An IEEE Press Classic Reissue
ANTENNA THEORY AND DESIGN Revised Edition
Robert S. Elliott University of California Los Angeles, California
IEEE .4ntennas & Propagation Society, Sponsor
IEEE PRESS
A JOHN WILEY & SONS, INC., PUBLICATION
Copyright
2003 by the Institute of Electronics & Electrical Engineers. All rights reserved
Published by John Wiley & Sons. Inc.. Hoboken. New Jersey. Published simultaneouslv In Canada. This book was previously published by: Pearson Education Inc., formerly known as PrenticeHall. Inc. No part of this publication may be reproduced. stored in a retrieval system or transmitted in any form or by any means. electronic, mechanical. photocopying, recording, scanning or otherwise. except as permitted under Section 107 or 108 of the 1976 United States Copyright Act. without either the prior written pemiission of the Publisher. or authorization through payment of the appropriate percopy fee to the Copyright Clearance Center. Inc.. 222 Rosewood Drive, Danvers. MA 01923. (978) 7508400. fax (978) 7504744. or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department. John Wiley & Sons, Inc.. I I I River Street. Hoboken. NJ 07030. (201 ) 748601 1. fax (201) 7486008. email:
[email protected] Limit of LiabilityIDisclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book. they make no representation or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives or written sales materials. The advice and strategies contained herein may not be suitable for your situation. You should consult with a professional where appropriate. Neither the publisher nor author shall be liable for any loss of profit or any other commercial damages, including but not limited to special. incidental. consequential. or other damages. For general information on our other products and services please contact our Customer Care Department within the U.S. at 8777622974, outside the U.S. at 3 175723993 or fax 3 175724002 Wiley also publishes its books in a variety of electronic formats. Some content that appears in print. however. may not be available in electronic format.
Library of Congress CataloginginPublication Data is available.
ISBN 0471 449962 Printed in the United States of America.
IEEE Press 445 Hoes Lane Piscataway, NJ 08854 IEEE Press Editorial Board Stamatios V. Kartalopoulos, Editor in Chief M. Akay J. B. Anderson R. J. Baker J. E. Brewer
M. E. ElHawary R. J. Herrick D. Kirk R. Leonardi M. S. Newman
M. Padgett W. D. Reeve S. Tewksbury G. Zobrist
Kenneth Moore, Director o f lEEE Press Catherine Faduska, Senior Acquisitions Editor IEEE Antennas and Propagation Society, Sponsor APS Liaison to IEEE Press. Robert Mailloux
To the memory of Tom Taylor
foreword to the revised edition
The purpose of the IEEE Press Series on Electromagnetic Wave Theory is to publish books of longterm archival significance in electromagnetics. Included are new titles as well as reprints and revisions of recognized classics. The book Antenna Theory and Design, by Robert S. Elliott is one such classic. In the case of antennas and Robert S. Elliott, I should like to be personal. Much of the material that forms the basis of Antenna Theory and Design I studied as a graduate student under Bob Elliott's guidance at UCLA in the late 1950's and early 1960's. This material became the fundamental background for me during my tenyear antenna design and development career at Hughes Aircraft Company and, what was then, North American Rockwell. The notes I compiled in his courses later became the foundation for two antenna courses when I nioved on to the University of Arizona in 1968. Antenna theory can be studied, assimilated, and then written down in a textbook with little practical experience. Antenna design is another matter entirely. Bob Elliott's career has been in actuality two careers in one. He has contributed significantly to antenna and microwave component design and development at Hughes Aircraft Company and Rantec Corporation while also forming and leading a strong, internationally recognized antenna and microwave program at UCLA. The book, Antenna Theory and Design, reflects the breadth and depth of coverage that such a background would suggest. As a result, the book is useful to academics and also to practitioners in industry and government laboratories. Professor Elliott has been an internationally wellknown contributor to electromagn e t i c ~for many years. He is universally regarded among his peers and students as an electromagnetic scholar. As an example, his clear and groundbreaking exposition on electromagnetics and its relationship to the special theory of relativity appears in the widelyregarded book, Electromagnetics; History, Theory, and Applications. This scholarly work was added to the IEEE Press Series on Electromagnetic Wave Theory in 1993. Professor Elliott is a Fellow of the IEEE (1961). Prior to his retirement from active teaching, he was the Hughes Distinguished Professor of Electromagnetics at UCLA. Among his teaching awards, he was elected Best Teacher, UCLA CampusWide ( 1 983) and has been elected Best Teacher, UCLA College of Engineering, four times. Among his many professional honors, he was elected a Fellow of the National Academy of Engineering (1988). The IEEiE Antennas and Propagation Society (APS) awarded him the APS
Foreword to the Revised Edition
Distinguished Achievement Award (1985). In addition, he has received two APS Prize Paper awards. In 2000, he was awarded an IEEE Third Millennium Medal. I have received many comments from Bob Elliott's colleagues and former students since beginning this reissue project. The one that most typifies this book is, "Many of the insights in his text are originally his and are still considered the fundamental way of looking at things." It is with pleasure that I welcome this classic book into the series. Donald G. Dudley University of Arizona Series Editor IEEE Press Series on Electromagnetic Wave Theory
preface to the revised edition
This textbook first appeared in 1981 an4 although it has been out of print for the past decade, a continuing demand has led to the decision that it be reissued. Like Electromagn e t i c ~its , predecessor in this Classic Series, it seems to have become something of a collector's item. The primary appeal is apparently due to the fundamental treatment of both theory and design for a wide variety of antenna elements, arrays, and feeding systems. The serious reader will find in this text the basic coverage of all aspects of the antenna discipline needed as background for someone desiring to pursue a career in electronic systems, or as preparation for advanced study leading to a desired career as an antenna engineer. The decision to reissue has provided an opportunity to eliminate errors that have been discovered in the final printing of the original text. Several colleagues and former students kindly contributed to the compiling of a list of errata. Most of the errors found were minor, a few were more serious, notably the entries in Tables 7.6 and 7.7. J.H.Anderson verified all suggested corrections and assembled the errata in a common format, thus facilitating their removal. His help is warmly acknowledged. The opportunity was also taken to update the references in Sections 5.14 and 8.13 because of seminal advances in the design of slot arrays and in the synthesis of shaped antenna patterns. The author wishes to thank the IEEE Press for its decision to reissue Antenna TheoI?,and Design and trusts that their faith in this project will not go unrewarded. Robert S. Elliott Los Angeles
contents
Foreword to the Revised Edition ix Preface to the Revised Edition xi Preface xix
I SOURCEFIELD RELATIONS SINGLE ANTENNA ELEMENTS 1 The FarField Integrals, Reciprocity, Directivity
1
3
Introduction 3 Electrostatics and Magnetostatics in Free Space 4 The Introduction of Dielectric, Magnetic, and Conductive Materials 7 TimeVarying Fields 10 The Retarded Potential Functions I1 Poynting's Theorem 13 The StrattonChu Solution 17 Conditions at Infinity 21 Field Values in the Excluded Regions 25 The Retarded Potential Functions: Reprise 26 The Far Field: Type I Antennas 27 The Schelkunoff Equivalence Principle 31 The Far Field: Type IL Antennas 36 The Reciprocity Theorem 39 Equivalence of the Transmitting and Receiving Patterns of an Antenna 41 1.16 Directivity and Gain 46 1.17 Receiving Cross Section 48 I . I 8 Polarization of the Electric Field 53 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 1.10 1. l l 1.12 1.13 1.14 1.15
2 Radiation Patterns of Dipoles, Loops, and Helices 58 2.1 Introduction 58 2.2 The CenterFed Dipole 58 2.3 Images in a Ground Plane 65 2.4 A Monopole Above a Ground Plane 67 2.5 A Dipole in Front of a Ground Plane 68
xiv
2.6 The Small Current Loop 69 2.7 Traveling Wave Current on a Loop 71 2.8 The EndFire Helix 73 3 Radiation Patterns of Horns, Slots and Patch Antennas 79
3.1 3.2 3.3 3.4 3.5 3.6 3.7
Introduction 79 The OpenEnded Waveguide 79 Radiation from Horns 83 CenterFed Slot in Large Ground Plane 86 WaveguideFed Slots 88 Theory of WaveguideFed Slot Radiators 91 Patch Antennas 99
II ARRAY ANALYSIS AND SYNTHESIS 111 4 Linear Arrays: Analysis 113 4.1 Introduction 113 4.2 Pattern Formulas for Arrays with Arbitrary Element Positions 114 4.3 Linear Arrays: Preliminaries I1 7 4.4 Schelkunoff's Unit Circle Representation 128 5 Linear Arrays: Synthesis 141
5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 5.10 5.1 1 5.12 5.13 5.14
Introduction 141 Sum and Difference Patterns 142 DolphChebyshev Synthesis of Sum Patterns 143 Sum Pattern Beamwidth of Linear Arrays 148 Peak Directivity of the Sum Pattern of a Linear Array 153 A Relation Between Beamwidth and Peak Directivity for Linear Arrays 157 Taylor Synthesis of Sum Patterns 157 Modified Taylor Patterns 162 Sum Patterns with Arbitrary Side Lobe Topography 165 Discretization of a Continuous Line Source Distribution 172 Bayliss Synthesis of Difference Patterns 181 Difference Patterns with Arbitrary Side Lobe Topography 185 Discretization Applied to Difference Patterns 187 Design of Linear Arrays to Produce NullFree Patterns 190
6 Planar Arrays: Analysis and Synthesis 196 6.1 Introduction 196 6.2 Rectangular Grid Arrays: Rectangular Boundary and Separable Distribution I9 7 6.3 Circular Taylor Patterns 213 6.4 Modified Circular Taylor Patterns: Ring Side Lobes of Individually Arbitrary Heights 218 6.5 Modified Circular Taylor Patterns: Undulating Ring Side Lobes 221 6.6 Sampling Generalized Taylor Distributions: Rectangular Grid Arrays 225
6.7 Sampling Generalized Taylor Distributions: Circular Grid Arrays 230 6.8 An Improved Discretizing Technique for Circular Grid Arrays 233 6.9 Rectangular Grid Arrays with Rectangular Boundaries: Nonseparable TsengCheng Distributions 237 6.10 A Discretizing Technique for Rectangular Grid Arrays 243 6.1 1 Circular Bayliss Patterns 250 6.12 Modified Circular Bayliss Patterns 256 6.13 The Discretizing Technique Applied to Planar Arrays Excited to Give a Difference Pattern 256 6.14 Comparative Performance of Separable and Nonseparable Excitations for Planar Apertures 261 6.15 Fourier Integral Representation of the Far Field 265
Ill SELFIMPEDANCE AND MUTUAL IMPEDANCE, FEEDING STRUCTURES 275 7 SelfImpedance and Mutual Impedance of Antenna Elements 277
7.1 7.2 7.3 7.4 7.5 7.6 7.7 7.8 7.9 7.10
Introduction 277 The Current Distribution on an Antenna: General Formulation 278 The Cylindrical Dipole: Arbitrary Cross Section 281 The Cylindrical Dipole: Circular Cross Section, Hallen's Formulation 284 The Method of Moments 286 Solution of Hallen's Integral Equation: Pulse Functions 287 Solution of Hallen's Integral Equation: Sinusoidal Basis Functions 294 SelfImpedance of CenterFed Cylindrical Dipoles: Induced EMF Method 297 SelfImpedance of CenterFed Cylindrical Dipoles: Storer's Variational Solution 305 SelfImpedance of CenterFed Cylindrical Dipoles: Zeroth and First Order Solutions to Hallen's Integral Equation 308
7.1 1 SelfImpedance of CenterFed Cylindrical Dipoles: KingMiddleton
SecondOrder Solution 314 7.12 SelfImpedance of CenterFed Strip Dipoles 321 7.13 The Derivation of a Formula for the Mutual Impedance Between Slender Dipoles 325 7.14 The Exact Field of a Dipole: Sinusoidal Current Distribution 329 7.15 Computation of the Mutual Impedance Between Slender Dipoles 332 7.16 The SelfAdmittance of CenterFed Slots in a Large Ground Plane: Booker's Relation 336 7.17 Arrays of CenterFed Slots in a Large Ground Plane: SelfAdmittance and Mutual Admittance 342 7.18 The SelfImpedance of a Patch Antenna 344 8 The Design of Feeding Structures for Antenna Elements and Arrays 351
8.1 8.2 8.3 8.4
Introduction 351 Design of a Coaxially Fed Monopole with Large Ground Plane 352 Design of a BalunFed Dipole Above a Large Ground Plane 355 TwoWireFed Slots: Open and CavityBacked 359
xvi
8.5 8.6 8.7 8.8 8.9 8.10 8.12 8.13 8.14 8.15 8.16 8.17
Coaxially Fed Helix Plus Ground Plane 361 The Design of an Endfire Dipole Array 363 YagiUda Type Dipole Arrays: Two Elements 368 YagiUda Type Dipole Arrays: Three or More Elements 373 FrequencyIndependent Antennas: LogPeriodic Arrays 375 Ground Plane Backed Linear Dipole Arrays 386 8.1 1 Ground Plane Backed Planar Dipole Arrays 390 The Design of a Scanning Array 393 The Design of WaveguideFed Slot Arrays: The Concept of Active Slot Admittance (Impedance) 397 Arrays of Longitudinal Shunt Slots in a Broad Wall of Rectangular The Basic Design Equations 402 Waveguides: The Design of Linear WaveguideFed Slot Arrays 407 The Design of Planar WaveguideFed Slot Arrays 414 Sum and Difference Patterns for WaveguideFed Slot Arrays; Mutual Coupling Included 418
IV CONTINUOUS APERTURE ANTENNAS 427
9 Traveling W a v e A n t e n n a s 429 9.1 9.2 9.3 9.4 9.5 9.6 9.7 9.8 9.9 9.10
Introduction 429 The Long Wire Antenna 430 Rhombic and VeeAntennas 432 DielectricClad Planar Conductors 437 Corrugated Planar Conductors 440 Surface Wave '~xcitation442 Surface Wave Antennas 446 Fast Wave Antennas 453 Trough Waveguide Antennas 464 Traveling Wave Arrays of QuasiResonant Discretely Spaced Slots [Main Beam at 8, = arccos(Plk)] 467 9.1 1 Traveling Wave Arrays of QuasiResonant Discretely Spaced Slots (Main Beam Near Broadside) 474 9.12 Frequency Scanned Arrays 476
10 Reflectors a n d Lenses 482 10.1 10.2 10.3 10.4 10.5 10.6 10.7
Introduction 482 Geometrical Optics: The Eikonal Equation 483 Simple Reflectors 490 Aperture Blockage 495 The Design of a Shaped Cylindrical Reflector 498 The Design of a Doubly Curved Reflector 504 Radiation Patterns of Reflector Antennas: The Aperture Field Method 508 10.8 Radiation Patterns of Reflector Antennas: The Current Distribution Method 518 10.9 Dual Shaped Reflector Systems 521
10.10 Single Surface Dielectric Lenses 525 10.1 1 Stepped Lenses 529 10.12 Surface Mismatch, Frequency Sensitivity, and Dielectric Loss for Lens Antennas 532 10.13 The Far Field of a Dielectric Lens Antenna 534 10.14 The Design of a Shaped Cylindrical Lens 536 10.15 Artificial Dielectrics: Discs and Strips 538 10.16 Artificial Dielectrics: Metal Plate (Constrained) Lenses 542 10.17 The Luneburg Lens 545 APPENDICES 557 A. Reduction of the Vector Green's Formula for E 559 B. The Wave Equations for A and D 562 C. Derivation of the Chebyshev Polynomials 564 D. A General Expansion of cosm v 567 E. Approximation to the Magneticvector Potential Function for Slender Dipoles 569 F. Diffraction by Plane Conducting Screens: Babinet's Principle 573 G. The FarField in Cylindrical Coordinates 581 H. The Utility of a Csc, 8 Pattern 585 Index 587
A ninemonth sequence in antenna theory and design is offered on a yearly basis at the author's institution. The first and second quarters are open to seniors and firstyear graduate students; the third quarter is at the graduate level. The sequence presupposes a background at the intermediate level in electromagnetic theory and a knowledge of introductory transmission line theory, including Smith charts and waveguide modal analysis. The present book has evolved from the lecture notes for the antenna sequence. It has been the author's experience, in teaching this sequence for the past five years, that the various topics which seemed to provide a balanced treatment were not to be found at an introductory level in a single textbook currently available. Further, some recent developments, the importance of which is widely recognized, were only available in the research literature. Student frustration over nonuniformity of notation from article to article and over the economic hardship associated with buying a multiplicity of texts that would only be partially used, provided the original motivation for the lecture notes. The editing of these notes by successive groups of students is appreciated, and it is hoped that their criticisms have benefited the final product. The topic coverage has been influenced by the author's experience and by the needs of local industry in the Los Angeles area. The reader will find emphasis on microwave antennas, particularly on arrays for use in radar and communication systems. The practical applications of such antennas have grown to occupy a major portion of the field, so it is hoped this emphasis will find wide appeal. However, other topics have not been neglected, as can be observed from the Table of Contents. The text is divided into four parts. Part I commences with a review of electromagnetic theory and then proceeds to the establishment of integral relations between a collection of sources (the antenna) and the radiated field caused by these sources. A convenient division of antennas into two types emerges from this development. The first type, for which the actual sources are known quite well, includes dipoles, loops, and helices, and their pattern characteristics are studied in turn. The second type, for
Preface
which the closein fields are known with reasonable accuracy, can be analyzed in terms of equivalent sources. This category includes horns, slots, and patches, all of which are considered in some detail. Part I1 is concerned with the analysis and synthesis of one and twodimensional arrays. The antenna elements studied in Part I form the constituent parts of these arrays, and focus is on the pattern characteristics. The synthesis procedures of Dolph and Taylor are introduced and extended to pattern requirements involving arbitrary side lobe topography. In Part 111 the emphasis is shifted to the impedance properties of antenna elements, used either singly or in arrays. Halltn's integral equation formulation of the selfimpedance of a cylindrical dipole is developed and extended to strip dipoles. Several types of solution are studied, including those obtained by the method of moments and by functional expansion. Babinet's principle is used to extend these results to slots. Mutual impedance, so important in the design of arrays, is formulated with the aid of the reciprocity theorem and then calculated for the most commonly used antenna elements. All of this information of selfimpedance and mutual impedance is then employed in the design of feeding structures for single elements and for linear and planar arrays, including those which scan. Part IV is devoted to antennas with continuous (or quasicontinuous) apertures. Long wire antennas such as the rhombic and V are studied and the properties of many surface wave structures are analyzed. These include slow wave types such as dielectricclad and corrugated ground planes and fast wave types, notably leaky waveguides. The book concludes with an introductory treatment of reflectors and lenses, antenna types to which many of the principles of optics can be applied. The three courses that form the antenna sequence at the author's institution span three months each, with four hours of lecture offered per week. The first course covers Chapters 1 , 2, and 4 plus the first six sections of Chapter 5, the first fifteen sections of Chapter 7, the first twelve sections of Chapter 8, and the first three sections of Chapter 9. It thus concentrates on wire antennas (dipoles, monopoles, loops, and helices) after introduction of the fundamentals. The second course covers Chapter 3, the remainder of Chapters 7, 8, and 9, and all of Chapter 10. It emphasizes aperture antennas (slots, patches, reflectors, and lenses). The third course is devoted to pattern synthesis and relies on the last half of Chapter 5 and all of Chapter 6, plus some of the current literature. For someone wishing to give a balanced offering of antenna topics in a one semester course, a combination which should prove satisfactory would contain Sections 1. l through 1.6, Sections 1.10 through 1.18, Sections 2.1 through 2.6, Sections 3.1 through 3.6, Sections 4.1 through 4.4, Sections 5.1 through 5.3, Section 7.8, Sections 7.13 through 7.15, Sections 8.1 through 8.6, Sections 10.1 through 10.5, and Sections 10.10 through 10.1 1. This would provide exposure to the fundamentals, to wire antennas, to aperture antennas, to the elements of array theory, to the problem'of feeding arrays in the presence of mutual coupling, and to the application of geometric optics to the design of reflector and lens antennas. Various friends have been kind enough to read portions of the manuscript and
offer their comments. The author wishes to acknowledge his indebtedness to Professors N. Alexopoulos, C. Butler, D. G. Dudley, G. Franceschetti, Y. T. Lo, C. T. Tai, and P. G. Uslenghi, and to his industrial colleagues J. Ajioka, V. GalindoIsrael, W. H. Kummer, and A. W. Love. Among the many students who have uncovered errors and assisted in modifications of the text, the efforts of D. Kim and J. Schaffner deserve explicit mention. A special and warm expression of gratitude is reserved for my longtime colleague and friend, Alvin Clavin, Manager of the Radar Laboratory at Hughes Canoga Park. He had the confidence to offer me consulting work at Hughes when I had been away from the field for a decade, thus rekindling my interest in the subject. This tribute extends to the entire Hughes organization, which has been so generous in supporting many of the antenna research efforts which have found their way into the pages of this book. My association with the engineering staff at Hughes has been rich and valuable, and particular gratitude must be expressed for the counsel of Louis Kurtz and George Stern. The computer assistance given me at Hughes by Ralph Johnson and Annette Sato is also gratefully acknowledged.
ROBERT S. ELLIOTT Los Angeles
sourcelf ield relations single antenna elements This initial part of the text, consisting of three chapters, is concerned first with establishing the general relations between a collection of sources (the antenna) and the radiated field produced by those sources (the farfield pattern). The source/field formulas are then used to deduce the pattern characteristics of the most commonly encountered antenna elements (dipole, loop, helix, horn, slot, and patch). These radiators will be seen to be ideally suited to many applications in which a single element will suffice. They have the added advantage of being useful in arrays, a subject which is discussed in Part 11.
1
the fa$neUd inu~n$gals~ ,,er ,hci,,
1 .IIntroduction
This chapter is concerned primarily with establishing formulas for the electromagnetic field vectors E and H in terms of all the sources causing these radiating fields, but at points far removed from the sources. The collection of sources is called an utltenna and the formulas to be derived form the basis for what is generally referred to as antenna pattern analysis and synthesis. A natural division into two types of antennas will emerge as the analysis develops. There are radiators, such as dipoles and helices, on which the current distribution can be hypothesized with good accuracy; for these, one set of formulas will prove useful. But there are other radiators, such as slots and horns, for which an estimation of the actual current distribution is exceedingly difficult, but for which the closein fields can be described quite accurately. In such cases it is possible to replace the actual sources, for purposes of field calculation, with equivalent sources that properly terminate the closein fields. This procedure leads to an alternate set of formulas, useful for antennas of this type. The chapter begins with a brief review of relevant electromagnetic theory, including an inductive establishment of the retarded potential functions. This is followed by a rigorous derivaticn of the StrattonChu integrals (based on a vector Green's theorem), which give the fields at any point within a volume V in terms of the sources within V and the field values on the surfaces S that bound V. This formulation possesses the virtue that it applies to either type of antenna, or to a hybrid mix of the two. Simplifications due to the remoteness of the field point from the antenna will lead to compact integral formulas, from which all the pattern characteristics of the different types of antennas can be deduced. A general derivation of the reciprocity theorem is presented; the result is used to demonstrate that the transmitting and receiving patterns of an antenna are identical. The concept of directivity of a radiation pattern is introduced and a connection is estab
The FarF~eldIntegrals. Reciproc~ty,Directivity
lished between the receiving cross section of an antenna and its directivity when transmitting. The chapter concludes with a discussion of the polarization of an antenna pattern. A. REVIEW O F RELEVANT ELECTROMAGNETIC THEORY'
It will generally be assumed that the reader of this text is already familiar with electromagnetic theory at the intermediate level and possesses a knowledge of basic transmission line analysis (including the use of Smith charts) and of waveguide modal representations. What follows in the next several sections is a brief review of the pertinent field theory, primarily for the purposes of introducing the notation that will be adopted and highlighting some useful analogies2 Throughout this text MKS rationalized units are used; the dimensions of the various source and field quantities introduced in the review are listed on the inside of the front cover. 1.2 Electrostatics and Magnetostatics in Free Space
A timeindependent charge distribution
expressed in couloumbs per cubic meter, placed in what is otherwise free space, gives rise to an electrostatic field E(s, y, z ) . Similarly, a timeindependent current distribution J(x, Y,z )
(I.lb)
expressed in amperes per square meter, produces a magnetostatic field B(x, y, z). T o heighten the analogies between electrostatics and magnetostatics, it is sometimes useful to refer to the "reduced" source distributions
in which 6 , is the permittivity of free space and p i ' is the reciprocal of the permeability of free space. Coulomb's law can be introduced as the experimental postulate for electrostatics and described by the equations ]The reader who prefers to omit this review should begin with Section 1.7. zThe pairing of B with E (and thus of H with D), the use of p i ' , the introd,~ctionof reduced sources, and the parallel numbering of the early equations in this review all serve to emphasize the analogies that occur between electrostatics and magnetostatics. This is done in the belief that perception of these analogies adds significantly to one's comprehension of the subject. See R. S. Elliott, "Some Useful Analogies in the Teaching of Electromagnetic Theory," IEEE Trans. on Education, E22 (1979), 710. Reprinted with permission.
1.2 Electrostatics and Magnetostatics in Free Space
5
(c,
in which R is the directed distance from the source point q , () to the field point (x,y, z), and F is the force on a charge q placed at (x, y, z), due to its interaction with the source system p ( t , q, 5). Similarly, the BiotSavart law can be introduced as the experimental postulate for magnetostatics and is represented by the equations
One can show by performing the indicated vector operations on (1.4a) that
VxE=O
(1.5a)
In like manner, the curl and divergence of (1.4b) yield
Equations 1.5 are Maxwell's equations for static fields. Integration of (1.5b) and use of the divergence theorem gives Gauss' law, that is,
js
E dS
=
(5)
dV
=
total reduced charge enclosed
(1.6a)
Similarly, integration of (1.5~)and use of Stokes' theorem yields Ampere's circuital law: fc
B
dl
=
Is(5)
dS
= total
reduced current enclosed
(1.6b)
In like manner, integration of (1.5a) and (1.5d), followed by the application of Stokes' theorem or the divergence theorem results in the following relations.
The FarField Integrals, Reciprocity, Directivity
From (1.7a) it can be concluded that E (x, y, z) is a conservative field and that $ E dl between any two points is independent of the path. Equation 1.7b permits the conclusion that the flux lines of B are everywhere continuous. Equation 1.4a can be manipulated into the form
in which
is the electrostatic potential function. In like manner, Equation 1.4b can be rewritten in the form
where
is the magnetostatic vector potential function. One can see that the reduced sources (1.2) play analogous roles in the integrands of the potential functions (1.8a) and (1.8b), as well as in the integrands of the field functions (1.4a) and (1.4b). There is no compelling reason to introduce either D or H until a discussion of dielectric and magnetic materials is undertaken, but if one wishes to do it at this earlier stage, where only primary sources in what is otherwise free space are being assumed, then it is suggestive to write
with the subscripts on D and H denoting that the medium is free space. Then it follows logically from (1.5) that
V*Do=p
VXH,=J
and from (1.6) that
f Do f Ho
dS
=
C
dl
=
5 p dV 5 J dS
=
v
S
+
total charge enclosed
=
total current enclosed
(1.12a) (1.12b)
S
Equations 1.12 are the forms in which one is more apt to find Gauss' law and Ampere's circuital law expressed. It is apparent from (1.12) that Do and H, play analogous roles in the two laws.
1.3 The lntroduct~onof D ~ e i e c t r ~ Magnet~c. c, and Conductive Mater~als
7
When flux maps are introduced, (1.12a) leads to the conclusion that the lines of Do start on positive charge and end on negative charge. If one chooses to defer the introduction of D and H until materials are present, a flux map interpretation of (1.6a) includes the idea that the lines of E start on reduced positive charge and end on reduced negative charge. It has already been noted in connection with equation (1.7b) that the flux lines of B are continuous. Since H, differs from B only by a multiplicative constant, the flux lines of H, are also continuous. 1.3 The Introduction of Dielectric, Magnetic, and Conductive Materials
The electrostatic behavior of dielectric materials can be explained quite satisfactorily by imagining the dielectric to be composed of many dipole moments of the type p = l,qd, in which q is the positive charge of the oppositely charged pair, d is their separation, and 1, is a unit vector drawn from q to +q. If P(x, y, z ) is the volume density of these elementary dipole moments, one can show3 that their aggregated effect is to cause an electrostatic field given by
with S the dielectric surface and V its volume. In (1.13a), Vs operates on the source point and V, operates on the field point. Similarly, the magnetostatic behavior of magnetic materials can be explained in terms of a collection of current loops with magnetic moments of the type m = l,na21, where nu2 is the area of the loop, I is the current, and 1, is a unit vector normal to the plane of the loop in the righthand sense. If M(x, y, z) is the volume density of these elementary loops, one can show4 that their aggregated effect is to cause a magnetostatic field given by
In the more general situation that there is a primary charge distribution p(x, y, z) somewhere in space and secondary (or bound) charge distributions P,,on the dielectric surface and  V P throughout its volume, the total electrostatic field is E = El E,, with E, given by (1.4a) and E, given by (l.13a). No additional information would be conveyed by using Do = E,,E in this situation. However, it is extremely usefulS to
.
+
3See, for example, R. S. Elliott, Electromagnetics (New York: McGrawHill Book Co., Inc., 1966), pp. 33037. dElliott, Electromagnetics, pp. 4047. slbid., pp. 33940.
The FarF~eldIntegrals, Reciprocity. D i r e c t ~ v ~ t y
generalize the concept of D through the defining relation
This insures the desirable feature that
at all points in space (both within and outside the dielectric), thus permitting the assertion that the flux lines of D start and stop on primary charge alone. If there are no primary charges inside the dielectric, the D lines are continuous there. Outside the dielectric, (1.14a) reduces to D = e , E , which is consistent with (1.10a). Since V x E = V x El V x E,, and since El and E, are both expressible as the gradient of a scalar function, it follows that in this more general situation of primary and secondary charge distributions,
+
However, one can see from the defining relation (1.14a) that V x D = V X P and thus the generalized D, unlike E, may not be an irrotational field everywhere. Many dielectric materials are linear (or nearly so), in the sense that P = x,E,E holds, where X , is a constant called the dielectric susceptibility. When this can be assumed, Equation 1.14a reduces to
+
where e is the permittivity of the dielectric medium. The quantity 6 / 6 , = 1 X , is more useful and is known as the relative permittivity, or dielectric constant. Similarly, in the more general situation that there is a primary current distribntion J(x, y, z) somewhere in space and secondary (or bound) current distributions M x 1, on the surface of the magnetic material and V x M throughout its volume, the total magnetostatic field is B = B, B,, with B1 given by (1.4b) and B, given by (1.13b). No additional information would be conveyed by using H, = p,'B in this situation. However, it is extremely useful6 to generalize the concept of H through the defining relation (1.14b) H = pilB  M
+
This insures the desirable feature that
6op. cit., Elliott, Electrornagnetics, pp. 40810.
9
1.3 The lntroductlon of Dielectr~c.Magnetic, and Conduct~veMaterials
at all points in space (both within and outside the magnetic material) thus permitting the assertion that H is irrotational except at points occupied by primary sources. Since V B = V B, V B,, and since B, and B, can both be expressed as the curl of a vector function, it follows that in this more general situation of primary and secondary current distributions,
+
However, one can see from the defining relation (Equation 1.14b) that V H = V M, and thus the generalized H, unlike B, may have discontinuous flux lines. Most magnetic materials are nonlinear, but in the exceptional case that linearity can be assumed, M is linearly proportional to B and Equation 1.14b reduces to
in which X, is the magnetic susceptibility and p is the permeability of the magnetic material. Equations 1.15 are Maxwell's equations for static fields when dielectric and magnetic materials are present. They are supplemented by Equations 1.14, one of which links E, D, and the secondary sources P, with the other linking B, H, and the secondary sources M. The integral forms of (1.15a) and (1.15~)lead to fsD fc H
dS
=
primary charge enclosed
dl = primary current enclosed
(1.17a) (1.17b)
Thus the generalized D and H satisfy Gauss' law and Ampere's circuital law, respzctively, in terms of the primary sources alone. This is their principal utility. On the other hand, E and B enter into a calculation of the force on a charge q moving through the field. In the most general static source situation (primary and secondary charge and current distributions), Equations 1.3, 1.4, and 1.13 combine to give
which is the Lorentz force law. When conductive materials are present and Ohm's law is applicable,
at points occupied by the conductor, with o the conductivity of the material.7 70p. cit., Elliott, Electromagnetics, pp. 47381.
1.4 TimeVarying Fields
If the sources become timevarying, represented by p(x, y, z, t) coulombs per cubic meter
(1.20a)
J(x, y, z, t) amperes per square meter
(1.20b)
and are assumed t o exist in otherwise empty space, then Equations 1.5 need to be generalized. Faraday's EMF law and the continuity equation linking charge and current lead t o the result that
in which c is the speed of light and E(x, y, z , t ) and B(x, y, z, t ) are now functions of
time as well as space. Equations 1.21 are Maxwell's equations in their most general form for primary sources in empty space. If one uses (1.10) and the fact that p o ~ , c = Z 1, these equations convert readily to the more familiar set
If dielectric, magnetic, and conductive materials are present and are represented by timevarying dipole moments, current loops, and drifting electron clouds, respectively, if the defining relations in (1.14a) and (1.14b) are extended t o apply when the fields and secondary sources are timevarying, and if Ohm's law (1.19) is still valid in the timevarying case (and all of these are good assumptions in practical situations), then Maxwell's equations becomes
sop. cit., Elliott, Electromagnerics, pp. 39394, 464, 509.
11
1.5 The Retarded Potential Functions
where now D and H have their generalized meanings, as given in the supporting Equations 1.14, and J is linked to E by (1.19) at all points occupied by conductor. 1.5 The Retarded Potential Functions
In antenna problems, one desires to find the field values at a point in terms of all the timevarying sources that contribute to the fields. This implies an integration of (1.22) or (1.23), a relatively difficult undertaking that will be deferred until Section 1.7. A simpler but less rigorous approach will be followed in this section, in which E and B are not found directly, but are found instead through the intermediation of potential functions whose relations to the sources are obtained intuitively. Let the timevarying sources be given by (1.20) and be assumed to exist in a finite volume V in otherwise empty space. Then Maxwell's equations in the form (1.21) are point relations that connect E(x, y, z, t) and B(x, y, z , t) to the sources. Since V B = 0, it is permissible to introduce a new vector function A(x, y, z , t ) by the defining equation B=VXA
(1.24)
Because the divergence of the curl of any vector function is identically zero, it is apparent that (1.24) automatically satisfies (1.21d). If (1.24) is inserted in (1.21a), one obtains
where the dot over A implies timedifferentiation. Since the curl of the gradient of any scalar function is identically zero, the most general solution to (1.25) results from the introduction of a new scalar function @(x, y, z, t ) such that
Equation 1.26 not only satisfies (1.21 a) but, taken in conjunction with (1.24), provides a solution for E and B if the newly introduced functions A and Q, can be related to the sources. This can be done by forcing (I .24) and (1.26) to satisfy the two remaining Maxwell equations, that is, (1.21b) and (1.21c), notably the equations containing the sources. If (1.24) and (1.26) are used in (1.21), the result is that
Equation 1.27 is a hybrid secondorder differential equation (hybrid in the sense that it contains both A and @) and as a consequence would be extremely difficult to solve.
The FarField Integrals. Reciprocity, Directiv~ty
Fortunately, a simplification is possible because, up to this point, only the curl of A has been specified, and a vector function is not completely defined until some specification is also placed on its divergence. It is convenient in this development to choose
for then (1.27) reduces to
Equation 1.29 is an inhomogeneous secondorder differential equation in the unknown function A, with the negative of the reduced current distribution (which is assumed to be known) playing the role of driving function. It is variously called the Helmholtz equation or the wave equation, the latter name arising because the solutions to (1.29) away from the sources are waves that travel at the speed of light. The task remains to insure that (1.24) and (1.26) satisfy the remaining Maxwell equation (1.21b). Substitution gives
This is also a hybrid differential equation, but use of (1.28) converts it to
Thus A and cD satisfy the same differential equation, the only difference being the driving function; in (1.30) it is the negative of the reduced charge distribution (which is assumed to be known) which appears and governs 0. The development has now reached the point that if (1.29) and (1.30) can be solved for A and cD, then (1.24) and (1.26) can be used to determine E and B, and the goal will have been achieved. A solution of (1.30) can be inferred from the limiting electrostatic case. If the p(x, y, z), then (1.25) and (1.30) sources cease to vary with time so that p(x, y, z, t ) reduce to E =
[email protected] (1.31)


