=2 K A 0 1 @h ~ " #
~ e ~e ~e ~ e ¼ ~s1=2 @ A ~ 1 1 @ A1 1=2 2 ~ e ¼ ~s1=2 @ A2 þ~s1=2 1 A ~ 1 þ~s1 1 K ~ þ~ s ¼ s A K 0 2 1 @h ~ @h ~ ~ 2~d2 2~d2 @ h 2~d2
2 ~e ~e ~ e ¼ ~s1=2 @ A3 þ ~s1=2 1 A ~ e þ ~s1 3 A ~ e þ 1 ~s1 @ A1 A 3 2 1 @h ~ 2 @~j22 2~ d2 8~ d22 " # ~ e ~ e ~e ~e 1 1 @ A 3 3=2 @ A 1 3=2 @ 2 @ A 1=2 @ A3 2 1 1 ~s ~s þ þ þ ~s ¼ ~s @h ~ @h ~ @h ~ ~ 2 @~j22 @ h 2~ d2 8~ d22 " # 2 ~ ~ 2 þ 1 ~s1 K ~ 1 þ 3 ~s3=2 K ~ 0 þ 1 ~s3=2 @ K0 ¼ ~s1=2 K 2 @~j22 2~ d2 8~ d22
(6:90)
(6:91)
(6:92)
Finally, Equations 6.79 through 6.86 and Equations 6.90 through 6.92 must be returned to the frequency and time domains: Frequency domain: X ~_ s A 0 N
i¼1
þ
þ
~ (2D) K ~_ 0 d~j2 ¼ 0 G
(6:93)
Li
~_ 0 d~j2 ¼ ~I_ i K
(6:94)
Li
~_ e ¼ (2j)1=2 K ~_ 0 ¼ 1 j K ~_ 0 A 1 2 N þ N þ ~ (2D) _ e X ~_ e @ G _ s X ~ (2D) ~_ ~ ~ ~ d~j2 A1 A1 G K1 dj2 ¼ ~cA1 ~ @ n i¼1 i¼1 Li
þ Li
(6:95) (6:96)
Li
~_ 1 d~j2 ¼ 0 K
(6:97)
181
Implementation of SIBCs for the BEM: Low Frequency Problems
~_ e ¼ (2j)1=2 K ~_ 1 þ (2j)1 1 K ~_ 0 ¼ 1 j K ~_ 0 ~_ 1 j K A 2 2 2~ d2 4~d2 X ~_ s A 2 N
i¼1
þ
X ~_ e ~ (2D) K ~_ 2 d~j2 ¼ ~cA G 2 N
i¼1
Li
þ
þ Li
~ ~_ e @ G d~j2 A 2 @~ n (2D)
(6:100)
_ 2~ ~_ e ¼ (2j)1=2 K ~_ 2 þ 1 (2j)1 K ~_ 1 þ 3 (2j)3=2 K ~_ 0 þ 1 (2j)3=2 @ K0 A 3 2 @~j22 2~ d2 8~ d22 " # ~_ 0 1 j ~_ j ~_ 3(1 þ j) ~_ 1 þ j @2K ¼ K1 K0 K2 2 8 @~j22 4~ d2 32~ d22 X ~_ s A 3 N
i¼1
þ
X ~ (2D) K ~_ e ~_ 3 d~j2 ¼ ~cA G 3 N
i¼1
Li
þ
(6:99)
~_ 2 d~j2 ¼ 0 K
Li
"
(6:98)
þ Li
~ (2D) ~_ e @ G d~j2 A 3 @~ n
~_ 3 d~j2 ¼ 0 K
#
(6:101)
(6:102)
(6:103)
Li
Time domain: ~s A 0
N þ X i¼1
þ
~ (2D) K ~ 0 d~j2 ¼ 0 G
(6:104)
Li
~ 0 d~j2 ¼ ~Ii K
(6:105)
Li
~^ b ~e ¼ @ K ~0 * T ~ b ¼ K ~0 * T A 2 2 1 @~t N þ N þ X X ~ (2D) s (2D) ~ e ~ ~ ~ e @G ~ ~ d~j2 G K1 dj2 ¼ ~cA1 A A1 1 @~ n i¼1 i¼1 Li
þ
(6:106) (6:107)
Li
~ 1 d~j2 ¼ 0 K
(6:108)
Li
1 ~ ~^ b ~^ b ~e ¼ @ K ~1 * T ~1 * T ~b þ 1 K ~b ¼ K ~0 * T T K A þ 0* 3 2 2 3 2 @~t 2~ d2 2~d2
(6:109)
182
Surface Impedance Boundary Conditions: A Comprehensive Approach
~s A 2
N þ X i¼1
~e ~ (2D) K ~ 2 d~j2 ¼ ~cA G 2
i¼1
Li
þ
Li
~ (2D) ~ e @G d~j2 A 2 @~ n
(6:111)
~0 @ ~ ~b 1 ~ ~b 3 ~ ~ b 1 @2K ~b ¼ K2 * T2 þ T K1 * T3 þ K0 * T4 þ 2 * 4 2 ~ ~ ~ 2 @~t @ j2 2d2 8d2 " # 2~ 1 ~ ~ 3 ~ ~ 1 @ K0 ~^ b ~ b b b ~ ^ ^ ^ ¼ K2 * T 2 þ K1 * T 3 þ K0 * T 4 þ T 2 @~j22 * 4 2~ d2 8~ d22 ~s A 3
N þ X i¼1
~e ~ (2D) K ~ 3 d~j2 ¼ ~cA G 3
Li
N þ X i¼1
þ
(6:110)
~ 2 d~j2 ¼ 0 K
Li
"
~e A 3
N þ X
Li
~ (2D) ~ e @G d~j2 A 3 @~ n
~ 3 d~j2 ¼ 0 K
#
(6:112)
(6:113)
(6:114)
Li
Equations 6.93 through 6.114 have the same structure of the left-hand sides and, therefore, they incorporate all advantages listed at the end of Section 6.2.1 for Equations 6.42 through 6.63. 6.2.3 Common Representation As has been shown in Sections 6.2.1 and 6.2.2, integral equation formulations in terms of A–K and E–H formalisms are so similar that it is natural to find a common representation for them. Thus the purpose of this section is derivation of an integral equation formulation that can be used in terms of ~ e and both formalisms. First, we will represent the equations relating E m . ~ ~ e and K ~ m1 using the ~ m1 @ H @~t, m ¼ 1, 2, 3, and the equations relating A j2
m
surface impedance operator-function F defined in Equations 3.84 through 3.88. In the particular two-dimensional low-frequency case, the operator-function F of the order of Rytov’s approximation reduces to the following form: Dimensional variables: 8 > 1 _ 3 1 @ 2 f_ 1=2 _ > > f þ ( jvsm)1 f þ ( jvsm)3=2 2 f_ þ ( jvsm)3=2 < ( jvsm) 2d2 2 @j22 8d2 F(2D) [ f ] ¼ 2 > > f Tb þ 1 f Tb þ 3 f Tb þ 1 @ f T b TD > : * 2 * 3 * 4 2 2 * 4 8d22 @j2 2~d2
FD
(6:115)
183
Implementation of SIBCs for the BEM: Low Frequency Problems
Nondimensional variables: 8 _ 2~ > > > (2j)1=2~f_ þ (2j)1 1 ~f_ þ (2j)3=2 3 ~f_ þ (2j)3=2 1 @ f < h i > 2 2 2d2 2 @j2 8d2 ~(2D) ~f ¼ F 2~ > f 1 3 1 @ > ~ ~b ~ ~b ~ ~b > ~b > : f * T2 þ ~ f * T3 þ ~2 f * T4 þ 2 ~2 * T4 TD @ j2 2d2 8d2
FD
(6:116)
We introduce the operators L(2) m , m ¼ 1, 2, 3, as follows: Dimensional variables: ( L(2D) 1 [f]
¼
( jvsm)1=2 f_ f * T2b TD
FD
(6:117)
Here and below the index (2D) denotes quantities related to the twodimensional case. 8 1 1 _ > > f FD < ( jvsm) 2d2 (2D) (6:118) L2 [ f ] ¼ 1 > > : f * T3b TD 2d2 8 2_ > 3=2 3 _ 3=2 1 @ f > FD ( jvsm) f þ ( jvsm) > < 2 @j22 8d22 (2D) (6:119) L3 [ f ] ¼ 2 > > > 3 f Tb þ 1 @ f T b TD : * 4 2 2* 4 8d22 @j2 Nondimensional variables: h i ~ (2D) ~f ¼ L 1
(
_ (2j)1=2~f FD b ~f T TD *~
(6:120)
2
8 > (2j)1 1 ~f_ FD < h i > 2~ d2 ~ (2D) ~f ¼ L 2 1 > ~ ~ > : ~ f * T3b TD 2d2 8 _ 2~ > > > 3 (2j)3=2~f_ þ (2j)3=2 1 @ f < ~2 h i > 2 @~j22 ~ (2D) ~f ¼ 8d2 L 3 2 ~ > 3 ~ ~b 1 @ f ~b > > > : ~2 f * T4 þ 2 ~2 * T3 TD @ j2 8d 2
(6:121)
FD (6:122)
The operators L(2) m are related with the surface impedance operator-function F of various approximation orders as follows:
184
Surface Impedance Boundary Conditions: A Comprehensive Approach
Dimensional variables: Leontovich F(2D) [ f ] ¼ L(2D) 1 [f] Mitzner (2D) F(2D) [ f ] ¼ L(2D) 1 [ f ] þ L2 [ f ] Rytov (2D) (2D) F(2D) [ f ] ¼ L(2D) 1 [ f ] þ L2 [ f ] þ L3 [ f ]
(6:123) (6:124) (6:125)
Nondimensional variables: h iLeontovich h i ~ (2D) ~f ~ (2D) ~f ¼L F 1 h iMitzner h i h i ~ (2D) ~f þ ~pL ~ (2D) ~f ~(2D) ~f F ¼L 1 2
(6:127)
h iRytov h i h i h i ~ (2D) ~f ~ (2D) ~f þ ~ ~ (2D) ~f þ ~p2 L ~(2D) ~f ¼L F p L 1 2 3
(6:128)
(6:126)
Substituting Equations 6.120 through 6.122 into Equations 6.55, 6.58, and 6.61 for the E–H formalism and Equations 6.106, 6.109, and 6.112 for the A–K formalism, we obtain the following equations: E–H formalism: 2 3 ~ ~0
@ H @ ~ (2D) ~ j2 7 (2D) 6 e ~ ~ ~ ¼ L1 4 H0 L1 E1 ¼ 5 ~ j @t @t 2 2 3 2 3 ~ ~ ~0 ~1 @ H @ H j2 7 j2 7 (2D) 6 (2D) 6 e ~ ~ ~ E2 ¼ L2 4 5 þ L1 4 5 @~t @~t 2 3 2 3 2 3 ~ ~ ~ ~0 ~1 ~2 @ H @ H @ H j2 7 j2 7 j2 7 (2D) 6 (2D) 6 (2D) 6 e ~ ~ ~ ~ E3 ¼ L3 4 5 þ L2 4 5 þ L1 4 5 @~t @~t @~t
(6:129)
(6:130)
(6:131)
A–K formalism: ~ (2D) K ~ e ¼ L ~0 A 1 1 ~ (2D) K ~ (2D) K ~ e ¼ L ~0 L ~1 A 2 1 2 ~ (2D) K ~ (2D) K ~ (2D) K ~ e ¼ L ~0 L ~1 L ~2 A 3 3 2 1
(6:132) (6:133) (6:134)
Implementation of SIBCs for the BEM: Low Frequency Problems
185
~ and ~ Introducing nondimensional functions u v which can take the following forms depending on the formalism (E–H or A–K) used, we get ~ j =@~t EH formalism @H 2 ~¼ (6:135) u ~ K AK formalism ~ s EH formalism E ~ v¼ (6:136) ~ s AK formalism A ~ and ~ The functions u v can be expanded into power series in the small parameter ~ p: ~¼ u
1 X
~ ~m ; pm u
m¼0
~ v¼
1 X
~pm ~vm
(6:137)
m¼0
~ ~m , ~s , H The coefficients of expansions in Equation 6.137 are related with E m j2
~ s , and K ~ m as follows: A m
8 . ~
> > ( jvsm)1=2 > < @ji
FD
i¼1
2 > X > @fji > > : T2b * @ji i¼1
(6:166)
TD
8 2 X d3i di @ f_ji > > > ( jvsm)1 > < 2di d3i @ji i¼1 2 > X > d3i di @fji > > : T3b * 2di d3i @ji i¼1
TD
FD (6:167)
190
Surface Impedance Boundary Conditions: A Comprehensive Approach
" 8 2 X > 3d23i d2i 2di d3i 3=2 > > ( jvsm) > 8d2i d23i < h i > i¼1 ~ f ¼ L(3D) " 3 > 2 > X 3d23i d2i 2di d3i @fji > > Tb > þ : 4* @ji 8d2i d23i i¼1
f_ji 1 @ þ 2 @ji @ji 1 @ 2 @ji
@ 2 fji @j23i
!# @ 2 f_ji @ 2 f_ji FD þ @j23i @j2i !# @ 2 fji TD þ @j2i
(6:168) Here the following transformation was used: 2 X i¼1
@ 3 fj @ 3 f ji @ 3 fj þ 3i þ 2 2 3i 2 @ji @j3i @ji @ji @j3i
!
2 X @ ¼ j i¼1 i
@ 2 f ji @ 2 f ji þ @j23i @j2i
! (6:169)
The ‘‘nondimensional’’ versions of Equations 6.166 and 6.167 are the following:
h i ~f ¼ ~ (3D) ~ L 1
8 _ 2 @~ > X f ji > > > (2j )1=2 > < @~j i¼1
> 2 > X @~fji > > ~b * > :T 2 ~ i¼1 @ ji
FD
i
(6:170)
TD
8 _ 2 ~ > X > d3i ~di @~f ji > 1 > FD < (2j ) h i > ~~ ~ (3D) ~ i¼1 2di d3i @ ji ~ ~ L2 (6:171) f ¼ > 2 ~ ~fj ~ > X @ d d > 3i i i > ~b * > TD :T 3 ~~ ~ i¼1 2di d3i @ ji " !# 8 2 X > 3~ d23i ~ d2i 2~ di ~ d3i @~fji 1 @ @ 2~fji @ 2~fji 3=2 > > þ þ FD > (2j ) < h i > @~ji 2 @~ji @~j23i @~j2i 8~ d2i ~ d23i i¼1 ~ (3D) ~f ¼ ~ L " !# 3 > 2 2 2 2~ ~ ~ ~ ~ ~fj 1 @ @ 2~fj > X 3 d d 2 d @ @ f d > i 3i j 3i i i i i > b ~ * > þ þ TD :T 4 @~ji 2 @~ji @~j23i @ ~j2i 8~ d2i ~ d23i i¼1 (6:172) The operators L(3D) are related to the functions F as follows: m Dimensional variables:
2 h i X @ Leontovich ~ F f ji f ¼ @ji i¼1
(6:173)
Mitzner 2 h i h i X @ (3D) ~ ~ F fji f þ L2 f ¼ @ji i¼1
(6:174)
L(3D) 1 L(3D) 1
Implementation of SIBCs for the BEM: Low Frequency Problems Rytov 2 h i h i h i X @ (3D) ~ (3D) ~ ~ L(3D) F f f þ L f þ L f ¼ ji 1 2 3 @ji i¼1
191
(6:175)
Nondimensional variables: Leontovich 2 h i X @ ~h~ i (3D) ~ ~ ~ L1 F f ji f ¼ ~ i¼1 @ ji Mitzner 2 h i h i X @ ~h~ i ~f þ ~ ~f ¼ ~ (3D) ~ ~ (3D) ~ L p L F f ji 1 2 ~ ~a1 i¼1 @ ji Rytov 2 h i h i h i X @ ~h~ i (3D) ~ (3D) ~ 2 ~ (3D) ~ ~ ~ ~ ~ ~ L1 F f ji f þ~ pL2 f þ~ p L3 f ¼ ~ ~a1 i¼1 @ ji
(6:176)
(6:177)
(6:178)
Substituting Equations 6.176 through 6.178 in the integral equation (Equation 6.163), we obtain N ð ~ X f þ 2 i¼1 Si
! ! 3 N ð 3 h i ~ (3D) X X X @G ~ (3D) m ~ (3D) (3D) m ~ (3D) ~ ~ ~ ~ ~ ~ ~p Lm [rf] d~s ¼ ~ ~p Lm Hfil d~s G f G n Hfil @~ n m¼1 m¼1 i¼1 Si
(6:179)
~ in the form of expansions in the small parameter p: We represent the function f ~¼ f
1 X
~m pm f
(6:180)
m¼1
Substituting the expansions in Equation 6.180 into Equation 6.179 and equating the coefficients of equal powers of p, we obtain the following equations for the expansion coefficients: PEC-limit (m ¼ 0): ~0 þ ~cf
N ð X i¼1
Si
N ð ~ (3D) X ~ ~ (3) ~ ~ s d~s ~ 0 @G nH d~s ¼ f G @~ n i¼1
(6:181)
Si
Leontovich’s approximation (m ¼ 1): ~1 þ ~cf
N ð X i¼1
Si
N ð h i X ~ (3D) ~ ~ (3D) H ~ (3D) L ~ s rf ~ 1 @G ~ 0 d~s d~s ¼ f G 1 @~ n i¼1 Si
(6:182)
192
Surface Impedance Boundary Conditions: A Comprehensive Approach
Mitzner’s approximation (m ¼ 2): ~2 þ ~cf
N ð X i¼1
Si
N ð h s i ~ (3D) X
~ ~ (3D) H ~ (3D) rf ~ (3D) L ~ rf ~ 2 @G ~0 þ L ~ 1 d~s d~s ¼ f G 2 1 @~ n i¼1 Si
(6:183) Rytov’s approximation (m ¼ 3): ~3 þ ~cf
N ð X i¼1
Si
N ð h i ~ (3D) X
~ ~ (3D) H ~ (3D) rf ~ (3D) rf ~ (3D) L ~0 þ L ~1 þ L ~ 2 d~s ~ 3 @G ~ s rf d~s ¼ f G 3 2 1 @~ n i¼1 Si
(6:184)
Obviously, the integral equations (Equations 6.181 through 6.184) can be reduced to the two-dimensional case of long parallel conductors. In this are related to L(2D) as follows: case, the operators L(3D) m m ! 8 > @ f_ > 1=2 @ > FD < ( jvsm) @j2 @j2 L(3D) ½ rf ¼ d1 !1 1 > > @=@j1 ¼0 > ^ b * @ @f :T TD 2 @j2 @j2
L(3D) 2 ½rf d1 !1
" ¼
L(2D) 2
¼
L(2D) 3
@=@j1 ¼0
L(3D) 3 ½rf d1 !1
@=@j1 ¼0
"
@2f @j22 @2f @j22
" ¼
L(2D) 1
# @2f @j22
(6:185)
# (6:186) # (6:187)
Taking into account Equations 6.185 through 6.187, we adjust the integral equations (Equations 6.181 through 6.184) to the two-dimensional case in the following form: ~0 þ ~cf
N þ X i¼1
~1 þ ~cf
N þ X i¼1
Li
Li
N þ ~ (2D) X @G ~ ~ (2D) ~ ~ s d~j2 ~ ~ nH dj2 ¼ f0 G @~ n i¼1
N X ~ (2D) ~ 1 @G d~j2 ¼ f @~ n i¼1
(6:188)
Li
þ Li
" ~ (2D) ~ (2D) L G 1
# ~s ~0 @H @2f j2 d~j2 @~j2 @~j22
(6:189)
Implementation of SIBCs for the BEM: Low Frequency Problems
~2 þ ~cf
N þ X i¼1
Li
193
" # " #! N þ ~s X ~ (2D) ~0 ~1 @H @G @2f @2f j2 (2D) (2D) ~ (2D) ~ ~ ~ ~ L2 þ L1 d~j2 2 f2 dj2 ¼ G @~ n @~j2 @~j22 @~j2 i¼1 Li
(6:190) þ
~ (2D) ~ 3 @G f d~j2 @~ n i¼1 Li " # " # " #! N þ ~ s @2f X ~0 ~1 ~2 @H @2f @2f j2 (2D) (2D) (2D) ~ (2D) ~ ~ ~ L3 d~j2 ¼ 2 þ L2 2 þ L1 2 G @~j2 @ ~j2 @~j2 @~j2 i¼1
~3 þ ~cf
N X
(6:191)
Li
6.4 Properties of the Surface Impedance Function 1. The operators L(2D) and L(3D) admit separation of variables into space and time m m components. We assume that a function can be split into space and time components as follows: f (~ r, t) ¼ g(~ r)h(t) 2D case
(6:192)
~ f ð~ r, tÞ ¼ ~ gð~ rÞh(t) 3D case (6:193) h i (3D) ~ Then the operators L(2D) f , m ¼ 1, 2, 3, can be represented in m [ f ] and Lm (3D) u, and the the form of superposition of the space operators C(2D) m [u] and Cm ½~ time operators Vm [n]: (2D) L(2D) m [ f ] ¼ Vm [n]Cm [u], m ¼ 1, 2, 3 h i ~ L(3D) u, m ¼ 1, 2, 3 f ¼ Vm [n]C(3D) m ½~ m
(6:195)
C(2D) 1 [u] ¼ u
(6:196)
(6:194)
where
1 u 2d2
(6:197)
3 1 @2u uþ 2 2 @j22 8d2
(6:198)
2 X @uji @ji i¼1
(6:199)
C(2D) 2 [u] ¼ C(2D) 3 [u] ¼
u ¼ C(3D) 1 ½~
194
Surface Impedance Boundary Conditions: A Comprehensive Approach
u ¼ C(3D) 2 ½~
C(3D) u 3 ½~
2 X d3i di @uji 2di d3i @ji i¼1
( 2 X 3d23i d2i 2di d3i @uji 1 @ ¼ þ @ji 2 @ji 8d2i d23i i¼1
(6:200) @ 2 uji @ 2 uji þ @j23i @j2i
!) (6:201)
and ( V1 [h] ¼
1=2
V2 [h] ¼
FD
(6:202)
T2b
(sm) * h TD ( ( jvsm)1 h_ FD (sm)1 T3b * h
( V3 [h] ¼
( jvsm)1=2 h_
( jvsm)3=2 h_ (sm)
3=2
(6:203)
TD FD
T4b * h
(6:204)
TD
(3D) u, and the With nondimensional variables, the operators C(2D) m [u] and Cm ½~ time operators Vm [n] take the form:
~ (2D) ½u ~ ¼ u ~ C 1
(6:205)
1 ~ (2D) ½u ~ ¼ ~ C u 2 2~ d2
(6:206)
~ 3 1 @2u ~ (2D) ½u ~ ¼ ~þ u C 3 2 ~ ~ 2 @ j22 8d2
(6:207)
2 h i X @~ u ji ~ (3D) ~ ~ ¼ u C 1 ~ i¼1 @ ji
(6:208)
2 ~ h i X uj i di @~ d3i ~ ~ (3D) ~ ~ ¼ u C 2 ~ ~ ~ i¼1 2di d3i @ ji
(6:209)
( 2 h i X 3~ d23i ~ d2i 2~ di ~ u ji 1 @ d3i @~ (3D) ~ ~ ~ ¼ u C3 þ 2 2 ~ ~ @~ji 2 @~ji 8di d3i i¼1 h i ~1 ~ V h ¼
(
_ (2j )1=2 ~ h FD ~b * ~ T h TD 2
~ ji @ 2 u ~ ji @2u þ 2 ~ ~ @ j3i @ j2i
!) (6:210)
(6:211)
Implementation of SIBCs for the BEM: Low Frequency Problems
h i ~2 ~ V h ¼
(
195
_ (2j )1 ~ h FD ~b * ~ h TD T
(6:212)
_ (2j )3=2 ~ h FD ~b * ~ T h TD
(6:213)
3
h i ~3 ~ V h ¼
(
4
This property allows separation of variables into space and time components in the surface impedance function, F, using the perturbation technique. 2. The time operators Vm [n] have the following property: V2 [n] ¼ V1 ½V1 [n] and
V3 [n] ¼ V2 ½V1 [n] ¼ V1 ½V2 [n]
(6:214)
This property is easily proven in the Laplace domain where the operators in Equations 6.202 through 6.204 take the form: m ½ m V n ¼ T n,
m ¼ 1, 2, 3
(6:215)
where 1 (s) ¼ (sm)1=2 s1=2 ; T
2 (s) ¼ (sm)1 s1 ; T
3 (s) ¼ (sm)3=2 s3=2 T
(6:216)
From Equation 6.216, we obtain 1; 1T 2 ¼ T T
1 T 1 ¼ T 2 3 ¼ T 1 T 1T T
(6:217)
Thus
2 ½ 1 ½ 1 ½n 1 V 1n ¼ T 2 n ¼ T 1 T 1 V V n ¼ T n ¼ V
1 ½ 1 ½n 2 V 3 ½ 1n ¼ T 3 n ¼ T 2 T 2 V n ¼ T n ¼ V V
2 ½ 2 ½n 1 V 3 ½ 2n ¼ T 3 n ¼ T 1 T 1 V n ¼ T n ¼ V V
(6:218) (6:219) (6:220)
Application of the inverse Laplace transform identities to Equations 6.218 through 6.220 leads to Equation 6.214.
