(Fundamental Theories of Physics) V. Kulish - Hierarchical Methods - Hierarchy and Hierarchical Asymptotic Methods in Electrodynamics. Volume 1 - Kluwer Academic Publ (2002)

Hierarchical Methods
Fundamental Theories of Physics
An International Book Series on The Fundamental Theories of Physics:

Their Clarification, Development and Application
Editor:
ALWYN VAN DER MERWE, University of Denver, U.S.A.
Editorial Advisory Board:

JAMES T. CUSHING, University of Notre Dame, U.S.A.
GIANCARLO GHIRARDI, University of Trieste, Italy
LAWRENCE P. HORWITZ, Tel-Aviv University, Israel
BRIAN D. JOSEPHSON, University of Cambridge, U.K.
CLIVE KILMISTER, University of London, U.K.
PEKKA J. LAHTI, University of Turku, Finland
ASHER PERES, Israel Institute of Technology, Israel
EDUARD PRUGOVECKI, University of Toronto, Canada
TONY SUDBURY, University of York, U.K.
HANS-JÜRGEN TREDER, Zentralinstitut für Astrophysik der Akademie der
Wissenschaften, Germany
Volume 123
Hierarchical Methods
Hierarchy and Hierarchical Asymptotic
Methods in Electrodynamics, Volume 1
by
Victor V. Kulish
National Aviation University, Kiev, Ukraine and
Sumy State University, Sumy, Ukraine
KLUWER ACADEMIC PUBLISHERS

NEW YORK, BOSTON, DORDRECHT, LONDON, MOSCOW
eBook ISBN: 0-306-48061-1
Print ISBN: 1-4020-0757-4
©2004 Kluwer Academic Publishers

New York, Boston, Dordrecht, London, Moscow
Print ©2002 Kluwer Academic Publishers

Dordrecht
All rights reserved
No part of this eBook may be reproduced or transmitted in any form or by any means, electronic,
mechanical, recording, or otherwise, without written consent from the Publisher
Created in the United States of America
Visit Kluwer Online at: http://kluweronline.com

and Kluwer's eBookstore at: http://ebooks.kluweronline.com
Contents
Acknowledgments xv
Preface xvii
1. GENERAL IDEAS, CONCEPTS,
DEFINITIONS, AND OTHER
PRELIMINARY INFORMATION 1
1 ESSENCE OF THE HIERARCHICAL
ANALYTICAL–NUMERICAL METHODS 1
1.1 One Illustrative Example
of the Multi-Frequency Oscillation–Wave Systems 1
1.2 Essence of the Hierarchical Approach 4
1.3 Classification of the Hierarchical
Methods Discussed in the Book 7
2 BASIC CONCEPTS AND DEFINITIONS
OF THE OSCILLATION THEORY
OF WEAKLY NONLINEAR SYSTEMS 8
2.1 Nonlinear Oscillations and Nonlinear Systems 9
2.2 Hidden and Explicit Oscillation Phases 12
2.3 Resonances 13
2.4 Slowly Varying Amplitudes and
Slowly Varying Initial Phases.
Complex Amplitudes 20
2.5 Harmonic and Non-Harmonic Oscillations 21
3 BASIC CONCEPTS AND DEFINITIONS
OF THE THEORY OF WAVES
IN WEAK NONLINEAR SYSTEMS 23
3.1 Definition of the Wave 23
3.2 The Phase and Group Wave Velocities 24
3.3 The Phase of a Wave 26
v
vi HIERARCHICAL METHODS
3.4 Transverse and Longitudinal Waves 26

3.5 Surface and Volumetric Waves 27
3.6 The Concept of Dispersion 27
3.7 Waves with Negative, Zero and Positive Energy 28
4 OTHER BASIC DEFINITIONS AND
EQUATIONS OF PHYSICS 29
4.1 Linear Velocity and Linear Acceleration 29
4.2 Linear Momentum and Force 29
4.3 Curvilinear Motion 30
4.4 Rotation 30
4.5 Energy. Field of Forces 31
4.6 Motion Integrals 32
4.7 Canonical Variables. Hamiltonian Equations 33
4.8 Electromagnetic Field 33
4.9 Lagrange and Euler Variables 34
5 CONCEPT OF SYSTEM 35
5.1 What is a System? 35
5.2 State of the System 36
5.3 General Classification of the Systems 36
5.4 Some General Properties of Complex Systems 37
6 BASIC POSTULATES AND PRINCIPLES
OF THE GENERAL SYSTEM THEORY 38
6.1 Principle of Physicality 38
6.2 Integrity postulate 39
6.3 Autonomy Postulate 39
6.4 Principle of Ability to Model 40
6.5 Complementarity Postulate 40
6.6 Action Postulate 41
6.7 Uncertainty Postulate 41
6.8 Purposefulness Principle 42
6.9 Entropy and Negentropy 42
6.10 Complexity 45
2. HIERARCHY AND
HIERARCHICAL SYSTEMS 49
1 BASIC CONCEPTS OF THE THEORY
OF HIERARCHICAL DYNAMICAL SYSTEMS 49
1.1 Why Does Nature Need Hierarchy? 49
1.2 Two Approaches to the Theory
of Hierarchical Systems 50
Contents vii
1.3 Self-Modeling Principle 51

1.4 Main Idea of the Hierarchical Method 52
1.5 Again: What is Hierarchy Originally?
The Structural Hierarchy 53
1.6 Dynamical Hierarchy 54
1.7 Hierarchical Series in Dimensionless Form 58
2 HIERARCHICAL PRINCIPLES.
HIERARCHICAL DESCRIPTION 58
2.1 Hierarchical Principles 58
2.2 Dynamical Equation of the Zeroth Hierarchical
Level 60
2.3 Structural and Functional (Dynamical) Operators 62
2.4 Classification of the Hierarchical Problems 64
2.5 Hierarchical Tree 66
3. HIERARCHICAL ASYMPTOTIC
METHODS. GENERAL IDEAS 71
1 DETERMINED HIERARCHICAL SYSTEMS 72
1.1 Determined Hierarchical Systems 72
1.2 Averaging Operators 76
2 STOCHASTIC HIERARCHICAL SYSTEMS 78
2.1 Factors of Stochasticity in Dynamical Systems 78
2.2 Example of Hierarchical Model
of Stochastic System 79
3 WAVE RESONANT HIERARCHICAL SYSTEMS 81
3.1 Standard Hierarchical Equations
in the Case of Hierarchical Wave Problem 82
3.2 Classification of Problems 85
3.3 Krylov–Bogolyubov Substitution 87
3.4 Case b) 88
4 VAN DER POL’S METHOD 90
4.1 A Few Introductory Words 90
4.2 Van der Pol’s Method’s Variables 90
4.3 Truncated Equations and
Their Hierarchical Sense 92
5 METHODS OF AVERAGING. STANDARD VERSION 95
5.1 Bogolyubov’s Standard System 95
5.2 The Problem of Secular Terms 99
viii HIERARCHICAL METHODS
6 METHODS OF AVERAGING.
TWO-LEVEL SYSTEMS WITH
SLOW AND FAST VARIABLES 100
6.1 Two-Level Systems with Slow and
Fast Variables. General Case 100
6.2 Two-Level Systems with
Fast Rotating Phases 102
4. HIERARCHICAL SYSTEMS
WITH FAST ROTATING PHASES 107
1 HIERARCHICAL OSCILLATIONS 107
1.1 A Few Introductory Words 107
1.2 Formulation of the Hierarchical
Single-Particle Electrodynamic Problem 109
1.3 Classification of Oscillatory Phases
and Resonances. Hierarchical Tree 112
1.4 Reducing Hierarchical Multi-Level
Standard System to the Two-Level Form.
The Scheme of Hierarchical Transformations 117
2 THE CASE OF SIMPLEST TWO-LEVEL
SYSTEM WITH ONE ROTATING
SCALAR PHASE 119
2.1 Formulation of the Problem 119
2.2 Algorithm of Asymptotic Integration 120
2.3 Accuracy of Approximate Solutions 124
2.4 Asymptotic Integration of Initial Equations
by Means of Successive Approximations 127
2.5 Peculiarities of Asymptotic Hierarchical
Calculational Schemes Based on
the Fourier Method 128
3 CASE OF TWO FAST ROTATING SCALAR PHASES 130
3.1 Formulation of Problem 130
3.2 Solutions. Non-Resonant Case 131
3.3 Solutions. Resonant Case 134
4 THE CASE OF MANY ROTATING SCALAR PHASES 139
4.1 Formulation of the Problem 139
4.2 Algorithm of Asymptotic Integration 140
5 ALGORITHM FOR SEWING TOGETHER
RESONANT AND NON-RESONANT SOLUTIONS 143
5.1 Essence of the Problem 143
Contents ix
5.2 Sewing Together of Resonant and

Non-Resonant Solutions 144
5.3 Example for the Solution ‘Sewing’:
The ‘Stimulated’ Duffing Equation 146
WITH FAST ROTATING PHASES.
EXAMPLES OF PRACTICAL
APPLICATIONS 157
1 GENERAL PROPERTIES OF THE MODEL:
‘A CHARGED PARTICLE IN THE FIELD
OF A STANDING ELECTROMAGNETIC
WAVE’ AND EXAMPLES
OF ITS PRACTICAL REALIZATION 158
1.1 Systems for Transformation of
Optical Signals into Microwave Signals
as a Convenient Illustrative Examples 158
1.2 Formulation of the Problem of Electron Motion
in the Field of Two Oppositely Propagating
Electromagnetic Waves 164
2 MODEL ‘AN ELECTRON IN THE FIELD
OF TWO OPPOSITELY DIRECTED
ELECTROMAGNETIC WAVES’ AS A TWO-LEVEL
HIERARCHICAL
OSCILLATATIVE SYSTEM 167
2.1 Reducing Initial Motion Equations to
the Standard Forms with Two Rotation Phases 167
2.2 Zeroth Hierarchical Level.
Parametrical Resonance 170
2.3 Passage to First Hierarchical Level.
Nonlinear Pendulum 171
2.4 Nonlinear Pendulum.
The Miller–Gaponov Potential 176
2.5 Nonlinear Pendulum. The First Motion Integral 178
Exact Solutions and Analysis 178
2.7 Full Solutions of the Initial System 183
3 MODEL ‘AN ELECTRON IN THE FIELD
ELECTROMAGNETIC WAVES’ AS
A THREE-LEVEL HIERARCHICAL
3.1 Transition to the Second Hierarchical Level 185
x HIERARCHICAL METHODS
3.2 Duffing Oscillator 188

4 MODEL ‘AN ELECTRON IN THE FIELD OF
THREE ELECTROMAGNETIC
WAVES’ AS A FOURTH-LEVEL HIERARCHICAL
4.1 Stimulated Oscillation of a Charged Particle 192
4.2 Stimulated Oscillations of an Electron Ensemble 196
6. HIERARCHICAL SYSTEMS WITH

PARTIAL DERIVATIVES. METHOD OF
AVERAGED CHARACTERISTICS 207
1 SOME PRELIMINARY INFORMATION 208
1.1 Motion Equations 208
1.2 Field Equations 210
1.3 Some General Information about Equations with
Partial Derivatives 211
2 METHOD OF AVERAGED CHARACTERISTICS 213
2.1 Concept of the Standard Form 213
2.2 General Scheme of the Method 214
3 CHARACTERISTICS AND
THE METHOD OF CHARACTERISTICS 222
3.1 Method of Characteristics. The Scalar Case 222
3.2 Method of Characteristics. The Vector Case 227
4 EXAMPLE: APPLICATION OF
THE METHOD OF AVERAGED CHARACTERISTICS
FOR A SIMPLEST SYSTEM WITH
OSCILLATATIVE RIGHT PARTS 231
4.1 Initial Equations 231
4.2 Characteristics 232
4.3 Passage to the First Hierarchical Level 232
4.4 Back Transformations 234
5 HIERARCHICAL METHOD OF AVERAGED
QUASI-HYDRODYNAMIC EQUATION 236
5.1 Averaged Quasi-Hydrodynamic Equation 237
6 THE METHOD OF AVERAGED
CURRENT–DENSITY EQUATION 240
6.1 Averaged Current–Density Equation 241
Contents xi
7 HIERARCHICAL METHOD OF
THE AVERAGED KINETIC EQUATION 244
7.1 Averaged Kinetic Equation 244
7. EXAMPLE: APPLICATION OF
THE METHOD OF AVERAGED
CHARACTERISTICS IN NONLINEAR
THEORY OF THE TWO-STREAM
INSTABILITY 249
1 PROBLEM OF MOTION
OF A TWO-VELOCITY ELECTRON BEAM
IN GIVEN ELECTROMAGNETIC FIELDS 252
1.1 Statement of the Motion Problem 253
1.2 Averaged Characteristics for the Motion Problem 254
1.4 Integration of the Averaged Quasi-Linear
Equation for the Beam Velocity 261
2 FIELD PROBLEM. HIERARCHICAL
ASYMPTOTIC INTEGRATION OF
THE CONTINUITY EQUATION 262
2.1 Continuity Equation of
the Two-Velocity Electron Beam 263
2.2 Averaged Characteristics and
the Averaged Quasi-Linear Equation 264
2.3 Back Transformation 268
2.4 Characteristics of
the Averaged Continuity Equation 270
3 FIELD PROBLEM. APPLICATION OF
CHARACTERISTICS FOR ASYMPTOTIC
INTEGRATION OF
THE MAXWELL‘s EQUATIONS 271
3.1 Averaged Maxwell’s Equations 271
3.3 Solving the Averaged Quasilinear
Equation for the Electric Field 276
3.4 Truncated Equations for the Harmonic
Amplitudes of a Space Charge Wave 277
3.5 Some Commentaries for the Obtained Results 279
WITH PARTIAL DERIVATIVES.
SOME OTHER ASYMPTOTIC METHODS 283
xii HIERARCHICAL METHODS
1 MAIN IDEAS OF THE METHOD OF

SLOWLY VARYING AMPLITUDES 284
1.1 General Calculational Scheme of
the Slowly Varying Amplitudes Method 285
1.2 Simplified Version of the Slowly Varying
Amplitude Method. Example: Effect of
Parametric Amplification of a Wave 288
2 TRADITIONAL VARIANT OF
THE SLOWLY VARYING AMPLITUDES
METHOD. RIGOROUS VERSION 293
2.1 Case of Spatially One-Dimensional Model 294
2.2 Classification of Transversely Inhomogeneous
Models 306
2.3 Model with Moderate Inhomogeneity 306
2.4 Method of Parabolic Equation 308
3 MODERNIZED VERSION OF THE SLOWLY
VARYING AMPLITUDE METHOD 309
3.1 Field Problem 309
3.2 Current Density Problem 311
3.3 Current Density Problem in Framework of
the Kinetic Approach 313
4 METHOD OF HIERARCHICAL
TRANSFORMATION OF COORDINATES 314
4.1 Main Idea of the Hierarchical
Transformations 314
4.2 Hierarchical Equations 317
4.3 Averaged Operator 320
5 MITROPOL’SKII METHOD 322
5.1 Reduction of a Partial Differential Equation
to the Standard Form with
Fast Rotating Phases 322
5.2 Basic Solutions 323
5.3 Truncated Equations 324
6 EXAMPLES OF REDUCING OF
THE MAXWELL EQUATIONS TO
THE STANDARD FORM FOR
THE METHOD OF SLOWLY VARYING
AMPLITUDES 326
6.1 Kinetic Version 326
6.2 Quasi-Hydrodynamic Case 328
Contents xiii
7 EXAMPLE: THE TWO-STREAM

INSTABILITY IN A TWO-VELOCITY
ELECTRON BEAM. THE METHODS OF
AVERAGED QUASI-HYDRODYNAMIC
EQUATION AND SLOWLY VARYING AMPLITUDES 329
7.1 Statement of the Problem 329
7.2 Motion Problem. The Averaged
Quasi-hydrodynamic Equation 330
7.3 Motion Problem. The Back Transformations 340
7.4 Field Problem. The Method of
Slowly Varying Amplitudes 341
Appendices 349
Results of calculations in the second approximation 349
Index 351
This page intentionally left blank
Acknowledgments
The author thanks his son Vol.V. Kulish, all friends, colleagues,
present and former students, for their efforts in helping to make this
book an accomplished reality.
The book had been written with the support of Grant No. 1457 given
by the US Government via the Scientific Research Center of Ukraine.
xv
Preface
Everybody is current in a world surrounded by computer. Computers

determine our professional activity and penetrate increasingly deeper
into our everyday life. Therein we also need increasingly refined com-
puter technology. Sometimes we think that the next generation of com-
puter will satisfy all our dreams, giving us hope that most of our urgent
problems will be solved very soon. However, the future comes and illu-
sions dissipate. This phenomenon occurs and vanishes sporadically, and,
possibly, is a fundamental law of our life. Experience shows that indeed
‘systematically remaining’ problems are mainly of a complex techno-
logical nature (the creation of new generation of especially perfect mi-
croschemes, elements of memory, etc.). But let us note that amongst
these problems there are always ones solved by our purely intellectual
efforts alone. Progress in this direction does not require the invention
of any ‘superchip’ or other similar elements. It is important to note that
the results obtained in this way very often turn out to be more significant
than the ‘fruits’ of relevant technological progress.
The hierarchical asymptotic analytical–numerical methods can be re-
garded as results of such ‘purely intellectual efforts’. Their application
allows us to simplify essentially computer calculational procedures and,
consequently, to reduce the calculational time required. It is obvious
that this circumstance is very attractive to any computer user. The
situations discussed are typical for various physical areas (especially for
applied physics which has to do with fast oscillatative nonlinear prob-
lems) and engineering. The models often considered can not be solved by
conventional numerical (or analytical) methods straightforwardly. It is
necessary to have new algorithmic ideas. All known analytical–numerical
methods are intended for similar situations. The point is the special an-
alytical transformation of the initial differential equations to their so
called shortened form. Such shortened equations can be solved by use
xvii
xviii HIERARCHICAL METHODS
some well known conventional numerical or analytical methods. The hi-

erarchical version of the analytical–numerical methods is a specific case
in which the special hierarchical transformations are used. In situations
the hierarchical transformations are constructed on the basis of some
asymptotic calculational procedures, these methods are treated in the
same way as the hierarchical asymptotic analytical–numerical methods.
The main purpose of this book is to expund the basic principles
and peculiarities of hierarchical analytical–numerical calculational pro-
cedures. Electrodynamics seems to the a convenient field for vivid illus-
tration of the essence and specific features of the latter.
The book is intended, first, for students at undergraduate, gradu-
ate, and postgraduate levels who specialize in computer calculations in
physical electronics, plasma physics, radiophysics, optical and electronic
engineering, acceleration and various space technologies, etc.. However,
the book will be of interest for mature experts also, because a significant
part of it is devoted to a new hierarchical point of view as well at some
widespread calculational problems and electrodynamics, as a whole. Be-
sides that we give a number of new illustrative examples in applied
relativistic electrodynamics (free electron lasers and EH-accelerators,
mainly), which could be interesting to them also.
The book has a peculiar ‘hierarchical’ structure with three levels of
complexity. The first level requires knowledge of integral and differen-
tial calculus, series, and general physics at bachelor degree level. The
second level requires a basic knowledge of linear algebra and linear dif-
ferential equations theory (graduate level). The third level requires some
preliminary knowledge of nonlinear equations and plasma-like systems
(postgraduate level). The book is structured logically in such a man-
ner that each level forms a self-consistent educational system, i.e., each
lower level does not depend on the higher levels. Thus a reader at each
educational level can find his own closed set of topics.
Concerning the concept of hierarchy, in itself (the use of which is the
most characteristic feature of the class of numerical-analytical methods
discussed), we should recognize that it is the most unusual and mys-
terious thing in the modern science. At the same time, it is a trivial
phenomenon of everyday life, too. Indeed, one can be convinced that
there is a hierarchy in everyday life everywhere. We can affirm once
more that a person lives in a completely hierarchical world. Here each
occupies a fixed place in a social hierarchy and the results of all events
and situations depend essentially on its level in relevant hierarchical sys-
tem (or, as an ancient Latin proverb says, ‘Quod licet Jovi — non licet
bovi’). The hierarchical order of the regional, state, national, and family
levels reminds us of this systematically. In places of worship each recalls
PREFACE xix
that the Supreme Being has created our world with some hierarchy and
that the arrangements of the Universe, of the human nature, of man’s
social order follow similar patterns, etc.. It seems incredible, but we
have found some similarity between modern physics and engineering. In
particular, there is hierarchy in condensed matter physics, cybernetics,
and coding and systems theories, etc. [1–9]. Recently the hierarchy has
been studied in the theory of oscillations and waves [3,10,11], too. (It
should be mentioned that the oscillatory wave like dynamical systems
namely are the main objects of interest in this book).
The characteristic feature of the hierarchical system is that some of
their specific properties allow us to use the concept of hierarchy to elab-
orate various asymptotic calculational procedures. That is why it is
natural that in spite of the explicit calculational trend of the book we
discuss some of the main ideas of the general theory of dynamical sys-
tems, too (see Chapters 1 and 2).
In what follows let us especially draw the reader’s attention to the fol-
lowing topic. There is a widely accepted opinion that the basis of modern
hierarchical theory in physics has formed during the last twenty years
[1–9] and this theory has philosophical–cognitive significance mainly.
However, it is a relatively little known fact that some calculational as-
pects of the hierarchical approach really have bean practically from used
at least the third decade of the twentieth century (!) in electrodynamics,
and electrical and radio engineering. It is important especially to note
that specific philosophy, terminology, and concepts which are character-
istic for the modern hierarchical theory, are not used here traditionally.
That is why this observation remains some way removed from the at-
tention of the ‘traditional hierarchical’ experts.
It is interesting that researchers (for instance Van der Pol [12], Krylov
and Bogolyubov [13, 14], Leontovich [15], and many others) have, as a
rule, no interest in the philosophy of the problems concerned. Their main
interest had been in the realization of hierarchical like calculational pro-
cedures in practice, e.g., to develop new types of highly effective asymp-
totic calculational methods. In this connection we may mention the Van
der Pol method [12, 16], the set of averaging methods [11, 13, 14, 16, 17],
the slowly varying amplitude method [11, 15, 18], various types of meth-
ods based on different types of transformations (conformal mapping,
convolution, operational calculus, and so on), etc.. From the modern
(‘new’) point of view, these methods can be regarded as the simplest
(two-level) versions of the hierarchical asymptotic analytical–numerical
methods, because specific transformations used here have explicit hier-
archical nature.
xx HIERARCHICAL METHODS
The version of hierarchical theory [10, 11, 19–25], which is described

below, essentially differs from the above mentioned ‘traditional’ [1–9]
one. Firstly, because a specific set of the so called hierarchical principles
is put in its basis. Secondly, because the calculational approaches such
as Van der Pol’s [12,16], averaging [11, 13, 14, 16, 17], and slowly varying
amplitude methods [11, 15, 18] were used for factorization of these princi-
ples (in other words, for their mathematical description) [10, 11, 19–25].
It should be mentioned that further application of the proposed ‘new
hierarchical ideology’ [10, 11, 19–25] for modern electrodynamics non-
linear oscillatative wave resonant models, in turn, yielded to some non-
trivial sequences. First, this stimulates the ‘hierarchical’ grasping by the
mind of some well known electrodynamic problems, which were studied
earlier in the framework of traditional approaches [26–35]. Results of
such a ‘work of thought’ turn out to be unexpected. Including that we
were surprised by the formal schemes of wave electronic devices with
long-time interaction, on the one hand, and the Universe, social and
biological systems, on other hand, possess the same formal hierarchical
structure [10, 23]. Unfortunately, fundamental reasons for this resem-
blance are not yet understood. However, this indicates that (possibly)
the proposed ‘new hierarchical ideology’ [10, 11, 19–25], which was elab-
orated in the framework of the nonlinear electrodynamic problems, pos-
sesses a really more universal nature than it seems at first sight. It also
means that we can hope that the general ideas of this ideology can be
used successfully in other branches of science, too. For instance, in uni-
fied field theory, astrophysics, the general theory of dynamical social,
economical, and biological systems, cybernetics, etc.. So the basic inter-
est of the present book should not be limited to the calculational elec-
trodynamic problems only. Electrodynamics here predominantly gives
convenient examples which demonstrate the essence and peculiarities of
the application of hierarchical calculational algorithms.
It is important to also note another characteristic point of the prob-
lem discussed. We take in view the elaboration of a number of new
hierarchical calculational procedures because we use the proposed new
hierarchical paradigm for practical purposes [10, 11, 19–25]. These are
the method of averaged characteristics, the methods of averaging kinetic
and quasi-hydrodynamic equations, the method of hierarchical transfor-
mation of coordinates, etc. [10, 11, 19–25].
The book consists of two volumes. The first of them is devoted to the
description of various versions of the hierarchical asymptotic analytical–
numerical methods. Detailed illustrations of characteristic peculiarities
of application of these methods in real nonlinear problems of the rela-
tivistic electrodynamics (the theory of the Undulative Induction Accel-
REFERENCES xxi
erators (UNIACs or EH-accelerators) and Free Electron Lasers (FELs))

are given in Volume II.
In writing this book the author has striven to make it open for dis-
cussion, and in doing so he would also hope that experts interested in
electrodynamics and other areas of research would be interested in the
further development of the hierarchical ideology proposed.
References
[1] J.S. Nicolis. Dynamics of hierarchical systems. an evolutionary approach.

Springer-Verlag, Berlin-Heidelberg-New York-Tokyo, 1986.
[2] R. Rammal, G. Toulouse, M.A. Virasoro. Ultrametricity for physicists. Reviews

of Modern Physics, 58(7):765–788, 1986.
[3] H. Haken. Advanced Synergetic. Instability Hierarchies of Self-Organizing Sys-

tems and Devices. Springer-Verlag, Berlin-Heidelberg-New York-Tokyo, 1983.
[4] H. Kaivarainen. Hierarchical concept of matter and field. Earthpuls Press, 1997.
[5] B.M. Vladimirskij, L.D. Kislovskij. The outer space influences and biosphere
evolution, volume 1. In Series ‘Astronautics, Astronomy’, Znanije, Moscow,
1986.
[6] V.V. Druzshynin, D.S. Kontorov. System Techniques. Radio i Sviaz, Moscow,
1985.
[7] V. Ahl, T.F.H. Allen. Hierarchy theory. New York: Columbia Univ. Press, 1996.
[8] T.C. Marshall. Free electron laser. Mac Millan, New York, London, 1985.
[9] Interpretation the hierarchy of nature: from systematic pattern to evolutionary

process theories. Academic Press, 1994.
[10] V.V. Kulish. Hierarchical oscillations and averaging methods in nonlinear prob-
lems of relativistic electronics. The International Journal of Infrared and Mil-
limeter Waves, 18(5):1053–1117, 1997.
[11] V.V. Kulish. Methods of averaging in non-linear problems of relativistic electro-

dynamics. World Federation Publishers, Atlanta, 1998.
[12] B. Van der Pol. Nonlinear theory of electric oscillations. Russian translation.
Svyazizdat, Moscow, 1935.
[13] N.M. Krylov, N.N. Bogolyubov. Application of methods of nonlinear mechanics

to the theory of stationary oscillations. Ukrainian Ac. Sci. Publishers, Kiev,
1934.
[14] N.M. Krylov, N.N. Bogolyubov. Introduction to nonlinear mechanics. Kiev:

Ukrainian Ac. Sci. Publishers, 1937 (in Ukrainian); English translation: New
Jersey: Princeton, Princeton Univ. Press., 1947.
xxii HIERARCHICAL METHODS
[15] M.A. Leontovich. To the problem about propagation of electromagnetic waves

in the earth atmosphere. Izv. Akad. Nauk SSSR, ser. Fiz., Bull. Acad. Sci.
USSR, Phys. Ser., 8:6–20, 1944.
[16] N.N. Moiseev. Asymptotic methods of nonlinear mechanics. Nauka, Moscow,

1981.
[17] E.A. Grebennikov. Averaging method in applied problems. Nauka, Moscow,

1986.
[18] M.I. Rabinovich. On the asymptotic in the theory of distributed system oscil-
lations. Dok. Akad. Nauk. SSSR, 191:1253–1268, 1971. ser. Fiz., Sov. Phys.-
Doklady.
[19] V.V. Kulish. Nonlinear self-consistent theory of free electron lasers. method of
investigation. Ukrainian Physical Journal, 36(9):1318-1325, 1991.
[20] V.V. Kulish, A.V. Lysenko. Method of averaged kinetic equation and its using
in nonlinear problems of plasma electrodynamics. Fizika Plazmy (Sov. Plasma
Physics), 19(2):216-227, 1993.
[21] V.V. Kulish, S.A. Kuleshov, A.V. Lysenko. Nonlinear self-consistent theory of
superheterodyne and free electron lasers. The International Journal of Infrared
and Millimeter Waves, 14:3, 1993.
[22] V.V. Kulish. Hierarchical approach to nonlinear problems of electrodynamics.

Visnyk Sumskoho Derzshavnoho Universytetu, 1(7):3–11, 1997.
[23] V.V. Kulish, P.B. Kosel, A.G. Kailyuk. New acceleration principle of charged
particles for electronic applications. hierarchical description. The International
Journal of Infrared and Millimeter waves, 19(l):3-93, 1998.
[24] V.V. Kulish. Hierarchical method and its application peculiarities in nonlinear
problems of relativistic electrodynamics. general theory. Ukrainian Physical
Journal, 43(4):483-499, 1998.
[25] V.V. Kulish. Hierarchical theory of oscillations and waves and its application
for nonlinear problems of relativistic electrodynamics. In Causality and locality
in modern physics. Kluwer Academic Publishers, Dordrecht/Boston/London,
1998.
[26] S.A. Przybylki. Cache & memory hierarchy: A performance directed approach.
Morgan Kaufmann Publishers, New York, 1990.
[27] C. Brau. Free electron laser. Boston: Academic Press, 1990.
[28] P. Luchini, U. Motz. Undulators and free electron lasers. Oxford: Clarendon
Press, 1990.
[29] A.N. Kondratenko, V.M. Kuklin. Principles of plasma electronics. Energoat-

omizdat, Moscow, 1988.
[30] A.A. Ruhadze, L.S. Bogdankevich, S.E. Rosinkii, V.G. Ruhlin. Physics of high-
current relativistic beams. Atomizdat, Moscow, 1980.
REFERENCES xxiii
[31] R.C. Davidson. Theory of nonlinear plasmas. Mass: Benjamin, Reading, 1974.
[32] A.G. Sitenko, V.M. Malnev. Principles of plasma theory. Naukova Dumka,
Kiev, 1994.
[33] A.P. Sukhorukov. Nonlinear wave-interactions in optics and radiophysics.
Nauka, Moscow, 1988.
[34] N. Bloembergen. Nonlinear optics. Benjamin, New York, 1965.
[35] J. Weiland, H. Wilhelmsson. Coherent nonlinear interactions of waves in plas-

mas. Oxford: Pergamon Press, 1977.
Chapter 1
GENERAL IDEAS, CONCEPTS,

DEFINITIONS, AND OTHER
PRELIMINARY INFORMATION
1. ESSENCE OF THE HIERARCHICAL

ANALYTICAL–NUMERICAL METHODS
1.1 One Illustrative Example
of the Multi-Frequency Oscillation–Wave
Systems
We begin the discussion of the concept ‘hierarchical analytical–nu-
merical methods’ with a certain physical example. Below in this Volume
(and especially in Volume II) we will illustrate widely relevant features
of the studied calculational technologies with the simplest model of an
electronic system that is referred in a special literature to as free electron
laser (FEL) [1–4]. It is very convenient example for illustration of the
essence and application possibilities of the methods discussed. Fig. 1.1.1
can explain some general basic operational principle of a FEL amplifier.
The FEL amplifier discussed comprises a pumping system (for in-
stance, some periodic reversed (in the transverse plane) system of per-
manent magnets (H-ubitron pumping system), see item 1 in Fig. 1.1.1),
as one of the essential elements. Electrons 2 of the relativistic electron
beam 3 move through the working bulk of system 1 on some undulatory
trajectory 3. Simultaneously, electromagnetic signal wave 4 propagates
along the longitudinal direction (i.e., along the axis ) of electron 2 mo-
tion. Signal wave 4 is amplified by virtue of realization of some specific
physical mechanism (see in detail in Chapter 10, Volume II).
The corresponding initial system of differential equations is found to
be nonlinear in the considered case. Besides that, it contains some peri-
odic functions in the equation’s right hands sides (see Chapters 10 and
11, Volume II). It is important that the scale of their periods might
1
2 HIERARCHICAL METHODS
be essentially different from some characteristic scale parameter (the

longitudinal size of working bulk of the system L, for instance). The
simultaneous presence of the periods of essentially different scale is the
most characteristic mathematical peculiarity of the general FEL problem.
What are these different periods in our specific physical example? The
first one is connected with the above mentioned spatial period of elec-
tron oscillations that occur under the influence of the periodic-reversed
magnetic (H-ubitron) field (see Fig. 1.1.1). The second is characterized
by the period of electron oscillations under the influence of the field of
signal electromagnetic wave. Apart from that the electron can make the
cyclotron oscillations in the case when longitudinal focused magnetic
field is put along the work axis of the system. Various collective type
of oscillations can also take place within the electron beam, which could
be regarded as a plasma-like system, and so on [2–4]. Generally speak-
ing, we can ascertain that in real systems there exist many real physical
mechanisms which can cause electron oscillations with essentially differ-
ent periods.
Let us limit ourselves to the discussion of a simplest FEL model,
where only the accounting of the pumping and the signal oscillation pe-
riods are important. In this case we can separate three characteristic
scale parameters. The first one is connected with the above mentioned
spatial period of the pumping system (see Fig. 1.1.1). Really this pe-
riod might be equal, for instance, to a few centimeters [2–4]. The second
scale parameter is determined by the spatial periods of the signal wave
GENERAL IDEAS, CONCEPTS, DEFINITIONS 3
In experiments these periods are usually characterized by values

And finally we introduce some dynamical parameter
which characterizes the velocity of change of electron energy, wave
amplitudes, etc., caused by the interaction process. It may determine,
for example, the spatial scale of changing these parameters (for instance,
this could be some inverse gain factors). In a real experiment the follow-
ing strong inequality is usually satisfied: Or, in other
words, essentially different spatial scales, indeed, characterize the above
discussed oscillation periods. Let us assume for simplicity (where
L, as before, is the longitudinal size of the system working bulk).
Dynamical objects of the discussed class are referred in literature to
as multi-frequency nonlinear resonant oscillation–wave systems (MFN-
ROWS). Thus we might be sure that the FEL model discussed could
be regarded as the simplest example of the MFNROWS. The integra-
tion procedure of corresponding nonlinear differential equation system
that describes the MFNROWS represents the essence of the so called
multi-frequency nonlinear resonant oscillation–wave problem.
At first sight, any standard numerical method like to the Runge–
Kutta or Adams method [5, 6] could, in principle, be used for solution
of this problem. However, such a calculational method of ‘straight nu-
merical integration’ really turns out to be too difficult in practice. It
is clear especially in the case of the multi-frequency nonlinear resonant
oscillation–wave problem [6]. But let us discuss this topic in more detail.
It is obvious that we should provide corresponding calculational step
for some chosen difference scheme for the sake of guaranteeing the
required calculational precision. At the same time, it is quite clear that
in the FEL case discussed the calculational step should be smaller
than the smallest period of the oscillations. In practice it means that
The simplest analysis shows that the general calculational situation

in such a case really becomes very complicated.
Let us give some evident numerical estimation for the illustration of
these complications originating from the requirements concerning the
calculational precision. Therein we take into consideration the above
given typical magnitudes for characteristic FEL parameters
Besides that, let us remind
that the total length of the FELs in experimental practice is
approximately equal to [2–4]. Then, let
us rewrite the precision condition (1.1.1) in the following form:
or, taking into account the estimations chosen above for the FEL pa-
rameters, it is not difficult to obtain the estimation for a number of
calculational steps:
Inequality (1.1.3) means that we should take many more than

steps to perform the complete cycle of calculations! The practical (and
quite grievous) consequences of the latter estimation for practice are
evident. Hence we should look for some other approaches for solution of
the problems of discussed kind. The hierarchical asymptotic analytical–
numerical methods (hierarchical methods) can be regarded as one of the
most promising ways for this. As experience shows (see, for example,
Volume II), they are indeed the most efficient ways of overcoming the
type of difficulties discussed.
The essential point of the hierarchical approach is the use of the prop-
erty of hierarchy for separation of the corresponding problem scale pa-
rameters. These scale parameters are used as correspondingly small (or
large) parameters for construction of the required asymptotic solutions of
the problem. A number of hierarchical methods for constructing asymp-
totic solutions on such a basis are described below in Chapters 2–4, 6, 8.
1.2 Essence of the Hierarchical Approach

The main idea of any analytical–numerical method is some specific
preliminary treatment (transformation) of the initial equations. In the
particular case of the hierarchical methods a number of hierarchical
schemes are proposed for such treatment. The terminal purpose of
all these transformations is reduction of the initial equations into some
‘smooth’ form, i.e., to equations which do not contain any fast oscilla-
tions. In the theory of hierarchical methods (see below Chapters 2–3)
these equations are called truncated (shortened) forms. This means that
we should obtain some simplified set of equations in the final step of
the transformation procedures. The characteristic feature of the latter
is that the truncated equations are written with respect to some slowest
dynamical variables. Because of this we obtain the situation in which
the final ‘smooth’ (truncated, shortened) forms indeed do not comprise
any fast oscillations. We regard such equations as the equations of higher
hierarchical level. It is important that these equations are characterized
by the highest hierarchical scale parameters.
The main merit of these truncated equations (i.e., equations of high-
est hierarchical level) is their evident simplicity. Hence in contrast to
the situation with the initial system, we can apply conventional numer-
ical (or exact or asymptotic analytical) methods for the solution of the
truncated equations obtained.
Let us point out that in the case discussed all information about
all faster motions (oscillations) turns out to be ‘enciphered’ in the cor-
responding transformation relations. The characteristic feature of the
hierarchical methods considered in the book is that such transformation
relationships can very often be found in simple analytical forms.
The ‘truncated’ solution obtained cannot be used as a solutions for
the initial problem. For this we should perform preliminary, so called
inverse transformations. The latter is based on the joint use of the
transformation relationships mentioned and ‘truncated’ solutions. In
this way we can finally obtain the complete solutions of the initial prob-
lem. Therein these solutions contain all groups of the above mentioned
oscillation periods of different hierarchies.
Thus any calculational algorithm of the hierarchical type consists of
the three following stages:
a) accomplishing corresponding hierarchical transformation of the ini-
tial equations up the hierarchical levels, i.e. these are transformations
from the lowest hierarchical level to the highest one (obtaining the trun-
cated equations);
b) solving the equations of the highest hierarchical level (i.e., trun-
cated equations);
c) performing the inverse transformations of solutions for the highest
hierarchical level to the lowest (initial) level.
All these stages are illustrated in more detail below in Fig. 1.1.2.
In what follows let us illustrate once more this calculational of idea
by the simplest example of a FEL model represented in Fig. 1.1.1. But
this time we will try to do it mathematically.
In the view of what has been said the corresponding initial equa-
tion set for the electron motion in the FEL working bulk must contain
some periodic functions. In turn, these periodic functions should com-
prise various oscillation phases and their harmonics (see below about
the concepts ‘phase’ and ‘harmonics’). These phases are determined by
electron oscillations under common action of periodic pumping field (of
the spatial period and the periodic signal–wave field (spatial period
Therein the initial motion equations can be represented by some
nonlinear vector equation (see Fig. 1.1.2)
where is the corresponding vector (electron coordinates,

momentum, energy and so on), Z (. . .) is the corresponding vector-
function, and are the phases of the electron oscillations in pumping

and signal fields that correspond to the mentioned above spatial periods
and and is the laboratory time (for more details see paragraphs
1.2 and 1.3). Within the framework of the hierarchical calculational ide-
ology we can regard the initial equation (1.1.4) as the standard equation
of zeroth hierarchical level. Then we transform the variables in equation
(1.1.4) (as well as this equation, in itself) hierarchically into variables
(equation) of the -th hierarchical level. For this it is necessary to use
some specific transformation relationships similar to (see Fig. 1.1.2):
where and are new vectors and a new vector-function for the
hierarchical level (in the considered particular FEL example the
total number of hierarchical levels equals 2 or 3, — see Vol-
ume II, Chapter 9 for more details), and are some
vector-function and operator, whose construction methods are given be-
low in Chapters 2–4. As a result of accomplished transformations the
initial equation (1.1.4) can be rewritten in the form of a truncated equa-
tion of the hierarchical level (see Fig. 1.1.2):
It is clear that the above discussed calculational problem, concern-

ing the difficulties related with in choosing the integration step (see
(1.1.2) and corresponding commentaries), is actually solved in this case.
Indeed, the right part of the vector equation (1.1.7) is ‘smooth’ and con-
sequently essentially simpler for solving. Any standard numerical (or
analytical) method can be used for its integration. The inverse transfor-
mation of the obtained truncated solutions into solutions for the initial
(zeroth) hierarchical level is realized by means of the relationships like
to (1.1.5) (see Fig. 1.1.2).
Thus summarizing we can say that the essence of the hierarchical
approach consists in accomplishing a specific cyclical calculational pro-
cedure of the type: ‘an equation of the zeroth hierarchical level the
smoothed (truncated) equation of the upper hierarchical level trun-
cated solutions of this equation transformation of the obtained trun-
cated solutions on the zeroth hierarchical level’.
1.3 Classification of the Hierarchical