in which cD is now a timeinvariant function, that is, Q(x, y, z, t ) @(x,y, z). But if one returns to Section 1.2, it can be observed that (1.8a) and (1.31) are identical. Further, if the divergence of (1.8a) is taken and the result is combined with (1.5b), Equation 1.32 is reproduced, and its solution must be (1.9a), namely,
1.6 Poynt~ng'sTheorem
13
Thus the limiting (timeinvariant) solution to (1.30) is (1.33). How can this be used to deduce the general (timevariant) solution to (1.30) ? It can be argued that a change in the charge density at a source point (5, q, c) causes a distrubance which is not immediately felt at a field point (x, y, z), since that disturbance, traveling at the speed of light, must take a time interval Rlc to traverse the intervening distance R. Thus if one wishes to find the value of 45 at the point (x, y, z) at the time t, that is, @(x,y, z, t), one should use the charge densities a t the source points (5, q, c) at the earlier times t  (Rlc). This suggests that a solution to (1.30) might be
This is admittedly a highly intuitive argument, and a rigorous solution to this problem will be presented in the development beginning in Section 1.7. However, if (1.34) is inserted in (1.30), one finds that it is indeed a solution. By a similar argument it can be inferred that
Equations 1.34 and 1.35 are known as retardedpotential functions because of the use of retarded time in the integrands. In conformance with the names already given to their limiting forms in electrostatics and magnetostatics, 45 is called the electric scalar potential function and A is called the magnetic vector potential function. 1.6 Poynting's Theorem
One of the most useful theorems in electromagnetics concerns the power balance in a timevarying electromagnetic field. To introduce this theorem, let it be assumed that there is a system of impressed sources Ji that produces an electromagnetic field E', Bi, and that this impressed field causes a response system9 of currents Jr to flow, creating an additional field Er, Br. If all these sources are in otherwise free space, the impressed and response fields both satisfy Maxwell's equations in the form (1.21). The total current density and field at any point are therefore
9The decomposition of the total current system into impressed and response current densities is arbitrary, but often forms a natural division. For example, the currents that flow in a dipole may be considered to be a response to the impressed currents that flow in the generator and transmission line feeding the dipole.
T h e F a r  F ~ e l dIntegrals, Reciproclty. Directivity
If power is being supplied to the field, it must be at the ratet0
But from Maxwell's equations (1.22), J i = V x H ,   2dD Jr at
so that d 3 P = [E
. V x H, ;
Application of the vector identity
coupled with the use of (1.22) gives
As a consequence, (1.36) may b: rewritten as
This result gives the power balance in a volume element dV. The left side of (1.37) is the instantaneous power being supplied by the impressed sourczs to d V . The factor
is the time rate of change of density of stored energy. ' The factor E Jr represents the power density being absorbed from the field by the response current density J r . If, for example, the response current is flowing in a conductor, this term accounts for ohmic loss. Alternatively, if Jr is due t o freely moving charges, E Jr accounts for their change in kinetic energy. When the law of conservation of energy is invoked, it follows that the term V (E x H,) may be interpreted as the volume density of power leaving dV. This conclusion can be seen from another point of view by integrating (1.37). With the aid of the divergence theorem, one is able to write
loop. cit., Elliott, Electrottlagnetics, p. 283. 1
lop. rit., Elliott, Electronlngnetics,pp. 19395, 28384.
15
1.6 Ponyting's Theorem
The left side of (1.38) represents the entire instantaneous power being supplied bji all the sources. The first integral on the right side of this equation accounts for the time rate of change of the entire stored energy of the field. The second integral stands for the power being absorbed by the system of response currents. The last integral therefore represents the entire instantaneous power flow outward adross the surface S bounding the volume V. For this reason, one may define the Poynting vector as
and place upon it the interpretation that it gives in magnitude and direction the instantaneous rate of energy flow per unit area at a point. This is Poynting's theorem. Since the units of E and H , are volts per meter and amperes per meter, respectively, it is seen that the units of 6 are watts per square meter. Cases in which the currents and fields are varying harmonically in time occur so frequently and have such importance as to deserve special discussion. Expressing all quantities in the form of a complex spatial vector function multiplied by ejw', such as
one may write 6 = E X H, = $(E X =
z
+(Eej"' + E*ej"l)
X
(Xoej"' + XfC;fejw')
+ E* X X O )+ $(E X 3CfC,ej2"'+ E* x H t ) + :@e(E x H,)
3C8
:@e(E
X
3Ctej2"')
(1.40)
The term + @ e ( E x H r ) is independent of time and thus represents the timeaverage value of 6, giving
@ =
[email protected](E x H t )
(1.41)
The term & @ e ( E x H , ) contains the factor ej2"' and thus represents the oscillating portion of Poynting's vector. Therefore 6 may be interpreted at a point as consisting of a steady flow of energy density plus a flow which surges back and forth at frequency 20. Similarly 1~ ~2 z 0
:=

.
hEOE E +C0E E*
= +e0[+(Eej"'
+ E*ej"')
(Eejw' + E*ej"')]
+ + c 0 @ e ( E .E )
and + , y i l ~ Z,= ;i,yilB 1 B*
[email protected](B
B)
The terms + E , E E* and + p ; ' B B* are independent of time and reprzsent the timzaverage stored energies; their time derivatives are zero. The terms +t,@e(E E ) and  $ p i l @ e ( B B) oscillate at a frequency 2 0 and they represent the variablecomponents of the stored energy. Finally, E
Jr = * @ e E
Jr*
[email protected] Jr
16
The FarFleld Integrals. Reciprocity, Directivity
Here again, the term $(ReE Jr* represents the timeaverage power density being absorbed by the response currents; the term E Jr oscillates at a frequency 2 0 and represents the energy density being cyclically absorbed and released by the response currents. With this formulation, Equation 1.38 may be rewritten in two parts. The timeaverage power balance is seen to be
while the timevariable part, oscillating at a frequency 20, may be written
Thus, on the time average, the sources supply power only to that component of the response currents in phase with the electric field, represented by the first integral in (1.42), and to the net energy flow out of the volume V across the surface S. In addition, the sources may have to furnish energy and take it back at the cyclic rate 2 0 if the right side of (1.43) is not zero. However, in many practical circumstances, the individual integrals in (1.43) may not be in phase, but may be adjusted purposely so that they cancel each other, thus "matching" the generator. B. INTEGRAL SOLUTIONS OF MAXWELL'S EQUATIONS I N T E R M S O F THE SOURCES
The next four sections and two related appendices are devoted to a rigorous solution of Maxwell's equations in integral form, giving the fields at any point within a volume V in terms of the sources within V and the field values on the surfaces S that bound V. One advantage to this development, beyond its rigor, is that the results are in a perfect form to delineate approaches to the two types of antennas mentioned in the introduction, namely those on which the current distribution is known quite well (such as dipoles and helices), and those for which the closein fields are known quite well (such as slots and horns). Another advantage of the development is that it delivers the retarded potential functions as an exact consequence of the central result^.'^ lzSome authors, in contradistinction to using the StrattonChu formulation (which gives E and B directly as integrals involving the sources), prefer to present a rigorous proof that the retarded potential functions A and @ are given by the integrals shown in (1.34) and (1.35). Then E and B follow from (1.24) and (1.26). That approach is comparable in complexity to the StrattonChu development, and suffers from the ultimate disadvantage of requiring an ad hoc introduction of fictitious magnetic sollrces without rigorous validation. The concept of fictitious magnetic sources arises naturally from the StrattonChu solution, and their results provide a sound basis for Schelkunoff's equivalence principle. See Section 1.12.
17
1.7 The StrattonChu Solut~on
However, the reader who is not interested in delving into the complexities of this development, and who is satisfied with the intuitive introduction of the retarded potential functions given in Section 1.5, may wish t o move directly to Section 1.11. This can be done without any loss of continuity. 1.7 The StrattonChu Solution
Since Maxwell's equations are linear in free space, no loss in generality results from assuming that time variations are harmonic and represented by ej"'. The angular frequency o may be a component of a Fourier series or a Fourier integral, thus bringing arbitrary time dependence within the purview of the following analysis. Accordingly, iff ( x , y, z, t ) is any field component or source component, it will be assumed that f ( x , y, z, t ) =: f ( x , y, z)ejwr. Further, it will be assumed that all of the sources are in what is otherwise free space. This does not preclude the presence of a dielectric material if it is represented by a P dipole moment distribution, nor the presence of a magnetic material if it is represented by an M magnetic moment distribution, nor the presenc: of a metallic conductor if it is viewed as consisting of a positive ion lattice and an electron cloud, coexisting in free space. With dielectric or magnetic materials present, P = J, and J, = V X M are the bound current density contributions to the total current density J. In the case of the metallic conductor, the electrostatic fields of the lattice and cloud are assumed to cancel each other, thermal motions are assumed to be random with a null sum, and only the oscillatory motion of the electron cloud is germane, making a contribution aE to the total current density J, with a the conductivity of the metal. All of these assumptions concerning the representation of electrical behavior of materials are valid in the practical realm of the actual materials used to construct most antennas. For this reason the ensuing analysis has wide applicability. Maxwell's equations (1.21), for timeharmonic sources in otherwise free space, can be written in the form
Since cZp,eo = 1, the result if the divergence of the second of these equations is taken is the continuity relation
V
J
=
jop
(1.45)
In all five of the above equations, the time factor ejor is suppressed and the fields are complex vector functions, as is the current density. The charge density is a complex scalar function.
The FarF~eldIntegrals, Reciprocity. D i r e c t ~ v ~ t y
If the curl of either (1.44a) or (1.44b) is taken and then (1.44b) or.(1.44a) is used to eliminate E or B, one obtains the vector wave equations
in which k = w/c is called the propagation constant, for a reason that will emerge shortly. These last two equations can be integrated through use of a technique first introduced by Stratton and Chu, and based on a vector formulation of Green's second identity. Consider a region V, bounded by the surfaces S , . . S,, as shown in Figure 1.1. Let F and G be two vector functions of position in this region, each continuous and having continuous first and second derivatives everywhere within V and on the boundary surfaces Si. Using the vector identity
FIG. 1.1 Notation for Vector Green's Theorem. 1 3 5 . A. Stratton and L. J. Chu, "Diffraction Theory of Electromagnetic Waves," Phys. Rev., 56 (1939), 99107. Also, see the excellent treatment in S. Silver, Microwave Antenna Theory and Design, MIT Rad. Lab. Series, Vol. 12 (New York: McGrawHill Book Co., Inc., 1939), pp. 809. The present development is a reproduction, with permission, of what appears in R. S. Elliott, Electromagnetics (New York: McGrawHill Book Co., Inc., 1966), pp. 27280 and 5348, and differs from Silver's treatment principally in the nonuse of fictitious magnetic currents and charges.
1.7 The StrattonChu Solut~on
and letting A
If A
=
=
G and B
F while B
==
=
V x G, one obtains
V x F, then
When the difference in these results is integrated over the volume V, one obtains
If 1, is chosen to be the inwarddrawn unit normal vector from any boundary surface Siinto the volume V, use of the divergence theorem gives
This result is the vector Green's theorem. Suppose that the fields E and B of (1.46) and (1.47) both meet the conditions required of the function F in V ; let G be the vector Green's function defined by
in which a is an arbitrary constant vector and R is the distance from a n arbitrary point P ( x , y, z ) within V to any point (5, q, within V or on Si. As defined by (1.49), G satisfies the conditions of the vector Green's theorem everywhere except at P. Therefore, one can surround P by a sphere C of radius 6 and consider that portion V' of V bounded by the surfaces S , . . . S,, C. Letting E = F, one finds that
c)
j
V'
(E V, x
vs x
ya

ya
V, x V, x E) d~ (1.50)
I
. SN,x
( y a x V, x E

E x V, x y a )
c),
1, d S
in which, since y is a function of (x, y, z ) as well as (l,q, it is necessary to distinguish between differentiation with respect to these two sets of variables by subscripting
The FarField Integrals, Reciprocity. D~rectivity
the operators so that
and
It is shown in Appendix A that both sides of this equation may be transformed so that a is brought outside the integral signs, with the following result:
Since a is arbitrary, it follows that the integrals on the two sides of the above equation can be equated, yielding
+ (1,
x E) x Vsy
 jwy(1,
x B)] dS
(1.52)
where, for convenience, the surface integral over the sphere C is displayed separately. It is further shown in Appendix A that the right side of (1.52) reaches the limit 4nE(x, y, z), with (x, y, z ) the coordinates of the point P, as C shrinks to zero. Therefore the limiting value of (1.52) is
This important formula gives E at any point in the volume V in terms of the sources within V plus the field values on the surfaces that bound V. By letting B = F, one may proceed in a similar fashion to deduce a companion formula for B(x, y, z). Alternatively, the curl of (1.53) may be taken and then (1.44a) used to obtain B. By either procedure, one finds that
Equations 1.53 and I .54 comprise a solution of Maxwell's equations in terms of the timeharmonic charge and current sources within V and the field values on the boundary surfaces Si. 1.8 Conditions at Infinity
Let it now be assumed that the surface S , of Figure 1.1 becomes a large sphere of radius (TI centered at the point P. Initially, (TI will be taken great enough to enclose all the sources J and p of the fields; ultimately will be permitted to become infinitely large. Under these circumstances, consider the contributions to (1.53) and (1.54) of the surface integrals over S,. If 1, is a unit vector directed outward along the radius of the spherical surface S,, so that 1, = I,, one may write for the appropriate part of (1.54) [*(I. SN
C
x E)
+ (I, x B)
X
Vsy
+ (1.
B) VSV] d s
Similarly, the appropriate part of (1.53) becomes

If (TI W , since the surface of the sphere increaszs as (TI2, the surfacz integral in (1.55) will vanish if (1.57) lim (TIB is finite a+
Similarly, the surface integral (1.56) will vanish if lim (RE is finite m
.
The FarField Integrals. Reclproc~ty,Directivity
Relations 1.57 through 1.60 are known as the Sommerfeld conditions at injnity. Expressions (1.57) and (1.59) are commonly called the jniteness conditions (Endlichkeit Bedingungen) and Expressions 1.58 and 1.60 are customarily called radiation conditions (Ausstrahlung Bedingungen). The finiteness conditions require that E and B diminish as @  I , while the radiation conditions require that they bear the relation to each other found in wave propagation in regions remote from the sources. (See Section 1.1 1 .) It is now possible to demonstrate the extremely important result that real sources, confined to a finite volume, always give rise to fields that satisfy the Sommerfeld conditions. T o see this, consider Equations 1.53 and 1.54 when the only boundary surface is the large sphere S,, with radius that will be permitted to become infinitely large. It shall be assumed that the real sources J and p are finite and confined to a finite volume V,. With the surface S, becoming an infinite sphere, the volume V in (1.53) and (1.54) also becomes infinite, but no convergence difficulties arise with the volume integrals because the sources are all within V,. If one borrows from the results of Section 1.6, the fields over S, will consist of outgoing waves with power density E x H, watts per square meter. S i n e the surface area of S, is increasing as ( R 2 , if there is even the most minute loss in V , the law of conservation of energy requires that E and H, diminish more rapidly than @  I , and thus Conditions 1.571.60 are satisfied. One can then conclude that in an unbounded region, B(x, y, z ) and E(x, y, z ) are given solely by the volume integrals that appear in (1.53) and (1.54). A check on this conclusion for the limiting case of no loss in V may be obtained through an ordering of the terms that comprise the volume integrals. T o see this, assume that there are n o bounding surfaces except the infinite sphere S,, and that the surface integrals involving S, in (1.53) and (1.54) are zero. Then, for this situation, Equations 1.53 and 1.54 reduce to
where the second version of the integrand in (1.61) has been achieved with the aid of the continuity equation (1.45). It can now be ascertained whether or not E and B, when computed from (1.61) and (1.62), satisfy Sommerfeld's conditions at infinity. Let an arbitrary point in V , be selected as the origin and let r be the vector drawn from the origin to the field point P(x, y, 2 ) ; the vector drawn from the source element to P will be labeled R. Then (J
.V,) V,yl
= (J
[ ( +
Vs) 1, jk
YkR1
1.8 Conditions a t l n f ~ n ~ t y
in which spherical coordinates (r, 8', 4') centered a t P have been used and
Performing the indicated differentiations, one obtains
The functions yl, V,y, and ( J Vs)Vsyl are all seen to involve polynomials in the variable RI. Retain for the moment only firstorder terms; then substitution in (1.61) and (1.62) gives
But R
+
[(x  O2 ( y  q)' = [(r sin 8 cos 4 =
+
(Z

c)2]112
o2+ (r sin 8 sin 4

11)'
+ (r cos 8 
()2]1/2
in which now conventional spherical coordinates (r, 8, 4) centered at the origin have been introduced. As P becomes remote, R can be expressed in the rapidly converging series
R
=
r
(c sin 8 cos 4 +
sin 8 sin 4
+
C O 8) ~
+ O(r')
(1.65)
Similarly, R1
r  ~ + O(r2)
lim l R= 1. ,*
and thus as r becomes very large, Equations 1.63 and 1.64 may be written
+
in which 2 == 5 sin 8 cos 4 $ q sin 8 sin $ ( cos 8. If one were to go back and include all the terms in the expressions for Vsyl and (J Vs)V,yl, they would alter the results in (1.66) and (1.67) only a t the level ofO(r'). Therefore these two expressions for B and E may be taken as exact. In considering Expressions 1.66 and 1.67 with respect to the Sommerfeld conditions, one notices that the terms of O(r2) and below satisfy all four conditions and
The FarField Integrals, R e c ~ p r o c ~ tD~rectivity y.
thus concern may be focused on the explicit firstorder terms. But lirn r B I
jk lirn =4n r
ejk'
and, since the volume integral is a function of the source coordinates and the angular direction to P, but not of r, this limit is finite. A similar argument establishes that lim r E is also finite and thus both finiteness conditions are satisfied. 1
Further,
The integrand in (1.69) is identically zero and therefore Condition 1.60 is satisfied. In like manner, Condition 1.58 is also found to be satisfied. This supports the argument that any system of real sources confined to a finite volume V o gives rise to an electromagnetic field at infinity that satisfies Sommerfeld's conditions, that the surface integral over an infinite sphere S, gives a null contribution, and that in an unbounded region the electromagnetic field at any point P, near or remote, is given precisely by (1.61) and (1.62). Suppose now that parts of the volume V o are excluded from V by the finite, regular closed surfaces S , . . . Si . . . . These surfaces may exclude some of the sources from V or not, but their presence does not alter the results at infinity. However, now the more general expressions in (1.53) and (1.54) apply, and one may conclude by saying that these expressions are valid even if the volume V is infinite, so long as real sources in a finite volume are assumed. If the volume V is infinite, the surface at infinity need not be considered. This solution for E and B, given by Equations 1.53 and 1.54, is in a form that is convenient for the purpose of drawing a distinction between two types of radiators. Type I antennas will be taken to be those for which the actual current distribution is known quite well, such as dipoles and helices. Type I1 antennas will be those that have actual current distributions which would be difficult to deduce, but which could be enclosed by a surface over which the fields are known with reasonable accuracy. These include horns and slots. For type I antennas, there will be no volumeexcluding surfaces and (1.53) and (1.54) will contain only volume integrals. For type I1 antennas, the volumeexcluding surfaces (usually only one) will be chosen to surround all the actual sources so that there are none to be found in the remaining part of space V. Thus for type I1 antennas, (1.53) and (1.54) will contain only surface integrals. In the developments that follow later in this chapter, it will be seen that it is useful to replace the field values occurring in the integrands of these surface integrals by equivalent sources. Thus for the remainder of this book, type I radiators will be referred to as actualsource antennas and type I1 radiators will be called equivalentsource antennas.
1.9 Field Values in the Excluded Regions
Because of its bearing on the analysis of type I1 (equivalentsource) antennas, it is important to consider the values of the fields E and B at points inside the excluding surfaces shown in Figure 1.1. In particular, let the field point (x, y, z ) lie anywhere in the volume V , which has been surrounded by the closed surface S,. A simple application of the general results in (1.53) and (1.54) gives
Another way to view this situation is to imagine that V , is the volume region comprising the collection of field points and that S , is the sole surface, performing the function of excluding all the rest of space. From this viewpoint, a second application of the general results in (1.53) and ( I .54) yields
The negative signs in front of the surface integrals in (1.72) and (1.73) are occasioned by the fact that now the normal to the surface S, is oppositely directed. If the difference between these two sets of formulas for the fields within V , is formed, one obtains
The FarField Integrals, Reciprocity. Directivity
The right sides of (1.74) and (1.75) are seen to be exactly the same as the right sides of (1.53) and (1.54).Therefore one can conclude that if the range of the field point ( x , y , z ) is unrestricted, when (x, y, z ) lies within V , Equations 1.53 and 1.54 will give the true fields E and B. However, when ( x , y, z ) lies outside V , Equations 1.53 and 1.54 will give a nu!l result. 1.10 The Retarded Potential Functions: Reprise
If the volume V is totally unbounded, Equations 1.53 and 1.54 give
Since V,y/ =  V,y, and since J and the limits of integration are functions of (C, q, 0, but not of ( x , y, z ) , these integrals may be written in the forms
Therefore it is convenient to introduce two potential functions by the defining relations
@(x,y , z , t ) =
J" P(&
dV
in which the time factor ej"' has been reinserted and ejkR/Rhas been substituted for y. The function A is called the magnetic vector potential function and cD is called the electric scalar potential function. Since k = ole, one may write
[ . (t  31
exp [ j ( o t  k R ) ] = exp
[email protected] 1.1 1 The FarField: Type I Antennas
27
Therefore each current element in the integrand of (1.80) and each charge element in the integrand of (1.81) makes a contribution to the potential at (x, y, z) at time t which is in accord with the value it had at the earlier time t  Rlc. But this is consistent with the idea that it takes a time Rlc for a disturbance to travel from (5, q, to (x, y, z). For this reason, (1.80) and (1.81) are often called the retardedpotentials. From (1.78) and (1.79),
c)
in which the subscripts on the del operators have been dropped, since A and @ are functions only of (x, y, 2 ) and not also of (5, q, The differential equations satisfied by A and @ may be deduced by taking the divergence of (1.82) and the curl of (1.83), which leads to
c).
These relations are valid whether J and p are harmonic functions of time or more general time functions representable by Fourier integrals. A proof may be found in Appendix B. All of the results in this section can be seen to be consistent with those obtained in Section 1.5 by a different line of reasoning. C. THE FARFIELD EXPRESSIONS FOR TYPE I (ACTUALSOURCE) ANTENNAS
In antenna problems, one is interested in determining the fields at points remote from the sources. This introduces several simplifications in the field/source relations, as can be seen in the development in the next section. 1. I 1 The FarField : Type I Antennas
The typical situation for an actualsource antenna is suggested by Figure 1.2. The sources are assumed to be oscillating harmonically with time at an angular frequency o and to be confined to some finite volume V. There are no sourceexcluding surfaces S,. For convenience, the origin of coordinates is taken somewhere in V. It is desired to find E and B at a field point (x, y, z) so remote that R ))) max [tZ q 2 Said another way, the maximum dimension of the volume V that contains all the sources is very small compared to the distance from any source point to the field point.
+ + c2]1/2.
T h e FarF~eldIntegrals, Rec~procity,D!rectivity
Jx FIG. 1.2 Notation for FarField Analysis.
Because (x, y, z ) is outside of V and thus is a sourcefree point, it follows that Maxwell's equations (1.44) reduce to
As seen either in the development of Section 1.5 or Section 1.10, B can be related to the timeharmonic current sources by the equation
B = V X A
(1.87)
in which
1 wave number and R the distance from the source point with k = o ! c = 2 ~ / the (5, q , [) to the field point (x, y, z). From (1.86) and (1.87) it follows that, at source
29
1 .I 1 The FarF~eld: Type I Antennas
free points ( x , y, z), E
=
( c 2 / j w )V x V x A
For this reason it is not necessary to find 0.The charge distribution on the antenna need not be known for farfield calculations; the current distribution will suffice. The procedure is reduced to finding A from (1.88) and E and B from (1.89) and (1.87). The distance between source point and field point is given by
R
=
[(x

5)'
+(y

n)'
+
(Z 
[)2]1/2
+ (r sin 8 sin C$  r,~)' + (r cos 8 = [r2 2 4 5 sin 8 cos 4 + q sin 8 sin C $ + ( cos 8 ) + r2 + q2 + = r  (5 sin 6' cos I$ + q sin 8 sin I$ + cos 8 ) + O(rI) =
[(r sin 8 cos $  5)'
()2]1/2

c2]1/2
(1.90)
in which the last result is obtained via a binomial expansion. If (1.90) is inserted in (1.88) and terms of O(r2)are neglected, one obtains the farfield approximation
in which
6: = 5 sin 8 cos q5
+ q sin 8 sin $ +
cos 8
(1.92)
The distance 6: can be interpreted as the dot product of: (I) the position vector drawn from the origin to (5, a, c ) ; and (2) a unit vector drawn from the origin toward ( x , y, z). The result in (1.91) can be given the interpretation that A(x, y, z, t ) is expressible as the product of an outgoing spherical wave
a n d the directional weighting function
The radiated power pattern of the antenna, given by the function 6 ( 8 , $ ) watts per square meter can be expressed in terms of this weighting function a ( $ , 4 ) . To see this relation, one can first perform the curl operations indicated by (1.87) and (1.89). When this is done and only the terms in r  I are retained, it is found thatI4
14The subscript zero has been dropped on H as a simplification, since it is unambiguously clear that the region is free space.
The FarF~eldIntegrals, Rec~procity,D ~ r e c t ~ v ~ t y
in which 1. is a unit vector in the radial direction and 77 = ( p , / ~ , )=' 377 ~ ~ ohms is the impedance of free space. The transverse part of A is AT = l,Ae t 1,A,. It can be concluded from a study of (1.95) and (1.96) that the radiated E and H fields are entirely transverse, that E differs from AT only by a multiplicative constant, that H is perpendicular to E, and that
The complex Poynting vector yields an average power density which can be written (see Section 1.6) @(8, $)
1
= Tale(E
x H*)
It is customary to call that part of the radiation pattern associated with E, the 8polarizedpattern, or the vertically polarizedpattern, and to call that part of the radiation pattern associated with E# the $polarized pattern, or the horizontally polarized pattern. From (1.95) and (1.98), it can be seen that these two patterns are given by the functions
Often one is interested only in the relative power densities being radiated in different directions (8, $), in which case the factor i[k2q/(4nr)2]can be suppressed. Since the unit vectors in spherical and cartesian coordinates are connected by the relations 1,
=
1, cos 8 cos $ t 1, cos 6' sin $  1, sin 8
1,
=
1, sin $
+ 1, cos $
it follows that the transverse components of (1.94) can be written in the forms
a d o , 4)
=
1[cos 8 cos B JAC q, i ) + cos 8 sin $ ~ ~ (q,5 ,  sin
a+(o74)
=
(1.101)
6' Jz(5, q, l)]ejked5 dq d l
jv [sin
$ JJ5, % i )
+ cos $ Jy(5, q, i)lejkcd5 dq d(
( 1.102)
These two equations are the key results of this development and form the basis of pattern analysis and synthesis for actualsource antennas. If one starts with known current distributions, a, and a+can be determined from (1.101) and (1.102) and then used in (1.99) and (1.100) to deduce the radiation patterns. This is the analysis prob
1.1 2 The Schelkunoff Equivalence Pr~nc~ple
31
lem. Conversely, if desired patterns are specified, (1.101) and (1.102) become integral equations in the soughtfor current distributions. This is the synthesis problem. The results of this section can be summarized by saying that when one is doing pattern analysis of a type 1 (actualsource) antenna, the steps t o follow are these.
1. Place the known current distribution in (1.101) and (1.102) and determine ao(e, $1 and a,(e, $1. 2. If the farfield power patterns are desired, use (1.99) and (1.100). Then I a,(8, $) l2 and I a @ , 4) I"il1 give the vertically and horizontally polarized relative power patterns, respectively. 3. If the E and H fields are desired, use (1.95) and (1.96). For pattern synthesis, (1.101) and (1.102) become integral equations in the unknown current distribution with a,(O, $) and a,(B, $) specified.I5
D. THE FARFIELD EXPRESSIONS FOR TYPE II (EQUIVALENTSOURCE) ANTENNAS16 A distinction has already been made between antennas for which the actual source distribution is known to reasonable accuracy and those for which it is not. In the latter case, it is fortunately often true that the fields adjacent to the antenna are fairly well known; it is then useful t o surround the antenna by surfaces that exclude all the real sources. If the StrattonChu formulation is used, the fields E(x, y, z)ejwt and B(x, y , z)ejwrcan then be determined from Equations 1.53 and 1.54 with only surface integrals involved. An alternate (and equivalent) approach that is rich in physical insight is one in which substitute sources are placed on the surfaces enclosing the antenna. These sources must be chosen so that they produce the same fields at all points exterior to the surfaces as the actual antenna does. The next two sections are concerned with developing this alternate approach. 1.I 2 The Schelkunoff Equivalence Principle
The concept of equivalent or substitute sources is an old and useful idea that can be traced back to C. Huyghens,17 but the development t o be presented here is patterned after S. A. Schelkunoff.18 1 soften i t is a vexing problem to specify the phase distribution of as and a+since all that may really be desired is some specified ao(O,4)I or la,(B, $)I. In such cases, one can search for that phase distribution of a, and adwhich results in the simplest physically realizable current distribution. This can be a much more formidable synthesis problem. 16Reading the material in Part D of this chapter can be deferred without any loss in continuity until Chapter 3 is reached. 17C. Huyghens, TrairP de la LutiliPre, 1690 (English translation: Chicago: The University of Chicago Press, 1945). 18s. A. Schelkunoff, "Some Equivalence Theorems of Electromagnetics and their Application to Radiation Problems," Bell System Tech. Jour., 15 (1936), 921 12.
The FarField Integrals, Reciprocity, Directivity
In pursuing this idea, one finds that if the equivalent sources are to reproduce faithfully the external fields, electric sources alone will not suffice. It is necessary to introduce fictitious magnetic sources. In anticipation of this, consider the situation in which real electric sources (p, J ) create an electromagnetic field (El, B,) and magnetic sources (p,, J,) create an electromagnetic field (E,, B,). The properties of these fictitious magnetic sources are so chosen that Maxwell's equations are obeyed in the form given below. Away from the sources, no distinction can be made that would allow one to determine which type of source had given rise to either field. The two sets of sources and fields satisfy V
x
El
=
joB,
(a)
VXE,=J"~~B Po '
2
(e)
The divergence of (1.103e) combined with (1.103h) reveals that V J, = JwP,. ' In other words, the manner in which the magnetic sources have been introduced insures that the continuity equation applies for magnetic as well as electric sources. In a development paralleling what is found in Section 1.5, it is useful once again to introduce potential functions, this time by means of the defining relations B1=VxA
(a)
E,=VXF
(b)
(1.104)
As before, A will be called the magnetic vector potential function; by analogy, it is appropriate to call F the electric vector potential function. Equation 1.104a insures compliance with (1.103d); similarly, (1.104b) is in agreement with (1.103g). Equations 1.103a and 1.103f then lead to
from which
with O and cD, called the electric and magnetic scalar potential functions, respectively. If the total fields are E = El E, and B = Bl B,, then
+
+
1.1 2 The Schelkunoff Equivalence Pr~nciple
Equations 1.103b and 1.103e can next be converted to the forms
If the divergences of A and F are selected to satisfy
then these hybrid equations reduce to
V 2 A+ k 2 A =
J pi1