6.5 Boundary Element Formulations for Two- and Three-Dimensional Problems in Invariant Form Let an external source produce quasisteady currents flowing through the conductors so that
196
Surface Impedance Boundary Conditions: A Comprehensive Approach
Two-dimensional case: ( Ii (t) ¼
sp
Ii exp ( jvt) sp u0 (t) I ~ i
Time-harmonic regime Pulsed regime
, i ¼ 1, 2, . . . , N
(6:221)
Three-dimensional case: ~ Ii ð~ r, tÞ ¼
( sp ~ Ii exp ( jvt) sp ~ Ii ~ u0 (t)
Time-harmonic regime Pulsed regime
,
i ¼ 1, 2, . . . , N (6:222)
where ~ u0 is a known nondimensional time-dependent function that is the sp sp same for all conductors and Ii or ~ Ii , i ¼ 1, 2, . . . , N, are the magnitude (in two-dimensional case) or magnitude and direction (in three-dimensional case) of the total current, respectively. The index ‘‘sp’’ stands for spatial. From Equations 6.221 and 6.222 it follows that Two-dimensional case: 8 sp > jvI exp ( jvt) Time-harmonic regime dIi (t) < i ¼ d~ u (t) > dt : Iisp 0 Pulsed regime dt
(6:223)
Three-dimensional case: 8 sp > jv~ I exp ( jvt) Time-harmonic regime r, tÞ < i @~ Ii ð~ ¼ ~ sp du0 (t) > @t :~ Pulsed regime Ii dt
(6:224)
Taking into account Equations 6.221 and 6.222, the filamentary magnetic field in Equations 6.158 and 6.159 can be represented as 8 ~ ~s ~ r exp 2j~t Time-harmonic regime
sp ~ ~ s ~ ~ ~r, ~t ¼ H > ~ :H ~s ~ r ~ u0 ~t Pulsed regime sp ~ ~ ~s ¼ H sp
N ð ~ X ~r ~ ~r0 ~ ~ 0 ~I sp ~ ~ r dl i ~ ~0 3 i¼1 ~ ~ r r Li
(6:225)
(6:226)
197
Implementation of SIBCs for the BEM: Low Frequency Problems
Introducing the following nondimensional time-dependent functions ~um , m ¼ 1, 2, 3: ~um ~t ¼
(
~ m exp 2j~t ¼ (2j )m=2 exp 2j~t V ~ m ~u0 ~t ¼ T ~ m ~t ~*~u0 ~t V
Time-harmonic regime Pulsed regime
, m ¼ 1, 2, 3
(6:227) and using Equation 6.214 gives h i ~1 ~ ~2 ~ ~1 V ~ 1 ~u1 ~ u2 ¼ V u0 ¼ V u0 ¼ V
(6:228)
h i h i ~2 ~ ~ 1 ~u0 ¼ V ~3 ~ ~ 3 ~u0 ¼ V ~1 V ~1 ~ ~2 V ~ 2 ~u1 ~ u3 ¼ V u0 ¼ V u0 ¼ V u2 and ~ u3 ¼ V (6:229) ~ m , for which the integral equations in Equa~m , v ~m , and f We seek functions u tions 6.140 through 6.155 and Equations 6.188 through 6.191 in the twodimensional case and Equations 6.181 through 6.184 in the three-dimensional case have to be solved, in the following form: Two-dimensional case: ~ ~r, t ¼ u ~ sp ~m ~ um (t); r ~ u m ~
~ ~r , t ¼ ~ ~ um (t), vsp r ~ vm ~ m ~
m ¼ 0, 1, 2, 3
(6:230)
Three-dimensional case: ~ sp ~ ~m ~ ~r, t ¼ f um (t), r ~ f m ~
m ¼ 0, 1, 2, 3
(6:231)
~ (3D) can be represented in the following form ~ (2D) and L Then the operators L m m according to properties proven in the previous section: Two-dimensional case: (2D) sp sp ~ (2D) ½u ~1 ~ ~ (2D) u ~ ~0 ¼ V ~0 ¼ ~u1 C ~0 u L u0 C 1 1 1 (2D) sp sp ~ (2D) ½u ~2 ~ ~ (2D) u ~ ~0 ¼ V ~0 ¼ ~u2 C ~0 u u0 C L 2 2 2
~ (2D) ½u ~1 ~ ~ (2D) u ~ (2D) u ~1 ¼ V ~sp ~sp u1 C L ¼ ~u2 C 1 1 1 1 1 (2D) sp sp ~ (2D) ½u ~3 ~ ~ (2D) u ~ ~0 ¼ V ~0 ¼ ~u3 C ~0 u L u0 C 3 3 3
(6:232) (6:233) (6:234) (6:235)
198
Surface Impedance Boundary Conditions: A Comprehensive Approach (2D) sp sp ~ (2D) ½u ~2 u ~ (2D) u ~ ~1 C ~1 ¼ V ~1 ¼ ~u3 C ~1 L u 2 2 2 (2D) sp sp ~ (2D) ½u ~1 ~ ~ (2D) u ~ ~2 ¼ V ~2 ¼ ~u3 C ~2 u u2 C L 1 1 1
(6:236) (6:237)
Three-dimensional case: h s i i h i (3D) h ~ s ~ ~ ~ (3D) H ~1 ~ ~ (3D) H ~ ~ rf ~ rf ~ s rf ~0 ¼ V ~ sp ¼ ~u1 C ~ sp L H u C 0 sp sp 1 1 1 0 0 (6:238) h s i h i h i
sp sp ~ (3D) ~ (3D) ~ s s ~ (3D) H ~2 ~ ~ ~ ~ rf ~ ~ ~0 ¼ V ~ ~ ~ u L H H r f r f C C ¼ u 0 2 2 sp sp 0 0 2 2 (6:239)
sp sp ~ (3D) rf ~1 ~ ~ (3D) rf ~ (3D) rf ~1 ¼ V ~ ~ u1 C L ¼ ~u2 C (6:240) 1 1 1 1 1 h s i i h i (3D) h ~ s ~ ~ ~ (3D) H ~3 ~ ~ (3D) H ~ ~ rf ~ rf ~ s rf ~0 ¼ V ~ sp ¼ ~u3 C ~ sp H L u0 C sp sp 3 3 3 0 0
(3D)
~ (3D) rf ~2 ~ ~ (3D) rf ~ ~1 ¼ V ~ sp ¼ ~u3 C ~ sp L u1 C rf 1 1 2 2 2
(3D)
~ (3D) rf ~1 ~ ~ (3D) rf ~ ~2 ¼ V ~ sp ¼ ~u3 C ~ sp u2 C L rf 2 2 1 1 1
(6:241) (6:242) (6:243)
Substituting Equations 6.225 through 6.227 and 6.232 through 6.243 into Equations 6.140 through 6.155, 6.181 through 6.184, and 6.188 through sp ~sp 6.191, we obtain the following equations for the spatial components ~vm , u m sp (two-dimensional case), and w ~ m (two- and three-dimensional cases): Two-dimensional case: E–H and A–K formalisms: sp ~ v0
N þ X i¼1
þ
~ (2D) u ~ ~sp G 0 dj2 ¼ 0
(6:244)
Li
~ ~(i) ~sp u 0 dj2 ¼ Isp
(6:245)
Li sp ~ v1
N þ X i¼1
N sp X ~ (2D) u ~ (2D) u ~ ~sp ~0 cC G 1 dj2 ¼ ~ 1 i¼1
Li
þ Li
~ ~sp u 1 d j2 ¼ 0
þ Li
~ (2D) (2D) sp @G ~ ~1 d~j2 (6:246) u C 1 @~ n
(6:247)
199
Implementation of SIBCs for the BEM: Low Frequency Problems
sp ~ v2
N þ X i¼1
sp sp ~ (2D) u ~ (2D) u ~ (2D) u ~j2 ¼ ~c C ~sp ~ ~0 d þ C G 2 1 1 2
Li
N þ X ~ (2D) (2D) sp sp @G ~ ~ (2D) u ~ ~0 d~j2 u C þ C 1 1 2 ~ @ n i¼1
(6:248)
Li
þ
~ ~sp u 2 d j2 ¼ 0
(6:249)
Li sp ~ v3
N þ X i¼1
sp sp sp ~ (2D) u ~ (2D) u ~ (2D) u ~ (2D) u ~ ~sp ~2 þ C ~1 þ C ~0 c C G 3 dj2 ¼ ~ 1 2 3
Li
N þ ~ (2D) (2D) sp X sp sp @G ~ ~ (2D) u ~ (2D) u ~ ~ ~0 d~j u C þ C þ C 1 2 3 2 1 ~ @ n i¼1
(6:250)
Li
þ
~ ~sp u 3 d j2 ¼ 0
(6:251)
Li
H–f formalism: ~ sp þ ~cf 0
N þ X i¼1
~ sp þ ~cf 1
N þ X i¼1
Li
Li
N þ ~ (2D) X ~ ~ (2D) ~ ~ s d~j2 ~j2 ¼ ~ sp @ G nH d f G sp 0 @~ n i¼1
(6:252)
Li
2 3 ~ ~s @ H N þ 2 ~ sp (2D) ~ X sp @ f0 7 ~ j2 ~ (2D) 6 ~ (2D) C ~ sp @ G d~j2 ¼ f G 5dj2 1 1 4 ~ @~ n @ j2 @~j22 i¼1 Li
(6:253) 3 1 2 " # ~ ~s @ H N þ N þ 2 ~ sp 2 ~ sp (2D) X X ~ sp @ f @ f @ G j2 0 7 1 C ~ ~ (2D) ~ (2D) 6 ~ (2D) B ~ sp þ ~ sp ~cf G d~j2 ¼ f 5þ C Adj2 @C 2 4 1 2 2 @~ n @~j2 @~j22 @~j22 i¼1 i¼1 0
Li
Li
3 2 ~ ~s sp @ H N þ N þ 2 (2D) ~ ~ X X sp @ f0 7 j2 ~ (2D) 6 ~ (2D) B ~ sp þ ~ sp @ G ~cf d~j2 ¼ f G @C 5 3 4 3 3 ~ @~ n @ j2 @~j22 i¼1 i¼1 0
Li
Li
"
2 ~ sp ~ (2D) @ f1 þC 2 @~j22
#
"
2 ~ sp ~ (2D) @ f2 þC 1 @~j22
#
(6:254)
!
d~j2 (6:255)
200
Surface Impedance Boundary Conditions: A Comprehensive Approach
Three-dimensional case: H–f formalism: ~ sp þ ~cf 0
N ð X i¼1
~ sp þ ~cf 1
N ð X i¼1
~ sp þ ~cf 2
N ð X i¼1
Si
Si
Si
N ð ~ (3D) X ~ ~ (3) ~ ~ s d~s ~ sp @ G ~ n H d s ¼ f G sp 0 @~ n i¼1
(6:256)
Si
N ð h i X ~ (3D) ~ ~ (3D) H ~ (3D) C ~ s rf ~ sp @ G ~ sp d~s d~ s ¼ f G sp 1 0 1 @~ n i¼1
(6:257)
Si
N ð h i ~ (3D) X
sp @ G ~ ~ (3D) rf ~ (3D) H ~ (3D) C ~ s rf ~ ~ sp þ C ~ sp d~s G f2 d~s ¼ sp 2 1 0 1 @~ n i¼1 Si
(6:258)
~ sp þ ~cf 3
N ð X i¼1
N ð h i X ~ (3D) ~ ~ (3D) H ~ (3D) C ~ s rf ~ sp @ G ~ sp ~ d s ¼ f G sp 3 0 3 @~ n i¼1 Si Si
~ (3D) rf ~ (3D) rf ~ sp þ C ~ sp d~s þC 2
1
1
2
(6:259)
We have obtained three sets of integral equations: Equations 6.244 through 6.251 for two-dimensional problems in terms of E–H and A–K formalisms; Equations 6.252 through 6.255 for two-dimensional problems in terms of the H–f formalism, and Equations 6.256 through 6.259 for three-dimensional problems in terms of the H–f formalism. The first equation in each set provides the solution in the perfect electrical conductor limit. The other integral equations give the corrections of the order of Leontovich, Mitzner, and Rytov’s approximations. We emphasize here that the formulation in Equations 6.244 through 6.251 does not depend on the formalism. In other words, the E–H and A–K formalisms in the two-dimensional produce the same integral equations for the spatial components. Note that the E–H and A–K approaches give the Fredholm surface integral equations of the first kind whereas the H–f approach leads to the Fredholm surface integral equations of the second kind. The latter is preferable from the numerical implementation point of view since the Fredholm surface integral equation of the first kind is the so-called ill-posed problem and requires special regularization algorithms [3]. Returning to dimensional variables using the scale factors given in Table 6.1 enables us to write Equations 6.244 through 6.259 in the following final form:
201
Implementation of SIBCs for the BEM: Low Frequency Problems
TABLE 6.1 Nondimensional Quantities and Their Scale Factors Nondimensional Quantity ~sp u k , k ¼ 0, 1, 2, 3
Scale Factor
Unit
I*D(kþ1)
A m(kþ1)
mI*t1 Dk
V m(kþ1)
I*Dk
A mk
I*Dk
A mk
~s H sp ~ (2D) G
I*D1
A m1
~ (2D) =@~ n @G ~ (3D) G
Nondimensional
Nondimensional
D1 D1
m1 m1
D2
m2
sp ~ vk , k ¼ 0, 1, 2, 3, E–H formalism ~sp v k , k ¼ 0, 1, 2, 3, A–K formalism ~ sp , k ¼ 0, 1, 2, 3 f k
~ (3D) =@~ @G n d~j2
d~s ¼ d~j1 d~j2 @=@~j1 , @=@~j2 ~ (2D) ½ , k ¼ 1, 2, 3 C k ~ (3D) ½ C k
D
m
D2
m2
D1
m1
kþ1
m(kþ1)
k
mk
D
, k ¼ 1, 2, 3
D
~ k, k ¼ 2, 3, 4 T d~t (in time convolution product)
(k3)=2
s(k3)=2 s
t t
Two-dimensional case: E–H and A–K formalisms: sp v0
N þ X i¼1
þ
sp
G(2D) u0 dj2 ¼ 0
(6:260)
Li sp
(i) u0 dj2 ¼ Isp
(6:261)
Li sp
v1
N þ X i¼1
N sp X sp G(2D) u1 dj2 ¼ ~cC(2D) u0 1 i¼1
Li
þ
þ Li
@G(2D) (2D) sp C1 u1 dj2 (6:262) @~ n
sp
u1 dj2 ¼ 0
(6:263)
Li sp
v2
N X i¼1
þ
sp sp (2D) sp G(2D) u2 dj2 ¼ ~c C(2D) u u þ C 1 2 1 0
Li
N þ X sp @G(2D) (2D) sp C1 u1 þ C(2D) u0 dj2 2 @~ n i¼1 Li
(6:264)
202
Surface Impedance Boundary Conditions: A Comprehensive Approach þ
sp
u2 dj2 ¼ 0
(6:265)
Li sp
v3
N þ X i¼1
sp sp (2D) sp (2D) sp G(2D) u3 dj2 ¼ ~c C(2D) u u u þ C þ C 2 1 0 1 2 3
Li
N þ X sp @G(2D) (2D) sp (2D) sp C1 u2 þ C(2D) u u þ C dj2 1 0 2 3 @~ n i¼1 Li
þ
sp
u3 dj2 ¼ 0
(6:266)
(6:267)
Li
H–f formalism: ~cfsp 0 þ
N þ X i¼1
~cfsp 1 þ
N þ X i¼1
sp
f1
Li
sp f0
Li
N X @G(2D) dj2 ¼ @~ n i¼1
@G dj2 ¼ @~ n (2D)
N þ X i¼1
Li
þ
~ s dj2 nH G(2D) ~ sp
(6:268)
Li
3 2 ~s @ H 2 sp sp @ f0 7 j2 6 G(2D) C(2D) 5dj2 1 4 @j2 @j22 (6:269)
3 1 2 " # ~s @ H N þ N þ 2 sp 2 sp (2D) X X sp @ f @ f @G j2 6 B sp (2D) 0 7 1 C ~cfsp dj2 ¼ f2 G(2D) @C(2D) 5 þ C1 Adj2 2 4 2 þ @~ n @j2 @j22 @j22 i¼1 i¼1 0
Li
Li
(6:270)
~cfsp 3 þ
N þ X i¼1
sp
f3
Li
N X @G(2D) dj2 ¼ @~ n i¼1
þ G(2D) Li
2 3 1 " # " # ~s sp sp sp @ H 2 2 2 sp @ f0 7 @ f1 @ f2 C j2 B (2D) 6 (2D) þ C(2D) 5 þ C2 @C 3 4 Adj2 1 2 2 @j2 @j2 @j2 @j22 0
(6:271) Three-dimensional case: H–f formalism: ~cfsp 0 þ
N ð X i¼1
Si
sp
f0
N X @G(3D) ds ¼ @~ n i¼1
ð Si
~ s ds nH G(3D) ~ sp
(6:272)
Implementation of SIBCs for the BEM: Low Frequency Problems
~cfsp 1 þ
N ð X i¼1
N ð X
~cfsp 2 þ
i¼1
sp
f1
Si
N X @G(3D) d~s ¼ @~ n i¼1
(3D) sp @G
f2
Si
@~ n
d~s ¼
N ð X i¼1
ð
h i ~ s rfsp ds H G(3D) C(3D) sp 0 1
203
(6:273)
Si
h i
~ s rfsp þ C(3D) rfsp ds H G(3D) C(3D) sp 2 1 0 1
Si
(6:274) ~cfsp 3 þ
N ð X i¼1
¼
Si
N ð X i¼1
sp
f3
@G(3D) ds @~ n
h i
~ s rfsp þ C(3D) rfsp þ C(3D) rfsp ds H G(3D) C(3D) sp 0 1 2 3 2 1
(6:275)
Si
where the operators C(2D) and C(3D) m m , m ¼ 1, 2, 3, are given in Equations 6.196 through 6.198 and Equations 6.199 through 6.201, respectively. We now formulate the procedure for calculation of the distribution of the electromagnetic field over the conductor’s surface, using Equation 6.26 as initial data. The formulation consists of the following three steps: E–H and A–K formalisms (two-dimensional problems): sp
sp
rÞ and um ð~ rÞ, m ¼ 0, 1, 2, 3, by solving 1. Obtain the space functions vm ð~ the surface integral equations (Equations 6.260 through 6.267). 2. Obtain the functions Es, As, Hj2, and K using the following formulae: 8 ! 3 X > > sp (1m)=2 sp > > m n0 þ exp ( jvt) nm ( jv) > < m¼1 s ! E ¼ 3 > X > d~u0 sp ~ > sp b > nm Tmþ1 * > m n0 u0 þ : dt m¼1 8 ! 3 X > > sp m=2 sp > > exp ( jvt) nm ( jv) n0 þ > < m¼1 s ! A ¼ 3 > > > nsp ~u þ X nsp Tb ~ > > m mþ1 * u0 0 0 :
Time-harmonic regime
(6:276) Pulsed regime
Time-harmonic regime
(6:277) Pulsed regime
m¼1
8 ! 3 X > > sp m=2 sp > > exp ( jvt) um ( jv) u þ > < 0 m¼1 ! Hj 2 ¼ K ¼ 3 > X > sp ~ > sp b ~ > u þ u T u u > 0 0 m mþ1 * 0 :
Time-harmonic regime
(6:278) Pulsed regime
m¼1
3. Obtain the functions Ee and Ae using Equations 3.132 and 3.118, respectively.
204
Surface Impedance Boundary Conditions: A Comprehensive Approach
H–f formalism (two- and three-dimensional problems): ~ s using the formula: 1. Calculate the field H sp N ð ~ r ~ r0 1 X s (i) 0 ~ ~ r Þ Isp ð~ dl Hsp ¼ 4p i¼1 j~ r ~ r 0 j3
(6:279)
Li
rÞ, m ¼ 0, 1, 2, 3, by solving the surface 2. Obtain the space functions fsp m ð~ integral Equations 6.268 through 6.271 in the two-dimensional case or Equations 6.272 through 6.275 in the three-dimensional case. 3. Obtain the scalar potential using the formula: f¼
sp u0 f0 ~
þ
3 X
! b fsp m Tmþ1
* ~u0
(6:280)
m¼1
At this point it is worth re-emphasizing the main advantages of the formulation developed: 1. The form of the integral equations, including the right-hand side, is independent of the time dependence of the incident current and is determined solely by the geometric parameters of the given system of conductors. Therefore, by solving the integral equations for space components just once for a given system of conductors, and multiplying the result by the corresponding time components we obtain solutions for any time dependence of the incident current. 2. The integral equations for the approximations differ only in the form of the right-hand side and can be solved by the same programmed routine; therefore, new computational complications do not arise beyond those involved in solving the problem in the well-known perfect electrical conductor limit.
6.6 Numerical Examples To illustrate the theory discussed in the previous sections, consider the skin and proximity effect calculation in a system of two long parallel copper conductors of circular cross section, shown in Figure 6.1. As was already mentioned in Section 6.2, the problem can be considered as two-dimensional in the xy-plane of the conductor’s cross sections. The characteristic scale, D, of the problem is the radius, d, of each conductor (equal to 21 mm) and the material of the conductor is copper. The distance between the centers of
Implementation of SIBCs for the BEM: Low Frequency Problems
205
the conductors is the parameter of the problem and is denoted as L. In view of the double axial symmetry of the conductors system, only a half-contour of the cross section of one of the conductors need to be considered so that the coordinate, j, varies in the range between 0 and pD as shown in Figure 6.1. sp r), given Figures 6.2 through 6.5 show distributions of the functions um (~ in Equations 6.260 through 6.267. To visualize the proximity effect, each of these figures contains two curves corresponding to L equal to 3D and 8D. sp The function u0 is the surface current density (or the tangential magnetic field) at the conductor’s surface in the PEC limit. Due to the proximity effect, the surface current density is higher at the inner side of the conductors carrying oppositely directed currents. From Figure 6.2 it is clearly seen that the proximity effect is stronger when the conductors are located closer to each other. The PEC limit can be used unless the skin layer is very thin, i.e., in the initial moments of time, when the electromagnetic field diffusion into the conductor can be neglected. Expansion of the skin depth induces ‘‘redistribution’’ of the magnetic field ‘‘along’’ the conductor’s surface. Obviously this phenomenon cannot be described within the PEC limit, but it is taken into sp account by the second term of expansion, u1 (Leontovich’s approximation). The effect of this process is to smooth out nonuniformities in the surface sp sp current density distribution so that u1 is negative where u0 is maximum and sp sp u1 is positive where u0 is minimum, as can be seen in Figure 6.3. It should sp also be noted that the magnitude of u1 is higher in the case of strong proximity effect. From Figures 6.4 and 6.5, which show the distributions of 18
u 0sp (A/m)
14
10
6
2
L = 3D L = 8D 0
0.8
1.6
2.4
ξ/D FIGURE 6.2 sp Distribution of the zero-order term u0 over half the contour of the cross section of one conductor for L ¼ 2D and L ¼ 8D.
206
Surface Impedance Boundary Conditions: A Comprehensive Approach
40
u1sp (A/(ms1/2))
0
–40
–80
–120
L = 3D L = 8D 0
0.8
1.6 ξ/D
2.4
FIGURE 6.3 sp Distribution of the first-order term u1 over half the contour of the cross section of one conductor for L ¼ 2D and L ¼ 8D.
L = 3D L = 8D
1200
u2sp, (A/(ms))
800
400
0
–400
0
0.8
1.6 ξ/D
2.4
FIGURE 6.4 sp Distribution of the second-order term u2 over half the contour of the cross section of one conductor for L ¼ 2D and L ¼ 8D.
Implementation of SIBCs for the BEM: Low Frequency Problems
207
5,000
–5,000
sp
u3 (A/(ms3/2))
0
–10,000
–15,000
L = 3D L = 8D 0
0.8
1.6 ξ/D
2.4
FIGURE 6.5 sp Distribution of the third-order term u3 over half the contour of the cross section of one conductor for L ¼ 2D and L ¼ 8D. sp
sp
u2 (Mitzner’s approximation) and u3 (Rytov’s approximation), respectively, it follows that each coefficient of expansions tends to compensate the effect of the previous term. Another conclusion is that the magnitudes of the terms grow with the index number. This is not unusual in asymptotic expansions. Next we demonstrate the evolution of the surface current density given by Equation 6.278 for the particular case of D ¼ 0.03 m and L ¼ 0.12 m (Figure 6.1). The system of conductors with these parameters will be used again in Chapter 9 for experimental validation of the SIBCs. The transient is obtained by closing a circuit breaker so that the resulting source current is exponential. Figure 6.6 shows the actual current pulse whereas the timedependent functions Tkb , k ¼ 0, 1, 2, 3, are shown in Figure 6.7. Figure 6.8 shows the distributions of the surface current density K obtained using the PEC, Leontovich’s, Mitzner’s, and Rytov’s boundary conditions compared with data obtained using a commercial FEM software for two instants of p ¼ 0:3) (left-hand side figure) and t ¼ 12 103s (p ¼ 0:43) time: t ¼ 6 103 s (~ (right-hand side figure). From these results it can be concluded that the SIBC formulation allows an efficient and accurate simulation of the test case. The hypothesis of PEC gives definitely worse results, whereas increasing the order of the SIBC formulation is closer to the FEM solution and to the experimental measurements, considering the uncertainty in the latter. However, it is unclear a priori, until what times the BEM–SIBC formulation may be used. We will take up this issue in Chapter 9 where we develop an analytical formula that gives approximate limits of applicability of the surface impedance concept.
208
Surface Impedance Boundary Conditions: A Comprehensive Approach
I (A)
120
80
40
0
0
0.01
0.02 t(s)
FIGURE 6.6 Current waveform.
Temporal component
2
1 T0 T1 T2 T3 0
0
0.01
0.02 t(s)
FIGURE 6.7 Time-dependent functions Tkb , k ¼ 0, 1, 2, 3, for the current pulse shown in Figure 6.6.
209
Implementation of SIBCs for the BEM: Low Frequency Problems
PEC Leontovich Mitzner Rytov FEM
1200
K (A/m)
1000
1200
1000
800 800 600 600 400 400
0
0.8
1.6 ξ/D
2.4
0
0.8
1.6 ξ/D
2.4
FIGURE 6.8 Distribution of the surface current density for t ¼ 6 103 s (~p ¼ 0.3) (left) and t ¼ 12 103 s (~ p ¼ 0.43) (right).
6.7 Quasi-Three-Dimensional Integro-Differential Formulation for Symmetric Systems of Conductors The two-dimensional formulations considered in Section 6.2 are valid only for very long conductors. In this section, we consider a system of N parallel finite length conductors with constant cross sections, connected by wires to a current source I(t) [10]. We choose a fixed Cartesian coordinate system with the x- and y-axes in the plane of the conductor’s cross sections and the z-axis parallel to the conductors’ axes. The system is assumed to satisfy symmetry about the x- and y-axes (the simplest example of such a configuration is shown in Figure 6.9). The current distribution in the conductors is assumed to be symmetric about the y-axis and antisymmetric about the x-axis. This particular case is important from the methodological point of view since it gives another application of the perturbation technique in multiconductor skin effect problems. As was shown in previous sections, the distribution of the magnetic scalar potential over the surface of the conductors, assuming small penetration depth of the field in the conductors, is described by an integral equation of the form
210
Surface Impedance Boundary Conditions: A Comprehensive Approach
y ξ
I(t) α r zc
x
z
h
D
FIGURE 6.9 A system of two current-carrying conductors of circular cross sections connected in series by wires.