Methods Discussed in the Book
Accordingly with what was said in the Preface there are a lot of meth-
ods which can be classified as hierarchical. In this book we confine our-
selves by discussing only those asymptotic methods which are effective
for solving weak nonlinear wave oscillatory (including resonant) multi-
frequency electrodynamical problems. Let us classify these methods.
The results of such a classification are presented in Fig. 1.1.3.
As can be easily seen, all methods discussed can be divided into those
which are destined for obtaining solutions of exact differential equations,
and whose objects of description are the systems with partial derivatives.
The Van der Pol, Bogolyubov, and Bogolyubov–Zubarev methods are
related to the methods of the first group (see Chapters 3, 4 and 5).
These methods are widely known in nonlinear mechanics [7–9]. Besides
that, their hierarchical versions (which are foreseen for situations when
the number of hierarchical levels is more than two) are set forth, too.
An essentially different situation takes place for the methods of the
second group. The latter are expounded in Chapters 6–8. Here only
the method of slowly varying amplitudes can be considered as tradi-
tional. The other four methods described (the methods of averaged
kinetic and quasi-hydrodynamic equations, hierarchical transformation
of coordinates, and the method of averaged characteristics, namely) are
relatively new, and therefore little known for the time being for wide
circle of experts. Nevertheless, as will be shown below in the corre-
sponding examples (see Chapters 7, 8 and Volume II), they can be very
effective for solving of a number of complex nonlinear wave resonant

electrodynamic problems, especially in the case of plasma-like systems.
2. BASIC CONCEPTS AND DEFINITIONS

OF THE OSCILLATION THEORY
OF WEAKLY NONLINEAR SYSTEMS
The nonlinear oscillation problem is the most widespread in physics
and engineering. In general, one might be convinced that the entire sur-
rounding world has an explicit oscillatory nature. Indeed, Earth rotates
around its own axis and around the Sun simultaneously; most existing
machines (e.g., your car) contain rotating parts, within all electronic
devices we can see oscillation or rotational electron motion, etc.. Ac-
cording to de Broglie’s hypothesis, all surrounding matter has a wave
nature (waves are complex simultaneous oscillations in space and time
— see the next Subsection for more details). Therefore methods of solv-
ing the oscillation and wave problems (including the hierarchical ones)
are quite popular in most areas of science. Then we take into consider-
ation that further in Volume II we will illustrate the applied aspects of
the hierarchical methods mostly by various examples, which have to do
with nonlinear oscillations and waves in electrodynamical systems. That
is why it seems expedient to give some preliminary information about
the general theory of oscillations. This mainly concerns the basic ideas,
definitions, and concepts used in the following Chapters.
Let us start with the simplest concepts ‘nonlinear oscillations’ and
‘nonlinear oscillatative systems’.
2.1 Nonlinear Oscillations and Nonlinear Systems

Conventionally, the mathematical pendulum is the simplest and most
evident example of the nonlinear oscillatative system. The scheme of
the mathematical pendulum is illustrated in Fig. 1.2.1.
Here the material point of mass oscillates near the vertical axis
It is not difficult to obtain the differential equation for describing the
particle’s motion. As is known from any standard course of general
physics, the generalization of the second law of dynamics (see below
subsection 1.4) for rotational motion can be written in the form:
where is the of moment of the force vector
is the free fall acceleration vector, is its space radius-

vector, is the material point mass, is the of the angular
acceleration vector
is the moment of inertia with respect to the is the vector

of the turning angle, is the unit vector along the rotation axis is
collinear to the normal to the drawing plane), is the magnitude

of turning angle, is the laboratory time. The corresponding set of
dynamical equations (1.2.1)–(1.2.3) can be rewritten in the form:
where is the pendulum length. After corresponding transformations

equation system (1.2.4) can be reduced to the equation of nonlinear
pendulum
where the physical meaning of the constant
will be clarified below. The expression (1.2.5) represents the wide class of
mathematical objects which are called nonlinear equations. Generaliz-
ing, we can say that all equations which contain some nonlinear functions
in their right part (in the case above we have the trigonometric sine),
could be classified as nonlinear equations. Besides the differential equa-
tions the integral and mixed integral–differential nonlinear equations are
known also. However, we will take an interest in the differential nonlin-
ear equations only.
All physical systems whose dynamics are described by nonlinear equa-
tions of any type are called nonlinear dynamical systems. The nonlinear
pendulum (see Fig. 1.2.1 and equation (1.2.5)) represents their simplest
partial type which are called nonlinear oscillation dynamical systems.
In the theory of nonlinear systems, the concepts of linear equation
and linear system are also used. They play a very important role in the
theory of nonlinear oscillation dynamical systems. In particular, these
objects are used as the so called comparison equations and comparison
systems (generating equations or systems, also). The concepts of weakly
nonlinear equations and weak nonlinear systems can also be defined by
means of the latter. Further, we illustrate this in more detail.
Let us consider the pendulum to do small oscillations In
what follows we expand the sine in (1.2.5) in power a series in :
Accounting the first term in (1.2.7) only, we reduce the equation

(1.2.5) to the form of linear differential equation (differential equation
of linear oscillations):
The solution to (1.2.8) can be obtained easily using Euler’s substitu-

tion. Simple calculations yield:
where is the angular amplitude, is the phase of lin-

ear oscillation, and is the initial phase of oscillations. The physical
meaning of the constant is obvious: it is the cyclic frequency of linear
oscillations of the angular coordinate It is related with to period of
oscillation
In what follows we take into account the second term in the expan-
sion (1.2.7). After routine calculations it is not difficult to obtain the
following nonlinear differential equation:
where The nonlinear pendulum equation (1.2.5) as well as

(1.2.6) are very popular in the literature dedicated to nonlinear oscilla-
tions. Equations that are similar (with respect to mathematical struc-
ture) to (1.2.10) are called Duffing equations [10]. Analytical solutions
for the latter are also well known (see, for instance, [10]). These equa-
tions are interesting because they describe the simplest cases of the weak
nonlinear oscillations. This type of oscillations is characteristic for the
weak nonlinear dynamical systems.
Let us further pay some more attention for the weak nonlinear oscil-
lations and systems. The equation (1.2.9) had obtained supposing the
series (1.2.7) converges. In the given case this condition can be written
in the obvious form:
Taking this into consideration we can determine the weak nonlinear

systems as the systems for which some condition like (1.2.11) holds. In
other cases, we can say about the moderately nonlinear systems
or the essentially nonlinear systems In the book,
we confine ourselves to studying the weak nonlinear systems only. Wide
literature is now dedicated to studying the moderately nonlinear and
essentially nonlinear systems (see, for instance [11]). However, it should
be mentioned that their theory requires essentially different calculational

and ideological approaches, which are not the object of our interest here.
Let us also point out the following important circumstance. The char-
acteristic feature of the theory of weak nonlinear oscillations is the wide
use of concepts of oscillation amplitude and oscillation phase . But,
strictly speaking, these concepts had been introduced for the case of
linear oscillations only (see (1.2.9) and corresponding commentaries).
Hence simple dissemination of such ‘linear’ terminology under the weak
nonlinear theory is not a rigorous procedure. However, we have not any
contradiction in this case. The point is that the physical behavior of
the weak nonlinear systems at relatively small observational intervals
(for instance, during one or a few oscillations) turns out to be almost
the same as a ‘purely linear’ equivalent system. The essential difference
appears only in the course of a large number of oscillations. In other
words, we can regard the weak nonlinear system as some slowly evolving
(equivalent) linear system. It means that the system amplitude and fre-
quency in this case are slowly varying parameters (variables) of the weak
linear system. Correspondingly, the phase of oscillations in this situation
can be regarded as fast changing on background of the slowly varying
parameters (variables). This physical peculiarity we will widely use fur-
ther to construct various hierarchical calculational algorithms with fast
rotating phases.
2.2 Hidden and Explicit Oscillation Phases

The concept of oscillation phase (including fast rotating phase) is
widely used in the theory of weak nonlinear oscillations in a number
of effective hierarchical analytical–numerical calculational algorithms.
Therefore, let us further discuss shortly some general peculiarities of the
‘phase’ concept.
We begin with the concept of hidden oscillation phase. Its simplest
illustration could be given by using the example above discussed of the
mathematical pendulum (see Fig. 1.2.1). The oscillation phase in this
case can be represented in the form:
(see (1.2.9) and corresponding commentaries). The main feature of the

hidden phases is that they are not related to any evident periodicities of
the acting system forces. Indeed, the main force acting (Earth’s gravita-
tion) in the case mentioned of the mathematical pendulum is not periodic.
So the oscillation period of the latter,
does not depend on any external periodicity.

In contrast, the explicit oscillation phases are always determined by
some external periodicities. We may take the free electron laser (FEL)
as a convenient illustration example (see Fig. 1.1.1). The periodicities
of the pumping magnetic and signal electromagnetic fields (and their
corresponding acting forces) determine the periodicity of the electron
motion. The electron oscillation phases in this case formally coincides
with the field oscillation phases. Thus periodicity of the considered
oscillation system (FEL) is expressed evidently, in contrast to the case
of hidden phases.
Let us note that in the general case oscillation periods in the weak
nonlinear oscillations always depend on the time or the coordinate or
on both of them simultaneously. As we will see later, this circumstance is
very important for constructing corresponding hierarchical procedures,
as well as for physical analysis, too.
Then it should also be noted that according to the definitions given
above of slowly and fast varying values, the slow and fast oscillation
phases could be determined also. This means that some hierarchy with
respect to the changing velocity of the oscillation phases can take place
in the case of multi-frequency systems. The systems of this kind we
classify as the hierarchical oscillation systems.
2.3 Resonances
In what follows let us discuss the concept of ‘resonance’. We dis-
tinguish the proper and stimulated (induced) oscillations. The proper
oscillations occur in the case of a system with no external influence.
One can be convinced that the oscillations of a mathematical pendu-
lum are proper, because the acting force (Earth gravitation) is a part
of the system considered. I.e., this force can be treated as an internal
one. The stimulated (induced) oscillations take place in the situations
some external periodic force exerts influence on the system. According
to the definition given, we can regard the model of FEL (see Fig. 1.1.1)
as a system with stimulated oscillations. The electron oscillations un-
der action of the magnetic pumping system could be treated as proper
oscillations. On the other hand, the action of an electromagnetic signal
wave on electron motion could be regarded as a stimulated (induced)
influence. The stimulated (induced) system oscillations are connected
with this influence.
Thus the dividing oscillations into proper and stimulated, strictly say-
ing, are relative. This can be explained by it depending on the defining
elements (and forces) which can be included in the system considered,

and which are considered as external ones.
The interesting feature of the stimulated oscillations is the pheno-
menon of resonance. It could seem incredible that until today there is
no universally recognized and rigorous definition of the concept of res-
onance. Often any sharp increasing the oscillation amplitude (for the
changing the frequency of an external force) is called the resonance (see,
for instance, [12]). However, in electrodynamics there are number of
rather ‘strange’ phenomena (such as the plasma beam instability or the
two-stream instability — see Chapter 7, and Chapter 11 in Volume II).
They are also characterized by rather sharp increase of the oscillation
amplitude, but have no ‘truly resonant nature’. Similar phenomena are
also known in hydrodynamics (the Helmholtz instability [13]) and other
areas of science. The point is that all forces acting in phenomena of this
type have an explicitly expressed proper nature. As a rule, in electrody-
namics such processes are called instabilities [14]. Therefore following
the tradition [14] we consider resonance to be a specific instability. But
this does not mean that any instability can be treated as a resonance.
The distinguishing feature of the resonance is that here one of acting
stimulated force always is external one.
In this book we accept the definition of resonance as a sharp increasing
oscillation amplitude with the following condition held
where the phases correspond to proper and stimulated oscillations.

Thus the characteristic feature of resonance is a closeness of velocities of
phases’ change of the proper and stimulated oscillations. This definition
is essential in the theory of oscillations used later in the book.
The physical meaning of definition (1.2.13) can be illustrated by the
FEL model (see Fig. 1.1.1). In the given case the self-evident definition of
the proper phase (i.e., the evident phase of oscillations of some electron
under magnetic pumping field) is
where is the pumping wave number, is the spatial pe-

riod of oscillations of the pumping field (see Fig. 1.1.1), is the
electron longitudinal coordinate, is its initial pumping oscillation
phase.
In what follows let us determine the phase of electron stimulated os-

cillations. We use for this the conventional definition for the phase of
plane monochromatic signal wave. Then explicit expression for the elec-
tron phase of stimulated oscillations formally coincides with the signal
wave phase:
where is the cyclic frequency, is wave period,

is signal wave number, is signal wave length, is light
velocity in vacuum is initial wave phase. The
difference is only the longitudinal coordinate is slowly varying func-
tion (in contrast to the case of plane electromagnetic wave in itself,
where variables and are independent values). Substituting (1.2.14)
and (1.2.15) into (1.2.13), we obtain classical the resonant condition for
the FELs with pumping by undulative magnetic field (H-ubitron pump-
ing) [1–4]:
where is the velocity of the electron on longitudinal direction.

In the essentially relativistic case the formula (1.2.16) can be
rewritten as [2–4]:
where is the relativistic factor, is the en-

ergy, is the rest mass of the electron. Obtaining (1.2.17), we sup-
posed The formulas (1.2.16), (1.2.17) explain, neces-
sity to use essentially relativistic electron beams in FELs.
Indeed, it can easily be seen from (1.2.17), (1.2.18) that the possibil-
ity of obtaining high frequencies (including the optical range fre-
quency) determined by the range of electron energy Let us
accomplish some estimations for illustration of the latter affirmation.
Therein we will use corresponding results of numerical estimations, per-
formed earlier in this Section (see Subsection 1.1). Including we accept:
(that corresponds to the cyclic frequency
According to (1.2.17) it means
that the ratio can be attained
in the case of relativistic electron beam only
[2–4].
The resonances of the type (1.2.16), (1.2.17) are classified in theory

of nonlinear oscillations as the parametric resonances [15–17]. The pe-
culiarity of the parametric resonances is that here both resonant phases
are represented by some explicit phases only. The other possible
variant in which one of phases is represented by a hidden phase in the
resonant condition is called the quasilinear resonance [15–17].
We discussed before the so called paired (two-multiple) resonances
only , when two different phases take part in resonant process (1.2.13). It
should be mentioned that, in principle, the three-multiple, four-multiple,
and so on, resonances can be realized, too [15, 16, 18]. For instance, the
well known four-wave parametric resonances in plasmas, nonlinear optic
mediums, etc. [19–21] can be treated as physical examples of resonances
of the three-multiple type, and so on. In a general case of a
resonance the condition (1.2.13) can be generalized in the following form
where are the numbers of oscillation harmonics, are

the oscillation phases, and is the small parameter of the
problem. One can easily be convinced that the definition (1.2.18) in par-
ticular case of two-multiple resonance
reduces to the simplest form (1.2.13).
In view of (1.2.18) the quantity
can be regarded as a slowly varying function. In literature, it is called

the slowly varying combination phase. Similarly, we can also introduce
the fast varying combination phase:
where are the harmonic numbers, therein

however at least one of numbers has opposite sign with respect
to the analogous number In this case the general resonant condition
(1.2.18) can be written more elegantly
In the framework of the hierarchical theory of oscillations and waves

[16] the condition (1.2.21) gives the criterion for classifying resonances
according to the system’s hierarchical levels. Namely, in the case of
the FEL model the fast combination phase is related to the zeroth
hierarchical level, whereas the slow combination phase characterizes
the dynamical variables of the first hierarchical level [15, 16, 18]. But let
us discuss this problem in more detail.
In the general case of the multi-frequency model we can have, in prin-
ciple, a few combination phases may be related with the
first hierarchical level. These phases formally describe several resonances
realized in parallel and simultaneously. Such systems are called multi-
resonance systems. From the ‘point of view of the zeroth hierarchical
level’ (that characterized by the fast combination phases
all slow combination phases are characterized by the same order in
their velocities of varying. However, essentially different situations oc-
cur in the zeroth and first hierarchical levels, correspondingly. We take
in view that the first level comprises only slow phases without any
‘fast background’ of fast phases Hence different combination phases
begin ‘compete’ (with respect to their velocities) amongst themselves.
Therefore amongst these slow combination phases, the more and less
slow phases can be distinguished. Some of these phases can be regarded
as new ‘super-slow’ and ‘quasi-fast’ ones, similarly with situation in he
zeroth hierarchical level. In turn, some of such ‘quasi-fast’ phases can
form new ‘super-slow’ combination phases in the same manner (see
further Chapter 4 and 5 for more details). We can regard this as a re-
alization of peculiar ‘slow resonances’, i.e., the resonances of the next
(i.e., second) hierarchical level, and so on. Further, the described cal-
culational procedure can be repeated many times until no oscillation
phases (resonant as well non-resonant) remain at the considered stage
of the calculational.
Let us mention that, generally, we can distinguish two types of the
multi-resonance systems. The first is characterized by bound (in par-
ticular, coupled ) resonances. In this case one of oscillation phases can
take part in a few resonances simultaneously [17, 22, 23]. The coupled
resonances always are resonances of the same hierarchy. In contrast,
the non-bound resonances (which are characteristic for systems of the
second type) can belong to the same hierarchical level as well as to dif-
ferent hierarchical levels. In detail the concept of hierarchy in nonlinear
multi-frequency oscillatory systems is discussed in Chapter 4. The cor-
responding examples are given in Chapter 5. The general ideology of

arbitrary hierarchical dynamical systems is described in Chapter 2.
Then we turn again to discussion of resonant condition (1.2.21). At
the first sight it could be seem that the essence of the latter essentially
differs from the analogous condition (1.2.18). Nevertheless, it is not
really so. The point is that both quantities in (1.2.21), (1.2.18): and
respectively, are characterized by an approximately equal scale of
varying velocity, Hence, formal differences are not essential
and they do not influence in strictness of corresponding calculational
accomplished on such basis.
The sense of the latter affirmations can be obviously illustrated at
the example discussed above of the FEL model (see Fig. 1.1.1 and cor-
responding comments). This particular case is described by following
parameters:
The changing velocity of the phase in this case is essentially lesser

than changing velocity of the phase if the condition
is satisfied, in the case with resonance condition (1.2.13) held. Hence,

velocities of changing and are of the same order in the small
parameter
Then let us define the concepts of ‘resonant point’ and ‘vicinity of
the resonant point’. For this, it is convenient to introduce the concepts
of ‘resonant function’ and ‘resonant curve’. We assume nonlinear reso-
nant system is passing through a resonant state during some interaction
process. The function
we consider as the resonant function. Further, we again use the model of

FEL. Conventionally, the effectiveness of the energy exchange between
of the electron and the electromagnetic signal wave for the time interval
in the theory of electron devices with long-time interaction
is characterized by single-particle efficiency of the interaction
where is theefficiency, is the energy of the electron as function

of time at and Analysis shows that the qualitative characteristic of
dependency of the efficiency (1.2.25) on time could be explicitly resonant.
To illustrate the latter we change variables in the function
(for instant, using the definition of the resonant function (1.2.24))
and pass to corresponding dependency The latter is described
graphically by the resonant curve. A typical form of resonant curve is
given in Fig. 1.2.2.
The sense of the resonant point and resonant vicinity is obvious from
this figure. Namely, the resonant point corresponds to the point of
maximum The vicinity of the resonant point is determined by the
interval on which the efficiency varies from
through the point until
We mention that the realization of slow braking of the electron motion
is a characteristic peculiarity of the FEL interaction mechanism [2–4].
Therein the transformation of kinetic energy of the electron into the
signal energy of the electromagnetic wave occurs. This means that the
velocity of the longitudinal electron motion in the situation considered is
a slowly varying function of time As follows from definitions (1.2.14),
(1.2.15), (1.2.19), (1.2.20), (1.2.24) the resonant function in the
general case is a slowly varying function of time too. The latter
appears in Fig. 1.2.2 as a slow motion of the working point A along
the resonant curve from the right side to the left side. In principle the
passing of working point A through the resonant point can

occur immediately as well as through its nearest vicinity. The resonant
condition (1.2.21) can violate in time when the point A moves too far
from the resonant point (and the resonant vicinity, correspondingly).
In such a case the electron interaction with the electromagnetic field
ultimately loses its resonance character and, therefore, the effectiveness
(1.2.25) becomes too low. In the specialist literature, the above physical
mechanism described is called saturation of amplification [2–4].
Let us point out that the physical peculiarities described of resonant
systems essentially influence the mathematics of the problem. Namely,
as shown below, calculational algorithm for the given class of calcula-
tional problems essentially depends on the kind of interaction mech-
anism, i.e., the resonant or the non-resonant (see for details in Chap-
ters 7–11). Therefore specific algorithm use is needed for sewing resonant
and non-resonant parts of solutions [17].
2.4 Slowly Varying Amplitudes and

Slowly Varying Initial Phases.
Complex Amplitudes
Let us discuss some peculiarity of description methods of the weak
resonant nonlinear systems. As mentioned above, the classical form of
the linear solutions like (1.2.9) is, generally, illegal in the weak nonlinear
case. It is also mentioned that such a solution representation, nonethe-
less, could be disseminated at the weak nonlinear models, if some addi-
tion specific suppositions are accepted. These ‘addition specific suppo-
sitions’ concern new ‘extended’ definitions for the oscillation amplitude
and initial phase. The difference with the linear case like to (1.2.9) is
that at that time the oscillation amplitude and initial phase are found
to be slowly varying functions on time So, the corresponding weak
nonlinear solutions can be written in the following ‘modernized’ general
form
where is the component of some vector

is component of the slowly varying amplitude vector
is component of the slowly varying initial oscillation phases vector
is the (permanent) cyclic frequency. It is easily be convinced that
some other version of introducing the slowly varying values is possible,
too:
where is the slowly varying cyclic frequency, is the permanent

(!) initial phase. Both discussed representations (1.2.26) and (1.2.27), as
experience shows are, in principle, equivalent. However, the first version
is more widespread in practice, inasmuch as really more convenient.
Thus the basic weak nonlinear mathematical problem could be formu-
lated as a determining the slow dependencies and (or
and respectively. One of main interest in this book is description
of methods for the constructing such kind dependencies.
Apart from the real amplitude and phase the so called
complex slowly varying amplitudes are used for practical calculations.
The definition of complex slowly varying amplitude can easily be ob-
tained from (1.2.26) utilizing the well known de Moivre formula:
Using (1.2.28) in (1.2.26), we obtain the representation for (1.2.26)

via the slowly varying complex amplitudes:
where ‘clear’ oscillation phase is defined as
exp is the slowly varying complex amplitude; the

designation i.e., it is the complex
conjugated term.
Thus the weak nonlinear oscillations formally could be described by
means the concept of slowly varying amplitudes and phases. Therein
the oscillations similar to (1.2.26), (1.2.27), and (1.2.29) are called in
references the quasi-harmonic oscillations, because, as mentioned above,
they are not ‘true’ harmonic oscillations. But let us discuss the concepts
of harmonic and non-harmonic oscillations in more detail.
2.5 Harmonic and Non-Harmonic Oscillations

Thus we introduce further the concepts of harmonic and non-
harmonic oscillations.
According to the classical definition an oscillation is a motion process
in which the object studied systematically comes back to its initial state.
This definition includes a very wide spectrum of various types of oscilla-

tions, including the periodic, non-periodic, nearly periodic, conditionally
periodic; harmonic and non-harmonic oscillations, etc.. In what follows
we briefly discuss a few of these concepts which will be needed below.
One says about the periodical oscillations in the case only if the ‘sys-
tematical’ process in the system is strictly periodical. Earlier, in our
illustration examples with the linear pendulum (see Fig. 1.2.1 and cor-
responding commentaries) we considered that the action of the external
and the internal forces causes the oscillation processes, having pure pe-
riodical. In a simplest case they can be described by some harmonic (see
definitions (1.2.9), (1.2.26), (1.2.27), (1.2.29)) functions. This particular
case of the periodical oscillations is called the harmonic oscillations.
In what follows we use the model of free electron laser (FEL —
Fig. 1.1.1 and corresponding explanations). As can easily be seen in
Fig. 1.1.1, the electron in the working bulk of FEL moves under the
action of two periodic fields. They are the pumping static (permanent)
magnetic H-ubitron field with space period and the electromagnetic
signal wave field, which is characterized by the space period and the
temporary period Therefore the electron moves along some
complex trajectory, having no fixed period (as it is mentioned in Subsec-
tion 1.1, we have at least two periods in the considered FEL case). Thus
the two-periodic oscillations are realized in the discussed illustration ex-
ample. In the general case of many oscillations we can speak about the
(multi-periodic) oscillations (here
The important peculiarity of nonlinear systems is their ability to
transform any initially harmonic oscillations into non-harmonic (but,
periodic) oscillations. For example, in the case of the nonlinear pen-
dulum the linear oscillations take place for relatively small amplitudes
only (see solutions (1.2.9)). A number of higher harmonics occurs with
the growth of the oscillation amplitude. This characteristic process can
be illustrated by nonlinear pendulum. Here the oscillation process is
described by the equation (1.2.5). In the simplest case of linear ap-
proximation the solution of (1.2.5) can be given by a sine-like harmonic
function (1.2.9). In the second, third, and so on orders of approximation,
oscillation harmonics appear in the solution (see further Chapter 5 for
more details). The nonlinear pendulum represents the evident example
of so called multi-harmonic periodic oscillations.
Similarly, higher harmonics are generated really in FEL by pumping
and signal oscillation electron phases, simultaneously (see Volume II, for
instance). Besides that, by virtue of nonlinear nature of the oscillation
process various combination oscillations (see definition of the combina-
tion phases (1.2.19), (1.2.20)) could be exited, too. In this case we get
a complex case of the multi-periodic multi-harmonic oscillations.
3. BASIC CONCEPTS AND DEFINITIONS

OF THE THEORY OF WAVES
IN WEAK NONLINEAR SYSTEMS
3.1 Definition of the Wave
The definition of the wave is based on the concept discussed above of
oscillation. Namely, a wave is oscillations occurring in time and space
simultaneously. Let us show this by the simplest example of a sound
transversal wave in a solid medium (see Fig. 1.3.1 and Fig. 1.3.2).
Under such an interaction neighboring material points oscillate too.
Let us note that any perturbation caused by such interactions propagates
in the medium with some finite velocities. Therefore, initial oscillation
phases of different material points are found to be shifted in time (see
curve 2 in Fig. 1.3.1).
Let us consider that we have to do with waves in some transversally
unbounded homogeneous stationary non-dissipated elastic medium. In
addition, we choose within this medium some small part of it which can
be treated as the material point at coordinate In what follows,
this material point starts to oscillate with the frequency
(where is the temporal oscillation period). These oscillations occur in
the transverse plane under action of some external perturbation. The
discussed state of system is illustrated by curve 1 in Fig. 1.3.1.
Then we consider that all medium points always interact with one
another. According to the above suppositions this interaction should be
elastic.
We marked above already that initial phases of different point oscil-

lations are shifted in time. Therein relative magnitude of this shift de-
pends essentially on each point position in each instant of time. Hence
all oscillations of other points occur with some ‘time lag’. This lag is
larger the more remote the point is. The physical picture discussed is
illustrated in Fig. 1.3.2. It shows explicitly the lag of oscillations of the
second material point placed at
Thus oscillations of different medium points have different initial oscil-

lation phases because of the lag-effect. Therefore, some spatial distribu-
tion of elementary oscillators (i.e., the medium points) over their initial
oscillation phases occurs. Let us photograph mentally this distribution
at two different time instant and We can see this spatial distribu-
tion has an explicit oscillatory nature (see Fig. 1.3.2). However, contrary
to the situation depicted in Fig. 1.3.1, these oscillations have an explic-
itly expressed spatial nature. Let us remember that the time instants
and are chosen arbitrary. This means the spatial oscillation process
is realized simultaneously with the temporary oscillations. So the wave
indeed is a superposition of two simultaneous oscillation processes, one
of which occurs in time, and the other is realized in space.
The spatial oscillation period is called the wavelength. The
value is named the wave number. Therefore the definition for the wave
amplitude does not distinguish from the above given definition for the
oscillation amplitude (see formulas (1.2.26)–(1.2.29) and corresponding
commentaries).
3.2 The Phase and Group Wave Velocities

Thus between oscillations in two spatial points and these is a
time lag Consequently the propagation velocity of in-
teractions between material points can be determined in the following

manner
with all notations given above. However, it should be mentioned that

the notation which is determined by such a method, describes, in
the general case, two different physical values. The first of them is the
motion velocity of a fixed phase point on the wave front. And the second
is the velocity of energy transfer by the wave. (Let us recall which the
wave front is a surface in space that is formed by points with equal
oscillation phases). Both these values in the simplest case considered
coincide. However, these velocities can be different in the case of some
disperse anisotropy arbitrary medium. They are called the phase and
group velocities, respectively. We must distinguish these concepts in the
general case.
The following rigorous definition for the phase velocity is accepted:
where is the wave vector,

are the unit vectors along the axes and respectively;
is the wave number, are the corresponding
modules of the vector components. The group velocity could be
defined as:
Let us note that the directions of vectors of the phase and group
velocities coincide only in the simplest case of isotropic medium. In
the general case these vectors are directed arbitrary. Both these cases
can occur in the above discussed FEL model. For instance, the phase
and group velocities of the signal wave in traditional FEL H-ubitron ar-
rangements, as a rule, are coincided [2–4]. However, some other rather
exotic physical situations could be realized in some Dopplertron type of
FELs. Their characteristic design feature is the use intense electromag-
netic waves (including microwaves) as FEL pumping [1]. Therefore the
situations with opposite directions of the phase and group velocities can
occur really [16, 24, 25]. The physical models of such type are discussed
in Chapters 10–13, Volume II.
3.3 The Phase of a Wave

Thus each oscillation point in the considered model medium can
be characterized by proper oscillation phase (see Fig. 1.3.1)
Therefore each next point oscillates owing to the lag-effect with

the shifted phase:
where all designations are given above. However, we should member

that choice of the observation point is arbitrary. Hence, it could be
written for any oscillating point of the medium
where as before, is the wave number, is the wave-

length (see Fig. 1.3.2).
Taking into consideration the definitions for phase velocity (1.3.2), we
can generalize the definition for the wave phase (1.3.6) in the case of an
arbitrary wave process:
where is the radius-vector of observation point in the three-

dimensional space.
3.4 Transverse and Longitudinal Waves

We discussed earlier the examples in which the plane of particle os-
cillations in a medium is normal to the direction of wave propagation.
The waves of such a type are called transversal waves. In the opposite
situation (i.e., when the material points oscillate along the direction of
wave propagation) we have longitudinal waves. It should be mentioned
that rather exotic wave types with mixed electrodynamic structure can
exist in some complex electrodynamic systems. They can not be classi-
fied as either purely transversal waves or longitudinal ones. For example,
the L waves in some type of retarding electrodynamic systems can be
excited. The other example: the TM or TE waves in waveguides, etc..
The electromagnetic signal waves in FELs can, as a rule, be classified
as the transverse type waves, whereas the pumping wave in an FEL
Dopplertron can be purely transversal (laser radiation [26], for instance)
as well as complex mixed type of waves [1–4]. Space charged (plasma)
waves can be excited within the working bulk of relativistic electron
beams. Amongst these, both transversal and longitudinal waves can be

found. Corresponding examples of interaction of transverse and longi-
tudinal waves are discussed in Chapters 12, 13, Volume II.
3.5 Surface and Volumetric Waves

Surface and volumetric waves are distinguished by their physical na-
ture. Surface waves are characterized by surface physical mechanisms of
excitation. The total wave energy in this case is localized within nearest
to the surface layer of medium. The characteristic thickness of such a
layer roughly equals the length of the wave. Usually waves on water are
evidently surface waves. Volumetric waves propagate through volumes
whose characteristic size can essentially exceed the wavelength. The
light and sound waves serve as other obvious examples of volumetric
waves.
3.6 The Concept of Dispersion

Dispersion is characterized by presence of a dependency of the phase
wave velocity on frequency:
Taking this into account we can rewrite the definition (1.3.2) in more
general form
Relationships like to (1.3.9) are called the dispersion relations or the

dispersion laws. Solving the corresponding dispersion equation
can obtain them. Here is the dispersion function. The waves

whose dispersion law can be found from dispersion equations like to
(1.3.9) are defined as the proper waves of the considered system. For
instance the light waves are proper waves for vacuum, as an electrody-
namic system, and so on. The opposite relation
is known as the condition of improper waves’ existence. The stimulated

electron waves in plasmas of electronic beam or H-ubitron (magneto-
undulated) pumping field in FELs are examples of improper waves (see
numerous illustration examples in Chapters 12, 13, Volume II).
3.7 Waves with Negative, Zero and Positive

Energy
Performing corresponding calculational for the considered model of
sound wave in an elastic medium, we can obtain the well known expres-
sion for the wave energy density [27]:
where is the wave energy in an elementary volume is the

medium density; A and are the wave amplitude and the frequency,
respectively. It seems at first sight that the wave energy should be a
quite positive-determined quantity, because As is widely
known, the energy describes the ability of bodies to work. Therefore
the existing the ‘negative energy’ or the ‘zeroth energy’ seems to be
impossible. However, the concepts ‘the wave with negative energy’ and
‘the wave with zeroth energy’ are often used in the modern physics, too.
What is the matter? Let us shortly discus this problem in more detail.
As a simplest analysis shown, the mentioned seeming contradiction
carries purely terminological nature. The point is that, talking about
the wave energy we really always take in view the energy difference (!).
It is the difference between the medium energy with a wave and the
medium energy without the wave. Hence, strictly speaking, we should
write for the wave density in the general case the following more precise
definition instead (1.3.12):
It can easily be seen that the sign of the energy density (and the
wave energy, too) depends on the correlation of magnitudes of the two
different energies and in (1.3.13). In particular, we have
the ‘usual case’ of positive wave energy in the case
The opposite case describes the system with the negative wave energy.
If both energies are equal we have to do with the zeroth wave energy.
All these three types of waves are used further in the book. For in-
stance, the electromagnetic signal wave in FELs is characterized by the
positive energy always. The slow space charged wave (SCW) in electron
beam plasmas in parametric Raman’s FELs is the wave with negative
energy. In addition, at last, an example of a wave with zeroth energy
is shown by increased and dissipated waves in the beam models with
two-stream instability. In more detail, all these examples are discussed

in Chapters 12, 13, Volume II.
4. OTHER BASIC DEFINITIONS AND

EQUATIONS OF PHYSICS
A number of physical concepts and definitions have been used above.
All of them will be utilized in the following Chapters. However the
author speaks in view the situation, when a reader has not required
knowledge in the field of general physics. So, let us add in what follows
the portion of widely known physical information. Here a number of
concepts, definitions, and most popular formulae necessary in the future
are given (see also [27] or any suitable textbook fore more details).
4.1 Linear Velocity and Linear Acceleration

Let us start with the mechanics of material point. The (linear) in-
stantaneous velocity is the value
where is the radius-vector of the material point. The linear instanta-

neous acceleration could be determined as
Both them describe the kinematics of material point in three-

dimensional space.
4.2 Linear Momentum and Force

The concept of the linear momentum
is related to the dynamics. In Newtonian dynamics the force is the only

cause that determines the motion of a material point. The second law
of dynamics describes this relation mathematically:
where is the force. In the nonrelativistic case the second law can be
rewritten as
It should be mentioned that the discussed definitions for the momen-

tum and the force are valid in the nonrelativistic only. Some other
definitions should be given in the relativistic case.
4.3 Curvilinear Motion

In the case of curvilinear motion the concepts of the moment of force
the moment of momentum
and the moment of inertia of material point
are used to describe the dynamics of material point. Comparing ex-

pressions (1.4.6)–(1.4.8) with corresponding quantities for linear motion
one can see that is functional analog of the force the moment of
momentum corresponds to the linear momentum and the moment
of inertia I is the ‘curvilinear analog’ of the mass of material point
These concepts were used above to illustrate nonlinear pendulum (see
(1.2.1)-(1.2.4)).
4.4 Rotation
The simplest case of curvilinear motion is rotation. In the rotation
kinematics, besides the linear values, the angular values are introduced
also. Including, the angular displacement is:
where is the angle of a turn of the radius-vector of material point

is the unit vector normal to the plane of vector rotation (see Fig. 1.2.1
and corresponding commentaries). The angular velocity and the angular
acceleration (see also (1.2.1)–(1.2.4)) are defined for the linear motion
as:
The different scalar components of the vector of angular velocity

describe the different circular frequencies of rotations of
the material point around the rotation axes. This displays the physical
meaning some of algorithms, which are set forth further in Chapters 3,

4.
The correspondence between linear and angular quantities is deter-
mined by the relationships:
where is the vector along the rotation directed to the rotation

center, and are corresponding unit vectors. The
generalization of the second law of dynamics onto rotation can be given
in the form (1.2.1).
4.5 Energy. Field of Forces

The ability of the body to work
is called the energy. One distinguishes the kinetic and the potential
energy. The kinetic energy characterizes the motion dynamics of a body.
In general relativistic case it is
where all designations are given above. The value is called the rest
energy, is the rest mass. In the case of nonrelativistic motion the
expression (1.4.13) can be reduced to the following well known form:
The definition of the potential energy is related to the definition of

field of forces. The part of space for each point of which some vector
can be corresponded, is called the field of the vector
The second law of dynamics (1.4.4) can be expressed in the form
describing the motion of quasi-continuous flow of material points in the
field of forces:
You must distinguish the potential and vortex fields. The condition
of field potentiality can be written as
where L is the integration contour, is the differential element of this

contour. The potential field can be described by the potential function
The force in each point of the field can be expressed as the
negative gradient of the function
Below we will use both above mentioned field types. Therein the
electric field can exist in the potential form (it can be generated by
charged particles) as well as the vortex field (that is excited by a time-
varying magnetic field — see Volume II for more details).
Thus the kinetic energy is related to the body motion, whereas the
potential energy corresponds to the position of the body in a force field.
4.6 Motion Integrals

Characteristic mathematical peculiarity of any dynamical system is
the possibility of constructing the so called motion integrals. These are
such combination of dynamical parameters, which are conserved during
the system motion process. It is known that 3N – 1 motion integrals can
be constructed for any closed system consisting of N material points. It
is remarkable that only three from them are characterized by the prop-
erty of additivity. These integrals are connected with the corresponding
conservation laws. Including, the energy conservation law
the momentum conservation law
and the moment of momentum conservation law
All these motion integrals are related to the inherent properties of

the space-time and they are common for all dynamical systems. Apart
from the other particular motion integrals discussed, which could be
specific for a studied system, can be obtained, too. In Chapter 5 and in

Volume II we will widely use various motion integrals in analysis.
4.7 Canonical Variables. Hamiltonian Equations

Mechanics knows two versions of the problem of material point mo-
tion — Newtonian and Hamiltonian ones. Above we had to do with the
first of them only. As was mentioned, the Newtonian formalism treats
the force as the main motion cause that is represented by the second
law of dynamics (1.4.4). In contrast to from the Newtonian formalism,
the Hamiltonian formalism does not treat the force as a necessary basic
concept. The motion of the material point in this case is completely
described by the Hamilton function. The latter is the total energy writ-
ten in terms of momentum and coordinates. The description by means
of canonical variables is used here. Let us use the three-dimensional
coordinate and the canonical momentum as the canonical variables.
Hamilton function in this case can be written as
The dynamics of a material point in an external field of force is gov-

erned by the Hamiltonian equations
As mentioned above, the dynamics of a system considered can be com-

pletely described by the equations (1.4.22). These equations are widely
used further in the book (see, for instance, Chapter 5, and Chapters 9,
11–13 in Volume II.
4.8 Electromagnetic Field

The electric and magnetic fields are specific particular cases of the
electromagnetic field. Dynamics of the latter is described by the Maxwell
equations:
where and are the vectors of intensity (strength) and displacement

(induction) of the electric field, and are the vectors of strength and
induction of the magnetic field; and are the current density vectors
resulting from external and intrinsic sources of fields, and are the
space charge densities caused by similar sources, is the nabla operator.
The intensity vector of electric field is force that acts on some positive
unit charge (+q) in a given point of the electric field:
As mentioned earlier, there are two different types of the electric fields.
They are the scalar (potential) and the vortex electric fields. As follows
from equations (1.4.23), electric charges create the scalar electric fields,
whereas a time-varying magnetic field creates the vortex electric field.
The vector of electric displacement is connected with the intensity
vector by the following relationship:
where is the dielectric permittivity of medium. It is the auxiliary

value that facilitates description of the electric field within dielectrics
(magnetodielectrics).
Electric current (i.e., directed motion of charges) creates magnetic
field. Analogously to situation with the vector the vector of mag-
netic induction is the force characteristic of the magnetic field. The
intensity of magnetic field is an auxiliary quantity to describe mag-
netic field in substance
where is the magnetic permeability of the medium.
4.9 Lagrange and Euler Variables

Electrodynamic plasma-like systems are, as a rule, described in terms
that are usually accepted in the hydrodynamics and aerodynamics. For
instance, intensive electron beams are treated as flows of charged fluids
or gases. The Lagrange or Euler variables are used for such method of
description.
In the first case, a researcher considers an individual particle (elec-
tron or compact aggregation of electrons — large particle) and fixes
its position in any time instant Correspondingly, phases of the
electron oscillations (see above Subsection 1.2.2) ex-
pressed via the Lagrange coordinates are referred to as Lagrange
oscillation phases. Analogously, we can determine the linear Lagrange
velocity angular velocity etc..
In the second approach the researcher considers a spatial point with
the radius-vector and it registers the velocities of electrons passing
through this point in each time instant Therefore, each point of the
beam volume is put in relation to a certain point of the field of velocities
The velocity function is an Euler variable, as well as the
phase the angular velocity etc., could be regarded also
as the Euler variables.
Thus the Lagrange formalism is related with the one-particle descrip-
tion of the system, whilst Euler formalism is associated with its many-
particle description. This distinction will help us to avoid misunder-
standing in the analysis of the nature of quantities to be considered in
what follows. In the account below, we employ widely both: the La-
grange as well as Euler approaches.
5. CONCEPT OF SYSTEM
Most of the objects investigated in physics and engineering are var-
ious types of systems. Therefore, the knowledge of general features of
the systems allows to better understand the problems studied. It con-
cerns both: the mathematical description of the problem, as well as
the analysis of results obtained. Apart from that, peculiarities of some
systems (for instance, the hierarchical dynamical systems) allow us to
construct corresponding highly effective asymptotic calculational meth-
ods. In other words, the systems properties can be useful in solving the
mathematical part of the problem, too. Lastly, the set of concepts and
ideas considered in the general system theory is very interesting in itself.
So let us discuss some general features of dynamical systems.
5.1 What is a System?