(a)
V2F
+ k Z F= A pi1
(b)
(1.108)
(b)
(1.109)
Finally, (1.103~)and (1.103h) transform to
V Z @f
[email protected] =
k €0
(a)
[email protected],
[email protected],=Prn 0
The solutions for A and @ have already been given (see Section 1.10) and the solutions for F and @, are obviously similar. If the electric and magnetic sources are confined to reside in surfaces, then lineal current densities K amperes per meter and K, magnetic amperes per meter replace J and J,. In like manner, the areal charge densities p, coulombs per square meter and p,, magnetic coulombs per square meter replace p and p,. The potential functions are then given by
Suppose one desires to find the values that these surface sources should have in order to give a specified electromagnetic field external to S but a null field within S. As suggested by Figure 1.3a, let a contour C, be constructed such that the leg ab is just outside S and parallel to B,,,,; the leg cd is parallel to ab and just inside S ; both legs have infinitesimal lengths dl. Since (1.103b) and (1.103f) combine to give V x B = ( J / p i l ) (jw/cZ)E, integration of this result and the application of Stokes' theorem yields
+
in which S, is the membranelike surface stretched over the infinitesimal rectangular contour C,.
The FarF~eldIntegrals, Reciprocity. Directivity
True fields outside S
True fields outside S
Fig. 1.3 Determination of Equivalent Surface Sources
As the legs bc and da are shrunk toward the limit zero, with ab always outside S and cd always inside, the electric flux enclosed goes to zero, the current enclosed is Kdl, and the line integral in (I. 111) gives B,,,,dl, since there is no contribution from inside. In Figure 1.3a, K emerges from the paper if B,,,, is in the direction from a to b. One obtains the result that
with 1, a unit outwarddrawn normal vector. Similarly, if (1.103a) and (1.103e) are added and the result integrated, with the contour taken so that its leg ab is parallel to Eta,,, one finds that
Next, imagine that an infinitesimal pillbox has been erected, straddling S as shown in Figure 1.3b. I f the view in the figure were to be rotated 90°, one would see an
35
1 .I 2 The Scheikunoff Equ~valencePrinc~ple
infinitesimal disclike area dS, with the upper surface of the pillbox just outside S and the lower surface just inside S. Since (1.103~)and (1.103g) combine to give V E = plea, integration plus use of the divergence theorem yields
in which S, is the total surface of the pillbox, enclosing the volume V,. Because there is to be no field inside S , as the height of the pillbox is reduced toward the limit zero, with one pillbox face always on each side of S, in the limit Ends = (p,/e,) dS, with En the normal component of E. Thus
In like manner, if (1.103d) and (1.103h) are combined and this process is repeated, one finds that
Schelkunoff's equivalence principle in essence asserts that, if the equivalent sources given by (I. 1 12) through ( I . 1 15) are inserted in the potential functions (1. l lo), and the results are used in (1.106) and (1.107), the calculation of E and B will give the true fields at all points external to S and null fields at all points internal to S. It is not immediately obvious that this should be so, since all that has been done so far is to choose equivalent sources that would correspond to the situation that the true fields exist infinitesimally outside S and that no fields exist infinitesimally inside S , with no obvious indication that this will produce the proper field values at points further removed from S. However, Schelkunoff's assertion can be affirmed by following his suggested procedure. If the factor elo' is suppressed, if y replaces ejkR/R, and if equations (1.1 12) through (1.1 15) are substituted in (1.1 lo), the result is that
When these expressions for the potential functions are used in (1.106) and (1.107), and the vector transformations V,y = Vsy and V, x [ly(l, x E)] = V,y x (1, X E) = (1, x E) x Vsy are employed, one finds that
' Js[(In *E)Vsy t
(1, X E) x Vsy
1
+ (I.
E ( ~ , Yz ,)
=
B(x, Y , Z )
= 4n 
[ F ( l , X E)
X
B)
X

jwy(1. x B)] dS
Vsy L (I.
(1.1 17)
B ) v , ~ ] d~ (1 .I 18)
The FarF~eldIntegrals. Reciprocity, D ~ r e c t ~ v i t y
where V, and V, are the del operators for the field point variables and source point variables respectively; they have been defined by Equations 1.51. These integral solutions for E and B at the field point (x, 41, z), in terms of the field values over the surface S, are seen to be identical to the StrattonChu solutions 1.53 and 1.54 for the case that all the real sources have been excluded from the exterior volume V. Since it has already been shown in Sections 1.7 and 1.9, via a direct integration of Maxwell's equations, that (1.53) and (1.54) give the true fields at all points exterior to S, whereas they give a null result at all points interior to S, it follows that Schelkunoff's equivalence principle has been established. 1. I 3 The Far Field : Type II Antennas
In a development paralleling what was done in Section 1.11 for actualsource antennas, the potential expressions (I. 110) for equivalentsource antennas can be simplified if the field point (x, y, z) is remote from all the sources. The details need not be repeated, but the thread of the argument proceeds as follows. Away from the sources, (1.103b) and (1.103e) give c2
c2
E l =J W  V x B i =10 , v x ~ x ~ 1 1 B ~ =   V X E , =  V ~ V ~ F JW
JW
so that (1.106) and (107) simplify to
As before, one can dispense with the need to know the charge distributions if the fields are only sought at sourcefree points; knowledge of the current distributions, which determine A and F, is sufficient. The farfield forms of these vector potential functions can be written as the product of the outgoing spherical wave factor (1.93) with the directional weighting functions
When (1.119) and (1.120) are applied to the farfield forms of A and F and only the terms in r  * are retained, the result is
1.1 3 The Far F ~ e l d. Type II Antennas
37
with AT and FT the transverse components of the vector potential functions. Once again it can be noted that the far field E and H are both transverse to the radial direction and are perpendicular to each other, and that 1 E/H I = q. In this case of equivalent sources, the complex Poynting vector gives as the average power density 1 a ( e j 6 ) =
[email protected]{[/oAT 2
+ jk(lr X FT)] x [(:)(I.

k2q 1 21 (4nr)~ae([aT  c ( l r X %T)
=
lrk2"a,a~ 2(4~r)~
x A:)
+j o e , ~ ~ ] ]
+ apy + $(S,SZ + s g y )
(1.125)
A study of (1.123) reveals that the vertically polarized (E,) pattern is related to a, and 5,, whereas the horizontally polarized (E,) pattern is governed by a, and 5,. Thus the component patterns are given by the functions
r ) ~ ]be suppressed when only relative levels are of Once again, the factor + [ / ~ ~ q / ( 4 n can interest.
As before, the transverse components of a and 5 can be obtained by expanding (1.121) and (1.122) into components. This gives a ~ e9),
=
j [COSe cos Q K,(F, V,i ) + cos e sin Q KAC,V,i ) 
a,(e,
$1 =
5,(0,9)
=
5,
sin 0 K2(c, q, c)]ejkzdS
[sin
Q K,(c,
n, i )
js[cos 0 cos Q K,.(c, 
t cos Q Ky(5, n. i)lejkzd~
(1.128) (1.129)
n, i)+ C O e~ sin m Kym(c, V ,
sin 8 K,,(C[, q, [)]ejk":dS
(1.130)
The FarField Integrals, Reciprocity, Directivity
These four equations are the key results of this development and form the core of pattern analysis and synthesis for most equivalentsource antennas.lg If one starts with known equivalentcurrent distributions, a,, a,, 5,, and 5, can be determined from (1.128) through (1.131) and then used in (1.126) and (1.127) to deduce the radiation patterns. This is the analysis problem. Conversely, if desired patterns are specified, (1.128) through (1.13 1) become integral equations in the equivalent current distributions that are sought. This is the synthesis problem. The results of this section can be summarized by indicating the procedure for doing pattern analysis of a type I1 (equivalentsource) antenna. 1. Surround the antenna with a closed surface S over which the actual fields are known, at least to a good approximation. 2. Use (I. 112) and (1.1 13) to find the equivalent lineal current densities K({, q, [) and K,(t, tt, tl) on S. 3. Find aT(8, 4) and 9,(8,1$) from (1.128) through (1.13 1). 4. If the component power patterns are needed, use (I. 126) and (1.127) to determine them. 5. If the far fields E and B are required, use (1.1 19) and (1.120).
For pattern synthesis, (1.128) through (1.131) assume the roles of integral equations in the unknown equivalentcurrent distributions, with aT(8,$) and ST(B,4) specified.20
E. RECIPROCITY, DIRECTIVITY, AND RECEIVING CROSS SECTION OF AN ANTENNA
This penultimate part of Chapter 1 is concerned with the development of several concepts that have proven to be extremely useful in antenna theory. The first of these is the concept of reciprocity, based on a simple deduction from Maxwell's equations. The second (directivity) is a measure of the ability of any antenna to radiate preferentially in some directions relative to others. The last concept (receiving cross section) introduces a measure of the ability of an antenna to "capture" an incoming electromagnetic wave. 19Occasionally a design problem will be encountered in which the antenna is very long in one dimension and the sources are essentially independent of that dimension. It is then convenient to assume that the problem is two dimensional and use cylindrical coordinate expressions equivalent to (1.128) through (1.131). See Appendix G for the development of these expressions. 2oThe synthesis problem is actually quite a bit more complicated than this simple statement would suggest. Often it is only 6r,s(B, $) and (Pr,$(B, 4) that are specified. The division into a~(B,4) and ST(Br$) is immaterial to the desired result, but it may be critical in terms of physical realizability of a synthesized antenna. Another difficulty is that thephase of the farfield pattern is seldom specified. This offers the antenna designer an added degree of freedom, but complicates the synthesis problem. One should strive for a phase distribution of the farfield pattern that permits the simplest physically realizable antenna. This can be a formidable undertaking.
1 .I4 The Reciprocity Theorem
One of the most important and widely used relations in electromagnetic theory is the reciprocity theorem, which will be invoked many times in this text as various subjects are presented. A derivation of this theorem is based on the idea that either of two sets of sources, (Ja, J:, pa, p:) or (Jb,JL, pb, p:), can be established in a region, producing the fields (Ea, Ba) and (Eb, Bb), respectively. It is assumed that the two sets of sources oscillate at a common frequency. There may be dielectric, magnetic, and conductive materials present in which some or all of these sources reside, but if so the electromagnetic behavior of these materials must be linear. The equivalent situation of free and bound sources in free space will be used to represent the behavior of the materials, as a consequence of which Maxwell's curl equations in the free space form,
can be used to connect the fields and current sources for each set. Equations 1.132 are a restatement of (1.103) in combined form, with D = t 0 E and H = fiiLB.These curl equations can be dotted as indicated to give
Since V * ( E a X H b  E b X H a ) = H b  VX E a  E a . V X H b  H a . V + E b * V X Ha
xEb
it follows from (1.133) that
in which integration has been taken over a volume V large enough to contain all the sources of both sets, and in which the divergence theorem has been employed. Equation 1.134 is a statement of the reciprocity theorem for sources in otherwise empty space, but with the possibility that some might be bound sources representing the behavior of linear materials. Several special forms of this reciprocity relation have proven useful and can be described as follows.
The FarField Integrals. Reciprocity, Direct~vity
1. If S is permitted to become a sphere of infinite radius, with the sources confined to a finite volume V, the fields at infinity must consist of outgoing spherical waves for which E, = qH, and E, = qH,. Under these conditions the surface integral in (1.134) vanishes and one obtains
Equation 1.135 is a principal reduction of the reciprocity theorem, which is used in circuit theory to demonstrate a variety of useful relationships. It will be used in this text to establish the equality between transmitting and receiving patterns for arbitrary antennas and to develop a basic formula for the mutual impedance between antenna elements. 2. Another important reduction of the reciprocity theorem can be derived by returning to Equation 1.134 and considering the situation illustrated in Figure 1.4a.
Fig. 1.4 Geometries for T w o Applications of the Reciprocity Theorem
The volume V is enclosed between the surfaces S , and S,, with S , completely surrounding S , . If all the sources are excluded by S , , so that none of them lie in V , then the right side of (1.134) has a null value. And, if S , is once again permitted to become a sphere of infinite radius, the fields at infinity again consist of outgoing spherical waves for which E , = q H b and E , qH,, and the integral over S, in (1.134) vanishes. One is left with
1 .I 5 Equivalence of the Transmitting and Rece~vingPatterns of an Antenna
41
In (1.136), d S is drawn outward from V, but no change in (1.136) occurs if d S is instead drawn outward from V,, the volume enclosed by S , . Thus (1.136) can be interpreted by saying that if a surface S , is constructed to enclose all the sources of both sets in a finite volume V,, then the fields caused by these sources satisfy the relation in (1.136). Equation 1.136 will be used in Chapter 7 in the establishment of the induced EMF method for computing the selfimpedance of a dipole. 3. A variant on the previous reduction is suggested by Figure 1.4b. The closed surface S , excludes all the sources, that is, the volume V, is source free. Application of (1.134) to this situation once again gives (1.136). This result will be used in Chapter 3 in the derivation of a formula for the scattering from a waveguidefed slot. 1.15 Equivalence of the Transmitting and Receiving Patterns of an Antenna
The reciprocity theorem can be used to establish the very important result that the transmitting and receiving patterns of an antenna are the same. Consider the situation indicated by Figure 1.5, in which two antennas are sufficiently separated so that each
{x Fig. 1.5 Disposition of T w o Antennas in Each Other's Far Field
The FarF~eldIntegrals. Rec~procity.D ~ r e c t ~ v i t y
is in the farfield region of the other. Spherical coordinates are arranged to place antenna 1 at the origin and antenna 2 at the point (r, 8,$I). Both antennas can be as simple or complicated as one wishes, so long as they are composed of linear materials. It will be assumed that a transmitter is connected to one antenna via a suitable transmission line and a receiver is connected to the other antenna, also via a suitable transmission line.21 In accord with the notation used in Section 1.14, let the aset of sources occur when a transmitter is attached to antenna 1 and a receiver to antenna 2. The bset of sources will represent the situation when the positions of transmitter and receiver are interchanged. The combination of transmitter and receiver used in the bsituation need not be the same as in the asituation. It will be assumed that a cross section 1 can be found in the transmission line connecting antenna 1 to the transmitter (receiver) at which a single, clean propagating mode exists, and that similarly a cross section 2 can be found in the transmission line connecting antenna 2 to the receiver (transmitter) where a single, clean propagating mode exists.21 For the asituation, let electric and magnetic current sheets be placed at cross section 1 so that the fields on the antenna side are undisturbed, but so that, with the transmitter turned off, the fields on the transmitter side have been erased. From Equations 1.1 12 and 1.113, these port sources are given by
in which 1, points along the transmission line toward antenna 1 and Ea and Ha are evaluated in cross section 1 . In like manner, let electric and magnetic current sheets be placed at cross section 2 so that the fields on the antenna side are not altered, but so that, with the receiver turned off, the fields on the receiver side have been erased. These port sources satisfy
with 1, pointing along the transmission line toward antenna 2, and with Ea and Ha evaluated in cross section 2. The effective replacement of the transmitter and receiver by equivalent sources at ports I and 2 leaves intact all the asources and fields between these cross sections, including the radiation field transmitted by antenna 1 and received by antenna 2. In precisely the same manner, equivalent electric and magnetic current sheets can be found which, when placed at cross sections 1 and 2, can serve as proxies for the transmitter and receiver in the bsituation. These two sets of sources and the fields they produce satisfy the reciprocity theorem in the form of (1.135). The volume V over which the integration is to be performed must encompass all the original sources between the two cross sections plus the equivalent sources in the two cross sections. 2lAs
a special case of this analysis, the transmission lines may be lumped circuits.
1 .I 5 E q u ~ v a l e n c eof t h e T r a n s m ~ t t l n ga n d Recelvlng Patterns of a n Antenna
43
Consider any point between the cross sections that is occupied by sources. If these sources flow in conductive material,
with a the conductivity at that point. Similarly, if the material is dielectric,
with
xe the dielectric susceptibility at that point. And if the material is magnetic,
in which X , is the magnetic susceptibility a t that point. In (1.139) through (1.141), the parameters a, x,, and X , can be functions of position, depending on the composition and disposition of the materials that comprise the two antennas and their feeds, but, with the assumption that all materials are linear, these parameters are independent of the levels of the fields. Thus, for every source point between the cross sections, equal contributions are made t o the integrals on the two sides of (1.135). What remains are the contributions made by the equivalent sources in the cross sections. Equation 1.135 reduces t o
with S , and Sz the cross sectional surfaces at ports 1 and 2. By virtue of the set of relations of the type (1.137), this can be converted to the form
It is demonstrated in textbooks dealing with transmission line theoryzz that any propagating mode can be represented by a voltage wave and a current wave, defined so that Y, z)
=
V(z) g(x, Y)
(1.144)
Htang(x, Y, 2 )
=
I(z) h(x, Y)
(1.145)
Eta&,
with Z the propagation axis and with the functions g(x, y) and h(x, y) characteristic of the given mode. The level of these characteristic functions is adjusted so that
ZzSee, for example, S. Silver, Microwave Antenna Theory and Volume 12 (New York: McGrawHill Book Co., Inc., 1939), p. 55.
Design, MIT Rad.
Lab. Series,
The FarF~eldIntegrals, R e c ~ p r o c ~ tD y .~ r e c t ~ v ~ t y
with S a crosssectional surface. The functions V(z) and I(z) in (1.144) and (1.145) are called the mode voltage and mode current and are given generally by
with A and B constants to be determined by the boundary conditions, with P the propagation constant and Y o the characteristic admittance of the mode. When this representation is applied to the modes at S , and S,, one finds that
Substitution in (1.143) together with use of (1.146) gives
This is a key result of the analysis and can be interpreted as saying that the mode voltages and currents at the two ports satisfy the reciprocity theorem. Next, let Z , , be the impedance of antenna 1 referenced at port 1, and let Z,, be the impedance of antenna 2 referenced at port 2. Then
Further, let Z,, be the impedance of the receiver transformed to port 1 in the bsituation, and let Z,, be the impedance of the receiver transformed to port 2 in the asituation. Then
Vt
=
I?Z,,
V",
ZRZ
(1.151)
When (1.150) and (1.15I) are placed in (1.149), one finds that
The transformed receiver impedances are obviously independent of the direction ( 8 , $ ) from antenna 1 to antenna 2 and, since the two antennas are in far fields of the other, so too are the driving point impedances Z , , and Z,,. Thus
with K
 [I +
(ZR,/Z2,)1/[1 t (Z,,/Z1 ,)I, a constant.
1.1 5 Equivalence of the T r a n s m ~ t t ~ nand g R e c e ~ v ~ nPatterns g of an Antenna
45
If Vf is held fixed and I," is measured as a function of (0,4) while antenna 2 is moved along some programmed path on the spherical surface of radius r, the transmitting field pattern of antenna 1 is recorded. Reciprocally, if V,6 is held fixed and 1; is measured as a function of (0,$) while antenna 2 is moved along the same programmed path, the receiving field pattern of antenna I is recorded. But Equation 1.153 leads to the conclusion that
In words, the normalized transmitting field pattern and the normalized receiving field pattern of any antenna are identical. Some features of this proof are worth noting. No specification of the size, shape, or type of either antenna was necessary, nor were there any restrictions on the types of transmission lines feeding the two antennas, except that each should exhibit a single, clear propagating mode at the chosen ports. The materials of which the antennas and their feeds were composed were arbitrary except that they needed to be linear. It was not necessary for either antenna to be matched to its transmission line, nor was there any requirement that the transmitter or receiver be matched to either transmission line. Also, there was no restriction on the orientation of antenna 2 as it moved along its programmed path. It could be continuously reoriented to measure EB(O,$1, or E,(0, I$), or E,(0, $1, or some arbitrarily shifting polarization. All that is needed is for antenna 2 to replicate its orientation at each point along the path after it has shifted from receive to transmit. One can conclude from this that the proof is very general. Equation 1.154 establishes the equivalence of the transmitting and receiving field patterns of any antenna. A simple extension shows that this equivalence applies to the power patterns as well. If (1.154) is multiplied by its complex conjugate, the result can be used to deduce that
The quantities 11,"12R,,/2 and IIPIZ RR,!2 that appear in (1.155) are the powers absorbed in the receiver when antenna 1 is transmitting and receiving, respectively. Since each is linearly proportional to the power density of the waves passing the receiving antenna, it is proper to infer that they are measures of the transmitting and receivingpower patterns of antenna 1. With K' = (RR,/R,,) I KV,b/Vf IZ, one can write
Care must be taken in interpreting (1.156). For example, if antenna 2 is linearly polarized and always oriented as it moves along its programmed path, in order to receive or transmit only 0polarized waves, then ( 1.156) becomes
The FarField Integrals, R e c ~ p r o c ~ t Directivity y.
from which one can conclude that the normalized 0polarized component of the power pattern of antenna 1 is the same for receive and transmit. Similarly, if antenna 2 is linearly polarized but aligned to receive or transmit only $polarized waves, (I. 156) reduces to S2c(0, $1
=
K'Sc(e,
$1
(1.158)
and once again equivalence is demonstrated in the component power patterns for antenna 1. And if, for example, antenna 1 is linearly polarized with only an E, electric field, then (1.158) gives a null result, as it should. If antenna 1 does not radiate an E, field, it cannot detect an incoming E, field. The sum of Equations 1.157 and 1.158 shows that the total power patterns are equivalent : s::;,(e,
$1
=
~~s:;,,,(e,$1
(1.159)
Acceptance of the conclusion that the normalized total power patterns of any antenna are the same for transmit and receive, and thus that one need not determine both, still leaves a measurement difficulty that should be noted. This concerns the fact that not any antenna can be chosen to play the role of antenna 2, make one traverse of the programmed path, and a t each point in the path be oriented so that the received powers in (1.155) coincide with the power densities in (1.159). This will occur only if antenna 2 is polarizationmatched to antenna 1. For example, if antenna 1 is circularly polarized, antenna 2 must be circularly polarized in the proper screw sense in order to have the received powers in the a and bsituations that can be interpreted as the total radiated and received power patterns of antenna 1. However, if one is content to use as antenna 2 a linearly polarized antenna, make two traverses of the programmed path, one with 0orientation and the other with $orientation, and keep transmit power and receiver sensitivity stable, then the separate measurements give the component power patterns. Their sum gives the total power pattern, and Equations 1.157 through 1.159 indicate that it does not matter whether the measurements are made with antenna 1 transmitting and antenna 2 receiving, or vice versa. 1 .I 6 Directivity and Gain
Often a principal goal in antenna design is to establish a specified radiation pattern 6(0,$) watts per square meter through a suitable arrangement of sources. The specified pattern frequently embodies the intent to enhance the radiation in certain directions and suppress it in others. A useful measure of this is the directivity, which is simply the radiated power density in the direction (0, 4) divided by the radiated power density averaged over all directions; that is, D(e, $1
=
@(e,$1 (1/4xr2) Jon J:'@(0', @ ) r 2 sin 8' dBf d$'
(1.160)
47
1.16 Directivity and G a ~ n
Equation 1.160 contains the implications that the origin for spherical coordinates has been chosen somewhere in the immediate vicinity of the antenna, and that power densities are being evaluated on the surface of a sphere whose radius r is large enough to ensure being in the far field of the antenna. Jf the radiation intensity is defined by
then, since 6(0,$) is measured in watts per square meter, it follows that P(0,$) is measured in watts per steradian. Substitution in (1.160) gives the equivalent expression D(0, $)
= Jon
J
4np(e, $1 $11 sin 81doJ d$*
" P(P,
The value D(8, $) is a pure numeric. It will have a value less than unity in directions in which radiation has been suppressed, and a value exceeding unity where the radiation has been enhanced. If (e,, $,) is the direction in which the radiation intensity is greatest, then D has its largest value at ( d o , 4,) and D(Oo, 6,) is the peak directivity. In characterizing an antenna, one must be careful to distinguish between directivity and gain. Directivity is used to compare the radiation intensity in a given direction to the average radiation intensity and thus pays no heed to the power losses in the materials comprising the antenna. Gain includes these losses, and the definition of gain is therefore
in which P,,,is the total power accepted by the antenna from the transmitter, measured in watts. The denominator of (1.163) is the value, in watts per square meter, that the radiated power density would have if all the power accepted by the antenna were radiated isotropically. Since the power accepted is greater than the actual power radiated, the denominator of (1.163) is larger than the denominator of (1.160), and, as a consequence, G(0, Q) < D(B,$). Most antennas are constructed of linear materials; in this case, one may argue that .n
Pa,,= KL jo
2n
6'(01, $')r2 sin 0' dB1dm'
(1.164)
with K, a pure real constant that has a value somewhat greater than unity. When this is so, Equation 1.163 becomes
The FarField Integrals, Reciprocity, Direct~vity
The gain and directivity differ by a multiplicative factor that is independent of direction. In particular, the peak gain occurs in the same direction (O,, 6,) as the peak directivity. Often, gain and directivity are expressed in decibels (dB). From (1.165),
The gain in any direction is seen to be 10 logloK, decibels below the directivity in that direction; 10 logl0KLthus represents the power losses in the materials forming the antenna. For some applications it is useful to introduce the concept of partial directivity and partiul gain. As an example of how this is done, a return to Equations 1.98 through 1 .I00 or 1.125 through 1.127 helps to recall that
If this relation is inserted in Equation 1.160, it can be seen that it is possible to write
in which D1(8,$) = (l/4nr2)
l' j 0
@r,de,6) 2'@(8r,$')r2 sin 8' dBf d$'
(1.169)
0
and D"(8, $)
=
(l/4nr2)
I
jn 0
@r,+(e9 4) @(el,$')r2 sin 8' dB1d$'
(1.170)
0
are the partial directivities associated with the 8component and $component patterns, respectively. Similar definitions follow readily for the partial gains. An example of the utility of this concept would be when an antenna is to be designed to give peak radiation at an angle (8,, $,), but all the radiation should be 8polarized; any $polarized radiation is unwanted, but for practical reasons some may be unavoidable. In such a circumstance it is the peak partial directivity D1(OO,$,) that is a pertinent measure, not the peak total directivity D(B0, 4,). The division of the total power pattern into components can be done in other ways than the 8/$ partition indicated above. For example, the decomposition could equally well be into righthanded and lefthanded circularly polarized component power patterns. In that case one could identify righthanded and lefthanded partial directivities and gains. 1.I 7 Receiving Cross Section
A receiving antenna will absorb energy from an incident plane wave and feed it via a transmission line to its terminating impedance. A useful measure of its ability to do this results from introducing the concept of the absorption cross section of the antenna
1 .I 7 R e c e ~ v ~ nCross g Sect~on
49
or, as it is more commonly known, its eq~rivaientreceiving crosssectional area. If S is the power density of the incoming plane wave in watts per square meter and P, is the absorbed power in watts, then the equation
serves to define the receiving cross section, in square meters, as a function of the angle of arrival of the incoming signal. In order to have A,(8, $) be a maximum measure of the capture property of the antenna, it is customary to assume that the incomingplane wave is polarization matched to the antenna, and that the antenna is terminated by a matched receiver. With these assumptions, Equations 1.155 and 1.159 are applicable and one can write
An integration of (1.172) gives A,(&, $') sin 8' dB' d$'=
K'
lL2'
@~:,,(8', $')rZ sin 8' dBf d$'
(1.173)
If the ratio of (1.172) to (1.173) is taken, one obtains
in which D(8, 4) is the directivity of antenna 1 when it is transmitting, as given by is the average receiving cross section of antenna 1, defined by (1.160). Then
SoS, a,(el.40 n
i, =
Zn
sin
ef def d4'
is the same for It is a remarkable fact that the average receiving cross section all lossless antennas that are polarization matched. This can be demonstrated as follows. Consider again the situation of two antennas, depicted as in Figure 1.5, with antenna 1 transmitting and antenna 2 receiving in the asituation and the reverse occurring in the bsituation. T o obtain maximum power transfer, assume that in the asituation the transmitter attached to antenna 1 has an internal emf V , and an internal impedance that has been adjusted to equal Z,*,, with Z , , the driving point impedance of antenna 1. Similarly, in the bsituation, let the transmitter attached to antenna 2 have an internal emf V, and an internal impedance Z & , with Z z 2the driving point impedance of antenna 2. In the asituation, I f = V,/2RI, and the power delivered to antenna 1 is IfIZR,, = 1 V, 12/8R,,. If the losses in the antenna can be neglected, all of this
The F a r  F ~ e l dIntegrals. Reciproc~ty,D ~ r e c t ~ v ~ t y
50
power is radiated, with an average density in watts per square meter given by
When antenna 2 is located at the point (r, 8 , $), the power density in the wave arriving from antenna 1 is given in watts per square meter by
in which D , ( 8 , 4) is the directivity of antenna 1 . If use is made of (1.171) with S = @(8, $), it can be argued that the power absorbed by the receiver attached to antenna 2 is given by
P, = 1 V c4nr2 1 2 / 8 R l D,(B, Q)A,,,(B,$1 watts with A,,,(B,
4) the receiving cross section of antenna 2. Use of (1.174) converts this to
The power absorbed is also given by (1/2)@eI,"Z;*ZR2but, with a matched receiver, Z,, = Z;,, and thus f'r
= &11,"12 R22
(1.177)
When (1.176) and (1.177) are combined, the result can be written in the form
If this analysis is repeated for the bsituation one finds that
The currents and voltages in the two situations are related generally by Equation 1.152. In the circumstance being considered here, V ; = C Z , = ( V 8 / 2 R , ) Z , , , V,b = Z,bZ2, = (V8/2R2,)Z2,,Z R l = ZT1,and Z R 2= 25; as a consequence of this, (1.152) reduces to
,
I$
= If
Hence, upon comparing (1.178) and (1.179), one can see that

Ar,1

= Ar,z
(1.180)
1.1 7 R e c e ~ v ~ nCross g Sect~on
51
Since antennas I and 2 are completely arbitrary (except that they must be polarizationmatched), Equation 1.18 1 is a general result. The value of the constant for linearly polarized antennas can be deduced as follows: Let antenna 1 be completely arbitrary and located a t the origin, as shown in Figure 1.5, except that it is linearly polarized and has been oriented to transmit a n E field that has only a 8component. Antenna 2 is a single current element of length dl, located a t the point (r, 8 , 4 ) , and oriented parallel to 1, so that the two antennas are polarization matched. In the asituation, let antenna 1 be transmitting with antenna 2 absent. In the bsituation, the current element I,I,b dl (antenna 2) is present and radiating, and antenna 1 is receiving. Port 2 is taken to be the 8directed line segment of length dl located at (r, 8, 4). In this case the reciprocity relation (1.149) becomes
which can be rewritten in the form
,
Since fi = Vf/Z,when antenna 1 is transmitting, and I t when antenna 1 is receiving, (1.182) assumes the form
=

V,b/Z,,
=

Vf/Zrl
When (1.183) is multiplied by its complex conjugate, the result is
In the asituation, antenna I accepts an amount of power given by
from the transmitter and, if losses in antenna I are neglected, all of this power is radiated. The power density at (r, 8, $) in the asituation is, therefore,
When ( E; I2 is eliminated from (1.184) and (I. l85), one obtains the result that
The F a r  F ~ e l dIntegrals, Reclproc~ty,D i r e c t ~ v l t y
I n the bsituation, the receiver attached to antenna 1 absorbs the power
since Z,, = Z:,. This absorbed power can also be expressed in terms of the power density in the waves radiated by the current element and the receiving cross section of antenna 1. From Equations 1.99 and 1.101 the maximum23power density radiated by a single current element is
and thus the power absorbed by antenna 1 is also given by
If (1.187) a n d (1.189) a r e c o m b i n e d a n d t h e r e s u l t s o l v e d f o r I VP 12, f u r t h e r c o m b i n a t i o n
with (1.186) gives
and thus the universal value of the average receiving cross section for linearly polarized antennas is A2/4n. Equation 1.190 is a n extremely useful result. It permits computation of the optimum power level in a receiver which is attached to a n antenna of peak directivity D(0,, $,) when the power density in the incoming signal is known. This value is diminished slightly by the losses in the antenna. It is also diminished by the multiplicative factor (1  1 r I 2 ) when the receiver and the antenna are mismatched, with r the reflection c o e f f i ~ i e n t . ~ ~ F. POLARIZATION This concluding section of Chapter 1 is concerned with characterizing the polarization of a n electromagnetic field far from the sources which produce it. Such characterization is important in many practical applications. Prominent examples include the following. (1) For purposes of optimizing propagation through a selective medium (such as the ionosphere), or optimizing backscattering off a target, it may be desirable t o specify the polarization the wave should have. This places a constraint on the design of the transmitting antenna. (2) When a sum pattern is required to have a 2 3 I t is the maximum value that should be used since the current element is oriented so that its maximum power density is directed at antenna 1. W e e , for example, Silver, Microwave .Antenna Theory, and Design, pp. 5153.
53
1.1 8 Polarization of the Electric Field
specified polarization and low side lobes, it is important t o check that the antenna being proposed does not produce a crosspolarized pattern at a height which exceeds the desired side lobe level. This possibility exists, for example, with parabolic reflector antennas. (3) The polarization of an incoming wave may have t o be accepted, which places a constraint on the design of an antenna that will receive this wave optimally. (4) The polarization of an incoming wave may be unpredictable, in which case it may be desirable to design a receiving antenna which will respond equally t o all polarizations. T o be equipped to deal with these and similar problems, it is important t o be able to describe the polarization of an electromagnetic wave unambiguously. 1 .I 8 Polarization o f the Electric Field
It has been shown in Sections 1.1 1 and 1.13 that the far field of a transmitting antenna can be viewed as the product of a n outgoing spherical wave and a complex directional weighting function. For the electric field (which is conventionally used as the vehicle for describing polarization), this complex directional weighting function is given by (1.95) for type I antennas and by (1.123) for type I1 antennas. In either case, at a far field point (r, 9, $), the electric field can be represented by
when timeharmonic sources are used in the transmitting antenna. The functions E,(r, 9, $) and E,(r, 9, $) that appear in (1.191) are, in general, complex. If this is recognized by the notation
then it can be appreciated that what is really meant by (1.191) is that E(r, 9, $, t)
+ jE;) + 1,(EL $ jE&)]ejwr = l,(EL cos wt Et sin wt) + 1,(E& cos o t  E&sin o t )
=
&e[l,(E;
(1.193)

with EL, E t , E i , Ry all real functions of r, 9, and $. Equation 1.193 can be rewritten in the form
E
=
l,Acos(ot
+ a ) + l , B c o s ( o t + P)
(1.194)
in which A
u
=

+
J ( E ~ (E;;)~ E" arctan 2 E:,
B
p
+
J ( E ~ (E;)~ E" = arctan fE
=
,
With no loss in generality, the origin of time can be selected so that u E t = 0). Then (I. 194) becomes E
=
1,A cos or
+ 1,B cos ( a t 1 P)
=
0 (that is, (1.196)
The FarF~eldIntegrals, Reciprocity. Directivity
Equation 1.196 is a particularly convenient representation of the electric field for the purpose of identifying its polarization.
(a) LINEAR POLARIZATION If B = 0, the electromagnetic wave is said to be linearlypolarized in the 9direction. Similarly, if A = 0, the wave is 4polarized, But more generally, if /I = 0 but A , B # 0, the 9 and 4 components of the electric field are in phase. The polarization is then tilted, but it is still linear, as can be seen from the time plots of Figure 1.6a. Therefore the most general example of linear polarization occurs when E, and E, are in phase. (b) CIRCULAR POLARIZATION becomes E
=
If A = B and
A(1, cos o t
/I = 90°,
Equation 1.196
+ 1, sin o t )
(1.197)
In this case the magnitude of E is constant with time. The angle that E makes with the l e direction is ot and this angle changes linearly with time. The locus of the tip of E is a circle, as indicated in Figure 1.6b. For this reason, the field is said to be circularly polarized. The sequence in Figure 1.6b is drawn as though the observer were looking toward the transmitting antenna from afar, along a longitudinal line in the (9,4) direction. The progression of E with time is seen to be counterclockwise, which is the direction of rotation a righthand screw would have if it were being turned to progress in the direction of propagation. For this reason, (1.197) is said to represent a righthanded circularly polarized wave. If one were to write
E
=
A(1, cos a t

(1.198)
1, sin a t )
so that E, leads Eo by 90°, instead of lagging by 90" as in (1.197), then a lefthanded circularly polarized wave would be described. (c) ELLIPTICAL POLARIZATION The most general case of (I. 196) occurs when A # B, /I # 0. The magnitude of E is given by
If the time derivative of this function is set equal to zero, the extrema of I E(t) I can be identified. They occur at angles ot = 6 governed by tan 2 6
=