~cf(s) þ
N ð X i¼1
f(s0 )
Si
@G(3D) 0 ds ¼ X(s) @~ n0
(6:281)
where X is the known right-hand side, which depends on the order of approximation G(3D) is given in Equation 6.161 The integral on the left-hand side of Equation 6.281 is composed of integrals over the ends and sides of the conductors. In the system considered here, the scalar potential distribution is symmetric about the axis of antisymmetry of the current distribution and antisymmetric about the axis of symmetry of the current. For the system shown in Figure 6.9, these conditions have the form: f( x, y, z) ¼ f(x, y, z);
f(x, y, z) ¼ f(x, y, z)
(6:282)
The integral over the lateral surfaces of the conductors can then be written in the form N ð X i¼1
Si
! zð2 4 ð @G(3D) 1 X 1 kþ1 0 0 0 @ ds ¼ f (1) dj f(j , z ) dz0 @~ n 4p k¼1 @~ n R(3D) k L
(6:283)
z1
where the summation is over the quadrants of the xy-plane, L is the total path of integration, including all of the contours of the conductors cross sections in a given quadrant, the coordinate j runs along the contour (we assume that j is continuous and goes around all of the contours of the conductor cross sections in succession), z1 and z2 are the coordinates of the ends of the are calculated as conductors, and the functions R(3D) k
Implementation of SIBCs for the BEM: Low Frequency Problems h i1=2 2 R(3D) ðj0 , j, z0 , zÞ ¼ ðz0 zÞ þR(2D) ðj0 , jÞ k k
211
(6:284)
ðj0 , jÞ are the distances between the point j, which in view of the where R(2D) k symmetry of the problem goes over only the part of L lying in the first quadrant of the xy-plane, and the points j0 lying in the first, second, third, and fourth quadrants, respectively. To calculate these distances, let r(j) be the radius vector to a point on the contour of the conductor cross section in the xy-plane and let a(j) be the angle between the radius vector and one of the coordinate axes, say y (Figure 6.9). The shape of the cross section is uniquely determined by these functions; therefore, assuming they are known, we have 8 2 2 (2D) 0 > > ¼ R ð j , j Þ ¼ ðr(j)Þ2 þðrðj0 ÞÞ 2r(j)rðj0 Þ cosðaðj0 Þ a(j)Þ a > 1 1 > > 2 > > > < a2 ¼ R(2D) ðj0 , jÞ ¼ ðr(j)Þ2 þðrðj0 ÞÞ2 2r(j)rðj0 Þ cosðaðj0 Þ þ a(j)Þ 2 2 2 > (2D) 0 > a ¼ R ð j , j Þ ¼ ðr(j)Þ2 þðrðj0 ÞÞ þ2r(j)rðj0 Þ cosðaðj0 Þ a(j)Þ > 3 3 > > > 2 > > : a4 ¼ R(2D) ðj0 , jÞ ¼ ðr(j)Þ2 þðrðj0 ÞÞ2 þ2r(j)rðj0 Þ cosðaðj0 Þ þ a(j)Þ
(6:285)
4
Differentiating the function 1=R(3D) with respect to the normal to the lateral surface of the conductor and substituting the result into Equation 6.283, we obtain zð2 4 ð 1 X bk ðj, j0 Þfðj0 , z0 Þ kþ1 0 0 (1) dj 3=2 dz 4p k¼1 2 0 0 z1 ðz zÞ þak ðj, j Þ L
(6:286)
where the functions bk ¼ @ak =@~ n0 are found directly from (Equation 6.286): 8
0 @rðj0 Þ > 0 @rðj Þ > b ¼ 2 r ð j Þ r(j) > 1 > 0 > @~ n @~ n0 > > >
> 0 > @rðj0 Þ > 0 @rðj Þ > r(j) > < b2 ¼ 2 rðj Þ @~ 0 n @~ n0
0 > @rðj0 Þ > 0 @rðj Þ > b ¼ 2 r ð j Þ þ r(j) > 3 > > @~ n0 @~ n0 > > >
> 0 > @rðj Þ @rðj0 Þ > > þ r(j) : b4 ¼ 2 rðj0 Þ @~ n0 @~ n0
cosðaðj0 Þ a(j)Þ rðj0 Þ sinðaðj0 Þ a(j)Þ
@aðj0 Þ @~ n0
cosðaðj0 Þ þ a(j)Þ rðj0 Þ sinðaðj0 Þ þ a(j)Þ
@aðj0 Þ @~ n0
cosðaðj0 Þ a(j)Þ rðj0 Þ sinðaðj0 Þ a(j)Þ
@aðj0 Þ @~ n0
@aðj0 Þ cosðaðj Þ þ a(j)Þ rðj Þ sinðaðj Þ þ a(j)Þ @~ n0 0
0
0
(6:287)
Transforming to dimensionless variables using the characteristic linear dimension, D, of the cross section and the distance, Z, over which the magnetic field parameters vary significantly along the axis as the length scales in the j- and z-directions, respectively, the integral in Equation 6.286 takes the form:
212
Surface Impedance Boundary Conditions: A Comprehensive Approach
zð2 4 ð 1 X fðj0 , z0 Þ kþ1 0 0 0 (1) bk (j, j )dj 3=2 dz ~ 2 4pb 0 k¼1 z z z1 L þak ðj, j0 Þ ~ b
(6:288)
where ~ ¼ D=Z b
(6:289)
is a small parameter. The inner integral in Equation 6.288 is represented as an asymptotic series ~ Applying the method described in [11], and neglecting terms of order inb. ~ 3 and higher, we obtain the asymptotic results: O b zð2
z1
z 2 ~ nþ1 n ð2 0 X b f(j0 , z0 ) @ f z z n f ð j 0 , z0 Þ 0 0 dz ¼
3=2
3=2 dz n 2 ~ n! @z 2 b n¼0 z0 z z0 z z1 þ ak (j, j0 ) þ ak ðj, j0 Þ ~ ~ b b
(6:290) Evaluating the integrals analytically and substituting the results into Equation 6.288, we have 9 8 > > ð = < 4 X 1 z z1 z2 z kþ1 bk 0 fdj (1) þ h i h i 1=2 1=2 > ak > 4p ; : ðz z Þ2 þ b ~ 2 ak ~ 2 ak k¼1 ðz2 zÞ2 þ b 1 L 8 > 4 2 ð X ~ b @f 0 bk < dj (1)kþ1 h 4p @z ak > : k¼1 L
9 > =
z z1
z2 z i1=2 h i1=2 > ; ~ 2 ak ~ 2 ak ð z z1 Þ þ b ðz2 zÞ2 þb 2
ð 4 4 X ~2 ð @2f ~2 b b @f 0 X kþ1 0 ~ ln b dj dj (1) b ln a (1)kþ1 bk k k 8p @z2 4p @z k¼1 k¼1 L
~2
b 8p
ð L
þ
@ f 0 dj @z2 2
4 X
8 >
=
z z1 z2 z (1)kþ1 bk h i1=2 h i1=2 > > : ð z z1 Þ 2 þ b ; ~ 2 ak ~ 2 ak k¼1 ðz2 zÞ2 þ b
4 1=2 X ~2 ð @2f b 2 kþ1 0 2 ~ dj (1) b ln z z þ ð z z Þ þ b a 1 1 k k 8p @z2 k¼1
L
1=2 ~ 2 ak z2 z þ ðz2 zÞ2 þ b
(6:291)
213
Implementation of SIBCs for the BEM: Low Frequency Problems
~ ! 1 under the integral sign, we obtain Taking the limit b X ð 4 4 X ~ 2 ð @f 1 1 bk b 1 dj0 fdj0 ð1Þkþ1 ð1Þkþ1 bk a 4p z 4p @z z z z 2 1 k k¼1 k¼1 L
L
ð 2 4 4 ~2 ð @2f 0 X ~2 b b @ f 0X kþ1 ~ ln b dj ð 1 Þ b ln a dj ð1Þkþ1 bk k k 8p @z2 4p @z2 k¼1 k¼1 L
~2
b 8p
ð L
4 X
~2
b @ f 0 dj ð1Þkþ1 bk þ 2 8p @z k¼1 2
ð L
L
4 X @ f 0 ln ½ 4 ð z z Þ ð z z Þ dj ð1Þkþ1 bk 2 1 @z2 k¼1 2
(6:292) It follows from Equation 6.287 that 4 X
ð1Þkþ1 bk ¼ 0
(6:293)
k¼1
Substituting Equation 6.293 into Equation 6.292, we finally have 1 4p
ð L
fdj0
4 X k¼1
ð1Þkþ1
4 X ~2 ð @2f bk b 0 dj ð1Þkþ1 bk ln ak ak 8p @z2 k¼1
(6:294)
L
We next consider the integral over the ends of the conductors. The normal to the surface in this case is along the z-axis and the integral in Equation 6.282 takes the form: N ð X i¼1
Si
f
N ð 4 X @G(3) 1 X ds ¼ ð1Þkþ1 f(s0 ) h @~ n 4p i¼1 k¼1 Si
zend z ðzend
2
zÞ þ a k
ðs, s0 Þ
i3=2 ds
0
(6:295) where zend is z1 or z2. Introducing scale factors D and Z leads to the following form of the integral: N ð 4 X zend z X 1 0 ð1Þkþ1 f(s0 )
3=2 ds ~ 2 4pb i¼1 end k¼1 z z Si þ ak ðs, s0 Þ ~ b
(6:296)
Let us now estimate this integral. Depending on the distance between the end of the conductor and the observation point, the following cases are possible:
214
Surface Impedance Boundary Conditions: A Comprehensive Approach
~ the observation point is close to the end of the cona. jzend zj b: ~ it is not difficult ductor. Transforming to the variable y ¼ (zend z)=b, to show that the integral in this case is of the order O(1) (without taking symmetry into account). b. zend z: the observation point is far from the end of the conductor. Let ˆ(z) ¼
4 1X (1)kþ1
3=2 2 ~ b k¼1 zend z 0 þ ak ðs, s Þ ~ b
(6:297)
2 ~ 2 end ~ in Equation 6.297 Factoring out the small parameter b ak z z b and expanding ˆ(z) in a power series in this parameter, we obtain
4 4 end 3 X 2 3 ~ 2 end kþ1 2 ~ ~ (6:298) ˆ(z) ¼ b ak z z (1) 1 b ak z z þO b 2 k¼1 It follows from Equation 6.287 that 4 X
(1)kþ1 ak ¼ 0
(6:299)
k¼1
Therefore, the first two terms of expansion vanish in the sum over the quadrants as a result of symmetry, and ˆ(z) hence, the entire integral 6and, ~ . equation (Equation 6.296) is of the order O b Therefore, end results are important only in a small region near the ends. Far from the ends only the integro-differential operator equation (Equation 3 ~ , has to be kept in Equation 6.281. It can be 6.294), which is of the order O b shown that bk @G(2D) ¼ 4p ak @~ n0
(6:300)
where G(2D) is given in Equation 6.17. Hence, the integral equation (Equation 6.281), describing the scalar potential distribution over the surfaces of the ~ 2 as conductors can be written to within terms of the order O b ~cf(j, z) þ
N ð X i¼1
Li
fðj0 , zÞ
N ð 2 ~2 X @G(2D) 0 b @ fðj0 , zÞ @R(2D) (2D) dj R ln R(2D) dj0 ¼ X(j, z) 0 @~ n 2p i¼1 @~ n0 @z2 Li
(6:301)
215
Implementation of SIBCs for the BEM: Low Frequency Problems
The left-hand side of this equation differs from the left-hand sides of Equations 6.268 through 6.271 for the two-dimensional problem by the presence of the third term. Note that Equation 6.301 does not involve the end coordinates z1 and z2. Since Equation 6.301 is an integro-differential equation, boundary conditions in the variable z must be specified. These conditions can be obtained by setting z ¼ z1 or z ¼ z2 in Equation 6.291 and repeating the subsequent oper~ 3 , we obtain ations. Neglecting terms of the order O b ~cfjz¼zend þ
N ð X i¼1
Li
N ð 2 ~2 X @G(2D) 0 b @ fðj0 , zÞ @R(2D) (2D) fjz¼zend dj R @z2 @~ n0 2p i¼1 n0 z¼zend @~ Li
0
ln R
dj ¼ Xjz¼zend (6:302) where fzend ¼ f j, zend , zend ¼ z1 , z2 . Therefore, for a symmetric system of long parallel conductors of finite length and constant cross section, we have obtained an integro-differential equation (Equation 6.301) containing only contour integrals and differing from the two-dimensional equation by the presence of an additional term, instead of the standard integral equation (Equation 6.281), where the integration is over the surfaces of the conductors. To analyze Equation 6.301 we represent it in the following form using Equation 6.294: (2D)
ð ~cf(j, z) þ Li
~ 2 ð @ 2 fðj0 , zÞ b fðj , zÞE0 ðj, j Þdj E2 ðj, j0 Þdj0 ¼ X(j, z) (6:303) 2p @z2 0
0
0
Li
where E0 ðj, j0 Þ ¼
4 1 X bk (1)kþ1 ; ak 4p k¼1
E2 ðj, j0 Þ ¼
4 1 X (1)kþ1 bk ln ak 8p k¼1
(6:304)
and the summation is, again, over the quadrants. Replacing Equations 6.303 and 6.304 by their discrete forms in j, we obtain Dj
Mj X j¼1
Mj X @ 2 fj (z) ~ 2 Dj E00 ij fj (z) b (E2 )ij ¼ Xi (z), i ¼ 1, . . . , Mj @z2 j¼1
(6:305)
where
E00 ij ¼
ðE0 Þii þ ci =Dj, ðE0 Þij ,
Mj is the number of segments in the contour L
i¼j i 6¼ j
(6:306)
216
Surface Impedance Boundary Conditions: A Comprehensive Approach
4.5
4
3.5 FIGURE 6.10 Dependence of the smallest eigenvalue of a pair of parallel conductors of circular (solid curve) and square (dashed curve) cross section on the number of nodes Mj of the contour (D=h ¼ 1).
3 40
60
80
100
Mξ
The general solution of Equation 6.306 has the form: ~ fi (z) ¼ gi exp Bi z=b
(6:307)
where gi are the functions of the point on the contour 1 Bi are the eigenvalues of the matrix E22 E00 The solution in Equation 6.307 describes the z dependence of the potential distribution resulting from the finite lengths of the conductors, and the smallest eigenvalue, Bmin, determines the degree of this dependence. We note that the eigenvalues depend only on the configuration of the cross sections of the conductors and not on their lengths. The dependence Bmin ¼ Bmin(Dj) is shown in Figure 6.10 for a pair of parallel conductors with circular cross section and a pair with square cross sections. We see that the curves have asymptotes in the limit Dj ! 0. The calculations show that for most systems of conductors the smallest eigenvalue lies between 1 and 10. Therefore, the function (Equation 6.307) is significantly different from zero only at small distances (of the order of a fraction of a diameter) from the ends. The dependence Bmin ¼ Bmin(h) is shown in Figure 6.11 for a pair of conductors of circular cross section, where h is the distance between the conductors. It follows from the graph that as h increases (and therefore as the interaction between the currents in the conductors decreases) the region where end effects are important widens, and Bmin varies by about an order of magnitude between the limiting cases. Hence, the solution of the two-dimensional 1 problem for the smallest eigenvalue of the matrix, E22 E00 , gives an estimate of the error in neglecting the third term involving the small parameter on the left-hand side of Equation 6.301.
217
Implementation of SIBCs for the BEM: Low Frequency Problems
7.5
5
2.5
1
10
100
~
0
h
FIGURE 6.11 Dependence of the smallest eigenvalue of a pair of parallel conductors of circular cross section on the distance between them measured in cross-section radii.
As a conclusion we show the results of numerical implementation of the formulation (Equations 6.305 and 3.306) for a pair of parallel conductors of finite length connected in series by wires as shown in Figure 6.9. The wires are assumed to be connected to a current source which is so far from the conductors that it does not interact with them. The circuit is closed by a crosswire at z ¼ zc. Figures 6.12 and 6.13 show the distributions of the two components of the surface current density along the z-axis at the contour point j ¼ 0, obtained using the quasi-three-dimensional formulation (Equations 6.305 and 3.306), a standard three-dimensional boundary element formulation, and a two-dimensional boundary element formulation using
~
Kz
0.4
0
−0.4
0
zc
4
8
z
FIGURE 6.12 Longitudinal component of the surface current density as a function of z at the contour point j ¼ 0 for a pair of parallel conductors shown in Figure 6.9. Solid curve: standard three-dimensional boundary element formulation; dashed curve: quasi-three-dimensional formulation (Equations 6.305 and 3.306); dashed line: two-dimensional boundary element formulation in the approximation of long parallel conductors.
218
Surface Impedance Boundary Conditions: A Comprehensive Approach
0
~ Kξ
−0.1
−0.2
−0.3
−0.4
0
zc
4
8
z
FIGURE 6.13 Transverse component of the surface current density as a function of z at the contour point j ¼ 0 for a pair of parallel conductors shown in Figure 6.9. Solid curve: standard three-dimensional boundary element formulation; dashed curve: quasi-three-dimensional formulation (Equations 6.305 and 3.306); dashed line: two-dimensional boundary element formulation in the approximation of long parallel conductors.
the approximation of long parallel conductors. The crosswire is assumed to contact the conductors at a point whose contour coordinate is zero. We see that even in the most critical case, the quasi-three-dimensional formulation in Equations 6.305 and 3.306 qualitatively describes the correct behavior of the current and is much more accurate than the two-dimensional approximation.
References 1. C.A. Brebbia, The Boundary Element Method for Engineers, Pentech Press, London, 1978. 2. M.J. Tsuk and J.A. Kong, A hybrid method for the calculation of the resistance and inductance of transmission lines with arbitrary cross sections, IEEE Transactions on Microwave Theory and Techniques, 39(8), August 1991, 1338–1347. 3. R.B. Wu and J.C. Yang, Boundary integral equation formulation of skin effect problems in multiconductor transmission lines, IEEE Transactions on Magnetics, 25(4), July 1989, 3013–3015. 4. L. Di Rienzo, N. Ida, and S. Yuferev, Calculation of energy-related quantities of conductors using surface impedance concept, IEEE Transactions on Magnetics, 44(6), June 2008, 1322–1325.
Implementation of SIBCs for the BEM: Low Frequency Problems
219
5. S. Yuferev and N. Ida, Time domain surface impedance concept for low frequency electromagnetic problems—Part I: Derivation of high order surface impedance boundary conditions in the time domain, IEE Proceedings—Science, Measurement and Technology, 152(4), July 2005, 175–185. 6. S. Barmada, L. Di Rienzo, N. Ida, and S. Yuferev, The time domain surface impedance concept for low frequency electromagnetic problems—Part II: Application to transient skin and proximity effect problems in cylindrical conductors, IEE Proceedings—Science, Measurement and Technology, 152(5), September 2005, 207–216. 7. D. De Zutter, H. Rogier, L. Knockaert, and J. Sercu, Surface current modeling of the skin effect for on-chip interconnections, IEEE Transactions on Advanced Packaging, B, 30(2), May 2007, 342–349. 8. I. Mayergoyz, A new approach to the calculation of three-dimensional skin effect problems, IEEE Transactions on Magnetics, 19(5), September 1983, 2198–2200. 9. A.N. Tikhonov and V.Y. Arsenin, Solution of Ill-Posed Problems, Wiley, New York, 1977. 10. S. Yuferev and V. Yuferev, New quasi-three-dimensional integro-differential formulation of boundary integral equation method for the calculation of the skin and eddy currents in parallel conductors, Journal of Technical Physics, 39(1), 1994, 1–6.
7 Implementation of SIBCs for the Boundary Integral Equation Method: High-Frequency Problems
7.1 Introduction In Chapter 6, we have shown how the SIBC concept can be coupled with boundary integral equations to form an efficient method of solution for low-frequency applications. The concepts developed are now extended to include high-frequency applications. We start, by reintroducing the displacement currents in Maxwell’s equations. This means that the basic derivations in Chapter 3, including the surface impedance functions must be generalized to take into account the displacement currents in Maxwell’s equations. Using the vector Green’s function we first develop the electric field integral equation (EFIE) and the magnetic field integral equation (MFIE) representation of the fields based on the incident electric and magnetic fields. These equations are then coupled with the SIBCs previously developed to implement formulations of various orders of approximations in the time and frequency domains in a manner analogous to that in Chapter 6. The shift from low-frequency to high-frequency carries both advantages and penalties. Of course, in conductors, the skin depths tend to be thinner, resulting in more accurate representation of fields. In that sense, one of the main differences is that at high frequencies, it is also possible to treat lossy dielectrics—that is, the surface impedance may now be applied at the interfaces between lossy dielectrics and dielectrics (such as free space). On the other hand, the introduction of displacement currents in Maxwell’s equations results in negation of one of the most important of the advantages we found in the low-frequency transient representation—that of separation of variables. We saw in Chapter 6 that the variables in the surface impedance function given in Equations 3.82 through 3.84 admitted separation into time and space variables only by neglecting displacement currents and treating each term of Equations 3.82 through 3.84 independently. In the high-frequency case we cannot do that and the time convolution integrals cannot be precomputed and tabulated in advance. Yet, direct computation of the convolution integral in the time-domain integral equations is impractical due to the large computational time and storage requirements. To mitigate this difficulty we also include a discussion of methods for efficient 221
222
Surface Impedance Boundary Conditions: A Comprehensive Approach
implementation of the time convolution integrals in the formulations. More specifically, we show in Appendix 7.A.1 how Prony’s method may be incorporated to approximate the impedance by decaying exponentials and by doing so we arrive at a simple recursive evaluation of the convolution integrals. As in previous chapters we use an example to illustrate the theory developed. In this case, the problem of scattering from a cylinder is used as an example.
7.2 Integral Representations of High-Frequency Electromagnetic Fields Consider a homogeneous body Vc of finite conductivity ðe1 ¼ e, m1 ¼ m, s1 ¼ sÞ surrounded by a nonconductive medium V ðe2 ¼ e0 , m2 ¼ m0 , s2 ¼ 0Þ. The distribution of the electromagnetic field in the dielectric space can be described by the frequency-domain Maxwell equations in the following form (with suppressed time variation exp(jvt)): _ _ ~_ ~ J magn r~ E ¼ jvm0 H
(7:1)
_ _ ~_ ¼ jve0 ~ rH E þ~ J
(7:2)
_ r ~ E ¼ relec =e0
(7:3)
~_ ¼ rmagn =m r H 0
(7:4)
where _ ~ ~_ are the electric and magnetic field phasors E and H _ _ ~ J and~ J magn are the electric and magnetic current densities elec and rmagn are the electric and magnetic charge densities r The laws of conservation of charge can be written in the form: _ r ~ J ¼ jvrelec
(7:5)
_ r ~ J magn ¼ jvrmagn
(7:6)
Let V be a linear, homogeneous, isotropic medium so that e0 and m0 can be _ ~_ can be written as treated as scalars and the vector wave equations for ~ E and H _ _ _ _ E ¼ jvm0~ J r ~ J magn rr ~ E kv2 ~
(7:7)
Implementation of SIBCs for the BIEM: High-Frequency Problems
_ _ magn ~_ k2 H ~_ ¼ jvm ~ rr H þ r ~ J 0J v
223
(7:8)
where kv is the wave number. Further transformations will be performed using the vector Green’s theorem in the form: ð ð ~Q ~ r~ ~ rr~ ~ dv ¼ ~ PrQ P ~ ns ds Q P ~ PrrQ V
S
(7:9) ~ and ~ where Q P are vector functions of position with continuous first and second derivatives within and on S, where S is the boundary of V and the normal unit vector ~ ns is directed out of the dielectric space as shown in Figure 7.1. The boundary S can be decomposed into the surface Sc of the conducting body and the boundary Si located far from the body. ~ in the form: We choose the vector function Q ~ r ~ r 0 expðjkv j~ r ~ r 0 jÞ ~ r ~ r0 ~¼Q ~ ð~ ¼ Q r,~ r 0Þ ¼ Gð~ r,~ r 0Þ r ~ r 0j r ~ r 0j j~ j~ r ~ r 0 j2 j~ Gð~ r,~ r 0Þ ¼ r0 Gð~ r,~ r 0Þ ¼
r ~ r 0 jÞ expðjkv j~ 0 r ~ r j j~
r ~ r 0 j expðjkv j~ r ~ r 0 jÞ ~ r ~ r0 1 þ jkv j~ 0 0 r ~ r j r ~ r j r ~ r 0j j~ j~ j~
(7:10) (7:11) (7:12)
Here ~ r and ~ r 0 are the observation and source point position vectors, respect~ defined in Equation 7.10 meets the ively. It is easy to see that the function Q conditions of differentiability and continuity except in the case~ r ¼~ r 0 because of the singularity in the Green function G. Substitution of the electric field
ξ1 ξ2
B η
d C
A n n'
E inc
H inc
k
FIGURE 7.1 The problem of transient scattering from an infinitely long straight cylinder.
224
Surface Impedance Boundary Conditions: A Comprehensive Approach
for ~ P in Equation 7.9 leads to the following formula known as the Stratton– Chu formula [1]: ð relec _ _ magn ~ ~ jvm0 JG þ J rG rG dv e0 V ðh i _ _ ~_ G ~ ns H jvm0 ~ ns ~ E rG ~ ns ~ E rG ds (7:13) ¼ S
The vector operations indicated in Equation 7.13 are performed in the source coordinates. Treatment of the singularity in G requires some derivations and results in the following equation [2]: ð 1 relec 0 _ _ _ ~ jvm0~ r G dv0 JG þ~ J magn r0 G E¼ e0 4p q V ðh i 1 _ _ ~_ G ~ n0 H jvm0 ~ n0 ~ E r0 G ~ n0 ~ E r0 G ds0 (7:14) 4p q S
where q represents the absolute value of the solid angle subtended by S at ~ r and the ‘‘prime’’ is used to indicate a normal defined at the source point The surface integral in Equation 7.14 is the principal value integral over Si þ Sc. Invoking the duality of Maxwell’s equations, the magnetic field can be written in a similar form: ð rmagn 0 _ _ ~_ ¼ 1 jvm0~ J r0 G r G dv0 J magn G þ~ H e0 4p q V ðh i 1 _ ~_ r0 G þ ~ ~_ r0 G ds0 (7:15) n0 ~ þ jve0 ~ E Gþ ~ n0 H n0 H 4p q S
A situation of great interest occurs when Si recedes to infinity. For sources of bounded extent within V one finds that the radiation condition at infinity requires that the contribution from the integral over Si be due entirely from sources outside Si. We shall refer to this contribution as an incident field and subsequently write Equations 7.14 and 7.15 as ð 4p ~_ inc 1 relec 0 _ _ magn _ 0 ~ ~ ~ jvm0 JG þ J r G r G dv0 E E¼ e0 4p q 4p q V ðh i 1 _ 0 ~_ r0 G þ ~ ~_ r0 G ds0 (7:16) n ~ jve0 ~ E Gþ ~ n0 H n0 H 4p q S
Implementation of SIBCs for the BIEM: High-Frequency Problems ð 4p ~_ inc 1 rmagn 0 _ _ J magn G þ~ jvm0~ J r0 G r G dv0 H þ e0 4p q 4p q V ðh i 1 _ ~_ r0 G þ ~ ~_ r0 G ds0 n0 ~ þ jve0 ~ E Gþ ~ n0 H n0 H 4p q
225
~_ ¼ H
(7:17)
Sc
To render the integral equations solvable with respect to the tangential components of the electric or the magnetic fields on the surface Sc, the observation and source points should be positioned on the surface of the conducting body. Let Sc be smooth and take into account the fact that q ¼ 2p for ~ r on a smooth surface. Then, Equations 7.16 and 7.17 can be written in the forms: ð 1 relec 0 _ _ _ _ ~ Eb ¼ 2 ~ JG þ~ J magn r0 G E inc jvm0~ r G dv0 e0 2p V ðh i 1 _ _ ~_ b G ~ jvm0 ~ Eb r0 G ~ n0 ~ Eb r0 G ds0 n0 ~ n0 H 2p
(7:18)
Sc
ð rmagn 0 _ _ ~_ b ¼ 2 H ~_ inc þ 1 H J magn G þ~ jvm0~ J r0 G r G dv0 e0 2p V ðh i 1 _ ~_ b r0 G þ ~ ~_ b r0 G ds0 þ jve0 ~ Eb G þ ~ n0 H n0 H n0 ~ 2p
(7:19)
Sc
where the vector ~ n is directed into the dielectric space normal to the surface Sc. Frequently, the only sources of the electromagnetic field in practical problems of electromagnetic scattering are incident fields Einc or Hinc, and neither currents nor charges exist in the dielectric domain V. Thus the volume integrals vanish and Equations 7.18 and 7.19 reduce to the following forms: 1 _ _ ~ E inc Eb ¼ 2 ~ 2p
ðh
i _ _ ~_ b G ~ jvm0 ~ E b r0 G ~ n0 ~ Eb r0 G ds0 n0 H n0 ~
(7:20)
Sc
~_ inc þ 1 ~_ b ¼ 2 H H 2p
ðh
i _ ~_ b r0 G þ ~ ~_ b r0 G ds0 jve0 ~ Eb G þ ~ n0 H n0 ~ n0 H
(7:21)
Sc
The time-domain analogs of Equations 7.20 and 7.21 can be obtained using Fourier transform techniques. The transform pair which is employed in this derivation is given by _
1 ð
f (v) ¼
f (t) exp(jvt)dt; 1
_ 1 f (t) ¼ F f (v) ¼ 2p
1 ð
_
f (v) exp( jvt)dv
1
(7:22)
226
Surface Impedance Boundary Conditions: A Comprehensive Approach
Since the interchange ofÐ the order of integration is valid, the operations F (Fourier transform) and Sc ds0 can be interchanged. This leads to 1 ~ Eb (t) ¼ 2~ Einc (t) 2p
9 8 _ ð< _ _
_ _ _ = ~ b (v)G F ~ n0 ~ n0 ~ m~ E b (v) r0 G F ~ E b (v) r0 G ds0 n0 F jvH ; : 0 Sc
(7:23) 9 8 _ ð< _ _
_ _ _ = 1 ~ b (t) ¼ 2H ~ inc (t) þ ~ b (v) r0 G þ F ~ ~ b (v) r0 G ds0 n0 H n0 H H e0~ E b (v)G þ F ~ n0 F jv~ ; 2p : Sc
(7:24)
From Fourier transform theory it is known that if f(t) and d f(t)=dt approach zero as t approaches infinity, then _ d f (t) F jv f (v) ¼ dt
(7:25)
Using Equation 7.20 and the identity F½expðjvt0 Þ ¼ dðt t0 Þ
(7:26)
we obtain r ~ r 0j expðj(v=c)j~ r ~ r 0 jÞ 1 j~ ¼ d t F½Gð~ r,~ r 0, vÞ ¼ F (7:27) r ~ r 0j r ~ r 0j c j~ j~ r ~ r 0j ~ r ~ r0 r ~ r 0j ~ r ~ r0 j~ 1 @ j~ F½r0 Gð~ d t r,~ r 0, vÞ ¼ d t þ 3 c c @t c r ~ r 0j r ~ r 0 j2 j~ j~ (7:28) Substitution of Equations 7.27 and 7.28 into Equations 7.23 and 7.24 yields the time-domain EFIE and MFIE in the following forms: Electric field integral equation: ~b
ð(
R m0 @ 0 ~ b 0 ~b ~ @ 0 ~b ~ ~ ~ n n E H E n R @t0 R3 @t0 Sc ) ~ ~ @ 0 ~b ~ R R 0 ~b R ~ n E 2 ~ n E ds0 (7:29) cR R3 @t0 cR2 0 ~inc
E ¼ 2E
1 2p
t ¼tR=c
227
Implementation of SIBCs for the BIEM: High-Frequency Problems
Magnetic field integral equation: ~b
ð(
~ R R e0 @ 0 ~b 0 ~ b ~ @ 0 ~b ~ ~ n E þ ~ n H 2 n H 3þ 0 R @t R cR @t Sc ) ~ R @ 0 ~b ~ 0 ~b R ~ n H þ ~ n H þ ds0 (7:30) R3 @t0 cR2 0 ~ inc
H ¼ 2H
1 þ 2p
t ¼tR=c
where ~ R ¼~ r ~ r 0. Equations 7.29 and 7.30 can be represented in a more tractable form: 1 ~ Einc Eb ¼ 2~ 2p
) ð( h i ~ h i~ R m0 @ 0 ~ b 0 b 0 ~b R ~ ~ ~ ~ n H P n E P n E R @t0 R R 0
Sc
~ inc þ 1 ~ b ¼ 2H H 2p
ð( Sc
ds0
t ¼tR=c
) h i ~ h i~ e0 @ 0 ~b R 0 b 0 ~b R ~ ~ n H þP ~ n E þP ~ n H R @t0 R R 0
(7:31) ds0
t ¼tR=c
(7:32)
Here the operator P is defined as P½ f ð~ r 0, t0 Þ ¼
f ð~ r 0 , t0 Þ 1 @f ð~ r 0 , t0 Þ þ r,~ r 0Þ @t0 r,~ r 0 ÞÞ2 cRð~ ðRð~
(7:33)
7.3 SIBCs for Lossy Dielectrics At very high frequencies the skin depth may be thin even when conductivity is poor. This means that the surface impedance concept can be applied to lossy dielectrics as well. However, in this case the displacement current cannot be neglected in the governing equations for conducting regions. Therefore, Equations 3.1 and 3.2 must be replaced by the complete Maxwell’s equations in the form: r ~ E ¼ m
~ @H @t
~ ¼ s~ rH Eþe
@~ E @t
(7:34) (7:35)
228
Surface Impedance Boundary Conditions: A Comprehensive Approach
We represent Equations 7.34 and 7.35 in local coordinates ~j1 , ~j2 , h ~ with nondimensional variables as follows: ~j ~j ~j ~h E ~ @H @E p~ dk @ E h 3k k p ¼ (1)k ~ , k ¼ 1, 2 ~ ~k ~ @h ~ @~t ph ~ @~jk d ph ~ dk ~ 2 X
(1)i
i¼1
(7:36)
~ di
~h ~i @H @E ¼ ~ @~t ph ~ @~j3i di ~
! ~ ~j ~j ~j ~h H @H @E @H d3k k ~ 3k 3k k ~ p ¼ (1) Ejk þ ~a , ~ @h ~ @~t ph ~ @~jk ~ ph ~ d3k ~ d3k ~
(7:37)
k ¼ 1, 2 (7:38)
~ p2
2 X
(1)3k
i¼1
~ di
~j ~ @H 3i ~ h þ ~a @ Eh ¼E ~ @~t ~h ~ @ ~ji di p
(7:39)
Here the parameter ~a1 ¼ st=e
(7:40)
can be viewed as a loss tangent. Substituting Equations 2.44 and 2.45 and transferring into the Laplace domain, we modify Equations 7.36 and 7.37 to the following forms: ~j @E k ~ p @h ~
~j @E ~h E k þ ~k @~jk d
!