It seems incredible that until today there is no precise and clear the
system definition. To illustrate this we use one of the best definitions
known in the literature: “ We call the system an object of any nature
(or an aggregate of interacting objects of any nature, including objects
of different nature) that has explicit ‘system’ property (properties), that
does not belong to any individual part of the system for any method of
its decomposition, and that does not follow from properties of system
parts” [28].
Thus a specific feature of a system is that it has principally new
properties what do not reduce to those of any of its parts. For instance,
air has properties (wind, tornado, etc.) what its molecules do not have at
all. Another example, two young people decide to form their own family.
The problem arises inevitably: which of them should be the head of the
family? It is obvious that before the wedding this problem does not
appear. This example describes also the peculiar property of the so
called hierarchical systems to be discussed in the next Chapter. Here

in this Section we confine ourselves by the most general properties of
systems (including hierarchical ones). Therein we discuss the dynamical
systems most, keeping in mind that dynamical systems include static
systems as a particular case.
Let us note that separate parts of the system could have proper sys-
tem properties differing from those of the system as a whole. These
parts are called the subsystems. The association of systems is called the
supersystem. Hierarchical systems are a specific variety of supersystems
where each hierarchical level consists of some subsystems.
Any system, supersystem, or subsystem has inputs and outputs. Be-
sides that, any object of such a type is in some surrounding medium.
The system communicates with the surrounding medium by its inputs
and outputs. In the general case the communicating with the medium
is driven by energy, matter, information, etc..
Hence, the concept of system is relative. Each subsystem can be
regarded as a system (with respect to any its ‘subsubsystem’) or, equiv-
alently, each supersystem can be considered as a subsystem of corre-
sponding ‘supersupersystem’. As is shown in the following Chapters,
this system property is characteristic for the hierarchical systems.
What is the medium from the ‘system point of view’? The medium
in the system theory is some external surrounding the system inter-
acts with. Therefore, other parts of the system serve as corresponding
medium for each subsystem.
The open and closed systems are distinguished. The open system has
an exchange with the surrounding medium, whereas in the closed system
there isn’t any exchange.
5.2 State of the System

The ordered aggregate of system parameters that determine the evolv-
ing in the system is called the state of the system. For instance, the state
of some thermodynamically equilibrium gas is completely determined by
three parameters: gas pressure temperature T, and volume V.
5.3 General Classification of the Systems

All systems can be divided into four large classes:
a) the determined systems, where all possible processes are determined
within the interval (where is the time of the system life);
b) the stochastic systems, where the behavior of its elements has es-
sentially stochastic character);
c) the chaotic systems, in which there is dynamical chaos (for instance,

the motion of a dense plasma beam in the case of essential collisions);
d) the complex systems, i.e. the systems characterized by the essential
complexity (the complexity concept is given below in this Chapter). The
distinguishing feature of the complex system is that it consists of a large
number of structural elements.
The nature of the first three types of systems is rather obvious. How-
ever, it is not so evident essence of the complex systems. The latter are
different in principle from others and they have a number of interest-
ing and unusual peculiarities. Our Universe, human society, the human
organism, plasmas and plasma-like systems, and many other similar ob-
jects are the most vivid examples of the complex systems.
5.4 Some General Properties of Complex Systems

The most important properties of complex systems for understanding
their nature are:
1. The uniqueness. Each complex system exists in one copy only or
else it is a very rare object. For instance, we know only one Universe,
only one human society, etc.. Any doctor can say that every human
organism is unique and exists in Nature only in one copy. Each husband
can confirm that his wife and his mother in law, as complex systems, are
unique objects and each of them exists in Nature in one copy only, too.
2. The weak predictability. One can have as much as he like detail
information about the system elements or he may know its behavior at
some interval (–T , 0]. However, this does not allow him to recognize the
exact behavior of the complex system, as a whole, in the later interval
(0, ]. The examples of the Universe, human society, a human organism,
a wife and a mother in law illustrate this property rather evidently.
3. The negentropyness or the purposefulness. A complex system can
control (in some range) its own entropy (or negentropy) for any ran-
dom influences of the external medium. As is known (see below this
subsection) the entropy characterizes the chaos within the system. Cor-
respondingly, the negentropy is the measure of order within the same
system. In the dynamical sense, the negentropy dynamics characterizes
an aspiration of the system to achieve some purpose. The purposeful-
ness expresses the analogous thing; namely, the ability of the complex
system to conserve and to amplify some main dynamical process leading
to this purpose. Therefore both these concepts are very close.
As is demonstrated later, there are such types of processes in vacuum
electronics. For instance, the cooling effect of particle beams within the
EH-accelerators (see Chapter 9, Volume II) [16, 29–31]. A corpuscular
beam, as a complex system, behaves in this case in such a manner that
the beams ‘warmth’ is taken away from the system. Here, it seems that
we have a violation of the second law of thermodynamics. However, it
is not so, because any accelerator (including EH-ubitron one) is ther-
modynamically an open system. It means that the required energy for
the ‘cooling’ procedure is taken from the surrounding medium (in the
present case from external source).
Other similar examples are the effects of phase and polarization dis-
crimination (see Chapter 12, Volume II) [31–33]. In the first case the
initial slowly varying oscillation phases of electromagnetic waves always
strive for the same magnitudes during the nonlinear resonance interac-
tion, irrespective of their initial magnitudes. In the case of polarization
discrimination the wave polarization behavior is characterized by the
analogous trend.
6. BASIC POSTULATES AND PRINCIPLES

OF THE GENERAL SYSTEM THEORY
Systems in physics and engineering are rather specific objects of the
surrounding world and this specificity is reflected by some general prin-
ciples. We point out three such principles, namely:
a) the principle of physicality;
b) the ability to model principle;
c) the purposefulness principle.
6.1 Principle of Physicality

According to this principle, processes in systems are governed by the
principles of physics. It means that in the system theory any observed
effect always has its proper cause (that, however, can be unknown).
Hence, application of the known physical principles is quite sufficient
to explain any phenomena in systems. In other words, the idea of the
existence in Nature of some unknown nonphysical ‘mystical power’ does
not find any place in this field of human activity. It should be mentioned,
however, that the latter affirmation does not deny the existence of God.
It says only that God should be considered as a real physical object,
whose activity does not contradict the basic physical principle. Hence,
the concept of God might be formulated in terms of physics. Such an
attempt is undertaken in the next Chapter, in particular.
In turn, the principle of physicality includes a few basic postulates
[28]. Let us consider the most important of them.
6.2 Integrity postulate
Integrity postulate: any complex system should be regarded as a single

system entire. This postulate in the case of complex systems is based
on the existence of the above mentioned ‘system property’. Hence, any
decomposition of the system violates the system integrity, inasmuch as
such a decomposition leads to the vanishing of this property. Let us
illustrate this thought by the following example. Let there be some
volume of a gas with a number of molecules so this object
could be classified as a complex system. We divide this volume into N
separate parts. It is supposed that there is only one molecule within each
such partial volume. It is obvious, that the integrity of the system, as a
whole, is violated as the result of decomposition. Indeed, any separate
molecule has no collective properties of the gas. In principle we can
decompose a given gas volume into parts by another method, but the
condition therein, is satisfied for each partial separate
volume is the number of molecules in a part). In this case each
partial gas volume can be represented as some subsystem of many
molecules. However, the specific ‘whole system’ property of each such
subsystem differs from the system collective properties, as a whole.
Thus the system as a whole can never be considered as an exact

equivalent of any of its partial parts (with any of its subsystems). This
postulate can also be illustrated by the so called hierarchical systems.
In this case each level of the hierarchical system (i.e. each partial hier-
archical subsystem), on the one hand, has a resemblance to the whole
system (the hierarchical resemblance principle — see below in the next
Chapter for details). On the other hand, this resemblance does not have
the total scale system nature and has a particular specific character.
6.3 Autonomy Postulate
As a rule, every type of physical phenomenon corresponds to a certain

transformation group. Such a group generates a proper geometry for the
space problem. Therefore the general theory of systems also allows a ge-
ometrical interpretation. One postulates that there is some autonomy
(with respect to a given system) type of geometry describing a complex
system. This geometry should be an invariant for any system decompo-
sition. Each part of the system is described by the same geometry which
the system has as a whole. The latter phrase expresses the autonomy
postulate.
6.4 Principle of Ability to Model

A finite number of various models can describe any given complex
system. Each model can reflect certain peculiarity of the system only. In
other words, each complex system can be modeled by various methods.
In the general theory of systems this statement is called the principle of
ability to model. A number of consequences follow from it. The most
important is that each particular group of properties of some complex
system can be studied by specific theoretical model.
Let us mention that two types of such models can be constructed.
They are particular models and the total one. Particular models can
describe a part of the system properties only. However, the particular
models only have a practical significance. Turing’s well known theorem
explains this paradox. The point is that any total model is as complex for
studying as the system itself. Therefore only partial simplified models
can be used for practical purposes really. However, let us recall that
simplifying the model we, at the same time, simplify the system analysis
by loosing part of information about true system properties. So the
choice sing of the theoretical model is always connected with a search
for some compromise between its completeness, on the one hand, and
quantity of the lost information concerning its real properties.
The other system properties can be formulated as the following pos-
tulates.
6.5 Complementarity Postulate

Niels Bohr formulated the complementarity postulate in quantum me-
chanics. However, its application in the system theory shows generality
of its nature. The essence of the complementary postulate is the follow-
ing: a complex system reacts differently revealing its various properties
under the action of different external influences. In the general case,
these reactions demonstrate the system properties which can even ex-
clude each other. For instance, an electron has corpuscular properties
(photo-effect, Compton effect, etc.) in certain cases and it behaves like
a wave (diffraction of electron on a crystal lattice, tunnel effect, etc.)
in other situations. Hence the electron is a particle and a wave simul-
taneously. Such a situation seems impossible from the point of view of
‘ordinary common sense’. However, physical experiments prove this is
true. The unusualness of the discussed situation becomes clear if we take
into account the property of ability to model the complex systems.
We can illustrate such a physical picture by the following virtual ex-
periment. Let us suppose that the object studied (for instance, the mar-
ble statue of your supervisor) is in a box. Therefore we have a possibil-
ity of observing the system behavior through only a few small apertures
which are placed on different sides of the box. It is obviously that the
observed pictures which we can see through different apertures can be
essentially different. Each such picture (i.e., the supervisor projections)
demonstrates some separate group of the object properties only. There-
fore each such group can be described by some ‘proper’ theoretical (i.e.,
particular) model. Thus in spite of the initial single object we obtain,
eventually, a number of theoretical models describing different proper-
ties of the same model. It is obviously that some of these properties can,
in principle, exclude each other.
6.6 Action Postulate

All known complex systems of people’s everyday life and different
fields of science demonstrate one interesting peculiarity which can be
formulated in the form of an action postulate. Accordingly with this pos-
tulate, any complex system reacts to an external action in the threshold
way. There are a number of examples in physics and engineering that il-
lustrate this system property. For instance, an electron transition within
an atom occurs only when the energy of perturbed photon exceeds some
threshold value (experiments by Frank and Hertz, the short-wavelength
limit of breaking radiation, the red limit of photoeffect, etc.). In this
book the efficacy of this postulate is demonstrated by examples of the
threshold effects realized in FELs and EH-accelerators (see Chapters 9–
13, Volume II).
6.7 Uncertainty Postulate

This postulate was originally introduced in quantum mechanics
(Heisenberg’s uncertainty principle). Practice shows that this princi-
ple holds for macroscopic complex systems too. It is known here as the
uncertainty postulate. Let us point out once more that it is applicable to
both: quantum and classical complex systems. Its physical nature can
be clarified by the known ‘device problem’. The point is that we have
a possibility to obtain any information about a system only through a
corresponding measuring. However, any measuring process changes the
state of the treated system. Therefore, the error of measuring is always
finite. Hence, the measured state always differs from the ‘true’ one, i.e.,
the strictly ‘true’ behavior is uncertain, in principle. In the case of a
strong enough (with respect to corresponding characteristic scale) ex-
ternal ‘device’ influence, its true dynamics can be essentially different
from the measured one.
6.8 Purposefulness Principle

In the general system theory the purposefulness is a functional trend
to achieve corresponding state of the complex system or to conserve
(or to amplify) some system process. The sense of the purposefulness
principle is that complex system has specific purposefulness behavior.
It is the most incomprehensible and intriguing feature of real complex
systems, because neither modern physics nor engineering contains any
grounds for it. However, in practice, the purposefulness is always a result
of evolution of the whole system properties. Obvious examples are the
Earth biosphere, human society, the Universe, etc.. Realization of the
discussed principle in electrodynamic systems is presented by the cooling
effect in EH-accelerators (Chapter 9, Volume II), and the effects of phase
and polarization discrimination in FELs (Chapter 12, Volume II).
6.9 Entropy and Negentropy

As has already been mentioned , entropy determines the degree of
chaos in the system. Negentropy (negative entropy) is a measure of
order in the system. Three types of entropy are distinguished: physical
entropy, thermodynamic entropy, and informational entropy [10].
Let us define the thermodynamic and physical entropies in the usual
way accepted in general physics. We assume simplest system consisting
of two identical particles moving within the volume V (see Fig. 1.6.1).
The state of volume V with both particles characterizes the microstate
of discussed system. The system shown in Fig. 1.6.1 is characterized by
only one microstate.
Then let us assume the volume V is divided into two parts (see
Fig. 1.6.2). The particles can take their positions within identical vol-
umes and in four different ways. However, only two of them really
are different. Inasmuch as both particles (and the volumes are iden-
tical, the states of the system, as a whole, described by drawings a) and
d), should be considered equivalent. The same situation takes place with
the states shown in drawings b) and c). Therefore the system shown in
Fig. 1.6.2 can have two different microstates. Let us continue this imag-
inary experiment with the increasing number of particles. Doing this,
we can obtain some generalization for the case of system consisting of
N interacting particles and partial volumes. Doing this we obtain the
expression for calculational of the total numbers of microstates W:
The number of system microstates W (1.6.1) is called the thermody-

namic probability. It, in contrast to from the ‘usual’ one, is characterized
by the following inequality: W > 1. As follows from (1.6.1), for the sys-
tem consisting of K subsystems, the total thermodynamic probability of
the system as a whole can be given as
where is the partial thermodynamic probability characterizing

subsystem. The multiplication in (1.6.2) of partial thermodynamic
probabilities is very inconvenient for real calculations. However,
it is widely known that the logarithmic function transforms the mul-
tiplication procedure into addition. Hence using the logarithm we can
introduce instead the ‘multiplied’ thermodynamic probability W the log-
arithmic (thermodynamic) entropy S:
where is the Boltzmann constant. It is evident that, in contrast to the

thermodynamic probability W , the thermodynamic entropy is more
convenient for practical use:
In what follows we consider the concept of information entropy. Let

there be some system behaving as a source of information. We assume
this source is discrete with respect to the variety of possible steady
states. Each state has (usual!) probability of realization The un-
certainty in prediction of the system state, as a whole, in a certain time
can be given as [10]:
where is the information entropy. It attains the maximum:
It can be shown (see, for instance, [10]) that both entropies are related
by some constant factor. This means that, in principle, both these con-
cepts could be considered to be equivalent in the system theory. Let us
note, however, that the information entropy characterizes more general
property of the surrounding world than the thermodynamic or physi-
cal ones. The concept of information entropy will be used further in
Chapter 2 to discuss peculiarities of hierarchical systems.
Three fundamental laws form the basis of thermodynamics. The
first is the generalization of the energy conservation law for thermo-
dynamic systems (the first law of thermodynamics). The two other laws
are specific characteristics of thermodynamic system property. There
are known seven classical equivalent formulations of the second law of
thermodynamics. For instance, the formulation of Clausius is: the en-
tropy of a closed system in thermodynamic equilibrium can only increase
or remain unchanged. According to the third law of thermodynamics
(Nernst’s theorem), the entropy tends to a finite limit, vanishing for
zero system temperature.
It should be mentioned that the applicability of entropy to describe
complex systems is an open question in modern literature (see, for in-
stance [28]). This is explained by the discussed above peculiarities of
complex systems. The author’s position coincides with the position given
in [10].
6.10 Complexity
Complexity is rather clear intuitively, however it is relatively difficult
for the formal description. We distinguish structural and algorithmic
complexity. The first one characterizes the number of structural elements
of a system. The second one is related to the degree of development
of reciprocal connections between them. However, both concepts are
closely connected.
The structural complexity can be illustrated by examples of the theory
of electrical circuits. For instance, let us assume an electrical circuit
characterized by inputs and outputs. The dynamics of the system
can be given by some multi-terminal network (see Fig. 1.6.3).
We assume that the currents flow into the system through the
system inputs, and the currents flow through the outputs. There
are some interactions between currents of both types within the system
volume. Therefore each current can correlate (in the general case)
with any current
where and are the vectors with currents and as components;

is the matrix of transformation.
It is obvious that all information about the structural complexity of
the system considered is contained in elements of the transformation
matrix In the simple case when all elements are
vanishing, besides the diagonal elements each system input is
connected with one corresponding output only. In the limit case
the system transforms into a system of noninteracting ideal conductors.
So the structural complexity is minimum in this situation.
As can be shown, the algorithm describing such system turns out to be

the shortest one. If all matrix elements are random and all the
structural complexity is maximum. It is obvious that the corresponding
algorithm describing this model should be much longer. The minimum
length of corresponding algorithm, describing the system, characterizes
its algorithmic complexity. The algorithmic complexity is used most of-
ten in informatics and cybernetics, [10]. Below, in Chapter 2, talking
about the complexity, we will have in the mind the algorithmic complex-
ity.
References
[1] R.A. Silin, V.V. Kulish, Ju.I. Klymenko. Electronic device. Inventions Bulletin,
15 may 1991. Soviet Inventors Certificate, SU No. 705914, priority of 18.05.1972,
Published in non-secret USSR press after removal of relevant stamp of secrecy:.
[2] T.C. Marshall. Free electron laser. MacMillan, New York, London, 1985.
[3] C. Brau. Free electron laser. Academic Press, Boston, 1990.
[4] P. Luchini, U. Motz. Undulators and free electron lasers. Clarendon Press,
Oxford, 1990.
[5] F.B. Hildebrand. Introduction to numerical analysis. Dover Publishers, New

York, 1987.
[6] D. Kahaner, C. Moler, S. Nash. Numerical Methods and Software. Prentice-Hall

International, Inc, 1989.

1986.
[8] N.N. Bogolubov, Mitropolskii. Methods of averaging in the theory of nonlinear

oscillations. Publising House Academy of Science of USSR, Moscow, 1963.

1981.
[10] J.S. Nicolis. Dynamics of Hierarchical Systems. An Evolutionary Approach.

Springer-Verlag, Berlin, Heidelberg, New York, Tokyo, 1986.
[11] R.K. Dodd, J.C. Eilbeck, J.D. Gibbon, H.C. Morris. Solutions and nonlinear
wave equations. Academic Press, London, 1982.
[12] M.M. Khapaev. Asymptotic methods and equilibrium in theory of nonlinear

oscillations. Vysshaja shkola, Moscow, 1988.
[13] L.D. Landau, E.M. Lifshitz. Hydrodynamics, volume 6 of Theoretical Physics.

REFERENCES 47

[15] V.V. Kulish, S.A. Kuleshov, A.V. Lysenko. Nonlinear self-consistent theory
of superheterodyne and parametrical free electron lasers. The International
Journal of Infrared and Millimeter Waves, 14(3):451–560, 1993.
limeter Waves, 18(5):1053–1117, 1997.
[17] V.V. Kulish. Methods of averaging in nonlinear problems of relativisticelectro-

[18] V.V. Kulish, A.V. Lysenko. Method of averaged kinetic equation and its use in
the nonlinear problems of plasma electrodynamics. Fizika Plazmy, 19(2):216–
227, 1993.
[19] A.P. Sukhorukov. Nonlinear wave-interactions in optics and radiophysics.


mas. Pergamon Press, Oxford, 1977.
[22] S.S. Kohmanski, V.V. Kulish. To the nonlinear theory of free electron lasers
with multi-frequency pumping. Acta Phys. Polonica, A68(5):740, 1985.
[23] S.S. Kohmanski, V.V. Kulish. Parametric resonance interaction of electron in

the field of electromagnetic waves and longitudinal magnetic field. Acta Phys.
Polonica, A68:725–736, 1985.
[24] V.V. Butuzov, V.P. Zakharov, V.V. Kulish . Parametric instability of high
current relativistic electron flux in the field of dispersed electromagnetic waves.
Deposited in Ukrainian Scientific Research Institute of Technical Information,
Kiev, feb 1983.
[25] V.P. Zakharov, V.V. Kulish. Explosive instability of electron flux in the field of
dispersing electromagnetic waves. Ukrainian Physical Journal, 6:878–881, 1985.
[26] A.M. Kalmykov, N.Ja. Kotsarenko, V.V. Kulish. Possibility of transformation

of the frequency of the laser radiation in electron flux. Pisma v Zhurnal Tech-
nicheskoj Fiziki, 4(14):820–822, 1978. (Soviet: Letters in the Journ. of Technical
Physics).
[27] D.H. Menzel. Fundamental Formulas of Physics. Dover Publications, Inc, 1960.
[28] V.V. Druzshynin, D.S. Kontorov. System-techniques. Radio i Sviaz, Moscow,

1985.
[29] V.V. Kulish, P.B. Kosel, O.B. Krutko, I.V. Gubanov. Effect of cooling of rel-
ativistic beams of charged particles during of their acceleration in the crossed
eh-ubitron fields. Pisma v Zhurnal Tekhnicheskoi Fiziki, 22(17), 1996. (Letters

in Russ. Journ. Techn. Phys.).
particles for electronic applications. The International Journal of Infrared and
Millimeter Waves, 19(1):33–93, 1998.
[31] V.V. Kulish, P.B. Kosel, A.G. Kailyuk, I.V. Gubanov. New acceleration princi-
ple of charged particles for electronic applications. examples. The International
Journal of Infrared and Millimeter Waves, 19(2):251–329, 1998.
[32] V.P. Zakharov, V.V. Kulish. Polarization effects in parametric interaction be-
tween transverse electromagnetic waves and the high current electron flow. Zur-
nal Tekhnicheskoj Fiziki, 53(10):1904–1908, 1983. (Sov. Journal of Technical
Physics).
[33] V.P. Zakharov, V.V. Kulish, S.S. Kohmanski. Effect of phase discrimination
of electromagnetic signal in modulated relativistic electron flow. Radiotekhnika
i elektronika, 34(6):1162–1172, 1984. (Sov. Journal of Radioengineering and
Electronics).
Chapter 2
HIERARCHY AND
HIERARCHICAL SYSTEMS
1. BASIC CONCEPTS OF THE THEORY

OF HIERARCHICAL DYNAMICAL
SYSTEMS
What is hierarchy in itself? Why does Nature need a hierarchy? What
are specific features of hierarchical systems? We will try in this Chapter
to give the simplest answers to these (and some other similar) questions.
1.1 Why Does Nature Need Hierarchy?

We start with the questions: what is sense of the hierarchy in the
Nature and what it is originally? It could seem strange but the mod-
ern science has not strict and clear answers to these questions. Taking
this into consideration we discuss the following such explanations, which
seem most attractive amongst various others [1–4].
The complex systems, containing a large number of elements and
connections, should be essentially unstable. Therein this instability in-
creases with growth of the system complexity. However, it is well known
that there are many examples of complex systems in Nature which are
obviously stable. The explanation of this paradox is the following. As
mentioned above, all complex natural systems have the property of pur-
posefulness. Hence theoretically some special mechanisms (purposeful-
ness) providing and supporting the stable state of such systems should
exist. Experience shows that indeed, similar mechanisms exist really.
For instance, astrophysical analysis of the cosmic catastrophes on the
Earth during the last three billions years [5] shows that our biosphere
has the obvious property of purposefulness. Every time after a next
global catastrophe the Earth reacts in a specific way such that the influ-
49
ence of any negative results of this catastrophe on the biosphere turns

out to be minimal [5]. In other words, owing to purposefulness the stable
state is reached every time by the shortest path.
As analysis has shown, the concept of purposefulness is closely con-
nected with self-organization [3]. Therefore in speaking about purpose-
fulness we, at the same time, should take in view the self-organization.
But the question arises: what is the eventual result of realization of the
purposefulness and self-organization in complex systems from the struc-
tural point of view? The answer is very simple: it is the hierarchical
systems. The point is that amongst various known complex systems
only hierarchical ones are found to be stable [1,6]. This means that ex-
isting natural complex systems self-organize by such manner that their
hierarchical arrangement always is realized eventually. I.e., we can say
that the ‘ p u r p o s e ’ o f an evolution process in any natural complex system
is forming its hierarchical structure because only such structures can be
stable really.
1.2 Two Approaches to the Theory

of Hierarchical Systems
As we noted above, conventional version of hierarchical theory (see
[1–6]) is most popular now. The most essential feature of this version
is that all hierarchical levels of a considered dynamical system are de-
scribed by the same dynamical variables. I.e., these variables are com-
mon to each hierarchical level, as well as for the system, as a whole. As
a result, mathematical structures of relevant dynamical equations for
each hierarchical level in this case, as a rule, are found to be essentially
different.
Another (new) version of hierarchical approach differs from the con-
ventional one because a proper set of dynamical variables for each hi-
erarchical level is introduced [7–16]. Therein the proper variables are
chosen in such way that the mathematical structure of the dynamical
equations for every hierarchical level (in these new variables) is roughly
the same. We call this feature the self-modeling (self-resemblance) prin-
ciple [12, 13, 15,16]. The self-modeling principle is completed by some
set of other hierarchical principles.
Hence in our version of hierarchical theory essentially new basic con-
cepts are used. The self-modeling principle is a key point of this version.
Taking this into consideration we begin discussion of the new hierarchical
concepts with the self-modeling principle.
HIERARCHY AND HIERARCHICAL SYSTEMS 51
1.3 Self-Modeling Principle

First, it should be noted that really the hierarchical idea, in itself, is
very ancient. One can find relevant explicit examples in ancient Indian,
Greek, Roman, Egyptian, American mythologies [17,18], etc.. However,
a more vivid and perfect abstraction of its original sense was given by
cosmogonic ideas of the Kabbalah [19–23]. According to the latter our hi-
erarchical world is organized within the above mentioned ‘self-modeling
(self-resemblance) scheme’. This means that every hierarchical level of
the world system and the world, as a whole, has the same formal hierar-
chical structure. This structure is called the tree of life (hierarchical tree
of life). The general idea of this concept is illustrated in Fig. 2.1.1. Ac-
cording to established tradition [19–23] the tree of life depicts complex
topological combination of specific force centers (referred as to sefirot
(sefirs)) and the relevant net of dynamical connections, which are called
the ways (see Fig. 2.1.1).
It is considered that the tree of life is an elementary universal scheme

of everything existing in our world. I.e., the Universe, as a whole, a peo-
ple, an atom, an electronic device, and many other objects of the sur-
rounded world have the same scheme of arrangement (!). Using modern
scientific language we can say that all natural hierarchical systems in
the surrounding world satisfy the self-modeling principle. It means that

each higher hierarchical level should model each lower one and, at the
same time, the system as a whole, too. (In modern philosophy, this idea
is also known as ‘the holographic principle’).
It indicates that a single fundamental principle, which should be de-

scribed by a certain fundamental equation (or a set of equations), really
determines the Universe, as a whole, as well as its all hierarchical levels.
However, the ‘little problem’ remains here: how to find this equation
really? Today we are not ready to answer this question. But we hope
that the discussed version of hierarchical theory of dynamical systems
will give for us a suitable answer in future. Unfortunately, now the
hierarchical tree can not be factorized completely (i.e., mathematically
described) in the terms of modern physics. However, we can use today
simplified versions of the self-modeling principle for practical purposes.
Namely this approach is used in the present book. Here we confine
ourselves to studying some particular hierarchical dynamical problems
only. Including that we will try to use the general self-modeling idea for
studying some oscillatative and wave resonant electrodynamic objects.
Here the ideology and methods of constructing the specific hierarchical
calculational technology, which is based on the use of the self-modeling
principle, are the main objects of our attention.
1.4 Main Idea of the Hierarchical Method
The topic discussed is interesting mainly owing to the possibility of

treating any natural nonlinear dynamical system as some kind of hier-
archical systems. So some specific mathematical method can be used
for description of such systems. We bear in mind the set of analytical–
numerical methods which are based on the use of the hierarchical concept
discussed above, generally, and the self-modeling principle, in particular.
The main idea of such hierarchical methods can be expressed as us-

ing some hierarchical properties of the treated real dynamical system for
elaboration of relevant algorithms for asymptotic integration of equa-
tions governing its dynamics. Therefore in what follows we will discuss
the hierarchical properties of real dynamical systems, which are used
further in the hierarchical calculational algorithms. Formally to discuss
relevant hierarchical algorithms, we again come back to the sacramental
question: what is hierarchy originally? At that time, however, we are
interested in the quantitative aspect of this problem.
1.5 Again: What is Hierarchy Originally?

The Structural Hierarchy
The concept of hierarchy seems so obvious that there is no difficulty
in defining it. However, the original nature of the hierarchical concept
is much deeper and more complex than it looks at first sight.
We define hierarchy as specific system of preferences with respect to
certain hierarchical parameters. However, what is ‘specific system of
preferences’ ? Looking for the answer to this question, let us primarily
concern the simplest example of hierarchical system referred to as the
‘Russian matryoshka’. The traditional its version is shown in Fig. 2.1.1.
The matryoshka consists of the series of figures distinguished from
each other by their size only. Therefore these figures can be included
one within other; i.e., the largest matryoshka (Fig. 2.1.2) contains all
smaller matryoshkas.
The characteristic sizes of the matryoshkas can be arranged as a set

of hierarchical scale parameters. We form the above mentioned ‘sys-
tem of preferences’ with respect to these scale parameters Owing to the
obviously expressed resemblance of matryoshkas, we can use only one
characteristic size as the hierarchical scale parameter. This can be ex-
plained by that all matryoshkas possess the same spatial proportion. For
example, the height of the matryoshka can be regard as an univalent
(and rather convenient) hierarchical scale characteristic.
When let us open the assembled matryoshka in Fig. 2.1.2. As a result
we may discover that it is a hierarchical series of resembling separate
matryoshkas (Fig. 2.1.3). Thus we can consider the height of any
matryoshka as a peculiar structural hierarchical scale parameter We
can construct the hierarchical series for all these scale parameters (see
Fig. 2.1.3):
here is the total number of structural hierarchical levels of the con-

sidered system, is the normalization characteristic constant (for
instance, the height one of experimentalists). But in the discussed il-
lustration example we have to do with the static hierarchical system
only, i.e., the system without any motion. Hence, it can be said that the
system (matryoshka) possesses structural hierarchy.
In what follows we can disclose one important peculiarity of the hi-

erarchical system considered. Namely, every subsequent matryoshka
contains a fewer number of structural elements (atoms and molecules
of the matryoshka’s material). It means that every higher hierarchical
level-matryoshka is less complex than the lower one (the concept of com-
plexity is discussed above in Subsection 6.10). Generalizing, one can be
convinced that this property is common for all natural hierarchical sys-
tems with structural hierarchy. We especially fix our attention to this
feature because it will become necessary later.
1.6 Dynamical Hierarchy

All hierarchical systems known in nature and society, besides the
structural hierarchy, have dynamical (functional) hierarchy, too. This
type of hierarchy is characteristic for systems with motion. Yearly, in
the case of structural hierarchy, we had to do with hierarchy of structural
scale parameters. In this (second) case relevant dynamical parameters
play a similar role. To characterize dynamical hierarchy we use the hier-
archy of velocities of varying system dynamical parameters on time. The
Universe is a most convenient illustration of the concept of dynamical

hierarchy.
The Table 2.1.1 illustrates some dynamical features of the Universe.
Information pictured is interesting and somewhat unexpected. There
are both structural and dynamical hierarchies in given natural system si-
multaneously. Besides that, a number of hierarchical levels in both these
cases are equal, similarly to the tree of life (see Fig. 2.1.1). It should
be noted especially that this fact is characteristic for many natural hi-
erarchical systems such as human society, national and world economic
systems, military systems, various electronic systems, etc..
In principle, the hierarchical systems with different numbers of struc-

tural hierarchical levels and dynamical ones can be realized. However, it
should be pointed out especially that the systems with coinciding num-
ber of structural and dynamical hierarchical level are most widespread
in the Nature. Therefore, later we shall take an interest in systems of
such a type only.
Similarly to the matryoshka, a lesser quantity of structural elements
characterizes every higher hierarchical level of any natural dynamical
hierarchical system. For instance, the total number of elementary par-
ticles in the Universe are substantially more than the number of nuclei,
the number of star systems are more than the number of galaxies, and
so on (see Table 2.1.1). However, the number of structural elements in

a system in thermodynamics is connected with degree of chaos in the
system. It means that the lower is the occupied hierarchical level the
higher is the degree of chaos in the Universe, as a hierarchical system.
Similarly, in every human social structure the chaos increases with a
decrease of number of a social level, etc..
Then we pay attention to the important fact that different dynam-
ical variables (and, consequently, corresponding dynamical equations)
describe the dynamical processes at different hierarchical levels. Or, in
other words, a proper set of the dynamical equations exists for each hi-
erarchical level of the hierarchical system. In the methodological plane
this can be treated as an inevitable sequence of appearance of the above
discussed self-modeling principle, really acting in the Nature. Indeed,
in the general case the same form of dynamical equations for different
hierarchical levels can take place really in the case only, when the proper
dynamical variables for these levels are different.
From the physical point of view this can be explained by the fact that
different types of fundamental interactions play the main role on the
different levels of hierarchy. Table 2.1.2 gives evident illustration of this
natural phenomenon.
But let us discuss this phenomenon in more detail. As mentioned

above, every hierarchical level has its proper set of the fundamental inter-
actions that determine characteristic dynamical processes here. There-
fore one or two types of interactions prevail on each hierarchical level.

In particular, dynamics of processes at the first and second hierarchi-
cal level of the Universe, as a hierarchical system (elementary particles
and nucleus), is determined by the weak and strong interactions (see
Table 2.1.2). At the same time the electromagnetic interaction plays the
main role at the third level, and so on. It should be mentioned that
other interactions also play some role here. However, the proper mecha-
nism (or mechanisms) of fundamental interactions determines a general
physical picture of the given level. For example, gravitation interaction
practically does not exert any influence at the processes at first three
levels. But it completely determines the dynamics of main processes at
higher hierarchical levels.
Later we turn to the above mentioned affirmation that simultane-
ous presence of the structural and dynamical hierarchies, as a rule, are
characteristic for natural dynamical systems. We note that both: the
structural as well as dynamical characteristic scale parameters turn out
to be essentially different (in magnitude) for different levels of the sys-
tem (see Table 2.1.1). The higher the hierarchical level the slower the
velocity of characteristic processes, and the fewer the number of struc-
tural elements. This observation opens the possibility for using the ratios
of neighboring scale parameters as relevant expansion parameters of the
dynamical problem considered. This idea is put in the basis of the hi-
erarchical method. It turns out to be very effective for construction of
different versions of the hierarchical calculational algorithms.
Thus each hierarchical system level can be characterized by dynami-
cal scalar scale parameter is a total number of
dynamical hierarchical levels). This parameter characterizes the varying
velocities of the proper (to the given level) dynamical variables. Cor-
respondingly, analogously to the case of structural hierarchy (see, for
example (2.1.1)), the relevant dynamical hierarchical series can be con-
structed in the case of dynamical (functional) hierarchy, too:
In general, as mentioned already, the numbers and are different.