BZsin 2P A2
+ BZ
COS
2p
+
If 6, is the angle in the first quadrant which satisfies (1.200), then 6, = 6, n/2 also satisfies (1.200). Substitution of the angles 6, and 6, in (1.196) reveals both the direction and magnitude of each of the two extrema of E(t). The two directions are at right angles to each other and form the principal axes of the locus. It is left as an exercise to show
1.1 8 Polar~zat~on of the Electric F~eld
*
7 17
+
I
y
+
I
(a) Linear polarization
GlQoaa~ (b) Circular polarization
Q3366 (c) Elliptical polarization
wt=o
wt=n
4
;
wt=
I
3r
I= 4
at=,,
I
I.
Fig. 1.6
Phasor Plots of E Versus Time for Electromagnetic Waver, of Various Polariza
tions
that this locus is an ellipse.2s Its semimajor and semiminor diameters can also be found by substituting 6, and 6, in (1.199). This gives
2SThe components Eg and E6 that occur in (1.196) are analogous to the voltages applied to the two sets of deflecting plates in an oscilloscope in order to create a Lissajou figure on the screen.
The FarField Integrals. Reciprocity. Directivity
A typical plot of (1.196) is shown in Figure 1 . 6 ~ As . in the case of circular polarization, the direction of rotation of E can be either clockwise (lefthanded elliptical polarization) o r counterclockwise (righthanded elliptical polarization). This is determined by whether the phase angle j3 is lead or lag.
REFERENCES COLLIN,R. E. and F. J. ZUCKER, Antenna Theory, Part I (New York: McGrawHill Book Co., Inc., 1969).
R. S., Electromagnetics (New York: McGrawHill Book Co., Inc., 1966). ELLIOTT, ELLIOTT,R. S., "The Theory of Antenna Arrays," Chapter 1 in Volume I1 of Microwave Scanning Antennas, ed. R . C. Hansen (New York: Academic Press, 1966).
S., Microwave Antenna Theory and Design, MIT Rad. Lab. Series, Volume 12 (New SILVER, York: McGrawHill Book Co., Inc., 1939). PROBLEMS 1.1 Complete the StrattonChu derivation by letting F = B and repeating the analysis that was used in Section 1.7 to obtain B, thus establishing Equation 1.54 of the text. 1.2 Alternatively, take the curl of Equation 1.53 to find  j w B and in this manner verify Equation 1.54 of the text. 1.3 Use the expression for the curl of a vector in spherical coordinates and begin with Equation 1.91 in the form
Then use (l.lOb), (1.87), and (1.89) to deduce that, in the farfield
thus confirming (1.95) and (1.96). 1.4 Demonstrate the validity of equations (1.113) and (1.115) in the text. 1.5 Use equivalentsource Equations 1.112 through 1.1 15 in the retarded potential functions (1.1 10) and show in detail that the results agree with the surface integrals in the StrattonChu formulation for E(x, y, z ) and B(x, y, 2). 1.6 Begin with the farfield expressions (1.123) and (1.124) and show that the power radiated has a density given by (1.125). 1.7 Enumerate the theorems in circuit analysis that can be proven with the aid of the reciprocity relation (1.134). Sketch the proof of each. 1.8 An antenna A, when transmitting, radiates a circularly polarized field in the direction (8, $), which is righthanded. If antenna A is receiving an elliptically polarized electro
magnetic wave, incident from the direction (8, $), state the conditions of ellipticity which will maximize the received signal. State those which will minimize it. 1.9 Show that IE(t)l, as given generally by Equation 1.199, has as its locus an ellipse with axes that occur at angles 8 , and 62 = 6 , nj2, with these angles satisfying (1.200). What is the ellipticity ratio?
+
2
ndiatiouit patterns of dipoles, loop% d helices
2.1 Introduction
In this chapter the formulas that have been developed for the fields caused by an assumed known distribution of current will be applied to a succession of simple but practical radiators. These type I (actualsource antennas) include dipoles, loops, and helices. The centerfed dipole of length 21 will be taken up first, with emphasis on the two cases of greatest interest, when 21 112, and when 21 d. A voltage is applied across the gap, often by means of a twowire transmission line. The resulting current distribution on the pair of tubular conductors gives rise to a radiating field. If a good estimation can be made of this current distribution, the formulas of Section 1.1 1 can be used to deduce the field. One can gain insight to the current distribution by considering the case of a twowire transmission line that is opened out, as shown in Figure 2.2. Without any flare, the opencircuit termination causes a standingwave distribution of current, oppositely directed in the two conductors. Pairs of current elements, which are equal, opposite, and close together, radiate negligibly, which is the behavior of a good transmission
2.2 The CenterFed D p o l e
line. If the origin of the Xaxis is taken a distance I back from the end of the transmission line, the current distribution can be given by /(x, t )
=
I, sin [ k ( l  x)]ejw'
With a flare of 45", as shown in the second panel of Figure 2.2, the inductance and capacitance per unit length change with position along the flared segment, and thus so too does the characteristic impedance; however, to first order, the wave number is still constant at the freespace value k . For this reason, one can argue that the current distribution is little altered by the flare. This is still assumed to be the case in the third panel of Figure 2.2, where the current distribution is also shown as that of a standing wave with sinusoidal spatial distribution. Note that the pair of current elements, which had canceled each other's radiation tendencies in the first panel where they were oppositely directed and close, are more widely separated and reinforcing in the third panel, \.chich serves to illuminate why a dipole radiates. Modern methods, pioneered by the work of E. Hallknl and S. A. Schelkunoff2 and using powerful computational techniques such as the method of moments, have led to more precise knowledge of the current distribution on a cylindrical dipole, but the deviation from a sinusoidal function is found not to be great, and for pattern calculations can be ignored. (Compare with Section 7.6, and particularly Figures 7.8 and 7.9). With the dipole diameter
[email protected]> d, the real part of the impedance is close to 73 ohms.
Radiat~onPatterns of Dipoles. Loops, and Hel~ces
Now attention can be turned to the second special case. 2. The short dipole, 21 > w. A twowire line can be imagined to be feeding the slot at the central points P , and P,. With w ((( A, the slot itself resembles a section of twowire line, the two "wires" being semiinfinite ground planes with adjacent edges at x = &w/2, with these "wires" shorted at z = *I. A standing wave of voltage exists on this section of line
Fig. 3.6 CenterFed Slot i n a Large Ground Plane
3.4 CenterFed Slot In Large Ground Plane
such that the electric field in the slot is given to good approximation by
in which V,,, is the peak voltage. If the generator and twowire line which attaches to the points PI and P, are in y < 0, the fields in y > 0can be determined by the same technique used in connection with Figure 3.1. The actual large ground plane is modeled by a n infinite ground plane. A sourceexcluding surface S is constructed that consists of a boss S , in front of the slot, plus a n otherwise infinite plane S , inside the ground plane, plus an infinite hemisphere in y < 0. Use of the image principle reduces all the sources t o a lineal magnetic current sheet on S , given by K,
=

2pi11, x 1 , L II' sin [k(l
1,
it3
sin [ k ( ~

1 [I)]
11111
Equation 3.17 is identical in form to Equation 2.2. For this reason, a narrow centerfed slot in a large ground plane is often referred to as a magnetic dipole. The analysis of Section 2.2 can be repeated with the principal result that s,(e>
=
4pi1Vm [cos (kl cos 8) k sin 8

cos (kl)]
For a slot onehalf wavelength long, use of (1.123) and (1.124) gives
E$=
;''[ v
e ~ ( w k~r )
cos [(n/2) cos 81 sin 8
cos [(71/2) cos 81 which are in the same form as (2.8) and (2.9), but w ~ t hthe polarization rotated 90 . The field pattern shown in Figure 2.4 thus also applies for a halfwavelength slot in a lsrge ground plane. Since there is no A for this antenna, (1.125) yields
8 V 'I? cos2 [(n/2) cos 811 p,,4(e) ,= _ (471r)' sin28 j
[
If the presence of the transmitter and twowire feed in y < 0 can be assumed to have little influence on the farfield in y < 0, then (3.19) through (3.21) apply on both sides of the ground plane. Under this assumption, the total power radiated is
Radiation Patterns of Horns. Slots, and Patch Antennas
Since, for this case of 1 = 1214, the peak input voltage to the slot (at the terminals P,/P,) is Vm,the feeding transmission lines can be said to be delivering the power
to a conductance G,,, placed across its terminus. The value R,,, = l/G,,, is called the radiation resistance of the centerfed half wavelength slot. From (3.23),
R,,,
nv14 =
0.609
486 ohms
An interesting relation results if (3.24) is multiplied by (2.14), for then
This is a special case of Booker's relation4. It will be shown in Chapter 7 that
for any length 21 of a narrow dipole, as long as the complementary slot in a ground plane has the same length. The analysis undertaken in Section 2.2 for a short dipole could be repeated here for a short slot. All of the results are similar, with E and H interchanged; this is left as an exercise. The practical applications of a twowire fed slot, cut in a large ground plane and radiating into both halfspaces, are few. However, if the slot is "boxed in" on one side by a metallicwalled cavity and the dimensions of the cavity are properly chosen, the radiation in one halfspace is hardly affected, whereas in the other halfspace, it is virtually eliminated. The presence of the cavity affects the input impedance of the slot. This is a subject which will be treated in Chapter 8. 3.5 WaveguideFed Slots
Most antenna applications involving slots unify the feeding and radiating structures by placing the slots in one of the walls of a rectangular waveguide. This insures a nonradiating transmission line, permits precise machining of the slots, and provides a mechanically rigid structure. Usually the slots are arranged in arrays, which complicates the feeding because of mutual coupling. That subject will be treated in Chapter 4H. G . Booker, "Slot Aerials and Their Relation to Complementary Wire Aerials (Babinet's Principle)", J.I.E.E. (London), 93, part IIIA (1946), 62026.
89
3.5 WaveguideFed Slots
8. For now, the discussion will be limited to the behavior of a single slot, cut in one of the walls of a rectangular waveguide and excited by a TE,, mode. With axes chosen as shown in Figure 3.7, the normalized field components for a TE,, mode, traveling in the positive zdirection, are
Fig. 3.7 Rectangular Waveguide
The normalization in (3.27) consists of choosing the peak magnitude of the longitudinal component to be unity, but adjusting the phase by 90" through the presence of the factor j. This causes the transverse components to have pure real amplitudes, a convenience since they enter into the Poynting vector calculation of power flow. Figure 3.8a shows the electric field distribution at a fixed time, and Figure 3.8b
(a) Electric field and charge distribution
(b) Magnetic field
(c) Current flow
Fig. 3.8 Field and Source Distributions in a Rectangular Waveguide for
TElo Mode (From Electromagnetics by R. S. Elliott. Copyright 1966, McGrawHill. Used with permission of McGrawHill Book Company.)
Radiation Patterns of Horns, Slots, and Patch Antennas
illustrates the corresponding distribution of magnetic field. The charge and current distribution in the waveguide walls can be determined using the same procedure which led to Equations 1.112 and 1.114. Figure 3.8a shows the instantaneous distribution ~ the corresponding instanof charge on the upper broad wall and Figure 3 . 8 pictures taneous current distribution. Over time, these patterns will propagate longitudinally at the phase velocity of the TE,, mode. If a narrow slot is cut in one of the waveguide walls such that its long dimension runs parallel to a current line, the presence of the slot causes only a minor perturbation in the current distribution, and negligible coupling to outer space occurs. Thus the longitudinal slot on the center line in Figure 3.9a causes little disturbance, as can be seen by studying the current distribution in Figure 3 . 8 ~ Such . a slot is useful for making measurements of the Efield inside the waveguide, since a vertical probe can be inserted through this slot to sample the field. If the probe is permitted to move longitudinally, VSWR data can be obtained.
(a) Longitudinal slots i n broad wall
(b) Inclined slot i n broad wall
(c) Inclined slot i n narrow wall
Fig. 3.9 Some Practical Slot Configurations in the Walls of a Rectangular Waveguide (From Electromagnetics by R. S. Elliott. Copyright 1966, McGrawHill. Used with permission of McGrawHill Book Company.)
However, the longitudinal slot displaced from the center line in Figure 3.9a will interrupt Xdirected current; the more the displacement, the greater the interruption. The electric field developed in this slot has as one of its manifestations a displacement current which "replaces" the interrupted conduction current. This electric field can be represented by its equivalent magnetic current sheet, and can radiate into outer space. Similarly, the inclined broad wall slot in Figure 3.9b interrupts Zdirected current, the more so the greater the inclination. This is another candidate for use as a radiating element. Finally, the inclined slot in the narrow wall, shown in Figure 3.9c, will interrupt Ydirected current, the more the inclination, the greater the interruption. This slot is a third candidate for use as a radiating element. A desirable feature shared by all three of these radiatingtype slots is that there is mechanical control over the amount of radiation, through choice of the amount of displacement or inclination. What is needed is a set of design equations which reveal this connection, a problem that is addressed in the next section.
3.6 Theory of WaveguideFed Slot Radiators5
It is assumed that the reader is familiar with waveguide mode theory, and thus it will be stated without proof that the field components in rectangular guide can be expressed in the normalized form TE,, rnode
Hz = jH
02
p'Y"z
TM,, mode E,
=
I jE,,eT
Et
=
Este'Yoz
E,
=
Eatefy a z
Ht
=
&HateTycz
H,
=
+HateTy~'
In (3.28) the subscript u is shorthand for the double index mn, and nx nny Ha, = cos m a cOsT E,,

m n x . nny sin a sln b
The upper signs in (3.28) need t o be taken for propagation in the positive Zdirection; the lower signs for propagation in the negative Zdirection. The transverse field vectors are given by TE,, modes
TM,, modes
sThe results to be presented in this section were first obtained by A. F. Stevenson in a classic paper "Theory of Slots in Rectangular Waveguides", J. Appl. Phys., 19 (1948), 2438. However, the development follows an approach used by J. E. Eaton, L. J. Eyges, and G. G. MacFarlane, Microwave Antenna Tl~eoryand Design, ed. S. Silver, vol. 12, MIT Rad. Lab Series (New York: McGrawHill Book Co., Inc., 1949), Chapter 9.
Rad~ationPatterns of Horns. Slots. and Patch Antennas
This information needs to be applied to scattering off a slot cut in one of the walls of the waveguide. Without at this point specifying which type of slot it might be from among those illustrated in Figure 3.8, imagine that the slot is contained in the region bounded by z = z , and z = z,, with z , > z , . If the waveguide is assumed to be infinitely long and a TE,, mode is launched from z =  m , traveling in the positive 2direction, the incidence of this mode on the slot will cause a profusion of reactions. Backward and forward scattering of all TE,, and TM,, modes is possible. Radiation into outer space via the electric field set up in the slot is possible. If the waveguide walls are assumed to be perfectly conducting, and if the a and bdimensions are chosen so that all modes except TE,, are cut off, then a power balance can be written that will connect the slot's excitation to its displacement or inclination. This can be done because (1) the power contained in the incident wave is calculable; (2) the power radiated is also calculable if the electric field in the slot is known; and (3) the forward and backward scattered waves in the TE,, mode can be determined if the electric field in the slot is known. It is this last determination that completes the linkage in the power balance equation. To see the relation between scattering off the slot and the electric field distribution in the slot, consider two fields (El, HI) and (El, Hz), both timeharmonic at the common angular frequency w ,and both satisfying Maxwell's equations in a region V bounded by a closed surface S. Let S be a rectangular parallelopiped with end faces S , at z = z , and S , at z = z,; the remainder of S is a surface S , which is skintight against the four interior faces of the waveguide between z , and z,. With S a sourcefree region, the reciprocity theorem in the form of (1.136) is applicable and the two fields are connected by the relation
Let (El, HI) be the scattered field due to the interaction of the slot and the incident mode. (E,, H z ) does not relate to the actual situation, but its use is an artifice to obtain the scattering coefficient. It will be taken to be a single normalized mode, that is, a member of either the TE or TM families given in (3.28), traveling in the positive 2direction and designated by the subscript b = m'n'. The transverse components of the scattered field (El, HI) can be represented by
in which the summation is over all TE and TM modes. The forwardscattered mode amplitudes C, and the backwardscattered mode amplitudes B, are yet to be deter
3.6 Theory of Wavegu~deFedSlot Radiators
93
mined. It should be noted that (E,, HI) cannot be represented by (3.35) in the region z , < z < z , because of the presence of the slot. Since the tangential component of E, is identically zero over the entire subsurface S,, and since the tangential component of El is identically zero over all of the subsurface S , except that part occupied by the slot, it follows from (3.34) that
J,l, (El x H,)
dS
=
(3.36)
I , f I,
in which
and 12
=
j,(E, x H,

E l x H,)
dS
1
Because of the orthogonal properties of these modes, the only contribution to I , comes when a = mn not only equals b = m'n', but also when the indices mn and m'n' refer to the same type of mode (both TE or both TM). The proof is left as an exercise, the result being that
The same argument applies to the evaluation of I,, except that in the special case the integrand is identically zero, and therefore I, 0 . When these results are placed in (3.36), a formula emerges from which the backscattered mode amplitude B, can be computed as follows. a = mn = b = m'n'

If this process is repeated with the only change being that (E,, H,) is assumed to be propagating in the negative Zdirection, one finds that
Radiation Patterns of Horns, Slots, and Patch Antennas
It is important to note that, although the denominators of (3.40) and (3.41) are equal, the numerators are not necessarily equal, because H, in (3.40) is associated with a +Z propagating mode, whereas the H, in (3.41) is associated with a 2 propagating mode. Equations 3.40 and 3.41 have wide applicability. As an illustration of their use, consider an offset longitudinal shunt slot in the upper broad wall, depicted in Figure 3.10. It will be assumed that the waveguide walls have negligible thickness and are composed of perfect conductor. The slot is rectangular with length 21 and width w where 21 >> w. The origin of coordinates has been taken so that the XYplane bisects the slot. The transverse dimensions of the waveguide are chosen so that only the T E , , mode can propagate.
Fig. 3.10 An Offset Longitudinal Slot in the Upper Broad Wall of a Rectangular Waveguide
The forward and backwardscattering off this slot in the T E , , mode will be determined with the aid of (3.40) and (3.41). First, it is a simple matter to show that
Second, the narrowness of the slot permits the assumption that
El
= 1XElX(C)
and thus
B,,
=
 ( n / ~ )cos ~ (nx , / a ) J ~ P , oab P~ 7
Il
V ( ( )ejfitocdl
in which V ( [ ) = wE,,(c) is the voltage distribution in the slot.
3.6 Theory of WaveguideFed Slot Radiators
In like manner, one can show that
CIO
 ( n / a ) k o s (nx, la)
=
jwpOBl
J:,
V(5) ejslocdc
The slot voltage distribution function V(c) depends on the manner in which the slot is excited. Let it be assumed that a matched generator is placed at a position z > I, so that the generator launches a TE,, mode of amplitude A , , in the guide propagation in the +Z direction. The TE,, modes scattered off the slot, of amplitudes B,, and C,,, cause no additional reflections because of the matched generator and the matched load. Despite the fact that the generatorlaunched TE,, mode incident on the slot has a phase progression across the slot, detailed analysis shows that, if 21 z 1,/2, the dominant component of V(5) is a symmetrical standing wave of the form
The similarity of (3.46) to (3.17) should be noted; it is as though the slot were essentially being excited at its center by a twowire line. With the approximation in (3.46) assumed, Equations 3.44 and 3.45 become B,,
=
C,, =
2 V , ( ~ / U )cos ~ (nx, la) sin [ k ( l j w ~ ~ B ~ ~ a b
01
cos B, ,i d~
nx B,, = C,, = 2Vm (COS~,~~cosk1)cosJ j w o ( P l,lk)ab a
(3.47)
It is important to observe that the assumed symmetry of V(c) resulted in the scattering off the slot being symmetrical, that is, B,, = C , , . This implies that the slot is equivalent to a shunt obstacle on a twowire transmission line. To see this, consider the situation suggested by Figure 3.1 1. A transmission line of characteristic admittance Go is shunted at z = 0 by a lumped admittance Y. The voltage and current on the
Fig. 3.11 A Shunt Obstacle on a Two.Wire Transmission Line
Radiation Patterns of Horns. Slots, and Patch Antennas
line are given by
The form selected for (3.48) is consistent with a matched generator being to the left of Y and a matched load to the right of Y in Figure 3.11. The boundary conditions are
which, when inserted i n (3.48), give
Thus B
=
C(the scattering is sjlrnrnetrical), and
The usefulness of (3.52) lies in the fact that, by analogy, if one can find the ratio 2Bl,/(Alo B l o )for the slot, one can then say that the slot has an equivalent normalized shunt admittance equal to that ratio. The slot is said to be resonant if Y/Go is pure real. Since no loss in generality results from taking A , , to be pure real, it follows that the resonant normalized conductance of the slot is given by
+
where B , , is perforce pure real also. What this implies is that, for a given displacement x l of the slot, it is assumed in (3.53) that the length 21 of the slot has been adjusted so that B l o is either in phase with, or out of phase with, A , (It will be seen subsequently that B , , is out of phase with A,,). Attention will now be restricted to this special case of a resonant slot.= The assumption of resonance permits a deduction from (3.53) via a power balance equa
,.
6The more general case of a nonresonant slot will be considered in Chapter 8.
3.6 Theory of WaveguideFed Slot Radiators
tion. The incident power is given by
Pi,,, =  me
S,.
( A ,,E
. 1, d S ,
,,,, x A:,Hh,.)
=
m4$l;~b
A , ,A:,
(3.54)
In like manner, one finds that the reflected and transmitted powers are
If use is made of the information that B , , pure real, then mpOP1Oab
4(nIal2
[A:,

B:,

(A,,
=
C , , and that all three amplitudes are
+
=
power radiated
But experiment shows that resonance occurs when 21 1,/2, in which case, if the upper wall of the waveguide is imbedded in a large ground plane, the radiated power is given by onehalf of (3.22). Thus
If V , is eliminated from (3.47) and (3.58), the result is G
G'
 
2 B l ~
A,,
7
B10
= 2.09
0 (cos P,,l ( P I
elk)
nx  cos kl)' cos2 2 a
(3.59)
When the substitution x = x ,  ( ~ 1 2 )is made in (3.59) and the approximation kl n / 2 is used, one obtains
in which x is the offset from the center line of the broad wall. Equation 3.60 is a celebrated result first obtained by A. F. Stevenson7; it indicates that the normalized conductance of a resonant longitudinal shunt slot in the broad wall of a rectangular waveguide is approximately equal to a constant times the square of the sine of an angle proportional to its offset. 7Stevenson, "Theory o f Slots."
Radiation Patterns of Horns. Slots, and Patch Antennas
Figure 3.12 gives a typical plot of experimental data showing resonant length of a longitudinal shunt slot versus offset. It can be seen that although the resonant length is offsetdependent, it stays close to the value 21 = L0/2 assumed in (3.60). ""&I Slot w i d t h = O.n"'C
1
111b11
a = O .900 i n c h *fin.
1
I
I
1
I
I
I
I
I
1
~ x ~ e r i m e n tpoints al determined by admittance measurements.
1
Experimental points f r o m radiation pattern measurements.
I
1
I
I
1
1
I
I
Slot displacement o f f waveguide centerline, x inches Fig. 3.12 Resonant Length versus Offset for Longitudinal Shunt Slot (After R. J. Stegen, "Longitudinal Shunt Slot Characteristics," Hughes Technical Memorandum No. 261, Nov. 1951, Hughes Aircraft Co., Culver City, California)
Figure 3.13 shows a plot of (3.60) versus experimental data. The agreement is quite good, serving to justify the approximations that were made in the theory. A more accurate analysis, which will give a better fit to the experimental data, will be presented in C h a ~ t e r8. The assumption that the voltage distribution in the slot is given by (3.46), plus the experimental information that 21 Z L0/2 for a practical range of offsets, means that the radiation pattern is insensitive to offset and the same as a halfwavelength dipole (with the polarization rotated 90"). Thus when the slot is imbedded in a large ground plane, the Hplane pattern is given by Figure 2.4 and the Eplane pattern is almost semicircular. But the power level in these patterns is governed by the offset of the slot through the factor sin2 nxla. Slot offset thus serves as a transformer, providing a means for controlling the radiation level through the amount of coupling to the incident feeding TE,, mode. This will prove to be a very useful feature when slot arrays are studied in Chapter 8. The procedure followed in this section can be repeated for the cases of inclined slots in the broad and narrow walls. The analysis, though lengthy, is not difficult if the foregoing is used as a guide, and these two cases are left as exercises.
3.7 Patch Antennas
Slot offset in inches Fig. 3.13 G,/Go versus Offset for Resonant Longitudinal Shunt Slot (After R. J. Stegen, "Longitud~nalShunt Slot Characteristics," Hughes Technical Memorandum No. 261, Nov. 1951, Hughes Aircraft Co., Culver City, California)
3.7 Patch Antennas
A radiating element with the attractive characteristic that it has a low profile is the patch antenna, illustrated in Figure 3.14. It consists of a thin metallic film bonded to a grounded dielectric substrate, and has the additional advantages of being lightweight, conformable, economical to manufacture, and easily wedded to solid state devices. The patch can be any shape, but the regular geometric shapes (such as rectangles or circular discs) are most commonly used. Feeding is achieved either via microstrip, as shown in Figure 3.15, or through use of a coaxial line with an inner conductor that terminates on the patch, as illustrated by Figure 3.16. The placement of the feed is important to the operation of the antenna. The flow of electromagnetic power in the patch antenna can be visualized easily. Guided waves transport the energy along the microstrip or coax to the feed point. The energy then spreads out into the region under the patch; some of it crosses the boundary of the patch, to be radiated into space. If the fields in this exit region can be
R a d ~ a t ~ oPatterns n of Horns, Slots, and Patch Antennas
Fig. 3.14 Components of a Patch Antenna (Feed Not Shown)
Fig. 3.15 A MicrostripFed Rectangular Patch
determined, equivalent sources may be placed on the boundary, from which the radiation pattern can be deduced. In practice, the permittivity of the dielectric layer is usually not great ( E / c ,< 4) and its thickness t is small, so the region under the patch behaves very much like a portion of a parallel plate transmission line. Waves that leave the feed point see almost an open circuit when they arrive at the perimeter of the patch and considerable reflection occurs, so that the fraction of the incident energy emerging to be radiated is small. This suggests that the patch behaves more like a cavity than a radiator, and pinpoints its principal disadvantagesthat it is not a highly efficient antenna and that
3.7 Patch Antennas
point
Fig. 3.16 A C o a x ~ a l l yFed C ~ r c u l a r
Disc Patch
feed
it is narrow band. The efficiency can be improved by using patches in arrays, but narrowbandedness puts a limitation on the applications. The assumption that the electromagnetic properties of the patch antenna can be deduced by viewing it primarily as a cavity (perhaps leaky cavity would be a better description) has been a fruitful one. Y.T. Lo and his coworkerss have been able t o obtain good correlation between experiment and a theory based on this assumption, both for pattern and impedance when t > 1, the phasors must fan out slightly beyond one sheet before they sum to zero and give a pattern null. Thus a tapered distribution suffers the penalty of some increase in beamwidth to the first null. However, this sacrifice is balanced by a compensating advantage. When the phasors have fanned out to occupy slightly more than one and onehalf sheets a secondary maximum is reached. But this maximum is contributed to principally by that third of the phasors representing the outermost elements, that is, by the phasors with the smaller amplitudes. Thus the secondary maximum, or first side lobe, is lower than it was in the case of the uniform current distribution considered earlier. Similarly, the tertiary maximum is lower, because it is contributed to principally by that fifth of the phasors representing the outermost elements, and so on. One reaches the
Fig. 4.6 Phasor Diagram for a Linear Array w i t h Tapered Excitation
123
4.3 L~nearArraysPreliminaries
important conclusion that side lobe level can be controlled by tapering the array excitation, at some cost in beamwidth. Several synthesis techniques for accomplishing this will be explored in Chapter 5. Note that if L ) 1,one can let 8:
=
8,
 A8; and
cos 8, cos A8;
8;' = 8,
+ sin 8, sin AO;
cos 8, cos AO',  sin 8, sin dB;'
+ A8': and write A.

cos 8,
=

cos 8,
= 
L
Since cos A8: s 1, cos A8;' g 1 , sin A8; r A&, and sin A8;' equations sum to give
1
L
AO;', these last two
This result is seen to embrace the earlier special case in which all the currents were equal and in phase, and which gave a beamwidth between nulls of 21/L. (In that case, 8, = n/2 and csc 19, = 1). Equation 4.26 indicates that the beam broadens as it is scanned off broadside, the beamwidth between nulls being governed by the projected length of the array transverse to the beam direction. Proceeding further with this case, when one permits the phasors to spread out so that they occupy one and onehalf sheets of the complex plane, secondary maxima occur, giving the first side lobe on each side of the main beam, and at a relative height of 13.5 dB. When the phasors occupy two full sheets, the second pair of nulls occurs. A spread over two and onehalf sheets produces tertiary maxima, and so on. Thus one finds a sequence of side lobes of steadily diminishing amplitudes, terminated when the limits ofreal space (8 = 0°, 180") are reached. A typical pattern computed from the magnitude of (4.23) is shown in Figure 4.7, with 2M 1 = 15, 411 = 112, and a, = 1112. One can observe that, with the beam tilted up, the side lobe heights are still symmetrical, but that there are more side lobes below the main beam than above. Once again, this is a plot of / a,(8) / in the halfplane $I = 0". The threedimensional pattern can be found by rotating Figure 4.7 about the Zaxis. Since Equation 4.23 is a series with a geometric progression, it can be summed to give
+
a,(e)
=
sin [(nL/L)(cos 8  cos 8,)] . sin [(nd/l)(cos 8  cos O,)]
t
Fig. 4.7 Polar Field Plot for a Linear Array Excited with Uniform Amplitude, Uniform Progressive Phase ; L = 7.51 ; Linear Scale
125
4.3 Linear ArraysPreliminaries
The proof is left as an exercise. Equation 4.27 is usually more convenient to use for calculating patterns such as the one shown in Figure 4.7. In this case also, if the phs~sorscan fan out so that adjacent phasors are 360" apart before the limits of real space are reached, a second main beam will occur. This will happen if kd cos 8'  u,
=:
&2n,
cos 8'
=
cos 8,
(Ild)
(4.28)
Equation 4.28 is seen to be a gereralization of (4.21). If 8, = n/2, one obtains 8' = a r c c o s ( i I / d ) as before. Howevt:r, if 8, assumes another value, and one wishes t o prevent the appearance of a second main beam, then it is apparent that the spacing d must be chosen so that
The second of these inequalities is more demanding and requires that
Equation 4.30 is the criterion for avoiding multiple beams in large scanning linear arrays. As an example of its use, one can see that if :he beam is scanned close to endfire, the elements must be spaced only onehalf wavelength apart if a second main beam is to be prevented from appearing. This discussion of a scanne8d beam can be readily enlarged to include the case of a tapered amplitude distribution together with a uniform progressive phase for the currents I,,.The main beam will still point at an angle 8, given by (4.24); it will be somewhat broader due to tapering, and the side lobes will be lower. (d) EXTENSION TO AN EVLN NUMBER OF ELEMENTSDIFFERENCE PATTERNS If the number of elements in an equispaced linear array is even, for instance, 2Nthe array can be laid out along the Zaxis so that the elements are at the positions *d/2, +3d/2, . . . , and the array factor in (4.10) becomes
From this point, the discussion of'the preceding parts of this section can be repeated with little change. If all the currents are equal and in phase, a,($)is represented by a family of phasors lying at the angles *y/2, &3y/2, . . . , with y = kdcos 0 as before. These phasors fan out to give, in sequence, nulls and side lobe heights, and the conclusions about side lobe levels, null positions, possible multiple beams, all still prevail, now with L = 2Nd. Earlier arguments can be repeated for tapered excitation and when a uniform progressive phase is introduced.
Linear Arrays Analysis
What adds to the interest in linear arrays with even numbers of elements is the opportunity to excite the two halves of the array out of phase with each other, something which is neither convenient nor desirable with an odd number of elements because of the awkward presence of the middle element. As an example of this possibility, let all the currents be equal in amplitude, but let I  , = I,, for all n. Then (4.31) gives a family of phasors that appear in the complex plane as shown in Figure 4.8.
Fan direction ,f /
\,Fan direction \
Fig. 4.8 Phasor Diagram for Difference Mode; Uniform Excitation
For 8 = n/2, half of these phasors lie along the positive real axis, the other half lie along the negative real axis, and the sum is zero; that is, there is a null in the pattern at broadside. As 8 departs from 7112 toward 0 (or n), these phasors fan out into the upper (or lower) half of the complex plane. When (2N  l)y1/2 g 37114, their phasor sum S is a maximum, given by S I j J T N . (Proof of this is left as an exercise). When (2N  l)y1/2 z 272, their phasor sum is zero. As the phasors fan out further onto the second sheet of the complex plane, this summation process is repeated, with only onethird of the phasors effectively participating. Thus secondary maxima of approximate heights fj f l N / 3 , . . . , are reached. A typical difference pattern for uniform but asymmetrical excitation is shown in Figure 4.9. Twin main lobes are found straddling 8 = n/2, with symmetrically decaying side lobes; the first pair is 9.5 dB below the level of the main lobes. Patterns of the type found in Figure 4.9 are more readily computed from
This result is obtained by summing (4.3 1) for the special case In= I_,= I , ,for all n. The high side lobe level (9.5 dB) of this pattern is due t c the choice of equal amplitudes. If a tapered current distribution of the type shown in Figure 4.10 were selected, the side lobe level could be reduced. The reader might wish to confirm this by sketching the phasor diagrams as they would appear for 8 values corresponding to
4.3 Linear ArraysPreliminaries
Fig. 4.9 Polar Field Plot for a 14Element Linear Array, d = 112, w i t h Uniform Asymmetric Excitation ; Broadside Difference Pattern ; Linear Scale
Fig. 4.10 Bar Graph of Taperecl Current Distribution t o Give Difference Pattern w i t h Reduced Side Lcbes
successive lobe maxima. A synthesis procedure for determining the proper taper to produce a specified side lobe level will be presented in Chapter 5. A uniform progressive phase can be given to the current distribution, with the two halves of the array still excited out of phase with each other. The effect on Equation 4.31 is to replace k d cos 8 by kd(cos 8  cos e,), with the current ratios pure real and 8, the null angle between the two principal lobes. The current distribution can, in this case, also have a taper to reduce side lobes. Patterns such as those shown in Figures 4.5 and 4.9 form a useful pair in radar applications. With the two halves of the array fed in phase (the sum mode), one obtains a pattern with a single main beam; this is useful for acquiring a target, but not too useful for telling exactly where the target is. However, if the target is close
Linear Arrays: Analys!s
enough and the excitation is switched to the mode in which the two halves of the aperture are excited out of phase (the difference mode), then, unless the target is exactly in the null between the two principal lobes of the difference pattern, a return signal will be detected in the radar receiver. This signal can be used either to tilt the array mechanically or to introduce a uniform progressive phase shift in the excitation, the result being to place the target in the central null of the difference pattern. The sensitivity of this process is high, so the target's angular position can be determined with considerable accuracy. 4.4 Schelkunoff's Unit Circle Representation
The development of the previous section can be reinforced and extended with the aid The synthesis techof an extremely useful formulation due to S. A. Schelk~noff.~ niques t o be considered in Chapter 5 also benefit from Schelkunoff's method of representation. Consider an equispaced linear array of N 1 elements (N can be even or odd),
+
laid o u t a l o n g t h e Zaxis. F r o m (4.7), t h e a r r a y f a c t o r c a n b e written i n t h e f o r m
In (4.33), a uniform progressive phase factor a, has been factored out of the current distribution and shown explicitly because many applications involve current distributions of this type. However, no loss in generality results from this, since the ratio I J I , appearing in (4.33) can still be complex. If one lets w = ejr (4.34)
ry
=
kd cos 6