~j E 1 k ~h ~ þ p ~2 ~ d dk 2
k
2 X i¼1
~ @H j3k ~p @h ~
~ ~ H @H j3k h þ ~d3k @ ~j3k
!
~h @E @~jk
!
~ , k ¼ 1, 2 ¼ (1)k ~p~sH j3k
~j @E ~ ~ 2 h i ~þ~ ¼ ~p~sH p p (1) h ~ @~j3i d3i i
~ ~ H 1 @H j3k h p~ h ~ þ ~d2 ~d3k @~j3k 3k 2
!
(7:41)
(7:42)
~ j , k ¼ 1, 2 p1 ð1 þ ~a~sÞE ¼ (1)k ~ k
(7:43) 2 X i¼1
(1)3i
~j @H h ~ ~ 3i ~þ~ p p2 ¼ ~p1 ð1 þ ~a~sÞE h ~ @~ji di
(7:44)
Representing the electric and magnetic fields in the form of Equations 3.25 and 3.26 in the small parameter ~ p and transforming Equations 7.41 through 7.44 in a manner similar to Equations 3.16 through 3.19, the following relations between the terms of expansions are obtained:
229
Implementation of SIBCs for the BIEM: High-Frequency Problems
PEC-limit: b b ~ ~ ~0 ¼ 0 ~0 ¼ H E j3k
(7:45)
h
First-order (Leontovich’s) approximation: b b ~s ~ ~ ~ 0 , k ¼ 1, 2 ~1 p pffiffiffiffiffiffiffiffiffiffiffiffiffiffi H ¼ (1)k ~ E 2 ~ ~ ~ s þ a s j3k jk b ~ ~ @ H b 0 2 X 1 ~ ji ~ ¼~ p pffiffiffiffiffiffiffiffiffiffiffiffiffiffi H 1 ~s þ ~a~s2 i¼1 @~ji h
(7:46)
(7:47)
Second-order (Mitzner’s) approximation: b b b ~ ~ ! ~s 1 dk d3k ~ ~ ~ ~ k ~ ~ p pffiffiffiffiffiffiffiffiffiffiffiffiffiffi H 1 ~ ¼ (1) ~ p pffiffiffiffiffiffiffiffiffiffiffiffiffiffi H , k ¼ 1, 2 E2 0 2 2 ~~ ~s þ ~a~s ~s þ ~a~s j3k jk jk 2dk d3k b b b ~ ~ ! 2 X 1 @ 1 di d3i ~ ~ ~ ~ ~ ~ H1 ~ p pffiffiffiffiffiffiffiffiffiffiffiffiffiffi p pffiffiffiffiffiffiffiffiffiffiffiffiffiffi H H2 ¼ ~ 0 2 2 ~~ ~ ~s þ ~a~s i¼1 @ ji ~s þ ~a~s h ji ji 2di d3i
(7:48) (7:49)
Third-order (Rytov’s) approximation: 2 b ~ ~3 E j3k
6 b b ~ ~ 6 ~ ~s 1 dk d3k ~ ~ ~ ¼ (1)k ~ ~p pffiffiffiffiffiffiffiffiffiffiffiffiffi H p pffiffiffiffiffiffiffiffiffiffiffiffiffi 6 H 2 1 6 2 ~~ ~ ~ ~ ~s þ ~a~s2 4 s þ a s jk jk 2dk d3k 0 ~ p2
b ~2 dk þ 2~dk ~d3k 3~d23k 1 1 B B ~ ~ B H 0 2 ~s þ ~a~s @ 2 8~d2k ~d23k jk
b ~ ~ @2 H 0 @~j2k
jk
þ
1 2
b 13 ~ ~ @2 H 0 C7 j3k C7 jk C7 @~jk @~j3k A7 5
b ~ ~ @2 H 0 @~j23k
(7:50)
2 6 b b b ~ ~ 2 X 1 @ 6 1 di d3i ~ ~ ~ ~ ~ ~3 ¼~ 6 H p pffiffiffiffiffiffiffiffiffiffiffiffiffi ~p pffiffiffiffiffiffiffiffiffiffiffiffiffi H H 2 1 6 2 2 ~~ ~ ~ ~ ~ ~ ~ ~ @ j s þ a s s þ a s h ji ji 2di d3i i4 i¼1 0 ~ p2
b ~2 di þ 2~di ~d3i 3~d23i 1 1 B B ~ ~ H B 0 ~s þ ~a~s2 @ 2 8~d2i ~d23i ji
b ~ ~0 @ H 2
@~j2i
ji
þ
1 2
b 13 ~ ~0 @ H C7 ji j3i C7 C7 @~ji @~j3i A7 5
b ~ ~0 @ H 2
@~j23i
2
(7:51)
230
Surface Impedance Boundary Conditions: A Comprehensive Approach
Returning to the frequency=time domains, we can represent Equations 7.46 through 7.51 in the forms: h i ~ b ¼ (1)3k @ F ~ H ~b , E j3k jk @~t ~b ¼ H h
k ¼ 1, 2
(7:52)
2 X @ ~h ~ b i F H ji ~ i¼1 @ ji
(7:53)
~ ~f ] generalizes Equations Here, the surface impedance operator-function F[ 3.86 through 3.88 and now covers both high- and low-frequency problems: 8 1 > > p pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ~fjk LD > >~ 2 > ~s þ ~a~s < h i > Leontovich ~ 1 j 1 ~ _ F f jk ¼ p ~ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ~f jk > > 2 1 þ 2j~a > > > > : ~ ~b ~ pfjk * T2 TD
FD,
k ¼ 1, 2
8 1 1 ~ ~dk ~d3k > > ~p pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ~f j ~p2 > f LD > > ~s þ ~a~s2 jk 2~dk ~d3k ~s þ ~a~s2 k > > > > h i < 1 1 ~dk ~ d3k ~_ ~f_ þ p~2 j ~Mitzner ~fj ¼ p~ 1 j pffiffiffiffiffiffiffiffiffiffiffiffiffiffi F fj k jk ~ ~ > 2 2 1 þ 2j~ a 1 þ 2j~a 2dk d3k k > > > > > ~dk ~d3k > > ~ ~b > ~fj * T ~ b TD : ~pfjk * T2 ~p2 3 k 2~dk ~d3k
FD, k ¼ 1, 2
(7:54)
(7:55)
8 ~ ~
1=2
> > ~f ~p2 ~s þ~a~s2 1 ~f dk d3k > ~ p ~s þ ~a~s2 > jk jk ~ ~ > > 2dk d3k > > > 0 1 > > > > @ 2~f j3k
3=2 ~d2k þ 2~dk ~d3k 3~d23k 1 @ 2~f jk 1 @ 2~f jk > 3 2 > ~ @f A LD > ~ p ~s þ~a~s þ > jk > 2 @~j2k 2 @~j23k @~jk @~j3k > 8~d2k ~d23k > > > > > > > ~dk ~d3k _ > 1j j _ > ~f >p ~ ð1 þ 2j~aÞ1=2~f jk þ p~2 ð1 þ 2j~aÞ1 > j > 2 2 > 2~dk ~d3k k > > > > 2 ~2 < h i > 1þj _ d~ þ 2d~k d~3k 3d 3k ~Rytov ~fj ¼ ~ p3 ð1 þ 2j~aÞ3=2 ~f jk k F k ~d2 ~d2 4 > 8 > k 3k > > > 1 > > _ 2_ 2_ > @ 2~f j3k > 1 @ ~f jk 1 @ ~f jk > > A FD, k ¼ 1, 2 þ > > > 2 @~j2k 2 @~j23k @~jk @~j3k > > > > > > > ~ ~ > > > ~ b ~p2 dk d3k ~fj * T ~b > p~fjk * T >~ 2 ~dk ~d3k k 3 > 2 > > > > ! > > ~~ ~2 ~2 > @ 2~fj3k 1 @ 2~fjk 1 @ 2~fjk > 3 ~ dk þ 2dk d3k 3d3k > ~ ~ b TD T p þ (7:56) f > j : k 2 @~j2k 2 @~j23k @~jk @~j3k * 4 8~d2k ~d23k
Implementation of SIBCs for the BIEM: High-Frequency Problems
231
~ ~ m and T ^ m were obtained using the inverse Laplace where the functions T transform and written in the form [3]: ~t
1 1=2 ~ b ~t pffiffiffiffiffiffiffiffiffiffiffiffiffiffi , ~a (7:57) exp ¼ T I0 2 2 2~ a ~s þ ~s ~a ~t ~t ~t
~s 1 ~^ b ~ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi , I0 exp þ U ~t ~a1=2 ¼ T I1 2 t 3=2 2 2~a 2~a 2~a 2~a ~s þ ~s ~a (7:58) ~t
1 ~ b ~t ¼T , 1 exp (7:59) 3 ~a ~s þ ~s2~a ~t ~s 1 ~^ b ~ ¼T exp , (7:60) 3 t 2 ~a ~s þ ~s ~a ~a ~t ~t
1 1=2~ ~ b ~t ~ tI (7:61) exp ¼T , 2 a 1 4 3=2 2~a 2~a ð~s þ ~s2~aÞ ~t ~t ~t ~s ~^ b ~ 3=2~ ~ I exp ¼T t I (7:62) , a 0 1 4 t 3=2 2~a 2~a 2~a ð~s þ ~s2~aÞ where In(~ x) is the modified Bessel function of order n and U(~t) is the ~^ b ~ b and T Heaviside function. The relations between the functions T m are m written in the form: ~b
b ~ ~ (0), ^ b ~t ¼ dTm (t) þ d ~t T T m 3 ~ dt
m ¼ 2, 3, 4
(7:63)
~ m are By setting ã ! 0 in Equations 7.57, 7.59, and 7.61, the functions T reduced to their ‘‘low-frequency form’’ Equations 2.189 through 2.191, respectively. Return to dimensional variables in Equations 7.52 through 7.56 finally yields @ h bi F Hjk , k ¼ 1, 2 @t 2 X @ h bi F H ji Hhb ¼ @ji i¼1
Ebj3k ¼ (1)3k m
F
Leontovich
f jk ¼
(
( jvsm)1=2 (1 þ jve=s)1=2 f_jk (sm)
1=2
f jk *
T2b
TD
FD
(7:64) (7:65)
,
k ¼ 1, 2
(7:66)
232
Surface Impedance Boundary Conditions: A Comprehensive Approach
8 dk d3k _ 1=2 > > (1 þ jve=s)1=2 f_jk ( jvsm)1 (1 þ jve=s)1 fj FD < ( jvsm) 2dk d3k k Mitzner F f jk ¼ k ¼ 1, 2 > d d3k > : (sm)1=2 fjk * T2b (sm)1 k fjk * T3b TD, 2dk d3k
(7:67)
8 > jve 1=2 _ jve 1 dk d3k _ > 1=2 1 > 1 þ ( jvsm) 1 þ ( jvsm) f fj > jk > > 2dk d3k k s s > > > > > > jve 3=2 _ d2k þ 2dk d3k 3d23k 1 @ 2 f_jk > > ( jvsm)3=2 1 þ f jk > > s 2 @j2k > 8d2k d23k > > > ! > < @ 2 f_j3k 1 @ 2 f_jk FRytov fjk ¼ FD, k ¼ 1, 2 þ 2 > 2 @j3k @jk @j3k > > > > > > d d3k > > > (sm)1=2 fjk * T2b (sm)1 k fj T b > > 2dk d3k k * 3 > > ! > > > > d2k þ 2dk d3k 3~d23k 1 @ 2 fjk 1 @ 2 fjk @ 2 fj3k > 3=2 > Tb TD (sm) f þ > j k : 2 @j2k 2 @j23k @jk @j3k * 4 8d2k d23k
(7:68)
where rffiffiffi s st st ¼ I0 exp e 2e 2e st T3b (t) ¼ 1 exp e rffiffiffi s st st T4b (t) ¼ 2t I1 exp e 2e 2e T2b (t)
(7:69) (7:70) (7:71)
7.4 Direct Implementation of SIBCs into the Surface Integral Equations Let us represent the frequency-domain equations (Equations 7.20 and 7.21) and time-domain equations (Equations 7.31 and 7.32) with local coordinates j1, j2, h: Electric field integral equation: 1 E_ bjk ¼ 2E_ inc jk þ 2p
ð
_ ~_ b G ~ eh H jvm0 ~ eh ~ Eb r0 G E_ bh ðr0 GÞjk ds0 jk
jk
Sc
(7:72)
233
Implementation of SIBCs for the BIEM: High-Frequency Problems
1 Ebjk ¼ 2Einc jk þ 2p
8 ð< Sc
9 ! h i ~ h i R = m0 @ R j k ~b P ~ ~ eh ~ Eb P Ebh e H : R @t0 h jk R ; R jk
ds0 t0 ¼tR=c
(7:73)
Magnetic field integral equation: _ b ¼ 2H _ inc 1 H j3k j3k 2p
ð
_ jve0 ~ Eb eh ~
j3k
Gþ
~_ b r0 G ~ eh H
j3k
0 _ b ðr0 GÞ þH h j3k ds
Sc
Hjb3k
8 ! ð< h i ~ R 1 e0 @ b b ~ ~ ~ ~ ¼ 2Hjinc E þ P e H e h h 3k j3k R 2p : R @t0 Sc
j3k
9 h i R = j 3k þ P Hhb R ;
(7:74) ds0 t0 ¼tR=c
(7:75)
where ~ ¼ ~ ~ ~ ~ ¼ ~ ~ ~ n0 H eh H; n0 H eh H
(7:76)
The superscript ‘‘b’’ denoting quantities on the surface V will be omitted from here on. According to the vector identities ~ f ¼ ð1Þk fj3k , eh ~ jk
k ¼ 1, 2
(7:77)
and ~ eh ~ eh ~ f ~ g ¼ ð1Þ3k ~ f jk
j3k
eh ~ gh ~ eh ~ f gj3k ¼ ð1Þ3k ~ f h
j3k
gh ¼ fjk gh
(7:78)
where ~ f and ~ g are any vectors. With the use of Equations 7.76 through 7.78 we represent Equations 7.72 through 7.75 in the following form: Electric field integral equation: 1 E_ jk ¼ 2E_ inc jk þ 2p
ðn Sc
Ejk ¼
2Einc jk
1 þ 2p
ð
o _ j G E_ j ðr0 GÞ E_ h ðr0 GÞ ds0 jvm0 (1)k H h jk 3k k
(7:79)
Rh Rjk m0 @Hj3k P Eh (1) P Ejk ds0 R @t0 R R t0 ¼tR=c k
Sc
(7:80)
234
Surface Impedance Boundary Conditions: A Comprehensive Approach
Magnetic field integral equation: ðn o 3k _ 0 _ j ¼ 2H _ inc 1 _ j ðr0 GÞ þ H _ h ðr0 GÞ H E jve (1) G þ H 0 j j3k h j3k ds 3k k 3k 2p Sc
Hj3k ¼ 2Hjinc 3k
1 2p
ð Sc
(7:81) Rh Rj3k e0 @Ejk þ P Hh (1)3k þ P Hj3k ds0 R @t0 R R t0 ¼tR=c (7:82)
The normal component of the electric field can be derived from Equation 7.2 and expressed in terms of the tangential components of the magnetic field: 2 2 X _j _j @H @H 1 j X 3i 3i ~_ ¼ 1 ~ E_ h ¼ eh r 0 H (1)3i ¼ (1)i @ji @ji jve0 jve0 i¼1 ve0 i¼1
(7:83) and
~ ðt0 Þ ~ ð t0 Þ ðt X ðt X @ H @ H 2 2 1 1 j3i j3i Eh (t) ¼ (1)3i dt0 ¼ (1)i dt0 e0 i¼1 e0 i¼1 @ji @ji 0
0
(7:84) Substitution of Equations 7.83 and 7.84 into Equations 7.79 and 7.80 gives ) ð( 2 _ X j k _ i @ Hj3i 0 0 _ _Ej ¼ 2E_ inc þ 1 ds0 jvm0 (1) Hj3k G Ejk ðr GÞh ðr GÞjk (1) jk k @j 2p ve0 i i¼1 Sc
Ejk ¼ 2Einc jk þ
1 2p
(7:85)
ð( (1)k Sc
Rh 1 Rjk P Ejk R e0 R
m0 @Hj3k : R @t0 ðt X 2 0
i¼1
9 = @ (1)i P Hj3i dt0 ; @ji
ds0
(7:86)
t0 ¼tR=c
The EFIEs in Equation 7.86 or the MFIEs in Equations 7.81 and 7.82 still contain more than one unknown and require an additional equation that can be obtained by taking into account the electromagnetic properties of the conducting domain Vc. Under conditions of the skin effect, the SIBCs in Equations 7.52 and 7.53, giving the tangential components of the electric field and normal component of the magnetic field at the interface, can be
235
Implementation of SIBCs for the BIEM: High-Frequency Problems
substituted in Equations 7.85 and 7.86 and Equations 7.81 and 7.82 as follows: Electric field integral equation: _j (1)3k jvm0 F H ¼ 3k ) ð( 2 _ X 0 m0 1 k _ k i @ Hj3i inc 0 _ _ ds0 2Ejk þ j v(1) Hj3k G þ (1) vF Hj3i ðr GÞh ðr GÞjk (1) ve0 m0 2p @ji i¼1 Sc
(1)3k m0
@ m0 F[Hj3k ] ¼ 2Einc jk þ 2p @t
ð
(
(7:87)
Rh @ (1)k @Hj3k þ (1)k F[P[Hj3k ]] 0 R @t R @t0
9 ðt X = 2 1 Rjk i @ 0 (1) P[Hj3i ]d t ; e0 m0 R @ji i¼1 0
Sc
ds0
(7:88)
t0 ¼tR=c
Magnetic field integral equation: _j ¼ H 3k
_ inc 2H j3k
ð(
1 2p
) 2 X @ 0 0 _ j ðr GÞ þ ðr GÞ _j GþH _ j ds0 v e0 m0 F H F H h j3k 3k 3k i @j i i¼1 2
Sc
(7:89) Hj3k ¼ 2Hjinc 3k
1 2p
ð( Sc
) 2 Rh Rj3k X e0 m0 @ 2 h b i @ F Hj3k þ P[Hj3k ] F[P[Hji ]] þ 0 R @t02 R R i¼1 @ji
ds0
t ¼tR=c
(7:90)
In the limiting case of a perfect electrical conductor, the tangential electric field on the surface Sc is assumed to be zero and Equations 7.87 through 7.90 reduce to the following equations: Electric field integral equation: ) ð( 2 X _j @ H m0 1 3i _j G (1)k H ð r 0 G Þ jk (1)i ds0 ¼ 2E_ inc jv (7:91) jk 3k 2p @j v 2 e0 m 0 i i¼1 Sc
9 8 ð< ðt X = 2 k m (1) @Hj3k 1 Rjk i @ 0 dt (1) P H 0 j3i ; 2p : R @t0 e0 m 0 R @ji i¼1 Sc
0
ds0 ¼ 2Einc jk t0 ¼tR=c
(7:92)
236
Surface Impedance Boundary Conditions: A Comprehensive Approach
Magnetic field integral equation: _j þ 1 H 3k 2p Hj3k þ
1 2p
ð
_ j (r0 G) ds0 ¼ 2H _ inc H h j3k 3k
Sc
ð
P[Hj3k ] Sc
(7:93)
Rh ds0 ¼ 2Hjinc 3k R t0 ¼tR=c
(7:94)
Either of these equations can be used to solve for the tangential magnetic field. The principal difference between them is that the EFIEs in Equations 7.91 and 7.92 are Fredholm integral equations of the first kind whereas the MFIEs in Equations 7.93 and 7.94 are Fredholm integral equations of the second kind. The latter is generally preferable from the computational point of view [4]. However, the final choice is also governed by the geometry of the scatterer. For example, the EFIE is ideally suited for thin cylinders [2]. Frequently, simulations of antenna problems are performed using the equivalent surface current ~ J s defined as ~ b ¼ ~ ~ b J s ¼ Hb , ~ nH eh H Js ¼ ~ jk j3k
k ¼ 1, 2
(7:95)
The EFIEs in Equations 7.91 and 7.92 and the MFIEs in Equations 7.93 and 7.94 for the PEC can be rewritten in terms of an equivalent surface current defined as follows: Electric field integral equation: m jv 0 2p
ð(
(1)k J_jsk G
Sc
) 2 X @ J_jsk 1 i 0 2 (r G)jk (1) ds0 ¼ 2E_ inc jk @j v e0 m0 i i¼1
9 8 ð< ðt X = 2 k @J s h i m (1) 1 Rjk jk i @ s 0 0 (1) P J dt jk ; 2p : R @t0 e0 m0 R @ji i¼1 0
Sc
(7:96)
ds0 ¼ 2Einc jk t0 ¼tR=c
(7:97) Magnetic field integral equation: 1 J_jsk þ 2p Jjsk þ
1 2p
ð Sc
ð
_ inc J_jsk (r0 G)h ds0 ¼ 2H j3k
(7:98)
h iR h P Jjsk ds0 ¼ 2Hjinc 3k R t0 ¼tR=c
(7:99)
Sc
237
Implementation of SIBCs for the BIEM: High-Frequency Problems
Note that in general, the equivalent surface current is not the same as the surface current density ~ K: The latter can be related with ~ J s as follows: 1 ð
K jk ¼
1 ð
Jjk dh ¼ 0
0
1 1 1 ð ð ð @Hh @Hj3k @Hh @Hh ~ dh ¼ ~ dh ¼ Hjb3k þ dh ¼ Jjsk þ dh nH jk @j3k @h @j3k @j3k 0
0
0
(7:100)
The second term vanishes only in the particular case of the PEC-limit. It is clear that the practical implementation of Equations 7.87 through 7.90 is a more complex task than Equations 7.91 through 7.94. In Section 7.5, we will modify Equations 7.87 through 7.90 to a form with the same structure of the left-hand side as in Equations 7.91 through 7.94 incorporating the PEC-condition.
7.5 Implementation Using the Perturbation Technique Both the EFIEs (Equations 7.87 and 7.88) and the MFISs (Equations 7.89 and 7.90) can be brought into the following form by moving the incident fields and the terms containing the surface impedance function to the right-hand side: Electric field integral equation: 1 jvm0 2p
ð(
) 2 _j X @ H 1 i 0 3i _j G (1) H (r G)jk (1) ds0 3k @j v2 e0 m0 i i¼1 k
Sc
0 1 B _ k ¼ 2E_ inc jk þ (1) jvm0 @F Hj3k þ 2p
ð
1 _ j (r0 G) ds0C F H A h 3k
(7:101)
Sc
1 m0 2p
9 8 ðt 2 ð< = (1)k @Hj3k 1 Rjk X i @ 0 (1) P[Hj3i ]dt ; : R @t0 e0 m 0 R @ji i¼1 0
Sc
0 B@ 1 k B ¼ 2Einc jk þ (1) m0 @ F[Hj3k ] þ @t 2p
ð Sc
Rh @ F[P[H ]] j 3k R @t0 0
ds0 t0 ¼tR=c
t ¼tR=c
1 C ds0 C A (7:102)
238
Surface Impedance Boundary Conditions: A Comprehensive Approach
Magnetic field integral equation: ð 1 _ j ðr0 GÞ ds0 ¼ 2H _ _ inc H Hj3k þ j3k h 3k 2p Sc ) ð( 2 X 1 @ 2 0 _ j G þ ðr GÞ _ j ds0 v e0 m0 F H F H j3k 3k i 2p @j i i¼1
(7:103)
Sc
Hj3k þ
1 2p
ð
P[Hj3k ]
Rh 0 ds ¼ 2Hjinc 3k R
Sc
ð(
1 2p
Sc
2 Rj3k X e0 m0 @ 2 @ F[P[Hji ]] 0 2 F[Hj3k ] þ R @t R i¼1 @ji
) 0
ds0
(7:104)
t ¼tR=c
Introducing the scale factors into the operator P, we obtain ! ~f f * h i * ~f @ ~f ~f f * 1 f D @ f * ~ h~i *
¼ 2 P ~f f ¼ ¼ þ þ P f ~ @ ~t0 t ~ 2 D2 cRD ~ 2 ct @~t0 D R D2 R
(7:105)
where h i ~ ~ ~ ~f ¼ f þ ~q @ f P ~2 @~t0 R 2D=(cv) FD ~q ¼ D=(ct) TD
(7:106)
(7:107)
Here the parameter ~q has a physical meaning as the electrical size of the problem. Equations 7.101 through 7.104 can be represented with non-dimensional variables using the scale factors in Table 3.1 in the following forms: Electric field integral equation: j 2p
ð(
) ~_ 2 X @ H j ~ 2 i 0 3i ~ (2~q) (r G) ~ _j G 2(1) H (1) d~s0 jk 3k ~ji @ i¼1 k
Sc
h ð h i i 1 ~_ ~_ ~_ inc þ (1)k F 0~ 0 ~ H ~ ~ F H ¼ 2E d s þ (r G) j3k j3k h jk 2p Sc
(7:108)
239
Implementation of SIBCs for the BIEM: High-Frequency Problems 9 8 t ð< 2 = ~j ~j ð X R (1)k @ H @ i 3k ~ H ~ j d~t0 ~q2 k (1) P 3i ~ ~ ; : R @~t0 R @~ji i¼1
1 2p
~ c ~t0 ¼~tR=~
0
Sc
~ inc þ (1)k ¼ 2E jk
d~s0
ð ~ Rh @ ~h ~ ~ i @ ~ ~ 1 0 F Hj3k þ F P Hj3k d~s ~ @~t0 2p @~t ~ c R ~t0 ¼~tR=~ Sc
(7:109) Magnetic field integral equation: ð
1 ~_ H j3k þ 2p
~_ inc 1 ~_ 0~ s0 ¼ 2H H j3k (r G)h d~ j3k 2p
Sc
ð( Sc
) 2 h i X @ ~h ~_ i ~_ 0~ ~ ~ H G þ (r d~s0 G) 2j~qF F H j3k ji j3k ~ i¼1 @ ji
(7:110) ~j þ 1 H 3k 2p
ð Sc
) ð( 2 2 2 ~j X ~h R ~q @ ~ ~ R 1 @ ~h ~ ~ i 0 inc 0 3k ~ ~ ~ P Hj3k d~s ¼ 2Hj3k F Hj3k þ F P Hji ~0 ~ ~ d~s ~ ~ @~t0 2 ~ ~ 2p R R R t ¼tR=~c i¼1 @ ji Sc
(7:111)
~ m , m ¼ 1, 2, 3, ~ using the operators L For further transformation we represent F as was done in Equations 6.170 through 6.172: h i Leontovich h i ~ 1 ~fj ~ ~fj F ¼ L k k
(7:112)
h i Mitzner h i h i ~ 1 ~fj þ ~ ~ 2 ~fj ~ ~fj F ¼ L p L k k k
(7:113)
h i Rytov h i h i h i ~ 3 ~fj ~ 1 ~fj þ ~ ~ 2 ~fj þ ~p2 L ~ ~fj F ¼ L p L k k k k
(7:114)
~ m in our high-frequency case are written in the form: where the operators L 8 1j 1 _ < h i > pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ~f bjk ~ 1 ~fj ¼ 2 1 þ 2j~ a L k > : ~f b T ~ b TD jk * 2
FD
8 ~ 1 d3k ~_ b dk ~ > >j f jk FD < 2 1 þ 2j~a ~ ~ h i > 2 d d k 3k ~ ~ L 2 f jk ¼ > ~ d3k ~b ~ b d ~ > > : k fj * T3 TD ~ ~ 2dk d3k k
(7:115)
(7:116)
240
Surface Impedance Boundary Conditions: A Comprehensive Approach
h i ~ 3 ~fj ¼ S L k
0 1 8 _ 2 _b 2 _b > ~d2 þ 2~dk ~d3k 3~d2 @ 2~f bj3k > 1þj 1 @ ~f jk 1 @ ~f jk _ > 3=2 k 3k b ~ @ A FD > þ (1 þ 2j~a) f jk > > 4 2 @~j2k 2 @~j23k @~jk @~j3k < 8~d2k ~d23k ! > > ~~ ~2 ~2 > @ 2~fjb3k @ 2~f b @ 2~fjbk > > ~b dk þ 2dk d3k 3d3k 1 jk þ 1 ~ b TD > T : fjk 2 @~j2k 2 @~j23k @~jk @~j3k * 4 8~d2k ~d23k
(7:117)
We now represent the magnetic field (unknown in both the EFIE and the MFIE) as power series in the small parameter ~p: ~j ¼ H k
1 X
~ ~j ~ pm H , k ¼ 1, 2 k
(7:118)
m
m¼0
Substituting Equation 7.118 into Equations 7.108 through 7.111, equating coefficients at equal powers of ~ p and taking into account Equations 7.112 through 7.117, the following frequency- and time-domain equations for the terms of expansions are obtained: Electric field integral equation: 9 8 ~ > > ~_ > > @ H > m ð> = < 2 X j j ~ k ~ 2 i 0 3i ~ ~ _ ~ d~s0 2(1) H m (1) G (2q) (r G)jk ~ > 2p > @ ji > > j3k i¼1 > > Sc : ; ~ ¼ Y_ EFIE m ,
1 2p
8 ð> < Sc
m ¼ 0, 1, 2, 3
(1) ~ > : R
k
(7:119)
9 > ðt X 2 = ~ j3k ~ i @ ~ 2 Rjk 0 ~m d~t ~q (1) P H ~ ~ji > j3i @~t0 R @ ; i¼1
~ ~m @ H
0
d~s0
~ c ~t0 ¼~tR=~
¼ Y~ EFIE m , m ¼ 0, 1, 2, 3
(7:120)
~ and Y~ EFIE are represented in the form: Here the right-hand sides Y_ EFIE m m PEC-limit (m ¼ 0): ~ ~_ inc , ¼ 2E Y_ EFIE 0 jk
k ¼ 1, 2
(7:121)
~ inc , ¼ 2E Y~ EFIE 0 jk
k ¼ 1, 2
(7:122)
Leontovich’s approximation (m ¼ 1): 9 8 " # > > ð " # = < ~ ~ 1 ~ ~_ ~_ k 0~ 0 ~1 H ~ L L H Y_ EFIE ¼ (1) d~ s þ (r G) 0 1 0 h 1 > > 2p ; : j3k j3k Sc
(7:123)
241
Implementation of SIBCs for the BIEM: High-Frequency Problems
Y~ EFIE ¼ (1)k 1
8 >
:@~t
~ ~ ~0 L1 H
j3k
)
ð ~ Rh @ ~ ~ ~ 1 ~0 L1 P H þ 0 ~ ~ j 2p R @t 3k Sc
d~s0
~ c ~t0 ¼~tR=~
(7:124) Mitzner’s approximation (m ¼ 2): 8 " # " # > ð < ~ ~ 1 ~_ EFIE ~ ~_ k ~ ~ _ þ L2 H þ Y2 ¼ (1) L1 H 1 0 > 2p : j3k j3k Sc 9 " #! > = ~ ~_ 0~ 0 ~2 H ~ þL d s (r G) 0 h > ; j3k Y~ EFIE 2
¼ S(1)
k
@ ~1 L
@~t
~ ~1 H
j3k
~2 þL
~ ~0 H
~1 L
"
~ ~_ H 1
#
j3k
(7:125)
j3k
ð ~ Rh @ ~ ~ ~ ~ 0 ~ ~ ~ ~ L1 P () H 1 þ L2 P H 0 0 ~ d~s ~ @~t0 j3k j3k R ~t ¼~tR=~c
1 þ 2p
Sc
(7:126) Rytov’s approximation (m ¼ 3): 8 >
: þ
ð
1 2p
"
~1 L
~ ~_ H 2
#
~2 þL
"
~ ~_ H 1
j3k
"
~ ~_ H 2
~3 þL
"
~ ~_ H 0
# ~2 þL
"
~ ~_ H 1
j3k
#
j3k
j3k
Sc
#
j3k
# ~3 þL
"
~ ~_ H 0
#!