However, as in the hierarchical model of Universe, their equality is more
typical in practice. Therefore, further we consider everywhere:
for simplicity, i.e., the total number of terms of both series is equal and
that each term of the series (2.1.1) has corresponding term b in the
series (2.1.2). Hence, one of two hierarchical series (2.1.1) and (2.1.2) is
enough to describe the hierarchy in a dynamical system.
1.7 Hierarchical Series in Dimensionless Form

We consider that the hierarchy in the system is evidently expressed
(like to the Universe; for instance — see Table 2.1.1), i.e., <<
Moreover, we suppose and Then
normalizing series (2.1.1) and (2.1.2) with respect to scale parameters
or we obtain relevant hierarchical series in the normalized
dimensionless form. Bearing in mind these assumptions, we rewrite the
series (2.1.1) in the so called ‘strong’ normalized dimensionless form
where is a normalized scale parameter of

hierarchical level. In the framework of the hierarchical method described
below, these parameters play the role of relevant expansion parameters
for the corresponding dynamical functions at level.
2. HIERARCHICAL PRINCIPLES.
HIERARCHICAL DESCRIPTION
2.1 Hierarchical Principles
The system of hierarchical principles is put in the basis of the dis-
cussed version of hierarchical method. These principles [10–16] are gen-
eralization of the above discussed self-modeling idea, and they summa-
rize known today experimental facts and observations. The five following
hierarchical principles can be formulated in this way, e.g. one general
hierarchical principle and four particular fundamental ones.
General hierarchical principle. Everything in the Universe is of hi-

erarchical nature. Indeed, it is difficult to find in nature or society any
object that, on the one hand, not belonging to relevant hierarchical level
of some hierarchical system and having no intrinsic hierarchical struc-
ture, on the other hand. The other four principles concern the structure,
general dynamical and thermodynamic features of the hierarchical sys-
tems.
The principle of information compression. The principle of in-

formation compression: each higher hierarchical level is always simpler
than the preceding one. The concept of system complexity is defined
above (see subsection 6.10). The Universe (see Table 2.1.1), human so-
ciety, and so on are evident examples of realization of this principle. Here
the smaller a number of structural elements (and corresponding connec-
tions between the elements), the higher a number of hierarchical level
of the system. Indeed, we always have one President (or other Head) of
a state, a few heads of other power branches, a few tens of ministers,

and so on. The millions of citizens are on the opposite end of the hier-
archical stairs. Similar examples can be found in the astrophysics, the
theory of long-time interacting electron devices [7–16] (see Chapter 4,
Subsection 1.3, and Volume II), etc..
The principle of hierarchical resemblance. The principle of hier-

archical resemblance (self-modeling principle or holographic principle):
each hierarchical level in its general properties represents the system as
a whole. In other words, the same type of dynamical equations de-
scribes as well every hierarchical level of the system and the system,
as a whole. As mentioned already, each hierarchical level is described
by proper set of dynamical variables. We regard this principle also as
a peculiar hierarchical invariance. The Boltzmann kinetic equation or
the quasi-hydrodynamic equation (or the equation for current density)
can serve as evident illustrations of appearance of the principle of hi-
erarchical resemblance. These equations have the same mathematical
structure at every hierarchical level, but every time with respect to an-
other (proper) set of dynamical variables (see relevant Chapters 6–8 and
Volume II).
The hierarchical analogue of the second thermodynamic prin-

ciple. The hierarchical analogue of the second thermodynamic princi-
ple: each higher hierarchical level has less information entropy than the
preceding one (the concept of information entropy is discussed above
in item 1.6.9). Numerous examples can be taken from our everyday
life. Indeed, one can see each lower hierarchical society level has greater
chaos. Because information from a lower level is compressed at each
higher hierarchical level that this principle has interesting consequence:
the information influence in dynamical system is directed from lower hi-
erarchical levels to higher ones, whereas the managing action is directed
oppositely. I.e., the managing influence is directed from up to down
(correlate these affirmations with the diagram in Fig. 1.1.2).
Hierarchical analogue of the third thermodynamic principle.

Hierarchical analogue of the third thermodynamic principle: the highest
level of a closed hierarchical system is characterized by vanishing infor-
mation entropy. As before, we can find the most vivid illustration in
everyday life. For instance: only one person (i.e. only one structural
element of a present system) occupies the top place in any social pyra-
mid (i.e. only one person can be, say, the President of the Ukraine at
the same time). It is obvious that the entropy of the subsystem (highest
hierarchical level as a subsystem) consisting of one element only must

be zero. So, the hierarchical principle here holds strictly. The principle
is applicable to any other natural and social hierarchical systems. How-
ever, one can readily be convinced that it holds especially ardently in
everyday social practice.
All above discussed hierarchical principles are applicable in their en-
tirety only to the closed natural hierarchical systems. There are many
another theoretical hierarchical constructions (which always turn out to
be purely artificial systems) those are open systems. These principles
can be satisfied only partially for such systems.
It may seem interesting and even incredible that all these principles
hold in multi-frequency nonlinear hierarchical electrodynamic systems
[7–16,24], too (Chapters 6–8, and Volume II).
2.2 Dynamical Equation of the Zeroth

Hierarchical Level
Let us assume that some dynamical system can be described by the
following exact differential vector equation
where is some vector and is the

relevant vector-function in Euclidean space
(the laboratory time, for example). The initial conditions are standard
ones
Concerning properties of vector-function Z and initial conditions, we

separate three different situations:
determined system, function Z satisfies the uniqueness of the solution to
equation (1.2.1) and initial conditions are given uniquely;
stochastic systems characterized by random initial conditions;
chaotic systems with properties caused by nonlinear nature of the sys-
tem and other causes. For instance, chaos can be generated by both:
nonlinear interactions of structural elements at a given hierarchical level
and a stochastic (or chaotic) component in the forming of hierarchical
structure of the system, as a whole.
Below (in this and the following Volumes) we shall discuss most of
these situations.
Now we take into account the general hierarchical principle. Let us
assume that our dynamical system possesses a hierarchical structure.
According to the principle of hierarchical resemblance each hierarchical

level of the system is described by some equation with resemble (with
respect to the expression (1.2.1)) mathematical structure. Therefore for
every hierarchical level we have proper (specific only for this level) set
of dynamical variables.
Further, we consider the hierarchy strongly expressed and the num-
bers of structural and dynamical hierarchical levels coincide. As above,
in the case of strong hierarchy (definition (2.1.3)) the set of the hierar-
chical structural and functional (dynamical) parameters is a set of small
parameters for corresponding asymptotic expansions.
Thus solving the hierarchical problem we first give an adequate formal
definition for the problem small parameters. To do this, the hierarchical
analogy of the second and third thermodynamic principles can be used.
According to the first, each higher hierarchical level has less information
entropy than the preceding one. One can prove the latter affirmation
equivalent to the assertion that each higher hierarchical level has less
characteristic velocities of the varying of dynamical variables than the
preceding one. The affirmation that characteristic velocities of dynami-
cal variables varying of the highest level of the closed hierarchical system
are zero is the consequence of the fourth hierarchical principle. In par-
ticular, the example of the Universe as a natural hierarchical dynamical
system (see Table 2.1.2) obviously illustrates both these consequences in
practice. Using these suppositions we formulate the general definition
of the hierarchical scale parameter of a problem for the level:
where is the vector of the proper variables of a hierarchical

level. In this way we can classify the initial set of variables according
to hierarchical levels. Possibility occurs to construct corresponding hi-
erarchical series of the type (2.1.3) (with the scale parameters (2.2.3)).
The initial dynamical equation (2.2.1) (the zero-level equation) generally
depends on all hierarchical small parameters (2.2.3) simultaneously:
Later we expand the right part of the equation (2.2.1) in a power

series with respect to small parameters Because hierarchy here is
expressed strongly (2.2.3), we can temporarily neglect the influence of
all other higher hierarchical levels but the nearest lower hierarchical level:
where << 1 is the scale parameter for the first hierarchical level.
The physical sense of the approximation is that influences of neigh-
boring hierarchical levels are accounted only. The model of the Universe
above can be used as illustration supporting this assumption. Indeed,
the Sun exerts the main influence upon the motion, for instance, of
the Earth. At the same time its influence upon the Moon’s motion is
inessential, whereas the motion of the latter is mainly determined by
influence of the Earth (see Fig. 2.1.2). However, other higher hierarchi-
cal levels also exert some small influence upon the given level (the other
planets and Sun weakly perturb the motion of the Moon in the grav-
itational field of the Earth). This calculational situation is illustrated
in Fig. 2.2.1. The illustration of the scheme of interactions between the
zeroth and other hierarchical levels of the dynamical system is given.
Taking into consideration the above formulated supposition about ac-
counting only neighboring interactions only, we obtain the system any
hierarchical level interacts with neighboring levels only. It means that
in such a case we can ‘travel’ through the hierarchical system only ‘step
by step’, from one level to neighboring next level and in back direction.
The calculational technology taking into account the influence of other
(higher) hierarchical interactions (see illustration in Fig. 2.2.1) can be
elaborated on the way of further developing of the approach discussed
here (for instance, by perturbations). In this book, however, we are not
interested in such calculational situations. We propose this problem for
the reader, as an teaching tusk.
2.3 Structural and Functional (Dynamical)

Operators
Now we introduce the concepts of the structural and functional (dy-
namical) operators regulating the structural and functional (dynamical)
hierarchies. To describe structural hierarchy one introduces the struc-
tural operator M
where is the vector-function defined as the right part of relevant

standard dynamical equation (of the type (2.2.5)) for hierarchical
level. In other words, every such function characterizes nonlinear dy-
namical properties of hierarchical level of a system.
Thus the structural operator M is responsible for the hierarchical re-
semblance in the discussed dynamical system. To obtain the functions
(for a known function ) various integral

transformations are used. For electrodynamic problems most promis-
ing are the averaging operator, the Fourier or Laplace transformations,
the convolution type transformations, the conformal mapping, and some
others. The differential or mixed differential–integral structural opera-
tors can be used for problems of another type. But in this book we will
discuss only integral operators (the averaging and Fourier transforma-
tions, correspondingly).
The transformation functional (dynamical) operator Û (a function in
the case of integral structural operators) describes the dynamics of con-
nections between proper variables of two neighboring hierarchical levels.
In other words, it provides the realizing of the hierarchical resemblance
(self-modeling) principle in the given system. We define the transforma-
tion functional (dynamical) operator as
As one can be convinced, the introduced operators (structural (2.2.6)

and functional (2.2.7)) also satisfy other hierarchical principles (the prin-
ciple of information compression, the hierarchical analogies of second and

third laws of thermodynamics, and the general hierarchical principle, re-
spectively).
Then we recall once more that according to the assumption
only interactions of neighboring hierarchical levels are taken into ac-
count everywhere. Correspondingly, other types of functional operators
describing different non-neighboring interactions in the system,
can be introduced in principle, too. However, as already mentioned,
in this book we study the simplest dynamical models with neighboring
interactions only.
2.4 Classification of the Hierarchical Problems

In the framework of the hierarchical theory presented in the general
case, the operators of hierarchy and are not independent
mathematical objects. In fact, one of them should always be consid-
ered given characteristic of the hierarchical system. In this connection
we distinguish two types of hierarchical problems:
1) The analysis problem — having structure of the operator
one must construct the functional operator i.e. to determine the
character of functional connections between hierarchical levels.
2) The synthesis problem — having the operator one must con-
struct the operator i.e., to synthesize the hierarchical structure of
the system.
Up to now the main attention in literature was paid only to the anal-
ysis problem. In should be noted that the theory and calculational tech-
nology of the synthesis problem is not being developed today. Therefore
in this book we also have to do with the analysis problem only.
In turn each mentioned problem can be divided into a direct and its
inverse problem. The straight problem (see Fig. 2.2.2) is characterized
by the formulating of a problem for the highest hierarchical level.
Further, the relevant functional dependencies for other lower hierarchical
levels are constructed by special calculational.
In the case of the inverse problem (see Fig. 2.2.3) the initial problem is
formulated for a certain hierarchical level conditionally called the zeroth
level. Then we ‘lift’ along the ‘vertical hierarchical axis up’ toward the
highest hierarchical level and, therefore transform this problem into some
higher (highest for example) hierarchical level.
Specific feature of the hierarchical methods which are discussed in this
book is that here both the above mentioned problems are used at the
same time. At the first stage of the solution algorithm the initial prob-
lem is transformed into some higher (highest, for instance) hierarchical
level (the straight problem — see Fig. 2.2.2). Then solving the problem
at this level we further ‘slip down’ the found solutions (the inverse prob-
lem — Fig. 2.2.3). In what follows we find required solutions for the
zeroth hierarchical level (see also Fig. 1.1.2 and relevant commentaries).
Namely this version of hierarchical calculational technology is the main
object of our interest in this book.
Then we turn our attention to the problem of numbering of hierar-

chical levels in a dynamical system. There are a few methods of the
numbering. In this book we used a peculiar method smaller numbers
correspond to levels that are more complex. Therein we conveniently
consider some level as the zeroth one. Numbering in such a way, we
can realize, in principle, negative numbers of hierarchical levels also. It
should be mentioned that the discussed case can not be
realized within the framework of other versions of numbering methods
[2].
2.5 Hierarchical Tree

To illustrate hierarchical structures rather obvious graphic method
is used. It is called the hierarchical tree method. Its graphic figures
are called the hierarchical trees [2] (it should not mistaken the concepts
of (hierarchical) tree of life (see Fig. 2.1.1) and hierarchical tree, cor-
respondingly). The simplest version of the hierarchical tree is shown
in Fig. 2.2.4. Here the specific hierarchical tree is given, where each
two structural elements of some hierarchical level form one element of
the level one step higher. We illustrate this graphical situation by the
following example.
Let it be some abstract military sub-unit each chief has two followers
only. On the top of hierarchical structure a general stays, who has two
colonels as direct subordinates. It is obvious (that is important) that
without these subordinates, the general ‘is not a general’ functionally.
Thus colonels are responsible to forming one step higher level — the level
of the general (unfortunately, generals in a real life very often forget this
rather important circumstance).
Further, we suppose that each colonel has lieutenant colonels repre-

senting the level of lieutenant colonels and governing two majors, and
so on. It is obviously that any real military chief has, as a rule, more
than two followers. However, in the given illustration example this is
not important. By this we only try to illustrate one of basic features
of all known hierarchical structures. Namely, some group (that is spe-
cific for the given system) of structural elements at any hierarchical

level forms one structural element at the hierarchical level,
and so on. This idea is illustrated in this example by the following man-
ner. The general without his subordinates ‘is not a general’ functionally,
because in such case he could be lose any possibilities to be a ‘general
leading activity’. Analogously, without major-subordinates colonels ‘are
not colonels’, and so forth.
Then we turn our attention to the highest point of the hierarchical
tree in Fig. 2.2.4. This point plays very important role for understand-
ing of physical nature of real dynamical hierarchical structures. It is
well known that any general has his own military authority. In turn,
this authority also has some higher authority, and so on. On the top of
the structure in Fig. 2.2.4 the point ‘the highest military authority’ de-
picts all possible (including, unknown) higher authority. It is important
that this point also represent some hierarchical system. However, the
structure of the latter can not be disclosed in the framework of lower its
part. Indeed, all structural element of the system pictured in Fig. 2.2.5
should known all down structure of managed ‘branch’ of the hierarchi-
cal tree. But any such element should know only one higher chief. In
turn, this chief knows ‘all down structure of managed ‘branch’ of the
hierarchical tree’, and so on. Or in other words, any information in the
considered hierarchical system is directed from lower hierarchical levels to
higher ones, whereas the managing action is directed oppositely (it is the
sequence from the hierarchical analogue of the second thermodynamic
principle).
Thus we can affirm that highest separate point in the discussed hi-
erarchical system embodies the ‘highest military power’ in this military
structure.
The analogous, in principle, situation is realized in our Universe, too.
The hierarchical tree of the Universe, given in Fig. 2.2.5, illustrates this
observation. It is obvious that contrary to the previous thought model a
number of structural elements at each hierarchical level can be essentially
much more than two. However, as mentioned above, this circumstance
is not important: both hierarchical trees really are similar.
Further we again turn to discussion of the problem concerned to the
highest point on the top of hierarchical tree. Analyzing all known today
natural complex hierarchical dynamical systems, one can be convinced
that all relevant hierarchical threes (without any exceptions!), analo-
gously to Fig. 2.2.4, Fig. 2.2.5, and Fig. 4.1.1, always contain a corre-
sponding separate point at the top. Hence we have weighty grounds to
affirm that this property is universal for our world and it is characteris-
tic for all hierarchical dynamical systems existing in the Nature. On the
other hand, our Universe also can be considered as a natural dynamical

complex system. Consequently, it, in principle, should also contain such
an element, in spite of the circumstance that it strictly is not yet fixed
in official astrophysics. On the other hand, the question arises: what is
the physical sense of the highest point in the diagram in Fig. 2.2.5 in
view of the treating of the analogous point in Fig. 2.2.4?
Thus following to this logic we inevitably come to the fundamental
problem: what hierarchical level (and the separate point on its) really
occupied higher the ‘metagalaxy’ level? What is the real physical sense
of these point and level? The author considers that this point (and level,
correspondingly) represents all unknown for us highest (with respect to
our world-hierarchic system) hierarchical levels. This can be treated
as a physical definition of the concept of God. Thus as a result of our
reasoning we obtain a possibility for formulating of the concept of God
in traditional terms of the physics.
It could seem incredible, but, as shown below in Chapter 4, analyzing
wave resonance (oscillation–resonance) processes in electronic devices
with long-time interactions [11, 12, 16] (see Fig. 4.1.1 and corresponding
comments) the analogous hierarchical trees can be obtained, too. In
this case the discussed highest separate point represents some source
of energy for acceleration of the electron beam that comes in input of
the system. Taking into consideration our above reasoning we can say
that functionally this source could be treated as a peculiar ‘god’ for this
REFERENCES 69
device-hierarchic system. In this connection it is interesting to note that

general hierarchical approach, which is used here (the theory of hier-
archical waves and oscillations [7–16, 24]), really have essentially wider
framework of utilization than it seems at first sight. As corresponding
analysis shown, the most natural hierarchical systems (including, Uni-
verse, human society, etc.) can be successfully described in the frame-
work of this wave-like (oscillatative-like) approach. Moreover, the hier-
archical asymptotic calculational methods, some of them are described
in this book below (see also [7–16, 24]), can be used for studying of these
objects also.
Thus the concept of hierarchical tree indeed turns out to be universal
and very effective means for description of various dynamical hierarchical
systems of different nature.
References
[1] J.S. Nicolis. Dynamics of Hierarchical Systems. An Evolutionary Approach.

Springer-Verlag, Berlin-Heidelberg-New York, Tokyo, 1986.
[3] H. Haken. Instability Hierarchies of Self-Organizing Systems and Devices. Ad-
vanced Synergetic. Springer-Verlag, Berlin-Heidelberg-New York- Tokyo, 1983.
[4] Kaivarainen. Hierarchical concept of matter and field. Earthpuls Press, 1997.
[5] B.M. Vladimirskij, L.D. Kislovskij. The outer space influences and the biosphere
evolution, volume 1 of Astronautics, Astronomy. Znanije, Moscow, 1986.
[6] V.V. Druzshynin, D.S. Kontorov. System-techniques. Radio i Sviaz, Moscow,
1985.
investigation. Ukrainian Physical Journal, 36(9):1318–1325, 1991.
227, 1993. (Sov. Plasma Physics).
superheterodyne and free electron lasers. The International journal of infrared
and millimeter waves, 14(3):451–568, 1993.
limeter Waves, 18(5):1053–1117, 1997.
Visnyk Sumskogo Derzshavnogo Universytetu, 1(7):3–11, 1997.
particles for electronic applications, hierarchical description. The International
Journal of Infrared and Millimeter waves, 19(l):33–93, 1998.
Journal, 43(4):83–499, 1998.
[14] V.V. Kulish, P.B. Kosel, O.B. Krutko, I.V. Gubanov. Hierarchical method and
its application peculiarities in nonlinear problems of relativistic electrodynam-
ics. theory of eh-ubitron accelerator of charged particles. Ukrainian Physical
Journal, 43(2):33–138, 1998.
[15] V.V. Kulish. Hierarchical method and its application peculiarities in nonlin-
ear problems of relativistic electrodynamics. single-particle model of cyclotron-
resonant maser. Ukrainian Physical Journal, 43(4):98–402, 1998.
to nonlinear problems of relativistic electrodynamics. Causality and locality
1998.
[17] Andre Nataf. Dictionary of the Occult. Wordsworth Editions Ltd, Hartfordshire,
1988.
[18] F. Kapra. Dao of physics. ORIS, St. Petersburg, 1994.
[19] M. Laitman. Kabbalah. Israel, 1991.
[20] D. Fortune. The mystical Qabalah. Alta Gaia Books, New York, 1989.
[21] N. Garbo. Cabal. New York-London, 1978.
[22] C. Ponce. Quest books. Kabbalah. The Thesophical Publishing House, Wheaton,
Illinois, USA Adyar, Madras, India, 1995.
[23] W. Parfitt. The New Living Qabalah. Element, Shaftesbury, Dorset Rockport,
Massachusetts, Queesland, 1995.
dynamics. World scientific Publishers, Atlanta, 1998.
Chapter 3
HIERARCHICAL ASYMPTOTIC
METHODS. GENERAL IDEAS
A specific feature of this Chapter is that it is recommended to be

read two times. The first time should help to ‘find one’s bearing on the
ground’. The second might be helpful for the obtaining entire picture
of the considered scientific direction. A reader (we fear) after acquain-
tance with the book might be too overloaded by the variety of proposed
calculational schemes and peculiarities of their practical application. So
we hope that this Chapter can help to generalize main impressions and
obtained knowledge in the field of the hierarchical theory and the hier-
archical methods, in particular.
We shall discuss below the main ideas and peculiarities of the hierar-
chical calculational method. It will be shown how specific calculational
ideology can be developed immediately from the above discussed charac-
teristic properties of hierarchical dynamical systems. In particular, the
general idea of hierarchical approach by direct and inverse problems of
analysis (see Figs. 2.2.1–2.2.5 and relevant commentaries) will be illus-
trated quantitatively (i.e., by means relevant mathematical description).
As mentioned above a specific combination of these problems allow for-
mulating the general calculational hierarchical problem. But let us make
a few remarks on this topic.
All hierarchical problems (and, correspondingly, hierarchical systems)
could be divided into oscillatative and wave ones. The first of them are
characteristic for so called lumped systems and they could be described
by relevant exact differential equations. Their typical examples are the
problems about motion of a charged particle in some external electro-
magnetic field (see example set forth below in Chapter 5). The equations
like to (2.2.4), (2.2.5) may be proposed as a general formulation of such
type problems.
71
Characteristic feature of the problems of second type is description

of the studied object by some system with partial derivatives. Such ob-
jects usually is classified as the distributed systems. The problems about
propagation of different nature waves in various mediums (including non-
linear) are described, as a rule, by the system with partial derivatives.
Examples of such type will be given in Chapter 7 and Chapter 9 (Chap-
ter 8, Section 7).
The problems of both type are the object of attention in this book.
They are deeply correlated one with other. Therein mathematicaly spe-
cific features of methods for their solution, as well as physical nature of
the considered objects determine this. That is why, in spite of their
unlikeness, the mathematical structure of methods for their solution
roughly is found to be the same. Therein some of methods for solution
of the second type problems (like to the method of averaged characteris-
tics — see Chapter 6), include immediately the methods for asymptotic
integration of ordinary differential equations as a key point of the calcu-
lational algorithm. Other methods (such as the slowly varying amplitude
method — see Chapter 8) use also the general ideology developed for
the problem of the first type, but for the system with partial derivatives
only.
Thus the main objects of our future discussion are the mathematical
description of the straight and inverse problems in framework of the
hierarchical theory of lumped and distributed oscillatative wave systems.
1. DETERMINED HIERARCHICAL SYSTEMS

1.1 Determined Hierarchical Systems
Let us begin our general discussion by the studying some hierarchi-
cal system described by initial dynamical equation like (2.2.4). It is
considered for simplicity that we deal with the determined hierarchical
systems (see the system classification given in the preceding Chapter).
Then, accepting the supposition discussed above about interacting only
neighboring hierarchical levels, we come to the ‘hierarchicised’ equation
for the zeroth level of type (2.2.5). We use the definition (2.2.7) (formal
solutions for equation (2.2.5)) for the connection between variables of
the zeroth and first hierarchical levels:
Differentiating expression (3.1.1) by the time we consider the dy-

namical equation for the zeroth hierarchical level (2.2.5). After transfor-
mations, we yield the so called generalized hierarchical equation of the
HIERARCHICAL ASYMPTOTIC METHODS. GENERAL IDEAS 73
first order for the unknown functional operator (transformation func-

tion)
The first term of equation (3.1.2) is product of Jacobi matrix

of the order and the column vector It is con-
sidered that the structural operator (see definition (2.2.6) and cor-
responding comments) is given. It means that, in view of definition
(2.2.6), the new dynamical function of the first hierarchical level
can be considered as a known.
In the coordinate form the equation (3.1.2) can be written as
Accordingly to the resemblance hierarchical principle, relevant dy-

namical equation for the next hierarchical level should has mathemati-
cal structure analogous to (2.2.5). But at that time it takes place with
respect to new (proper) dynamical variable set of the first level only:
The new dynamical equation for the first hierarchical level (3.1.4), in
correspondence with the above discussed hierarchical principles, should
be simpler that the initial equation for the zeroth hierarchical level
(2.2.5). Therefore, the equation of this type usually referred to as the
truncated equations. Relevant solutions of equation (3.1.4) allow to find
the dynamical (functional) operator (see equations (3.1.2), (3.1.3)),
and as the result, to construct the solutions for variables of the zeroth
hierarchical level (see definition (3.1.1)). Owing to application of the
hierarchical method, this calculational algorithm is found to be actually
simpler than in the case of solving the initial problem (2.2.4) immedi-
ately.
In the case the discussed dynamical system, besides the zeroth, has
two higher levels of hierarchy else (the first and the second, correspond-
ingly) , we could follow in the analogous way. Namely, we consider prob-
lem (3.1.4) as a new initial one (analogous to problem (2.2.4)). This
means that it could be solved by the analogous method (see also illus-
trations in Fig. 1.1.2 and corresponding commentaries):
Thus summarizing, we can formulate the general idea of the hierar-

chical method. This idea consists in the use of the fact that each new
truncated dynamical equation for a higher hierarchical level is sim-
pler than the preceding one. At the same time, it possesses the analogous
mathematical structure with the initial (and preceding) equation. (This
is a result of practical realization of the compression and resemblance
hierarchical principles — see the preceding Chapter). This means that
the solution procedure for the dynamical equation of the last hierarchi-
cal level should be essentially simpler than the analogous procedure for
initial equation of the zero level.
Further we propose some unexpected analogy of the discussed calcu-
lational scheme with one like situation in biology. In this case the trun-
cated equation of the highest hierarchical level (3.1.9) can be treated as
some peculiar ‘gene’ of the treated dynamical system. Whereas the com-
plete set of equations (3.1.8)–(3.1.10) we can treated as some ‘embryo’.
Within this ‘gene’complete set of properties of the original system (2.2.4)
are ‘enciphered’. This situation is like that, as in the ‘true’ gene the in-
formation about some organism is enciphered. Providing some ‘culture
medium’ for our ‘mathematical gene’ one can evolve it into the original
system (2.2.4) (similarly, for instance, to situation with developing of
the human embryo). In our formal case the functional operators
and the generalized hierarchical equations of the type (3.1.2), (3.1.7),
(3.1.10), etc. play the role of such peculiar ‘culture medium’ for devel-
oping our ‘equations embryo’ (3.1.10). It is clearly that in this case we
indeed can hope for simplifying of general solution procedure because
any embryo in nature always is simpler the whole organism.
Thus this analogy once more illustrates that all natural hierarchical
dynamical systems apparently have a common nature. Moreover we
might hope that these ‘common nature’ could be described in framework
of the proposed new hierarchical approach.
At last, we should mention that this method, generally, has some
difficulties. Most of them originate from peculiarities of integration
of generalized hierarchical equations (3.1.2), (3.1.3), (3.1.7), (3.1.10).
These difficulties relate to the special form of a chosen structural op-
erator. It is obvious that such calculational schemes have of practical
interest in the one case only. Namely, in the case when the total labor-
intensity of the joint solution procedure (for hierarchical dynamical equa-
tions (2.2.4), (3.1.4), (3.1.6), (3.1.9) and corresponding generalized ones
(3.1.2), (3.1.3), (3.1.7), (3.1.10)) is less than the procedure of solving
initial equation (2.2.1) ‘immediately’. This result can be attained re-
ally if some analytical algorithms could be constructed for the solving
generalized hierarchical equations like to (3.1.2), (3.1.3). This holds at
a practice for special structural operators M only. General situation is
simplified in the case of periodic (oscillatative) systems. A number of
structural operators of such kind can be constructed for such situations.
It should be mentioned, however, that most widely used such operators
are based on the well known Riemann theorem of integral averaged value
[1] :
Let us discuss some of structural (averaging) operators M, which can

be constructed on the basis of Riemann theorem (3.1.1).
1.2 Averaging Operators

Averaging (smoothing) operators satisfy the above mentioned labor-
intensity test in optimal manner. It is explained by the wide prevalence
of these operators in the problems of electrodynamics and mechanics.
Most often, averaging operators are used for some explicit function of
time space coordinate or of the and simultaneously. Formally,
these operators can be constructed as
the averaging operator on spatial coordinates, and
is the averaging operator on time. The condition of the function period-

icity in the latter case
is satisfied. In the first case (3.1.12) the function is supposed

on all components of vector within the
domain of definition i.e.,
where by the vector is denoted. If

the function Z is periodical only on a part of coordinates the
operator (3.1.12) can be determined as
if
It means that the averaging procedure is performed only on part of

the variables and that the function Z is periodical in these
variables only. While integrating in (3.1.13), the variables are

considered as relevant constants. Here and below in this subsection we
neglect hierarchical index
In spite of formal resemblance of both versions of the averaging op-
erator, their nature is essentially different. Namely, the averaging in
(3.1.13) is performed over the independent scalar variable (the labo-
ratory time, for instance). Whereas, in similar procedures (3.1.12) we
have dealing with the components of some vector But in the general
case, the latter s relevant function of variable (This functional de-
pendence is determined by dynamical equation (2.2.1)). Nevertheless,
the coordinates are considered as constants at performing relevant
averaging procedures in (3.1.12). It should be noted that procedures
of constructing of corresponding averaging operator with accounting of
dependencies of the type is possible also (see, for instance, [2–4]).
But below in this book we will not have to do with such situations.
In practice mixed version of function periodicity (in time
and coordinates of vector simultaneously) is realized. In view of
definitions discussed, the construction of averaging operator is obvi-
ous. However, let us mention one special case of the mixed averag-
ing operators. It realizes in situations, when acting forces are of wavy
form. The discussed in Chapter 1 free electron laser (FEL) (analyzed
also in Chapters 8–11, Volume II) is evident case of mixed problem.
Characteristic peculiarity of the latter is linear combination of time
and space coordinates. Such special combinations are
called scalar phases (see, for instance,
definition (1.3.7) in Chapter I). The function (written in
parametric form with respect to ) is periodical in components of vector
By here the following vector is denoted
where are components of the vector of cyclic frequencies and

are components of the wave vectors in space:
In problems of electrodynamics and mechanics the case of function

expandable in a Fourier series takes a special place. It means
that the Dirichlet–Jordan theorem [5] about development in a Fourier
series holds here. Then the Fourier series express the func-
tion Z in the domain For the last case the relevant Fourier expansion
can be written as
where is the norm of the q-dimensional harmonic

vector the components of which are integer numbers
is a scalar product of two q-
dimensional vectors:
According to these definitions, numbers play the role of harmonic

numbers and are Fourier amplitudes.
2. STOCHASTIC HIERARCHICAL SYSTEMS

The general characteristics of the oscillation-like problems have dis-
cussed above in the previous Section. In what follows let us dwell brief on
the hierarchical wave-like problems. Therefore the stochastic distributed
systemss are chosen as a convenient illustration example of such kind.
2.1 Factors of Stochasticity in Dynamical Systems

Let us discuss mathematical peculiarities of the description of stochas-
tic hierarchical systems. Within the framework of the presented ap-
proach, the stochastic properties of a model can be determined by three
different causes:
a) random character of initial conditions for scalar components of
vector for corresponding equations (2.2.1), (3.1.4), (3.1.6), (3.1.9) (for
instance, a spread in thermal particle velocities in plasma-like system
input);
b) stochastic dynamics in physical mechanisms described by functions
(collision processes in kinetic systems);
c) stochasticity of the constructing the hierarchical tree, i.e. by the
non-synonymy of relevant solutions for the transformation functions
Due to stochastic mechanisms, we have the transformation of initially

determined hierarchical tree into a stochastic ensemble of hierarchical
trees (about the concept of hierarchical tree see in Chapter 2, Subsec-
tion 2.5). Here we discuss only general ideas of possible stochastic ver-
sions of the presented hierarchical theory. Later (see, for example, Chap-
ters 12, 13, Volume II) we will consider and illustrate various particular
features of these systems in the framework of the method of averaged
kinetic equation (see Chapter 6).
2.2 Example of Hierarchical Model

of Stochastic System
For simplicity we consider that only the first two of above mentioned
processes determine stochasticity of a system. To describe this system
we construct distribution function
where are the components of and

are scale parameters. So the dynamics of phase volume is
determined by the single-particle dynamics of components In fact,
amongst these components there are space coordinates of particles and
their velocities.
The motion of each separate particle is described by single-particle
equation of the type (3.1.4), (3.1.6). In the stochastic model discussed
here , the fact stochasticity of motion is determined by only the first two
causes, i.e., random character of initial conditions for scalar components
of vector and stochastic dynamics in physical mechanisms described by
functions respectively. Besides that we assume ‘determined part’
can be separated from the function (see (3.1.4), (3.1.6)) in
additive way:
where the function describes the stochastic component of the particle

dynamics (for instance, stochastic addendum in the Langevin equation).
So, in the case the pure stochastic model realizes. In the opposite
case the stochasticity of the system is determined by stochasticity
of initial (boundary) conditions only. Further, we accept for simplicity
that discussed stochastic system has properties of a quasi-continuous
medium. We differentiate the definition (3.2.2) in time and change the
momentum differential into particular derivatives traditionally
In what follows we consider single-particle equation (2.2.5) and as-

sumption (3.1.2). As a result we obtain the so called kinetic equation:
Quite often the function describes the collision processes in

stochastic systems. Therefore, we call the right side of the equation
(3.2.4) the collision integral Then the equation (3.2.4) can be
rewritten in the following more compact form:
where is the collision scale parameter. In the case

of the 6-dimensional space (three space coordinates plus three compo-
nents of canonical momentum), we formally reduce (3.2.4), (3.2.5) to
traditional form of the Boltzmann equation.
Thus we see that stochastic dynamics of the system, as a whole, really
is determined by single-particle dynamics of its elementary parts. This
circumstance allows us to use the version discussed above of the hierar-
chical approach to investigating kinetic systems, too. Indeed, using the
formula (3.3.1) and definition (3.2.5) we reduce kinetic equation (3.2.4)
to the form with the total differential
where all notations are obvious in view of context of the about said. It is
seen that mathematical structure of equations (3.2.6) and, for example,
(2.2.5) is the same. Hence, hierarchical calculational scheme set forth
above can be applied for solving the stochastic problem, too. However,
there are here some methodical difficulties. The latter are connected
with some peculiarities of calculational procedures for back transforma-
tions from the system with total derivatives to the system with particular
derivatives. The essence of idea of these transformations and the tech-
nical ways of overcoming of the mentioned difficulties will be discussed
below in other Chapters. In particular, one example of such hierarchical
technology is described in this book in a form of the method of averaging
characteristics, method of averaged quasi-hydrodynamic equation, and
the method of averaged kinetic equation, respectively (see, for example,

Chapters 6–8 in this Volume and in Chapters 12 and 13, Volume II).
Here we only point out that generally problems of such a type are par-
ticular variety of the wider class of problems called the hierarchical wave
ones. Let us shortly discuss some characteristic properties of systems
described by such kind equations.
3. WAVE RESONANT HIERARCHICAL

SYSTEMS
As we mentioned above, some peculiar aspects of analyzing dis-
tributed (wave) systems require relevant modifications of hierarchical
methods. The mathematical specificity here is reflected by partial
derivatives in equations governing the processes in the system. There-
fore all problems of this kind can be divided into two separate groups.
Namely, those problems which can be reduced to systems of exact dif-
ferential equations, and those that cannot be reduced to such systems.
Some class of wave systems described by so called quasilinear equations
(with respect to the partial derivatives — see Chapter 6 for more de-
tails), including hyperbolic equations [6], quasi-hydrodynamic equation,
Boltzmann kinetic equation (see Chapters 6–8, and Chapters 12, 13 in
Volume) [7], etc., serve as explicit examples of the first kind. The list
of the second kind is much wider. We mention here the nonlinear wave
equations discussed by theoreticians, such as Korteveg–de Vries, Boussi-
nesq, modified Korteveg–de Vries, nonlinear Schrodinger, sine-Gordon,
and other equations [8]. Unfortunately, the averaging method in its clas-
sical form is not applicable in most cases of this type. The only excep-
tion is the class of weak nonlinear problems. In particular, the problems
like to the problem about wave propagation in the media with weak
non-linearity dispersion (about these concepts see in Chapter 1, Sec-
tions 2 and 3). Other numerous examples of such type could be found in
the physics of condensed matter, spectroscopy of atoms and molecules,
etc.. In electrodynamics problems of this kind were arose at first in
studying the radio wave propagation in layers of the Earth atmosphere.
M.A. Leontovich [9] proposed and accomplished an idea similar to the
Van der Pol [10] and averaging (Krylov–Bogolyubov) [6,11–20] methods.
The amplitudes of interacting waves are taken as the slowly varying vari-
ables (see Chapter 1, Subsection 2.4). Since the non-linearity is weak,
the solution formally reproduces the solution of the equivalent linear
system. The distinction is that wave amplitudes became slowly vary-
ing quantities and their dynamics is described by smoothed (truncated)
equations (equations of the first hierarchical level — see Chapter 8).
Similarly to the Van der Pol and Krylov–Bogolyubov approaches,

the evolution of the slowly varying amplitude method undergoes nec-
essary stages of generalization and mathematical substantiation. A.V.
Gaponov, M.I. Rabinovich, and L.A. Ostrovsky [21–23] have done this
most completely. However, practice shows the application of their ver-
sion in electrodynamic problems requires modification of the calcula-
tional procedure [2,24–26]. The latter should account the specific fea-
tures of physical conditions in plasma-like electrodynamic systems. For
instance, it concerns various effects of nonlinear generation of quasi-
stationary electromagnetic fields, realization of degenerate mechanisms
of wave resonant interactions, etc.. The modified version of such a type
is described and illustrated below in Chapters 8, 12 and 13.
3.1 Standard Hierarchical Equations

in the Case of Hierarchical Wave Problem
N.M. Krylov and N. N. Bogolyubov [11,12] elaborated (in the form
of the averaging method [6,11–19]) the first calculational approach for
asymptotic solution of nonlinear oscillation problems that we now regard
as hierarchical one. Bogolyubov’s algorithm, in itself, is the simplest
version of the two-level hierarchical calculational scheme (see Section 5).
So, hierarchical technologies developed here (see Chapters 4–8) can be
considered as relevant generalization of Bogolyubov’s ideas [6, 7,12–15,
19].
As mentioned previously, the essential peculiarity of the methods of
such class (as well as their hierarchical variety) is that they have to do
with exact differential equation, which are written in some specific forms.
In references these forms are called as standard systems (see, for exam-
ple, equations (1.1.4), (2.2.1), relevant materials of Chapter 4, and Sec-
tion 5). Such calculational situations are typical, as mentioned above, for
lumped systems. Apart from the lumped systems the distributed systems
are rather widespread in electrodynamics. As mentioned above, the use
of particular derivatives is most characteristic point for formal methods
of description of the distributed systems, from the mathematical point
of view. Therefore, at the first step of the constructing of relevant hi-
erarchical calculational procedures for the case of distributed systems
(i.e., the systems with particular derivatives) we should formulate also
the analogous concept of a standard system.
In what follows let us remind the above said that general ideology of
the hierarchical calculational approaches in both these cases are rather
close. This circumstance in some specific cases allows to consider the
calculational technologies developed for systems with exact differential
equations and systems with particular derivatives from some common
point of view. In other cases we have not such possibility. However,

different types of equations possess their different specific features that
can be expressed in the form of corresponded calculational procedures.
Therefore below we undertake the attempt to classify the wave reso-
nant problems with respect to this criteria. The concept of the general
standard form is put in the basis of such classification.
Thus we consider dynamical wave system described by the following
differential matrix equation, which we below consider general standard
form:
where A, B, C are square matrices of size is some

vector-function in Euclidean space with coordinates
i.e.
P is some linear differential operator in the space
is a given weak nonlinear vector-function, is some scalar vari-
able, for instance, the laboratory time. The equation (3.3.1) in the
coordinate form can be written as
where are the elements of the matrices A, B, C, are ele-

ments of the operator P; are the components of the vectors U
and R.
One may be convinced that Maxwell’s equations, the Boltzmann ki-
netic equation, the quasi-hydrodynamic equation, and many others can
be reduced to general standard form (3.3.1), (3.3.2) [6,13–18]. The rel-
evant examples are given below in Chapter 8 and Chapters 12, 13 in
Volume II). Unfortunately, we should state that up to today there is no
general algorithm for integration of equations (3.3.1), (3.3.2) immedi-
ately. Therefore researchers usually consider particular cases only. We
also follow this way.
At first we use hierarchical principles (see Chapter 2 for more details).
We begin with general hierarchical principle that the considered dynam-
ical system has a hierarchical structure. Relevant dynamical equations
for each higher level are simpler than for preceding one (the principle
of information compression). Moreover, each hierarchical level of the
system is described by equations of the same mathematical structure
(the principle of hierarchical resemblance). Therefore for each hierar-
chical level we have set of another dynamical variables proper to it (the
principle of information compression), and so on. Then we introduce

structural and functional operators.
The structural hierarchical operator can be introduced similar
to (2.2.6):
where is the given vector-function defined before (the right part of