(4.35)
a=
Equation 4.33 can be converted to the form a0(.)
=
5 LW. I. "0c ,vnI" IN
.=o I ,
=
(4.36)
IN
from which
From this, by the fundamental theorem of algebra,
5s. A.
Schelkunoff, "A Mathematical Theory of Linear Arrays," Bell Sysrern Tech. J., 22
(1943), 80107.
4.4 Schelkunoff's U n ~ C t ~ r c l eRepresentat~on
129
Equations 4.37 and 4.38 rweal several interesting things. First, I f(w)j differs from the magnitude of a,(@ by only a multiplicative constant and thus can serve as a surrogate for the array factor. St:cond, the array factor can be represented as a polynomial in w of degree one less than the number of elements in the equispaced array. Third, this polynomial has N roc~tswhich are linked to the array excitation, since the current ratios comprise the coeEcients of the polynomial. The placement of these roots in the complex wplane can be related to the field pattern in real 8 space, as will be seen in what is to follow. In this manner, the analysis and synthesis of array patterns due to equispaced arrays can be tied to a study of the properties of polynomials, a dist nct asset for the antenna designer. To develop this approach further, observe that as 8 varies in real space from 0 to n, the definitions in (4.34) and (4.35) require that y vary from y, = kd  a, t o yf = kd  a, and that w trace out a path along a unit circle in the complex plane, as illustrated in Figure 4.1 1. The total excursion of w is from eJr*to e"1, proceeding clockwise as 8 goes from 0 to n. Thus y, (read ystart) and yf (read y/finish) mark the initial and terminal points of the wexcursion, the angular extent of which is 2kd rad~ans.
Fig. 4 11 Schelkunoff's Unlt Circle
Further inspection of (4.38) reveals that if the roots w, are placed on the unit circle, in the range of w, then 1 f(w,)I = 0, and a pattern with N nulls will result. Alternatively, if all the roots w, are placed of the unit circle, or at least outside the range of w, a pattern devoid of r~ullswill b: produced. Both types of patterns have their uses. Synthesis of nullfree patterns is a more difficult topic and generally is beyond the scope of this introductory treatise, though some discussion of it will be found in Chapter 5. In what follows, attention will be focused on situations in which
L ~ n e a rArrays : Analysis
the roots w, have all been placed on the unit circle through proper choice of the current distribution. (a) UNIFORMLY EXCITED BROADSIDE ARRAYS If the excitation currents have equal amplitudes and a common phase, (4.35) reduces to y/ = kd cos 8 and (4.36) simplifies to
DeMoivre's theorem permits the conclusion that f (w) has roots at the positions
(Note that the root w = 1 is excluded because of the factor in the denominator). Thus a uniform amplitude/equiphase excitation of the equispaced array does put all the roots on the unit circle. As has already been seen in Section 4.3, this results in a pattern with lobes interspersed by nulls.
An illustration of (4.40) is given in Figure 4.12 for the case of a fiveelement array. The roots are found at the positions &2n/5 and 5 4 4 5 . The value of I f(w)I can be found by taking the product of the four distances d l , d,. d,, d,, which is an example of the use of (4.38). As the point w moves along the unit circle (which corresponds to permitting 0 to vary in real space), these four distances change, as does their product. Whenever w coincides with one of the roots w,, the distance d, = 0 and / f (w,) I = 0; that is, a null has been encountered in the array factor. This presupposes that the range of w includes the roots w,. For example, if d = 112, then y / , = n and tyf = n; all four of the roots are within the range of w.
Fig. 4.12 Root Positions on a Schelkunoff Unit Circle for a Uniformly Excited FiveElement Linear
131
4.4 Schelkunoff's U n ~ C~rcle t Representation
As w moves clockwise from e j K through e j O to e  j K , the product d,d,d,d, traces out half a side lobe, a null at w,, a full side lobe, a null at w,, a main beam, a null at w,, a side lobe, a null at w,, and finally another half side lobe, as illustrated in Figure 4.13. The nulls in real space can be determined from y, = 2nm/(N 1) = kdcos em, which is in agreement with the earlier results of Section 4,3.
+
Fig. 4.13 Polar Field Plot for a Uniformly Excited FiveElement Linear Array, d = ,I/Broadside :!; Sum Pattern; Linear Scale
If only a quick sketch of the pattern is needed, a graphical construction using the unit circle of Figure 4.12 (suitably enlarged) can be helpful. The height of a side lobe or main lobe occurs when w is approximately halfway between roots. The product d,d, . . . d, when w is at such halfway points gives the relative lobe heights. This can be determined with reasonable accuracy if care is taken in measuring the distances. Knowledge of these lobe heights and the null positions 8, is all that is needed to be able to produce a decent polar representation of the field pattern. The reader might wish to try this for the fiveelement array by constructing an enlarged version of Figure 4.12 and checking that the side lobe heights are  13.5 dB and 17.9 dB (in agreement with the conclusions of Section 4.3), and that the pattern resembles Figure 4.13. If d = 1, then y, = 2n and. yf = 2n; w ranges two full revolutions around the unit circle. The result for a fiveelement array is the pattern shown in Figure 4.14. If one wishes to avoid these extra main beams at 9 = 0, n, a suitable restriction for a fiveelement array would be to have v, coincide with w, and let w range more than
L~nearArrays: Analys~s
Fig. 4.14 Polar Field Plot for a Uniformly Excited FiveElement Linear Array, d = 1; Broadside Sum Pattern; Extra Main Beams at EndFire; Linear Scale
one full revolution around the unit circle, past w, and on to y/,, which is allowed to coincide with w,. In the more general case of N 1 elements, this means choosing
+
Equation 4.41 can be viewed as the maximum element separation for a uniformly excited broadside array if multiple main beams are to be avoided. For N large, it 1 = 5, the elements should be agrees with the result found in Section 4.3. For N no further than 0.81 apart. The pattern for this case is shown in Figure 4.15.
+
4.4 Schelkunoff's U n ~ Clrcle t R~presentat~on
Fig. 4.15 Polar Fielcl Plot for a Uniformly Exc~tedFiveElement Linear Array, d = 0 . 8 1 ; Broadside Sum Pattern; Maximum Spacing for Multiple Beam Avoidance; Linear Scale
It is instructive to compare the patterns of Figures 4.13 and 4.15. Both are for uniformly excited fiveelement arrays, and each has a single main beam a t broadside. But Figure 4.15 shows a narrower main beam and three additional side lobes. This is due to the increased length of the array (412 versus 2.512). (b) BROADSIDE ARRAYS WITH LOWERED SIDE LOBES The fact that (4.38) can be written in the form 1 f(w)j
N
=
Il dm,with dm the distance between w and u),, m=1
has already been noted in the discussion in this section. With w approximately halfway between the successive nulls w m and w,,,, the product of these distances gives the relative height of the mth side lobe. It follows that if these two roots are brought closer together, the height of this side lobe will be reduced. For broadside arrays, if the heights of all the side lobes are to be reduced, the roots must cluster closer around n, indicating that the main beam region on the unit circle (from w , t o w,) must be enlarged. In other words, the main beam is broadened as the price paid to reduce the side lobes. This tradeoff has already been noted in Section 4.3. As an example of this effect, consider again the fiveelement equispaced array. With uniform amplitude/equiphase excitation, this array could be represented by the Schelkunoff unit circle shown in Figure 4.12. For element spacings of d = 12/2,12, 0.812, the patterns of Figures 4.13 through 4.15 were obtained. All of these patterns have the same side lobe topography, since they arise from a common unit circle diagram. The innermost side lobes are at 1 1.9 dB and the next set is at 13.7 dB.
Linear Arrays: Analys~s
Suppose it is desired to produce patterns in which all the side lobes are at 20 dB. This would require displacing the four roots of Figure 4.12 so that they are closer together and clustered closer to n. This design problem can be solved graphically by trial and error. At present, the roots are at i72" and &144". Suppose the new positions &87" and & 149" are tried. (This places the roots 62" apart, instead of 72".) When a large unit circle is constructed with roots placed in these positions, the product of distance measurements gives the innermost side lobe at  18.5 dB and the other side lobes at 21.3 dB. This suggests that the roots w , and w , have been shifted too close together, but that the shift in w , and w 4 might be just about right. One could continue to try revised root positions, perhaps 1 8 7 " and 147", thus converging to the root positions that will give the desired 20 dB level for all side lobes. The interested reader might wish to pursue this and demonstrate that the proper root positions are f89" and 145.5". With the correct root positions known, one can return to (4.38) and write
If this result is compared to (4.37), it can be observed that the relative current distribution is 1
1.6
1.95
1.6
1
The central element is seen to be most strongly excited. The distribution has a symmetric taper, consistent with the discussion in Section 4.3. ( c ) SCANNED ARRAYS The only change, if an equispaced linear array is to have a uniform amplitude/uniform progressive phase excitation, is that one needs to return to the more general definition of y/ given in (4.39, with a, = kdcos 8, and 8, the central angle of the main beam. The Schelkunoff unit circle is unaffected; all the roots are where they were before. Only the starting and ending values of y are altered. They are now Y,
=
kd(i  cos e,),
ylf =
kd(i
+ cos 0,)
(4.43)
The total excursion is still 2kd, and the height of the main beam still occurs at y = 0. However, the fact that the main beam is scanned raises anew the question about multiple main beams. As an example, one can return to the case of the fiveelement array, for which the unit circle of Figure 4.12 applies. If the main beam is to point at 8 = 120" (that is, 30" beyond broadside) and if the element spacing is d = A12, then = 3 d 2 , $, =  d 2 , and the pattern is as shown in Figure 4.16. On the other hand, if the element spacing is d = 1,then y/, = n, y/, = 371, and the pattern assumes the shape shown in Figure 4.17. If one wishes to avoid the presence of a second main beam, it is
*,
4.4 Schelkunoff's U n ~ C~rcle t Representation
Fig. 4.16 Polar Field Plot for a FiveElement Linear Array Exc~tedw i t h Uniform Amplitude, Uniform Progressive Phase, d = 1212; Sum Pattern with Main Beam Scanned t o Oo = 120"; Linear Scale
apparent that the spacing should be chosen such that
as before, but now (4.43) imposes the requirement that
or that
L ~ n e a rArrays' Analys~s
Fig. 4.17 Polar Field Plot for a FiveElement Linear Array Excited with Uniform Amplitude, Uniform Progressive Phase, d =?, ; Sum Pattern with M a i n Beam Scanned to Bo = 120"; Extra M a i n Beam at B = 60"; Linear Scale
which is the criterion for avoidance of multiple beams if a uniformly excited linear array is scanned. For N large, this agrees with (4.30). For the fiveelement array, with 0, = 60°, the maximum spacing should be 81/15.
(d) DIFFERENCE PATTERNS For equispaced linear arrays of an even number of elements, f ( w ) has an odd number of roots. If one of these roots is placed at UI, = 0 and the others are arranged in complex conjugate pairs as suggested by Figure 4.18, a symmetrical difference pattern will result, with twin main beams straddling a null at O,, and with the same side lobe topography on both sides of the main beams. A special case of this occurs when I,, =  I _ , = 1, for all n, for which it has been seen that the pattern is given by (4.32). In addition to the central null at 0 = ~ 1 2(4.32) , indicates nulls at the positions
4.4 Schelkunoff's U n ~ Circle t Representat~on
21
cos 0,
=
mn
Since L = 2Nd, with 2N the number of elements in the array, and since y this result can be converted t o the form
w i t h Roots Placed t o Give Symmetric Difference Pattern
=
kd cos 0,
a
Thus the roots are equispaced on the unit circle, but they are all double roots except for the single root a t y = 0.This is an inefficient root placement. For example, the pattern shown in Figure 4.9 is for a 14element array, d = 112, with uniform asymmetric excitation, and shows only two and onehalf side lobes on each side of the twin main beams. I t has already been remarked that this pattern does not have
impressively low side lobes (the innermost pair of side lobes is only 9.5 dB below the height of the twin main beams). If the roots were repositioned appropriately on the unit circle, so that they occurred singly, there could be five and onehalf side lobes on each side of the pattern, with a concomitant lowering of the side lobe level at no expense in terms of broadening the twin main beams. Further lowering of the side lobe level could be accomplished by clustering the roots closer to y = n (with one root held at y = 0 to insure a difference pattern). This would be at the expense of some broadening of the twin main beams. Such designs can be achieved graphically by trial and error in the manner already described for the sum pattern. They can also be obtained by a synthesis technique that will be presented in Chapter 5. ( e ) SUPERGAIN ARRAYS Since the total excursion of w along the unit circle is 2kd, the intriguing possibility arises that one could make the interelement spacing d
138
Linear Arrays. Analys~s
smaller and smaller, simultaneously repositioning the roots on the unit circle so that they always remained within the range of w, and in such a way that the pattern in real space was unaltered. In this manner, for example, a sum pattern with a main beam of a prescribed beamwidth and side lobes of prescribed heights could be generated by an equispaced linear array of a specified number of elements, but with the total length of the array arbitrarily small. The current distribution in this compressed array will need to be investigated. It clearly will not be uniform, because the earlier analysis of that case revealed that, for a uniformly excited array, the nulltonull beamwidth of the main beam of a sum pattern is 212/L. (See Equation 4.1 8 and the related discussion.) As a specific illustration of this possibility of a reducedlength array, consider once again the fiveelement equispaced linear array, excited to give a broadside sum pattern with symmetrical side lobes. If the roots are placed on the unit circle at the positions ty, = *72", *144", it has already been seen that the excitation will be uniform. For d = 1212, the pattern will have a main beam plus one and onehalf side lobes on each side of it. Suppose next that the interelement spacing is reduced to a fraction f of 112 and that the roots are simultaneously repositioned to be at & y / , , i y 2 , with y, = 72f degrees. This will clearly keep the roots within the range of w. As a generalization of (4.42), f(w)
= (w2 
2w cos tyl f 1)(w2  2w cos Iy2
+ 1)
and thus the current distribution is 1
+ cos 2ty,) 2(1 + 2 cos ty, cos 2ty1) 1 2(cos ty, + cos 2ty,)
2(cos ty,
+
+
(4.46)
+
$ When y/, = 72", all of these currents are unity and (1)2]Rdenotes the ohmic losses, with R some appropriate ohmic representation of the resistivity and shape of an element. The field strength at the peak of the main beam is measured by the sum of the currents and therefore the total radiated power can be represented by (5)2K, with K a factor that depends on pattern shape. Assume that, with tyl = 72", the ohmic losses are 1 % of the power radiated. Then (5)2K = 100(5R) or K = 20R. Now assume that ty, = lo, that is, there has been a 72fold contraction in the length of the array and in the root placement on the unit circle. For this case, (4.46) gives, for the current distribution,
The ohmic losses have become 70R and the field strength at the peak of the main so the radiated power is 0.13764 x 1012K. The ratio of beam is only 0.371 x the power radiated to the ohmic losses is
139
Problems
T h e ohmic losses, which were assumed t o b e only 1 % of the radiated power a t d = 112 spacing, a r e in contrast a trillion times a s large a s t h e radiated power a t d = 11144 spacing. Even with a modest reduction in spacing t o d = 114, t h e o h m i c losses a r e f o u r times a s large a s the radiated power. This simple example serves t o illustrate the drastic penalty o n e must pay in loss of efficiency if reduction of length is contemplated f o r linear arrays. Further study shows t h a t mechanical a n d electrical tolerances become severe a n d frequency bandwidth is sharply curtailed a s the interelement spacing is contracted. F o r all these reasons, supergaining (as this process is called) has proven t o b e impractical.
REFERENCES AMITAY, N., V. GALINDOISRAEL, and C. P. Wu, Theory and Analysis ofphased Array Antennas (New York: WileyInterscience, 1972), Chapter 1. BACH,H., and J. E. HANSEN,''Uniformly Spaced Arrays," in Antenna Theory, Part I, ed. R. E. Collin and F. J. Zucker (New York: McGrawHill Book Co., Inc., 1969), Chapter 5. ELLIOTT,R. S., "The Theory of Antenna Arrays," in Microwave Scanning Antennas, Vol. 2, ed. R. C. Hansen (New York: Academic Press, 1966), Chapter 1. JORDAN, E. C., and K. G . BALMAIN, Electromagnetic Waves and Radiating Systems, 2nd ed. (Englewood Cliffs, N.J.: PrenticeHall, Inc., 1968), Chapter 12. WOLFF,E. A,, Antenna Analysis (New York: John Wiley and Sons, Inc., 1966), Chapter 6.
PROBLEMS 4.1 Begin with Equations 4.1 and 4.5 and show that a,(e, I$) can be written as the product of the array factor (4.10) and the element factor (4.1 1). 4.2 Assume that a uniformly spaced linear array of 2 N 1 elements is uniformly excited. 1 is very large, show that the height of the first side lobe occurs when the corIf 2 N :responding 2 N 1 1 phasors uniformly occupy one and onehalf sheets of the complex plane. D o this by letting the position of the outermost phasor, N y , be a variable. Approximate the phasor diagram by a continuum density of phasors and show that this first side lobe is 13.5 dB below the height of the main beam.
+
4.3
Show that Equation 4.27 is a transformation of equation (4.23).
4.4
Assume that a uniformly spaced linear array of 2 N elements is uniformly excited, but in the difference mode. If 2 N is very large, show that the height of the principal lobe occurs when the corresponding 2 N phasors have fanned out such that the outermost one is at the position (2N l ) y / 2 E 3x14. D o this by letting the position of the outermost phasor be a variable. Approximate the phasor diagram by a continuum density of phasors and show that the pair of side lobes that is closest in is approximately 9.5 dB below the level of the twin principal lobes. 
Linear Arrays: Analysis
4.5 4.6
Show that Equation 4.32 can be derived from (4.31) when I, Show the equivalence of Expressions 4.27 and 4.39.
=
I_,
= I,, for
all n.
4.7
A sixelement equispaced linear array is to be given uniform amplitude/equiphase excitation. Construct a suitably large Schelkunoff unit circle (say r = 4 inches) and mark the root positions. If d = 4 2 , find the null positions in &space. Use the fact that j f (w) 1 = dl . . .dS to determine the relative lobe heights, and make a rough polar plot of the field pattern.
4.8
For the sixelement array discussed in the preceding problem, use trial and error to determine the proper root positions to yield an array pattern with a single main beam and all side lobes at 20 dB. Find the corresponding current distribution.
4.9
For the sixelement array discussed in the two preceding problems, use trial and error to determine the proper root positions to yield a difference pattern with all side lobes at 20 dB. Find the corresponding current distribution. Note that this current distribution is substantially different from what you would get by reversing the phase of half of the array excitation found for the sum pattern in Problem 4.8.
4.10 It has been shown in the text that, for an equispaced linear array of 2N elements, a
difference pattern will result if the nulls are placed on the unit circle at the positions = 2nrnlN with rn = 0, il, . . . , i N 1. Show that this gives I, =  I  , = 1, for all n, for the current excitation, and that the pattern is represented by Equation 4.32.
W,

4.11 Design a sixelement equispaced array to give a sum pattern with its main beam at end
fire and all side lobes at 30 dB. This can be done graphically by trial and error. Then deduce the maximum spacing between elements if not even a trace of a second main beam is to be present at reverse endfire. 4.12 It has been shown in the text that if an equispaced linear array of 2N elements is excited uniformly but asymmetrically, a difference pattern results with the roots occurring on the unit circle at positions y/, = 2nrn/N, rn = 0 , i1 , . . . , 5 N  1. All roots are double except y o . This is an inefficient root placement. One way to correct this is to place the roots singly at $,,, = 2 r n ~ l ( 2 N l ) , m = 0, =kN  1 , . . . , 1. Show that if this is done, all side lobes are the same height as the twin main beams and that all currents in the array are zero except the end two, which are equal and opposite. (The result is called an interferometer pattern.) 4.13 With reference to the preceding problem, another possible root placement that avoids double roots is to let W, = n m / ( N l), rn = 0, 12, &3, . . . , & N . Show that this gives larger yregions on the unit circle for the twin main beams than it does for the side lobes. Plot the polar field patterns for the case 2 N = 14, d = 21.2, and compare with Figure 4.9 of the text. Find the current distribution and compare with Figure 4.10 of text.
+
5.1 Introduction
In the previous chapter the basic analysis of equispaced linear arrays was presented under the assumption that a known current distribution existed in the array and that one desired to find the resulting array pattern. A variety of practical distributions were assumed (uniform amplitude with and without uniform progressive phase, tapered amplitude with and without uniform progressive phase, the two halves of the array excited out of phase) and it was discovered that useful sum and difference patterns were caused by these distributions. In conjunction with the introduction of the Schelkunoff unit circle, the subject of synthesis was even touched on when a graphical trialanderror technique was suggested in which root placement could be systematically altered until a desired pattern was achieved. In the present chapter the synthesis problem will be addressed directly. In synthesis, one begins by specifying the desired array pattern. Since the discussion here is restricted to linear arrays,' the desired pattern must be a function of 8 alone and not 4, that is, in the form a,(@or so(@. But the class of functions a,(@or is large. It includes sum patterns with uniform side lobes, with symmetrically tapered side lobes, and with asymmetric side lobes. It includes difference patterns with the same variety of side lobe topographies. It includes patterns with neither nulls nor side lobes. For all of these patterns, the synthesis question is basically the same: Given a,(@or 5,(0), what is the requisite current distribution in an equispaced array? This question will be answered in succeeding sections of this chapter for some of the more widely used classes of prescribed patterns. Dolph's technique, which uses Chebyshev polynomials to deduce discrete current distributions that yield sum patterns with uniform side lobes, will be taken up first. Taylor's procedure, which accomplishes basically the same result but for continuous line sources, will also be 'This restriction will be lifted in Chapter 6 with the introduction of planar arrays.
L ~ n e a rA r r a y s : S y n t h e s ~ s
presented. A perturbation method, which can be used to modify either a Dolph or Taylor pattern in order t o produce sum patterns with arbitrary side lobe topographies, will be introduced and then extended t o applications involving difference patterns. The Woodward technique, which synthesizes patterns requiring nullfilling, will also be described. The chapter is not devoted entirely to synthesis procedures. The concepts of halfpower beamwidth and peak directivity of a linear array pattern are introduced and applied specifically t o the case of sum patterns, since these two quantities often form a key part of the design specifications on which the pattern synthesis must be based.
5.2 Sum and Difference Patterns Many applications of linear arrays involve the need to produce sum and difference patterns with the main beam of the sum pattern pointing at an angle go, with the twin main beams of the difference pattern straddling d o , and with both patterns exhibiting a symmetrical side lobe structure. When there are 2N elements in the array, equispaced by an amount d, the array factor can be written in the form2
Under the above stipulations, all the current amplitudes in (5.1) can be taken as pure real. For the sum pattern, I,,= I, and (5.1) becomes
5
s(8) = 2 n =
For the difference pattern, I,
cos [(2n

1)
1
=
and (5.1) takes the form
I,
+
An array with 2 N 1 elements (an odd number) is not suitable for the creation of a difference pattern because of the awkward presence of the central element. However, as seen in Chapter 4, it can be used to produce a sum pattern. Under the assumption of symmetrical side lobes, the pattern from such an array is given, as a reduction from Equation 4.13, by S(0)
=
1
+ 2 2 # cos[?n(~)(cos 8 n=l

cos 8,)
0
2For convenience a,(@ is used, but the results apply equally well for 5,(8).
5.3 DolphChebyshev Synthes~sof Sum Patterns
143
A major class of synthesis problems can be stated in terms of these equations. If a sum pattern with a specified side lobe topography is desired, how does one determine the current distribution I,,//, in (5.2) o r I,,//, in (5.4) to achieve the desired result? Or, if a difference pattern with a certain side lobe topography is desired, how does one find the current distribution I n / /in, (5.3) to bring this about 1Several sections of this chapter are concerned with answers to these questions.
5.3 DolphChebyshev Synthesis o f Sum Patterns The discussions of Chapter 4 revealed some useful information about the design of equispaced linear arrays, excited so as to give an array factor with one main beam plus side lobes (sum pattern). This information can be summarized as follows. 1. For 2 N f 1 elements, if the 2N roots are placed on the unit circle in complex conjugate pairs, a symmetrical sum pattern will result. If the positions of these root pairs are adjusted, the side lobe heights can be altered. T o reduce the level of the side lobes, the root pairs need to be clustered closer to ry = n,at the expense of broadening the main beam. 2. For 2N elements, if the 2N  1 roots are placed on the unit circle with one root at ry =  n and the remainder in N  1 complex conjugate pairs, a symmetrical sum pattern will result. If the positions of the root pairs are adjusted, the side lobe heights can be altered. T o reduce the level of the side lobes, the root pairs need to be clustered closer to y = n, at the expense of broadening the main beam. 3. With the roots occurring in complex conjugate pairs, f ( w ) is a polynomial with pure real coefficients. These coefficients appear in symmetrical pairs in the polynomial, thus evidencing the fact that the current distribution in the array is symmetrical in amplitude. With these observations as background, the problem of proper positioning of the root pairs can be addressed. If one argues that side lobes occur in spatial regions in which it is desirable t o suppress radiation, and assumes that the suppression of a// side lobes is equally important, then an optimum design is one in which all side lobes are at the same height. The reduction of a single side lobe further than this common level could only be at the expense of additional broadening of the main beam, and would deny the assumption that the region of this side lobe is no more important than the region of any other side lobe. This problem of seeking the proper root positions (and thus the proper array excitation) to give a sum pattern with uniform side lobes at a specitied height was solved by C. L. Dolph in a classic paper.' T o d o this, he took advantage of a useful property of Chebyshev polynomials, which are solutions of the differential equation
3C. L. Dolph, "A Current Distribution for Broadside Arrays Which Optimizes the Relationship between Beamwidth and Side Lobe Level", Proc. IRE, 34 (1946), 33548.
Linear Arrays : Synthesis
It is shown in Appendix C that, if the index m is an even integer 2N, then a solution to (5.5) is
If rn is an odd integer 2N  1, then
with
( :) the binomial coefficient r!/s!(r  s)!. Both (5.6) and (5.7) can be put in the
revealing form T,,,(u) = cos(m cosI u ) = cosh ( m cosh' u) =
(1)"' cosh(m cosh' juJ)
1sus1 u> 1 u< 1
(5.8)
which is easily verified by substitution in (5.5). Thus T,(u), with m an integer, is a function that oscillates in a cosinusoidal manner in the range 1 u 1 < 1 and then rises hyperbolically in 1 u ( > 1. It is this property of the Chebyshev polynomials that makes them so useful in antenna array design. Figure 5.1 shows the typical features of
Fig. 5.1 Chebyshev Functions (Reprinted from Microwave Scanning Antennas, Volume 2 , R. C. Hansen, Editor, Courtesy 1966 Academic Press.) of Academic Press, Inc.
0
5.3 DolphChebyshev S y n t h e s ~ sof Sum Patterns
145
Chebyshev functions. For m = 2N there is symmetry about u = 0, with N root pairs in I u 1 < 1 . Here T2,(1) = T z N (  1 ) = 1. There are no roots outside I u 1 = 1 , so just 1 the function rises rapidly, with steeper slope for larger values of 2N. beyond I u I For m = 2N  1 , there is antisymmetry about 8 = 0 , with a root at u = 0 and N  1 root pairs in / u 1 < 1, where Tz, , ( I ) = TzN,( 1) = 1. There are no roots beyond I u J = 1 , so the function rises rapidly, with steeper slope for larger values of 2N  1. These plots reveal why the Chebyshev polynomials were ideal for Dolph's purpose. If the variable u can be made t o correspond in some manner t o the real angle variable 9 , so that an appropriate segment of Tm(u)can be made to relate to a,(9), then a pattern with uniform side lobes will result. T o develop this correspondence, one can return to the basic equation for the array factor, assume a uniform progressive phase and symmetrical amplitude distribution, and write as alternate forms of (5.4) and (5.2)

in which, as before, y/
=
kd(cos 9

cos 9,)
Equation (5.9) applies for an odd number of elements and is equivalent to a polynomial of order 2N in the variable cos ( ~ 1 2 ) Equation . 5.10 applies for an even number of elements and is equivalent to a polynomial of order 2N  1 in the variable cos (y//2).Thus if one selects the transformation u
= U , COS W
(5.12)
2
then (5.6) and (5.9) can be equated for arrays with an odd number of elements, and (5.7) and (5.10) can be equated for arrays with an even number of elements. What this transformation accomplishes can be appreciated by returning to Figure 5.1. As 9 ranges from 0 to 9 , to n, and as yl ranges from y/, = kd(1  cos 9,) to zero to y / , = kd(l cos a,), u will range from u, u, cos (r/l,/2) to u, to uf = uo cos ($,/2). The pattern trace is shown by the arroweddotted path in Figure 5. la. If u, is chosen so that Tm(uo)= 6, with 20 log,, b the desired side lobe level, then a pattern will result consisting of a main beam a t the relative field height b plus a family of side lobes all a t the height unity. Dolph's design procedure can now be articulated. One begins by selecting the number of elements, which determines the degree of the Chebyshev polynomial that is to be used (m is one less than the number of elements). Next, it is necessary to find u, from Tm(uo)= 6, with b fixed by the desired side lobe level. Then, from (5.8), the roots of Tm(u)can be determined readily and are given by

+
up =
+ cos
(2p 
Linear Arrays : Synthesis
When these values are inserted in (5.12), the corresponding root positions y, on the unit circle can be computed. At this point f(w) is known in factored form and can be multiplied out to give the current distribution. As an example, consider a fiveelement array that is to be excited to give a sum pattern with all side lobes at 20 dB. Then it is T4(u) which should be used, and T,(u,) = b = 10. From (5.8), cosh (4 coshI u,)
=
10
which is satisfied by u, = 1.2933. Also, from (5.8), one finds that the roots of T4(u) are +0.9239 and *0.3827. Thus, from (5.12), the roots on the unit circle are at the positions
If the calculation that produced Equation 4.42 is repeated (where this same problem was being solved by a graphical trialanderror method), the present more accurate y values give
with the coefficients of the various powers of w representing the relative current distribution. With the uniform progressive phase factor a, embedded in the definition of w (compare Equations 4.34 through 4.39, the coefficients
represent the relative magnitudes of the currents in the fiveelement array. This implies that the DoIphChebyshev distribution is the same in magnitude regardless of where the main beam points. All that changes when the main beam pointing direction 8, is altered is the uniform progressive phase, which must be attached to the amplitude distribution. If the avoidance of extra main beams is of concern, the precautions noted in Section 4.4 must be observed. The interelement spacing should be chosen so that the yexcursion on the Schelkunoff unit circle only traverses the main beam region once. The case of an endfire DolphChebyshev distribution is worth special mention, and the example cited above can serve as a typical illustration. To place the main beam at endfire, y, = 0" and y f = (360°  88.82") = 27 1.18". The interelement spacing should not exceed
if even a vestige of a second main beam cannot be tolerated at reverse endfire. With this spacing, the uniform progressive phase is a= = kd = 2.369 radians and the
5.3 DolphChebyshev Synthes~sof Sum Patterns
endfire DolphChebyshev current distribution for this array is
It is possible to determine the Dolph current distribution for the general case without the intermediate steps of finding the roots up, yp, and wp followed by multiplying out the factors off (w). T o do this, one needs to insert (5.12) in (5.6) or (5.7) and make use of the relations
c0s2.
1
W 2

5 ( 2nn q l ) c o s ( 2 q 
2"2 
q = ~
1)Y 2
A derivation of these two formulas can be found in Appendix D . In Equation 5.14, E, = I if q = 0, otherwise 6, = 2.
With the use of (5.13) through (5.15), the expressions for the Chebyshev polynomials become
The coefficients of cos 2my/2 are in the same ratio in (5.9) and (5.16) for all rn, and likewise the coefficients of cos (2m  l)y/2 are in the same ratio in (5.10) and (5.17) for all m. Therefore the relative current distribution for an array with 2 N f 1 elements 1s
and, for an array with 2 N elements is
Although they look formidable, (5.18) and (5.19) are simple to program. T o use them, one still needs to start with the knowledge of the number of elements in the array and the desired side lobe level, so that u, can be determined. For arrays with a large number of elements, the time saved in using (5.18) or (5.19) is considerable when compared to the procedure that first determines the root positions.
5.4 Sum Pattern Beamwidth of Linear Arrays
Discussions in Chapter 4 have revealed that the angular extent of the main beam in a sum pattern is inversely related to the length of a linear array. Further, it has been seen that, for a given length array, the main beam broadens as the side lobe level is lowered. In synthesis problems, these relationships must be approached from the other end. Often the design specifications include statements about the desired beamwidth of a sum pattern, as well as the side lobe level. The designer must not only determine the requisite current distribution, but also the array length and number of elements, both chosen to avoid multiple beams (dl1 should not be too great) and supergaining (LIA should not be too small). Because of the importance of beamwidth as a design specification, it is desirable to sharpen these earlier discussions by introducing a more precise definition of beamwidth. The one normally used is that the beamwidth is the angular separation between 8 directions at which the radiated power density is down to onehalf its 1 element^,^ laid out symmetmaximum value. For an equispaced array of 2N rically along the Zaxis, let 8 = 8,  8, be this beamwidth, in which 8, and 8, are the two values of 8 which satisfy the relation
+
In (5.20), the current amplitudes I, are assumed to be pure real, the current phase progression is governed by a,, d is the interelement spacing, and 8, is the pointing angle of the main beam. The amplitude distribution I,,/I, can be described by a Fourier series, namely,
in which P is the highest spatial harmonic needed to represent the distribution. Attention will be restricted to sum patterns in which the side lobe topography is symmetric, which means that I,/I,, is also symmetric, and thus that a, = a_, is pure real for all p. Since a, = kd cos 8,, insertion of (5.21) in (5.20) gives
=
2
.=  P
sin {(zL!A)[cos 8,  cos 8, i(PAIL)]) sin {(nd/A)[cos8,  cos 8, (pl/L)]j
+
4A11 the results obtained in this section are equally valid for 2N elements.
(5.22)
149
5.4 Sum Pattern B e a m w ~ d t hof Linear Arrays
+
in which L = (2N I)d is the length of the array. It has been shownS that, for large arrays with conbentional distributions, Equation 5.22 can be transformed to
where
is a substitution variable from which the beamwidth can be deduced. Several cases will now be considered. CASE I : UNIFORM DISTRIBUTION This is the simplest case of all and extremely useful as a reference. Only a, has a value and (5.23) yields two solutions for K such that sin Kn = 0.707Kn and Kn = & 1.392, and therefore
from which it follows that the halfpower beamwidth is given by @
=
8,