~ d~s0 (r0 G) h j3k
9 > = > ;
(7:127) ( ¼ (1)k Y~ EFIE 3 þ
1 2p
@ ~ ~ ~ ~ ~ ~ ~ ~ ~0 þ L2 H 1 þ L3 H L1 H 2 j3k j3k j3k @~t ð ~ Rh @ ~ ~ ~ ~ ~ ~ H ~ ~1 þ L L P H 1 2 2 P ~ @~t0 j3k j3k R
Sc
~ ~ ~ H ~0 þ L3 P
~0 j3k
~ c t ¼~tR=~
) d~s0
(7:128)
242
Surface Impedance Boundary Conditions: A Comprehensive Approach
Magnetic field integral equation: ~ ~_ H m
ð ~ ~_ H m
1 þ 2p j3k
~ ~m H
1 þ j3k 2p
m ¼ 0, 1, 2, 3
(7:129)
~ Rh 0 d~s ¼ Y~ MFIE , m ¼ 0, 1, 2, 3 m ~ j3k R
(7:130)
j3k
Sc
ð
~_ MFIE ~ d~s0 ¼ Y (r0 G) , h m
~ P
~ ~m H
Sc
~ and Y~ MFIE are represented in the forms: Here the right-hand sides Y_ MFIE m m PEC-limit (m ¼ 0): ~ ~_ inc ¼ 2H Y_ MFIE j3k 0
(7:131)
~ inc ¼ 2H Y~ MFIE 0 j3k
(7:132)
Leontovich’s approximation (m ¼ 1): " ð( ~ ~_ ~1 H 2j~qL 0
1 ~ Y_ MFIE ¼ 1 2p
#
j3k
Sc
" #) 2 X ~ @ ~_ ~ ~ þ (r G) H L G d~s0 0 j3k ~ji 1 @ ji i¼1 0~
(7:133)
~ ) ð( 2 2 2 X ~ R 1 @ @ q j ~ ~ MFIE ~ ~ P ~ H ~0 ~0 L L Y~ 1 ¼ d~s0 þ 3k H ~ji 1 ~ @~t02 1 ~ j3k ji 2p @ R R ~ c i¼1 ~t0 ¼~tR=~ Sc
(7:134)
Mitzner’s approximation (m ¼ 2): 1 ~ Y_ MFIE ¼ 2 2p
ð(
~1 2j~q L
"
~ ~_ H 1
Sc 2 X @ þ (r G)j3k ~ji @ i¼1 0~
1 Y~ MFIE ¼ 2 2p
ð( Sc
~1 L
#
"
~2 þL
"
j3k
~ ~_ H 1
# ji
~2 þL
~ ~_ H 0
"
#!
~ G
j3k
~ ~_ H 0
#!)
d~s0
(7:135)
ji
~q2 @ 2 ~ ~ ~ ~ ~ ~0 L1 H 1 þ L2 H ~ @~t02 j3k j3k R
) 2 ~j X R @ ~ ~ ~1 P ~2 P ~ H ~ H ~1 ~0 L þL d~s0 þ 3k ~ ~ 0 ~ j j R i¼1 @ ji i i ~t ¼~tR=~c (7:136)
Implementation of SIBCs for the BIEM: High-Frequency Problems
243
Rytov’s approximation (m ¼ 3): " # " # " #! ð( ~ ~ ~ ~ ~ ~ ~_ ~ ~ ~ _YMFIE ¼ 1 ~ _ _ þ L2 H 1 þ L3 H 2j~q L1 H 2 G 0 3 2p j3k j3k j3k Sc " # " # " #!) 2
0 X ~ ~ ~ @ ~ ~ ~ ~_ ~ ~ ~ _ _ L1 H 2 þ L2 H 1 þ L3 H d~s0 þ rG j 0 3k ~ ji ji j3k i¼1 @ ji 1 Y~ MFIE ¼ 3 2p
ð( Sc
(7:137) ~q2 @ 2 ~ ~ ~ ~ ~ ~ ~ ~ ~0 L2 H 1 þ L2 H 1 þ L3 H 02 ~ ~ j j j3k @ t R 3k 3k
2 ~j X R @ ~ ~ ~ ~ 3k ~2 P ~ H ~1 ~1 L2 P H þL ~ ~ji ji ji R @ i¼1 ~ ~3 P ~ H ~0 d~s0 þL þ
ji
~ c ~t0 ¼~tR=~
(7:138)
It is well worth emphasizing here the following basic advantages of both the EFIE and MFIE formulations: 1. The integral equations for the terms of expansions differ only in the form of the right-hand side and can be solved by the same solution procedures. Therefore, new computational complications do not arise in comparison with solving the problem using the wellknown perfect conductor limit (m ¼ 0). 2. The convolution integrals in the integral equations are on the righthand sides only and can be calculated and tabulated before numerically solving the appropriate integral equation so that the computer resources required for the computation are greatly reduced as compared with the formulation developed in [5]. Representing the expansions in Equation 7.118 in dimensional form: H¼T
3 X
~k ¼ ~ pk H
3 X
Hk
(7:139)
~k ~k ¼ ~ pk H*H pk ID1 H Hk ¼ ~
(7:140)
~k ¼ ~ H pk (I)1 DHk ¼ dk (I)1 Dkþ1 Hk
(7:141)
k¼0
k¼0
where
we get
244
Surface Impedance Boundary Conditions: A Comprehensive Approach
Substituting Equation 7.141 and other scale factors into Equations 7.119 through 7.138, we finally obtain the EFIE and MFIE formulations with dimensional variables: Electric field integral equation: 8 9 ~_ > ð> @ H = < 2 m X 1 1 j3i k ~ i 0 _ ds0 ¼ Y_ EFIE jvm0 (1) H m G 2 (r G)jk (1) m > j3k 2p > v e0 m 0 @ji ; : i¼1 Sc
(7:142)
8 9 ~m > ðt X ð> 2
j3i @t0 2p > e0 m0 R @ji : R ; i¼1 0 Sc 0
ds0 ¼ YEFIE m
t ¼tR=c
(7:143)
Here the right-hand sides Y_ EFIE and YEFIE are represented in the forms: m m PEC-limit (m ¼ 0): ¼ 2E_ inc Y_ EFIE 0 jk ,
k ¼ 1, 2
(7:144)
¼ 2Einc YEFIE 0 jk ,
k ¼ 1, 2
(7:145)
Leontovich’s approximation (m ¼ 1): ( ) ð 1 ~ ~ k EFIE 0 0 _0 _0 þ (r G)h ds L1 H Y_ 1 ¼ (1) L1 H j3k j3k 2p
(7:146)
Sc
( YEFIE 1
¼ (1)
k
ð Rh @ @ 1 ~ ~0 L 1 H0 þ L1 P H 0 j3k j3k @t 2p R @t0
) ds
0
t ¼tR=c
Sc
(7:147) Mitzner’s approximation (m ¼ 2): 8 >
:
~_ þ L2 H 0
~_ H 1
j3k
j3k
þ L2
~_ H 0
(r0 G)h ds0
j3k
9 > = > ;
þ
1 2p
ð
L1
~_ H 1
j3k
Sc
(7:148)
245
Implementation of SIBCs for the BIEM: High-Frequency Problems (
YEFIE 2
@ ~1 ~0 þ L2 H L1 H j3k j3k @t ð Rh @ 1 ~1 ~0 þ L2 P H L1 P H þ 0 j3k j3k R @t0 2p k
¼ (1)
) ds0
t ¼tR=c
Sc
(7:149) Rytov’s approximation (m ¼ 3): ( ~_ _YEFIE ¼ (1)k L1 H 2 3
j3k
ð
1 þ 2p
L1
~_ H 2
þ L2
j3k
~_ H 1
j3k
~_ þ L2 H 1
þ L3
j3k
þ L3
~_ H 0
j3k
~_ H 0
j3k
) (r0 G)h ds0
Sc
YEFIE 3
8 < @ k ~ ~ ~0 þ L 2 H1 þ L3 H L1 H2 ¼ (1) : @t j3k j3k j3k ð
1 þ 2p
Sc
(7:150)
Rh @ ~ ~1 þ L2 P H þ L1 P H2 j3k j3k R @t0
~0 þ L3 P H
0 j3k
ds0
t ¼tR=c
9 = (7:151)
;
Magnetic field integral equation:
~_ H m
j3k
þ
1 2p
ð ~_ H m Sc
~m H
j3k
þ
1 2p
ð
j3k
~m P H
(r0 G)h ds0 ¼ Y_ MFIE , m
j3k
Rh 0 d~s ¼ YMFIE , m R
m ¼ 0, 1, 2, 3
(7:152)
m ¼ 0, 1, 2, 3
(7:153)
Sc
and YMFIE are represented in the forms: The right-hand sides Y_ MFIE m m PEC-limit (m ¼ 0): _ inc ¼ 2H Y_ MFIE 0 j3k
(7:154)
¼ 2Hjinc YMFIE 0 3k
(7:155)
246
Surface Impedance Boundary Conditions: A Comprehensive Approach
Leontovich’s approximation (m ¼ 1): 1 ¼ Y_ MFIE 1 2p
ð( v e0 m0 L1 2
~_ H 0
j3k
Sc
) 2 X @ ~_ G þ ðr GÞj3k ds0 L1 H 0 ji @j i i¼1 0
(7:156)
YMFIE 1
) ð( 2 X R e0 m0 @ 2 @ j ~0 ~0 þ 3k L1 H L1 P H 0 j3k ji R @t02 R i¼1 @ji
1 ¼ 2p
ds0
t ¼tR=c
Sc
(7:157)
Mitzner’s approximation (m ¼ 2):
Y_ MFIE 2
1 ¼ 2p
ð(
v e0 m0 L1 2
~_ H 1
j3k
~_ þ L2 H 0
j3k
G
Sc
) 2 X @ ~ ~_ _1 þ (r0 G)j3k þ L2 H ds0 L1 H 0 j j3k @j 3k i i¼1
YMFIE 2
1 ¼ 2p
ð( Sc
(7:158)
e0 m0 @ 2 ~ ~0 þ L2 H L1 H1 j3k j3k R @t0 2
) 2 Rj3k X @ ~ ~0 þ þ L2 P H L1 P H1 0 ji ji R i¼1 @ji
ds0
t ¼tR=c
(7:159) Rytov’s approximation (m ¼ 3): ð( ~_ v2 e0 m0 L1 H 2
1 ¼ Y_ MFIE 3 2p
j3k
þ L2
~_ H 1
j3k
þ L3
~_ H 0
j3k
G
Sc
) 2 X @ ~ ~ ~_ _ _ ds0 þ (r G)j3k þ L2 H 1 þ L3 H L1 H 2 0 j3k j3k j3k @ji i¼1 0
(7:160)
247
Implementation of SIBCs for the BIEM: High-Frequency Problems
YMFIE 3
1 ¼ 2p
ð( Sc
e0 m0 @ 2 ~ ~ ~0 þ L þ L H H H L 1 2 2 1 3 j3k j3k j3k R @t0 2
2 Rj3k X @ ~2 ~1 þ L2 P H L1 P H ji ji R i¼1 @ji ) ~0 ds0 þ L3 P H 0 ji þ
(7:161)
t ¼tR=c
where
(
L 1 f jk ¼
( jvsm)1=2 ð1 þ jve=sÞ1=2 f_jk 1=2
(sm)
f jk *
T2b
TD
8 1 dk d3k _ 1 > > < ( jvsm) ð1 þ jve=sÞ 2dk d3k fjk L2 fjk ¼ > d d3k > : (sm)1 k fj T b TD 2dk d3k k * 3
L3 fjk ¼
FD
(7:162)
FD (7:163)
8 > jve 3=2 _ d2k þ 2dk d3k 3d23k 1 @ 2 f_jk 3=2 > > 1þ fjk ( jvsm) > > s 2 @j2k 8d2k d23k > > > > > ! > > > > @ 2 f_j3k 1 @ 2 f_jk > > FD þ > > @jk @j3k 2 @j2 < 3k
> > d2k þ 2dk d3k 3d23k 1 @ 2 fjk > 3=2 > > f (sm) j > k > 2 @j2k 8d2k d23k > > > > > ! > > > @ 2 fj3k 1 @ 2 f jk > > > T b TD þ : 2 @j23k @jk @j3k * 4 (7:164)
The principal difference between the practical implementations of timedomain formulations for low- and high-frequency problems is that in the latter case, separation of variables into space and time components is impossible. Thus the time convolution integrals cannot be precomputed and tabulated in advance. Direct computation of the convolution integral in the time-domain integral equations is impractical due to the large computational time and storage requirements [6]. Techniques for efficient implementation of the time convolution integrals in Equations 7.144 through 7.164 are considered in Appendix 7.A.1.
248
Surface Impedance Boundary Conditions: A Comprehensive Approach
7.6 Numerical Example To illustrate the foregoing theory, we consider the problem of transient scattering from an infinitely long straight cylinder of circular cross section illuminated by an incident Gaussian pulse. In this case, the coordinate j1 is directed along the cylinder as shown in Figure 7.1 so that d1 ¼ 1. Since the radius d of the cylinder cross section is constant, d2 ¼ D and ~d2 ¼ 1. In the discussion of the numerical solution of the integral equations obtained above, we restrict ourselves by considering the transverse electric (TE) case. J s is nonUnder these conditions, only the circumferential j2-component of ~ zero and the single-component vector can be treated as a scalar. Representing the equivalent surface current as an asymptotic expansion in the small parameter ~ p: ~J s ¼
1 X
~~Jms p
(7:165)
m¼0
we write the following time-domain MFIEs for the terms of expansion using Equations 7.130, 7.132, 7.134, 7.136, and 7.138: ð ~ h s R ~ ~ ~J s þ 1 ¼ Y~ MFIE , m ¼ 0, 1, 2, 3 (7:166) P Jm m m ~ 2p R ~ ~t0 ¼~t~qR Sc
Y~ MFIE 1
~ inc Y~ MFIE ¼ 2H (7:167) j2 0 ð ð 2 2 ~ ~q @ ~s ~ b 1 1 @ 0 ~ ~J s * T ~ b Rj2 ~ J ¼ d s d~s0 P T * 0 2 ~ ~ @~t0 2 0 2 ~j2 2p R 2p R @ ~ ~ ~t0 ¼~t~qR ~t0 ¼~t~qR Sc
1 ¼ Y~ MFIE 2 2p
1 2p
Sc
~q2 @ 2 ~s ~ b 1 ~s ~ b J T þ J0 * T3 d~s0 ~ @~t0 2 1 * 2 2~ R d2 ~ ~t0 ¼~t~qR Sc ~ ð s b s R @ 1 j2 b ~ ~J * T ~ ~J * T ~ þ ~ P P 1 2 0 3 ~ R @~j2 2~ d2 0
(7:168)
ð
d~s0
(7:169)
~ ~t ¼~t~qR
Sc
) ð 2 2 ( 2~s ~ J @ q 1 @ 1 3 1 0 ~b ~J s * T ~J s * T ~J s * T ~b þ ~b þ ~b þ Y~ MFIE ¼ d~s0 T * 02 3 2 2 1 3 0 4 4 ~ @~t 0 2p R 2 @~j22 2~ d2 8~ d22 ~ ~t ¼~t~qR Sc ð( 1 @ ~ ~J s * T ~ ~J s * T ~ ~J s * T ~b þ 1 P ~b þ 3 P ~b P 2 2 1 3 0 4 2p @~j2 2~ d2 8~d22 Sc ) ~ j 1 @ 2 ~ ~s ~ b R 2 þ d~s0 (7:170) P J T * 0 4 ~ 2 @~j22 R ~0 ~ ~ t ¼t~qR
Implementation of SIBCs for the BIEM: High-Frequency Problems
249
From Equation 7.63 and Duhamel’s theorem it follows that b b d2 Tm @2 b (t) dTm (t) @f ð~ r,tÞ b Tm (t) * f ð~ r,tÞ ¼ f ð~ r,tÞ þ Tm (0) f ð~ r, tÞ þ * 2 2 @t dt dt t¼0 @t
(7:171)
Using Equations 7.57 through 7.62, we obtain ~b ~t ~t ~t ~a þ ~t d2 T 1 1 2 I exp I ¼ 0 1 2~a 2~a 2~a 2~a3=2 ~a ~a~t d~t2 ~b ~t d2 T 1 3 ¼ exp 2 ~a ~a d~t2 ~ b ~a ~t ~t ~t d2 T 4 ¼ 5=2 I0 exp ~a 2~a 2~a d~t2
(7:172) (7:173) (7:174)
Equations 7.166 through 7.170 were solved using the marching-on-time inc inc D was used for I where Jmax technique described in [2]. The scale factor Jmax inc is the maximum of J . The width of the incident pulse was taken to equal the radius of the cylinder, i.e., ~q ¼ 1. The time ~t ¼ 0 corresponds to the time that the pulse reaches the body. From the calculations it follows that two pulses, one on each side of the cylinder, travel while attenuating in amplitude, from the illuminated region to the shadow region of the scatterer [6]. Figure 7.2 shows the distribution
~s J0
1
0.5
0
–0.5
0
1
2
3
ξ2/D FIGURE 7.2 Distribution of the zero-order term ~J0s of the expansion of the surface electric current over half the cross section of the cylinder.
250
Surface Impedance Boundary Conditions: A Comprehensive Approach
of the function ~J0s ¼ ~J0s ~j2 obtained at time ~t ¼ 1. The contour coordinate ~j2 has its origin at point A and proceeds around the contour ending at point C (Figure 7.1). The function ~J0s is the solution of the problem for a perfect conductor and cannot take into account the redistribution of the current induced by the field diffusion into the body. The effect of this process is to smooth out nonuniformities in the surface current distribution along the surface of the body. The surface current acts in the directions shown by arrows in Figure 7.2. In our model, this redistribution is described by the next terms of the expansions of the surface electric current. The distribution of the first-order term ~J1s is shown in Figure 7.3 together with the function ~J0s . From Figure 7.3 it follows that the surface electric current will increase in the regions where the function ~J1s is positive and decrease where the function ~J1s is negative. The second-order term ~J2s gives a further correction by taking into account the radius of curvature of the surface of the body. Figure 7.4 demonstrates the distributions of the terms ~J0s , ~J1s , and ~J2s . It is seen from Figure 7.4 that each of these terms tends to flatten the peak in the distribution of the previous term. The coefficients ~J1s and ~J2s demonstrate the redistribution of the field in the direction normal to the body’s surface whereas the third-order coefficient ~J3s reflects the redistribution of the field along the surface of the body. The function ~J3s is shown in Figure 7.5. Figures 7.6 through 7.8 show the time-distributions of the terms ~J0s , ~J1s , ~J2s , and ~J3s at points A, B, and C, respectively. The final time-distributions of the surface electric current obtained in the PEC-approximation ( ~J s ¼ ~J0s ), in the p~J1s ), and in the Rytov approximation Leontovich approximation ( ~J s ¼ ~J0s þ ~ p~J1s þ ~ p2~J2s þ ~ p3~J3s ) at points A, B, and C are given in Figures 7.9 ( ~J s ¼ ~J0s þ ~ through 7.11, respectively. The parameter ~ p equals 1=2. From these figures
~s J0
~s ~s J0, J1
1
0.5
0
~s J1
–0.5 0
1
2
3
ξ2/D FIGURE 7.3 Distributions of the first-order term ~J1s (solid line) and zero-order term ~J0s (dashed line) of the expansion of the surface electric current over half the cross section of the cylinder.
Implementation of SIBCs for the BIEM: High-Frequency Problems
~s J0
1
~s ~s ~s J 0, J 1, J 2
251
0.5
~s J2
0
~s J1
–0.5 0
1
2
3
ξ2/D FIGURE 7.4 Distributions of the terms J2s (solid line), J1s (dotted line), and J0s (dashed line) of the expansion of the surface electric current over half the cross section of the cylinder.
0.4
~s J3
0.2
0
–0.2
–0.4
0
1
2
3
ξ2/D FIGURE 7.5 Distribution of the third-order term J3s of the expansion of the surface electric current over half the cross section of the cylinder.
it follows that the proposed time-domain SIBC of the high order of approximation allows the accuracy of the calculations to be about 10%–20% higher compared with solving the problem using standard time-domain SIBC of the order of the Leontovich approximation.
252
Surface Impedance Boundary Conditions: A Comprehensive Approach
1.5
~s J0 ~s J1 ~s J2 ~s J3
~ ~ ~ ~
J s0, J s1, J s2, J s3
1
0.5
0
−0.5 0
1
~ t
2
3
FIGURE 7.6 Time-distributions of the coefficients of expansion of the surface electric current at point A at the contour of the cylinder cross section.
~s J0 ~s J1 ~s J2 ~s J3
~ ~ ~ ~
J s0, J s1, J s2, J s3
1
0.5
0
–0.5 0
1
2 ~ t
3
FIGURE 7.7 Time-distributions of the coefficients of expansion of the surface electric current at point B at the contour of the cylinder cross section.
253
Implementation of SIBCs for the BIEM: High-Frequency Problems
1 ~s J0 ~s J1 ~s J2 ~s J3
~ ~ ~ ~
J s0, J s1, J s2, J s3
0.5
0
–0.5
0
1
2 ~ t
3
FIGURE 7.8 Time-distributions of the coefficients of expansion of the surface electric current at point C on the contour of the cylinder cross section.
1.5
PEC Leontovich Rytov
~s J
1
0.5
0
–0.5
0
1
2 ~ t
3
FIGURE 7.9 Time-distributions of the surface electric current, calculated in the perfect conductor limit, in the Leontovich approximation and Rytov’s approximation, at point A on the contour of the cylinder cross section.
254
Surface Impedance Boundary Conditions: A Comprehensive Approach
PEC Leontovich
~s J
1
Rytov
0.5
0
0
1
2 ~ t
3
FIGURE 7.10 Time-distributions of the surface electric current, calculated in the perfect conductor limit, in the Leontovich approximation and Rytov’s approximation, at point B on the contour of the cylinder cross section.
1
PEC Leontovich Rytov
~ Js
0.5
0 0
1
~ t
2
3
FIGURE 7.11 Time-distributions of the surface electric current, calculated in the perfect conductor limit, in the Leontovich approximation and Rytov’s approximation, at point C on the contour of the cylinder cross section.