(3.3.1)). This function characterizes nonlinear dynamical properties of
hierarchical level. It is obvious that the operator (3.3.3) satisfies
all hierarchical principles, too.
The transformation functional operator (some function in the given
particular case) describes the dynamics of connections between proper
variables of two neighboring hierarchical levels. We can determine it
analogously to (2.2.7):
Both operators of hierarchy are not independent mathematical ob-

jects. We consider some integral structural operator is0given. Then
we act on equation (3.3.4) by the differential operator defined by the left
side of equation (3.3.1). Further, we use the second hierarchical princi-
ple with equation (3.3.1) and definition (3.3.4). The general hierarchical
transformation equation can be obtained on this way
where required vector dependences and

can be found by definition (3.3.4); is inverted
operator these operators satisfy to the follow-
ing condition: where I is the unit vector. Here
also the corresponding operator after relevant transforma-
tions is written in the variables of hierarchical level. In this
case we choose it in the simplest differential form:
Therefore, with given structural operator (or the inverse operator

the equation (3.3.5) gives us relevant hierarchical transfor-
mation function (3.3.4). Further, we assume the function
weakly nonlinear and expandable in Fourier series.
It should be mentioned that these assumptions hold for wide class of
electrodynamic problems. One type of such problems formally correlates
with the problem of a charged particle motion (beam) in predetermined
electromagnetic fields [3,4,7,27–35]. The self-consistent dynamics of
plasma-like systems is the second example [2,7,24–26].
3.2 Classification of Problems

As mentioned above, the hierarchical algorithms can be conditionally
divided into two groups:
a) the algorithms reducible to integration of standard ‘single-particle’
equation (1.3.4);
b) the algorithms reducible to the so called Rabinovich standard form
[21–23].
Let us in what follows shortly to comment specific features of both
these cases.
Case a). This case realizes whether if some calculational procedure

for constructing of some equivalent set of exact differential equations can
be accomplished really. A number of procedures of such type are known
in references. They are the method of characteristics (see Chapter 6,
Section 3), the method of lines, the method of total differential (see
above expression (3.2.3) and corresponding commentaries), and some
others. Let us illustrate this idea in the simplest example of the method
of total differential. We suppose for this that two first terms in left side
of can be rolled up to the total derivatives because (3.2.3):
where I is unit matrix, the expression in the right side of the second
equation is the product of the diagonal Jacobian matrix of order
and column operator is a nonlinear dynamical vector func-

tion defining single-particle motion in the system. One may be confident
that the stochastic hierarchical problem discussed above can be reduced
to the form (3.3.1), (3.3.6). Using (3.3.6) after relevant transformation
equation (3.3.1) can be rewritten as
where is given nonlinear wave vector-function of

hierarchical level. Further, we introduce a new extended coordinate
vector for hierarchical level:
and a new extended nonlinear dynamical vector-function
Therefore the initial system (3.3.1) obtains the form of relevant ‘true’
single-particle equation like (3.3.1) or (2.2.1):
In the particular case we obtain ‘pure single-particle’ version

of the discussed hierarchical problem.
Thus a spatially distributed wave resonant dynamical problem de-
scribed by equation (3.3.1) in the case (3.3.6) can be reduced to quasi-
single-particle problem (3.3.10). This means that all calculational tech-
nologies, developed for asymptotic integration of the‘single-particle’ ex-
act equations (see Sections 1, 4–6, and Chapters 4, 5), can be successfully
used for solution of this type distributed problems. However, there are
some essential differences between the calculational algorithms in the
‘purely lumped’ and ‘distributed’ cases, respectively. Namely, it is read-
ily seen that equation (3.3.10) allows to find solutions for the function
in the form Whereas the initial form of initial ‘dis-
tributed’ equation (3.3.1) supposes the eventual solutions’ dependency
of the form This problem is known as the back
transformation one. Specific calculational technology is developed for
its solution. Let us shortly discuss the main ideas of such calculational
technology.
Constructing of the truncated equation for the hierarchical level
and solving it according to the general hierarchical scheme, we obtain

formal solutions for level in the form (see definition (2.2.7)
and corresponding commentaries):
Taking into account definitions (3.3.8) and (3.3.9), we rewrite solution

(3.3.12) as
where all notations are evident. Further, we expand the function

in a series in small vicinity of the point
where is the difference between and which can be

found from the corresponded expansion of the function in small
vicinity of the point
In what follows, we perform the analogous procedure with hierarchical
levels and and so on. Moving by this method ‘backward’,
we eventually can reach the ‘last’ zeroth level
After performing expansion in a series (3.3.15) in small vicinity of the

point we obtain the looked for dependency
3.3 Krylov–Bogolyubov Substitution

As shown below, the physical situation with the function
in the right part of equation (3.3.10) periodical (or nearly periodical, or
conditionally periodical) is most interesting. Therein experience shows
that the most promising versions of hierarchical method can be con-

structed within averaging methods (in more detail such type calcula-
tional algorithm is described below in Section 5). Let us say a few words
about such calculational scheme based on Krylov–Bogolyubov substitu-
tion. This is interesting for another reasons, too. The point is that
historically N.M. Krylov and N.N. Bogolyubov (and many followers af-
ter) [6,12–19] namely began studying the averaging methods. The latter
can be regarded as the first practically effective realization of the hier-
archical idea in practical calculational algorithm.
We assume that the function in (3.3.10) is periodical and slowly
varying
where the ‘prime’ is ignored further for simplicity. In this case, the
functional operator can be found from the Krylov–Bogolyubov sub-
stitution:
Relevant general hierarchical equation for the vector-function

(3.3.2) can be written in some another (with respect to (3.3.2)) form
Here, we would like especially to stress that equation (3.3.18), as it

will be shown below in Section 5, can be solved analytically [11, 12].
Due to this, the application of hierarchical scheme based on substitution
(3.3.17) turns out to be very effective.
3.4 Case b)
But let us turn to the problem discussed above in Subsection 3.2.
Assume that function R in standard form (3.3.1) can be developed in a
Fourier series. Separate out the zeroth Fourier harmonics
where is the oscillatory part of the function R. Then we consider

that function R can be represented as a series in powers of the highest
hierarchical small parameter
With (3.3.19), (3.3.20) we can write the standard equation (3.3.1) as

a set of two equations
where are relevant expansion coordinates,

Each of equations (3.3.21) describes essential different class of the ob-
jects of studying. In particular, the first of the latter governs a ‘smooth’
(quasi-stationary) dynamics of the system. For example, it might con-
cern a generation of some ‘smooth’ quasi-static electric and magnetic
fields within a plasma-like relativistic system (see, for instance, Chap-
ters 12, 13 in Volume II). Standard numerical methods or exact or ap-
proximate calculational methods can be used for solving this part of the
problem.
The second of equations (3.3.21) describes some nonlinear dynamics of
waves in the system, including the wave resonant interactions of different
kind. Characteristic feature of this problem is presence a number of
fast oscillations. At that, the general situation is complicated by the
non-linear and resonant nature (quite often) of these wave interactions.
The problems of such types can be solved by the use of hierarchical
analytical–numerical methods.
There are three versions of solution to the discussed problem are de-
scribed further in the book:
1) Application of the slowly varying amplitudes method (see Chap-
ter 8, Sections 1–3).
2) Application of the method of averaged characteristics (see Chap-
ter 6, Section 2).
3) Application of the hierarchical transformations of coordinates
method (see Chapter 8, Section 4).
Therefore accordingly with the above said, we can distinguish two
versions of the slowly varying amplitude method:
a) traditional version (see Chapter 8, Sections 1, 2) that is rather
popular in acoustics, nonlinear optics, radiophysics, etc.;
b) modernized version (see Chapter 8, Section 3), which takes into

account specific peculiarities of non-linear wave resonant electrodynamic
problems of plasma-like systems.
Following any of these paths we can obtain some truncated (short-
ened) equations of the first hierarchy and relevant set of transformation
relationships. Accordingly to the general hierarchical idea further we
again use the described calculational scheme, separating the following
(in a hierarchical series) small parameter But at that time this doing
with respect to the truncated equations of the first hierarchy, and so on.
Ultimately, after exhausting terms in the relevant hierarchical series, we
obtain some truncated equations for the highest hierarchical level. Solv-
ing these equations and fulfilling corresponding back transformations we
eventually obtain complete solution of the initial problem.
4. VAN DER POL’S METHOD

4.1 A Few Introductory Words
The mathematical averaging procedure has been employed in celestial
mechanics as early as in Gauss’ time [19]. It became with time a basic
element of the algorithm for the approximate solution of the systems
of nonlinear differential equations. But this happened only after the
Flemish radio engineer Van der Pol had applied the averaging idea to
propose a simple refined method of solving the nonlinear problems of
electrical and radio engineering [10].
In spite of being insufficiently substantiated, the method has been
widely employed by practical engineers, while mathematicians thought
it was not worth their attention, in the same manner as they had
treated Heaviside’s papers on operational calculus. N.N. Bogolyubov
and N.M. Krylov initiated the crucial changes in 1930s. They employed
Van der Pol’s ideas to develop a rigorous mathematical approach to the
analysis of nonlinear ordinary differential equations of some specific type.
Having separated the slow and fast variables, they presented the solution
in the asymptotic series form, with the first term reproducing the rele-
vant Van der Pol result [12,13]. Later on, the approach has been devel-
oped and substantiated by Ukrainian mathematicians of Bogolyubov’s
school (Yu.A. Mitropolsky, B.I. Moiseenkov, A.M. Samoilenko and oth-
ers) [6,14,15].
4.2 Van der Pol’s Method’s Variables

Let us suppose that some lumped oscillatative system is described by
the following second order equation:
where is the spatial coordinate, is some constant (say, linear cyclic

frequency — see its concept above in Chapter 1, Section 2), is the
small parameter of the problem, is a given (in the general case,
nonlinear) function of coordinate and velocity Here the symbol‘dot’
is total (material) derivative on time It is importantly
that the requirements to the function are not too strict — it can
be, e.g., discontinuous. The condition implies that the system
(3.4.1) must be weakly nonlinear (about this concept see in Chapter 1,
Section 2). It means that it describes nearly linear oscillations.
Then we introduce the generating equation:
which immediately follows from (3.4.1) for As readily seen, the

generation equation (3.4.2) is an ordinary linear differential equation
(compare with (1.2.8)) whose solution is well known to be (see, for in-
stance, (1.2.9))
where is the amplitude, and
is the linear oscillatative phase (see Chapter 1, Section 2). The variables
and are referred to as Van der Pol variables. Van der Pol’s main idea
is to assume that the solution of the nonlinear equation (3.4.1) is formally
described by the same expression as the solution of the linear equation
(3.4.2). The difference is that the amplitude is slowly varying function
of The physical meaning of the term slowly varying function is that
the amplitude weakly changes during one oscillation period
(for more details about this concept see in Chapter 1, Section 2). The
result becomes appreciable only after a long time which includes
many oscillation periods If the amplitude is slow,
then the phase for reasons given above, is the fast variable. It is
reasonable to assume that the weak nonlinearity does not modify the
general behavior of the solutions though gives rise to small nonlinear
corrections in the relevant weakly nonlinear equation. In what follows
we derive such equations from the slow amplitude and the fast phase.
We require that the new functions and satisfy, along (3.4.3),
the condition
We differentiate (3.4.5) with respect to and substitute the result in

(3.4.1), to obtain:
Moreover, we differentiate (3.4.3) with respect to and make use of

the condition (3.4.5), to obtain:
Thus we have derived a set of two differential equations for and

Solving the equations with respect to and reduces the system
(3.4.6)-(3.4.7) to the form:
Expressions (3.4.8), (3.4.9) evidently illustrate concepts discussed

above of slowly varying and fast variables. Indeed, as readily seen the
ration of velocities of changing variables and is proportional to the
small parameter
Taking into consideration the definition for the hierarchical small pa-
rameters (2.2.3) we can consequently conclude that these variables are
related to different hierarchical levels of the system. Besides that, we
can detect that the Van der Pol’s description discussed above could be
treated as simplest and earliest version of the hierarchical approach in
the theory of oscillations and waves.
4.3 Truncated Equations and

Their Hierarchical Sense
The set of equations (3.4.8), (3.4.9) is nothing but an alternative ver-
sion of the initial equation (3.4.1) and, in principle, it has no evident
advantages over (3.4.1). However, the form (3.4.8), 3.4.9 suggests the
way of how to find the approximate solutions. Following Van der Pol, we
replace the slow varying amplitude by its time averaged value, using
averaging operator (3.1.13), i.e.,
For the phase we substitute:
Here the symbol ‘overline’ is averaging symbol. Then we carry out

the time averaging in (3.4.8) and (3.4.9) with regard for (3.4.11) and
thus obtain an approximate system of equations:
where
Equations (3.4.13), (3.4.14) are referred to as truncated. In terms of

the hierarchical ideology they are the equations of the first hierarchical
level. Small parameter in this case, as mentioned already, can be
treated as the hierarchical scale parameter of the first hierarchical level
(compare definitions (3.4.10) and (2.2.3), respectively). Thus we have
the possibility of being convinced that the Van der Pol method is indeed
the simplest realization of the hierarchical methods. Schematically idea
of the Van der Pol method is illustrated in Fig. 3.4.1.
Deriving the system of truncated equations (3.4.13), (3.4.14) (i.e., the
equations of the first hierarchical level) is the most important point of
the Van der Pol method. This system possesses simpler mathematical
structure (principle of information compression) than both the initial
equation (3.4.1) and the system (3.4.8), (3.4.9), since equations (3.4.13)
and (3.4.14) can be disconnected. Indeed, here one can solve the first
equation independently and write the solution for the phase in quadra-
tures. As a result, solving the truncated equations actually is simpler
than any procedure of integrating (3.4.1) directly. Moreover, in many
problems of electrodynamics the researcher is interested in obtaining the
rate of change of the phase rather than the phase in itself.
The quantity has physical meaning of the nonlinear cyclic oscilla-

tion frequency of the system. In particular, as will be shown in Chap-
ter 8, Volume II, the dependency (to be more exact,
is basic for realization of the elementary physical mechanism, which
provide the operation of many electronic devices. Including, Free Elec-
tron Lasers (FEL), Cyclotron Resonance Maser (CRM), etc.. With
being given, the slow dependence is immediately determined by
truncated equation (3.4.14), i.e.,
which makes this approach even more attractive for the researcher.
The idyllic situation is violated, however, by some difficulties. In
particular, it had been unclear for a long time whether the solutions ob-
tained by the Van der Pol method are sufficiently accurate, i.e., to what
extent the exact solutions of the initial equation (3.4.1) are equivalent to
those of the truncated systems (3.4.13), (3.4.14). This made the essence
of the mathematicians’ main object against the Van der Pol method
[36]. This aspect has disturbed the practical engineers far less. They
have been quite satisfied with the observation that in most cases the re-
sults obtained by this method had been in fairly good agreement with the
experimental data. Few rare cases of obvious discrepancies did not influ-
ence the favorable attitude. However, the negative experience accumu-

lated as the range of application of the method expanded. The problems
were solved by N.M. Krylov and N.N. Bogolyubov, then Yu.A. Mitropol-
sky, D.I. Zubarev, A.A. Andronov, L.S. Pontriagin, A.N. Tichonov,
V.N. Volosov, E.A. Grebennikov and other [6,12–19]. Van der Pol’s
ideas obtained a new mathematically rigorous form and made a founda-
tion for an extended mathematical theory, Bogolyubov’s approach being
its part. The latter made it possible to develop efficient asymptotic
integration algorithms of the hierarchical type for a wide spectrum of
differential equations.
5. METHODS OF AVERAGING. STANDARD

VERSION
The main object of application of the hierarchical versions of the av-
eraging methods is the mathematical model, which can be described by
the following vector equation:
where is some vector and is the

relevant vector-function in Euclidean space
(for instance, the laboratory time), are the hierarchical small param-
eters. In the present Section and below in Chapter 4 we, predominantly,
will have to do with particular varieties of this model, where only influ-
ence of neighboring hierarchical levels really is important. In this case,
as shown above in Chapter 2, Subsection 2.4, the general hierarchical
problem (3.5.1) can be reduced to a successively catching hold of two-
level hierarchical systems (see illustration in Fig. 2.2.1). As a result we
eventually obtain that the basic element of any hierarchical problem of
the type (3.5.1) is the algorithms of asymptotic integration of the sim-
plest two-level hierarchical system. Therefore, here and below in next
Chapter we will pay the main attention to investigation of two-level hi-
erarchical systems of various types. Our studying we begin with the so
called Bogolyubov’s standard system [19].
5.1 Bogolyubov’s Standard System

Traditionally the particular variety of the general systems (3.5.1)
is referred to as Bogolyubov’s standard system. Here we accept for sim-

plicity: Let us construct the asymptotic algorithm for its in-
tegration, basing on averaging operator (3.1.13). This means that the

function for the first hierarchical level can be written in the
form:
We assume that the function is on t, i.e.,
for all
Accordingly with the general hierarchical scheme (see above Section 1)
the dynamical equation for the first hierarchical level should have similar
to (3.5.2) mathematical structure:
The connection of the proper variables of the zeroth and first hi-
erarchical level we find in form of the Krylov–Bogolyubov substitution
(transformation — see also (3.3.17) and commentaries)
The generalized hierarchical equation of the first order (3.1.2) for the
unknown functional operator (transformation function) in this case
can be obtained in the form:
Then we expanse the function Z in a Fourier series (see (3.1.19) and

relevant comments):
Using the expansion (3.5.8), (3.5.9) we rewrite general hierarchical

equation (3.5.7)
Further we accept the following supposition: the function is

considered analytical with respect to within domain In this case
we have a possibility of constructing a formal solution for [19].
We will look for change of variables (3.5.6) in the form
Then we substitute the change (3.5.11) in general hierarchical equa-

tion (3.5.10) equalizing here terms of equal order (with respect to the
small parameter ) in left and right sides, respectively. As a result after
relevant calculations we obtain the following catching hold of system of
equations
Very important feature of system (3.5.12) is a possibility to obtain its

approximate analytical solutions:
For obtaining of total solution of initial problem (3.5.2) (zeroth hier-

archical level) we must:
a) to substitute solutions (3.5.13) into (3.5.11);
b) to find relevant solutions of the dynamical equation of the first
hierarchical level (3.5.5) and to substitute their into (3.5.11);
c) to coordinate initial conditions.
The last item means that we should to choose the arbitrary functions
by such way the equation is satisfied
Thus it is shown that in view of above set forth general idea of the
hierarchical approach (see Sections 1–3), the Bogolyubov method in-
deed can be regarded as a particular variety of hierarchical method.
More obviously this thought is illustrated in the diagram represented in
Fig. 3.5.1.
In principle we might also have some other algorithm for the discussed
version of the averaging method [19]. Such algorithm can be realized if
the idea of generating equation (see (3.4.2) and comments) is used. (In
reference such equations are also called to as equations of comparison
[19]). In this case the structural operator can be chosen in the following
manner
Respectively, the dynamical equation of the first hierarchical level can

be written as
Further calculational scheme is the same. However, it should be men-

tioned that in a general case the form of asymptotic solution here turns
out to be some differ of the solutions (3.5.11)–(3.5.13) [19].
5.2 The Problem of Secular Terms

Unpleased peculiarity of the above discussed versions of the calcula-
tional scheme on the average method is presence in asymptotic solutions
(3.5.13) secular terms of the type This problem can be solved
(completely or particularly) by using of the modernized version of the
calculational scheme. Let use to choose the following form of the struc-
tural operator
where are some unknown (a this stage of the calculational

procedure) functions, as before, is the zeroth term of a Fourier
series of type (3.5.8). Then we follow according to the above described
calculational scheme. As a result, instead system (3.5.13) we obtain the
modernized system
The next stage of the modernized version of the calculational scheme

is the determining of the unknown functions The idea of this
procedure follows from comparison of systems (3.5.12) and (3.5.18), re-
spectively. It is understood that the functions can be found, for
example, from the condition of elimination of relevant secular terms by
corresponding functions As a result of realization of this idea we
could obtain the following definitions for these functions
The method described above of elimination of the secular terms we

will use also below in other versions of the averaging method.
6. METHODS OF AVERAGING.
TWO-LEVEL SYSTEMS WITH
SLOW AND FAST VARIABLES
6.1 Two-Level Systems with Slow and
Fast Variables. General Case
Bogolyubov’s standard system of the type (3.5.2) represents the sim-
plest object for application of the averaging methods. Historically such
systems have been the first fully fledged realizations of the hierarchical
ideas [12,13]. Later a few other more general versions of such realiza-
tions have been developed [6,14–19]. Let as illustrate one of theses ‘more
general versions’ at an example of so called system with slow and fast
variables.
HIERARCHICAL ASYMPTOTIC METHODS GENERAL IDEAS 101
The Bogolyubov’s standard system (3.5.2) can be regarded as a par-

ticular case of the more general system
Indeed, assuming
we easily come to the standard form (3.5.2). Thus, as is readily seen,

Bogolyubov’s standard form (3.5.2) is the particular case of general sys-
tem (3.6.1), when the function is slowly varying function on
time But such a particular case is not the only one
possible. Let us discuss some other particular case. Namely, we have in
view the situation, when different velocities of changing of components
of different groups of the vector takes place. In other words, we might
say that the obviously expressed dynamical hierarchy of initial variables
with respect to the velocities of their changing is a specific characteristic
of this case. For simplicity we choose the case of two-level hierarchy:
where represents the slow part of the vector and represents the
fast part of it. The standard system of such kind can be presented in
the form
where X are the vectors, Y and are the

vectors, The vectors Y and are determined
within some space
The vector comprises, as components, the scalar cyclic frequencies

(angular velocities of rotation) Hence we might say
that equations (3.6.4) describes a multi-frequency two-level hierarchical
system with slow and fast variables. Respectively, the vector is referred
to as the vector of slow variables, and the vector is the vector of fast
variables.
It is not difficult to show that, as mentioned above, Bogolyubov’s

standard system (3.5.2), in itself, is the simplest particular case of the
system with slow and fast rotating phases. Indeed, we rewrite (3.5,2) in
the parametric form (with respect to the parameter and for
to obtain
i.e., the standard system (3.5.2) can be considered to be equivalent to

the standard form with slow and fast phases (3.6.4) in the following
particular case
Then we should to point out that the general algorithm for obtaining
of analytical solutions of the discussed type systems is not known today
[19]. Therefore, at a practice some particular cases of the systems similar
to (3.6.5) are use, as a rule. The systems with fast rotating phases
represent one of such particular cases. Owing to their specific features
namely these systems are found to be most interesting for practice.
6.2 Two-Level Systems with

Fast Rotating Phases
The particular case of the systems with slow variables and fast
phases (3.6.4) is referred as to the system with fast rotating phases if
In this case the fast variable is called as to the fast rotating phase
Thus the two-level multi-frequency hierarchical standard system with

fast rotating vector phase can be written as
REFERENCES 103
Besides that another version for the form of writing of hierarchical

standard system with fast rotating phases also will be used in this book
below
where is the large scale parameter of the problem. The various

algorithms of asymptotic integration of the systems similar to (3.6.9),
(3.6.10) (an some more complex) are set forth below in the following
Chapter.
References
[1] M.M. Khapaev. Asymptotic methods and stability in theory of nonlinear oscil-
lations. Vysshaja Shkola, Moscow, 1988.
limeter Waves, 18(5):1053–1117, 1997.

Visnyk Sumskogo Derzshavnogo Universytetu, 1(7):3–11, 1997.
particles for electronic applications, hierarchical description. The International
Journal of Infrared and Millimeter waves, 19(l):33–93, 1998.
[5] G.A. Korn, T.W. Korn. Mathematical handbook for scientists and engineers.
McGraw Hill, NY, 1961.
[6] Y.A. Mitropolsky, B.I. Moseenkov. Lectures on the asymptotic applications to

the solution of partial differential equations. Mathematics Institute, Kiev, 1968.

dynamics. World Scientific Publishers, Atlanta, 1998.
[8] R.K. Dodd, J.C. Eilbeck, J.D. Gibbon, H.C. Morris. Solutions and nonlinear
wave equations. Academic Press, London, 1982.

in the earth atmosphere. Akad. Nauk SSSR, Phys. Ser.(8):6–24, 1944. ser.
Fiz. (Bull. Acad. Sci. USSR, Phys. Ser.).
[10] B. Van der Pohl. Nonlinear theory of electric oscillations. Russian translation.
Svyazizdat, Moscow, 1935.
to nonlinear problems of relativistic electrodynamics. Causality and locality
1998.
[12] N.M. Krylov, N.N. Bogolyubov. Application of the methods of nonlinear me-
chanics to the theory of stationary oscillations. Ukrainian Academic Science
Publishers, Kiev, 1934.
[13] N.M. Krylov, N.N. Bogolyubov. Introduction to nonlinear mechanics. Ukrainian
Academy Science Publishers, Kiev, 1937. English translation: Princeton, New
Jersey: Princeton Univ. Press., 1947.
[14] N.N. Bogolubov, Ju.A. Mitropolskii. Methods of averaging in the theory of
nonlinear oscillations. Publising House Academy of Science of USSR, Moscow,
1963.
[15] N.N. Bogolubov, D.N. Zubarev . Asymptotic approximation method for the
system with rotating phases and its application to the motion of charged parti-
cles in magnetic fields. Ukranian Mathem. Zhurn, 7:201–221, 1955. Ukrainian
Mathem. Journal.
[16] V.I. Arnold. Applicability conditions and error estimates for the averaged
method applied to the resonant systems. Dok. Akad. Nauk. SSSR, 161(1):9,
1965. Sov.Phys.-Doklady.
[17] A.A. Andronov, A.A. Vitt, S.E. Khaikin. Theory of oscillations. Fizmatgiz,
Moscow, 1959.
1981.
1986.
[20] S.S. Kohmanski, V.V. Kulish. To the classic single-particle theory of free elec-
tron laser. Acta Physica Polonica, A66(6):713–740, 1984.
[21] A.V. Gaponov, L.A. Ostrovskii, M.I. Rabinovich. One-dimensional waves in
nonlinear disperse media. Vysh. Uchebn., 13(2):169–213, 1970. Ser. Ra-
diofizika (Sov. Radiophysics).
[22] M.I. Rabinovich, V.I. Talanov. Four lectures on the theory of nonlinear waves
and wave interactions. Leningrad, 1972.
[23] M.I. Rabinovich. On the asymptotic in the theory of distributed system os-
cillations. Dok. Akad. Nauk. SSSR, 191:253–1268, 1971. ser. Fiz. (Sov. Phys.-
Doklady).
REFERENCES 105
[24] V. V. Kulish. Nonlinear self-consistent theory of free electron lasers. method of

the nonlinear problems of plasma electrodynamics. Fizika Plazmy, 19(2):216-
227, 1993. (Sov. Plasma Physics).
Journal, 43(4):83–499, 1998.
[28] V.V. Kulish, P.B. Kosel, O.B. Krutko, I.V. Gubanov. Hierarchical method and
its application peculiarities in nonlinear problems of relativistic electrodynam-
ics. theory of eh-ubitron accelerator of charged particles. Ukrainian Physical
Journal, 43(2):33–138, 1998.
[29] V.V. Kulish. Hierarchical method and its application peculiarities in nonlin-
ear problems of relativistic electrodynamics. single-particle model of cyclotron-
resonant maser. Ukrainian Physical Journal, 43(4):98–402, 1998.
[30] S.S. Kohmanski, V.V. Kulish. To the nonlinear theory of free electron lasers.
Acta Physica Polonica, A68(5):749–756, 1985.
[31] S.S. Kohmanski, V.V. Kulish. To the nonlinear theory of free electron lasers
with multi-frequency pumping. Acta Physica Polonica, A68(5):741–748, 1985.
[32] S.S. Kohmanski, V.V. Kulish. Parametric resonance interaction of an electron
and the field of electromagnetic waves and the longitudinal magnetic field. Acta
Physica Polonica, A68(5):725–736, 1985.
[33] V.V. Kulish, I.V. Dzedolik, M.A. Kudinov. Movement of relativistic electrons
in periodically reversed electromagnetic field. Deposited in Ukrainian Scientific
Research Institute of Thechnical Information, Kiev, 23 jul 1985. Uk-85. Part I.
110 pages.
[34] V.V. Kulish, I.V. Dzedolik. Movement of relativistic electrons in periodically
reversed electromagnetic field. Deposited in Ukrainian Scientific Research In-
stitute of Thechnical Information, Kiev, 20 sep 1985. Uk-85. Part II. 54 pages.
[35] V.S. Jakovlev, V.V. Kulish, I.V. Dzedolik, V.G. Motina, S.S. Kohmanski . Gen-
eration of energy by relativistic electrons moving in the field of two electro-
magnetic waves in presence longitudinal magnetic field. PrePrint of University
Electrodynamics Academy of Sciences of Ukraine, 1983. 41 pages.
[36] L.I. Mandelshtam and N.D. Papaleksi. On substantiation of an approximate
method of solving differential equations. Zh.Eksp.Teor.Fiz, 2:220–234, 1934.
Sov. Phys.-JETP.
Chapter 4
WITH FAST ROTATING PHASES
The hierarchical systems with fast rotating phases are the most ‘pop-
ular’ (amongst other asymptotic hierarchical methods) in electrodynam-
ics applications. Firstly, because the physical models which can be de-
scribed by such equations, are widespread in a practice. Secondly, be-
cause, as mentioned above in Chapter 3, Section 6, analytical algorithms
of asymptotic integration of systems of this type can be constructed
without excessive difficulties. Taking this into consideration we will pay
special attention in this and the following Chapters for studying the
systems with fast rotating phases of different types.
1. HIERARCHICAL OSCILLATIONS
1.1 A Few Introductory Words
We begin this Section with discussion of some basic ideas of the theory
of hierarchical oscillations in electrodynamic systems. Therefore here we
will take an interest, mainly, in the mathematical part of the problem.
This problem, as will be understood latter, can quite often be reduced to
the asymptotic integration of the hierarchical systems with fast rotating
phases.
In what follows it should be pointed out that the purely mathematical
part in training the theory of hierarchical systems, in itself, is not the
only source of the principal difficulties for a student or an expert. Let
us turn the reader’s attention to another side of the discussed object.
As mentioned above in the Preface, this book is intended first of all
for students and experts concerned with physical electronics, accelera-
tion devices, plasma electrodynamics, radiophysics, etc.. Initially the
idea of the book originated from the lecturing experience gained by the
107
author when delivering specialized lecture courses in vacuum physical

electronics. The obtained the experience turned out to be somewhat un-
expected. Namely, the author was astonished to discover that students
easily cope with the mathematical aspects of the hierarchical method
(basic concepts, calculational algorithms, etc.) but find serious difficul-
ties in dealing with the two practical aspects. Moreover, very often the
author finds the same situation even in discussions with mature experts.
Taking this into consideration, let us briefly comment on the initial na-
ture of these difficulties.
The first difficulty is to formulate the problem in the terms of hierar-
chical methods. In particular, it is to reduce the initial sets of equations
to one of those mentioned above in Chapter 3, Sections 5 and 6 standard
forms. As noted earlier in Chapter 3, the asymptotic integration algo-
rithms of this method are valid only for the so called standard systems.
The point is that the general procedure of reducing arbitrary systems
of equations to standard forms is not known. Quite efficient special ap-
proaches, which have been proposed for particular groups of problems,
require deep understanding of fine qualitative distinctions in the occur-
rence of physical processes under consideration. As a result, to formulate
the problem adequately, at least in electrodynamics, has turned out to
be a skill that is not so easy to acquire. Therefore below we begin exposi-
tion of the hierarchical calculational technologies (including those based
on the hierarchical systems with fast rotating phases) from the discus-
sion of the problem of adequate mathematical description (factorization)
of electrodynamic systems in the terms of hierarchical methods.
The second difficulty concerns the physical analysis and interpretation
of the results obtained. The terminology of the hierarchical method is
rather special. The picture of the phenomena to be studied turns out
to be described in terms which are not traditional for ‘ordinary’ electro-
dynamics. That is why one needs some training in order to understand
a physical picture described in ‘hierarchical terms’. For example, this
might concern an evolution of a slow combination phase of the charged
particle. This evolution is really associated with particle nonlinear dy-
namics in the vicinity of some resonance. Or another example: a certain
combination of slowly varying complex wave amplitudes describes self-
consistent variations of wave polarizations owed to the nonlinear wave,
etc.. A number of special methodical examples are given below interac-
tion (see Chapters 7, 8 and Chapters 9–13, Volume II) for illustration of
methods of overcoming these kind of difficulties. The author advises for
the reader to study in detail these illustrations for better understanding
of methodical and calculational peculiarities of the methods of discussed
type.
HIERARCHICAL SYSTEMS WITH FAST ROTATING PHASES (1) 109
1.2 Formulation of the Hierarchical

Single-Particle Electrodynamic Problem
Taking into consideration the above said in item 4.1.1, we start with
discussing of some general features of the hierarchical version of classical
single-particle electrodynamic problem. It is the problem of periodical
(quasi-periodical) motion of a charged particle in some electromagnetic
field in the discussed case.
We accept the superposition that the acting electromagnetic field con-
sists of the wave and non-wave parts, correspondingly. with that the
latter (non-wave part) is assumed slowly varying, non-periodic quasi-
stationary and quasi-homogeneous fields. Thus the arrangement of the
electric and magnetic fields acting on the charged particles can be rep-
resented in the form:
where and are the electric field intensity and magnetic induction
field. and are slowly varying complex
amplitudes of electric and magnetic induction fields of partial waves
(proper and stimulated); are slowly varying
quasi-stationary and quasi-homogeneous electric and magnetic induction
fields, and are wave harmonic numbers ( ).
The quantities
are phases, is the angular frequency, is the wave vector, is the

laboratory time and is the position vector. In the case of proper waves
we have the dispersion relations between and All
these concepts are discussed in Chapter 1 (see Chapter 1, Sections 2
and 3 for more details).
In terms of general hierarchical theory (see Chapters 2 and 3), assign-
ing form (4.1.2) to the fields determines form of the chosen structural
operator (see (2.2.6)).
Further below (in this Volume and in Volume II) we illustrate some
practical features of the calculational procedures studied by using exam-
ples of various electrodynamic systems. The peculiarity of the latter is
that they are always related to some objects like charged particle beams
or plasmas fluxes which move within electromagnetic fields. Usually,

the single particle and multi-particle (including collective) problems are
distinguished.
In this book we are interested in electron, ion or plasma beams of
two extreme types: low and high intensity beams, respectively [1–3].
The characteristic feature of beams of the first type is that intensity of
Coulomb interaction between the charged particles is relatively low. In
fact, such beams are just streams of nearly independent drifting charged
particles. Hence we have a possibility to study of a separate particle in-
teraction with the fields. The single-particle theory of motion of charged
particles in electromagnetic fields is a result of such study. In turn the
single-particle problem is the main point of this theory. The particu-
lar case of the multi-particle problem can be realized here by simply
summing up of all separate single-particle interactions.
In contrast, the particle–particle Coulomb interactions can be rather
intense in high intensity beams. At sufficiently high densities the beams
show a collective behavior and they are treated as a quasi-continuous
flow of drifting charged (or quasi-neutral) plasmas. Hence, in this case
the charged particle beam should be regarded as a whole electrodynamic
object.
It is obvious that owing to the differences discussed above of the beams
of both types the mathematical descriptions of these objects should olso
be different. Below in this Chapter we give hierarchical algorithms for
treating of various single-particle systems. Some methods of asymptotic
treating of the high density beams will be set forth below in Chapter 5.
Thus the beam’s motion of the first type (rarified plasma beams)
can be considered as the motion of aggregate of individual particles
(Lagrange formalism). In this case, the beam motion can be described
by the single-particle equations in the Hamiltonian form (1.4.22):
or by the Lorenz equation, which can be regarded as a particular case

of the Newton equation (1.4.4) (because where is
the Lorenz force):
Here is the Hamiltonian of a

charged particle of type is its charge ( for electron and
for an ion, e is electron charge, Z is charge number of the

ion), is the rest mass of particle of sort is vector-potential
and is scalar potential of the electromagnetic fields [4]; is
the non-dimensional particle velocity; here is the velocity, c is light
velocity in vacuum, is the canonical momentum, is the position
vector of particle and are defined in Chapter 1, Section 4 already;
is the relativistic factor (here ).
The hierarchical part of the problem consists of two main stages. The
first is the reducing the equations (4.1.3) or (4.1.4) to one of standard
forms. The second is the asymptotic integration of the standard equa-
tions. In the book we study initial sets (4.1.3), (4.1.4) reduced to the
standard form hierarchical system with fast rotating phases [5–7]. The
simplest two-level versions of such systems we above discussed shortly
in Chapter 3, Section 6 (see equations (3.6.9), (3.6.10)). In the general
case of hierarchical oscillation systems the standard system with
fast rotating phases can be written in the following form [5–7]
where scale parameters are the relevant terms of the strong

hierarchical series (see for comparison (2.1.3)):
and x is the vector whose components are slowly varying variables only,
are partial vectors of different hierarchy of the vector of fast rotation
(revolving) phases are slowly varying part of velocity vector of fast
rotating phases, are relevant vector-functions, is hierarchical level
number, and m is the number of the highest hierarchical level.
In view of the setting forth in Section 2, initial equations (4.1.3),
(4.1.4) in standard form (4.1.5) mean that some ‘hierarchicisation law’
(structural operator (2.2.6)) of the given dynamical system is de-
termined. Real form might be found, from experiments or other consid-
erations.
Within context of the definition (2.2.4), large hierarchical parameters
in (4.1.5), (4.1.6) can be determined as
where is component of vector and is q-th component of vec-

tor x. Here we consider the rate of variation of not exceeding rates of
variations of other slow variables and the component is the ‘slowest’
of other fast components of the vector Thus with general definition
of hierarchical scale parameter (2.2.4) we affirm that variables (in
the case ) and represent two neighboring different hierarchical
levels.
1.3 Classification of Oscillatory Phases

and Resonances. Hierarchical Tree
Accordingly with the general principles of the hierarchical approach
we should to reduce initial systems (4.1.3) or (4.1.4) to the standard form
(4.1.5). Therefore, the first step consists in determining all elements of
‘slow vector’ and ‘fast phase vector’ in (4.1.5). In other words, it
consists in classifying total set of variables of systems (4.1.3) and (4.1.4)
as slow and fast ones, respectively.
The Hamiltonian and variables and as a rule,
might be considered as slow variables. In contrast, classification of the
Lagrange phases of particle oscillations (see about the concepts of La-
grange and Euler variables in Chapter 1, Subsection 4.9) is more com-
plicated. In this case we should separately find the set of phases forming
component basis of vectors and Besides that, we must divide ro-
tating phases of particle oscillations into those associated with explicit
and hidden periods of oscillation (see definitions in Chapter 1, Subsec-
tion 2.2). The phases of the first group are related with phases of the
particle oscillations in the wave type of fields. For example, they are
particle oscillation phases in undulated fields of various types, in the
fields of electromagnetic waves or in the fields of proper and stimulated
beam waves, and some others (see Chapters 6, 7, 9). The periodicity of
these fields is responsible for the periodic character of electron motion
with respect to the same wave phases (4.1.2)
The second case is much more complicated. It can be associated with

phases of periodic particle motion in non-periodical quasi-stationary
intrinsic beam fields, external focusing and beam-forming fields, etc..
Procedure separating hidden periods of arbitrary functions is known in
mathematics. For instance the Lantzos’ method [8, 9], Grebennikov’s
method [10] and others. There are also semi-empirical approaches based
on qualitative peculiarities of a studied physical picture [6, 7].
We assume the procedure of finding of all particle oscillation phases
(hidden and explicit) has been completed. Analyzing the rates of par-
ticle phase varying we find that their total set involves both types of
phases: slow components of vector x and fast components of vector
Moreover, some nonlinear combinations of two, three or more Lagrange
fast phases produce slowly varying functions. Each slow phase resulting
from this ‘combination’ action ( in the general case) corresponds
to some physical mechanism of particle resonance. We distinguish quasi-
linear and parametric resonance (relevant definitions in Chapter 1, Sub-
section 2.3). Quasilinear resonances are characterized by two-phase (or
in the general case) combinations where one of phases only is
associated with explicit period of the system. The magnitude of stimu-
lated wave force acting on the particle is always linearly dependent on
wave field amplitude in the lowest order in some small parameter. There-
fore these resonances are quasilinear . Examples: cyclotron resonance;
various types of synchrotron resonances; etc..
Parametric one-particle resonances correspond to cases when all
phases forming slow phase combination are of wave nature. For instance,
parametric resonances of third-, fourth-, and higher orders [11–18] can
realize in free electron lasers (FELs), in parametric electronic devices
(Adler’s lamp, parametric electron-wave lamps), etc. [17, 18]. In general,
slow and fast combined phases of coupled-pair parametric or
quasilinear resonances can be defined as
where are sign functions, and are numbers of wave field