8
cos
cosI [cos 8,  0.443
1 1 L
!
cos 8, t 0.443 L
in the range 0 < Q0 < d 2 , 8, 2 0. As the main beam is scanned from broadside (8, = n/2) to endfire (8, = 0), a cross section of the beam takes on a succession of positions, as indicated in Figure 5.2. As the conical beam closes toward endfire, a position is reached a t which 8, = 0, and from this position to endfire, there is no halfpower point on one side of the beam. For this reason 6 ,= 0 can be called the scan limit.Equation 5.25 will not give a real value for 8 , beyond this limit. When the endfire position is reached, the concept of beamwidth once again takes on meaning. Equation 5.26 is still valid and one can write
The beamwidths given by (5.27) and (5.28) are plotted in Figure 5.3 as functions of array length and scan position. These curves will prove useful beyond the present case of uniform amplitude excitation, as will be seen shortly. sR. S. Elliott, "Beamwidth and Directivity of Large Scanning Arrays", Appendix B, Microwave Jorrrnal, 6 (1963), 5360. Also in Microwcrve Scanning Antennas, ed. R . C. Hansen, Vol. 2 (New
York : Academic Press, 1966), Chapter I .
Endfire
Fig. 5.2 Conical Beam Shape versus Scan
Each of these expressions for beamwidth (Equation 5.27, which is valid to within one beamwidth of endfire, and Equation 5.28, which is valid at endfire) has a n approximate form when L >> A. Using smallangle expansions, one obtains
0 B
A csc 8, L
=
0.886
=
2[0.886+]
(at or near broadside)
(5.29)
1,2
(at endfire)
For L 2 512, Equation 5.29 is in error by less than 0.2% at broadside and is in error by less than 4 % when the main beam has been scanned to within two beamwidths of endfire. For L 2 51, Equation 5.30 is in error by less than I CASE 2: DOLPHICHEBYSHEV DISTRIBUTION It has been shown in the literature6 that when the current distribution is chosen so that the pattern is equivalent
o,.
6R.S. Elliott, "An Approximation t o Chebyshev Distributions," IEEE Trans. Antennas Propagat., API 1 (1963), 7079.
5.4 S u m Pattern Beamw~dthof L~nearArrays
Array length L/X in wavelengths Fig. 5.3 HalfPower Beamwidth versus Linear Array Length and Scan Angle for Uniform Excitation (Reprinted from Microwave Scanning Antennas, Volume 2. R . C. Hansen, Editor, Courtesy of Academic Press, Inc. 1966 Academic Press.)
0
to the Chebyshev polynomial T,,(u,
It follows from (5.31) that (2N level. Thus
cos y//2), then the Fourier coefficients are given by
+ ])a,

T,,(u,)
=
b, with 20 log,, b the side lobe
For large arrays, and for side lobe levels in the range from 20 decibels to 60 decibels, only a, and a , are significant in determining the beamwidth. If (5.8) and (5.31)
Linear Arrays : Synthesis
are used in conjunction for p (2N
=
1, it is found that a , is given quite precisely by
+ l)a, = cosh [(arccosh b)2  n2]li2
(5.33)
With all other Fourier coefficients insignificant in the calculation of beamwidth, (5.23) becomes for this case
Inspection of (5.34) reveals that the solution for K when a , # 0 differs from the solution for K when a , = 0 only through the presence of the ratio a,/a,. But (5.32) and (5.33) indicate that a,/a, depends on side lobe level but not on scan position nor the number of elements in the array. Thus it becomes convenient to introduce a beambroadening factor f which is simply the ratio of the halfpower beamwidth when a given array is excited DolphChebyshev to the halfpower beamwidth when it is uniformly excited. Computations of beamwidth from (5.34), with a , and a , given by (5.32) and (5.33), can be compared to the computations that produced Figure 5.3. The result is the fcurve shown in Figure 5.4. The fnumber can be interpreted as the cost in beambroadening to convert all the side lobes to a common height and reduce them to a specified level. The extended utility of Figure 5.3 can now be appreciated. If one wishes to determine the beamwidth of a linear array of normalized length L/R excited to give
15
20
25
30
35
40
45
50
55
60
Side lobe level in dB Fig. 5.4 BeamBroadening versus Side Lobe Level for Linear Arrays with DolphChebyshev Excitation
5.5 Peak Directivity of the Sum Pattern of a Linear Array
153
a sum pattern with the main beam at O,, a reading from Figure 5.3 will give the halfpower beamwidth for uniform amplitude excitation. For DolphChebyshev excitation with a specified side lobe level, that reading should be multiplied by the fnumber read from Figure 5.4. One can also work backwards with the aid of these two figures to deduce the array length needed when the scan position, beamwidth, and side lobe level are specified for a DolphChebyshev sum pattern. 5.5 Peak Directivity of t h e Sum Pattern of a Linear Array The peak directivity D ( O o )of the sum pattern produced by a linear array is a frequently encountered design specification. It can be deduced as a special case of the general definition of directivity, given as equation ( 1 . 1 6 0 ) and repeated here for convenience: D(O,4)
1
=
4Rr2
[
@(O, 4 )
(5.35)
l ' @ ( O f , Q ' ) r 2 sin 0 ' dO1 dm1
The power density in the sum pattern of a linear array is given by7
The array and element factors that appear in ( 5 . 3 6 ) have previously been defined by Equations 4 . 6 through 4.1 1. If the array is large, the element patterns broad, and the side lobe level of the sum pattern low, the principal contribution to the integral in the denominator of ( 5 . 3 5 ) is in the neighborhood of the main beam. When this is the case, the factor a e , , a ~ ,a,,,a$,,can be brought out in front of the integral and given its value at ( e , , 6 ) . If this is done,
+
D(O0)=
q, 1 /nTj:w>a(~)5:(~)
sin B do d$
which further simplifies to
2 , , ( e 0 ) ,,*( 00)
D(oO)= j n " , a ( ~ ) \ b (sin ~) B ~ O
(5.38)
Equations 5.37 and 5 . 3 8 represent the peak directivity of the a r r a y factor, or what is the same thing, the peak directivity of the sum pattern when the element factor is assumed to be isotropic. They are approximate and cannot be used with good accuracy for small arrays. 7Equation 5.36 assumes a type I (actualsource) array. The development can be duplicated exactly when the array is represented by an equivalent magnetic current distribution.
154
L ~ n e a Arrays: r Synthes~s
Continuing with the assumption that the array is large (LIIZ )) I), one can return to the definition 4.35 and write
Use of these relations in (5.38) gives
,~ reduces very simply to If d = 112 or any multiple t h e r e ~ f (5.41)
which is a most interesting formula in several respects. The directivity given by (5.42) is a measure of the coherence of radiation from the linear array. The numerator is proportional to the total coherent field, squared, whereas the denominator is proportional to the sums of the squares of the individual fields from the various elements. Furthermore the peak directivity, as expressed either by (5.41) or (5.42), is seen to be independent of scan angle. On the face of it this seems surprising, since it has already been observed that the main beam broadens as it is scanned away from broadside, a manifestation which usually signifies lowered directivity. However, for a linear array, as the conical beam is scanned toward endfire, the "cone" occupies a smaller solid angle in space, an effect that just cancels the beambroadening. Although Equation 5.42 is independent of scan angle, it is not independent of current distribution. If one uses the Fourier series description of the excitation embodied in (5.21), it is evident that
so that
For halfwave spacing, L
=
(2N
+ 1)12/2,so that (5.42) can be rewritten as
T h i s restriction will be lifted shortly.
155
5.5 Peak D ~ r e c t ~ vof ~ t the y Sum Pattern of a L n e a r Array
> I and 6/12 0, the other pointing in the halfspace z < 0.
Planar Arrays: Analysis and Synthes~s
Almost invariably, the element pattern will be selected to give negligible radiation in the halfspace z < 0 (through use of a ground plane, for example). There is then left a single main pencil beam, pointing in the direction (O,, (5,), with 0, and (5, satisfying two equations that can be deduced from (6.7) and (6.8), namely,
For given spacings d, and d,, and given interelement phase shifts a, and a,, Equations 6.9 and 6.10 give a unique pointing direction (O,, (5,) in z > 0. Radiation patterns from planar arrays, exhibiting this feature of a single pencil beam, are called sum patterils.
Equation 6.10 can be used as the criterion for avoiding the situation that a, and a, contain conical main beams that do not intersect. This situation would just be reached if sin2 0, = 1. Thus, the elliptical relation
limits the range of a, (or a,) for specified values of kd,, kd,, and a, (or a,). I n the remainder of this analysis, the existence of a single main pencil beam will be assumed.
( c ) BEAMWIDTH OF THE SUM PATTERN Since the significant side lobes are in the two cones defined by (6.7) and (6.8), these cones are the pattern cuts which should be taken to determine the side lobe level. However, the profiles of the main pencil beam obtained in these two cuts are not, in general, due to two orthogonal slices through the pattern. (For example, if the pencil beam lies close to the XYplane at (5, = n/4, these two cuts are almost coincident.) Thus it is desirable to define beamwidth in another fashion, one which will reveal more information about the structure of the pencil beam. In what is to follow, it will be shown that the  3 dB contour of the pencil beam is approximately elliptical. The beam cross section is suggested in Figure 6.2. At a given large distance r from the planar array, the size and shape of this elliptical contour are dependent on the pointing direction (Oo, (5,), as is the tilt of the axes of the ellipse. The two orthogonal planes which contain, respectively, one or the other of the ellipse axes, plus the origin, may be used to define the pattern cuts in which the beamwidth is measured. These two orthogonal measurements of halfpower beamwidth then serve to specify the major and minor diameters of the elliptical contour, and thus give an indication of the size and shape of the beam cross section. From (6.6), the central point in the main beam has the intensity
6.2 Rectangular Grid Arrays: Rectangular Boundary and Separable D i s t r ~ b u t ~ o n
z
Fig. 6.2 Orthogonal Beamwidthsof a Pencil Beam (Reprinted from Microwave Scanning Antennas. Volume 2, R. C. Hansen. Editor, Courtesy of Academic Press, Inc. 1966
0
Academic Press.)
In a nearby direction 8, a(e,
+ 68,$,
$ 695, the intensity will be down by 3 dB if
+ s e , 4, 4 s 4 ) = o.707a(e0,4,)
Nr
=
N,
=
c I,I,
0.707~
Nv
Nv
9 I, enp(jmkd,[sin(8, + 68) c o s ( ~ ,+ 695)  sin 8, cos $,I} x 9 I. exp(jnkdy[sin(8, + 68) sin($, + 6 0 sin 8, sin $,I)
(6.12)
N,

Nu
For large arrays, 68 and 695 are small, and (6.12) reduces to
0.707
Nz
Nv
N,
Nv
C C I,I,
NZ
==
C I,
exp(jmkd,[cos 8, cos 95, 68  sin 8, sin 95,695]}
N,
x
9 In exp(jnkdY[cos8, sin 95, 68 + sin 8, cos 0, d$]}
(6.13)
Nv
The right side of (6.13) consists of a family of phasors, symmetrically spread out in the complex plane, much as in the case of the linear array described in Section 4.3.
Planar A r r a y s : Analysis a n d Synthesis
This sum is to be 0.707 times the sum of the phasors when they are aligned. For conventional distributions, the phasors do not need to fan out very far for this to occur. The outmost phasor normally does not reach a position of more than 7712 radians. Thus, if one lets R,
=
kd,[cos 8, cos 4,68  sin 8, sin 4,641
R,
=
kd,[cos 8, sin 4 , 6 8
and
+ sin 8, cos 4, 641
+
the phase factor exp[j(mR, nR,)] can be expanded in a power series that will converge reasonably rapidly even for the largest values of m and n. When this is done, (6.13) becomes
Since the distributions I, and I,,have been assumed to be symmetrical, all summations which contain m or n to an odd power are zero. Thus, through third order,
The two sums that appear on the right side of (6.14) can be evaluated by considering the situation in which the beam lies in either the XZ or the YZplane. When the XZplane is chosen, 4, = 0, and (6.14) becomes
When the pencil beam lies in the XZplane, it is caused by the intersection of: (1) a conical beam that makes an angle (7712)  8, with the Xaxis; and (2) a conical beam that makes an angle (7712) with the Yaxis. The pattern cut in the XZplane is therefore identical to the one that would be obtained if there were only a single linear array laid out along the Xaxis. But this pattern contains two points that lie in the 3 dB contour of the pencil beam, namely the points (8 = 8, &$8,, 4 = 0), where 8, is the halfpower beamwidth of the Xdirected linear array when its conical main beam makes an angle (n/2)  8, with the positive Xaxis. For this reason, the couplet (68 = &Ox, 64 = 0) must satisfy (6.15), which gives 0.586
NZ
Nv
C C I,I,  N . Nu
Nz
=
(&kd, cos 8, 8JZ C N,
Nv
C mZI,I,, Nv
(6.16)
6.2 Rectangular G r ~ dArrays: Rectangular Boundary and Separable Distr~bution
For the special case 8,
=
0, (6.16) yields
in which @,, is the broadside beamwidth of the Xdirected linear array. Similarly, with the pencil beam placed in the YZplane, one finds that
If these two results are inserted in (6.14), rearrangement gives
Since (68, 64), and thus (R, R,), defines the set of pointing directions in which the field is 0.707 times the peak value, Equation 6.19 can be viewed as describing the 3 dB contour on the pencil beam. T o see this more clearly, let u and vaxes be erected along lines of longitude and latitude on the sphere of radius r, as shown in Figure 6.2. Then
Substitution of these variables in (6.19) gives
+
(Ucos 8, sin $,, v cos 4,)' (u cos 8, cos 4,  v sin 4,)' 4(r8*o/2)2 (r@Yo/2)2

(6,20)
This can be recognized as the equation of an ellipse in (u, v)space. Introduction of the axes u' and 71' via the rotation p (compare with Figure 6.2), such that u
=
u' cos p
+ v' sin p,
v
=
u ' sin p
+ v'
cos
p
permits (6.20) to be written in the form
In (6.21) d,, and d,, are the diameters of the ellipse measured along its two principal axes. The rotational angle P is given by tan 2P
==
(1
2 cos 8, sin 24, $ 8 ~ o ) / ( @f o 6'y2,)] sin2 8,
+ cos2 8,) cos 24, $ [(dl;, P
(6.22)
At a constant zenith angle 8,, rotates smoothly through 90" as 4, changes through 90". Individual expressions for d,, and d,, are unwieldy, but their product is given by
204
Planar Arrays: Analys~sand Synthes~s
the simple expression d,,d,, = r 2 sec 8 , 8xoi9yo
Thus the area of the ellipse is independent of 4,. If the pencil beam lies in the XZplane in the direction (8,, O), the u' and v'axes are aligned with the u and vaxes and (6.20) gives d, = d,. = r sec 8 , Ox,,
dy = d,,, = reyO
(6.24)
In the pattern cuts containing the uaxis or the vaxis, the halfpower beamwidths are therefore 8,, =
6 = Q x 0 sec 8 , r
and
8. = 4 r
=
Oy0
(4,
=
0)
Similarly, if the pencil beam lies in the YZplane in the direction (8, n/2), the ufaxis points in the vdirection and the v'axis is aligned with u. For this case (6.20) gives do = d,. = r8,0 (6.26) d, = d,, = r sec 8 , 8,,, In the pattern cuts containing the uaxis or the vaxis, the halfpower beamwidths are now 8.
8,,,
=
d. = Oy0 sec 8 , r
and
8.
=
% = 8,, r
( 4 =)
(6.27)
To use either Equation 6.25 or Equation 6.27, one needs first to determine Oy0,and the zenith pointing angle 8,. For uniform distributions, Ox, and 8,,
can be determined by using L,/A and L,,/A and reading the appropriate beamwidth off the broadside curve of Figure 5.3. For DolphChebyshev distributions, these beamwidths need to be modified by the $factor read from Figure 5.4. After this it is a simple matter to determine 8, and 8,. It is useful to define an areal beamwidth B by the relation
Through use of (6.23) this becomes
I:():( B
= 
 = 8x08yosec 0,
The areal beamwidth, which is a measure of the area inside the 3 dB contour of the pencil beam cross section, is seen to be independent of 4,. As one would expect, it has the same functional dependence on 8 , that the projected aperture does. The general effect of scanning a pencil beam can be constructed as suggested in exaggeration by Figure 6.3. At broadsidebroadside, the cross section of the beam
6.2 Rectangular Grid Arrays Rectangular Boundary and Separable D ~ s t r ~ b u t ~ o n
F i g . 6.3 Beam Shape versus Scan Posit~onfor a Pencil Beam (Reprinted from Microwave Scanning Antennas, Volume 2, R. C. Hansen, Editor, Courtesy of Academic Press, Inc. 0 1966 Academic Press.)
is approximately elliptical (position PI)with dimensions proportional to L;' and L;'. As the beam scans in the XZplane, the beam cross section elongates in that direction (position P,).Scanning in the YZplane causes elongation in the other beam dimension (position P,).For a constant angle 8 , from the zenith, as the beam ismoved from 4 , = 0 to 4 , n/2, the two halfpower beamwidths smoothly change and the elliptical cross section smoothly rotates, these two effects combining in such a way that the areal beamwidth remains constant. Thus for narrow pencil beams from large rectangular grid arrays with rectangular boundaries, if the distribution is separable, the entire subject of beamwidth can be based on the results previously obtained for linear arrays. The relations derived in this section are quite good to within several beamwidths of the limiting condition of no main beam a t all, defined by (6.1 1).

( d ) PEAK DIRECTIVITY OF T H E SUM PATTERN The peak directivity of this type of planar array (compare with Section 1.16) is given by
D 
4na(eo, 40)@*(807 $0) a(8, sin 8 dB dm
:I ioZzm)a*(B,4 )
in which it is assumed that the element pattern is such as to eliminate the total pattern in the halfspace 8 , : n/2 but is broad enough to be ignored in 8 ( n/2. It has been
Planar Arrays: Analysis and Synthesis
shown in the literature3 that, for large arrays which are not scanned closer than several beamwidths to endfire, an approximate reduction of (6.30) is D
=n
cos 8,
2LJA
C (apiao)"
2Ly/IZ Cbq/bo>2
in which the Fourier coefficients a, and bp describe the aperture distribution in the X and Ydirections, and L, = (2N, l)d, and Ly = (2N l)dy are the dimensions of the array. This equation has the simple interpretation that the maximum peak directivity of a planar array is
+
+
D = nD,Dy cos 8,
(6.32)
in which D, and Dy are the directivities of the two linear arrays. The factor cos 8, accounts for the decrease in projected aperture with scan. Unlike the directivity of a linear array, which was found to be independent of scan angle, the directivity of a planar array is dependent on the zenith coordinate 8,. However, it is independent of the azimuthal coordinate 4,. Because of the form of (6.31), many of the remarks that have been made about the directivity of linear arrays can be applied as well to this type of planar array. For aperture distributions with a uniform progressive phase and a symmetric amplitude, (6.31) indicates that maximum directivity results from the choice of uniform excitation. DolphChebyshev distributions suffer from a gain limit for very large arrays, and the curves of Figure 5.5 are applicable to such planar arrays.
(e) A RELA TION BETWEEN BEAM WIDTH AND PEAK DIRECTIVITY As in the case of linear arrays, one finds from (6.29) and (6.31) that, for this type of planar array, peak directivity and the reciprocal of areal beamwidth depend linearly on the area of the planar aperture. Elimination of L,Ly/IZZfrom these two expressions results in
where f, and fy are the beam broadening factors for the linear arrays of 2N, f 1 and 2Ny f 1 elements that comprise the two dimensions of the array. The quantity in brackets is unity for a uniform distribution, and is essentially unity for a DolphChebyshev distribution until gain limiting sets in. Thus, for these practical aperture distributions,
S. Elliott, "Beamwidth and Directivity of Large Scanning Arrays", Appendix D, M i c r a R. C. Hansen, vol. 2 (New York: Academic Press, 1966), Chapter 1. 3R.
wave Journal, 7 (1964), 7482. Also, Microwave Scanning Antennas, ed.
207
6.2 Rectangular G r ~ dArrays: Rectangular Boundary and Separable Distribution
in which the areal beamwidth is now expressed in square degrees rather than square radians4 It is important to remember that, in (6.34), both quantities are measured at the same tilt angle 8,.
(f) SUM AND DIFFERENCE PATTERNS To this point in the discussion, the array factor given by Equation 6.6 has been interpreted under the assumption that the normalized current distributions I, and I,, were symmetrical, the result in z > 0 being a sum pattern consisting of a pencil beam and a family of side lobes. In such circumstances, an alternate expression for the array factor is
+ 2 c I. cos m y , NZ
~ ( 8 , $ )= ( I
NY
(6.35)
m= 1
in which
y,
=
kd, sin 8 cos $  a,,
y,
=
kd, sin 8 sin $  a,
(6.36)
For the case of an even number of elements in each dimension, 2N, by 2Ny, (6.35) is replaced by
In this latter case, two difference patterns can be generated, one by causing I, = I,, while leaving I,, = I,, the other by doing the reverse. The first condition gives ~ ~ (+)8 =, 4 j [mE= I I, sin
2m
1
( T ~ x ) ]
[9 I. cos n=
2n
1
( T ~ y ) ]
(6.38)
I
while the second condition gives
I n t h e q5
=
O0, 180" p l a n e , a),(Q, q5) gives t h e p a t t e r n of an Xdirected l i n e a r a r r a y ,
l cos 4x2)
=
sin (nx, la) sin v(x,) cos [(Plk)v(x I )I  cos v(x I )
(8.104)
Equation 8.104 will insure the desired slot voltage distribution. In addition, to get an input match,

For a 0.900 inches and v = 9.375 GHz, one finds that P/k these values in (8.104) and (8.135), simultaneous solution gives
=
0.714. If one uses
8.1 5 The D e s ~ g nof Linear WaveguideFed Slot Arrays
41 3
x,
=
x,
=
0.082 in.
21,/A = 214/A= 0.487
g ( x , ) = g(x,)
x,
=
x3
=
0.180 in.
21,/1
g(x,)
= 213/A= 0.502
=
0.093
= g(x,) =
0.407
(8.106)
These are the starting lengths and offsets of the slots in the array. If there were no mutual coupling, they would give the desired slot voltage distribution and an input match. With these assumed lengths and offsets, one is able to compute Z i j for the equivalent dipole array, using Equations 7.155 and 7.156. It is found that the initial values of mutual impedance are Z,,
=
0.37 Pj8.39
Z,,
=
0.67
+j0.47
Z,,
=
1.49 $ j1.28
Z,,
=
2.88  j7.81
From this, one can compute the following initial values of Z,b:
Since K ,
=
480 for this waveguide size and frequency, Equation 8.85 becomes
::
480 f,' [480f,'lg(x,)h(y,)l 1 2::
y Go
The objective is to select ( x , , y , ) and (x,, y,) so that
under the .pattern restriction that
and under the input admittance restriction that
With the aid of Equations 8.99 through 8.102, a trialanderror solution of (8.109)
The Design of Feed~ngStructures for Antenna Elements and Arrays
through (8.112) is found to be x , = 0.086 in.
x,=0.176in.
y,
=
1.0125
y2=1.0099
21,112
=
0.4933
2I2/L=0.5058
Yf Go
=
0.0988 (8.1 13)
'," = 0.4010
G,
These results can be iterated. The lengths and offsets in (8.1 13), when used in (7.155) and (7.156), will produce an improved set of Z,,. However, if this is done, one finds that the new Z: are
and these values are so close to the previous set that negligible further change will be found if the iteration process is carried further. Thus (8.113) can be accepted as the design with external mutual coupling taken into account. A comparison of (8.1 13) and (8.106) reveals that there is a 5 % change in the offset of slots 1 and 4, a 24 % change in the offset of slots 2 and 3, and a 1 % lengthening of slots 2 and 4. These changes may seem small enough that one could argue in this application that mutual coupling be neglected. But the effect of these changes on aperture distribution and input admittance are significant, as shall be seen in Section 8.17. The smallness of these changes can be traced to the fact that these slots, being in a common waveguide, are almost endfire to each other, so that mutual coupling is lower and.falls off faster than when the slots are broadside. One can anticipate a bigger problem with mutual coupling in planar arrays, as will be seen in Section 8.16. As a final comment on the design of linear slot arrays, the modern trend is to make the bdimension of the waveguide smaller to save on weight and depth. However, this lengthens the slots and places adjacent ends of successive slots closer together, thus increasing the mutual coupling, and making it even more essential that its effect be included.
8.16 The Design of Planar WaveguideFed Slot Arrays When a family of waveguidefed linear slot arrays is arranged as shown in Figure 8.39, a pla.nar array results. This introduces a new variable into the design procedure. The individual waveguides containing the radiating slots (hereafter referred to as branch line waveguides) may be excited in a variety of ways. Perhaps the most common is to run a main line waveguide transversely across the back of the array and use coupling slots to energize the branch lines. Another method is to use a corporate feed, consisting of a set of Tjunctions which serve to split the power in a sequence of steps down to the level of the individual branch lines. But whatever method is used, the mode voltages in the various branch line waveguides can be adjusted in
8.1 6 The D e s ~ g nof Planar Wavegu~deFedSlot Arrays
Fig. 8.39 A Planar Array of Longitudinal Shunt Slots
relative level by the coupling mechanism. This is an additional parameter that can be exploited in the design procedure. It is convenient to go to double subscript notation in the design of planar arrays. Thus the basic design equations become
in which K, and K , are still given by (8.86) and
in which
is the normalized slot length, and (x,,, 21,") are the offset and length of the mth slot in the nth branch line waveguide. In most respects, the design of a planar slot array proceeds exactly as for a linear slot array, which was described and illustrated in the previous section. One assumes that every branch line array is resonantly spaced (this restriction will be lifted in Chapter 9). This implies that the mode voltages are given by V,,
= (
I)"'V,
(8.1 18)
The D e s ~ g nof Feed~ngStructures for Antenna Elements and Arrays
in which V,, is the reference mode voltage in the nth branch line. For this reason (8.1 14) can be rewritten in the form
Y2"  K , Go
I f,, I sin v ( x , , ) G
V"
since the alternation in direction of offset causes the compensating relation
When (8.1 15) and (8.1 19) are used to design a planar array, one must assume an initial set of slot lengths and offsets in order to compute an initial set of ZA, values. But one must also assume an initial branch line mode voltage distribution in order to use (8.1 19). It may be that, as the design proceeds and a series of iterations converges on a final set of slot lengths and offsets, one finds that the set of offsets in a particular branch line is inconveniently small or large (outside the trustworthy range of experimental design data). This can be altered by a change in the V , distri
bution. If one wishes to increase the average offset in a branch line without increasing the slot voltage level, this can be accomplished by lowering the coupling to that branch line, which serves to lower that particular branch line mode voltage. Considerable adjusting back and forth is usually needed in the course of the design in order to insure that the final spread of offsets in the branch lines is in an optimum range, and that all the coupling coefficients between the main line feeding structure and the branch lines are also in an optimum range. As a consequence of this adjustment, it is most unlikely that the sum of the normalized active admittances in any branch line is unity; the branch lines do not have to be matched in order to achieve a match in the main line. A simple illustration of the design of a planar slot array, one which dramatically demonstrates the strong effect of mutual coupling, involves the twobyfour slot antenna shown in Figure 8.40. It was desired to excite this array so that all eight slot
Fig. 8.40 A TwobyFour Array of Longitudinal Shunt Slots
41 7
8.1 6 The D e s ~ g nof Planar W a v e g u ~ d e  F e dSlot Arrays
voltages would be equal and in phase. Additionally, in order to insure that all slot offsets would be in the favorable dynamic range, the condition
was imposed. Waveguide dimensions a = 0.924 inch, b = 0.123 inch, and t = 0.025 inch were used, together with a slot width of 0.064 inch. Experimental design data in the form of curves similar to those found in Figure 8.35 through 8.37 was available at v = 8.930 gigahertz and polynomial expressions were fitted to the data, resulting in expressions akin to Equations 8.99 through 8.102. Equations 8.1 15 and 8.11 9 were then used to determine the proper slot lengths and offsets, with V, = V , taken to be the mode voltage distribution. The procedure was identical to the one outlined for the onebyfour array in Section 8.15, except that a computer program was used instead of hand calculations. The results are shown in Table 8.8. TABLE 8.8
Slot Number mn
Lengths and offsets for twobyfour slot array Offset x,, (Inches)
Length 21,. (Inches)
A study of this table of slot lengths and offsets reveals several interesting and surprising things. First, there is a 2: 1 range in slot offsets. (Were one to ignore mutual coupling or assume it was the same for each slot, all offsets would be the same.) Second, no slot in this array is selfresonant; each slot is detuned appropriately to make the individual active admittance resonant. Third, there is a quadrant 1 to quadrant I11 and quadrant I1 to quadrant 1V symmetry to the lengths and offsets, but no symmetry about the Xaxis nor about the Yaxis. This can be traced to nonsymmetrical effects caused by staggering the offsets and always occurs in array designs that yield symmetrical patterns. And fourth, it is clearly evident that the presence of broadside neighbors has substantially increased the effect of mutual coupling. At first it might seem puzzling, for example, that the required offsets of the 2, 1 slot and the 3, 1 slot are so radically dissimilar when the slot voltages are to be the same. However, a return to Figure 8.40 indicates that the environment of the 2, 1 slot is significantly different from the environment of the 3, 1 slot.
The Design of Feed~ngStructures for Antenna Elements and Arrays
An experimental determination of the active input admittance to each branch line waveguide yielded the results
which were 5 % and 3 % from the design values. The experimental Hplane pattern at 8.930 GHz, with the array embedded in an 8inch by 10inch ground plane, is shown as the solid curve in Figure 8.41. The theoretical pattern (dotted curve) is also shown for comparison.
Angle f r o m broadside, degrees Fig. 8.41 The HPlane Pattern of the TwobyFour Slot Array Depicted i n Figure 8.40; Comparison of Theory and Experiment
8.17 Sum and Difference Patterns for WaveguideFed Slot Arrays; Mutual Coupling Included
If Y:/Go is eliminated from (8.84) and (8.85), one obtains for a linear slot array
With the aid of (8.88) and (8.89), this can be rewritten as
x V: sin k~,,,z,,, (2)v,,f, N
rn= 1
=
(8.124)
8.17 Sum and Difference Patterns for Wavegu~deFedSlot Arrays: Mutual Coupling Included
41 9
Equations 8.124 can be identified as a set of simultaneous linear equations (with the slot voltages as independent variables) that can be put in the matrix form
in which it has been recognized that V ,f, = V 1 f,1, with V the reference mode voltage in the linear array. The common factor (KZ/Kl)Vhas been suppressed in going from (8.124) to (8.125). If the length and offset of every slot is specified, all the impedance terms in the 2matrix of (8.125) are known, as are all the elements If, ( in the column matrix. An inversion of (8.125) will give the slot voltage distribution for this set of lengths and offsets. As an example of the use of (8.125), consider again the fourelement linear slot array analyzed in Section 8.15. Because of symmetry considerations, for that case (8.125) takes the form
which reduces to
For the starting lengths and offsets given in (8.106), the mutual impedances are given by (8.107) and
As a consequence of (8.128), the matrix (8.127) becomes
+
r(71.07 j0.47) (5.16  j7.11)1 V ; sin k l , l i(5.16 j7.1 I) (74.73  j7.81)1[Y; sin k l , ]
0.1184l = [0.2564]
(8.129)
The Des~gnof Feed~ngStructures for Antenna Elements and Arrays
Inversion gives V ; = 0.001415 111.6" 
V,S = 0.003293 17.9" 
(8.130)
so that the slot voltage ratio is
The reader will recall that the desired ratio is 2.00 10". One can see that, whereas there is only a 5 % or less error in the starting lengths and offsets, this results in a 16.5 % magnitude error and a 3.7" phase error in the slot voltage distribution. For the slot voltages given in (8.130) and the selfimpedances and mutual impedances listed in (8.128) and (8.107), one can determine that
Use of (8.85) reveals that
and thus Yr; = 0.9604 + j0.1438 C
i= I
Go
= 0.97 18.5" 
The input admittance has a susceptive component which is 15 % of the conductance. There is a mismatch of 3 % in magnitude and 8.5" in phase. Were one to repeat these calculations for the final offsets and lengths given in (8.1 13), it would be discovered that the slot voltage distribution is correct in both amplitude and phase. Confirmation of this assertion is left as an exercise. For a planar slot array, elimination of Y;,/G0 from (8.1 15) and (8.119) gives V;. sin kl,, Z:,
=
(8.134)
a result which is identical to (8.123) except for the use of double subscript notation and the added feature that the mode voltage V , may differ from branch line to branch line. Equation 8.134 can be expanded to give
with the constant (K,/K,) suppressed. If the slot lengths and offsets are known, the impedance matrix appearing in (8.135) is known, as are the column matrix elements V , 1 f,, 1. Inversion of (8.135) will give the slot voltage distribution.
421
8.1 7 Sum and D~fferencePatterns for Wavegu~deFedSlot Arrays Mutual C o u p l ~ n gIncluded
A common use of (8.135) is in application to planar arrays, which are divided into four quadrants and fed to produce sum and difference patterns. In such cases it is convenient to use the indexing scheme shown in Figure 8.42. Because of the quadrant I to quadrant 111 and quadrant I1 to quadrant IV symmetry in such arrays (already noted for the twobyfour array discussed in Section 8.16), it is unnecessary to apply (8.135) to the entire array. If there are MbyN slot modules,23for the sum pattern Vjq = V h + p , , +  q and (8.1 35) becomes M/2
N
p1
ql
C C
+ Z&+
V;, sin klpq(Z,P:
lp,N+lq
)
=
V"l fm.
I (8.136)
2 (Zpattern: I I r n g E
I < ~ < N )
Symmetry conditions indicate that, in (8.36), V,, = V N + , _ , . If the slots are assumed to be parallel to the Xaxis in Figure 8.42, then the Eplane and Hplane difference patterns correspond to the quadrants being excited as shown in Figure 8.43. For A, (that is, the Eplane difference pattern), V;, =
Fig. 8.42 Indexing Notation for Slot Arrays Showing Individual Modules; M = 6, N = 4
23This does not necessarily mean a rectangular array, since some of the modules may be "empty," such as when corner slots are eliminated so that the array will fit in a circular boundary.
T h e D e s ~ g nof F e e d ~ n gStructures for Antenna Elements a n d Arrays
( b ) E plane difference channel
(a) Sum channel
(c) Hplane difference channel
Fig. 8.43 Quadrant Excitations for Sum and Difference Patterns o f Planar Arrays
 V&+,
,. For this case (8.135) becomes
For A, (that is, the Hplane difference pattern), it is also true that V;q =
 V&+,,,,+, _,. For this case (8.135) becomes M/2
N
p=1
q=l
V;, sin klpq(Z,P: Z"+ rnn 1  p , N + '  4
att tern:
(A,
M 1<m = V.lfm,I (8.138) 1 5 n < N)
However, in contrast to the A, case, in (8.138) the relation V,, = V,+,, applies. As an illustration of the use of these formulas, if the slot lengths and offsets for the twobyfour array discussed in Section 8.16 are used to compute the 2:; and Lj entries needed in (8.136) through (8.138), inversion gives the slot voltage distributions listed in Table 8.9. When these excitations are used to compute the patterns, the TABLE 8.9 Slot voltage distributions for twobyfour array Slot N o .
Slot Voltage V;,
mn
C
1,1 2,1 3, 1 4, 1 1,2 2,2 3,2 4,2
1.002 j0.002 I .OOO  j0.006 1.002  j0.005 1 .OM  j0.003 1.004  j0.003 1.002  j0.005 1 .OOO  j0.006 1.002 j0.002
+
+
AE
+
0.231 j0.063 0.968 j0.203 0.031 j0.139 0.791 j0.128 0.791  j0.128 0.031  j0.139 0.968  j0.203 0.231  j0.063
+ +
+
423
References
result for the sum channel is shown in Figure 8.41. The results for the difference channels are shown in Figure 8.44. Element factors have been included in all these patterns. The difference patterns exhibited in Figure 8.44 clearly illustrate a basic difficulty in the design of planar arrays for use in sum and difference applications. If the slots are excited by a common feeding structure for all three channels, and if one designs the feeding structure t o obtain a good sum pattern, one must accept some poor difference patterns. It is similarly true that, if one were to design the feeding structure in order to produce a good A, pattern, for example, then the resulting 2 and A, patterns are inferior. The only way now known to overcome this deficiency is to use separate feeding structures for the three channels, but this is extremely complicated and costly.
Legend:
Fig. 8.44 AH and A, D~fferencePatlerns for the TwobyFour Slot Array of Fig. 8.40
REFERENCES JORDAN, E. C. and K . G. BALMAIN, Elecl'romugnetic Waves and Radiating Systems, 2nd ed. (Englewood Cliffs, New Jersey: PrenticeHall, 1968), Chapter 15. K I N GR , . W. P., Tables of Antenna Charucteristics (New York: IFI/Plenum, 1971), pp. 715.
The Design of F e e d ~ n gStructures for Antenna Elements a n d Arrays
KRAUS,J. D., Antennas (New York: McGrawHill Book Co., Inc., 1950), Chapter 7 MITTRA,R., "Log Periodic Antennas," Antenna Theory: Part II, ed. R . E. Collin and F. J. Zucker (New York: McGrawHill Book Co., Inc., 1969), Chapter 22. WOLFF,E. A., Antenna Analysis (New York: John Wiley and Sons, Inc., 1966), Sections 5.5, 8.6, and 9.5.
PROBLEMS 8.1 Repeat the monopole design of Section 8.2, assuming that the gap problem can be ignored, and that the Taitype impedance expression (7.71) is applicable. 8.2 What is the impedance bandwidth (1.5 : 1 VSWR criterion) for the monopole designed in Problem 8.1 ? 8.3 Repeat the dipoleoveragroundplane design of Section 8.3, using the improved KingMiddleton approximation for selfimpedance. Equations 7.110 and 7.1 11 can be used if a computer is available, or data can be read from Figure 7.18. 8.4 A rectangular slot is cut in a large ground plane of negligible thickness. If the slot is 1 in. wide and is to be centerfed by a twowire line, what should its length be for resonance at 300 M H z ? What is its input resistance at resonance? 8.5 If a cavity is used to box in one side of the slot of Problem 8.4 and if the depth of the cavity is less than 1,/4, would you expect to have to lengthen or shorten the slot to reachieve resonance ? 8.6 What physical argument would you use to explain the low input VSWR over a wide frequency range for a groundplanebacked helix? 8.7 Find the optimum spacing of a twodipole array, with both elements driven, in order to achieve an endfire array pattern with maximum directivity but only one main lobe. Assume dipole radii a = 0.00321, and determine the lengths of the two dipoles in order to achieve a match with a twowire line at a reasonable impedance level. Assume series coupling to the line, as in Figure 8.9. 8.8 How would the design of the dipole array in Problem 8.7 be altered if only one element were driven ? 8.9 For a threeelement YagiUda array, find the optimum lengths of reflector and director if the driven element is 0.4751 long, if the interelement spacing is 0.151, and if all three dipoles have a radius a = 0.0032A. 8.10 A refinement of the explanation for the need to reverse the feeding at successive elements in a log periodic array of the type shown in Figure 8.27 results from assuming a "locally periodic" behavior along the structure. This permits the identification of transmissive, active, and reflective regions (akin to the directordriven elementreflector model adopted in Section 8.9). Assume this more extensive model and show that the feeding portrayed in Figure 8.26a is incorrect, and that the proper method of feeding is as shown in Figure 5.26b. [Compare with P. E. Mayes, G . A. Deschamps, and W. T. Patton, "Backwardwave Radiation from Periodic Structures and Application to the Design of FrequencyIndependent Antennas," Proc. IRE, 49 (1961), 962.1 8.11 A linear array of six balunfed dipoles stands 114 in front of a large ground plane. The
dipoles are 0.62 on centers and in the endtoend, or tandem orientztion. If a 20 dB SLL DolphChebyshev broadside array pattern is desired, together with an input match to 25 ohms, design a coaxial feed for this array. Your design should include a specification of the length of each dipole and the characteristic impedance of each balun section. Assume that all dipoles have a radius a = 0.00322. 8.12 Repeat Problem 8.11 for a sixbysix array. 8.13 Assume that the master input to the sixbysix array of Problem 8.12 is fitted with a perfect magic T. Find the Eplane difference pattern and the corresponding input impedance. 8.14 Repeat Problem 8.13 for the Hplane difference pattern. 8.15 With the phaseshifters in the branch lines, repeat the analysis of Section 8.12 and show the effect of mutual coupling on pattern and input impedance as the main beam is scanned 5 10" from broadside. 8.16 Design a resonantly spaced threeelement longitudinal shunt slot array in standard Xband guide. The frequency of operation is to be 9.375 GHz and the excitation is to be uniform, with an input match. 8.17 Repeat Problem 8.16 for a threeb;ythree array.
nv
continuous aperture antennas
Q
traveling wave antennas
9.1 Introduction
A traveling wave antenna is one in which the radiating aperture and feeding structure are intimately contiguous, if not continuously connected. As the name implies, the aperture distribution has features similar to those of a traveling wave; the amplitude of excitation may be tapered, but the phase progression is uniform, or nearly so. Traveling wave antennas may be one or twodimensional. Examples of the former are long wires and their derivatives (Vees and rhombics), polyrods, and leaky waveguides. Examples of the latter include corrugated and dielectricclad surfaces (both planar and curved) and arrays of leaky waveguides. This chapter offers an introduction to the analysis and design of some of the practical types of traveling wave antennas. It begins with a discussion of the long wire, followed by an extension to rhombics and Vees. Structures which will support slow waves are then analyzed (notably grounded dielectric slabs and corrugated surfaces) and the launching and termination o f these waves is considered, leading t o a n inter
pretation of the behavior of slow wave antennas. After this, leaky waveguides are introduced, with particular attention given to long continuous slots in either the narrow or broad wall of a rectangular waveguide (the latter offset from the center line), and to the quasicontinuous case of many closely spaced, nonresonant transverse slots in the broad wall (serrated rectangular waveguide). A design procedure is developed that will yield the aperture geometry necessary to achieve a desired pattern together with an input match. Trough waveguides are considered next and their beamscanning capabilities are explored. The chapter closes with a discussion of arrays of longitudinal shunt slots in the broad wall of a rectangular waveguide. The design procedure developed in Chapter 8 for resonantly spaced arrays (d = 1,/2) is extended to apply to nonresonant spacing, which results in a traveling wave excitation. Mutual coupling is generally severe in the traveling wave case and this effect is included in the analysis.
Traveling Wave Antennas
Traveling wave antennas typically have good input impedance characteristics (the reflected wave is effectively suppressed by some means) and therefore the emphasis in this chapter will be on pattern characteristics. 9.2 The Long Wire Antenna
One of the simplest of the traveling wave antennas is the long horizontal wire a distance h above the earth, fed at one end against ground and perhaps terminated at the other end in a matched load, as shown in Figure 9.1. If h is not negligible compared to a wavelength, this wire and its image do not behave like a twowire transmission line, but rather comprise an efficient radiating system. The traveling wave of current proceeding outward along the wire is attenuated due to the radiation, and thus the power absorbed in the matched load may be reduced to an acceptable level by making the wire long enough. Indeed, the leakage may be sufficient to obviate the need for a terminating load. The net current distribution on the wire is then not a standing wave, but is essentially an outward traveling damped wave. If the earth is highly conductive, the image current lies a distance h below the XYplane and is also an outward traveling wave, 180" out of phase with the current on the wire. The pattern due to wire plus image can be obtained by multiplying the element pattern of the wire by the array factor sin(kh cos 8). Assume that the current distribution on the wire can be represented adequately by
+
with y = a j p the complex propagation constant. Equations 1.101 and 1.102 can be used to determine the element pattern of the wire, and give
I
Earth's surface in z = 0 plane
Fig. 9.1 A Long, Terminated Horizontal Wire Antenna ; EndFed (Beverage Antenna)
9.2 The Long W ~ r eAntenna
in which