255
Implementation of SIBCs for the BIEM: High-Frequency Problems
Appendix 7.A.1: Efficient Evaluation of Time Convolution Integrals Consider again the problem of two-dimensional transient scattering from an infinitely long straight cylinder of circular cross section shown in Figure 7.1. In this case the time-domain SIBC (Equation 7.52) relating the tangential electric and magnetic fields can be written in the form (details are given in Section 7.6): ^ b * Hj T ^ b * H j2 T ^b * E j1 ¼ T 2 3 4 2 d2
H j2 1 @ 2 H j2 þ 8d22 2 @j22
! (7:A:1)
where m1=2 s st st st b ^ I1 I0 exp d(t) þ T2 (t) ¼ e 2e 2e 2e 2e st b 1 ^ T3 (t) ¼ e exp e pffiffiffiffiffiffi 1 st st st b ^ T4 (t) ¼ t(e em) I0 I1 exp 2e 2e 2e
(7:A:2)
(7:A:3)
(7:A:4)
Let us set E ¼ Ej1 and H ¼ Hj2 to simplify further derivations. Consider the implementation of the second term on the right-hand side. ðt ðt s b 0 ^b 0 0 1 ^ T3 * H ¼ H(t )T3 (t t )dt ¼ e H(t0 ) exp (t t0 ) dt0 e 0
(7:A:5)
0
Assuming H(t) is piecewise linear in time, Equation 7.A.5 in the discrete time domain with a time step Dt can be written as
^ b * Hj ¼ e1 T 3 nDt
n X l¼0
lDt ð
(l1)Dt
s exp (nDt t0 ) e
H(lDt) H((l 1)Dt) 0 dt (7:A:6) H((l 1)Dt) þ (t (l 1)Dt) Dt 0
256
Surface Impedance Boundary Conditions: A Comprehensive Approach
After integrating over t0
^ b * Hj ¼ s1 T nDt 3
n X l¼0
lDt ð
(l1)Dt
þ H((l 1)Dt)
s n h s i e exp (nDt t0 ) H(lDt) 1 þ exp Dt 1 e sDt e h e s e io exp Dt 1 þ sDt e sDt
(7:A:7)
Now, a recursive form can be obtained by taking the nth term out of the summation and rewriting the remaining summation in terms of ^ b * Hj T (n1)Dt 3 s n h s i e 1 ^ b * Hj ¼ T ^ b * Hj T Dt þ s exp Dt 1 exp H(nDt) 1 þ nDt (n1)Dt 3 3 e sDt e h e s e io exp Dt 1 þ (7:A:8) þ H((n 1)Dt) sDt e sDt
Equation 7.A.8 can be represented in another form: ^ b * Hj ¼ P1 H(nDt) þ P2 H((n 1)Dt) þ P3 T ^ b * Hj T nDt (n1)Dt 3 3
(7:A:9)
where h s i e exp Dt 1 P1 ¼ s1 1 þ sDt e s h e e i P2 ¼ s1 exp Dt 1 þ sDt e sDt s P3 ¼ exp Dt e
(7:A:10)
Therefore, if the surface impedance function is represented by a decaying exponential, the convolution summation can be recursively updated, avoiding the need for a complete time history of the field components. This ^b ^ b * Hj provides an approach to implement the terms T (n1)Dt and T4 * Hj(n1)Dt , 2 b b ^ ^ namely approximate the functions T2 and T4 by a series of exponentials which in turns allows efficient evaluation of the convolution integral using recursion: Q X (m)
^ T(t)
(m) c(m) exp v t , i i
m ¼ 2, 4
(7:A:11)
i¼1
^b . where Q(m) is the number of terms in the approximation of the function T m One of the most accurate methods for obtaining an exponential approximation
Implementation of SIBCs for the BIEM: High-Frequency Problems
257
to an exact function or to a data set is Prony’s method [7]. It has been used to extrapolate transmission line matrix (TLM) and finite-difference time-domain (FDTD) computed waveforms for microwave circuits by a number of investigators [8–10]. Implementation of time-domain SIBCs using Prony’s method was also considered in [11–13]. In describing Prony’s method we follow [12]. The method fits an exponential approximation of the form: f (N)
Q X
ci (mi )N
(7:A:12)
i¼1
to a function f(x) by sampling at Nmax equally spaced points, N ¼ 0, 1, 2, . . . , Nmax 1. This is a two-step procedure. The first step is to evaluate the values of m. The set of equations: fQþ1 þ fQ a1 þ fQ1 a2 þ þ f1 aQ ¼ 0 : : : : : : : :
: :
: :
: :
: :
: :
(7:A:13)
fNmax 1 þ fNmax 2 a1 þ fNmax 3 a2 þ þ fQ aQ ¼ 0 is solved using a least-square error algorithm for the coefficients a. The values of m are then found as the roots of the polynomial: mQ þ a1 mQ1 þ a2 mQ2 þ þ aQ1 m þ aQ ¼ 0
(7:A:14)
The second step is to find the coefficients c. The set of equations:
2 c1 (m1 )2 þ c2 (m2 )2 þ þ cQ mQ ¼ f2
(7:A:15)
c1 (m1 )Nmax 1 þ c2 (m2 )Nmax 1 þ þ cQ (mQ )Nmax 1 ¼ fNmax 1 is solved for the coefficients c using a least-square algorithm. ^ b defined in Equations 7.A.2 and 7.A.4, respectively, ^ b and T The functions T 2 4 depend on the medium properties and must be reapproximated every time the parameters of the medium are changed. We normalize these functions by introducing the variable: q¼
s t 2e
(7:A:16)
258
Surface Impedance Boundary Conditions: A Comprehensive Approach
So that ^ norm (q) ¼ (I1 (q) I0 (q)) exp (q) T 2
(7:A:17)
^ norm (q) ¼ q(I0 (q) I1 (q)) exp (q) T 4
(7:A:18)
^ norm are related with T ^ b and T ^ b , respectively, as ^ norm and T The functions T 2 4 2 4 follows: 1=2 s ^ norm 2e ^ b (t) ¼ m T (7:A:19) d(t) þ q T 2 e 2e 2 s ffi 1 ^ norm 2e ^ b (t) ¼ 2(spffiffiffiffiffi (7:A:20) T T em ) q 4 4 s Since the normalized functions in Equations 7.A.17 and 7.A.18 are independent of the properties of the medium, the rational approximation has to be performed only once. Application of Prony’s method yields the coefficients ^ norm and 20 terms to approximate c(m) and v(m) i i : 6 terms to approximate T2 ^ norm [14]. Tables 7.A.1 and 7.A.2 show the coefficients c(m) and v(m) for T ^ norm , T m 4 i i norm ^ norm ^ and T m ¼ 2, 4. The analytical and exponential approximations of T 2 4 are shown in Figures 7.A.1 and 7.A.2, respectively. Finally, evaluation of the integrals: Q X (m)
^ norm * H ¼ T m
ð Q X (m)
v(m) t i c(m) * H(t) ¼ i e
i¼1
i¼1
t
(m) 0 c(m) exp v (t t ) H(t0 )dt0 , m ¼ 2, 4 i i
0
(7:A:21) is done by the following recursion: ^ norm * Hj ¼ T nDt m
Q(m) X (m) (m) ^ norm P(m) * H 1i H(nDt) þ P2i H((n 1)Dt) þ P1i Tm i¼1
(n1)Dt
, m ¼ 2, 4
(7:A:22) TABLE 7.A.1 ^ norm Exponential Approximation Coefficients T 2 I
c(2) i
v(2) i
1
2.46e-8
8.14e-6
2
5.14e-6
3.71e-4
3
3.04e-4
0.0648
4
0.088
0.0686
5
0.139
0.0481
6
0.849
1.66
259
Implementation of SIBCs for the BIEM: High-Frequency Problems
TABLE 7.A.2 ^ norm Exponential Approximation Coefficients T 4 c(4) i
I
v(4) i
1
1.01
0.60
2
0.41
0.81
3
6.35
1.01
4 5
4.68 9.89
1.20 1.39
6
7.77
1.58
7
2.19
1.80
8
14.50
2.01
9
9.98
2.21
10
13.75
2.39
11
2.12
2.60
12 13
4.34 21.7
2.79 3.01
14
28.36
3.21
15
24.68
3.38
16
29.03
3.61
17
8.81
3.82
18
1.51
4.01
19
10.88
4.20
20
24.31
4.41
Tˆ 2norm
–1
–0.5
0
0
2.5
5
7.5
10
ϑ= σ t 2ε FIGURE 7.A.1 ^ norm function and its exponential approximation. Solid and dashed lines denote analytical The T 2 solution and exponential approximation, respectively.
260
Surface Impedance Boundary Conditions: A Comprehensive Approach
0.2
Tˆ 4norm
0.1
0
–0.1 0
2.5
5 ϑ= σ t 2ε
7.5
10
FIGURE 7.A.2 ^ norm function and its exponential approximation. Solid and dashed lines denote analytical The T 4 solution and exponential approximation, respectively.
where P(m) 1i P(m) 2i
¼ ¼
c(m) i v(m) i c(m) i v(m) i
" 1þ "
1 v(m) i Dt
1 v(m) i Dt
Dt 1 exp v(m) i
exp
(m) P(m) 3i ¼ exp vi Dt
v(m) i Dt
1þ
#
1
!# (7:A:23)
v(m) i Dt
Note that a different approach was considered in [15], namely: the normal^ norm is approximated in the Laplace domain using a series of ized function T 2 first-order rational functions. Then, the inverse Laplace transform yields exponentials. Both approaches lead to similar recursions.
References 1. J.D. Stratton, Electromagnetic Theory, McGraw-Hill, New York, 1941. 2. A.J. Poggio and E.K. Miller, Integral equation solutions of three-dimensional problems, in R. Mittra (Ed.), Computer Techniques for Electromagnetics, Pergamon Press, Oxford, U.K., 1973.
Implementation of SIBCs for the BIEM: High-Frequency Problems
261
3. M. Abramowitz and I.A. Stegun, Handbook of Mathematical Functions, Dover Publications, New York, 1972. 4. A.N. Tikhonov and V.Y. Arsenin, Solution of Ill-Posed Problems, Wiley, New York, 1977. 5. F.M. Tesche, On the inclusion of loss in time-domain solutions of electromagnetic interaction problems, IEEE Transactions on Electromagnetic Compatibility, 32(1), February 1990, 1–4. 6. C. Bennett, Jr. and W. Weeks, Transient scattering from conducting cylinders, IEEE Transactions on Antennas and Propagation, 18(5), September 1970, 627–633. 7. F.B. Hildebrand, Introduction to Numerical Analysis, Dover, New York, 1974, pp. 457–462. 8. J.L. Dubard, D. Pompei, J. Le Roux, and A. Papiernik, Characterization of microstrip antennas using the TLM simulation associated with a Prony-Pisrenko method, International Journal of Numerical Modeling, 3, 1990, 269–285. 9. W.L. Ko and R. Mitra, A combination of FDTD and Prony’s methods for analyzing microwave integrated circuits, IEEE Transactions Microwave Theory and Techniques, 39, 1991, 2176–2181. 10. J.A. Pereda, L.A. Vielva, and A. Prieto, Computation of resonant frequencies and quality factors of open dielectric resonators by a combination of the finitedifference time-domain (FDTD) and Prony’s methods, IEEE Microwave and Guided Wave Letters, 2, 1992, 431–433. 11. J.H. Beggs, R.J. Luebbers, K.S. Yee, and K.S. Kuz, Finite-difference time-domain implementation of surface impedance boundary conditions, IEEE Transactions on Antennas and Propagation, 40(1), 1992, 49–56. 12. J.G. Maloney and G.S. Smith, The use of surface impedance concepts in the finitedifference time-domain method, IEEE Transactions on Antennas and Propagation, 40(1), January 1992, 38–48. 13. A. Taflove and S.C. Hagness, Computational Electrodynamics: The Finite Difference Time-domain Method, 2nd edn., Artech House, Norwood, MA, 2000. 14. N. Farahat, The implementation of high order approximation SIBC for FDTD method, PhD dissertation, The University of Akron, Akron, OH, August 2000. 15. K.S. Oh and J.E. Schutt-Aine, An efficient implementation of surface impedance boundary conditions for the finite-difference time-domain method, IEEE Transactions on Antennas and Propagation, 43(7), 1995, 660–666.
8 Implementation of SIBCs for Volume Discretization Methods
8.1 Introduction Following implementation of low- and high-order SIBCs for the boundary integral equations in Chapters 6 and 7 we now turn our attention to implementation of SIBC for volume discretization methods. We start with the FDTD and implement the SIBCs developed in Chapter 3 for high-frequency applications. The contour-path method, based on the integral forms of Faraday and Ampere’s laws, is used for the implementation because it allows the grid to conform to the boundary and by doing so eliminates one of the main issues in use of the FDTD method—the use of regular grids. The fundamentals of the contour-path FDTD method are outlined separately in Appendix 8.A.1. This approach reduces the errors caused by discretization of boundaries and is well suited for incorporation of SIBCs. The conditions of applicability of the coupling between the FDTD and SIBC are discussed and examples to its use given to demonstrate the implementation. The finite integration technique (FIT) is another volume discretization method related to the FDTD method. Unlike the FDTD method which solves for the electric and magnetic fields, the FIT method operates on circulations of the electric field (voltages), circulations of the magnetic field (magnetomotive forces) and fluxes of the electric and magnetic field. In that sense it may be viewed as a generalized FDTD method for the solution of Maxwell’s equations in their integral form. As with the FDTD method, the FIT method is implemented for high-frequency problems using low- and high-order SIBCs in Cartesian and tetrahedral grids. The relations between the FIT and FDTD grids are discussed as well as the limits of applicability and the performance of the method is evaluated through an example. Perhaps the best-known volumetric discretization method is the finiteelement method (FEM) because of its extensive use in all engineering disciplines. It is however fundamentally different than the previous two methods in that Maxwell’s equations must be transformed and adapted to the method before discretization and solution can commence. This in turn has implications on the implementation of the SIBCs. Also, the FEM has a diversity of element types and methods of formulation. For the sake of brevity we will discuss here only the formulation in the ‘‘weak’’ sense using tetrahedral edge 263
264
Surface Impedance Boundary Conditions: A Comprehensive Approach
elements although the method and approach are general enough to be useful for other types of elements and formulations. Numerical examples complete this chapter as well.
8.2 Statement of the Problem Consider a nonconducting domain V(e1 , m1 , s1 ¼ 0) surrounding a conducting body Vc (e2 , m2 , s2 ). Let S be the surface of Vc and, consequently, the interface between the conducting and nonconducting regions (Figure 8.1). The external boundary of the region V is denoted as Sinf . The distribution of the electromagnetic field in both regions can be described by Maxwell’s equations in the following form: Nonconducting region V: ~ ¼ e1 rH
@~ E @t
r~ E ¼ m1
(8:1)
~ @H @t
(8:2)
Conducting region Vc: @~ E ~ ¼ s2~ rH E þ e2 @t r~ E ¼ m2
(8:3)
~ @H @t
(8:4)
We assume that the electromagnetic field in V is induced by external sources. Let the angular frequency v or duration t of the pulse of the incident field be such that the skin depth d is much smaller than the characteristic size D of the δ
V
S
σ=0 Vc D H in c FIGURE 8.1 A general geometry. A conducting region surrounded by a dielectric.
E in c Sinf
σ>0
265
Implementation of SIBCs for Volume Discretization Methods
conductor, i.e., the condition in Equation 2.1 is satisfied. Suppose our main interest is focused on the field in V. Then it is natural to replace Equations 8.3 and 8.4 by the SIBCs (Equation 3.132) applied on S and by doing so eliminate Vc from the numerical procedure so that only V has to be discretized.
8.3 Finite-Difference Time-Domain Method Without any loss of generality we restrict ourselves to a two-dimensional problem in the xz-plane by assuming no variation along the y-axis (Figure 8.2). In this case, Equations 8.1 and 8.2 take the forms: @Hx 1 @Ey ¼ @t m1 @z @Hy 1 @Ez @Ex ¼ @t @z m1 @x
(8:5) (8:6)
@Hz 1 @Ey ¼ @t m1 @x
(8:7)
@Ex 1 @Hy ¼ @t e1 @z @Ey 1 @Hx @Hz ¼ @t @x e1 @z
(8:8) (8:9)
x Ez(i, k – 1/2)
Ex(i – 1/2, k – 1)
Hy(i – 1/2, k – 1/2)
Ez(i – 1, k – 1/2) FIGURE 8.2 The original (Cartesian) FDTD boundary cell in two dimensions.
Ex(i – 1/2, k)
z
266
Surface Impedance Boundary Conditions: A Comprehensive Approach
@Ez 1 @Hy ¼ @t e1 @x
(8:10)
Equations 8.5 through 8.10 can be decoupled into transverse electric TEy and transverse magnetic TMy modes as follows: TEy mode:
TMy mode:
@Hy 1 @Ez @Ex ¼ @t @z m1 @x
(8:11)
@Ex 1 @Hy ¼ @t e1 @z
(8:12)
@Ez 1 @Hy ¼ @t e1 @x
(8:13)
@Ey 1 @Hx @Hz ¼ @t @x e1 @z
(8:14)
@Hx 1 @Ey ¼ @t m1 @z
(8:15)
@Hz 1 @Ey ¼ @t m1 @x
(8:16)
Independence of these modes allows one to decompose an arbitrary propagation in isotropic media into these two modes and solve them separately. Here we consider the TEy case whereas the TMy is analogous. Discretizing the domain V with rectangular cells and applying Yee’s algorithm [14], we write the following finite-difference approximations of Equations 8.11 through 8.13 for the ith cell in the x-direction and the kth cell in the z-direction at the nth time step: nþ1=2
n1=2
(i 1=2, k 1=2) Hy (i 1=2, k 1=2) Dt 1 Enz (i, k 1=2) Enz (i 1, k 1=2) Enx (i 1=2, k) Enx (i 1=2, k 1) ¼ m1 Dx Dz (8:17)
Hy
n Enþ1 x (i 1=2, k 1) Ex (i 1=2, k 1) Dt nþ1=2
¼
1 Hy e1
nþ1=2
(i 1=2, k 1=2) Hy Dz
(i 1=2, k 3=2)
(8:18)
Implementation of SIBCs for Volume Discretization Methods
267
n Enþ1 z (i 1, k 1=2) Ez (i 1, k 1=2) Dt nþ1=2
¼
1 Hy e1
nþ1=2
(i 1=2, k 1=2) Hy Dx
(i 3=2, k 1=2)
(8:19)
where Dx and Dy are the rectangular cell sizes Dt is the time step If the kth cell is a boundary cell adjacent to the interface (this case is shown in Figure 8.2), the electric field is determined at the interface so that the quantity Enx (i 1=2, k) in Equation 8.17 is given by the SIBC (Equation 3.132). The simplest approximation of Equation 3.132 in the discrete space and time according to Yee’s algorithm can be written in the form: h i ^ H n (i 1=2, k) Enx (i 1=2, k) ¼ F y h i h i ^ Hynþ1=2 (i 1=2, k 1=2) þ F ^ Hyn1=2 (i 1=2, k 1=2) F 2 (8:20) Substituting Equations 8.20 into 8.17, we obtain i Dt ^h nþ1=2 (i 1=2, k 1=2) ¼ Hyn1=2 (i 1=2, k 1=2) F Hy 2m1 Dz i Dt n Dt ^h n1=2 Ez (i, k 1=2) Enz (i 1, k 1=2) (i 1=2, k 1=2) þ F Hy m1 Dx 2m1 Dz Dt n E (i 1=2, k 1) (8:21) þ m1 Dz x
Hynþ1=2 (i 1=2, k 1=2) þ
Since the high-order SIBC takes into account the curvature of the body we use the generalized FDTD approach for curved surfaces (contour-path method [7] described in Appendix 8.A.1). In this method, the boundary cells adjacent to the surface of the scatterer are distorted in order to conform to the arbitrary curved surface of the scatterer. Figure 8.3 shows a nonrectangular boundary cell for the generalized FDTD formulation. The magnetic field updating Equation 8.17 can be written in the following form: Hynþ1=2 (i 1=2, k 1=2) ¼ Hyn1=2 (i 1=2, k 1=2) Dt n þ gEz (i, k 1=2) fEnz (i 1, k 1=2) sEntan (i 1=2, k) þ hEnx (i 1=2, k 1) m1 A (8:22)
where g, f, s, h are the side lengths of the distorted cell A is the area of the patch
268
Surface Impedance Boundary Conditions: A Comprehensive Approach
g
x
Ez (i, k – 1/2) Ex (i – 1/2, k – 1)
Eτ
h Hy (i – 1/2, k – 1/2)
S
Ez (i – 1, k – 1/2) z f FIGURE 8.3 Contour-path FDTD boundary cell in two dimensions.
Replacing Entan (i 1=2, k) by the SIBC in Equation 8.20, we obtain i sDt ^h nþ1=2 Hynþ1=2 (i 1=2, k 1=2) þ (i 1=2, k 1=2) ¼ Hyn1=2 (i 1=2, k 1=2) F Hy 2m1 A i Dt 1 ^h n1=2 þ (i 1=2, k 1=2) gEnz (i, k 1=2) fEnz (i 1, k 1=2) sF Hy m1 A 2 (8:23) þ hEnx (i 1=2, k 1)
^ of the surface impedance In our two-dimensional case, the time derivative F function is written in the form: 2 ^ b * Hy þ 1 @ Hy ^ b * Hy T ^ Hy ¼ T ^ b * Hy T (8:24) F 2 3 4 d 8d2 2 @s2 where ~ d is the local radius of curvature of the interface and the time-dependent b , m ¼ 2, 3, 4, are given by Equations 7.69 through 7.71. Substitutfunctions Tm ing Equation 8.24 into Equation 8.23, we obtain Hynþ1=2 (i 1=2, k 1=2) þ
) 2 nþ1=2 (i 1=2, k 1=2) ^ b sDt ^b 1 T ^b 1 T ^ b 1 @ Hy Hynþ1=2 (i 1=2, k 1=2) * T T * 2 4 2m1 A d 3 8d2 4 2 @s2
Dt ¼ 1=2, k 1=2) þ gEnz (i, k 1=2)fEnz (i 1, k 1=2) hEnx (i 1=2, k 1) m1 A 2 n1=2 (i 1=2, k 1=2) ^ b 1 ^b 1 T ^b 1 T ^ b þ 1 s @ Hy sHyn1=2 (i 1=2, k 1=2) * T T * 4 2 2 d 3 8d2 4 4 @s2 Hyn1=2 (i
(8:25)
Implementation of SIBCs for Volume Discretization Methods
269
The following representation is used to approximate the term @ 2 Hy @s2 in discrete space: @ 2 Hyn @s2
" 1 # 2 n @s 1 @ Hy @s 1 @s ¼ þ þ þ @x2 @y2 @y @x@y @x @y ( n ) @Hy @ 2 s @s 2 @Hyn @ 2 s @s 2 @Hyn @ 2 s @s 2 @Hyn @ 2 s @s 2 þ þ þ @x @x2 @x @y @x@y @y @x @x@y @x @y @y2 @y @ 2 Hyn
@s @x
1
@ 2 Hyn
(8:26)
Although Equation 8.25 is suitable for computer implementation, it has the following disadvantages: nþ1=2
1. It cannot be solved with respect to Hk1=2 by analytical methods so that the finite-difference algorithm becomes ‘‘implicit’’ for the ‘‘boundary’’ cell whereas it is ‘‘explicit’’ for other cells of the grid. 2. The full convolution with all past field components is included in Equation 8.25. This renders its evaluation for large problems impractical. These difficulties can be overcome by application of recursive formulas (described in Appendix 7.A.1) for fast and economic computation of timeconvolution integrals. In addition, the number of integrals on the left-hand side of Equation 8.25 can be reduced using the perturbation technique developed in the previous chapters. Using the scale factors introduced in Equation 3.52, we represent Equation 8.22 with nondimensional variables as follows: ~ n (i, k 1=2) ~ n1=2 (i 1=2, k 1=2) þ ~q g~E ~ nþ1=2 (i 1=2, k 1=2) ¼ H H y y z ~ Q ~ n (i 1=2, k) þ ~ ~ n (i 1=2, k 1) ~ n (i 1, k 1=2) ~sE hE (8:27) ~f E z tan x where ~ is the Courant number Q ~q is the electric dimension of the body defined in Equation 7.107 The values g, f, s, and h in Equation 8.27 have been normalized by the scale factor A=D where qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi D ¼ DxDz Dx2 þ Dy2
(8:28)
270
Surface Impedance Boundary Conditions: A Comprehensive Approach
Switching in Equations 8.20 and 8.24 to nondimensional variables, we obtain the SIBC in the following form: ~ nþ1=2 þ H ~ nþ1=2 þ H ~ yn1=2 ~ yn1=2 1~ ~ ~ tan ¼T ^ b * Hy ^ b * Hy ~ E p T 2 3 ~ 2 2 d ! nþ1=2 n1=2 2 ~ nþ1=2 2 ~ n1=2 ~ ~ H H H H @ @ 1 1 b y y y y ~ ^ * þ þ þ ~ p2 T 4 @~s2 @~s2 4 4 16~ d2 16~ d2
(8:29)
We represent the magnetic field as power series in the small parameter ~p: ~ nþ1=2 ~ nþ1=2 þ ~ ~ nþ1=2 þ ~p2 H ~ nþ1=2 (i 1=2, k 1=2) ¼ H pH H y y1 y2 y3
(8:30)
Substituting Equations 8.29 and 8.30 into Equation 8.27 and equating the coefficients of equal powers of ~ p, provides the following equations for the terms of expansion: ~ p0 : ~b * H ~ nþ1=2 þ ~s~q T ~ nþ1=2 ¼ H ~ n1=2 (i 1=2, k 1=2) H 2 y y1 y1 ~ 2Q ~q ~ n ~^ b ~ n1=2 ~ n (i 1, k 1=2) ~s T ~ gEz (i, k 1=2) ~f E þ z 2 * Hy ~ 2 Q ~ n (i 1=2, k 1) þ~ hE (8:31) x ~ p1 : ~s~q 1 ^ ~ b ~ nþ1=2 ~b ^ ~ nþ1=2 þ H ~ n1=2 ~ nþ1=2 þ ~s~q T * H ¼ * H H T y 3 y2 y2 y1 ~ 2 ~ ~ 2Q 2Q d
(8:32)
~ p2 : ~ nþ1=2 H ~ yn1=2 ~s~q ^ ~s~q 1 ~ ~s~q ~^ b H b y1 ~ b ~ nþ1=2 nþ1=2 nþ1=2 ^ ~ ~ T 3 * Hy2 T 2 * Hy3 T4 * þ ¼ þ Hy3 ~ ~ ~ ~ 2Q 2Q 4Q d 4~d2 4~d2 ! ~ nþ1=2 @ 2 H ~ yn1=2 @2H y1 þ þ (8:33) @~s2 @~s2 Returning to dimensional variables, the proposed formulation is written in the form: nþ1=2
Hynþ1=2 (i 1=2, k 1=2) ¼ Hy1
nþ1=2
þ Hy2
nþ1=2
þ Hy3
(8:34)
271
Implementation of SIBCs for Volume Discretization Methods
sDt b nþ1=2 T H ¼ Hyn1=2 (i 1=2, k 1=2) 2m1 A 2 * y1 Dt n n s ^b H n1=2 þ gEz (i, k 1=2) fEnz (i 1, k 1=2) T m1 A 2 2* y o þ hEnx (i 1=2, k 1)
nþ1=2
Hy1
nþ1=2
Hy2 nþ1=2
Hy3
þ
þ
þ
sDt ^ b sDt 1 ^ b nþ1=2 nþ1=2 T2 * H2 T3 * Hy1 ¼ þ Hyn1=2 2m1 A 2m1 A d
sDt ^ b nþ1=2 sDt 1 ^ b nþ1=2 sDt ^ b T2 * H3 T T3 * Hy2 ¼ 2m1 A 2m1 A d 4m1 A 4 * nþ1=2
Hy1
4d2
nþ1=2
n1=2 @ 2 Hy1 Hy þ þ 4d2 @s2
n1=2
@ 2 Hy þ @s2
(8:35) (8:36)
! (8:37)
The first equation is the implementation of the Leontovich SIBC. The second and third equations provide corrections by taking into account the curvature of the body’s surface and the field variation in the tangential direction. We emphasize that Equations 8.35 through 8.37 are different only in the righthand sides. Note that the left-hand sides of Equations 8.35 through 8.37 involve the time-convolution product, therefore, these equations cannot be solved with nþ1=2 (i 1=2, k 1=2) by analytical methods. This makes the respect to Hy scheme implicit for the ‘‘boundary’’ cell and reduces efficiency of Yee’s algorithm (which is explicit for all other cells). However, Equations 8.35 through 8.37 admit further transformations due to the fact that the electric field in V decreases when we approach the conductor’s surface. Since Enx (i 1=2, k) is related to the interface whereas Enx (i 1=2, k 1) is related to the dielectric medium, we can assume that Enx (i 1=2, k) Enx (i 1=2, k 1)
(8:38)
Thus, Equation 8.23 contains terms of different orders of magnitude, which means we can transform this equation using the perturbation technique. As the first step we transfer to nondimensional variables by choosing appropriate scale factors. Let Ev be the characteristic scale for variation of the electric field in all nodes of the boundary cell (located in the dielectric region V) except at the interface. Thus we can write ~ n (i 1=2, k 1) Enx (i 1=2, k 1) ¼ Ev E x
(8:39)
~ n (i, k 1=2) Enz (i, k 1=2) ¼ Ev E z
(8:40)
272
Surface Impedance Boundary Conditions: A Comprehensive Approach
~ n (i 1, k 1=2) Enz (i 1, k 1=2) ¼ Ev E z
(8:41)
As in previous chapters, the sign ‘‘’’ denotes nondimensional values. Using Equation 8.39, the scale factor Ek for the electric field Enx (i 1=2, k) at the interface can be represented in the form: Ek* ¼ ~yEv
(8:42)
~y ¼ E*k =Ev 1
(8:43)
~ n (i 1=2, k) ¼ ~yEv E ~ n (i 1=2, k) Enx (i 1=2, k) ¼ Ek* E x x
(8:44)
nþ1=2
where ~y is the small parameter. Since the quantities Hy (i 1=2, k 1=2) n1=2 (i 1=2, k 1=2) are of the same order of magnitude, they should and Hy have a common scale factor Hv . That scale factor can be defined from Equation 8.6 as follows: Hv ¼ Ev Dt=(m1 Dz) nþ1=2
(8:45) n1=2
Using Equation 8.45, the values Hy (i 1=2, k 1=2) and Hy k 1=2) can be represented in the form:
(i 1=2,
~ nþ1=2 (i 1=2, k 1=2) Hynþ1=2 (i 1=2, k 1=2) ¼ Hv H y ¼
Ev Dt ~ nþ1=2 H (i 1=2, k 1=2) m1 Dz y
(8:46)
~ n1=2 (i 1=2, k 1=2) Hyn1=2 (i 1=2, k 1=2) ¼ Hv H y ¼
Ev Dt ~ n1=2 H (i 1=2, k 1=2) m1 Dz y
(8:47)
With nondimensional variables, Equation 8.23 takes the form: h i ~^ ~ nþ1=2 ~ n1=2 (i 1=2, k 1=2) ~ nþ1=2 (i 1=2, k 1=2) þ ~y 1 F (i 1=2, k 1=2) ¼ H H Hy y y 2 h i ~^ ~ n1=2 ~ n (i, k 1=2) E ~ n (i 1, k 1=2) 1 F ~ n (i 1=2, k 1) þE (i 1=2, k 1=2) þ E Hy z z x 2
(8:48)
~ ynþ1=2 (i 1=2, k 1=2) for which the solution is We represent the function H sought in the form of expansions in the small parameter ~y: ~ nþ1=2 (i 1=2, k 1=2) ¼ ~ h0 þ ~y~h1 þ O ~y2 H y
(8:49)
273
Implementation of SIBCs for Volume Discretization Methods
Substituting Equation 8.49 into Equation 8.48 and equating the coefficients of equal powers of ~y, we obtain the equations for ~h0 and ~h1 : ~ ~ n1=2 (i 1=2, k 1=2) þ E ~ n (i, k 1=2) E ~ n (i 1, k 1=2) h0 ¼ H y z z h i 1~ ^ H ~ n1=2 (i 1=2, k 1=2) þ E ~ n (i 1=2, k 1) F y x 2 h i 1~ ~ ^ ~ h1 ¼ F h0 2
(8:50) (8:51)
In effect, we developed a two-step technique whereby ~h0 is calculated using Equation 8.50 and then ~ h1 is calculated using Equation 8.51. In dimensional variables, Equations 8.50 and 8.51 take the forms: Dt n Ez (i, k 1=2) Enz (i 1, k 1=2) m1 Dz i Dt ^h n1=2 Dt n F Hy E (i 1=2, k 1) (i 1=2, k 1=2) þ (8:52) 2m1 Dz m1 Dz x
h0 ¼ Hyn1=2 (i 1=2, k 1=2) þ
h1 ¼
Dt ^ F½h0 2m1 Dz
(8:53)
and Hynþ1=2 (i 1=2, k 1=2) ¼ h0 þ h1
(8:54)
Note that we do not actually know the exact values of ~y and Ev . Both of them are required only at the stage of derivation and are obviously not included in the dimensional formulation (Equations 8.52 through 8.54). All we really need to know is that the condition in Equation 8.38 is satisfied (and, consequently, Equation 8.43 holds). The representation in Equations 8.52 through 8.54 has clear physical meaning, namely h0 is the magnetic field calculated under the assumption that medium 2 is a perfect electrical conductor (PEC) and the electric field at the interface is zero; h1 is the correction allowing for finite conductivity of medium 2. Finally, combining Equations 8.34 through 8.37 and Equations 8.52 through 8.54, we obtain nþ1=2
Hynþ1=2 (i 1=2, k 1=2) ¼ Hy0 nþ1=2
Hy0
nþ1=2
þ Hy1
nþ1=2
þ Hy2
nþ1=2
þ Hy3
Dt gEnz (i, k 1=2) m1 A fEnz (i 1, k 1=2) þ hEnx (i 1=2, k 1)
(8:55)
¼ Hyn1=2 (i 1=2, k 1=2) þ
(8:56)
274
Surface Impedance Boundary Conditions: A Comprehensive Approach sDt ^ b nþ1=2 T2 * Hy0 þ Hyn1=2 (i 1=2, k 1=2) 2m1 A sDt ^ b nþ1=2 nþ1=2 T2 * Hy1 ¼ þ Hyn1=2 (i 1=2, k 1=2) Hy2 2m1 A sDt 1 ^ b nþ1=2 þ þ Hyn1=2 (i 1=2, k 1=2) T3 * Hy0 2m1 A d sDt ^ b nþ1=2 nþ1=2 T2 * Hy2 ¼ þ Hyn1=2 (i 1=2, k 1=2) Hy3 2m1 A sDt 1 ^ b nþ1=2 þ T3 * Hy1 þ Hyn1=2 (i 1=2, k 1=2) 2m1 A d ! nþ1=2 nþ1=2 n1=2 n1=2 Hy0 @ 2 Hy0 Hy @ 2 Hy sDt ^ b T þ þ þ 4d2 4d2 @s2 @s2 4m1 A 4 * nþ1=2
Hy1
¼
(8:57)
(8:58)
(8:59)
It is easy to see that Equations 8.55 through 8.59 are fully explicit because the term containing the impedance function is now on the right-hand sides of the equations. Let us derive the conditions of applicability of the technique. We will perform the derivation in the frequency domain considering propagation of a uniform plane wave from region 1 (z < 0) to region 2 (z > 0) (one-dimensional problem). Because of reflection, the following waves travel in both regions: Incident wave: þ Eþ 1 ¼ E10 exp (jb1 z)
(8:60)
E 1 ¼ E10 exp (jb1 z)
(8:61)
þ Eþ 2 ¼ E20 exp (g2 z)
(8:62)
Reflected wave:
Transmitted wave:
where pffiffiffiffiffiffiffiffiffiffi b1 ¼ v m1 e1 ¼ 2p=l;
g2 ¼
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðs2 þ jvÞjvm2
(8:63)
Since the total electric field is continuous at z ¼ 0, the reflection coefficient can be represented in the form: þ G ¼ E 10 =E10 ¼
h2 h1 h2 =h1 1 ¼ h2 þ h1 h2 =h1 þ 1
(8:64)
Implementation of SIBCs for Volume Discretization Methods
h1 ¼
pffiffiffiffiffiffiffiffiffiffiffiffi m1 =e1 ;
h2 ¼
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi jvm2 =ðs2 þ jve2 Þ
275
(8:65)
In the conducting region we have s2 ve2
(8:66)
h2 =h1 1
(8:67)
so that
With these, the reflection coefficient can be represented in the form: G (s 1)(1 s) þ O s2 ¼ 1 þ 2s þ O s2
(8:68)
where sffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffisffiffiffiffiffiffiffiffiffi jvm2 e1 m2 e1 jve2 1 þ j 1 ¼ s¼ ¼ pffiffiffi Bk2 1 s2 m1 m1 e2 s2 2 rffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffi s2 k2 ¼ ¼ tan u ve2 rffiffiffiffiffiffiffiffiffiffi m 2 e1 B¼ m 1 e2
(8:69) (8:70) (8:71)
and tan u is the loss tangent in medium 2. Using Equation 8.68, we write the total electric field in medium 1 in the form: þ E1 (z) ¼ Eþ 10 expðjb1 zÞ þ E10 expð jb1 zÞ ¼ E10 ðexpðjb1 zÞ þ G expð jb1 zÞÞ h i pffiffiffi 1 pffiffiffi 1 ¼ Eþ exp ð jb z Þ þ 1 þ 2 Bk þ j 2 Bk ð z Þ exp jb 1 1 10 2 2 n h pffiffiffi io p ffiffi ffi 1 ¼ Eþ ½cos b1 z sin b1 z þ j 2Bk2 þ 1 sin b1 z þ cos b1 z 10 2Bk2
(8:72) Setting z ¼ 0 in Equation 8.72, we obtain the total field at the interface: pffiffiffi 1 E1 (0) ¼ Eþ 10 2Bk2 (1 þ j)
(8:73)
The ratio of Equations 8.72 and 8.73 gives the variation of the field in medium 1 with respect to the field at the interface:
276
Surface Impedance Boundary Conditions: A Comprehensive Approach
E1 (z) ¼ E1 (0)
2 cos
pffiffiffi pffiffiffi 2p 2p 2p z 2Bk2 sin z þ j 2Bk2 þ 2 sin z l1 l1 l1
(8:74)
Our goal is evaluation of the electric fields on the opposite sides of the ‘‘boundary’’ cell (see Figure 8.2). Usually, more than 10 cells per wavelength are taken in computations. Thus 2pz=l1 1
(8:75)
2p Expanding sin 2p l1 z and cos l1 z in a Taylor series and taking into account Equation 8.75, we represent Equation 8.74 in the form:
E1 (z) E1 (0)
2p pffiffiffi 2p pffiffiffi 2 2Bk2 z þ j 2Bk2 þ 2 z l1 l1
(8:76)
Setting z ¼ Dz, we obtain E1 (Dz) 1 E1 (0)
if
Q¼
ffi pffiffiffiffiffiffiffiffiffiffirffiffiffiffiffiffiffiffiffi m1 e2 2p tan u dz 1 m 2 e 1 l1
(8:77)
The error e due to the representation in Equation 8.49 is derived directly from the condition in Equation 8.77 and written in the form: e ¼ Q2 ¼ (tan u)1
m2 e1 l1 2 m2 ¼ m1 e2 2pDz vs2 ðm1 DzÞ2
(8:78)
Next we give a few numerical examples to illustrate the techniques. First consider the one-dimensional problem of normal incidence of a TM plane wave at the plane boundary of a lossy dielectric half-space with the following material parameters: e1 ¼ e2 ¼ e0 ; m1 ¼ m2 ¼ m0 ;
s1 ¼ 0;
s2 6¼ 0
(8:79)
The excitation is defined as follows: Excitation ¼ 1000 sin (2pt=5:56E-8),
0 < t < (5:56E-8)=2s
(8:80)
The parameters of the computational mesh are chosen as Dx ¼ Dz ¼ 5 101 m; Dt ¼ 1:18 109 s; Nz-space steps ¼ Nx-space steps ¼ 100; Ntime steps ¼ 300
(8:81)
Figure 8.4 shows the distributions of the magnetic field at the interface obtained for a material with low loss tangent ( tan u ¼ 10). The curves were
277
Implementation of SIBCs for Volume Discretization Methods
H (A/m)
6 5
σ = 0.01 S/m
4
Q = 0.6
Exact solution FDTD–SIBC FDTD–PEC
tan θ = 10
3 2 1 0 –1 110
120
130
140
150
160
170
180
t /Δt FIGURE 8.4 Distribution of the magnetic field at the interface for low-conductivity scatterer.