harmonics, and and are numbers of electron oscillation harmonics
in these fields (see definitions (1.2.19), (1.2.20) and relevant discussion).
Hence, the lowest order of relevant amplitudes of stimulated waves in
the case of parametric resonances is quadratic. Therefore, we classify
parametric resonances as nonlinear ones.
According to the hierarchical principles, which are set forth in Chap-
ter 2, we define components of scale parameter tensor according to re-
ciprocal rate of varying of fast to slow combined phases (4.1.9)
It means that slow and fast combined phases are variables of

different hierarchical levels. We assume the rates of change of velocities
are of the same order (in magnitude) as rates of velocities of other slow
variables (see (4.1.7)).
Then we analogously take into account the non-resonant phases of
oscillations. The only difference is that instead of slow and fast combined
phases we choose ordinary slow and fast phases. On the next stage
of calculational, we construct hierarchical series of components of both
tensors of scale parameters in the form (4.1.6).
Hierarchical series (4.1.6) are the essential point of the theory called
the theory of hierarchical oscillations. In this theory each term of se-
ries (2.1.3), (4.1.6) corresponds to a single oscillation or resonance or
a group of oscillations and resonances of equivalent hierarchy. In its
physical nature this series follows from the assumption accepted above
on influences between neighboring hierarchical levels only (see Fig. 2.1.1
and corresponding comments).
For discussed oscillatory hierarchical systems relevant hierarchical
trees (similar to those displayed in Fig. 2.2.4, Fig. 2.2.5) can be con-
structed. Analogously to the case of hierarchical military unit (see
Fig. 2.2.4) here we consider for simplicity that each two structural ele-
ments of lower hierarchical level form one structural element of the next
higher level. In given thought case it means that all resonances are two-
fold only and all phases of each hierarchical level take part in resonant
process. Therefore a phase resonant hierarchical tree can be constructed
(Fig. 4.1.1). As well as in the case of military unit (Fig. 2.2.4) we ob-
tain relevant analogy with the Cayley’s tree, which are known in the
theory of fractal systems, theory of codes, etc. [19, 20]. On the one
hand, this circumstance can be treated as confirmation of correctness
of developed hierarchical theory. On the other hand, such an analogy
is unexpected from physical point of view and even somewhat mystical.
Indeed, we have only one charged particle in some electromagnetic field,
and, therein obtain the structure coinciding with the characteristic ones
for explicit multi-particle objects (Universe, military units, other social
and biological systems, etc.)!
Then we discuss the hierarchical tree (4.1.1), from the point of view of
the hierarchical principles formulated above. We see that a few
phases’ (they are two phases only in this particular case) forms each
combined phase of the next ( )-level. This means that the higher
the hierarchical level the fewer are the number of oscillatative phases.
The latter can be considered as obvious illustration of the compression
hierarchical principle.
The hierarchical resemblance principle holds owing to specific math-

ematical construction of equations for each hierarchical level. As noted
above, all these equations are similar. Characteristic velocities of chang-
ing of variables of different levels are higher the lesser number of hier-
archical level is (hierarchical analogy with the second thermodynamic
principle). It is readily seen that any oscillations are not on the highest
hierarchical level: It means that on
the highest hierarchical level information entropy is equal zero. Hence,
hierarchical analogy with the third thermodynamic principle takes place
here, too, and so forth.
However, the following observation seems much more interesting.
Namely, total energy of the highest level for electronic devices with long
time interaction (which possess hierarchical structure like that shown
in Fig. 4.1.1) exists in the form of kinetic energy of the initial quasi-
stationary beam and the energy of the initial quasi-static electric and
magnetic fields only. It is important that in the chosen model relevant
sources of electron beam and of the initial quasi-static fields are be-
hind the system. It means that only external quasi-stationary sources
can drive the closed hierarchical system through the highest hierarchi-
cal level. At that the energy is transmitted in the closed hierarchical
system from higher levels to lower one and a reaction (information) of

hierarchical levels at this influence transmits in the opposite direction.
It is characteristic that these external sources occupy peculiar place

on the top of hierarchical pyramid. Namely, it is the separate point on
the top of specific vertical line that connects its with the rest (lower)
part of the hierarchical tree. Here we could be draw parallels with the
cases of hierarchical tree discussed above in previous Chapter. For exam-
ple, a similar picture holds in the case of the Universe as a hierarchical
structure (see Fig. 2.2.5). Indeed, comparing these two cases we see the
stunning and unexpected analogy in the hierarchical structures of Uni-
verse and electronic devices: in both cases, we detect the same separate
characteristic point at the top of the hierarchical tree. (In hierarchical
model of the Universe, this top point corresponds to some unknown en-
ergy source of the primary Big Bang). Here also energy flows in the
system through the specific vertical line connecting the highest hierar-
chical level (‘Big Bang point’) with the hierarchical level ‘metagalaxy’.
(The fact that in the discussed electrodynamic hierarchical system of
Fig. 4.1.1 we consider only pair interactions is not essential.) More-
over, we obtain the same, in principle, situation analyzing military units
also, as is seen from comparison of Figs. 2.2.4 and 4.1.1. In the case of
Fig. 2.2.4, the point on the top of hierarchical tree is the ‘source of power
of highest commanders’. Hence we can ask whether we can consider the
similar point of Fig. 2.2.5 as the ‘source of power of highest comman-
ders’. Unfortunately, today we cannot satisfactorily treat this strange
analogy. We only say that we deal with one of obvious manifestations of
the discussed in Chapter 2 self-modeling principle. To our regret, such
an explanation entails many more new questions than it gives answers.
The fact of analogy with the second thermodynamic principle can be
illustrated by the same way, too.
With the aid of Fig. 2.2.4 we show that in our model the generation
of hierarchical levels with negative numbers is possible, too. It
takes place in the realization of special case of self-consistent systems For
instance, it occurs if nonlinear generation of highest resonant harmonics
and wave modes is realized. Such generation process really breaks down
when imposing real restrictions. The latter can be determined by pe-
culiarities of system design, satisfaction of the conservation principles,
etc., or by realization of the effect of hierarchical degeneration.
1.4 Reducing Hierarchical Multi-Level

Standard System to the Two-Level Form.
The Scheme of Hierarchical Transformations
Thus because of relevant transformations initial system (4.1.3) or
(4.1.4) reduces to the hierarchical standard form (4.1.5). Later, we de-
scribe the algorithm of asymptotic integration of the system (4.1.5).
Within the definition of scale parameter (4.1.10), mathemati-
cal meaning of hierarchical series (4.1.6) is equivalent to the statement
that the system has the hierarchy over velocities of characteristic fast
oscillation (rotation) phases of particles, i.e.
with all evident definitions. Apart from the scale parameter of the first
hierarchical level (the leading term in hierarchical series (4.1.6)), all
the other fast phases are relatively slow quantities. This ob-
servation is useful to construct required hierarchical algorithm of asymp-
totic integration reducing complex total system (4.1.5) to essentially
simpler two-level hierarchical form of the type (3.6.10):
where is the expanded vector of slow variables (here all partial fast
phases excepting the first one are considered as its com-
ponents), and so on. We now write the Krylov–Bogolyubov substitution
for the system with rotating phases (the dynamical hierarchical opera-
tor (2.2.7)). But at that time we do it with respect to corresponding
components of the vectors and
where averaged variables and are the components of relevant

vectors obtained from truncated (shortened)
system of equations for next (first) hierarchical level:
The are functions

calculated below. As we will show later, equations (4.1.12) and (4.1.14)
are equivalent with respect to general mathematical structure. It means
that hierarchical resemblance principle holds. Analogously, the infor-
mation compression principle also is satisfied, because equation (4.1.14)
describes the simplest physical situation — non-averaged fast vector
phases absent in (4.1.14). By virtue of the same cause, hierarchical
analogy of the second thermodynamic principle here also holds.
On the next stage of hierarchical analysis, we use discussed hierarchi-
cal features of system (4.1.14). Namely, we separate out the following
group of fast scalar phases from ‘temporary components’ of slow vari-
ables of the vector These phases define the dynamics of the vector
of partial fast phases (see (4.1.11)). The rest components of vector
and components of averaged phase serve as a basis forming the
vector of slow variables of the next (second) hierarchy:
where is the part of the vector in which the vector

of fast phase is separated. In view of above reasoning the
meaning of new functions and the others is evident. An
algorithm of their calculational is described below.
Then we treat system (4.1.15) in terms of hierarchical algorithm. I.e.,
we separate out the variables from vector (this group of variables
describes the dynamics of vector phase ), and so on. The procedure
repeats cyclically until no terms remain in hierarchical series (4.1.6) and

(4.1.11), i.e., as long as we take into account all the partial vector-phases
that form complete vector of the total fast phase As an eventual
result we obtain m-fold averaged (truncated, shortened) equation set
(( ) is total number of hierarchical levels) that can be solved easily.
This property is the main advantage of discussed version of hierarchical
method making it very attractive for applications. Approximate non-
averaged solutions of initial equations (4.1.5) are obtained by successive
use of relevant inverse transformation formulae (4.1.13).
2. THE CASE OF SIMPLEST TWO-LEVEL

SYSTEM WITH ONE ROTATING
SCALAR PHASE
2.1 Formulation of the Problem
As shown above in Subsection 1.4, the procedure of asymptotic inte-
gration of the general multi-level system (4.1.5) in some special case can
be essentially simplified. Namely, it can be reduced to accomplishing a
hierarchical chain of integrating of simplified two-level systems (4.1.12),
(4.1.14), etc.. Hence the constructing of relevant algorithm of asymp-
totic integration of a two-level system should represent next step of the
discussed calculational technology. We begin discussion of algorithms of
such type with the simplest case, when the fast rotating phase
in (4.1.12) turns out to be a scalar function. Then we rewrite the equa-
tion (4.1.12) in the form through small parameter (see
equation (3.5.9)). As a result we obtain:
where and we have neglected the ‘prime’

symbol. It is obvious that for scalar phase the function Y is scalar,
too.
Without loss of generality we put the period T equal to More-
over, we assume that functions are differentiable with respect
to the phase as many times as is the number of terms retained in the
asymptotic series expansion.
Later we follow with calculational scheme discussed above in Chap-
ter 3, Section 5. It means that the problem is to find the substitution of
the form of the type (3.5.11), which would make it possible to separate
slow and fast variables. Taking into consideration the problem of secular
terms discussed above (see Chapter 3, Subsection 5.2 for more details) let
us write the Krylov–Bogolyubov substitutions as the asymptotic series
with the unknown functions and to be defined to be in what follows.

The equation for the first hierarchical level we also represent in the form
of the asymptotic series:
The coefficients and are unknown, to be found later along with

the functions and In order to specify the problem, we impose addi-
tion restrictions on the unknown functions and Namely, we assume
them to be bounded functions of for The similar assump-
tion we accepted above (with respect to time t only — see Chapter 3,
Subsection 5.2) for solution of the secular terms problem.
2.2 Algorithm of Asymptotic Integration

In a manner described in Chapter 3, Section 1, we differentiate the
representation (4.2.2) with respect to t and make use of the system
(4.2.1) and (4.2.3). Moreover, we note that
After some straightforward algebra the relevant generalized hierarchi-

cal equation of the first order of the type (3.1.2) reduces to the expansion
We equate the coefficients of the equal order terms with respect to

in the right and left hand parts of (4.2.5) to obtain a system of inter-
connected equations which determine the required functions, i.e.,
Here
(the functions D 1 and D2 are obtained by expanding in Taylor

series with respect to ); the components of the relevant vectors
and all derivatives are calculated for It may be easily
verified that in the general case the system (4.2.6) may be written as
We shall find the unknown functions and by means of successive

integration of the system (4.2.6). Each subsequent equation is solved
taking into account the solution of the previous one, but not vice versa.
A complicating point is that equations (4.2.6) and (4.2.7) contain the
unknown functions and This difficulty may be overcome in
view of the condition that both functions and are bounded for
So let us consider the first equation (4.2.6). We integrate it over

(bearing in mind that the derivative with respect to is partial) to find
the formal solution in quadratures to be given by
where is an unknown function (‘integration constant’). When

calculating the integrals of the form (4.2.8) here and henceforth, we
assume the variable to be constant. We see that the solution (4.2.8)
still contains an unknown function We note that the function
is periodic with respect to (by virtue of periodicity see
(4.2.6)). Let us calculate the average over the period of the
integrand
In order to find the value of the function we employ the bound-

edness condition for the function
We consider the limit
to find that for condition (4.2.10) is satisfied only provided that
Making use of (4.2.12) and (4.2.9) we obtain the required expression

for the function A1, i.e.,
This, in turn, makes it possible to find
The mathematical structure of the arbitrary function suggests

that the latter should be chosen in accordance with the initial (bound-
ary) conditions of the problem. For example, if the conditions are
then In principle the solutions can be normalized to the

function in many other ways. In particular, if the initial sys-
tem is Hamiltonian, then one can require that the system for the first
hierarchical level (4.2.3) should be Hamiltonian too [21].
Now, let us find other unknown functions contained in the transforma-
tions (4.2.2) and representations (4.2.3). In a manner similar to solving
(4.2.14), we obtain
and so forth. This result may be generalized. Thus we have

Substituting (4.2.19) and (4.2.20) (with regard for (4.2.18)) in (4.2.2)

yields the required transformations and, actually, the solution of the
problem. Of course, provided the solutions of the system of truncated
equations (4.2.3) are known.
Thus as we have already mentioned, the essence of the Bogolyubov
method is to integrate the simpler (truncated) equations for the first
hierarchical level (4.2.3) rather than complicated equations (4.2.1). In
other words, to find the solutions of the truncated equations is sufficient
for the asymptotic integration of the initial system (4.2.1) to be com-
pleted. The accuracy of the solutions is determined by the number n
of terms retained in the series (4.2.2) and (4.2.3). Let us consider this
aspect in more detail.
2.3 Accuracy of Approximate Solutions

As well known the accuracy problem is important for application of
hierarchical asymptotic methods. Let us illustrate its peculiarities with
discussed averaging method.
Suppose number n of terms retained in asymptotic representation sim-
ilar to (4.1.14) is fixed, i.e.,
Integrating (4.2.21) over t yields components of vector We sub-

stitute the latter for the phase in second equation (4.2.2) to obtain
Now we treat solutions as approximations of ‘true’ solutions

and Then solutions of initial equation (4.2.1) can be written as
A reasonable question arises of how high is the accuracy of the approx-

imations? We remind the reader that the uncertain degree of accuracy
of the approximate expressions (4.2.21), (4.2.22) has been main point of
the mathematicians’ objections against the Van der Pol method.
Thus let us estimate accuracy of approximations (4.2.23). Since
according to (4.2.21), is of the order of
then in view of the mathematical structure of equation (4.2.21) the es-

timate for can be written as
The asymptotic integration procedure implies that is a slowly

varying function of Therefore, its accuracy is given by
We recall that calculating the derivative of a slowly varying function

increases the order of magnitude of the result by Therefore integrating
a slowly varying function decreases its order of magnitude by as well.
Hence, we estimate the accuracy of terms contained in the integrand of
(4.2.22) as
Thus for the quadrature (4.2.22) we have

This means that if in approximation is found within the ac-

curacy of then is found within This conclusion is important
and must be in mind when estimating calculational accuracy. In the
lowest approximation (n = 1), truncated equations are
(the so called zeroth approximation — see examples in Chapters 6, 7,

9). In the general approximations of solutions and are
It is important to mention the correspondence between the time inter-

val for which the solutions thus obtained are valid, and the number
of approximation in the averaging method:
This means that if solutions are obtained in the approximation

then physical processes described by them can be adequately interpreted
only for interval (4.2.30). Ignoring this fact leads to mistakes (see, for
instance, Sections 4 and 5). If some physical effect is predicted theoreti-
cally, one has to verify whether criterion (4.2.30) is satisfied. Otherwise
the validity of scientific results is doubtful.
Then we turn to the problem of accuracy of Van der Pol solutions
(3.4.11), (3.4.12). When equations (4.2.29) (and, respectively, (4.2.28)
for ) are compared to the relevant results, (3.4.11)–(3.4.14), ob-
tained by the Van der Pol method, it may seems that equation (3.4.14) is
written with excessive accuracy with respect to (4.1.13). However, this
is a premature conclusion. The matter is that, when deriving (3.4.11)–
(3.4.14), we assumed i.e., the integral can be calculated
exactly, the error in the coefficient lowers to and so on. As
a result we have
For n = 1 (4.2.31) reduces to the Van der Pol results (3.4.11)–(3.4.14),

i.e.,
Thus the above mentioned contradiction really does not take a place.
2.4 Asymptotic Integration of Initial Equations

by Means of Successive Approximations
In the preceding items we discussed asymptotic integration algorithm
for initial set (4.2.1) based on the hypothesis that both functions X and
Y are n times differentiable in Now, following [22,23] we show this
assumption is not fundamental and employed for convenience only.
Again, we consider the set (4.2.1). We write the Krylov–Bogolyubov
substitution in the standard form
The relevant system of equations of the first hierarchical level (4.2.3)

is given by
Substituting (4.2.32) and (4.2.33) in (4.2.1) yields

We analyze (4.2.34) by means of successive approximations. The stan-

dard (for this method) iteration procedure implies
where is iteration number,
We compare systems (4.2.35) and (4.2.7) and find them to be of similar

mathematical structure. Repeating above sequence of operations results
in solutions similar to (4.2.19), (4.2.20).
2.5 Peculiarities of Asymptotic Hierarchical

Calculational Schemes Based on
the Fourier Method
We have shown that asymptotic integration of initial system (4.2.1)
reduces to solving partial differential equations (4.2.7). The latter can
be written in generalized form
We recall that is a periodic function of with the period

It is peculiar of the integration procedure applied to the equations
of this type that the function is unknown and found from the
condition that the solution is bounded for
The Fourier version of averaging method is the most convenient one.
Therefore just it is taken as a basic version further in this book.
We substitute a Fourier series for i.e.,
where is the Fourier amplitude defined in conventional way. Let

us consider the zero harmonic of the Fourier expansion (4.2.38). Its
amplitude is equal to the mean value of the function for the
period T, i.e.,
Then equation (4.2.37) can be written as
Let us expand the function F in a Fourier series. Within the context

of (4.2.40), we have
where is the ‘integration constant’,
As follows from this result, Indeed, function F is bounded

for only provided the secular term contained in (4.2.41)
vanishes that is possible for only. Thus we obtain the known
result from the preceding analysis, i.e.,
As shown below in this book, the procedure described is efficient when

applied to electrodynamics.
3. CASE OF TWO FAST ROTATING SCALAR

PHASES
In Section 1 of this Chapter we have formulated the general calcula-
tional hierarchical problem. As was shown, this problem mathematically
is the development of an algorithm of asymptotic integration of hierar-
chical standard systems of the type (4.1.5). It should be mentioned that,
in the general case, arbitrary number of components of the total vector
of fast rotating phases is characteristic feature of these systems. Above
(in Section 2) we discussed the simplest variant of such hierarchical sys-
tem, where the vector of fast rotating is represented by one scalar phase
( ) only.
Then we emphasize that systems with number of rotating phases
possess some essentially new peculiar features (in comparison to the dis-
cussed case ). In the first place this concerns the possibility of
realization of resonances in such multi-phases (multi-frequency) objects.
Apart from that some specific features also characterize the non-resonant
multi-phases systems. Unfortunately the general scheme of asymptotic
integration of multi-phases hierarchical systems similar to (4.1.5) (which
is given below in Section 4) seems too cumbersome for a beginner. That
is why we begin studying of the multi-phases (multi-frequency) hierar-
chical systems with discussion of simplest system where the phenomenon
of resonance can be realized. As analysis has shown, this is the system
with two rotating scalar phases and one scalar large parameter.
3.1 Formulation of Problem

Thus we consider a system with two fast scalar phases )
and one scalar large parameter given by the following
standard form:
where and X are vectors, are scalars. Contrary to the

situation studied in the previous Section, now X and are two-period
functions with respect to phases with periods
In electrodynamic problems coefficients as a rule, are usually equal

to 1. Hence we can put for simplicity. The generalization to
the arbitrary case is straightforward and does not involve any
difficulties (see below the general algorithm described in Section 4).
We write the system of truncated equations (system of the first hier-
archical level, according to our terminology) in the form analogous to
(4.2.3), i.e.,
According to general procedure of the averaging method, we find the

change of variables reducing the system (4.3.1) to the form (4.3.3). For
this we use the substitution in the standard form given by
If solutions of the equation of the first hierarchical level (truncated

equations) (4.3.3) are known, then substitutions (4.3.4) play the role of
formal solutions of the initial system (4.3.1). Therefore, the main task
in this case is to find unknown transformation functions and
as well as the auxiliary functions and
3.2 Solutions. Non-Resonant Case

We carry out a procedure similar to that set forth in Section 2 and thus
find the first-approximation ( ) equations for the transformation
functions to be given by
It is not difficult to show that in the subsequent approximations,

equations of similar mathematical structure can be obtained. The
generalized mathematical construction of the the Krylov–Bogolyubov
system of equations can be written similarly to (4.2.37), i.e.,
As contrast to (4.2.37), however, in the case discussed the fast phase is

represented by the two-component vector ( ), and, correspondingly,
we have the two-periodic (with respect to the periods )
functions F and V. Let us employ the Fourier procedure described
above in Subsection 2.4. We expound the functions and X in double
Fourier series, i.e.
Then we substitute (4.3.7) and (4.3.8) in the first equation (4.3.5) and
assume that the coefficients of similar exponential functions are equal.
Thus we obtain
The condition that the function is bounded for is satisfied

if the secular terms in (4.3.7) vanish, i.e.,
In view of (4.3.11), relation (4.3.10) yields
where
The other expansion terms of the first approximation can be calcu-

lated in the same manner. Thus we obtain
where are the coefficients of the Fourier expansions of the functions

the arbitrary functions should be found from
the initial (boundary) conditions, is the twice-averaged
(over the phases and ) function
We note that the choice of the functions does not influence

the accuracy of the solutions obtained.
Let us analyze the mathematical structure of the obtained solutions

(4.3.14). We see that they contain the specific resonance denominators
of the form
We remind the reader that the problem in the discussed case of two
scalar phases has been formulated under the assumption that both phase
rotation velocities are commensurable, i.e.,
This looks clearer if the following circumstance is taken into consider-

ation. By virtue of the characteristic structure of equation (4.3.1) in the
discussed case the difference in the scales of rates of change of the slow,
on the one hand, and all fast variables, on the other hand, is determined
by the same large parameter of the problem
where is the q-th component of the vector Then we remind the

reader that the harmonic number in (4.3.15) is an algebraic quantity, i.e.,
Besides that we remind that the angular velocities
of the fast phases rotation (frequencies ) are algebraic quantities,
too. At the same time in principle some denominator in (4.3.15) in some
points can be equal to zero, i.e.,
This is the well known problem of small denominators [10]. Equation

(4.3.18) determines the resonance condition (physical essence of this con-
cept is described in Chapter 1, Subsection 2.3 in more detail). It is clear
that if this condition is satisfied then the solution (4.3.14) is divergent.
This infers that the above calculational procedure is inapplicable in the
resonance case (4.3.18). In order to describe the resonance states of the
system, the above constructed asymptotic integration algorithm should
be some modified. Let us discuss one of methods of such modification.
3.3 Solutions. Resonant Case

Let us find the asymptotic solutions, which would adequately describe
the dynamics of the system (4.3.1) in the neighborhood of the resonance
point (4.3.18), including this point itself. In order to describe the reso-
nance state, we introduce the it mismatch to be given by
Solving equation (4.3.19) for yields the resonant point

Then we expand the functions in the vicinity of the point The
analysis of the expansion shows: Thus in the resonance case
the large parameter (4.3.17) may be also interpreted as the measure of
the approach to the resonance state of the system in the near-resonance
region (see also Fig. 1.2.2 and corresponded comments).
The resonances of two types are distinguished - the main resonance
and the resonances associated with harmonics. The main resonance
occurs for The harmonic resonances occur for or
or when both inequalities are satisfied simultaneously:
1.
The case of constant frequencies. Let us begin our studying with

the case of main resonance for the constant frequencies. It means that
the supposition that the frequencies are constants is satisfied.
If the main resonance occurs in such model, we may write
Within the context of (4.3.20), the initial system (4.3.1) may be

rewritten as
We introduce a new variable
which is referred to as the combination phase (see also (1.2.22), (4.1.9)

and relevant comments). We differentiate (4.3.22) with respect to time
t and make use of (4.3.21). Thus we obtain
Inasmuch as phase rates of change have been assumed to be close

(see (4.3.20)), their difference can be small. Both
the new variable and the new function entering (4.3.23) are slowly
varying quantities under the resonance conditions. Then we can rewrite
the system (4.3.21) taking into account (4.3.22) and (4.3.23). As a result
we obtain
It is readily seen from the latter system that, as a result, of the ac-
complished transformations:
a) the variable can be regarded as a component of the new slow
variables
b) the functions and are periodic with respect to one fast
phase only (rather than two fast phases and of the system
(4.3.1)). In other words, the mathematical properties of the system
(4.3.24) are similar to those of the system (4.2.1), in which one should
pass to the form initial equations with large parameter (instead of the
form with small one ). Hence we can employ the relevant asymptotic
integration algorithm, which is described above in Section 2.
The case of slowly varying frequencies. Further let us turn

our attention to the more general case of slowly varying frequencies
At first we consider the main resonance for
bearing in mind that in the general case the functions
are algebraic quantities. Taking this into account, the main resonance
condition similar to (4.3.20) may be written in the case discussed in a
form like (4.3.20):
where is the sign function (because sign { } =

+1). Formally, there seems to be no difference between the conditions
(4.3.25) (for ) and (4.3.20). However, a rather important distinc-
tion exists. It is associated with the very possibility that variations of
the frequencies can be slow under nonlinear resonance interactions
in the system. This means that evolution of the system can be such
that the system can slowly get in and out of the resonance state (see
also the Fig. 1.2.2 and relevant comments). For each interval of values
of we have to check whether the solutions obtained really contain the
resonance denominators (4.3.15). The results of the check up determine
whether the non-resonance or resonance asymptotic integration proce-
dure should be employed. In the general case, when solutions of both
types occur (associated with different stages of interaction), we have to
join the solutions together in terms of some special procedure [7] (see
below in Subsection 3.4).
Then we turn to discussion of the algorithm of asymptotic integration
in the discussed resonance case.
At the first stage the procedure of separating out the slowly varying
combination phase could be modified in order generalize it. The general
‘multi-phases’ procedure of this type is described in monograph [23]. In
our case of two scalar phases only it reduces to the following simplified
form. We introduce the trivial non-degenerate linear transformation,
which relates the fast phase vector to the vector of combination
phases to be given by
Since the combination phase has been shown to be slow and, hence,
is associated with a component of the vector of slow variables
In turn, the combination phase by virtue of the definition (4.3.26),
must be considered as a fast one. The relations between the old and
new variables in such situation are given by the following simplest linear
transformations:
Then we substitute (4.3.27) into (4.3.1), make use of (4.3.26), and

carry out some calculations. As a result we obtain
At the next stage of the calculational procedure we introduce the

obvious notation
As a result of the transformations performed we avoid the problems

concerned with the small denominations: the system (4.3.28) reduces
to the standard form with a single fast rotating phase similar to (4.2.1)
(where we should accomplish the change and connected with
this relevant transformations), i.e.,
But the algorithm for solving such systems is described above in Sec-
tion 2. This means that the above formulated resonant problem is solved.
Further we turn our attention to the case of harmonic resonance.
Analogously to (4.3.25) the relevant harmonic resonance condition (i.e.,
in the case or or simultaneously) may be
written as
We define the slow and fast combination phases by the analogous

method:
where, as before, . The relation between the

old and new variables in now given by
We carry out substitutions analogous to those considered in the pre-

vious case, and again obtain a system of the form (4.3.30).
It is to be noticed that in practice the definition of harmonic concepts
sometimes causes misunderstanding. In electrodynamic problems one
has to distinguish the harmonics of wave fields in which the electron

moves, from the harmonics of its nonlinear oscillations in these fields (see
above the materials of Subsection 1.3). In the asymptotic integration
procedure, harmonics of external wave fields are taken into account in
the definitions of fast phases, i.e.,
where are the numbers of external wave field harmonics.

The mentioned above quantities and are harmonics of electron oscil-
lations in these fields. The relevant combination phases (4.3.32) should
be then written in the form (4.1.9). The integration procedure for the
initial resonance system (4.3.1) remains in principle similar.
4. THE CASE OF MANY ROTATING SCALAR

PHASES
4.1 Formulation of the Problem
Generalize the algorithms described above in the case of arbitrary
number of the rotating phases (see system (4.1.12) and other relevant
materials of Subsection 1.4). Introducing the vector of fast phases in
the form we rewrite initial equation (4.1.12) in
the following manner
where all notations are self-evident in the above. We look for a solution
to system (4.4.1) in the form of dynamical hierarchical operators (4.1.13):
where: are elements of vectors and and are numbers

of components in the latter. In accepted notations one can rewrite the
equations of the first hierarchical level (4.1.14) as
4.2 Algorithm of Asymptotic Integration

Strict determination of unknown functions in (4.4.2), (4.4.3) is am-
biguous. This is owed to arbitrariness in attributing different terms of
the series. Following the above calculational schemes discussed we can
eliminate this arbitrariness assuming all and are free of null
(in Fourier harmonics. Thus we postulate the whole averaged mo-
tion is described by and
Let us resume the differentiation of (4.4.2) and, taking into account
(4.4.3), substitute the obtained result into (4.4.1), equating the coeffi-
cients of equal powers Therefore we obtain the infinite sequence of
relations
Hereafter, the averaging sign is omitted for simplicity. Taking into

account (4.4.4), it is easy to see that functions are periodic in
Using this, we expand them into Fourier series of multiplicity m:
Here is the period associated with phase

are the relevant Fourier harmonics. Proceeding from similar
considerations as well as (3.1.18), we represent unknown function
as
Upon substituting (4.4.8), (4.4.8) into (4.4.4) and equating the co-
efficients with equal exponents, we obtain expressions for amplitudes
and functions
where functions in limits on satisfy normalization condition
Accordingly, definitions for coefficients (4.4.11) can be rewritten

as
Taking into account the condition of absence of null harmonics on

and for the latter we finally formulate the definition different
from (4.4.9) only by exclusion of terms in Similarly
to (4.4.11), the second approximation equation is solved, etc.. From the
structure (4.4.10), in particular, it follows that
i.e., there are no resonances amongst components of the vector of fast

phases. (Failure of (4.4.14) means that at the stage of classification of
phases (and their linear combinations) not all slow phases have been
singled out). Following the procedure, we obtain expressions for other
unknown functions, in particular:
where means the averaging in all fast phases (similar to (4.4.13));

are factors of expansions in Fourier series of the functions of
this form.
5. ALGORITHM FOR SEWING TOGETHER

RESONANT AND NON-RESONANT
SOLUTIONS
5.1 Essence of the Problem
Thus depending on satisfaction of a condition like to (4.4.14) we could
have two of principle different calculational situations in the considered
oscillatative systems. The first is the non-resonant case. It takes place
if condition (4.4.14) is satisfied. The calculational algorithms described
above in Sections 2 and 4 are dedicated for the studying of systems of
this type namely. The other situation corresponds to the resonant case.
In such a case the relevant resonant condition can be constructed also
(see, for example, Subsection 3.3).
Up to now we have discussed the models, which are characterized by

resonant or non-resonant states, but no both they in the same model.
However, the situations, when both (i.e., the non-resonant as well as the
resonant) states could be realized in the same system are also rather typ-
ical for practice. Most often this takes place in the nonlinear dynamical
systems, which pass consequently through non-resonant and resonant
states during their evolution on time or space (see Fig. 4.5.1 and illus-
tration example shown in Fig. 1.2.2). It is readily seen that we can
separate three characteristic zones for the complete solution in the
discussed case (see Fig. 4.5.1).
The first is the zone which corresponds to non-resonant part of the so-
lution. Characteristic feature of the latter is that the considered physical
process here carries an explicitly non-resonant nature. According to the
general theory, which was set forth above in Subsection 2.3, Chapter 1,
such non-resonant states can be determined by some conditions such as
the following
where and are corresponded fast and slow combinative phases (see
definition (4.1.10) and relevant comments). As shown in preceding Sec-
tion 4, condition (4.5.1) can be assumed to be equivalent to condition
(4.1.14):
where h* is the so called resonant mismatch (see also (4.3.20), (4.3.25),

(4.3.31), and relevant explanations).
The resonant part of solution (the resonant zone) is characterized by
satisfying of some resonant condition. In the framework of developed
hierarchical theory such condition could be formulated in one of the
forms like to (1.2.13), (1.2.18), (1.2.21) or (4.1.7). For instance:
or
correspondingly.
In view of the above, it is obvious that in the general case the non-
resonant and resonant solutions have essentially different mathematical
structures, because they are obtained in different mathematical ways.
Thus the problem of sewing them should be solved in order to obtain
the complete solution
5.2 Sewing Together of Resonant and

Non-Resonant Solutions
A specific feature of the sewing procedure in the case of hierarchi-
cal asymptotic methods is the presence a peculiar ‘gray’ zone between
the non-resonant and resonant zones (see Fig. 4.5.1). Strictly speak-
ing, neither non-resonant nor resonant solutions are
invalid within this zone. But, on the other hand, it is obvious that the
non-resonant solutions should transform into the resonant ones within
the ‘gray’ zone. In principle, more than one such ‘gray’ zones can ex-
ist in a general multi-resonance case. We may assume that some point
where the non-resonant solutions transform into resonant one im-
mediately, should exist for each such ‘gray’ zone . It is obvious,
that we should have one such point only in the simplest one-resonant
particular case illustrated in Fig. 4.5.1.
In what follows, let us turn to the problem of formulating the sewing
condition. It is assumed that a vector of slow variables
and some vector of fast rotating
phases characterize the considered system. Ac-
cordingly with the above said, we should satisfy the following
sewing conditions for realization of the sewing procedure in any ‘gray’
zone:
Here the subscript corresponds to the non-resonant parts of the

complete solution, and the subscript describes relevant variables of
the resonant solutions. As before is the time when transformation of
non-resonant solutions into resonant ones within a zone occurs (see
Fig. 4.5.1). Taking into consideration the properties discussed above of
the resonant and non-resonant condition, the following ‘gray’ condition
can be formulated (here is

some constant). Therewith, the condition
is satisfied for the time interval and the condition
is valid for the time interval

The system’s evolution in the inverse direction (i.e., from the resonant
state to the non-resonant one) can be described in an analogous manner.
Then let as remember that some unknown functions and (see

solutions (4.3.14), for instance) were obtained for the above described
procedures for constructing resonant and non-resonant solutions. It is
important to draw the reader’s attention that in the discussed ‘sewing’
case the relevant scheme of determining explicit view of these functions
is some different. (Because the combinative phase is a component
of the vector of slow variables in the resonant case, and, at the same
time, analogous fast phase plays a role of component of the vector of
fast rotating phases in the non-resonant case — see Sections 2–4 in
detail). Hence this circumstance should be taken into account in ‘sewing’
conditions (4.5.5). As a result of utilization of the calculational schemes
described in Sections 2–4 and the above noted peculiarities, conditions
(4.5.5) can be transformed into corresponding system of equations with
respect to the functions and
Thus we obtain two sets of the vector arbitrary functions and
as a result of the accomplished calculations (one set we have for the
resonant solutions, and the other set related with the non-resonant ones).
One of them could be obtained by the method described in Sections 2–4.
The second set is determined by ‘sewing’ condition of (4.5.5). This gives
the method of solving the considered ‘sewing’ problem.
It might be seem that the above described calculational algorithm
look too abstract for immediate practical realization. Let us illustrate it
further in more detail by the simple example of the ‘stimulated Duffing
equation’ (see below equation (5.4.3) and corresponding commentaries,
and [10]).
5.3 Example for the Solution ‘Sewing’:

The ‘Stimulated’ Duffing Equation
The Duffing equation for stimulated oscillations, as is well known (see,
for instance, [10]) can be written in the form:
where and are some constants, as earlier, is the

large problem parameter. Let as transform equation (4.5.9), using the
Van der Pol variables and (see Chapter 3, Subsection 4.1):
As a result the following equations for and (equations of the

zeroth hierarchical level) could be obtained:
Accomplishing the obvious exchanging of variables

it is not difficult to rewrite equations (4.5.11) in the following more
symmetric form:
It is readily be convinced that equations (4.5.12) could be regard as a

standard system with scalar slow variable and two fast rotating phases
and The analogous (but more general) system had been stud-
ied in Section 3 (see equations (4.3.1)). Hence all calculational schemes
described there for considering system (4.5.12) can be used here. Later
we will make use this circumstance in our analysis. At that according
to that set forth in Section 3, we assume that non-resonant as well as
resonant states could be realized in the same (considered) physical sys-
tem that is described by equations (4.5.12). This means that behavior
of the system at some time-intervals might be described by the resonant
solutions, and, therein, the non-resonant solutions characterize the sys-
tem evolution at other time-intervals. So let us begin our studying with
discussion of the non-resonant states (or state) of the system.
Non-resonant part of the complete solution. It is considered that

the non-resonant condition like to (4.5.2) is realized (
— see (4.3.2) and corresponding comments)
where are any integers for which

Accordingly with the algorithm described in Section 3 the Bogolyubov
substitutions (4.3.4)
should be accomplished at the next stage of our calculational procedure.

Here the relevant averaging values are determined by the equations like
to (4.3.3):
Let us limit ourselves by calculating the functions

in the first approximation only. As a result we can find
for the function
The function we obtain from the equation

After calculations it is not difficult to obtain also (see (4.3.14)):
In what follows, we can accept for simplicity: Analo-

gously, relevant calculations for the functions and yield:
The equation
could be found for the determining of the function Solving it we

obtain:
We also accept, as well as in the case (4.5.18):

Accomplishing the analogous calculations it is readily be obtained:
Resonant part of the solution. Let us consider the case of main

resonance (see (4.3.25)). Then, the resonant condition
(4.5.4) can be rewritten in the form ( ):
Further, using the procedure of reducing the resonant problem to the

non-resonant one, which is described above in Subsection 3.3, we should
pass to the slow combinative phase (4.3.22)
and to accomplish the exchanging
As a result, equations (4.5.12) could be transformed into the following

form:
It is obviously that the new variable is a slowly varying function and

the system (4.5.27) carries the non-resonant nature. Therefore, for the
finding of asymptotic solutions of system (4.5.27) we can apply the ‘non-
resonant’ algorithm described in Subsection 3.2. So Krylov–Bogolyubov
substitutions (4.3.4)
can be consider as a first step of this stage of our calculational proce-

dure. The second stage, analogously to the non-resonant case, is the
constructing the equations of the firs hierarchical level (i.e., equations
for averaging values)
After corresponding calculations we obtain the following expressions

for the function (the first Bogolyubov approximation
( )):
The Krylov–Bogolyubov equation
can be found for the function Solving (4.5.32) we have:

Analogously, we obtain for the functions and v(1):
Thus we have obtained the solutions for resonant and non-resonant

parts of the complete solution. The next stage of the considered calcu-
lational procedure is ‘sewing’ of these solutions within the ‘gray’ zone
(see Fig. 4.5.1).
The sewing the resonant and non-resonant parts of the solu-

tion. Let as suppose that the frequency is a slowly varying on time
value (i.e., This means that the considered physical system
passes through non-resonant and resonant states during its temporary
evolution. Therefore, we should to ‘sew’ the resonant and non-resonant
parts of solution for obtaining some discontinuous complete only solu-
tion. The key point of corresponding ‘sewing’ algorithm is the ‘sewing’
condition (4.5.5). In the case considered condition (4.5.5) can be rewrit-
ten in the form:
where, in view of definitions used in (4.5.5) all designations are self-

evident.
However, let us turn the reader attention to one important circum-
stance. The resonant solutions (4.5.30)–(4.5.37) are written by the us-
ing the concept of slow and fast combinative
phases. On the one hand, we did not need to utilize analogous combina-
tive phases for description of the non-resonant part of complete solution.
Instead this we had used the fast phases immediately. We have
some contradiction of designations for the resonant and non-resonant
parts of solution. For the sake of levelling the designations for descrip-
tion of both parts of solutions further we introduce the concept of fast
combinative phase (apart from the fast phase ) for the
non-resonant part (!) of solution. This gives a possibility to accomplish
the transformations like to ( ) ( ) in non-resonant solutions
(4.5.16)–(4.5.23). The ‘sewing’ conditions (4.5.38) reduce to the form:
Substituting the above obtained solutions for the non-resonant and

resonant parts into (4.5.39) we eventually obtain the system of equa-
tions with respect to the unknown functions and
Let us determine these functions in the first Bogolyubov
approximation. Utilizing (4.5.39), Krylov–Bogolyubov substitutions
(4.5.14), and solutions obtained for non-resonant and resonant parts,
it is not difficult to construct the system for determining the unknown
functions and
It should be pointed out that here we confine ourselves by accounting

only three first conditions (4.5.39). This is explained by that accounting
of other three conditions (4.5.39) (which contain derivatives) do not
exert remarkable influence at the complete solution, obtained in the
first Bogolyubov approximation.
Further we use the observation (see Fig. 4.5.1) that we can accept for
the ‘sewing time’
REFERENCES 155
(here we assume that the frequencies and as well as other values,

are non-dimension ones). As a result of accomplished transformations
we can write relevant expressions for the looked for functions
and
In what follows, we substitute obtained expressions (4.5.44) into res-

onant solutions (4.5.33), (4.5.35), and (4.5.37). In turn let us substi-
tute the got by such manner expressions into Krylov–Bogolyubov sub-
stitutions (4.5.14). This yields the resonant part of solutions which are
‘sewed’ with non-resonant ones in the point The conditions
(5.4.3) could be used for determining of this point.
Thus the above stated problem of ‘sewing’ resonant and non-resonant
solutions is solved.
References
[2] R.C. Davidson. Theory of nonlinear plasmas. Benjamin, Reading, Mass, 1974.
[3] A.G. Sitenko, V.M. Malnev. Principles of the plasma theory. Naukova Dumka,
Kiev, 1994.
[4] L.P. Landau, E.M. Liftshitz. Theory of field. Nauka, Moscow, 1974.
[5] V.V. Kulish. Nonlinear self-consistent theory of free electron lasers, method of
227, 1993. Sov. Plasma Physics.
[8] M.G. Serebriannikov, A.A. Pervozvansky. Discovery of hidden periodicities.