e  L ( y  jk sin Bcos $)
1 (y  jk sin 8 cos 4)
Io
This field is a figure of rotation about the wire and is given either by (9.2) for 4 = 0 (XZplane) or by (9.3) for 8 = n/2 (XYplane). In practical applications (wires composed of good conductors in an air environment), p z k; u is due almost entirely to radiation leakage and due hardly at all to ohmic losses. Further a 1, it is a hyperboloid. In both cases, the point source is at the left focus. Practical lens antennas can be conceived based on this information. Figure 10.31a shows a dielectric lens cross section in which the inner contour is a circle and the outer contour is an ellipse, with the exterior region free space. The center of the circle and the left focus of the ellipse are coincident at the source point (line) of the primary rays. These rays pass undiffracted through the inner surface, but are diffracted by the outer surface and emerge collimated. A translation parallel to the Zaxis of the contour shown in Figure 10.31a creates an elliptic cylinder lens (n < I), while a rotation of the contour about the Xaxis produces an ellipsoidal lens (n < 1). Similarly, Figure 10.31b shows a dielectric lens cross section in which the inner contour is a hyperbola and the outer contour is a straight line. The primary rays are diffracted at the inner surface and are collimated in the lens. They suffer no further
Focus
Focus
(a) Elliptical lens
(b) Hyperbolic lens
Fig. 10.31 Cross Sections of Dielectric Lenses that Can Collimate Secondary Rays
529
10.1 1 Stepped Lenses
diffraction upon emergence from the surface, so they remain collimated. A translation of the contour shown in Figure 10.31b results in a hyperbolic cylinder lens (n > I), while a rotation of the contour generates a hyperboloidal lens (n > I). There has been an implied assumption in this development that the boundary conditions on the electromagnetic waves at the interface will cause the secondary rays to be horizontal, that is, that the ray trace OAB in Figure 10.30 will satisfy Snell's law of refraction. It is a simple matter to show that, if the equation describing the interface is given either by (10.83) or (10.91), this condition is satisfied. The proof, which is left as an exercise, is similar to what was done in Section 10.3, where it was demonstrated that Snell's law of reflection is obeyed by a parabolic cylinder antenna with a line source placed at its focus. Lenses of the types shown in Figure 10.31 are called singlesurface lenses because all of the diffraction occurs at one interface. Two surface lenses can also be devised, but their analysis is somewhat more involved and will not be undertaken here. ' 10.11 Stepped Lenses
If the angle subtended by the feed from the extremities of the lens is large, as it invariably is in practice, the thickness of the lens (in the xdimension) varies markedly from center to extremity, as can be seen for the elliptic and hyperbolic lenses shown in Figure 10.31. This causes the lens to be bulky and heavy, a disadvantage which can be overcome by stepping. The basic idea is suggested in Figure 10.32. Zones have been created in either the inner or outer lens surface by stepping the thickness. This must be done so that the rays which pass through different zones are still collimated and in phase in the secondary aperture x = d. If the zones are labeled 0, 1,2, 3 , . . . , counting from the edge of the lens in toward the center, stepping of the different contour segments of the lenses displayed in Figure 10.27 can be accomplished as follows: ( a ) ELLIPTICAL LENSCIRCULAR
SEGMENT (FIGURE 1 0 . 3 2 ~ ) Let
r,
and
r , be the radii of the zeroth and mth zones. An equiphase secondary aperture distribution will be achieved if
with k o and k the wave numbers in free space and in the dielectric. This relation simplifies to
in which n is the refractive index of the dielectric and 1, is the freespace wavelength. I9For an introduction to the subject, see E. A. Wolff, Antenna Analysis (New York: John Wiley and Sons, Inc.. 1966), pp. 46871.
Focus
Focus
(a) Elliptical lens
( b ) Hyperbolic lensstepped
straight line
stepped circle
€0
Focus
Focus
(c) Elliptical lensstepped ellipse
( d ) Hyperbolic lensstepped hyperbola
Fig. 10.32 Cross Sections of Stepped Dielectric Lens
10.1 1 Stepped Lenses
531
Hence the radii steps are equal. The extent ofeach zone is determined by the minimum and maximum thicknesses required for structural strength. (6) HYPERBOLIC LENSSTRAIGHT LINE SEGMENT (FIGURE 10.326) Let x, and x, be the longitudinal dimensions of the zeroth and mth zones. The necessary condition is that
k(.u,

x,)

k,(.u,

x,)
2
2nm
from which
Once again the steps are equal and the extents of the different zones are governed by specification of the minimum and maximum thicknesses of the dielectric material. (c) ELLIPTICAL LENSELLIPTICAL SEGMENT (FIGURE 10.32~) For this case, the condition (10.81) needs to be replaced by
k p,
k,(d

p, cos $)

k , a . k,(d

a )  2nm
which reduces to n cos 4)

with 1,the wavelength in medium I . Equation 10.94 is in exactly the same form as (10.81), with the intersection of the ellipse and the Xaxis occurring at a  m l , / ( l  n) rather than at a. Hence (10.94) represents a family of confocal ellipses, all with their far foci at the origin and with equal increments in their stepped xintercepts. The extent of each zone is dictated by the spread of thicknesses specified for the dielectric lens. ( d ) HYPERBOLIC LENSHYPERBOLIC
SEGMENT (FIGURE I0.32d)
For
this case, (10.81) is once again replaced by (10.94). Simplification gives p,(n cos $

I)

Comparison with (10.81) reveals that (10.95) is a family of confocal hyperbolas, all with their far foci at the origin, and with equal increments in their xintercepts. As in the other cases, zonal extents are governed by thickness specifications. The success of this stepping technique depends on the insignificance of scattering at zone boundaries. Obviously, the greater the extent of each zone in wavelengths, the less serious will be this scattering effect. The four stepped contours shown in Figure 10.32 can be used to generate cylindrical lenses, for use with line sources, by translation parallel to the Zaxis. Alter
532
Reflectors and Lenses
natively, they can be used to generate rotationally symmetric lenses, for use with point sources, by rotation about the Xaxis. 10.12 Surface Mismatch, Frequency Sensitivity, and Dielectric Loss for Lens Antennas
The analysis of single surface lenses in Section 10.10 dealt with the primary and secondary rays that progress in a forward direction toward the secondary aperture at x = d. However, at any interface between dielectric media with different constitutive parameters, reflection can also occur. For the elliptical lens shown in Figure 10.31a, the primary rays are normally incident at the inner lens surface, which is either a portion of a circular cylinder or sphere. Locally, the reflection which occurs is the same as though the interface were planar (geometrical optics approximation). It is shown in many standard textsz0 that the reflection coefficient of a plane wave normally incident on an infinite dielectriclair interface is independent of polarization and given by
All of the reflected rays from this cylindrical (spherical) lens surface come to a focus at the feed. If the feed is well matched to free space when the lens is absent, the reflection coefficient measured in the feed when the lens is present, due to this one source of reflection, is given by (10.96). For example, if a polystyrene lens for which E / E , = 2.56 is used then n = 1.16 and = 0.23. The input VSWR at the feed terminals would be 1.6, and 5 % of the primary power would be returned to the feed under these conditions. The situation is more complicated than just described because there is also reflection off the elliptical cylinder (ellipsoid) that comprises the second surface of the lens depicted in Figure 10.31a. Because the direction of the normal to this surface varies from point to point on the surface, the aggregate back scattering effect is nonfocused. Reflection from this second surface primarily appears as a contribution to the total field in the halfspace behind the feed. Higherorder reflections within the lens can usually be ignored. As in the analogous problem of reflection from a paraboloid, there is some depolarization caused by backscattering off an ellipsoidal lens surface. This effect is essentially not present in the case of the long line source and elliptical cylinder lens. An analysis of the surface mismatch for a hyperbolic lens is similar but slightly more complicated. With reference to Figure 10.31b, backscattering off the hyperbolic cylinder (hyperboloid) is nonfocused and contributes to the total field behind the feed, with some depolarization in the case of the hyperboloid. Reflection off the planar surface is collimated and, if n is not too different from unity, most of it is ZQSee,for example, E. C. Jordan and K. G. Balmain, Electromagnetic Waves and Radiating Systems, 2nd ed. (Englewood Cliffs, New Jersey: PrenticeHall, Inc., 1968), pp. 14344.
10.1 2 Surface M~smatch.Frequency S e n s ~ t ~ v ~and t y , D ~ e l e c t r ~Loss c for Lens Antennas
533
transmitted through the hyperbolic surface and focused at the feed. In such cases the reflection coefficient given by (10.96) is applicable as an approximation in the case of the hyperbolic lens. When the focused backscattering caused by dielectriclair mismatch at one of the lens surfaces causes an unacceptably high input VSWR at the feed port, a remedy is to use a matching section in the lens. This consists of a quarterwavelength thick layer of dielectric with refractive index n' = n'12, bonded to the lens surface which is causing the focused reflected rays. Such mismatch correction is obviously frequencysensitive. All of the foregoing remarks are unaltered if surface mismatch is considered for the stepped lenses shown in Figure 10.32. However, stepping introduces another effect which is not present in the basic, unstepped, uncorrected lenses of Figure 10.31. Those prototype lenses are frequencyindependent, to the extent that geometrical optics approximations are valid. This is not the case when stepping is introduced. The point is readily appreciated upon return to Equations 10.92 through 10.95, each of which shows that the step increments are A/(1  n), with A the wavelength in one or the other of the two media. If the steps are properly dimensioned at the design wavelength Ad, they will not be correct for a slightly different wavelength A , = Ad AA. When there are M zones in the stepped lens surface, the path length difference for rays that go through the innermost and outermost zones is
+
at the two wavelengths. The change in path length difference due to a change in frequency is 8
=
AL,

ALd
=
( M  ])(A,  Ad) = ( M  1)AA
and thus a measure of the bandwidth is
A frequently used criterion for pattern degradation due to phase errors across the aperture is that the wavefront should not have a curvature of more than oneeighth wavelength from center to edge (6 l j 8 ) . In this case,

It is clear from this relation that the larger the lens aperture, and thus the more zones needed if stepping is employed, the more narrow band the lens becomes. Another effect that can be estimated in judging the performance of a dielectric lens is the attenuation in the dielectric. The complex permittivity can be expressed as
534
Reflectors and Lenses
The ratio e"/el, which is small for a good dielectric, is often given in the form tan
6
(10.100)
with tan 6 called the loss tangent of the dielectric. Values of e' and tan 6 can be easily obtained by measurement and are usually provided by the manufacturer over the frequency band of interest. The complex propagation constant is given by y =a
+ jp
= [jw,u,(jof0ff)(l  j
tan 6)]1/2= jnk(1  j tan
(10.101)
with k = 2n/rZo the freespace wave number and n = (e')'I2 the refractive index. If the loss tangent is small, the attenuation factor a is given to good approximation by nk
n tan 6 nepers per wavelength
(1 0.102)
For a zoned lens, the average thickness t is given roughly by
and hence the attenuation in the dielectric can be estimated by the formula t = 27.3
nI
tan 6 dB
(10.104)
As an example, for a stepped polystyrene lens used at Xband (10 GHz), e' and the loss estimate is 0.02 dB. and tan 6 = 4.3
.
=
2.54
10.13 The Far Field of a Dielectric Lens Antenna
The same technique that was used in Section 10.7 to find the far field of a reflector antenna can be employed to determine the far field of a dielectric lens antenna. With reference once again to Figure 10.30, assume a line source and hence a lens boundary formed by translating the curve AA,A1 parallel to the Zaxis. If I($) watts per radianmeter is the primary pattern and P(y) watts per square meter is the secondary power distribution in the aperture plane x = d, then I($)d$ = P ( y ) dy. But y = p sin $ and hence dy
=
p cos $ d$
+ sin $ dp =
From (10.82), d p an(n  1) sin 6 @  (ncos6 
10.1 3 T h e F a r F ~ e l dof a Dielectr~cLens Antenna
and thus I(')
=
a ( n 1) n cos Q  I
(EM$ + n cos Q

I
)P(,)
from which (
(n cos Q a n  I )(n
)


o r Q) ICQ>
Equation 10.105 applies whether n < 1 (elliptic cylinder lens) or n > 1 (hyperbolic cylinder lens). The field amplitude of the equiphase distribution in the plane x = d is given by P1lZ(y).With the polarization specified, the equivalent Huyghens sources can be determined and the far field computed. If a point source is used, so that the lens boundary is generated by rotating the curve A A , A 1 of Figure 10.30 about the Xaxis, a similar analysis can be used to obtain the secondary aperture distribution. With spherical coordinates (p, $, y ) erected at the focus, and I($, y/) the radiation pattern of the feed, a power balance gives I($, W)sin $ d$ dyl
=
P(y, Y)Ydy dW
(10.106)
where (y, y) are polar coordinates in the plane x = d and P(y, yl) is measured in watts per radianmeter. Since y = p sin 4, Equation 10.106 reduces to
As before, dy can be related to dQ, the result being that
P(" )'
=
(n cosQ  1)' a2(n  il2(n  cos Q)I($> yl)
Equation 10.107 applies whether n < I (ellipsoidal lens) or n > I (hyperboloidal lens). As before, P1lZ(y,y) gives the field amplitude of the equiphase distribution in the plane x = d, so all the information needed to calculate the farfield pattern is embodied in (10.107). Plots of Equations 10.105 and 10.107, under the assumption of an isotropic primary source, are shown in Figure 10.33 for n = 0.5 (elliptical lens) and n = 2 (hyperbolic lens). It can be observed that the action of a hyperbolic lens is to impose more taper on the primary distribution, whereas an elliptic lens lessens the taper. Because of the resulting penalty in aperture efficiency, hyperbolic lenses are not attractive for applications in which the lens would subtend an angle at the feed much greater than 2Q,,, = 60". On the other hand, if an elliptic lens has a subtended angle 2Q,,, such that cos Q,,, < n, the pole in (10.105) and (10.107) can cause design tolerance difficulties unless the taper in the primary distribution dominates this effect. A specific example of the use of these results is posed in Problems 10.21 and 10.22 at the end of this chapter.
Reflectors and Lenses
6 degrees Fig. 10.33 Normalized Secondary Aperture EField Distributions for Cylindrical and Rotational Dielectric Lenses; Isotropic Primary Radiation Assumed
10.14 The Design of a Shaped Cylindrical Lens
As in the case of reflector antennas, it is possible to alter the shape of a lens so that the secondary rays are not collimated, but instead have an angular distribution corresponding to a more general secondary pattern. The technique will be illustrated for the hyperbolic cylinder lens of Figure 10.31b, with the planar exit surface modified to decollimate the secondary rays in some desired manner. The situation is suggested by Figure 10.34. A family of horizontal rays travels through the lens and the one shown impinges on the exit surface at the point (x, y), making an angle i with the normal t o the surface at that point. The refracted ray departs at an angle r = i t9 to the normal, with t9 the angle this ray makes with the horizontal.
+
10 1 4 The D e s ~ g nof a Shaped C y l ~ n d r ~ c aLens l
Fig. 10.34 A Hyperbol~cCylinder Lens Modifled t o Give a Shaped Secondary Pattern
A unit normal vector at (x, y ) is given by
w ~ t hds a d~splacementalong the exlt contour to the ne~ghbor~ng polnt (x dy). It follows that
T
rix,
J, 1
1, sin i
=:
In
X
1,

d.~
(10.109)
lz cis
Similarly, if a unit vector 1, parallel to the exiting ray is defined by 1,
=
1, cos 8 i I , sin 8
then 1, sin r
sin e zdv
cos 8
ds
Reflectors a n d Lenses
With the exit region free space, Snell's law of refraction gives sin 0 dy
sin nr =sin i
+ cos 0 dx dx
which can be recast in the form dx
=
n
sin 0 cos 0 dy

If the connection between 0 and y is known, this result can be integrated to give
n
sin 0(y1) cos O(yl) dy '

thus establishing the shape of the exit contour as the function x(y). As before, the function 0(y) can be deduced from
with P ( y ) the power distribution in the lens, related to the feed pattern by Equation 10.105; the function W(0) watts per radianmeter is the desired farfield secondary pattern. A design problem illustrating the use of this technique is posed in Problem 10.25 at the end of this chapter. 10.15 Artificial Dielectrics: Discs and Strips
A major practical disadvantage of lens antennas composed of homogeneous, isotropic dielectrics (such as polystyrene) is their weight. Even with stepping, such antennas are so heavy that their use is precluded in all but some groundbased applications. For this reason considerable interest has been shown in the development of inhomogeneous materials with low density and high effective permittivity. Several artificial dielectrics (as such materials are called) have been devised with properties that make them well suited for use in lens antennas. An early candidate was a composite consisting of a lightweight, low permittivity host material (such as polyfoam) in which was imbedded a regular threedimensional array of conducting spherical particles. This is a medium which can be analyzed with little difficultyz1 and one finds that the equivalent permittivity exceeds 6 , for practical sphere sizes and spacings. However, the equivalent permeability is less than p, and this offset results in a refractive index that rises only to a maximum zlSee, for example, J. Brown, "Lens Antennas," Antenna Theory, Part 11, ed. R. E. Collin and F. J. Zucker (New York: McGrawHill Book Co., Inc., 1969). Chapter 18, pp. 1058.
539
10.1 5 A r t ~ f ~ c ~D~electr~cs al DISCSand Str~ps
value of 1.27 when the spheres are touching. This value of n is too low for most lens antenna applications. However, the analysis of an array of conducting spheres reveals an important fact. Were the spheres to become spheroids, flattened in the direction of propagation of an electromagnetic wave passing through the medium, the permeability reduction would be lessened. With total flattening, so that the metallic spheres are replaced by metallic discs lying in equispaced planes perpendicular to the direction of propagation, the permeability effect vanishes. Not so the favorable permittivity effect. The increase in permittivity caused by the presence of the metallic discs can be substantial, and refractive indices as high as two or three can be achieved with practical disc sizes and spacings. The disc dielectric is shown in Figure 10.35a. By symmetry, its refractive index is independent of the polarization of a normally incident electromagnetic plane wave. Because of this it can be used as a lens in conjunction with feeds that are linearly polarized (either horizontally or vertically), circularly polarized, or elliptically polarized. Metal strips lying in planes perpendicular t o direction of propagation
Metal disks lying in planes perpendicular t o direction of propagation
/
Propagation
Fig. 10.35 Disc and Strip Dielectric Media (From Antenna Theory, Part 11, Chapter 1 8 by J. Brown. Copyright 1969, McGrawHill. Used with permission of McGrawHill Book Company.)
An artificial dielectric that behaves similarly to the disc array is the strip dielectric, shown in Figure 10.35b. It can be designed to have an effective permittivity in the useful range by proper choice of the dimensions a, 6, and b'. However polarization is restricted to the case that E is perpendicular to the strip axis. An analysis of the behavior of a strip dielectric medium relies on several tech
Reflectors and Lenses
niques that have been used in earlier chapters. First, if a uniform plane wave is normally incident on the strips, as suggested in Figure 10.35b, the same electric field will be induced in every gap in a common transverse plane. The forward and backward Efield scattering from this plane will be symmetrical (See Appendix F ) . This is equivalent to a shunt obstacle in a transmission line (compare with Equation 3.52). The problem is similar to scattering from a capacitive iris in a rectangular waveguide and can be analyzed by a quasistatic method if b/A< 1. With Iosses ignored, the normalized equivalent shunt admittance is found to be purely susceptive and given by22
It follows that the strip dielectric medium of Figure 10.35b is equivalent to a periodically loaded transmission line, with shunt capacitive elements placed a units apart, as shown in Figure 10.36a. That the elements should be capacitive is due to the size restriction on b/2, which causes the local Efield to store energy electrostatically.
Fig. 10.36 A Periodically Loaded Transmission Line Equivalent t o a Strip Dielectric
Because of the lossless periodic loading of the transmission line, the mode voltages at corresponding points in successive sections differ only by a phase constant 4. The same is true for the mode currents. Hence, with reference to Figure 10.36b, one can write V, I,

=
V , cos ka  j l , Z , sin ka
=
 j V , G , sin ka t I , cos ka
V,

= :
I,
V,ej"
+JB V ,

1,ej+ + jBV,e'i"
(10.1 16)
in which ( V , , I , ) are the mode voltage and current at cross section 1, and so on. The cross sections are chosen so that the first is infinitestimally to the right of a shunt element; the second and third straddle the next shunt element, infinitesimally to ZzSee, for example, N. Marcuvitz, Waveguide Handbook, Vol. 10 of MIT Rad. Lab. Series (New York: McGrawHill Book Co., Inc., 1951), pp. 13867.
10.15 A r t ~ f i c ~ Dielectrics al : Discs and Strips
541
each side of it. Equations 10.116 combine the relations between input and output voltages on an obstaclefree stretch of transmission line and the continuity conditions at a shunt element. If (10.116a) is solved for I, and the result placed in (10.1 16b), a simple manipulation gives cos q!I
=
cos ka

B 2'3,
sin ka
(10.1 17)
With the dimensions a, b, and b' specified, Equation 10.115 can be used to compute BIG, and then Equation 10.1 17 will yield the value for q!I, the phase delay per section. The equivalent index of refraction can be defined by forming the ratio of this phase delay to the phase delay that would occur in the absence of the strips, namely ka. Hence,
It is useful to consider the form taken by (10.1 17) when k is small (long wavelength). Then cos q!I
=
cos ka
cos nka
zI

J(nk~)~
1  & ( k ~ ) ~sin ka
ka
and substitution in (lo. 117) gives
If BIG, is replaced in (10.1 19) by the right side of (10.1 15) and k, becomes exact and one obtains

0, the relation
with no the "static" (zero frequency) value of the refractive index. It can be observed that no is calculable from (10.120) once the dimensions of the strip dielectric are specified. Equations 10.1 15, 10.1 18, and 10.120 can be combined and substituted in (lo. 1 17) to give the result cos nku
=
cos ka

J(ni

I)ka sin ka
(10.121)
which is a particularly useful equation from which to deduce the index of refraction n. Plots of n versus a/A for several values of no are displayed in Figure 10.37. Cutoff occurs when cos nka reaches  1 , but it can be seen that values of n significantly greater than unity are achievable with practical strip dimensions.
Reflectors and Lenses
Fig. 10.37 The Refractive Index of Strip Dielectrics
The curves of Figure 10.37 are also applicable to the disc dielectric if the proper value of no is used. The computation of BIG, for an array of discs is not simple but static measurements on a sample of the medium, placed in an electrolytic tank, can provide an experimental value.23 Since disc and strip dielectric media require normal passage of the electromagnetic wave, and since they present a refractive index greater than unity, they are suitable for use as a lens antenna of the type shown in Figure 10.31b. The strip dielectric is appropriate for the hyperbolic cylinder/linesource case, whereas the disc dielectric can be used either as a hyperbolic cylinder lens or a hyperboloid lens. Stepping is feasible. Both of these artificial dielectrics exhibit loss tangents which are significantly greater than that of a typical homogeneous isotropic dielectric, principally because of the finite conductivity of the obstacles, edge effects if the obstacles are not carefully made, and adhesive losses. However, with controlled construction, the losses are quite acceptable in lens applications and the great savings in weight is a considerable advantage. 10.16 Artificial Dielectrics: M e t a l Plate (Constrained) Lenses
Another type of artificial dielectric which has been widely used in lens antenna applications is shown in Figure 10.38. It consists of an array of equispaced thin metal plates, each lying in a plane which contains E and the direction of propagation. A host material is not needed to hold the plates in place, since they can be connected together by transverse metal rods normal to E. Thus a rigid construction is easily obtained, and some of the loss components present in disc and strip dielectrics are eliminated. 23s. B. Cohn, "Artificial Dielectrics for Microwaves," Proc. Symposium on Modern Advances in Microwave Techniques (Polytechnic Institute of Brooklyn, 1955).
10.1 6 A r t ~ f ~ c D ~~ a el l e c t r ~ c sMetal Plate (Constrained) Lenses
Fig. 10.38 A M e t a l Plate Artificial Dielectric
Since E = 0 on the walls of the metal plates, for I12 < a < I , the electromagnetic waves which pass between the plates must be in the form of the dominant mode with phase constant j? given by the relation
As a consequence, the index of refraction is
It can be observed from (10.123) that the refractive index of a metal plate dielectric is less than unity, rising from a value of zero when a = 212 to a value of 0.886 when a = I . This wide latitude in the choice of a value for n is tempered by the fact that, as n departs further from unity, the surface mismatch between a metal plate dielectric and the adjacent air medium is aggravated. Stated differently, the arriving electromagnetic wave has a locally uniform electric field which must be transformed to a field that is locally a sequence of halfsinusoids. The more of these halfsinusoids there are per unit length in the H direction, the greater the mismatch. Because n i : 1, the basic lens type to which the metal plate dielectric is easily adapted is the elliptical shape, shown in contour in Figure 10.39. Translation gives an elliptic cylinder lens, suitable for use with a line source; rotation gives an ellipsoidal lens, suitable for use with a point source. Parallel, equispaced slices through
Reflectors and Lenses
Focus
Fig. 10.39 Central Cross Section of a Metal Plate Lens
the solid figures caused by translation or rotation give the templates for the individual metal plates which will comprise the lens. A second set of metal plates can be added to the structure of Figure 10.39, arranged so that they are perpendicular to the first set. This is suggested in Figure 10.40. Since the second set of plates is transverse to E, their presence does not affect the refractive index for vertical polarization. But the second set can be designed to
H Fig. 10.40 A n Eggcrate Artificial Dielectric
545
10.1 7 The Luneburg Lens
achieve the same result with a horizontally polarized wave. The resulting structure is called an eggcrate dielectric. With the same plate spacing in both dimensions, collimation of circularly polarized primary radiation, as well as polarization diversity, can be achieved. Stepping of these metal plate lenses can be accomplished in the manner described in Section 10.1 1 for homogeneous, isotropic dielectrics. Because the waves are constrained to travel between parallel plates, these structures are often called constrained lenses. With an elliptic cylinder or ellipsoidal lens boundary, wave diffraction naturally aids this constraint, but the constraint is present regardless of the shape of the boundary, and other boundary shapes are sometimes employed. Some depolarization occurs with an ellipsoidal metal plate lens. This effect is absent in the elliptic cylinder case.
10.17 T h e Luneburg Lens A class of lens that has proven extremely useful in antenna applications (and also in scattering work) is characterized by a refractive index which is variable, but spherically symmetric. By this one means that if the origin of coordinates is chosen at the center of the lens, then n = n(r). The nature of rays in such a medium can be deduced with the aid of Figure 10.41a. A ray MPQ is shown, and r is the directed distance from the origin O to P; 1, is a unit tangent vector at P.
Fig. 10.41 The Differential Geometry of a Ray
Consider the rate of change of the vector r x [I,n(r)] along the ray. One can write
with s a measure of distance on the ray. Because drlds = I,, the first term on the right in (10.124) vanishes. Also, by virtue of (10.15) the second term becomes r x Vn. Since n(r) is a spherically symmetric function,
Reflectors a n d Lenses
as a consequence of which the second term on the right side of (10.124) also vanishes. Hence r X [l,n(r)]
= constant
(10.126)
The implication of this result is that each ray is a plane curve which lies in a plane containing the origin. Along a ray, m(r) I sin q5 1
(10.127)
= K,
with the angle 4 defined as in Figure 10.41a and K , a constant. Equation 10.127 is often called Bouquer's formula. The angle q5 can be connected to the differential geometry of a ray with the help of Figure 10.41b. The right triangle LNP is such that sin
=
LN
=
r dB