obtained using the PEC-condition (dotted line), exact formula (dashed line), and Equations 8.55 through 8.59 (solid line). The error is significant as predicted by Equation 8.78, (Q < 1). For a material with high loss tangent tan u ¼ 103 the actual error of 2.4% is in agreement with the estimate e ¼ Q2 ¼ 2:2% given by Equation 8.78 as is shown in Figure 8.5. 6 5
H (A/m)
4
Exact solution σ = 1 S/m
FDTD–SIBC FDTD–PEC
tan θ = 1000 Q = 6.6
3 2 1 0 –1 110
120
130
140
150
160
170
180
t /Δt FIGURE 8.5 Distribution of the magnetic field at the interface for higher-conductivity scatterer.
278
Surface Impedance Boundary Conditions: A Comprehensive Approach
6
Exact solution
5
FDTD–SIBC FDTD–PEC
σ = 1 S/m 60° incidence
4
H (A/m)
3 2 1 0 –1 –2 110
120
130
140
150
160
170
180
t /Δt FIGURE 8.6 Distribution of the magnetic field at the interface for higher-conductivity scatterer at oblique incidence.
The problem becomes two-dimensional in the case of oblique incidence (let the incident angle be equal to 608). The magnetic field was computed in the middle of the dielectric surface for both the exact and proposed methods. Figures 8.6 and 8.7 show the magnetic field computed in the middle of the dielectric surface for materials with low and high loss tangent, respectively, using low- and high-order SIBCs. Another example is scattering of an incident TEy wave by an infinitely long conducting cylinder as shown in Figure 7.1. The excitation function is the following:
t 1 Excitation ¼ 1000 exp 16 7:071 108
2 ! (8:82)
The diameter of the cylinder is 12 cells which is almost one-third of the average wave length of the Gaussian pulse (l ¼ 3 108t=2). t is the duration of the Gaussian pulse. Distorted cells are used adjacent to the cylinder surface to allow the FDTD cells to conform to the curved surface. Figures 8.8 and 8.9 show the total magnetic field at point A (in Figure 7.1) obtained using loworder (Leontovich’s approximation) and high-order (Rytov’s approximation) SIBCs for different values of the parameter ~ p. From these figures it follows
279
Implementation of SIBCs for Volume Discretization Methods
6
Exact solution
σ = 0.01 S/m 5
FDTD–SIBC
60° incidence
FDTD–PEC
4
H (A/m)
3 2 1 0 –1 –2
110
120
130
140
150
160
170
180
t /Δt FIGURE 8.7 Distribution of the magnetic field at the interface for low-conductivity scatterer at oblique incidence.
5
σ = 0.01 S/m
High-order SIBC Low-order SIBC PEC
4
H (A/m)
3 2 1 0 –1 –2
100
200 t/Δt
FIGURE 8.8 Magnetic field at point A, low conductivity.
300
280
Surface Impedance Boundary Conditions: A Comprehensive Approach
5
σ = 0.1 S/m
High-order SIBC Low-order SIBC PEC
4
H (A/m)
3 2 1 0 –1 –2
100
200
300
t/Δt FIGURE 8.9 Magnetic field at point A, higher conductivity.
that the difference between results obtained using low- and high-order SIBCs is not visible at very small values of ~ p.
8.4 Finite Integration Technique A generalized FDTD scheme for the solution of Maxwell’s equations in their integral form is known as the FIT [1]. In both methods, staggered dual grids are used for approximation of the electric- and magnetic-related parameters. An example of the orthogonal dual mesh used in FDTD and FIT is shown in Figure 8.10. However, the FDTD method operates in terms of the electric and magnetic fields whereas FIT employs circulations and fluxes as state variables. In FDTD, the nodes, where electric and magnetic fields are calculated, are located in the middle of edges and in the middle of facets, respectively. In FIT, the state variables are the so-called grid voltages and grid fluxes related to edges and facets [2]: ð e¼ Le
~ E d~l; h ¼
ð Lh
~ d~l; H
ð d¼ Sd
~ ~ D nds;
ð b¼ Sb
~ B ~ nds
(8:83)
281
Implementation of SIBCs for Volume Discretization Methods
b2 B΄
B
e2 A
A΄
B0virt
e1
B1virt
b0
n
E 1virt
C
b1 D΄
e3
e4
x
D
b4
z y
FIGURE 8.10 Cartesian computational cell used in the FIT and FDTD methods.
where Li and Si are the cell’s edge and facet, respectively the subscript denotes the type of grid (primary or dual) the vectors d~l and ~ n are directed along the edge and normal to the facet, respectively Faraday’s law in integral form applied to the facet ABCD of the computational cell shown in Figure 8.10 can be written with the variables in Equation 8.83 as follows: ð
~ E d~l ¼
Le
ð ~ @B ~ nds @t
,
e1 þ e2 þ e3 þ e4 ¼ db0 =dt
(8:84)
Sd
The left-hand side contains all voltage components related to edges adjacent to the considered facet, combined with proper signs 1 according to their orientation. Material relations between e and d are obtained by introduction of a virtual continuous component Evirt at the intersection point of the dual and primary grids [1]. Since the normal vector to the dual facet is collinear with the direction of the intersecting primary edge, we obtain ð e¼
~ E d~l Evirt Le
(8:85)
Le
ð d¼ Sd
e~ E ~ nds eeff Evirt Ad
where
eeff ¼
1 Ad
ð eds Sd
(8:86)
282
Surface Impedance Boundary Conditions: A Comprehensive Approach
The discrete permittivity expression can now be written in the form: Ð
e~ E ~ nds d Sd eeff Ad ¼ Ð ~ ~ Le e E dl
(8:87)
Le
Similarly, a virtual magnetic flux density Bvirt is allocated as a continuous normal component at the center of the primary facet so that ð b¼ ð h¼
~ B ~ nds Bvirt Sb
(8:88)
Sb virt m B d~l m1 Le eff B 1~
where
m1 eff
Lh
1 ¼ Lh
ð
m1 dl
(8:89)
Lh
and Ð
B d~l m1~ m1 Lh h Lh eff ¼ Ð ~ Ab b B ~ nds
(8:90)
Sb
We now represent the SIBC relating the electric and magnetic fields in terms of the circulations and fluxes as state variables of the FIT. Suppose the facet ABCD of the Cartesian cell shown in Figure 8.10 belongs to the dielectric= conductor interface. To truncate the mesh, additional equations relating e1 , e2 , e3 , e4 , and b0 , and containing material properties of the conductor are needed. An SIBC relating the normal and tangential components of the magnetic flux at the interface is obtained directly from Equation 7.65 and written in the form: Bz ¼ Bh ¼
2 X @ @ @ F Bji ¼ F½Bx þ F By @ji @x @y i¼1
(8:91)
Here the surface impedance function F is given by Equations 7.66 through 7.68 for the Leontovich, Mitzner, and Rytov approximations, respectively. Applying vector identities to Equation 7.65, we obtain Bz ¼ r
h i ~ n ~ F ~ n ¼~ n r ~ n ~ F
(8:92)
Implementation of SIBCs for Volume Discretization Methods
283
where the components of the vector ~ F are defined as follows: Fx ¼ F½Bx ;
Fy ¼ F By
(8:93)
Performing integration on both sides of Equation 8.91 over the facet ABCD and applying vector identities, we obtain b0 ¼
ð ð ð h i ~ n r ~ n ~ F ds B ds ¼ r ~ n ~ F ~ n ds ¼ ~ z
Sb0
Sb0
(8:94)
Sb0
Application of Stoke’s theorem to the last equation yields b0 ¼
þ 4 ð X ~ ~ n ~ F d~l ¼ n ~ F d~l k¼1
L
(8:95)
Lk
From Figure 8.10, it follows that ~ n ~ F
ABCD
ax þ F½Bx ~ ax þ Fx~ ay ¼ F By ~ ¼ Fy~ ay
(8:96)
where ~ ax , ~ ay , and ~ az are unit vectors of the global Cartesian coordinates system shown in Figure 8.10 and ~ n ¼~ az . Substituting Equation 8.96 into Equation 8.95 and taking into account that ax ; d~lAB ¼ dlAB ~ ay ; d~lDA ¼ dlDA ~
d~lBC ¼ dlBC ~ ax ; d~lCD ¼ dlCD ~ ay (8:97)
we obtain ð ð 9 Le1 > ~ > ~ n ~ F d~l ¼ FDA dlDA ¼ F Bvirt ¼ F ½ b L > e 1 1 1 > y Sb1 > > > LDA LDA > > ð ð > > virt > L e 2 ~ ~ ~ > ~ n F dl ¼ FAB dlAB ¼ F B2 Le2 ¼ F½b2 > > > x Sb2 > = LAB LAB ð ð Le > > ~ > ~ n ~ F d~l ¼ FBC dlBC ¼ F Bvirt Le3 ¼ 3 F½b3 > 3 > y Sb3 > > > LBC LBC > > ð ð > > virt L > e4 ~ ~ ~ > ~ n F dl ¼ FCD dlCD ¼ F B4 Le4 ¼ F½b4 > > > x Sb4 ; LCD
LCD
(8:98)
284
Surface Impedance Boundary Conditions: A Comprehensive Approach
where Le1 ¼ DA; Le2 ¼ AB; Sb1 ¼ SADD0 A0 ;
Sb2 ¼ SABB0 A0 ;
Le3 ¼ BC;
Le4 ¼ CD
Sb3 ¼ SBCC0 B0 ;
Sb4 ¼ SCDD0 C0
Substitution of Equation 8.98 into Equation 8.96 yields the SIBC (Equation 8.91) in terms of the state variables of FIT: b0 ¼
4 X Lek F½bk S k¼1 bk
(8:99)
Substituting Equation 8.99 into Faraday’s law (Equation 8.84) we obtain the analog of Equation 7.64: ek ¼
Lek @ F½bk Sbk @t
(8:100)
The relations in Equations 8.99 and 8.100 are the desired SIBCs in terms of circulations of the electric field and magnetic flux. To illustrate and verify our results consider a canonical two-dimensional example, for which the analytical solution is known [3]: a line current I(t) placed at point ð0, ys Þ radiating over a half-space (Figure 8.11). Even though the high-order SIBC accurately models the curvature of the surface, the test case has been chosen with a planar surface in order not to introduce errors caused by the staircase effect (whereby the curved surface is modeled using a Cartesian grid). As a matter of fact, the modeling of curved surfaces by means of high-order SIBC and FIT would require a conformal scheme. In
x Bx By
(x0, 0)
P
Ez y
I(t) (0, ys) ε0, μ0, σ = 0
FIGURE 8.11 Geometry of the test problem.
ε2, μ0, σ
285
Implementation of SIBCs for Volume Discretization Methods
the proposed example the improvement in accuracy provided by the SIBC of high order of approximation is only due to the modeling of the variation of the electromagnetic field along the surface, which is not modeled by loworder SIBC of the Leontovich type. The computational domain (the dielectric half-space) is discretized into a 100 100 Cartesian grid made of one layer of three-dimensional Yee cubic cells with side length D ¼ 0.015 m. Mur’s first-order absorbing boundary conditions are used at the other boundaries [4]. The following current pulse is considered (Figure 8.12): I(t) ¼
t t0 ðttt 0 Þ2 e t
(8:101)
with t ¼ 40Dt, t0 ¼ 12Dt, where the time step Dt was chosen as Dt ¼ Dx=2c0 . The computed results were compared with the exact solution of the problem in the time domain, obtained from the solution in the frequency domain reported in [5] by means of the inverse fast Fourier transform algorithm. The Fourier transform of I(t) of Equation 8.101 was derived analytically as pffiffiffiffi j p 2 jvt0 v2 t2 t ve e 4 I(v) ¼ 2
(8:102)
0.4
I (A)
0.2
0
–0.2
–0.4 0
FIGURE 8.12 The current pulse I(t).
1
2 t (ns)
3
4
286
Surface Impedance Boundary Conditions: A Comprehensive Approach
120 FIT/SIBC Analytical
E (V/m)
80
40
0
–40 1
2
t (ns)
3
4
FIGURE 8.13 Electric field at the observation point for ys ¼ 20D; xs ¼ 10D; s ¼ 10 S=m.
and then substituted in the expression of the electric field at the observation point [5]:
Ez ðv, x0 Þ ¼
c0 m0 k0 I(v) 2p
1 ð
0
qffiffiffiffiffiffiffiffiffiffiffiffi 2 pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp jk0 ys 1 j2 cosðk0 x0 jÞdj 1 j2 þ e2r j2 (8:103)
where k0 ¼ v=c0 . The electric field at point ðx0 ,0Þ is computed using FIT with the low-order SIBC, FIT with the high-order SIBC, and the analytical solution (Figures 8.13 through 8.18). As can be seen from Figures 8.13 and 8.14, the low-order SIBC is sufficiently accurate when conductivity is relatively high (s ¼ 10 and 1 S=m). At low values of conductivity (s ¼ 0.1 S=m) and when the filamentary source is near the surface, the SIBC of low order does not provide accurate results and high-order SIBCs should be used (Figures 8.15 through 8.17). Figure 8.18 shows the particular case of xs ¼ 0 when the tangential derivatives are zero due to symmetry and, consequently, low- and high-order SIBCs give the same results. As a conclusion of this section we show how the conditions in Equations 8.99 and 8.100 can be generalized for tetrahedral grids. Let the facet ABC of
287
Implementation of SIBCs for Volume Discretization Methods
40
FIT/SIBC Analytical
E (V/m)
20
0
–20
1
2
t (ns)
3
4
FIGURE 8.14 Electric field at the observation point for ys ¼ 20D; xs ¼ 10D; s ¼ 1 S=m.
200
Analytical SIBC low order SIBC high order
E (V/m)
100
0
–100
1
2
3 t (ns)
FIGURE 8.15 Electric field at the observation point for ys ¼ 10D; xs ¼ 20D; s ¼ 0:1 S=m.
288
Surface Impedance Boundary Conditions: A Comprehensive Approach
120 Analytical SIBC low-order SIBC high-order
80
E (V/m)
40
0
–40 1
2
3
t (ns)
4
FIGURE 8.16 Electric field at the observation point for ys ¼ 10D; xs ¼ 30D; s ¼ 0:1 S=m.
120 Analytical SIBC low-order SIBC high-order
E (V/m)
80
40
0
–40 2
3
4 t (ns)
FIGURE 8.17 Electric field at the observation point for ys ¼ 30D; xs ¼ 50D; s ¼ 0:1 S=m.
5
289
Implementation of SIBCs for Volume Discretization Methods
Analytical SIBC low-order SIBC high-order
200
E (V/m)
100
0
–100
1
2
3
t (ns)
4
FIGURE 8.18 Electric field at the observation point for ys ¼ 30D; xs ¼ 0; s ¼ 0:1 S=m.
the tetrahedron, shown in Figure 8.19, be a part of the conductor=dielectric interface. In this case the SIBCs can be written in the form: bABC ¼
3 X Lek F½bk sin zk , S k¼1 bk
ek ¼
k ¼ 1, 2, 3
Lek @ sin zk F½bk Sbk @t
B
2
C
1 D 3
A
FIGURE 8.19 Tetrahedral computational cell.
(8:104) (8:105)
290
Surface Impedance Boundary Conditions: A Comprehensive Approach
where b1 ¼ bABD b2 ¼ bBCD b3 ¼ bACD Le1 ¼ AB Le2 ¼ BC Le3 ¼ CD Sb1 ¼ SABD Sb2 ¼ SBCD Sb3 ¼ SACD z1 is the angle between facets ABC and ABD z2 is the angle between facets ABC and BCD z3 is the angle between facets ABC and ACD
8.5 Finite-Element Method The FEM also uses discretization of space in tetrahedral grids, but, unlike FDTD and FIT, the FEM requires that Maxwell’s equations be transformed before discretization. Usually, the FEM is used in the frequency domain so we also will consider the case of single-frequency excitation. The governing equation for the magnetic field in the dielectric domain V (Figure 8.1) is obtained directly from Equations 8.1 and 8.2 and written in the form: h i ~ þ jvmH ~ ¼ r ~ J inc (8:106) r ð jve1 Þ1 r H where ~ J inc denotes the incident current. Following the method of weighted residuals, we multiply Equation 8.106 ~ m , m ¼ 1, 2, . . . , M, and integrate over by a set of vector weighting functions W V as follows: ð ð ð h i ~ m dv þ jvmH ~ m dv ¼ r ~ ~ m dv ~ W ~ W r ð jve1 Þ1 r H J inc W V
V
V
(8:107) To represent Equation 8.107 in the ‘‘weak’’ form, we integrate by parts the first term on the left-hand side of Equation 8.107 and apply the vector identity: ð ð h i 1 ~ ~ m dv ~ ~ rW r ð jve1 Þ r H Wm dv ¼ ð jve1 Þ1 r H V
V
ð
SþSinf
ð jve1 Þ1
h
i ~ m ds (8:108) ~ ~ rH n W
Implementation of SIBCs for Volume Discretization Methods
291
The surface integral in Equation 8.108 is written in the following form using Equation 8.1: ð
ð jve1 Þ1
ð h i ~ m ds ~ m ds ¼ ~ ~ ~ n~ E W rH n W
SþSinf
(8:109)
SþSinf
Substituting Equation 8.109 into Equation 8.108, we obtain the ‘‘weak’’ for~ such that mulation: find the magnetic field H ð
D E 1 ~m ~ m þ jveH ~ m dv þ ~ ~ rW ~ W n~ E ,W rH S jvm
V
ð ¼
~ m dv r ~ J inc W
D E ~m ~ n ~ E ,W
(8:110)
Sinf
V
~ m , and where holds for square integrable test functions W D E ð ~ ~ u, l ¼ ~ u ~ lds
(8:111)
S
S
A truncation technique must be applied to treat the second term on the righthand side of Equation 8.110. The magnetic field inside V can be expanded with a set of vector func~ m: tions W ~¼ H
M X
~n Hn(1) W
(8:112)
n¼1
where M is the number of degrees of freedom. Substitution of Equation 8.112 into Equation 8.110 and using the frequency domain SIBC in Equation 7.64 yields the following matrix equation for the expansions coefficients: [A] H (1) ¼ J inc ð Amn ¼
(8:113)
D E 1 ~n r W ~ m þ jveW ~ m dv þ ~ ~m ~n W PW rW , W (8:114) n S jvm
V
ð ~ m dv; J inc W Jminc ¼ r ~ V
m ¼ 1, 2, . . . , M, n ¼ 1, 2, . . . , N
(8:115)
292
Surface Impedance Boundary Conditions: A Comprehensive Approach
where the following notation has been introduced: 8 > < 1 j 1þj d2 3d23l d2l 2dl d3l ~ f 1 ~ ~ þ mvd ~ fi þ d d1 fi f d Pi ¼ i 3l l > jl jl jl jl 2u 4u 2ju 8d2l d23l : 0 þ
2
fi @2 ~
d B jl þ @ 2ju @j23l
19 > fi @2 ~ = jl jl C þ 2 A @jl @j3l > @j2l ;
fi @2 ~
(8:116)
The matrix equations for implementation of the Leontovich and Mitzner SIBCs can be easily obtained from Equations 8.113 through 8.115 by neglecting the terms containing d2 and d3 , respectively. Consider implementation of SIBCs using first-order edge finite elements ~ m and ~ Fk have the forms: [6,7]. The vector functions W ~ m ¼ l(m) rl(m) l(m) rl(m) W 1 2 2 1
(8:117)
(k) (k) (k) ~ Fk ¼ q(k) 1 rq2 q2 rq1
(8:118)
The barycentric functions li and qm can be represented in the form: l(m) i
¼ ð6Velem Þ
1
a(mi) 0
þ
!