[9] L. Lancos. Practical methods of applied analysis. Fizmatgiz, Moscow, 1961.

[10] E.A. Grebennikov. Introduction to the theory of the reasonable systems. Izda-
telstvo MGU, Moscow, 1987.
[11] A.P. Sukhorukov. Nonlinear wave interactions in optics and radiophysics.


[14] L.A. Vainstein, V.A. Solnzev. Lectures on Microwave electronics. Sov. Radio,
Moscow, 1973.
[15] V.I. Gaiduk, K.I. Palatov, D.M. Petrov. Principles of microwave physical elec-
tronics. Sov. Radio, Moscow, 1971.
[16] A.F. Alexandrov, L.S. Bogdankevich, A. A. Ruhadze. Principles of plasma elec-
trodynamics. Vyschja Shkola, Moscow, 1978.
[18] B.E. Zshelezovskii. Electron-beam parametrical microwave amplifiers. Nauka,

Moscow, 1971.
[19] A.I. Olemskoi, A.Ya. Flat. Application of the factual concept in the condensed-
matter physics. Physics-Uspekhy, 163(12):1–104, 1993.

particles for electronic applications. The International Journal of Infrared and
Millimeter Waves, 19(1):33–93, 1998.

1981.
Chapter 5
WITH FAST ROTATING PHASES.
EXAMPLES OF PRACTICAL
APPLICATIONS
Previously in this book we have set forth various versions of the

asymptotic hierarchical algorithms and relevant calculational schemes.
But experience shows that most of these calculational schemes are not
simple for an average reader to understanding. Moreover, as the author
is convinced, most professional physicists have heard about averaging
methods, but only some of them know how to use them at practice. The
main cause of such a situation is some psychological unusualness of the
structure, the basic idea as well as the special ‘language’ of the methods.
This could explain the main cause of the fact that hierarchical (including
averaging) methods now are not too widespread in physics, in general,
and in electrodynamics in particular. On the other hand, the methods
discussed demonstrate an obvious highly practical effectiveness. Per-
sonal author pedagogical experience shows that the best teaching effect
can be achieved in such situation only when all above discussed method-
ical ideas were illustrated by means of many calculational details and
evident applied examples.
Observation formulated above predetermined the main idea and the
structure of the book, as a whole. Indeed, the most part of its contents
comprises titles of examples, which demonstrate various practical cal-
culational peculiarities of the methods. In particular, Volume II, as a
whole, serves this purpose. Besides that, some illustration examples are
given in this (first) Volume. The difference is only that the examples
described in Volume II are mainly intended for professional physicists,
postgraduate, and master level students, who have essential preliminary
training in physics and mathematics. In contrast to this, the illustration
examples given in Volume I are intended for readers whose training is
not so advanced. Therefore characteristic feature of the examples given
157
below is the relative simplicity of the considered models and presence of

a lot of illustrative calculational details. The model ‘charged particles in
the field of standing electromagnetic waves’, which is studied below in
this Chapter, is chosen as a basis for such simplified illustration example.
1. GENERAL PROPERTIES OF THE MODEL:

‘A CHARGED PARTICLE IN THE FIELD
OF A STANDING ELECTROMAGNETIC
WAVE’ AND EXAMPLES
OF ITS PRACTICAL REALIZATION
1.1 Systems for Transformation of
Optical Signals into Microwave Signals
as a Convenient Illustrative Examples
The systems which can be described in framework of the model ‘a
charged particle in the field of a standing electromagnetic wave’ are
widespread in electronics, radiophysics, and plasma electrodynamics.
For instance, one might find them amongst various systems of laser
plasma diagnostic and control, systems for transformation, generation,
and amplification of electromagnetic waves [1–8], etc.. For instance, as
it will be shown further in Chapter 10, Volume II, the models of free
electron lasers (FELs) in some cases can be reduced to such model [4–8].
So let us discuss the system for transformation of optical signals into
microwave signals [5, 9–11], as a convenient illustration example. As the
analysis shows, the chosen model can represent this system in satisfac-
tory way. The devices of this type can be used as receivers of electromag-
netic signals of sub-millimeter-infrared (IR) ranges, as systems for dy-
namical parameter control of especially intensive laser beams in vacuum,
etc.. It is interesting to note that the solid state analogs of this device
are also known [11–13]. They are IR receivers of electromagnetic signals
constructed on the basis of some types of piezo-semiconductors. More-
over, they could be studied by using the discussed hierarchical methods.
So, the calculational technology set forth below can, in principle, be
applied for analysis of some solid state systems, too.
Let us restrict ourselves by discussion of only two versions of this
system design. These versions are presented in Figs. 5.1.1 and 5.1.2.
We begin our discussion with the design-version represented in Fig. 5.1.1.
The system shown in Fig. 5.1.1, which intended for transformation of
submillimeter-IR range signals into microwave signals, operates in the
following way. Transforming electromagnetic signal 1 with frequency
enters the signal input 2. Heterodyne signal 10 with frequency
enters the heterodyne input 9. Both inputs are placed in such man-
ner that the transforming and heterodyne signals have the possibility
to propagate within the same system working bulk in mutually oppo-
site directions along the system optical axis. Electron gun 3 generates
non-relativistic electron beam 5, which also moves along this axis. So
electron beam 5 moves in the working bulk of the device in the field of
two opposite directed transverse running electromagnetic waves Namely,
they are transforming 1 and heterodyne 10 waves with frequencies
and respectively. Owing to nonlinear properties of plasmas of beam
5 longitudinal Langmuir electron wave with the differential frequency
(difference frequency electron wave) is excited. This wave
also propagates in the direction of electron beam motion. The frequen-
cies and (that belong to submillimeter-IR range) are chosen close
to each other, so that the differential frequency lies within
the microwave range (i.e.,
The process of excitation of an electron beam wave, according to prin-

ciples of the physical electronics [1–4], could be interpreted as the effect
of electron beam modulation. The beam modulation in the discussed case
occurs under action of the field of two oppositely propagated transverse
electromagnetic waves. The Langmuir electron waves, to the physical
point of view, are analogous to the usual sound waves in a gas, because
the beam plasmas could be regarded as a peculiar flux of a charged gas.
Thus some (first) part of the device really works as a modulation sec-
tion. Besides that, the device contains some other design elemets, such
as retarding system 7. The difference frequency electron wave which
is excited within the modulation, transforms into the electromagnetic

microwave signal with the same frequency in retarding system 7.
This signal we regard as an output transformed electromagnetic signal
for the discussed device.
As the analysis shows (see, for example, [1–4]), effectiveness of any
modulation process always depends strongly on certain correlation be-
tween different characteristic dynamical parameters of a system. In our
case they are magnitudes of the frequencies on the one hand, and
the velocity of the beam as a whole, on the other hand. Qualitatively
this peculiarity can be illustrated in the following way. Propagating in
the opposite directions running electromagnetic waves and form a
drifting standing electromagnetic wave. It is well known that amplitude
maximums and minimums of an ‘ordinary’ (i.e., no drifting) standing
waves always stand on some fixed spatial points. In contrast, the analo-
gous points of the drifting standing wave drift along the system optical
axis with the phase velocity
(see for more details about the concept of drifting standing wave in Chap-
ter 10, Volume II). Here c is, as before, the velocity of light in vacuum
(the general concept of the phase velocity is discussed also in Chapter 1,
Section 3). It can easily be seen that in our case the phase velocity
always, because, as mentioned above, Moreover, for
(i.e., for the case of non-relativistic drifting ve-
locities can occur. In the particle case the velocity
and hence the drifting standing wave transforms into the ‘usual’
standing wave.
Then, we remind that in the considered device the electron beam 5
also moves, as a whole, along the optical axis. At that the direction
of the drifting wave in the case considered is the same as the direction
of electron beam 5. Such direction of the drifting wave is attained by
the mentioned special selection of the frequencies In
particular, the case is possible when the so called strong synchronism
appears. This takes place when the drifting wave velocity on the
one hand, and the electron beam velocity on the other hand, are
found to be equal
As the analysis show [5, 11], the highest effectiveness of the electron
beam modulation (i.e., excitation of the difference frequency electron
wave could be achieved in this case namely. After analyzing the

condition (5.1.2) one can be convinced that this is the condition for real-
ization of the so called parametrical resonance. (In detail the concept of
parametrical resonance is discussed earlier in Chapter 1, Subsection 2.3
and Chapter 4, Subsection 1.3). Such work regime is chosen as a basic
for the modulation section of the device shown in Fig. 5.1.1.
It should be mentioned that the theoretical model discussed of the
device modulation section formally coincides, in principle, with the sim-
plest model of free electron laser (FEL) with Dopplertron (including,
laser) pumping [14, 15] (see also Chapters 10–13 Volume II). The hetero-
dyne electromagnetic wave can be interpreted as the peculiar pumping
wave, which is purposed in FELs for achieving additional amplification
for the signal (transforming) wave However, this amplification in the
case considered is really inessential because the typical set of parameter
for the considered transformer (the electron beam here is non-relativistic
and weak-current). We shall neglect the influence of this amplification
effect.
In what follows we turn the reader’s attention to the fact that an-
other design element, whose work effectiveness also depend’s essentially
on the beam velocity presents in the device. This is the retarding
system 7 which, as mentioned above, transforms the electron wave
into the transformed output electromagnetic signal with the same fre-
quency In fact, this design part can be treated as a specific section
of the Travelling Wave Tube (TWT) (about this type of electron devices
see, for instance, in [1,2]). Hence a corresponding synchronism condi-
tion can also be formulated for this transformer section that includes
retarding system 7. It can be written as an approximate equality of
the phase velocity of the transformed electromagnetic wave and the
beam velocity
Then, we note that very often requirements for the beam modulation
process (5.1.2) and for the maximum of transformation effectiveness of
the retarding system 7 (5.1.3) cannot be satisfied simultaneously. Two
special pairs of electrodes 4 and 6, correspondingly, are introduced in
the device shown in Fig. 5.1.1 for correction of this drawback. Earlier
we talked about the modulation section but never concretizes its de-
sign. These electrodes, and part of the electron beam between them,
form the modulation section discussed above. Owing to the electrodes
we have a possibility to change beam velocity in the working bulk
of the modulation section and, consequently, to provide satisfaction of
optimal conditions (5.1.2). Or, in other words, we can provide in this
way optimal conditions for excitation of the difference frequency electron

wave in modulation section. At, satisfying the optimal condition for
transformation of this electron wave into the electromagnetic microwave
signal within the retarded system 7 (5.1.3) is provided by virtue of the
last pair of electrodes 6. Usually the case of essentially differing syn-
chronous conditions (5.1.2) and (5.1.3) is realized in the design versions
intended for work in the submillimeter-IR range. It should be mentioned
especially that the presence of running electromagnetic waves 1 and 10
(see Fig. 5.1.1) within the working bulk of retarding system 7 in such
situations does not have any significant influence at effectiveness of the
transformation process.
So the characteristic feature of the discussed design-version (see

Fig. 5.1.1) is the possibility of spatial separating of the processes of beam
modulation and electron wave transformation, respectively. Hence both
these processes can be studied separately owing to this feature. Later
we shall widely use the latter peculiarity.
The distinctive, in principle, situation takes place for the second type
of the discussed device, whose design scheme is shown in Fig. 5.1.2. Its
characteristic feature is the possibility of simultaneous satisfaction of
the conditions (5.1.2) and (5.1.3) in the same device working bulk. This
means that, from an engineering point of view we can omit here the
electrode systems like 4 and 6 in Fig. 5.1.1. As a result, the design as a
whole simplifies.
Thus the difference between the devices shown in Fig. 5.1.2 and
Fig. 5.1.1 is only the following. The systems of electrodes for control-
ling of the beam velocity are not foreseen in the device in Fig. 5.1.2,
because of primordial coinciding both synchronous conditions (5.1.2)
and (5.1.3), respectively. This means automatically that the modula-

tion and transformation processes in this case occur simultaneously and
this occurs in the same working bulk of the device. Or, more exactly,
both processes happen within the working bulk of the retarded system
5. However, the found design peculiarity changes essentially the general
physical situation in the device. Namely, in contrast to the previous
design version (see Fig. 5.1.1) we do not have in this case any possibil-
ity for spatial separation of both discussed processes (modulation and
transformation, respectively). Both these processes in the second case
represented in Fig. 5.1.2 cannot be studied separately.
Generalizing the above we can state that in spite of the seeming de-
sign resemblance of both represented (in Fig. 5.1.1 and Fig. 5.1.2) devices
they are characterized by essentially different operation physics. This
means automatically that their theoretical models are also different. In
the first case we have to do with the connection in series of two separate
models-sections. The modulation and transformation processes, respec-
tively, are characteristic for them. Therein the theoretical model: ‘an
electron beam in the field of two oppositely propagating electromagnetic
waves’ corresponds to the modulation (i.e., first) section. Whereas the
model ‘electron beam in the field of three electromagnetic waves’ are char-
acteristic for the second case. All this will be taken into consideration
during analysis of physical processes in the discussed systems.
As follows from what was said above, all working physical mechanisms
in the discussed devices really possess obviously multi-particle nature.
At the same time, the analysis shown that used usually electron beams
could be classified as a weakly intensive. This implies that reciprocal
particle interactions (Coulomb’s and magnetic, in the general case) can
be neglected in the framework of such system theory. Hence, complete
physical picture here is formed as a result of simple supposition of many
individual no interacted between themselves particles, which, in turn,
interact with the external heterodyne and signal electromagnetic wave
fields. The concept of single-particle interaction electrons with the elec-
tromagnetic fields is accepted further as a basic one.
Thus we begin our illustration process with the simplest single par-
ticle problems. Or, in more detail, we will illustrate the hierarchical
asymptotic calculational technology to the single particle theory of the
devices discussed above. Therefore as already mentioned above, we have
a possibility of construct two different, in principle, theoretical models
in the discussed situation. The first one describes physical processes,
which take place in the modulation section of the system, like that is
shown in Fig. 5.1.1. (Let us recall once more that we classify it as the
model with electron oscillations in the field of two oppositely directed
running electromagnetic waves with frequencies The second one

is the model with electron oscillations of in the field of three running
electromagnetic waves (with frequencies and respectively).
1.2 Formulation of the Problem of Electron

Motion
in the Field of Two Oppositely Propagating
Electromagnetic Waves
Let us consider in more detail the dynamics of electron motion in
the single-particle model of the first type. According to reasons dis-
cussed above we assume that the synchronous condition (5.1.2) is sat-
isfied. Then we perform the transition to the co-ordinate system
which moves along with the phase velocity of the drifting stand-
ing wave This means that the drifting standing wave in the case
of system transforms into an ‘ordinary’ standing wave, whose max-
imums and minimums of oscillations amplitude are spatially fixed. In
turn, the fact of satisfying condition (5.1.2) means that the velocity of
the electron beam, as a whole, is equal to zero in the co-ordinate system
Hence concerning the initial averaged velocity of a single
chosen electron in the system we can assume that it is also equal to
zero.
We describe the electron motion in the field of two oppositely running
electromagnetic waves in terms of the Hamilton equation (see above
Subsection 1.4.7)
where and are the Hamiltonian and

canonical momentum of an electron in the electromagnetic field of stand-
ing electromagnetic wave [16], is the laboratory time, is
the position vector (radius-vector) of the electron. Here and below ev-
erywhere we omit the primes that indicate that the quantities under
consideration are written in the system
As has been already mentioned above, any standing wave can be rep-
resented in the form of two running electromagnetic waves propagating
in mutually opposite directions. In our case we suppose for simplicity
that these running waves are plane and linearly polarized:
where is the vector-potential of the standing wave; are the ampli-

tudes of electric components of the running waves; are
their wave phases; are the sign functions which
determine the directions of the running wave propagation along
(in our case of two opposite running waves: are
the wave vectors, is the unit vector along the are the cyclic
frequencies of the running wave, are the wave numbers (analo-
gously with the situation with frequencies, we consider
are the initial phases oscillations of the running waves; is the unit vec-
tor along the are the cyclic frequencies of the running wave.
Here we should mention that in the framework of the chosen coordi-
nate system the cyclic frequencies of both running electromagnetic
waves are found to be equal: whereas they are not equal
in the laboratory coordinate system K. The physical sense
of this mathematical result will be discussed below in Volume II. Here
we only note that Doppler effect can explain this phenomenon. The
transforming signal wave (which propagates along the positive direction
of the — see below in Fig. 5.1.3) reduces its frequency at the
transition into the coordinate system At the same time, oppositely
directed heterodyne wave raises its frequency. As a result in the system
both frequencies turn out to be equal.
The graphical illustration of the considered model-scheme is given in
Fig. 5.1.3. Relevant definitions for all used concepts were given before
in Sections 1.2 and 1.3.
Analyzing the initial system (5.1.4) from the point of view of defini-
tion (5.1.5), one can surely be convinced that equation (5.1.4) possesses
an integral of motion (conservation law) with respect to transverse com-
ponent of the canonical momentum:
Indeed, inasmuch as any explicit dependencies of the Hamiltonian

a determined by the explicit dependency
really the Hamiltonian explicitly depends on the spatial coordinate

and laboratory time only
Due to this the partial derivatives and turn out to be

equal to zero. Then, taking into consideration this and the second of
equations (5.1.4) we can write the equation for the transverse component
of the canonical momentum:
where is the transverse radius-vector, is the unit vector

along the axis The other values were determined above.
It is obvious that for known initial transverse momentum
relevant exact solution for the transverse coordinates can be found
easily:
However, concerning the longitudinal component of motion, the cal-

culational situation is found to be essentially more complex. The point
is that there is no any possibility to obtain any exact solutions and
in this case. One of ways for obtaining of such asymptotic (i.e.,
approximate) solutions will be shown below.
2. MODEL ‘AN ELECTRON IN THE FIELD

A TWO-LEVEL HIERARCHICAL
OSCILLATATIVE SYSTEM
2.1 Reducing Initial Motion Equations to
the Standard Forms with Two Rotation
Phases
Now we shall try to apply the hierarchical method for solution of
the longitudinal part of the discussed electron motion problem. We
start with classification of the variables and reducing the initial set of
equations to some standard hierarchical form.
Taking into account results (5.1.9)–(5.1.11), the initial system (5.1.4),
can be rewritten in the following manner:
Analyzing system (5.2.1) together with definition (5.1.5), one can be

convinced that here two fast rotating phases of the electron oscillations
could be separated. Formally (in view of periodicity of the right hand
parts of equations (5.2.1)) these phases coincide with the oscillation
phases of the running electromagnetic plane waves:
Next, let us turn to the problem of identification of slow variables.

The variables and should be considered as the slow ones from
physical reasons. For instance, we take into consideration that accord-
ing to the well known definition the velocity of changing of the electron
energy in time is the power of acting forces (see Section 1.4). It is obvi-
ous that this power in a real physical situation must be always limited.
Analogously, the velocity of changing of the canonical momentum is the
force that acts on the electron and it also should be limited, and so on.
Bearing in the mind the above mentioned reasoning, let us perform
some transformations in system (5.2.1). First, we transform definition
for field of the standing wave (5.1.5) in the complex form, taking into
account the definitions (5.2.2):
where is the imaginary unit, the values
are the complex amplitudes (see corresponding definitions above in Sec-

tion 1.2). Let us assume that the phases in (5.2.3) are the electron
oscillation phases. Then, substituting (5.2.3), (5.2.4) in the definition
for Hamiltonian (see comments for equations (5.1.4)), we reduce the first
two equations of system (5.2.1) to the form:
In what follows, differentiating equations (5.2.2) by and after some

relatively simple transformation we obtain
Here the well known relationships of relativistic dynamics [17]:
are used. Here are the components

of the vector of electron velocity We have also used the property
that the Hamiltonian is the particle energy expressed through canonical
variables (see relevant definitions in Section 1.4 and standard textbooks
on the General Physics (see, fore instance, [16]).
Thus analyzing the mathematic structure of system (5.2.5)–(5.2.7) we
conclude that the initial system (5.2.1) can be reduced to the standard
form with two fast rotating phases like (4.3.1)
where the vector is constructed from the slow variables:

the vector-function X is constructed from left hand sides of equations
(5.2.5). As it is easily seen (compare the last two of equations (5.2.9)
and (5.2.6), (5.2.7), correspondingly) that we have two possible versions
of determining of functions and in the considered specific
case. We can accept for the first version the concept of slowly varying
nonlinear frequency
In the case of the second version we can write
Let us chose version (5.2.10) as the more convenient for our purposes.
But it is interesting to point out that, as the analysis shows, the choice
of that or other above discussed versions does not exert any influence at
the terminal calculational result.
It is necessary turn the reader attention also at some another specific
calculational feature of discussed type problems. Namely, here, accord-
ing to the definition given above of problem large parameter we should
determine it as a ratio of fast to slow variables. We have two different
variants for such definition:
respectively. Here the functions and are deter-

mined by the right hand parts of equations (5.2.5)–(5.2.7). The smaller
of parameters (5.2.11) is chosen as the problem large parameter
Thus it is clearly seen that the problem large parameter in our
particular case turns out to be a slowly varying function on time. That
is why in calculational practice the large parameters in an explicit form
are, as a rule, not separated out. Usually their peculiar ‘hidden’ form is
used, when they are present in relevant equations in an implicit form.

The relationships like ones in (5.2.11) are used for estimation of order of
relevant terms in equation right hand sides. These relationships are also
used for numerical current control of accuracy of the used calculational
asymptotic procedures at their different stages.
It should be remind that the algorithm of asymptotic integration of
systems similar to (5.2.9) has been described before in Chapter 4, Sec-
tion 3. So, let us find relevant solutions using this algorithm, and perform
their short physical analysis.
2.2 Zeroth Hierarchical Level.

Parametrical Resonance
Analyzing the discussed system we can be convinced that the resonant
states could be realized here (see condition (5.1.2) and corresponding
comments). The latter conclusion is one of the sequences of synchronous
condition (5.1.2), which had been imposed above. As has already been
mentioned, this condition coincides with relevant result, which can be
got from the parametrical resonance condition (see corresponding discus-
sion in Subsections 1.2.3 and 4.1.3, respectively). Following the general
algorithm of asymptotic integration (see above Section 4.3)), we should
introduce in such situation the slow and fast combinative phases (4.3.26)
It is obvious that the phase is a slow variable (because the electron

coordinate is a slowly varying on time function), whereas the phase
is fast one. After performing differentiation of equation (5.2.12) by
time, and taking into account the relationship
(which, in turn, follows from expressions (5.2.8)) it is not difficult to

obtain the corresponding differential equation for the slow combinative
phase
Analogously the differential equation for the fast combinative phase

could be got in the similar way:
As a result of the accomplished transformation we reduce the initial

standard system (5.2.5)–(5.2.7) to the standard system with one fast
rotating phase:
Indeed, having constructed new extended vector of slow variables

and new vector-function we can formulate equation system
(5.2.17) in the ‘classical’ standard form with one fast rotating phase
(4.3.30)
where
It is interesting to point out that the system obtained (5.2.17) can

be classified as a system with constant rotation frequency (the difference
between the systems with changing and constant frequencies is discussed
in detail above in Chapter 4, Section 3).
2.3 Passage to First Hierarchical Level.

Nonlinear Pendulum
According to the basic idea of the hierarchical method we should
pass from zero to the first hierarchical levels at the next stage of our
calculational procedure. For this the following the algorithm described

above in Chapter 4, Section 3, we use a transformations similar to (3.3.4)
where we should construct relevant truncated equation of the type (see

(3.3.3)) for determining the averaged values
Thus the main problem on this step is determining the unknown trans-
formation functions and the functions
in right hand parts of truncated equations (5.2.21). Let us solve this
problem following with the algorithm discussed above. Therein we con-
fine ourselves by calculations in the first approximation only of the av-
eraging method Therefore, we will keep only the terms in right
hand parts of equations not higher than cubic order, with respect to am-
plitudes of the running electromagnetic waves The latter supposition
can be substantiated by the following observation. The values
describe the amplitude of oscillations of the normalized (on the light

velocity c) transverse electron velocity in the field of the running
electromagnetic wave (here is the normalized electron energy
— relativistic factor)). In the case when the non-relativistic case of
electron motion takes place the following evident criterion is
satisfied
In relativistic electrodynamics the values are referred to as ac-

celeration parameters. In principle, the definition (5.2.22) we can also
regard as a convenient formula for normalization of the amplitudes
Hence, speculating about the ‘keeping in right hand parts of equations

the terms only not higher than cubic order, with respect to amplitudes
of the running electromagnetic waves we actually have in mind the
accelerator orders of accelerative parameters Therefore, these param-
eters are used as relevant expansion parameters in many electrodynamic
problems.
Thus following with the above mentioned algorithm, we can for-
mally write the solutions of the problem (5.2.17) in the form of Krylov–
Bogolyubov substitutions
where the averaged values can be found from the truncated equations
The characteristic expression for the function can be in the fol-

lowing form:
where
Analyzing equation (5.2.25) and taking into account definition (5.2.31)

one could easily be convinced that the averaged Hamiltonian (i.e., the
Hamiltonian of the first hierarchical level) is conserved in time
i.e. the expression (5.2.32) we can regard as the motion integral of

the first hierarchical level. This can be explained by the fact that the
function according to (5.2.31), does not depend on time explicitly.
Therein averaged combination phase depends linearly on the averaged
longitudinal coordinate
which easily can be obtained by simple averaging of definition (5.2.12).

Than, we differentiate equation (4.2.29) by time taking into con-
sideration motion integral (5.2.32), expression (5.2.33), and equation
(5.2.26). As a result we have:
Introducing the definition
we can write down result (5.2.34) in the more elegant form:
Let us discuss brief the results we obtained. Comparing (5.2.36) and

(1.2.5), it is not difficult to see that equation (4.2.35) formally reproduces
the analogous equation for the nonlinear pendulum (see Chapter 1, Sub-
section 2.1 for more details). But let us recall that the equation (1.2.5)
has been obtained for the case of a material point motion in some poten-
tial well. The homogeneous gravitational field determines the forming
of this potential well in that case (see Fig. 1.2.1 and relevant commen-
taries). We also recall that, as is well known, the gravitational fields
related to so called potential fields. Besides that, we can regard this
field as a proper field of the zeroth hierarchical level of the system pre-
sented in Fig. 1.2.1.
Following with this analogy we can introduce formally for the model
considered the concept of quasi-potential field. In contrast to the case
with ‘true nonlinear pendulum’ (see Fig. 1.2.1) this field belong to the
first hierarchical level. Action of this field provides nonlinear electron os-
cillations on the first hierarchical level, which is similar with oscillations
of the nonlinear pendulum on the zeroth level in potential gravitational
field (see Fig. 1.2.1).
It should be mentioned that the concept of proper field of a hier-
archical level looks, at list, unusual and somewhat whimsical. Therefore,
let us discuss it in more detail. We revert for this to the discussion about
hierarchical nature of our Universe (see Chapter 2, Sections 1 and 2; and
also Tables 2.1.1, 2.1.2, Fig. 2.2.5, and corresponding discussion). Be-
sides that, we will compare some hierarchical properties of the Universe
and the considered model ‘an electron in the field of electromagnetic
standing wave’, as hierarchical dynamical systems. It is not difficult to
see that general situation in both systems is similar, which in itself, looks
incredibly in principle. Analogously to the situation in the considered
model, some proper set of fields also exists on each Universe hierarchi-
cal level. But, in contrast to the considered model, all proper fields on
all Universe hierarchical levels carry explicitly expressed fundamental
nature. For example, weak and strong interactions are proper for the hi-
erarchical level ‘elementary particles’, electromagnetic fields are proper
for the atoms and molecules, gravitational field is proper for the planet
and star systems, and (it is possible) for the galactic systems, and so on.
But let us once more turn to the topic of fictitious (effective) and
fundamental fields in the physics. The quasi-potential field is a ficti-
tious (effective) one. It is obvious that it cannot be considered as the
fundamental one. We regard the quasi-potential field as one of many
displaying of a ‘hidden’ acting the electromagnetic wave fields (that be-
long, as noted, to the zeroth hierarchical level) on the first hierarchical
level.
However, we should recognize that, talking about the fundamental
nature of Universe fields, we have in the mind the fact that they are
really some postulates only. Indeed, all these fields appear as a result of
many experimental observations and they can not be obtained strictly
in framework of some general theoretical approach. (We avoid here
discussion of basic problems of the unified field theory). Or, in other
words, by this we would like say the following: it is not known surely to-
day how the proper (fundamental) fields of higher hierarchical levels are
connected with proper (also fundamental) fields of the lower hierarchi-
cal levels. The specific peculiarity of the hierarchical structure given by
the considered model ‘electron in the field of standing electromagnetic
field’ is the following. We achieved here the revealing of some specific

physical mechanism of forming the proper field of first hierarchical level
(quasi-potential field) by the fields of the zeroth level (electromagnetic
wave fields). I.e., we reveal, in fact, the specific mechanism of correla-
tion between the fields of the zeroth (fundamental fields) and the first
(non-fundamental ones) hierarchical levels.
Basing on the above formulated observation, let us propose the follow-
ing hypothesis. We suppose that all fundamental fields in the Universe
are formed analogously with the above-described mechanism of forming
quasi-potential field in our particular model with electron in the standing
wave. This means that all proper (fundamental) fields of higher hierar-
chical levels in the Universe appear as a result of displaying of acting of
proper (also fundamental) fields of the lower hierarchical levels. It could
be assumed that the mechanism of this displaying could be described by
some ‘generalized averaging procedure’, similar to that used above for
obtaining the quasi-potential field. We expect also that the two main
problems will arise in this case. The first is connected with the finding
of proper sets of dynamical variables for each hierarchical level. The sec-
ond is the constructing of relevant dynamical operators, which regulate
the rules of transacting between the hierarchical levels. The hierarchical
principles discussed earlier in Chapter 2 could serve as a basis for solving
of these problems.
Beyond doubt, the topic about the developing of a ‘new hierarchical
unified field theory’ is very interesting. But let us again revert to the
‘sinful earth’, i.e., to our model ‘an electron in the field of standing
electromagnetic wave’.

The Miller–Gaponov Potential
Let us temporarily interrupt the discussion of considered calculational
problem. Instead this we will turn to some interesting physical pecu-
liarities of our model. Namely, we will discuss the concept of effective
potentials (Miller–Gaponov potentials).
A.M. Miller and his student A.V. Gaponov (future Soviet academi-
cian) [18, 19] were the first who turned attention to the discussed pecu-
liarity of motion of charged particles in the field of two opposite directed
electromagnetic waves. They have proposed the concept of effective po-
tential field for description of the quasi-potential field. Corresponding
definition for the potential of such field can easily be obtained if we re-
call the well known correlation between a force acting on the charged
particle and the potential function (see expression (1.4.17) and
relevant commentaries)
In what follows we turn to equation (5.2.36) and the second law of

dynamics (see (1.4.4))
Taking into account (5.2.26), (5.2.37), and (5.2.38) we can introduce

the concept of the potential function (potential energy) of electron in
the quasi-potential field Accordingly with (5.2.37), the effective
force can be expressed via the potential function in the following way:
Integration by in (5.2.39) yields
where is some arbitrary function. As relevant analysis show, it

could be accepted for standard boundary and initial conditions for the
potential function
Taking into consideration expression (5.2.31), the potential function

in the considered case can be written eventually in the following
form:
i.e., here we obtained for the constant the expression
It should be mentioned that sometimes in references the potential

function is referred as Miller–Gaponov potential. To be exact, this
is not quite correct, because according to classical definition, the poten-
tial of an electromagnetic field should be defined in the following manner:
(in our case But, really it is not important because
the noted problem have purely philological nature. Therefore, here and
on we will understand definition (5.2.41) as the Miller–Gaponov poten-
tial. We note that in tradition references the Miller–Gaponov potential
wells are called the buckets.
2.5 Nonlinear Pendulum. The First Motion

Integral
Let us continue the discussion of the calculational peculiarities of the
considered model.
The problem of nonlinear pendulum is classical and it is represented
rather widely in references. We confine ourselves further by only the dis-
cussion of some characteristic methodical steps of the solution procedure
and short analysis of the obtained solutions.
We begin with the so called first integral of motion that can be ob-
tained from the equation of nonlinear pendulum (5.2.36). Let us multiply
for this equation (5.2.36) by the value It is not difficult get after
obvious simple transformations:
or, after integration, the sought first motion integral can be found:
Let us clarify the physical sense of this integral and the constancy of
H in our case. Substituting definition (5.2.13) into (5.2.44) and multi-
plying this equation by the relativistic electron mass and
the coefficient we transform expression (5.2.44) to the form of
the energy conservation law:
Thus as can be obviously seen, the first motion integral (5.2.44) by it-
self, from physical point of view, has the sense of the energy conservation
law for the first hierarchical level (5.2.32). To obtain equation (5.2.45)
we also used the relationships (5.2.13), (5.2.41), and the definition of the
averaged longitudinal electron velocity

Exact Solutions and Analysis
Thus as shown above, equation (5.2.44) can be regarded as a nor-
malized form of conservation law (5.2.45). Hence it is convenient to
introduce formally the concept of normalized ‘energy’
and the normalized ‘potential function’ [19]
As it can be easily seen, function represents the sequence

of maxima and minima of normalized potential energy of the electron
(see Fig. 5.2.1a). We also might say that this is a periodic sequence of
the potential wells. Such type wells are referred to as the buckets in
the applied electrodynamics. Following this tradition, we will use such
terminology further in this book.
Then let us find and analyze the solutions of equations (5.2.36). The
initial conditions are
We introduce the energy parameter by means of the expression [19]
The using of the energy parameter is very convenient for analysis

of the physical processes in the model ‘an oscillating electron in the
bucket’. Dynamics of these processes is illustrated in obvious way in
Fig. 5.2.1. The dependence of the normalized potential function V on
the oscillation phase is shown in Fig. 5.2.1a), and the phase
portrait of the electron motion is pictured in Fig. 5.2.1b).
The three typical cases of electron motion are illustrated there. E.g.,
the case corresponds to the oscillations of electrons localized in the
bucket (i.e., trapped), is associated with the electron drift (transit)
motion, and gives rise to the motion along the separatrix (see
Fig. 2.2.1b). The solutions of the equation of nonlinear pendulum similar
to (5.2.36) are well known (see, for example, [19]). In our designations
they may be written as
Making use of the Fourier expansions for the elliptic functions cn[...]
and dn[...], we rewrite solutions (5.2.50) in the form
which determines the dependence of after being integrated over Here,

the following designation has been introduced:
where
are full elliptic integrals. The nonlinear electron oscillation frequency is

then given by [19]
We can see that the oscillation frequency of an electron in the

bucket depends on its normalized energy H, i.e., the electron in this
case is a non-isochronous oscillator. We shall show in what follows that
this peculiarity can lead to some nontrivial consequences.
Thus the problem of electron motion in the field of two oppositely
directed electromagnetic waves for the first hierarchical level is reduced
to the well known problem of nonlinear pendulum. Therefore, all con-
clusions and solutions thereof can be employed to analyze the processes
of interest in the model under consideration.
Let us continue the discussion of the characteristic features of electron
motion on the first hierarchical level. For the sake of convenience we
introduce the number
Apart from that we employ the asymptotic of the elliptic integral

that is described by the expressions
where is the energy of the electron that moves along the sepa-
ratrix. We make use of (5.2.55) for to find that
Hence electron oscillations are harmonic near the bucket bottom which
corresponds to the case of small oscillations. As normalized energy H
increases, the nonlinear nature of the energy dependence of the oscil-
lation frequency (H) is revealed to greater extent (see (5.2.54)). As
a result the motion is appreciably modified. Near the separatrix, we
have and the oscillation frequency tends to zero:
The plot of velocity versus time, resembles
a sequence of solitons (see Fig. 5.2.2).
The distance between the neighboring solitons (bumps) is approxi-
mately equal to and each soliton width is close to This
means that the number N, defined by equation (5.2.54), is a measure of
relative pulse duration of the sequence of pulses–solitons. The analysis of
the Fourier amplitudes of electron oscillation harmonics (making use of
the solutions of (5.2.51)) reveals rich oscillations spectrum in the vicinity
of separatrix. The amplitudes of many first harmonics here being almost

equal (see Fig. 5.2.3) [19]. It should be noted that in this case the num-
ber N describes the number of characteristic oscillation harmonics. The
two above mentioned marginal cases and make it pos-
sible to analyze the electron oscillation dynamics in the bucket for the
whole energy range from H = 0 to In the case of drift motion
(for the picture looks symmetric, too. Electron oscillations
near the separatrix have many harmonics as well. When moving from
the separatrix the oscillation type changes towards the harmonic law.
Thus summarizing the discussed above results, we can conclude that

a proper, rather peculiar, and somewhat whimsical, physical picture
characterizes the first hierarchical level of the considered system. Gen-
eralizing this observation we can claim that each hierarchical level of
any dynamical hierarchical system always possesses a proper specific
physics. Other obvious illustrations of this conclusion we can obtain
analyzing such hierarchical object as, for example, our Universe (see
materials Chapter 2, Section 1 for more details). We should stress that
general physical picture of zero (initial) hierarchical level is formed as a
superposition of the physical pictures of all hierarchical levels. Let us

illustrate this phenomenon further using the illustration example of the
considered model.
2.7 Full Solutions of the Initial System

Taking into consideration the Bogolyubov substitutions (4.3.34) and
solutions (5.2.50), and after corresponding calculations, we can write the
full solutions of the considered problem for the zeroth (initial) hierarchi-
cal level:
where solutions for the averaged values are obtained in the form
It is easily seen that the mathematical arrangement of solutions

(5.2.57)–(5.2.66) obviously describes the slow electron oscillations (that
correspond to the first hierarchical level) on background of the fast oscil-
lations (the zeroth hierarchical level). This means that the total oscilla-
tion process itself is found to be rather complex. In spite of the use of the
hierarchical approach giving a real possibility for separating the oscilla-
tions of different hierarchical level. This allows to study these oscillations
separately, that essentially simplifies the general physical analysis.
Following the general idea of the hierarchical method (see Fig. 1.1.2),
the above used calculational can be shown as it was done in Fig. 5.2.4.
At the first sight, according to the scheme in Fig. 5.2.4 the model
studied should be classified as a two-level hierarchical oscillatative sys-
tem (zeroth and first levels, respectively) only. However, that is not
really correct because the system comprises one more (the highest) hi-
erarchical level, too. This level was not used in the described above
calculational scheme, because the equations of the first hierarchical level
allow exact solutions. In the hypothetical case we did not find such solu-
tions. Relevant asymptotic solutions should be constructed in this case
with accounting the second hierarchical level (see below Subsection 5.3).
This means that we would separate additionally new fast oscillatative
phase in such calculational scheme and construct corresponding equa-
tions of the following (second) hierarchical level and so on.
As has already been mentioned above, this second hierarchical level is

used nowhere in the discussed calculational procedure before. According
to the given explanations this had happened because in the special case
considered we have a possibility of obtaining exact solutions of the first
hierarchical level problem. This means that we need to use in this par-
ticular case the second hierarchical level for obtaining the full solutions
of the initial problem. But in fact the situations of such type (when the
problem of some lower hierarchical level has an exact solution) are not
typical in practice. Moreover, they are rather rare. That is why further
in the following Subsection we will give another version of calculational
scheme for the considered problem. At that time we will conditionally
assume that exact solutions for the first hierarchical level are not known.
We will look in this case for approximate solutions which can be obtained
by using of the hierarchical method.