40)
(10.128)
The combination of (10.127) and (10.128) gives
Integration yields
which is the equation of a ray in a medium with .a spherically symmetric refractive index. With this as background, consider the situation suggested by Figure 10.42. A lens of radius a is shown for which n(r) is a monotonically decreasing function of r, with n(a) = 1. A ray P , Q , Q,P2 is indicated. It leaves from P I at an angle a with respect to the Xaxis and travels the straight line path P,Q, in the homogeneous air medium. However, the path Q , Q , is curved because the lens is inhomogeneous. It was demonstrated in Section 10.2, as an interpretation of Equation 10.20, that rays always bend toward the region of higher refractive index. Thus with n(r) increasing toward the lens center, Q , Q , bends as shown. The ray then travels a straightline path Q,P, in air, intersecting the Xaxis at the point P,. By virtue of (10.126), this ray is a plane curve. Imagine the surface generated by rotating P , Q , Q,P, about the Xaxis. The intersection of this surface with any plane containing the Xaxis is also a ray because of the spherical symmetry of the lens. A sheath of rays can thus be envisioned, all of which leave P , at the conical angle a and all of which come to a focus at P,. Next, imagine another sheath of rays that leave P, at a drfSerent conical angle a'. Will these rays also come to a focus at P,, or at some other point? Investigation
10.1 7 The Luneburg Lens
Fig. 10.42 The Cross Section of a Spherically Symmetric Lens
of this question by R. R. LuneburgZ4disclosed that, if n(r) is properly chosen, all rays that leave P I and enter the lens can be brought to focus at the common point P,. This class of lenses appropriately bears his name. In its most common form, the Luneburg lens is designed so that the focal point P I is on the surface of the lens and the focal point P, is at infinity. This provides the practical advantage that radiation from a point source can be converted to a plane wave emerging from the lens in a direction diametrically opposite to that of the feed. The form which the refractive index function n(r) must take in order to satisfy the Luneburg condition, namely that all rays which leave P , and enter the lens come to a common focus at P,, can be deduced with the aid of Equations 10.127, 10.129, and 10.130. First, since the point P I is outside the lens and thus' in a region where n
=
1, t h e c o n s t a n t K , is given s i m p l y b y
K,
=r,
sina
(10.131)
It follows that K , has a positive value for all rays that enter the lens, but a value which is dependent on the angle at which the ray departs from P I . The range of K, is over the interval [0, a]. Second, a study of Figure 10.42 reveals that there is a point that has coordinates which can be designated by (r,, 0,) where the ray comes closest to the origin, and that this point is inside the lens. For 0 < Om,drld0 < 0, and for 0 > Om, drld8 > 0. Hence in Equations 10.129 and 10.130, the minus sign should be used in the range 0 I 8 I 0, and the plus sign should be used in the range 0, I 0 I n. Two special 24R. K. Luneburg, The Mathematical Theory of Optics (Berkeley: University of California Press, 1964), pp. 16488. This is a reproduction of Dr. Luneburg's lecture notes at Brown University. The original issue of these notes had misspelled his name as Luneberg.
Reflectors and Lenses
applications of (10.130) then give
The refractive index function n(r) must be chosen so that these two equations hold for all values of the constant K, in [0, a]. Their difference yields
Since
it follows that iff (K,) is defined by
then (10.134) can be rewritten in the form f (K,)
=
&[n + arcsin K
2 "1
+ arcsin K2  2 arcsin 2 ] a r2
(10.136)
The defining relation in (10.135) can be transformed into an integral equation ~~: of a known type by the following m a n i p ~ l a t i o n Let
As a result, (10.135) and (10.136) become K dr' and
f( K ) =
+b+
arcsin K r1
+ arcsin r2K  2 arcsin K ]
(10.138)
2sThe development from this point follows closely the original presentation by Luneburg, Theory of Oprics, pp. 18487.
10.17 The Luneburg Lens
Next, introduce the variable z by the definition
which converts (I 0.137) to the form f
=
O Kdz J p ( )  K'
The exact relationship between p and z depends on n(r), but since n(r) is a monotonic function, so too is p(z). The ranges are 0 < p 1 and CC < z < 0. The form of p(z) is therefore as suggested in Figure 10.43. If the function T(p) is defined by
Figure 10.43 The Function p ( r )
then T(p) is the magnitude of the abscissa 0 causes the transformation from (10.139) to
in Figure 10.43. This introduction
The lower limit on the integral in (10.141) can be explained by returning to (10.127) and noting that
since $, = n/2. The integral in (10.141) can be inverted by the following stratagem: Let g(K) be defined by
Reflectors and Lenses
in the interval 0 < K 5 I . Multiply (10.143) by 2(K" respect to K from p to I. This gives
p2)1/2and integrate with
With the order of integration interchanged, (10.144) becomes
Let the integration variable z be defined by
K Z = (s2  p2)z
+ p2
so that
2K dK = (s2  p2) dz This converts (10.145) to
so that
The result (10.146) can be applied to the function
n  arcsin K $(k) = 2 which can be recognized as part of (10.138). Thus $(K) can be written in the integral form
SK '
m(K) =
Kdp p J ~ 2
 P2 =
SK
This equation is of the type of (10.143) with T(p)
Kd(ln p) JK2  p 2
=
In p. Thus from (10.146),
Next, consider the entire function f(K) given alternatively by (10.138) and (10.141). Use of (10.146) yields
551
10.17 The Luneburg Lens
T(p)
T(I)

=
where this time T(p)
=
K
In p + .s(p)
=
dK +arcsin  (10.150) r i JK2
?

p2
In r'. Since T(1) = 0, if the symbolism
is adopted, (10.1 50) becomes In ( r
(?'d?I ( L ?
[(:)I=
=III n
w p 

+ ~ o p , ~
This equation, together with the defining relation p = nr' determines the required refractive index function n(r') in parametric form. For the special important case that r , = a and r 2 = oo, o(p, m) = 0 and ( p , ) l
1
1 arcsin dv v J v 2  p2
When this result is combined with (10.1521, one finds that
This requirement could be satisfied by a medium with permeability everywhere equaling that of free space but with dielectric constant that decreases parabolically from a value of 2 at r = 0 to unity at r = a. In practice this is achieved by using a family of concentric shells, each with a constant refractive index to approximate (10.154) in steps. At least ten steps need to be used to obtain good performance. It is difficult to design a useful feed which has a phase center that can be placed directly on the surface of the Luneburg lens. If the first focal point is at r , > a but the second focal point is still at r 2 oo, Equation 10.152 gives

In [n(rl)]

w(p, r')
=
I
1" [I
arcsin (Klr;) dK
=
+ j1 (P/~:)']
rit

arcsin (Klr;) dK JK2  p 2
Numerical solution of (10.155) will yield ~ ( r ' )E.. A . WolffZ6has provided results of such calculations in the form of a set of curves that are reproduced in Figure 10.44. 26WoIff, Antenna Analysis, p. 496.
Reflectors and Lenses
Normalized radius r' = rla Fig. 10.44 The Required Radial Variation of Dielectric Constant i n a Luneburg Lens (From Antenna Analysis, by E . A. Wolff. Copyright 1966, J o h n Wiley and Sons, Inc. Used w i t h permission.)
The secondary aperture distribution of a Luneburg lens antenna can be deduced with recourse to Figure 10.45. The ray which leaves the point source at P , and enters the lens at Q , emerges from the lens at Q,, parallel to the Xaxis. The coordinates of Q , are (a, 0). From (10.127) and Figure 10.45, r,
sin a
=a
sin $
=
a sin 8
(10.156)
Let I(a, p) be the radiation intensity of the point source placed at P , . Then I(a, p) sin a da d p is the tubular power flow toward the lens. If (p, P) are polar coordinates in the secondary aperture plane .r = constant and P(y, P) is the power density in that plane, then I(a, /I) sin a da d p But y
=a
=
P(y, p)p dy d p
(10.157)
r , cos u da
(10.158)
sin 8, and use of (10.156) gives y = r , sin a
dy
Substitution in (10.157) produces the relation
=
10.1 7 The Luneburg Lens
Fig. 1 0 . 4 5 A Luneburg Lens Focused at Infinity
The square root of (10.159) gives the amplitude of the equiphase secondary aperture distribution. From this the equivalent Huyghens sources and the secondary pattern can be deduced. A Luneburg lens can also be used as an efficient backscatterer by covering as much as half of its outer surface with a reflector, as suggested by Figure 10.46. This
will cause an incoming wave, incident from any direction in a halfspace, to be
Fig. 1 0 . 4 6 A Reflector
Luneburg
Lens
Reflectors and Lenses
reradiated with a secondary pattern with a main beam pointing in the incident direction. Maximum efficiency occurs if the direction of arrival of the incident rays causes them t o be brought t o a focus a t the central point of the reflecting surface.
REFERENCES BORN,M. and E. WOLF,Principles of Optics (New York: Pergamon Press, 1959), Chapter 3. COLLIN,R. E. and F. J. ZUCKER,ed., Antenna Theory: Part II (New York: McGrawHill Book Co., Inc., 1969), Chapters 1618. DESIZE,L. K. and J. F. RAMSAY, "Reflecting Systems," Microwave Scanning Antennas, Vol. I, ed. R. C. Hansen (New York: Academic Press, 1964), Chapter 2. JASIK,H., ed., Antenna Engineering Handbook (New York: McGrawHill Book Co., Inc., 1961), Chapters 12, 14, and 25. LOVE,A. W., ed., Reflector Antennas, A Collection of Significant Journal Papers (New York: IEEE Press, 1978). LUNEBURG, R. K., Mathematical Theory of Optics (Berkeley: University of California Press, 1966). RUSCH,W. V. T. and P. D. POTTER,Analysis oj'Reflector Antennas (New York: Academic Press, 1970), Chapters 2 and 3. SILVER,S., Microwave Antenna Theory and Design, Vol. 12, MIT Rad. Lab. Series (New York: McGrawHill Book Co., Inc., 1949), Chapters 4, 5, 11, 12, and 13. WOLFF,E. A,, Antenna Analysis (New York: John Wiley and Sons, Inc., 1966), Chapters 7 and 10.
PROBLEMS 10.1 Show that in a homogeneous medium the polarization of the geometric optics field remains constant along a ray. 10.2 Use the results of Section 10.2 to demonstrate Fermat's principle: The ray from a point P I to a point P2 is the curve along which the phase delay is a minimum with respect to infinitesimal variations in path. 10.3 Show that the law of reflection is satisfied for the paraboloidal reflector defined by Equation 10.32 when primary rays emerge from its focal point and all secondary rays are assumed to be collimated parallel to the Xaxis.
10.4 Design a shaped cylindrical reflector to produce the secondary pattern shown in Figure 10.13. Let $, = 20" and $2 = 80" and use the primary pattern I($) = cos[3($?1  50°)]. 10.5 Design a shaped cylindrical reflector to produce the secondary pattern shown in Figure 10.13. Let = 0" and $2 = 60" and use the primary pattern I($) = cos[3($  30°)]. Use a crisscross relationship between primary and secondary rays, that is, $ = 0" corresponds to 6 = 21°, . . . ,$ = 60" corresponds to 6 = 0". Do you see an advantage to such a design?
Problems
555
10.6 If 6($) is a constant so that all secondary rays are collimated, show that Equation 10.43 gives the equation of a parabolic cylinder. 10.7 Design a doubly curved reflector to produce the secondary pattern shown in Figure 10.13. Let 4, = 0" and (b2 = 60" and assume I($) = cos[3($  30°)]. Use an iterative procedure to determine the backbone curve, and terminate when the last iteration gives a p($) which differs from the penultimate p($) by less than 0.1 %everywhere. 10.8 Assume that the feed for the reflector designed in Problem 10.7 is vertically polarized. If the reflector is to be constructed of parallel vertical metallic plates 114 on centers, 118 thick, and 112 deep, find the shape of each plate. Assume that the Hplane pattern is to have a central main beam with a 5' halfpower beamwidth and symmetric side lobes. 10.9 Repeat the calculations for the illustrative example of Section 10.7 with the primary radiation down 5, 15, and 20 dB at the reflector edges. Determine thereby the secondary aperture efficiency and aperture blockage as functions of spillover. 10.10 Determine the secondary aperture distribution for the shaped cylindrical reflector of Figure 10.15. Try a sequence of values for p(OO)until the halfpower beamwidth on the ground side agrees with specification. 10.11 Find the equation of a paraboloid in spherical coordinates centered at the focus. 10.12 Show that, for a paraboloidal reflector, the relation between primary feed pattern I($, y ) and secondary aperture distribution P(r, y ) is given by
The geometry of Figure 10.6 applies: $ is the angle a primary ray makes with the negative Xaxis and (r, y ) are polar coordinates in the plane x = a. 10.13 Assume a paraboloidal reflector with anf/D ratio of 0.5 and with D = 101 is fed by a vertically polarized horn that has a radiation pattern given by Equations 3.8 and 3.9. Find the proper values of a11 and b/A to cause a  10 dB spillover at the reflector edge in each principal plane. Then use the result of Problem 10.12 to determine the secondary aperture distribution. From this, find the equivalent Huyghens sources and then compute the farfield pattern in the two principal planes. 10.14 Add in the effect of aperture blockage to the results of Problem 10.13, assuming a perfect void in the secondary aperture distribution due to total absorption by the horn of the secondary field incident on the horn mouth. 10.15 Repeat Problem 10.13 using the reflector current distribution calculated under geometrical optics assumptions. 10.16 For an unmodified Cassegrain antenna system, find the relation between an axially symmetric feed pattern and the secondary aperture distribution. 10.17 Show that Snell's law of refraction is satisfied at the elliptic cylinder surface shown in cross section in Figure 10.31a, with the equation of that surface given by (10.83). 10.18 Show that Snell's law of refraction is satisfied at the hyperbolic cylinder surface shown in cross section in Figure 10.31b, with the equation of that surface given by (10.58). 10.19 Make an accurate scale drawing of the surfaces of an ellipsoidal lens for which f/f,
=
Reflectors and Lenses
2.56 if a l l = 5 and y,,,/A = 5. Show two alternate weightsaving designs in which you have stepped one or the other of the lens surfaces. 10.20 Repeat Problem 10.19 for a hyperboloidal lens. 10.21 Consider an elliptic cylinder lens antenna for which n = 0.5, with ym,,/l = 5 and dm,, = 45". If the primary radiation intensity is given by
44) =
sin2
(T
(y
sin 4 )
sin 4 )
and if b / l is chosen to have a  10 dB spillover, find the secondary aperture distribution. From this, determine the equivalent Huyghens sources and the secondary pattern. 10.22 Repeat Problem 10.21 for a hyperbolic cylinder lens antenna with n
=
2.0.
10.23 An ellipsoidal lens antenna for which n = 0.5 has a span of 101 and $,,, = 28". If the primary radiation intensity is rotationally symmetric and given by I ( 6 ) = cos 34 watts per steradian, find the secondary aperture distribution. Then deduce the equivalent Huyghens sources and the secondary pattern. 10.24 Repeat Problem 10.23 for a hyperboloidal lens antenna with n
2.0. 10.25 Design a shaped cylindrical constrained (metal plate) lens to produce the csc2 0 secondary pattern shown in Figure 10.13. Use n = 0.5. =
10.26 Design a metallic strip artificial dielectric with practical dimensions at S band (3 GHz) to give an equivalent relative permittivity of 4.0. 10.27 Find an expression for the refractive index function of a Luneburg lens when the foci P, and P2 are symmetrically disposed with respect to the lens center. What specific form does this expression take if the foci are on the lens surface?
appendices
rednetion of the vector green's fornula for E In Chapter 1 the vector Green's theorem is used to establish the relation (1.50), namely,
=
Is ,.,, ,...
(ya x V, x E  E x Vs x ya)
I, dS
This equation can be transformed in the following manner: Using the ninth vector identity listed on the inside of the back cover, one may write Vs x Vs x y a However, Vs
va
=
yVs
=
V,(Vs
.a + a
ya)  V5ya Vsy
=
a
V,v
since a is a constant vector. Also
because y satisfies the scalar wave equation Viy
Employing both (A.4) and (1.46), one obtains
+ k 2 y = 0. Thus
('4.1)
(A.2)
A Reductron of the Vector Green's Formula for E
Use of the third vector identity (inside of back cover) gives
so that the left side of (1.50) becomes
in which the divergence theorem has been employed. The constant vector a may also be taken out in front of the integral sign on the right side of (1.50). Since, with the aid of the fifth vector identity (inside of back cover) and the triple scalar product, one can write [E x (V, x ya)] 1,
=
[E X (V,y x a)] 1, =[(I, x E) x V,y]
[ya x V, X E l  1,
=
a
joy(a X B ) * I , = j o y a * (1, x B)
it follows that
But (A.5) and (A.6) are modified forms of the left and right sides of (1.50), so they are equal to each other. And since this is true for any arbitrary constant vector a, it follows that the integrals themselves must be equal. Thus
=
jS . jX[(I. I..
SN
[(I.
E)Vsy
E)Vsy
+ (1.
+ (1.
X
X
E) X VSV  j o y ( l n X B)1 dS
E) x V S ~jwy(1. X
Bll dS
(A.7)
where, for convenience, the surface integral over the sphere I: has been split off. Consider this integral. On the surface of the sphere C one has
A Reduct~onof the Vector Green's Formula for E
561
If dS1 is the element of solid angle subtended at P by a surface element d S on C, then the right side of (A.7) can be written
Since both integrals in (A.9) are well behaved at P, it follows that 4n
lim 1 = 1irn 60
60
ejk6
IO4'E dS1 = E(P)
dR
=
4lrE(P)
(A. 10)

Next consider the limit, as 8 0, of the left side of (A.7). Obviously the surface integrals are well behaved because P is restricted to be a point within V and thus is not on any of the bounding surfaces S,. As V' + V, the volume integral is also well behaved. To see this, spherical coordinates may be introduced centered at P. Then dV = RZ sin 0 dR dB d$. Since y/ and Vsy/ contain terms involving R' and R2 only, the contribution of the volume element at R = 0 to the volume integral in (A.7) is finite. Therefore the limiting value of (A.7) is
(A. 11) ' I 
in which (x, y , z) are the coordinates of the point P, and it is to be remembered that a time factor e j w r has been suppressed.
be wave eqnnathns for A and 0
In Chapter 1 the potential functions A and 4 were introduced by the defining relations (1.80) and (1.81):
Upon taking the divergence of (B.l) one obtains
since J is not a function of (x, y, z). But
in which use has been made of the third vector identity listed on the inside of the back cover. If Vs J is replaced by  j o p in accordance with (1.45), (B.3) may be written
after the divergence theorem has been employed. Since S may be made large enough to encompass all the sources without containing any of them in its surface, the second integral in (B.4) vanishes and one is left with the conclusion that
B The Wave Equat~onsfor A and @
563
Through application of the Fourier integral, if J and p are general functions of time, one sees that
these integrals being natural extensions of (B. 1) and (B.2). Because linear superposition has been employed, it follows that A and cf, as given by (B.6) and (B.7) also satisfy (B.5). Further, the fields E and B arising from the sources J ( t , 7, C, t) and tl, C, t) satisfy E=V@A (B.8) B  V X A
(B.9)
These equations are restatements of (1.82) and (1.83) but for the more general potential functions (B.6) and (B.7). If one takes the divergence of (B.8) and the curl of (B.9), the result is
which, with the aid of (B.5) and (B.8) become
(B. 10) (B. 11) Thus both A and cf, satisfy the same type of differential equation, the solutions being given formally by (8.6) and (B.7).
derivation of the & Chebyshev polynowials p
Chebyshev's differential equation is
To find a solution, assume Tm(u)can be expressed as a power series, namely,
and then Tk(u) =
2 nanun
and
T;(u)
=
C n(n  l ) a , ~ "  ~ n=2
n= l
Substitution in ( C . l ) gives
When the coefficients of u raised to the various powers are separately equated to zero, one obtains 2az
+ m2uo= O
+ (m2  l ) a , = 0 (n + 2)(n + l)a,+, + (rnz  n2)an= 0 6a,
n
=
O
(C.4)
n
= 1
(c.5)
n 22
(c.6)
From (C.6), the recursion relation
arises, and is seen to truncate for n
=m
if m is a positive integer.
C D e r ~ v a t ~ oofn the Chebyshev Polynorn~als
565
For m an even integer, (C.4) and (C.7) in conjunction indicate that all the even coefficients can be expressed in terms of a,, with the highest nonzero coefficient being a,. If in such circumstances a , is arbitrarily set equal to zero all the odd coefficients are zero by virtue of (C.7). The result is a solution to ( C . l ) that is a polynomial of degree m containing one arbitrary constant a, and only even powers of u. Similarly, if m is an odd integer, (C.5) and (C.7) taken together reveal that all the odd coefficients can be expressed in terms of a , , with the highest nonzero coefficient being a,. If in such circumstances a, is arbitrarily set equal to zero all the even coefficients are zero by virtue of (C.7). The result is a solution to (C.1) that is a polynomial of degree m containing one arbitrary constant a , and only odd powers of u. For m an even positive integer, (C.4) and (C.7) give
Manipulation of (C.8) yields
a, = ( I)" ' a ,
If m
a',,,
=
~ ( 2m ) ( :
2N and n
=
,
I ) .. .
2
+ n 2
2
2

. .(? E + 2
2
n! 
2n'. then
22"' (2n)
( I)"'a9
 ( NN+ n' I ) !


I)!
N! ( N  n')!
From (C.2),


,
N "'0
c (I)"' Y
a,
a 0
4 n' (2n') ! ( N  n') !
.
N ( N t n ' ) ! (2u)2.T N ? n' ( 2 n f ) ! ( N n')!
( C .10)
C Derivation of the Chebyshev Polynomials
If one chooses a, so that TzN(0)= (
then from (C. 10)
When this value of a, is inserted in (C.10), the result is (C. 1 1) Equation C . l l is the general expression for an evendegree Chebyshev polynomial. Similarly, if m is an odd positive integer,
an = (l)(nl)/za
1
(rn+n2)~~~(m+3)(m+l)(ml)(m3)~~~(m[n~]) n! (C.12)
If the substitutions m = 2N  1 and n gives
= 2n' 
1 are made, manipulation of (C. 12) (C. 13)
From (C.2), N
T,,,(u)
=
o l C (1)("l) n= 1
22112 +
 (N;
l ; I) uznI
(C. 14)
It is customary to choose a , so that the slope of TZN,(u)is (2N  l)(l)Nl at the origin. When this done, N
TzN1 ( ~= ) nC (=1
(C. 15)
Equation C.15 is the general expression for an odddegree Chebyshev polynomial.
D
a general expansion of cosm v
The development will be restricted to the case in which m is an integer. Since
The binomial expansion gives
If m is odd, there is an even number of terms in this sum, occurring in pairs, such that
With the substitution m
=
2n'
1
1, (D.3) becomes

2 (2n''
n'1
~ o s ~ ~v ,= l5,
l) cos (2n1 2n
Finally, with the additional substitution I = n'


i)v
n, one obtains
Equation D.5 is the general expansion of cosmv for m an odd positive integer. If one returns to (D.2) and asserts that m = 2n' is an even integer, then
D A General Expansion of cosm v
Now there is an odd number of terms in the sum, composed of pairs plus a single term, such that c0szn,v
If the substitution I
=
= n'
p1 (2n' )

1
+
c (2):'
n' 1
cos 2(nr  n)v
"=o
n is once again used, then
cos2",w
=
1
"'
(
2n'
 F, 2 2 " ~ = ~ n'I
)
COS 21v
in which 6 , = 1, 6 , = 2 for 12 1. Equation D.8 is the general expansion of cosmv for m an even positive integer.
E
approximdion to the magnetic vector potential f i n d o n for slender dipoles1 In Chapter 7, the potential function
is encountered in Equation 7.18. Here K,(z1) is the lineal current density flowing on the outer cylindrical surface of the dipole (which is assumed to be composed of a perfect conductor and to have a length 21). Because of the circular cross section and the $symmetric method of feeding the dipole, K, is not a function of I$. The quantity R that occurs in (E.l) is the distance between the source point (x', y', z') and the field point (x, y, z), both of which lie on the outer cylindrical surface of the dipole. If the source point and field point are expressed in cylindrical coordinates by (a, $, z') and (a, p, z), respectively, with a the dipole radius, then
R
=
[2a2  2a2 cos (4
By symmetry, a(z) is independent of
8 = 0 in (E.2).

p) + ( Z  z ' ) ~ ] ~ ' ~
(E.2)
P, so no loss in generality accrues from setting
It is desired to determine how closely (E. I) is approximated by
in which I(zf) = 2na K,(zf) is the total axial current and
'The proof presented here is patterned after one which can be found in J. Galejs, Antennas in Inhomogeneous Media (Oxford: Pergamon Press, 1969), $2.5.
E Approximation to the Magnetic Vector Potential Function for Slender Dipoles
Equation E.3 represents the magnetic vector potential function evaluated at the field point (a, 0 , z ) when the current is concentrated on the Zaxis. With ka (( 1 and a a (for example, b = 10a). The first and third integrals respectively of these two expansions are equal and thus the two potential functions differ primarily because of the second integrals, that is, a(z)  a,(z) r
1"
Kz(z')a d$ dz/ (E.7) I(zt) dz'
+
In the range z  b Iz' 5 z b, k R and kr are small and ejkR and ejkr can be replaced by 1  j k R and 1  jkr. The integrals associated with the factors jkR and jkr in (E.7) cancel, leaving
It is reasonable to assume that KZ(zt)is slowly varying in the interval 2b and thus
E Approxirnat~onto the Magnetic Vector Potential Function for Slender Dipoles
Execution of the z'integrations gives
When the $ integration is performed on the term y = 0 in the first part of (E.10), one finds that the result exactly cancels they = 0 term in the second part of (E.lO). Thus In (b + Jb2  !2{21nlK
a , ( z ) = 

+ 2a2(1

cos 4) d$ (E.ll)
Equation E. I I can be put in the more compact form 2"
b
+ Jb2 + 2a2(1 b

cos4)
+ JPT2
(E. 12)
d$
from which it can be recognized that the integrand consists of the logarithm of a number that is never far from unity. Indeed, a power series expansion gives In
b
+ J b 2 + 2a2(1 b
+
J
cos $)

a2
2cos$)+
m 2 cos $)
..
I (E. 1 3)
When (E.13) is used in (E.12), the integration is simple and gives (E. 14) If one returns to (E.6), it is evident that U,(z) receives its principal contribution from the second integral and that
(E. 15) A good measure of the adequacy of the approximation is the ratio z
)

1
a ,(z)
)
(albI2  4 1n (2b/a)
(E. 16)
If b = 10a, this ratio has the value 0.0008. Even for b = 4a it is only 0.0075, less than a 1 "/, error. Thus it is acceptable to use (7.25) as an approximation for (7.24) if the
E Approximation t o the Magnetic Vector Potential Function for Slender Dipoles
computations include an integration that extends over a length of the dipole of at least plus and minus several wire diameters. This is possible for all values of z except those close to z = 51, but in those small regions Z(z) is negligible and the value of a(z) is small, so the error is not serious.
F
dihetiorm by phne cormdncting sueens: Babhel's principle Wire grids or arrays of slots in a ground plane excited by primary radiators such as horns, or even by plane waves caused by remote sources, can be designed to have useful antenna characteristics. A powerful integral equation technique which permits deduction of the scattering off such planar obstacles when excited by very general primary sources has been formulated by E. T. Copson.' An important result arising from Copson's formulation is a rigorous statement of Babinet's principle for complementary conducting screens. Consider an infinite, perfectly conducting ground plane of negligible thickness in which an arbitrary collection of arbitrarily shaped holes has been cut, as suggested by the first screen in Figure F.1. The holes have surface areas S , , S,, . . . , S,, to which reference will be made by the abbreviation Z,. The metallic portion of the screen will be designated by 2,. Imagine that sources in z < 0 cause an electromagnetic field distribution throughout space (in the absence of the screen) designated by (Ei, Hi). The sources induced in the screen cause a scattered field (E", H" and the total field at any point in space is given by
The total field in z > 0 (see Section 1.12) can be determined from the equivalent sources
with (t,q, 0+) any point in a plane in z > 0, parallel to and infinitesimally close to the screen. The potential functions for these equivalent sources are (see Section 1.12) IE. T. Copson, "An Integral Equation Method for Solving Plane Diffraction Problems," Proc. Roy. Soc. London, 186A (1946), lO(r18.
F Diffract~onby Plane C o n d u c t ~ n gScreens. Bab~net'sPrinciple
/
/
/
/
/'
I I \ \
/
/ I I \
I
I
.:..
, / Screen
2
Fig. F.1 Complementary Thin Conducting Screens
+
+
in which y = e  j k R / Rwith R = [(x  5)2 ( y  1)' z2]',". Equations 1.106 and 1.107 can be used to establish that, for any point ( x ,y , z ) in z>o, do, Ex = dF,  j o A x  dz dx
E
=  do, 
dz
dFy dx
, dF,dy ,
H,
=
pl O
Hz =
6,
dAY
dt
d@m dz

+
joroF,
d@m

e0
1
~ A Y dx
dx
Po I
aa, d~
575
F Diffraction by Plane Conducting Screens: Babinet's Pr~nclple
Since all of the potential functions are even in z, their first derivatives with respect to z must be odd in z. When the field components in (F.4)are evaluated at (x,y, z), the result must be zero, because these equivalent sources were chosen so as to give null fields in z < 0. But this means that, for any point (x, y, z) in z > 0, the first term on the right side of any equation in (F.4)is equal to the sum of the following two terms. For this reason (F.4) can be rewritten in either of two distinct forms.
E .z =  2 j o A
 2  do dx E =  2 j w A ,  2 
do, E , =   do,
d~
dz
It can be observed that (F.5)gives the field components entirely in terms of the equivalent magnetic sources, whereas (F.6) involves only the electric sources.
FORMULATION FOR &/I;, SMALL If the collection of holes C, is a small fraction of the entire screen, it is clear from a study of (F.3)that it is advantageous to select the set in (F.5)because then the pertinent potential functions require only integration over C,. It will be assumed in what follows that screen 1 fits this description and that (F.5)will be used to represent the fields in z > 0. With the primary sources in z < 0, the determination of the total fields in z < 0 cannot be achieved quite so directly. However, one can observe that
The scattered field components that occur in (F.7)are even in z, that is,
as a consequence of which the total field components appearing in (F.7)are also even in z. Therefore the scattered fields anywhere in z < 0 can be deduced through use of the equivalent magnetic sources
F Diffraction by Plane Conducting Screens: Bab~net'sP r ~ n c ~ p l e
576
which give rise to the potential functions
( F .10)
=
am(&
)',
( F .1 1)
Z )  @L(x,Y , Z )
The potential functions F:, F;, and Wm in (F.9) through ( F . l l ) are those that would apply for the backscattered fields in z < 0 if the screen contained no holes, for then
E:(C, tf, 0 ) =  E X q,O),
E;(C, q , 0 ) =
q , O),
HXC, 4 , 0 ) =  H : ( t , q , 0 ) The total field in z < 0 can be found by operating on the potential functions F:, Fi, and Qk as prescribed by Equations F.5 in order to get the scattered fields and then adding the components of the incident field. This gives
Hx = HI:
+ 2joeoF, + 26
In (F.12), EO = Ei XI = 0.
Hy = H:
+ Er, H0 = Hi + IT is
+ 2 j o r o F y+ 26,d@m dy the total field in z
(F. 12)
< 0 for the case
577
F D ~ f f r a c t ~ oby n Plane Conducting Screens: Bablnet's Pr~nciple
To summarize the results to this point, one can say that the total field in z > 0 can be obtained in terms of the potential functions Fx, Fy, and @, via (F.5), and that the total field in z < 0 can be obtained in terms of the potential functions Fx, Fy, and @, via (F.12) if one adds the total field that would exist in z < 0 if the screen were closed. There still remains the task of finding the aperture distribution Ex(t, q, O f ) , E,(cf, q, 0+), and H,(t, q, 0+) so that the potential functions are known. Since E,, Hx, and Hy must be continuous across the holes C,,it follows from (F.5) and ( F . 12) that if ( x , y, z) is a point in the screen occupied by a hole, then
( F . 13)
Because E:(x, Y , 0 ) =
Y , 01,
H:(x, y, 0 ) = HXx, y, O),
H;(x, y, 0 ) = H:(x, y, 0 )
Equations F. 13 reduce to
( F . 14)
and thus the integral equations linking the unknown aperture field to the incident field are
(F.15)
in which
e j k r Yo
= 
r
r =
[(x  O2
+ ( Y  q)']
with both the source point ( t , q, 0 ) and the field point ( x , y, z ) lying in 1,
( F . 16)
F D~ffractionby Plane Conduct~ngScreens: Bab~net'sPrinciple
578
FORMULATION FOR Z,/Z, LARGE If the collection of holes X I is a large fraction of the entire screen, Equations F.15 are difficult and costly to solve for the aperture distribution. It then proves useful to return to the alternate expressions for the fields, embodied in (F.6), and proceed as follows. Let the sources of the incident field be in z < 0. The total field in z > 0 is (Ei Em,Hi H'),and the scattered field in z > 0 can be found from (F.6) if
+
+
(F. 17)
It is important to note that the integration in (F.17) extends only over X,, that is, the material portion of the screen. This is a consequence of the fact that, because the screen has negligible thickness, H:, H;, and E: are identically zero in XI. The total fields in z > 0 are therefore given by
(F. 18) Upon reflection, one can conclude that (F.18) also applies in z < 0. The reason for this can be seen by examining one of the potential functions. The scattered field in z < 0 can be found in terms of the equivalent sources
which gives rise to potential functions such as (F. 19) But H;(t, q, 0) = H;(O
(F.24)
zConceptually, the relationship between the two primary fields given by (F.21) can be achieved as follows: Let (Ji, p i ) be the sources for (E:, Hi). Replace these sources by a fictitious set (J,, p,) which give the same field (Ei,Hi) everywhere. Finally, replace the magnetic sources by electric sources (J;, p i ) such that J:  m O J m and == ~f,p,. The sources (J:, will cause a field (El, which satisfies (F.21).

pi
pi)
HI)
F Diffraction by Plane Conducting Screens: Babinet's P r ~ n c ~ p l e
Similarly, when (F.23) is substituted in (F. 12), comparison with (F. 18) establishes that
EIKH,=E;
K
and H , +  E , = H ; 1
z