3 X
a(mi) l xl
l¼1
qj(k)
¼ ð3Sfacet Þ
1
(nj) b0
þ
3 X
,
i ¼ 1, 2
(8:119)
,
j ¼ 1, 2
(8:120)
! (nj) b l xl
l¼1
where xi are the global Cartesian coordinates that can be expressed in terms of the local coordinates jk using the transformation matrix a: xi ¼ xoi þ
3 X
aji jj ,
i ¼ 1, 2, 3
(8:121)
j¼1
where xoi are global coordinates of the origin of the local coordinate system. Thus Equations 8.119 and 8.120 can be represented in the forms: l(m) i
¼ ð6Velem Þ
1
~a(mi) 0
þ
3 X l¼1
q(k) j
¼ ð3Sfacet Þ
1
~b(nj) þ 0
3 X l¼1
! ~a(mi) l jl
! ~b(nj) j , l l
, i ¼ 1, 2
(8:122)
j ¼ 1, 2
(8:123)
293
Implementation of SIBCs for Volume Discretization Methods
where (l) ~a(l) 0 ¼ a0 þ
3 X
(l) o ~(l) a(l) k xk ; b0 ¼ b0 þ
k¼1
~b(l) ¼ j
3 X
3 X
bk(l) xok ; ~a(l) j ¼
k¼1
3 X
ajk a(l) k ;
k¼1
ajk b(l) k , l ¼ 1, 2, 3
(8:124)
k¼1
Substitution of Equations 8.122 through 8.124 into Equations 8.117 and 8.118 ~ m and ~ allows representation of the functions W Fk in the final forms: ~ m ¼ ð6Velem Þ2 W
3 X
~ en
n¼1
~ Fk ¼ ð3Sfacet Þ
3 X 2
A(m) on
þ
3 X
! A(m) in ji
i¼1
0 ~ el @B(k) ol þ
3 X
(8:125)
1
A B(k) jl jj
(8:126)
j¼1
l¼1
where ~(m1) ~b(m2) ~b(m2) ~b(m1) , i ¼ 0, 1, 2, 3, j ¼ 1, 2, 3 ~a(m1) ~(m1)~a(m2) ~a(m2) ; B(m) A(m) ij ¼ ai j j i ij ¼ bi j j i (8:127) The first, second, and third terms in Equation 8.116 are implemented in the same way. This means that implementation of Mitzner’s SIBC does not lead to any additional computational difficulties compared to the standard Leontovich SIBC. The fourth term in Equation 8.116 contains the integrals of the following kind: ð f ji Sfac
@ 2 fji ds; @ji @jj
ð f ji Sfac
@ 2 f ji ds; @j2i
ð f ji Sfac
@ 2 fji ds, @j2j
i ¼ 1, 2, j ¼ 1, 2
(8:128)
~ m and ~ Fk . From Equation 8.127 it directly follows that where ~ f denotes W A(m) ll ¼ 0;
B(m) ll ¼ 0,
l ¼ 1, 2, 3
(8:129)
By substituting Equation 8.129 into Equations 8.125 and 8.126, we obtain ~m @ W @jl
jl
¼ 0;
@ ~ Fk @jl
jl
¼ 0, l ¼ 1, 2
(8:130)
294
Surface Impedance Boundary Conditions: A Comprehensive Approach
It is easy to see that evaluation of the integrals in Equation 8.128 by taking into account Equation 8.130 leads to zero so that the last term in Equation 8.116 disappears. Thus, the use of the first-order vector approximating functions does not allow implementation of Rytov’s SIBC accurately. That is, it does not take into account the tangential variation of the magnetic field over the surface of the conductor. Higher-order vector functions are needed for this purpose.
Appendix 8.A.1: Basics of Contour-Path FDTD Method The Yee algorithm was originally formulated in rectangular Cartesian grids [14]. In these grids, curved boundaries must be approximated by a staircase model. In certain problems this approach leads to significant error. Many alternatives have been suggested and developed to modify the original FDTD formulation in order to accurately model curved boundaries [4,8–11]. Among them are orthogonal subgridding (use of fine cells near the boundary and coarse meshing elsewhere), general nonorthogonal gridding and local conformal grids. The contour-path FDTD method is one of the methods under the category of local conformal gridding. This method, developed by Jurgenns et al. [12] uses standard orthogonal Yee cells everywhere except in the vicinity of the curved boundaries. The cells immediately adjacent to the boundary of the scatterer are distorted to conform to the boundary. The method is based on Faraday’s and Ampere’s laws in integral forms instead of differential forms in the original FDTD. The basics of the method is briefly described in the following. Faraday’s law in integral form: ð þ @ ~ H ~ nds ¼ ~ E d~l (8:A:1) m @t s
c
can be applied along a contour c as shown in Figure 8.A.1 assuming that the electric field is constant along each edge of the contour [13]: ð @ m Hz i,j,k ds ¼ l1 Ex i,j1=2,k þ l2 Ey iþ1=2,j,k l3 Ex i,jþ1=2,k l4 Ey i1=2,j,k @t s
(8:A:2) then with the assumption of the magnetic field being constant inside the surface s and applying the central differences for time derivatives: 0 m@
nþ1=2
n1=2
Hz ji,j,k Hz ji,j,k Dt
1
ð A ds ¼ l1 Ex þ l2 Ey iþ1=2,j,k l3 Ex i,jþ1=2,k l4 Ey i1=2,j,k i,j1=2,k s
(8:A:3)
295
Implementation of SIBCs for Volume Discretization Methods
Hx(i–1/2, j, k+1/2)
Ex(i, j+1/2, k) Hz(i–1, j, k) Ey(i–1/2, j, k)
ds
l4
l1
l3 Hz(i, j, k)
l2
Ey(i+1/2, j, k)
dl
Ex(i, j–1/2, k)
Hx(i–1/2, j, k–1/2) FIGURE 8.A.1 Faraday’s law in integral form.
Rearranging Equation 8.A.3 gives nþ1=2 n1=2 Dt l1 Ex i,j1=2,k þ l2 Ey iþ1=2,j,k l3 Ex i,jþ1=2,k l4 Ey i1=2,j,k Hz i,j,k ¼ Hz i,j,k mA (8:A:4) It is easy to see that Equation 8.A.4 reduces to the standard Cartesian Yee algorithm in the particular case of l1 ¼ l3 ¼ Dx, l2 ¼ l4 ¼ Dy, and A ¼ DxDy. Analogously, as shown in Figure 8.A.2, the integral form of Ampere’s law e
ð ð @ ~ ~ d~l E ~ nds ¼ H @t s
(8:A:5)
c
Ex(i–1/2, j, k+1/2)
Hx(i, j+1/2, k) Ez(i–1, j, k)
Hy(i–1/2, j, k) l4
ds l1
Hx(i, j–1/2, k) Ex(i–1/2, j, k–1/2) FIGURE 8.A.2 Ampere’s law in integral form.
l3 Ez(i, j, k)
l2 Hy(i+1/2, j, k) dl
296
Surface Impedance Boundary Conditions: A Comprehensive Approach
is applied to the contour c and the assumption of the magnetic field being constant along the contour yields e
ð @ Ez i,j,k ds ¼ l1 Hx i,j1=2,k þ l2 Hy iþ1=2,j,k l3 Hx i,jþ1=2,k l4 Hy i1=2,j,k (8:A:6) @t s
Approximating the time derivatives with the central differences scheme finally gives nþ1 n nþ1=2 nþ1=2 nþ1=2 Dt l1 Hx i,j1=2,k þ l2 Hx iþ1=2,j,k l3 Hx i,jþ1=2,k l4 Hy i1=2,j,k Ez i,j,k ¼ Ez i,j,k þ eA (8:A:7)
References 1. T. Weiland, A discretization method for the solution of Maxwell’s equations for six-component fields, Electronics and Communications AEU, 31(3), 1977, 116–120. 2. T. Weiland, Advances in FIT=FDTD modeling, Proceedings of 18th Annual Review of Progress in Applied Computational Electromagnetics Conference (ACES), Monterey, 2002, pp. 1–14. 3. G.S. Smith, On the skin effect approximation, American Journal of Physics, 58(10), 1990, 996–1002. 4. G. Mur, Absorbing boundary conditions for the finite-difference approximation of the time-domain electromagnetic field equations, IEEE Transactions of Electromagnetic Compatibility, EMC-23(4), November 1981, 377–382. 5. J.G. Maloney and G.S. Smith, The use of surface impedance concepts in the finite difference time-domain method, IEEE Transactions on Antennas and Propagation, 40(1), January 1992, 38–48. 6. M.L. Barton and Z.J. Cendes, New vector finite elements for three dimensional magnetic field computation, Journal of Applied Physics, 61(2), April, 1987, 3919–3921. 7. A. Bossavit, A rationale for edge elements in 3-D field computations, IEEE Transactions on Magnetics, MAG-24(1), January 1988, 74–79. 8. T.V. Yioultsis, N.V. Kantartzis, C.S. Antonopoulos, and T.D. Tsiboukis, A fully explicit Whithey elements—Time domain scheme with higher order vector finite elements for three-dimensional high frequency problems, IEEE Transactions on Magnetics, 34(5), September 1998, 191–202. 9. J.P. Berenger, A perfectly matched layer for the absorption of electromagnetic waves, Journal of Computational Physics, 114, 1994, 185–200. 10. K.S. Kunz and R.J. Luebbers, The Finite Difference Time-Domain Method for Electromagnetics, CRC Press, Boca Raton, FL, 1993.
Implementation of SIBCs for Volume Discretization Methods
297
11. C.S. Chan, H. Sangami, J.T. Elson, and R.F. Bowers, Conformal finite-difference time-domain methods, in Time-Domain Methods for Microwave Structures, T. Itoh and B. Houshmad, eds., IEEE Press, Piscataway, NJ, 1998, pp. 181–197. 12. T.G. Jurgens, A. Taflove, K. Umashankar, and T.G. Moore, Finite-difference timedomain modeling of curved surfaces, IEEE Transactions on Antennas and Propagation, 40(4), April 1992, 357–365. 13. A. Taflove, Computational Electrodynamics, the Finite-Difference Time-Domain Method, Artech House, Norwood, MA, 1995. 14. K.S. Yee, Numerical solution of initial boundary value problems involving Maxwell’s equations in isotropic media, IEEE Transactions on Antennas and Propagation, 14(3), May 1996, 302–307.
9 Application and Experimental Validation of the SIBC Concept
9.1 Introduction The purpose of this chapter is twofold. The first is to discuss the validation of the SIBCs developed in the previous chapters. Of particular interest here is the experimental validation of SIBCs because analytical and numerical validations have been incorporated in Chapters 4 through 8 as part of the formulation and development of the various SIBCs. The second, perhaps more important purpose is to describe a simple methodology for the selection of appropriate SIBCs for use with any particular problem the user may need to address and, indeed, to allow a simple decision process by which the user can first decide if SIBCs are in fact appropriate for the problem at hand and then to decide on the order of approximation needed. This should aid in the practical use of the foregoing SIBCs. We start with the second of these goals.
9.2 Selection of the Surface Impedance Boundary Conditions for a Given Problem In the previous chapters, we classified the various SIBCs by their order of approximation. This classification afforded a certain understanding of their behavior, especially with regard to the ability of an SIBC to take into account curvatures and tangential variations of the fields. We said, and it is worth repeating here, that the zero-order approximation (PEC limit) allows no penetration of fields into the material, the first-order approximation (Leontovich condition) allows only perpendicular penetration into the conductor based on the skin depth condition, the second-order approximation (Mitzner’s condition) allows for surface curvatures, and the third-order approximation (Rytov’s condition) allows as well for tangential variations in the field. This classification clearly indicates that the errors in the Rytov approximation are fourth order in skin depth or, as we have indicated previously, in the small parameter ~ p (see Equation 2.19 as well as Table 3.1). 299
300
Surface Impedance Boundary Conditions: A Comprehensive Approach
The correct choice of SIBCs for a given problem is a very important, practical consideration the user is often faced with. The proper selection affects not only the results obtained but also the effort and time involved in solving the problem. The required order is not always clear a priori, especially in transient applications. If the order of approximation of the SIBCs used in the computation is inadequate, the computational results will be inferior in that large errors will be present, something that is difficult to gauge when reliable experimental or numerical validation data are not available. On the other hand, application of high-order SIBCs for the calculation of very thin skin layers does not provide any gain in accuracy of the results and leads to useless computational expenses. Fortunately, the approach we have taken in the formulations of the previous chapters, especially the reliance on the Rytov expansion method and the reliance on the characteristic values of the physical problem, lends itself to a simple and reliable selection of the SIBC. In this section, we look at some simple universal relationships between the characteristic values of the problem and the order of approximation so that the SIBCs best suited for a given problem can be selected in a simple and practical way. The methodology is applicable to all formulations and SIBCs described in the previous chapters including the time-domain, frequencydomain, linear, and nonlinear SIBCs. An example is included to illustrate the methodology. 9.2.1 Characteristic Values of the Problem Any nonstatic electromagnetic problem involves two basic scales: the characteristic dimension D of the body’s surface and the characteristic time t. These are the required values in the methodology to follow since they affect the applicability of the SIBC, as was amply demonstrated in previous chapters. We use these values as the input to our methodology. The characteristic time of a problem is defined as the ratio 2=v for timeharmonic incident fields or as the incident pulse duration t for a pulsed source. The characteristic dimension D is defined as (see Section 2.2 and Table 3.1): D ¼ minðd1 , d2 Þ
(9:1)
where d1 and d2 are the radii of curvature of the surface coordinate lines in the case of smooth bodies. In the more general case of multiple conductors and sources, the condition in Equation 9.1 may be expanded as D ¼ min(d1, d2, d3, d4), where d3 might be the distance to the nearest source and d4 the distance to the nearest conductor in the configuration. The main point is that D is the smallest ‘‘feature’’ of the physical configuration of the problem that may affect the fields on the surface of the conductors and, hence, in the skin layer.
Application and Experimental Validation of the SIBC Concept
301
Using D and t, we defined the characteristic skin depth d and characteristic dimension l of the field variation along the body’s surface as follows (see Equation 2.1): rffiffiffiffiffiffiffi t d¼ ms
(9:2)
l ¼ ct
(9:3)
where c is the velocity of light. If the material properties are nonlinear, the nonlinear characteristic permeability m and conductivity s must be used in Equation 9.2. This may be the case when a nonlinear BH-curve is given or when conductivity is temperature dependent. The conditions of applicability of the surface impedance concept can now be written in terms of D, d, and l: lD lD
or
d 1 D
(9:4)
D D ¼ 1 l ct
(9:5)
~ p¼
or ~q ¼
We now take ~ p and ~q as the basic parameters of the problem (see also Chapter 2). The conditions (Equations 9.4 and 9.5) hold for all SIBCs. The various representations in the previous chapters were often written in terms of these parameters. Therefore, they are convenient in developing the methodology for SIBC selection.
9.2.2 Asymptotic Expansions Although the arguments above and those to follow apply to all SIBCs we developed, and certainly to all formulations, we will nevertheless use as an example the E–H formulation from Section 3.3. Specifically, we can rewrite Equation 3.47, which gives the electric field in the frequency and time domains, in the Rytov approximation, as follows: In the frequency domain, 2 0 _b 2~ ~~ ~2 ~2 dk ~ 1 j ~_ b ~ d3k _ _ b dk þ 2dk d3k 3d3k 1 @ H jk b k b 2 j @~ ~ ~ 4 Ej3k ¼ (1) (1 þ j)~p H jk ~ H jk H p þ~ p j k 2 2 2 @~j2k 2~ dk ~ 8~ d2k ~ d3k d23k 13 ~_ b ~_ b @2H 4 1 @2H jk j3k A5 þO ~ p , k ¼ 1, 2 (9:6) þ 2 @~j2 @~jk @~j3k 3k
302
Surface Impedance Boundary Conditions: A Comprehensive Approach
In the time domain, " ~b ~~ ~2 ~2 ~ @2H d ~ b ~b d3k ~ b ~ b dk ~ jk k b ~ ~ b dk þ 2dk d3k 3d3k 1 p Ej3k ¼ (1) ~p Hjk * T1 ~ p2 H Hjk * T2 ~ j k 2 @~j2k d~t 2~ dk ~ 8~ d2k ~ d3k d23k ! # 2 ~b ~b @2H 4 1@ H j3k jk ~b þ O ~ T þ (9:7) p , k ¼ 1, 2 2 * 2 ~ ~ ~ 2 @ j3k @ jk @ j3k
The approximation in Equations 9.6 or 9.7 allows us to replace the condition in Equation 9.4 by ~ p4 1
(9:8)
The representations in Equation 9.6 and 9.7 clearly show the various approximation orders. The zero-order (PEC) representation (~p ¼ 0) gives zero electric field, as required. The first-order approximation (~p2 ¼ 0) gives the Leontovich approximation: ~_ b þ O ~p2 ~ b ¼ (1)k (1 þ j)~ pH E j3k jk
(9:9)
d ~ b ~ b ~ b ¼ (1)k ~ p Hjk * T2 þ O ~p2 E j3k ~ dt
(9:10)
The second-order approximation (~ p3 ¼ 0) gives the Mitzner approximation: " # 1 j ~_ b ~ dk ~d3k _ k b b ~ ~ H jk p H jk ~ p þ O ~p3 Ej3k ¼ (1) (1 þ j)~ ~ ~ 2 2dk d3k "
~b E j3k
# ~ d ~ b ~ b d3k ~ b ~ b dk ~ Hjk * T2 ~ Hjk * T1 þ O ~p3 ¼ (1) (1 þ j)~ p p d~t 2~ dk ~ d3k
(9:11)
k
(9:12)
The third-order (Rytov) approximation in Equations 9.6 and 9.7 clearly include the zeroth-, first-, and second-order approximations as well as the third-order corrections needed for the representation of tangential fields. 9.2.3 Methodology Now we can evaluate the ranges of the characteristic values for which the SIBCs (Equations 9.6 through 9.12) are best applicable. The conditions (Equations 9.4 and 9.5) of applicability of the SIBCs involve the two characteristic scales (D and t) and the two parameters of the problem (~p and ~q); therefore, the scales are ‘‘uniquely’’ expressed by these parameters.
Application and Experimental Validation of the SIBC Concept
303
From Equation 9.8 it follows that the approximation errors in the PEC limit, the Leontovich SIBCs, the Mitzner SIBCs, and the Rytov SIBCs are ~p, ~p2, ~ p3, and ~ p4, respectively. Based on this observation, we can define approximate ranges of the parameter ~ p, for which the SIBCs of these classes can be best applied: 1. For the PEC approximation, ~ p < 0:06
(9:13)
2. For SIBCs in the Leontovich approximation, ~ p ¼ 0:06 7 0:25
~ p2 0:003 7 0:06
(9:14)
3. For SIBCs in the Mitzner approximation, ~ p ¼ 0:25 7 0:4
~ p3 0:02 7 0:06
(9:15)
4. For SIBCs in the Rytov approximation, ~ p ¼ 0:4 7 0:5
~ p4 0:03 7 0:06
(9:16)
The range of the parameter ~q can be defined as ~q < 0:06
(9:17)
With the definitions in Equations 9.13 through 9.17, the approximation error due to using the specific SIBCs will not exceed 6%. These relations may be used as follows. From Equations 9.2, 9.4, and 9.5 we can write pffiffiffi t D ¼ pffiffiffiffiffiffiffi ~ p sm
(9:18)
D ¼ ct~q
(9:19)
and
Now, by substituting the extreme values of the parameters p~ and ~q from Equations 9.13 through 9.17 into Equations 9.18 and 9.19, the desired ranges of the scales D and t can be obtained for a given problem.
304
Surface Impedance Boundary Conditions: A Comprehensive Approach
A small disadvantage of the relations in Equations 9.18 and 9.19 is that they are written in terms of specific material properties. To obtain more universal relations we introduce the following nondimensional variables: ~ ¼ smcD D
(9:20)
~t ¼ smc2 t
(9:21)
With the variables in Equations 9.20 and 9.21, the functions in Equations 9.18 and 9.19 can be written in the following forms: ~ ¼ ~q~t D pffiffiffi ~ ¼ ~t D ~ p
(9:22) (9:23)
These are more convenient than the relations in Equations 9.18 and 9.19 because they are nondimensional. Now we can plot these for the extreme values of ~ p and ~q given in Equations 9.13 through 9.17. A representation of these values is shown in Figure 9.1. Region 1a in Figure 9.1 shows the application area of the SIBCs in the PEC limit. Region 1b is the application range in the Leontovich approximation. Similarly, the Mitzner approximation and the Rytov approximation ranges ~ t plane lies in are shown as 1c and 1d, respectively. If the point in the D~ regions 2 and 3, the surface impedance concept cannot be applied because the conditions in Equations 9.13 through 9.17 break down.
4
Log (D*)
1a
3 (q~> 0.06)
1b
3
1c ~ > 0.5) 2 (p 1d
2 3
4
5
6 Log (τ*)
FIGURE 9.1 ~ The (D)*(~ t)* plane.
7
8
9
Application and Experimental Validation of the SIBC Concept
305
Now we can summarize the methodology for selecting an appropriate SIBC for a given physical problem as follows: 1. Identify the characteristic values D and t, and the material properties s and m. ~ and ~t using 2. Calculate or estimate the nondimensional values D Equations 9.20 and 9.21. ~ t plane. 3. Find the appropriate point in the D~ 4. If this point lies in the regions 1a through 1d, choose the corresponding SIBCs from Equation 9.6 or Equation 9.7 or the PEC boundary conditions by selecting those terms indicated in Equations 9.13 through 9.17. It should be reemphasized that the functions shown in Figure 9.1 are ‘‘universal’’ because they do not depend on the properties of the conductors involved. To demonstrate the application of the foregoing methodology, we consider here a problem in the time domain made of a pair of identical parallel copper conductors with a circular cross section in which equal and oppositely directed pulses of current of magnitude 1 A flow from an external source as shown in Figure 9.2. The radius of each conductor and the distance between the conductors were taken equal to 0.1 m (characteristic value D ¼ 0.1 m). Under these conditions the current density has only one component directed along the conductors.
I
b
ξ
a
d I
FIGURE 9.2 Physical geometry of the problem.
306
Surface Impedance Boundary Conditions: A Comprehensive Approach
The SIBCs in Equations 9.10, 9.12, and 9.7 as well as the PEC condition were coupled with the surface integral equation using the formulation in Section 3.3 and solved using the boundary element method (BEM) as described in Chapter 6. To illustrate the method, the distributions of the surface current density over one half of the cross section of one conductor were calculated for the following current pulses: p ¼ 3.7 102 and ~q ¼ 3.3 107). From Equations 9.13 1. t ¼ 103 s (~ through 9.17 it follows that the PEC condition is suitable for this problem. Figure 9.3 shows that the use of an SIBC of the next order of approximation (Leontovich’s SIBC) will not provide a significant increase in accuracy of the results (the difference between the curves does not exceed 4%). p ¼ 1.2 101 and ~q ¼ 3.3 108). In this case, the 2. t ¼ 102 s (~ Leontovich SIBC seems to be optimal. From Figure 9.4 it follows that the difference between the curves obtained using the PEC and Leontovich conditions is about 15% whereas application of the Mitzner SIBC increases the accuracy by only 2%. ~ ¼ 3.7 101 and ~q ¼ 3.3 109). Figure 9.5 shows that 3. t ¼ 101 s (p the use of the Leontovich SIBC leads to unacceptable computational errors (about 18%). On the other hand, the difference between the curves obtained in the Mitzner and Rytov approximations does not exceed 3%. Therefore, in this problem it is necessary to use the Mitzner SIBC as the methodology predicts. 4
B C
3 K, A/m
B PEC limit C Leontovich’s SIBC
τ = 10−3 s ~ = 0.037 p
2
1 b
a 0.0
0.1
0.2
0.3
ξ, m FIGURE 9.3 Distribution of the surface current density over the conductor’s surface for t ¼ 103 s.
307
Application and Experimental Validation of the SIBC Concept
4 B
B PEC limit
C
C Leontovich’s SIBC D Mitzner’s SIBC
3 K, A/m
D τ = 10−2 s ~ = 0.12 p
2
1 b
a 0.0
0.1
0.2
0.3
ξ, m FIGURE 9.4 Distribution of the surface current density over the conductor’s surface for t ¼ 102 s.
4.0 B 3.0 K, A/m
E
B
PEC limit
C
Leontovich’s SIBC
D
Mitzner’s SIBC
E
Rytov’s SIBC
D 2.0
τ = 10−1 s ~ = 0.37 p
C
1.0 a 0.0
b 0.1
0.2
0.3
ξ, m FIGURE 9.5 Distribution of the surface current density over the conductor’s surface for t ¼ 101 s.
The methodology outlined above also facilitates the estimation of the approximation errors. Since the limit of applicability of a specific order of approximation is encapsulated in the condition in Equation 9.18 (or in Equation 2.1), we can rewrite this relation as follows:
308
Surface Impedance Boundary Conditions: A Comprehensive Approach m=2 ~ pm ¼ t= smD2 1
(9:24)
where the values m ¼ 1, 2, 3 correspond to the Leontovich SIBC, the Mitzner SIBC, and the Rytov SIBC, respectively, whereas ~p ¼ 0 corresponds to the PEC limit. We assume that the characteristics of the conductor’s material are constant and Equation 9.24 contains two parameters: t and D. Usually the duration of the pulse may vary whereas the radius of the conductor is constant. For such cases an approximation error em can be estimated from Equation 9.24 as follows: em ¼ am tm=2 ;
1=2 a ¼ smD2 ,
m ¼ 1, 2, 3
(9:25)
Note that the basic scale factors do not contain information about all the features of the problem. In particular, the definition of the time-scale factor as the duration of the pulse does not take into account the pulse shape. Another example is the space-scale factor D. It has been introduced as the characteristic size of the conductor’s surface but it ignores the proximity effect in problems involving more than one conductor. In other words, according to Equation 9.25, the pulse shape and mutual configuration of the conductors do not have a significant effect on the approximation error of the SIBCs. Consider again the example in Figure 9.2. We choose the duration t of the incident pulse to equal 2 ms so that the small parameter ~p is equal to 0.25, assuming D is the characteristic scale of the problem. To calculate the distributions of the surface current density for any instant of time using Equation 6.277, we need to define the shape of the current pulse (i.e., the function ~u0 ). Calculations were performed for the following two different pulses Ia(t) and Ib(t) to investigate the role of the pulse shape: ua (t); Ia (t) ¼ I0 (t=t)2 Ia (t) ¼ I0 ~ pffiffiffiffiffiffiffi ub (t); ~ ub ¼ t=t Ib (t) ¼ I0 ~
(9:26) (9:27)
Both pulses have the same duration, so that Ia (0) ¼ Ib (0);
Ia (t) ¼ Ib (t)
(9:28)
The final distributions of the surface current density at t ¼ t for both pulses are shown in Figures 9.6 through 9.11. Figures 9.6 through 9.8 show Ia (t) for L ¼ 2.5D, L ¼ 3D and L ¼ 8D respectively whereas Figures 9.9 through 9.11 shown Ib (t) for L ¼ 2.5D, L ¼ 3D and L ¼ 8D respectively. For comparison purposes, each figure also shows the ‘‘reference data’’ obtained using the boundary element formulation based on the time-dependent fundamental solution [1] without the use of SIBCs (the so-called original BEM) and a commercial finite element software. It is clearly seen that application of the PEC limit when the skin depth is not thin (~ p ¼ 0.25) causes an error of up to 25% (Figures 9.6 and 9.9). Use of the Leontovich approximation reduces the
Application and Experimental Validation of the SIBC Concept
25
PEC Leontovich Mitzner Rytov Original BEM FEM
20
K (A/m)
309
15
10
5 L = 2.5D 0
0
0.8
1.6
2.4
ξ/D FIGURE 9.6 Surface current density for Ia (t) ¼ I0 ~ua (t). The distance between the centers of the conductors is 2.5D.
18
PEC Leontovich Mitzner Rytov Original BEM FEM
K (A/m)
14
10
6 L = 3D 2
0
0.8
1.6 ξ/D
2.4
FIGURE 9.7 Surface current density for Ia (t) ¼ I0 ~ua (t). The distance between the centers of the conductors is 3D.
310
Surface Impedance Boundary Conditions: A Comprehensive Approach
10
PEC Leontovich Mitzner Rytov Original BEM FEM
K (A/m)
9
8
7
L = 8D
6
0.8
0
1.6
2.4
ξ/D FIGURE 9.8 Surface current density for Ia (t) ¼ I0 ~ua (t). The distance between the centers of the conductors is 8D.
25
PEC Leontovich Mitzner Rytov Original BEM FEM
K (A/m)
20
15
10
5 L = 2.5D 0
0
0.8
1.6
2.4
ξ/D FIGURE 9.9 Surface current density for Ib (t) ¼ I0 ~ub (t). The distance between the centers of the conductors is 2.5D.
Application and Experimental Validation of the SIBC Concept
18
PEC Leontovich Mitzner Rytov Original BEM FEM
14
K (A/m)
311
10
6 L = 3D 2
0
0.8
1.6
2.4
ξ/D FIGURE 9.10 Surface current density for Ib (t) ¼ I0 ~ub (t). The distance between the centers of the conductors is 3D.
10
PEC Leontovich Mitzner Rytov Original BEM FEM
K (A/m)
9
8
7
L = 8D
6 0
0.8
1.6
2.4
ξ/D FIGURE 9.11 Surface current density for Ib (t) ¼ I0 ~ub (t). The distance between the centers of the conductors is 8D.
312
Surface Impedance Boundary Conditions: A Comprehensive Approach
TABLE 9.1 Approximation Errors Calculated for the Problems Shown in Figures 9.6 through 9.11 and Given by Equation 9.25 Figure 9.6 26
Figure 9.7 17
Figure 9.8 4
Figure 9.9 34
Figure 9.10 23
Figure 9.11 6
Equation 9.25
Leontovich’s approximation (%)
11
3