A THREE-LEVEL HIERARCHICAL
3.1 Transition to the Second Hierarchical Level
Thus later we shall study the model with two oppositely directed
electromagnetic waves (i.e., standing electromagnetic wave) which we
regard as a three-level hierarchical oscillatative system. Let us solve
once more the considered problem taking into consideration its three-
level hierarchical nature.
The general calculational schemes of both hierarchical models (two-
and three-level, respectively), in principle, coincide. However, exact
coincidence takes place until we reach the stage of obtaining averaged
equations (5.2.25)–(5.2.30) only (see Fig. 5.2.5). Then the difference
concerns the method of solution of these equations. We have used in

the first case the method of exact solutions. In the second case (consid-
ered below) we shall look for approximate solutions of these equations
(5.2.25)–(5.2.30). We will apply for this the hierarchical calculational
scheme again for their asymptotic integration. Therefore, we can omit
the first step of the discussed calculational procedure in this case and
begin the further study with equation of nonlinear pendulum (5.2.36)
immediately.
Taking into account expansions in series (1.2.7), we introduce the
auxiliary function
where is the small parameter of the problem (really, as

corresponding analysis shows, the small parameter where
is the bigger of the accelerative parameters (5.2.23)). Separating the
function sin in definition (5.3.1), and substituting it into (5.2.36), we
can reduce this equation to the form
Here in (5.3.1) and (5.3.2) we performed the transition to the so called

extended combinative phase neglecting further the sign of
prime.
Then we must carry out the four following steps for accomplishing the
hierarchical procedure of asymptotic integration:
1) to separate oscillatative phases of the system;
2) to reduce the equation (5.3.2) to the standard form with rotating
phases similar to (5.2.18);
3) to construct the approximate solutions for the next (second, in this
case) hierarchical level;
4) to transform the obtained solutions into the lower (first) hierarchi-
cal level.
We take into consideration that corresponding formulae for transfor-
mation of the first level solutions to the zeroth (initial) hierarchical level
are given by relationships (5.2.57)–(5.2.61).
Concerning the procedure of separation of rotating phases on the first
hierarchical level, we should note the following. First, here we have only
one rotating phase. Second, this phase, in contrast with the situation
on the zeroth hierarchical level, turns out to be hidden (we had explicit
phases (5.2.2) in the preceding hierarchical level).
The Van der Pol variables (see in Chapter 3, Subsection 4.1) are ap-
plied for solution of the first problem (i.e., for the separating oscillatative
phases). As it was shown in Chapter 3, Subsection 4.1, they originate

from the solutions of a generating (linear) equation similar to (3.4.2):
i.e., we consider that due to weak nonlinear nature of oscillations we can

consider that period of nonlinear oscillations is close to a period of linear
system like (3.4.2).
We consider the amplitude to be slowly (super-slowly) changing
phase. Then, following the scheme described in Chapter 3, Subsec-
tion 4.1, we can transform equation (5.3.2) into the form that is similar
to (3.4.8), (3.4.9):
Equations (5.3.2) and (5.3.4), from mathematical point of view, are

equivalent. However, it can be easily seen that, in contrast to (5.3.2),
system (5.3.4) is written in the hierarchical standard form like in (4.2.1).
Consequently, further we can use the standard procedure for its asymp-
totic integration that is described earlier in Chapter 4, Section 2. Ac-
cording to this procedure we should be looking for formal solutions of
the problem (Bogolyubov substitutions) in the form like in (4.2.2):
where all notations are standard.

It is easy to see in (5.3.5) that the variable in the considered case
plays the role of some ‘super-slow’ oscillation phase for the second hi-
erarchical level. This means that the studied system indeed consists of
(at least) three hierarchical levels (the zeroth, first and second levels,
correspondingly — see Fig. 5.2.5). But the question arises: is this third
level higher for our system? Positive answer for this question could be
obtained we will prove that (hierarchical analogue of the
third thermodynamic principle — see Chapter 2, Subsections 2.1 and
4.1.3). Below (in the following Subsection) we will show that this really
takes place for our model.
The truncated equations (equations for the second hierarchical level)
can be constructed for determining of the averaged variables in substitu-

tions (5.3.5). The integration on the right hand side of equations (5.3.6)
can be easily accomplished if the integrands are represented as expan-
sions in a Fourier–Bessel series. Unfortunately, exact analytical solutions
for the obtained systems, as a whole, were not found yet. We are forced
in this situation to confine ourselves by the analysis of some simplified
particular cases only. Let us consider such a simplified example, which
is known in reference as the case of small oscillations
(for more detail concerning the classification of oscillatative models

see in Chapter 1, Subsection 2.1).
3.2 Duffing Oscillator

We can expand the function sin in (5.3.1) in the case (5.3.3) in a power
series
Let us confine ourselves by considering the cube term in (5.3.8) only.

In this case we can transform equation of nonlinear pendulum (5.3.2) in
the following manner
Then we perform some auxiliary transformations. They are intro-

duced as the normalized combinative phase x and the normalized time
where is the characteristic size of the interaction region (for instance,

distance between the electrodes 4 and 6 in Fig. 5.1.1), is the light
velocity in vacuum. Equation (5.3.9) after simple transformations can
be rewritten in the form
where
is the normalized frequency, and
is the Duffing parameter. As noted in Chapter 1, Subsection 2.1, equa-

tion (5.3.11) in references is known as the Duffing equation.
Using the Van der Pol variables
(here all notations are self-evident considering what was said above) we
transform equation (5.3.11) into the standard form similar to (3.4.8),
(3.4.9):
It is easy to be convinced that the hierarchical small parameter of the

problem
has the same order as the Duffing parameter Hence, the Duffing pa-
rameter can be used here as hierarchical problem small parameter.
Accomplishing numerical estimations for for typical characteristic sit-
uations (for instance, taking the device shown in Fig. 5.1.1 as a basis),
we can find that, indeed, the Duffing parameter might be considered
as the small parameter of the problem (really
Apart from that, considering condition (5.3.7) and the fact that
we take into account that
As we did before, we look for the formal solutions of system (5.3.15)

in the form of Krylov–Bogolyubov substitutions (see in more detail in
Chapter 4, Section 2)
where the averaging variables are determined by the truncated (short-

ened) equations (equations for the second hierarchical level)
Following with the standard calculational procedures described in

Chapter 4, Subsection 2.2 we obtain the set of equations similar to (4.2.6)
for determining the unknown functions
Solutions of system (5.3.19) can be found using the method that is set
forth in Chapter 4, Subsection 2. For solution of the problem of secular
terms (see Chapter 3, Subsection 5.2) we should accept the presump-
tion about absence of non-periodical terms in right parts of equations
(5.3.18). As a result we obtain
Integration of the first equation in system (5.3.19) yields
The expressions (5.3.20), (5.3.21) determine solutions obtained for

the first approximation of the averaging method. In analogous way the
solutions for the second approximation can be obtained:
Therein the analysis shows that for any approximation
This means that
always, i.e., the ‘super-slow’ phase of the second hierarchical level (see
definition (5.3.3))
indeed describes the highest hierarchical level of the considered system

(see the diagram in Fig. 5.2.5 and corresponding commentaries to for-
mula (5.3.5)). It is interesting to note that the velocity of changing of
the fast averaging phase in this situation is constant:
Thus above we have illustrated the represented calculational tech-

nology on the example of the single particle theory of the modulation
section of the device shown in Fig. 5.1.1. As it was found, such a system
can be treated (in general case) as a three level hierarchical oscillatative
system (see Fig. 5.2.5). The methodical peculiarity of this system is
that the first hierarchical level is formed by some parametrically reso-
nant oscillatative process, when both fast oscillatative phases (5.2.2) are
represented by evident ones. Whereas, the second hierarchical level is
formed as a result of the non-resonant electron oscillations, which are
characterized by the hidden phase (5.2.33).

OF
THREE ELECTROMAGNETIC
WAVES’ AS A FOURTH-LEVEL
HIERARCHICAL
4.1 Stimulated Oscillation of a Charged Particle
Analyzing the second design version of the discussed device (see
Fig. 5.1.2) we can discover that, in contrast with the above discussed
illustration example in Fig. 5.1.1, this device can be described as a
fourth-level hierarchical system. Characteristic feature of this system
is that here the second hierarchical level, as well as the first one, has
clearly expressed oscillative-resonant nature. According to the classifi-
cation performed in Chapter 4, Subsection 1.3, this resonance could be
determined as a quasilinear one. However, in spite of the obvious phys-
ical novelty, this model does not contain any new, of principle, calcula-
tional features, in comparison with the previously discussed illustration
example. We will confine ourselves by a short semi-qualitative analysis
only below in this Subsection.
So let us discuss the theoretical model that can be considered to be
equivalent to the device shown in Fig. 5.1.2. The graphical illustration
of this model is shown in Fig. 5.4.1. It can be easily seen that the
only difference from the previously discussed model shown in Fig. 5.1.3
is the presence here of the third electromagnetic wave This
is the transformed wave within the retarding system of the device (see
Fig. 5.1.2). Considering influence of this wave we should take into ac-
count that in the discussed system both resonance types (parametric
resonance determined by condition (5.1.2) and the quasilinear resonance

(5.1.3)) occur simultaneously. As we did before, we will perform the
analysis in the moving coordinate system We assume that the trans-
formed (third) electromagnetic wave in the coordinate system
can be approximated with a transverse retarded electromagnetic wave
propagating along axis (see Fig. 5.4.1). We use the method of artificial
magneto-dielectric (see this method is described further in Chapter 11,
Subsection 2.1, Volume II, and in reference [4]) for description of the
retarding property and electrodynamic arrangement of this wave.
The vector-potential of the electromagnetic field acting on electrons
can be represented in the following form:
where and are the cyclic frequency and wave

number of the transformed (third, retarded) electromagnetic wave in
the moving coordinate system The Hamilton equations (5.1.4) are
chosen as an initial basis.
Then we follow the calculational scheme that was set forth in the pre-
vious case (see Subsection 3.2). The difference is only in the following:
in the studied case we consider that the phase is a slowly varying vari-
able (for the zeroth hierarchical level, naturally). This supposition fol-
lows from the physical sense of synchronism condition (5.1.3). The fact
of satisfying the condition (5.1.3) means that any electron during of its
motion along axis ‘sees’ practically the same phase of the transformed
(third, retarded) electromagnetic wave Or, more exactly, this
phase could slowly change only during many electron oscillations with
fast phases This effect is referred to as the elementary mechanism
of the stimulated Cherenkov instability [4]. It is also put in the basis of
operation principle of traveling wave tubes (TWT) [5].
As a result of performed transformations we obtain an equation for
nonlinear pendulum similar to (5.2.36), where influence of the trans-
formed (third) electromagnetic wave is taken into account in the right
hand side of the equation of stimulated nonlinear pendulum
where f is the amplitude of corresponding normalized ‘force’ (it is not im-

portant to specify its expression for the discussed illustration example),
is the averaged phase velocity of the transformed
wave in the moving coordinate system
Analogously, taking into consideration expressions (5.3.1), (5.3.2),
(5.3.8)–(5.3.11) we could obtain the corresponding so called ‘stimulated’
Duffing equation. Passing to the Van der Pol variables (5.3.3) and per-
forming other relevant transformations, we can, in principle, come to the
standard system like to (4.3.1) or (5.3.15) (except the right hand side
only), obtained before for the case discussed above of ‘non-stimulated’
nonlinear pendulum (5.2.36).
Let us briefly illustrate the main idea of the procedure of asymptotic
solution of the ‘stimulated’ Duffing equation. We shall do it for the
discussed example of a four-level hierarchical oscillatative system. Fol-
lowing the calculational procedures described before in Subsection 3.2
the ‘stimulated’ Duffing equation can be obtained in the form
where we presume for the sake of simplicity that Then, using

the Van der Pol variables (5.3.14), we reduce equation (5.3.3) to the
standard form similar to (5.3.18)
where
All other notations were already given above in this Section.
Comparing systems (5.4.4) and (4.3.1) we can observe that both these
expressions possess the similar mathematical structure. This means that
here, in principle, two different ways of constructing the solutions system
(5.4.4) can be realized. The first is non-resonant case, when velocities
of the phases and changing are incommensurable values. In this
situation the calculational scheme described in Chapter 4, Subsection 3.2
can be used. In the second case, if the resonant condition
is satisfied, the resonant case (one version of the Cherenkov instability

[4], as in is mentioned above already) takes place. The calculational
scheme, which has described in Chapter 3, Subsection 4.3, allow us to
obtain asymptotic of the solutions of system (5.4.4). So, the ‘super-slow’
and quasi-fast
phases can be introduced. In the terms of concept of hierarchical three

this means that the following (second) hierarchical level is formed by the
slowly varying phase In turn, regarding the phase
as a basis for forming of the next (third) hierarchical level (with some
‘super-super-slow’ phase that represents the highest level of
the system — see below), we can conclude that the discussed model
could be treated as a fourth-level hierarchical system.
The hierarchical structure of this system is illustrated in Fig. 5.4.2.
4.2 Stimulated Oscillations of an Electron

Ensemble
In the preceding Subsection we considered behavior of a separate par-
ticle (electron) within the field of two electromagnetic waves under stim-
ulated influence of the third transversely directed electromagnetic wave
(single-particle models). But all practical electrodynamics systems of
the discussed type (see, in particular, the examples in Figs. 5.1.1, 5.1.2)
really have to deal with particle ensembles. I.e., such systems always con-
sist of many particles (multi-particle models). From the point of view
of general theory of systems (see earlier Chapter 1, Section 5) difference
between the single-particle and multi-particle models is in fact that the
first one can be treated as a subsystem of the second one. This means
that the multi-particle systems (ensembles) should have some ‘specific
system properties’. This, in principle, distinguishes it from any other
subsystem (e.g., electrons in this case, see Chapter 1, Subsection 5.1 for
more details).
Relevant analysis shows that the discussed multi-particle model (see
Fig. 5.4.1) really indeed possesses such ‘specific system properties’. One
of them is a possibility of realization of the so called grouping mecha-
nisms. In this case, owing to simultaneous realization of many nonlinear
single-particle interactions, the particle ensemble evolves into the form
of clearly expressed particle bunch. In reference to physical microwave
electronics [13, 15] the bunches of this type are quite often treated also as
peculiar electron clusters. Such bunch (cluster), as a whole, can behave

itself as a peculiar ‘large particle’.
As it will be shown below, the grouping mechanism discussed can
be achieved in our model, too. Including the case when the electron
oscillatative ensemble within two electromagnetic waves (transforming
and heterodyne ones, respectively) is exposed to the third (transformed,
retarded) electromagnetic wave (see Fig. 5.4.1). Physical features of this
process will be discussed somewhat below in more detail. Here we state
only that, in spite of obvious difference between the both (single- and
multi-particle, respectively) models, results of the single-particle analysis
could be effectively applied for analysis of the multi-particle systems.
First, for studying of the above mentioned process of the electron cluster
forming. Second, for description of dynamics of this cluster as a large
particle, i.e., the dynamics of the ensemble, as a whole.
So, let us illustrate some characteristic details of semi-qualitative tech-
nique of dynamics analysis for a multi-particle oscillatative particle en-
semble. We shall take the above discussed single-particle results as a
basis for such analysis. As before, we consider that this single-particle
dynamics is described in the terms of theory of hierarchical oscillations.
At that, it is assumed for simplicity that the considered multi-particle
oscillatative ensemble consists of many electrons that do not interact
with each other via the Coulomb repulsion and interact with the elec-
tromagnetic waves only.
Let us remind once more that the chosen system is characterized by
the presence of two resonances of essentially different physical nature.
The first is the parametric resonance (determined by resonant condi-
tions given by the last of formulae (5.2.20)). It can be realized on the
initial (zeroth) hierarchical level. Detailed quantitative and qualitative
multi-particle studies of this physical mechanism will be performed in
Volume II). Therefore, further in this Subsection we will pay our atten-
tion mainly to the discussion of the second resonant mechanism. The
characteristic feature of this mechanism is, as it mentioned above al-
ready, that it, in contrast with the first one, takes place on the first
(averaged) hierarchical level (see Fig. 5.4.2).
Thus taking into account what has been said, we can begin our study
beginning from the first hierarchical level immediately. According to the
previously performed single-particle analysis, we have two main types
of oscillatative motion on this level (see Fig. 5.2.1 and corresponding
comments). The first is the case that is determined by the condition
Here the particle moves ‘under maximums of the
curve (see Fig. 5.2.1). In the theory of nonlinear pendulum
such type of motion is known as the rotating motion of the nonlinear
pendulum. In the framework of the studying model, accepting the ter-

minology of physical microwave electronics, we classify these electrons
as the passage motion of particles [17]. The general motion process in
this case can be represented as a superposition of translation and oscil-
latative motions. As analysis shows, the averaged combinative phase
(which we classified above as the explicit one) characterizes the oscil-
latative component of this motion [17]. Therein the area of such particle
motion is, in principle, unlimited.
The second characteristic type of motion is the case
(see Fig. 5.2.1). In this situation ‘walls’ of the Miller–Gaponov potential
well (5.2.41) limit the aria of electron motion. In other words, here,
in contrast with the first case, the region of the particle motion turns
out to be limited. This means that the motion process, in itself, have
purely oscillatative nature. Here, these oscillations are determined by
the hidden phase (or ) which, in turn, is connected with the phase
of free oscillation of an electron in the bucket (see definitions (5.3.3),
(5.3.14), (5.4.6), (5.4.7) and the corresponding comments). Further we
limit ourselves by analysis of the second type model only.
Let us begin our semi-qualitative analysis with the unperturbed model,
i.e., the model in which the influence of the third (transformed) electro-
magnetic wave (see. Fig. 5.4.1) is neglected. It is considered that the
initial kinetic energy of all electrons is less than the ‘height’ of the Miller–
Gaponov well (see Figs. 5.4.3 and 5.4.4). Kinetic and potential
energies of each electron of the ensemble change in mutually opposite di-
rections (see Figs. 5.4.3 and 5.4.4). As it can be easily seen in Fig. 5.4.3,
electrons which increase their kinetic energy (Fig. 5.4.3a)) at the same
time decrease their potential energy (Fig. 5.4.3b), and vice versa. As
a result the total averaged energy for each electron is conserved in
time (the conservation law for the averaged energy — see (5.2.32) and
Fig. 5.4.4).
Then we assume that the system is exposed to an external disturbing
(third, transformed) electromagnetic wave that is characterized by the
frequency (which,in turn, is close to ) and the phase
We will refer to such a model as perturbed. The state of this model on
certain time is illustrated in Fig. 5.4.5.
For explanation of the illustration given in Fig. 5.4.5 let us remind
once again that different electrons of the ensemble oscillate along
with the differing initial phases
Besides that, we remember the widely known fact that any transverse
electromagnetic wave consists of crossed electric and magnetic fields.
Consequently, the electric and magnetic Lorenz forces (described, for
example, by equation (4.1.4)) should act on electrons in the bucket.
Simplest analysis show that in the nonrelativistic case we can neglect the
influence of the magnetic component of the Lorenz force in comparison
with the influence of its electric component (electric Lorentz force)
where is the electric field intensity of the third field, is the

amplitude of the latter. Here we remind that namely non-relativistic case
takes place in the discussed model (see above the model description given
before in Subsection 1.1). Graphical illustration of the electric Lorentz
force (5.4.8) in some cross-section with the coordinate is
shown in Fig. 5.4.6.
One of the results of an action of the electric Lorentz force is that some
electrons (similar to ones that have numbers 1 and 2) move along the
positive direction whereas other electrons (numbers 4 and 5) move along
its negative direction. In the considered ‘perturbed’ case, in contrast
to the situation with ‘unperturbed’ model (see Figs. 5.4.3 and 5.4.4),
this means that in any fixed cross-section with coordinate
part of electrons always decelerate under the influence of the electric
Lorentz force (electrons number 4,5 — see Fig. 5.4.5b). At the same
time, other particles (electrons number 1,2 and 3) accelerate under action
of this force. As a result, the electrons of the first group lose their
energy, whereas electrons of the second group increase the energy (see
Fig. 5.4.5a).
Then we recall that above in our model we have made the supposition
that the frequency of free electron oscillations is close to the wave fre-
quency (i.e., the following resonance condition is achieved:
This means that all electrons oscillate synchronously collinearly to the
vector of Lorentz force But, as was mentioned above, this occurs
with different (accelerative and decelerative) initial phases. Owing to
such synchronous arrangement of the system, each electron ‘sees’ ap-
proximately the same oscillatative phase of the transformed wave dur-
ing many of its oscillations. In this case it is usually said that the
quasi-stationary action of oscillatative electric wave field on the (also
oscillated) electrons of the ensemble [4] takes place. In vacuum physical
electronics such formulation of the resonant condition is tra-
ditionally called the principle of quasi-stationary interaction [4]. Such
treatment turns out to be very useful for analysis of so called electron
devices with long-time interaction [4].
However, the question arises: what consequences will the fact of re-
alization of the above mentioned long time mechanism have for general
oscillation dynamics of the ensemble? Later we will follow up nonlinear

dynamics of the electron oscillatative frequencies to clarify this problem.
First, we turn to analysis of the ‘unperturbed’ frequency of nonlinear
electron oscillations (5.2.53). Let us simplify the expression (5.2.53)
accepting that the case of small oscillations is achieved (see condition
(5.3.7)). Apart from that we consider that amplitude of the third (per-
turbed) electromagnetic wave is small, too. In such a situation we can
expect that perturbed oscillatative frequency should be close to un-
perturbed one Besides, we can expect that both these frequencies
are close to the ‘linear’ frequency (5.2.35). Then accomplishing rele-
vant expansion in a power series in (5.2.53) by the wave amplitude one

can find the following expression of rough estimation for the nonlinear
frequency
where is the amplitude of oscillation of the electron in the bucket

(potential well) along is the linear frequency (see definition
(5.2.35)), is, as before, the frequency of the standing wave (i.e., the
frequencies of the first and second waves in moving coordinate system
— see corresponding commentaries for definitions (5.1.5)), is the
light velocity in vacuum. Very important thing for further explanations
is the clarification of nature of the dependence of the amplitude
on the averaged kinetic electron energy
For this we use the solutions (5.2.58)–(5.2.61) and set of simplifying
suppositions accepted above. After accomplishing relevant calculations
for corresponding approximate expression the required dependence could
be obtained in the following form:
where is the electron rest mass. Substitution (5.4.10) into (5.4.9)

yields:
Let us turn the reader attention to the two following characteristic

features of the dependence (5.4.11). The first is that the nonlinear fre-
quency is close to the linear frequency i.e., The
second is inverse dependence of nonlinear frequency on the averaged
electron energy Namely, the bigger is the energy, the smaller is
the frequency. Or, in other words, the electrons with bigger energy are
characterized with a bigger oscillatative period.
Let us analyze shortly the dynamics of oscillatative phases of elec-
trons, taking into consideration the obtained results (5.4.9)–(5.4.11),
and accepted above supposition about small oscillations (see (5.3.17)).
First, we consider the model with unperturbed oscillations. Using solu-
tions (5.3.17) after not complex (but rather cumbersome) transformation
we can write the approximate expression
where is the phase of oscillations of the electron of the ensemble,

is the initial phase, is the averaged kinetic energy
of the electron. Above we have accepted that the initial energies
of electrons of the unperturbed ensemble are equal. This means, from
the point of view of relationships (5.4.11) and (5.4.12), that the time
dependencies for electrons can be represented with approximately
straight lines with equal inclination (see Fig. 5.4.7a). Then we discuss
the analogous dependencies for the perturbed model. In this case per-
turbed nonlinear frequency is also approximately equal to unperturbed
one
The difference is only in the following: in this case the averaged kinetic
energies of different electrons are found to be different. Apart from
that, they adiabatically slowly change in time. The electrons that are
additionally accelerated by electric field of the third (perturbed) wave
have bigger energies, and the decelerated electrons have smaller energies
(this situation is clearly illustrated above in Fig. 5.4.5a. The result of
this difference is somehow unexpected. Because the electrons with bigger
energies oscillate more slowly than the decelerated electrons which drive
oscillatative phases of the second ones to phases of the first ones. In more
detail this situation is shown in Fig. 5.4.7b. It can be easily seen that
the ultimate time comes when the slowly and fast oscillating electrons
meet in the vicinity of the same point, where they form the phase bunch
(phase cluster — see in Fig. 5.4.7b). During some time this cluster
remains relatively compact. This means that in this case it behaves as

a large particle.
Thus the process of self-organization (see above Chapter 1, Sections 5,
and 6) occurs in the studied perturbed electron ensemble owing to the
nonlinearity of the system. This example obviously illustrates the dis-
cussed above (in Chapter 1, Sections 5, and 6) so difficult to under-
stand ‘specific system property’. Indeed, collective nonlinear evolution
of the electron ensemble demonstrates the properties that not each of
separate electrons do not possess (in this case this is the property of
self-organization). Here, some initially non-organized electron ensemble
eventually organizes into an electron cluster.
The occurrence of the discussed process, self-organization, from the

practical point of view means that the ensemble as a whole obtain a
possibility to interact with the third (perturbing) electromagnetic wave
coherently, in spite of the fact that the initial system was not coherent
(see Fig. 5.4.5). The physical phenomena of such kind are called in vac-
uum physical electronics the mechanisms of phase grouping [1–4]. They
play very important role in work of many electronic devices such as gy-
roresonant systems (gyrotrons, cyclotron masers), free electron lasers,
traveling wave tubes, etc. (for more detail see Chapters 7 and 8, Vol-
ume II).
REFERENCES 205
At last, let us point out especially that the discussed mechanisms of

phase grouping, as the rigorous analysis shows, is not basic work effect
for the system shown in Fig. 1.1.2, although it could be realizes here
beyond doubt. However, as we had possibility be convinced, it is very
interesting, in itself. It is interesting as well from mathematical point
of view, and the physical one also. Beside that it can have independent
practical applications [17].
References
[1] L.A. Vainstein, V.A. Solntzev. Lectures on microwave electronics. Sov. Radio,
Moscow, 1973.
[4] V. V. Kulish. Methods of averaging in non-linear problems of relativistic electro-
[5] V.V. Kulish. On the theory of devices with difference-frequency signal separa-
tion in an electron beam. Sov. Microwave Electronics, 4:25–38, 1978. Super-high
Frequency Electronics.
Oxford, 1990.
[9] Ju.I. Klymenko, V.V. Kulish, P.V. Birulin. Transformer of frequency of an

optical radiation. Patent of USSR No. 520857 (Cl.H 03D 7/00). Priority of 01
April 1974.
[10] V.V. Kulish, Ju.I. Klymenko. Transformer of frequency of an optical radiation.
Patent of USSR No. 599719 (Cl.H 03D 7/00). Priority of 04 February 1976.
[11] V.V. Kulish. Aspects of the theory of charged particle interaction with electro-
magnetic wave fields. Ph.D. Diss. Theses, Kiev State University, Kiev, 1978.
[12] V.V. Kulish, Ju.M. Jermolaev. Demodulator. Patent of USSR No. 563882 (Cl.H
03D 9/00). Priority of 20 September 1974.
[13] V.V. Kulish, N.Ja. Kotsarenko. Optical frequency transformer. Patent of USSR
No. 615799 (Cl.H 03D 7/00). Priority of 04 June 1976.
[14] V.V. Kulish, N.Ja. Kotsarenko. Free electron laser. Patent of USSR (Cl. H 01
J 25/00; H 01 S 3/00). Priority 05 October 1978.
[15] A.M. Kalmykov, N.Ja. Kotsarenko, V.V. Kulish. Transformation of laser ra-
diation frequency in electron beams. Pis’ma Zh. Tecn. Fiz. (Sov. Tech. Phys.
Lett.), 14:820–824, 1978.
[16] L.P. Landau, E.M. Liftshitz. Field theory, volume 2. Nauka, Moscow, 1974.
[17] V.V. Kulish, A.V. Lysenko. Nonlinear selfconsistent theory of free electron
lasers. stimulated radiation of electrons oscillating in buckets. Ukrainian Phys-
ical Journal, 37(5):651–659, 1992.
[18] A.V. Gaponov, M.A. Miller. On the acceleration of charged particles in mov-
ing high-frequency potential well potential wells. Zh. Eksp. Fiz. (Sov.Phys. –
JETP), 34(3):751–755, 1958.
[19] G.M. Zaslavsky, R.Z. Sagdeev. Introduction in nonlinear physics. Nauka,
Moscow, 1988.
Chapter 6
HIERARCHICAL SYSTEMS WITH

PARTIAL DERIVATIVES. METHOD OF
AVERAGED CHARACTERISTICS
We have mentioned that nonlinear self-consistent resonant wave prob-

lems in electrodynamics challenge of many theorists over the last hun-
dred years (i.e., from Maxwell and Heavyside up to now) [1–10]. In the
general case mathematical description requires taking into account, at
the same time, nonlinearity in the right-hand parts of the correspond-
ing equations, partial derivatives in both their parts, the wave nature
of initial and boundary conditions, etc.. Up to now the number of ef-
fective methods the solving problems of such a type was limited. These
methods include the method of nonlinear dispersion equation [2–6,8–11],
the slowly varying amplitude method (see below Chapter 8) [4,5,12] and
some less widespread approaches. In studying the nonlinear mechanisms
of higher orders (cubic and higher with respect to amplitudes of oscilla-
tions) all these methods are either too inaccurate (the method of non-
linear dispersion equation, for instance) or limited in their application
(method of slow varying amplitudes and others).
Thus the problem of developing some new and more conventional
approaches and improving the traditional ones is rather topical for the
general theory of nonlinear oscillations and waves in electrodynamic sys-
tems. This Chapter (as well as the two following) is dedicated to the
presentation of such a kind of new calculational technology, which, in
turn, is based on the discussed earlier in the book hierarchical ideol-
ogy. The totality of these general ideas and calculational technologies,
in combination with the well known method of characteristics, are called
the method of averaged characteristics [13].
The method mentioned and its some particular realizations (such as
methods of averaging kinetic [14], quasi-hydrodynamic [15, 16], current–
density [15] equations), are described below in this Chapter. The ex-
207
amples of its practical application are given in the following Chapter

[17]. The plasma-like model of a two-beam relativistic electron system
is chosen there as a convenient illustration example.
As practice shows, some traditional methods (method of slowly vary-
ing amplitudes, for instance) in the case of plasma-like systems should be
also upgraded because such a type of objects studied are characterized
by a number of specific features. Chapter 8 is devoted to a discussion of
similar traditional and modernized versions of well known calculational
schemes. Except for this, there are some other relatively new analogous
versions, which are not now too widespread in modern electrodynamic
practice. They are described there too. The examples of practical ap-
plication of these calculational technologies are also given in Chapter 8.
This is done in the framework of nonlinear theory of two-stream insta-
bility.
1. SOME PRELIMINARY INFORMATION

1.1 Motion Equations
Let us consider the model of a plasma-like wave electrodynamic sys-
tem, for instance, an intense relativistic charged particle beam. We
consider that within the interaction volume of the plasmas both possi-
ble types of electromagnetic fields (as well intrinsic (including proper),
and the external ones) may exist.
It is well known that the Euler approach is preferable (see Chapter 1,
Subsection 4.9) for describing of plasma-like systems similar high inten-
sity plasma beams. In contrast to the Lagrange approach, it allows us
to observe the kinematic processes in an arbitrary spatial point that is
characterized by the position vector Therein magnitudes of the in-
stantaneous velocity and the canonical momentum of the particle,
passing through this point, are fixed. Therein each point of plasma beam
in the chosen model corresponds to some point in the velocity space, given
by a velocity field or in the momentum space So the
plasmas could be regarded as a stochastic complex system (about the
stochastic systems see in Chapter 3, Section 2) in the considered model.
The so called Boltzmann kinetic equation in canonical form is chosen
(see also (3.2.4)) for description of motion of a partial beam containing
of particles:
where is the distribution function of particles and is

the collision integral which describes interactions of the particles
METHOD OF AVERAGED CHARACTERISTICS (1) 209
with some particles, is corresponding scale parameter.

The kinetic approach (including the so called one-particle case (6.1.1)
[3]) is rather general description method for plasma-like (including elec-
tron beam relativistic) systems.
Thus the model discussed can be treated in framework of the consid-
ered approach as an obvious example of a hierarchical stochastic oscil-
latory ensemble (see Chapter 3, Section 2). It is important to note that
all particle motions in this case are described by the same single-particle
dynamical equations (4.1.3)) or (4.1.4). The motion of each particle is
characterized by random initial conditions. In our case thermal nature
of initial particle velocities and random electromagnetic interactions be-
tween the particles (which are interpreted here as particle ‘collisions’)
cause its stochastic nature.
In general, we can formulate the following affirmation: the considered
plasma-like system can be represented as a stochastic particle ensemble
described by the distribution function It will be shown be-
low that equations like (6.1.1) can be reduced to standard hierarchical
form like (4.1.5). The mathematical treatment of the beam motion pro-
cess could be possible in the framework of kinetic description by means
of the hierarchical method. However, another description method for
the beam motion problem can be realized, too. But let us discuss this
topic in more detail.
It might seem that the kinetic approach is the only appropriate
method for studying of any electrodynamic plasma-like problems. How-
ever, it is not correct always. The quasi-hydrodynamic approach can be
used also in the cases when certain conditions are held [2,3,6]. This
approach is based on the quasi-hydrodynamic equation as an initial one.
Motion of the plasma beam can be described in this case using model
of a flow of a two-component charged fluid [3]. A few versions of quasi-
hydrodynamic description are known in modern electrodynamics. Be-
low we will comment shortly some simplest way of obtaining the quasi-
hydrodynamic equation.
Let us accept some simplifying assumptions. In particular, we assume
that the pressure difference between output and input of a chosen beam
element can be neglected
transverse characteristic dimensions of the beam (for instance, trans-

verse size) are much greater than the Debye radius (see [2, 3, 6])
and the number of particles inside the sphere of radius is much greater
than unity, i.e.
Other appropriate conditions are held to maintain the above men-

tioned simplest two-fluid hydrodynamic model [3, 6] (here is the av-
eraged (unperturbed) particle density of the plasma beam).
Numerical analysis of collisions between particles in traditional vac-
uum electron devices and accelerators shows that their role more often
is inessential there. Having this in mind, we will consider below that
the collision processes in the studied model is governed by a qualitative
interest mainly. Employing the Bhatnahar–Gross–Krook collision model
[3, 6] allows to realize such a semi-qualitative description. In this way
we obtain the simplest form of relativistic quasi-hydrodynamic equation
(see for more details in [1, 2])
where is the velocity of the beam of as a whole, is the

effective particle collision frequency, is the beam
relativistic factor of particles is their charge, is their rest mass,
is the particle density and is the root-mean-square particle thermal
velocity.
1.2 Field Equations

It is necessary to add the Maxwell equations (1.4.23) for self-consistent
description of the beam motion:
where and are the standard designations for electric and

magnetic field vectors (see Chapter 1, Subsection 4.8 for more details);
and are the current density vectors arising from external and internal
sources of the fields, and are the space charge densities caused by
similar sources, and is the nabla operator.
In the case of kinetic approach the induced densities and are

related to the distribution function through the following
relations:
Using (6.1.6), one can obtain the continuity equation in the form:
In the case of quasi-hydrodynamic approach the densities and

can be expressed as [8,18,19]:
Here and are unperturbed and induced densities of correspond-

ing particles.
The joint integration of the Maxwell equations (6.1.6) with mo-
tion equations (6.1.1) and (6.1.5) enables solution of the complete self-
consistent problem of motion of plasma-like beam.
1.3 Some General Information about Equations

with Partial Derivatives
Let us recall some general concepts and definitions, which are usually
studied in standard course of differential equations with partial deriva-
tives.
Single differential equation. The following equation of general type

usually is replaced with the differential equation with partial derivatives
of a order:
This equation contains at least one order partial derivative of

a unknown function are independent
variables.
It is necessary to find a function such that the function
F becomes equal to zero after substitution of into (6.1.12).
Such a function is called the solution or the integral of a differential equa-
tion with partial derivatives. All partial solutions (excepting additional
solutions) can be obtained from the general integral.
It is well known that any general solution of exact differential equation
(equation with momentum derivatives) can be found with the accuracy
of some arbitrary constants. The specific feature of the equation with
partial derivatives is that instead of the arbitrary constants we obtain
some arbitrary functions. Use some initial or boundary conditions for
the function and its derivatives can define the latter. In the general
case the number of these arbitrary functions is equal to the order of
the differential equation. If we have independent variables then the
arbitrary functions are the functions of variables.
System of Equations with Partial Derivatives . The systems of

equations for a few unknown functions
can be written in the form
where is the current number of the

unknown functions, is the total number of the system equations.
In the case the system (6.1.13) is called determined one, and
in the cases or we have to deal with overdetermined or
underdetermined systems.
Each differential equation of type (6.1.12) or system of equations with
higher derivatives like (6.1.13) can be reduced to the system with only
first derivatives (system of the first order).
Conditions of compatibility. The system of differential equations

(6.1.13) allows obtaining the solutions only
in the case when the given functions and their derivatives satisfy
the conditions of compatibility. These conditions guarantee that the
differentiation of two or more equations (6.1.13) leads to coincidence of
the highest derivatives of the sought functions The conditions of
compatibility can be obtained by excluding of the functions and their
derivatives from some sequence of the equations. In turn, this sequence
could be obtained by differentiation of equations (6.1.13).
2. METHOD OF AVERAGED
CHARACTERISTICS
Let us start discussion of the hierarchical methods of asymptotic inte-
gration of differential equations with partial derivatives with the method
of averaged characteristics [13].
2.1 Concept of the Standard Form

As analysis shows, the method of averaged characteristics is suitable
for asymptotic integration of the systems like (6.1.13) that could be
described using so called standard form (or, the same, standard system
with partial derivatives — see also equation (3.3.1) and corresponding
commentaries)
where are square matrices of size is

some vector-function in Euclidean space with coordi-
nates i.e.
is a given weakly nonlinear periodical ( in
the general case) vector function, is some scalar variable, for instance,
the laboratory time. Therein it is considered that some hierarchy of the
dynamical values (in time or spatial coordinates) could be determined.
Then let us turn to the terminology of the theory of method of char-
acteristics (see also the following Sections for more details). It is not
difficult to prove that the standard form (6.2.1) can be transformed into
so called quasilinear (with respect to the derivatives!) homogeneous
equation
Further in this Chapter we shall be oriented on the studying the stan-

dard equations (and systems of equations) of form (6.2.2).
2.2 General Scheme of the Method

The calculational procedure of the method of averaged characteristic
includes three specific steps. The first step is the straight transformation
of an initial system of equations with partial derivatives to some
higher level. The main task of the second step is the solving of this
equation. Finally, the transformation of the obtained solution from this
hierarchical level to the initial (zero) level (back transformations)
is accomplished during the third step. In turn, each of the mentioned
steps contains a few smaller substeps (sub-stages). All these steps and
substeps are illustrated graphically in Fig. 6.2.1.
We begin with discussing the first step of the considered calculational
algorithm. It, as it can be easily seen in Fig. 6.2.1), includes two more
particular substeps. The main idea of the first substep is based on
some known properties of characteristics of initial differential equation
(6.2.2). Owing to these properties (see the following Subsection for more
details) we have a possibility of reducing the initial equation with partial
derivatives (6.2.2) to some system with momentum derivatives (exact
differential equations — see Fig. 6.2.1):
where all designations are self-evident. Corresponding essential simpli-

fication of the initial problem (6.2.2) can be attained as a result of the
transformations performed. This simplifying could be explained using
the fact that the integration procedure for the equations with momentum
derivatives like (6.2.3) usually is essentially simpler than the procedure
of finding solutions of corresponding equations with partial derivatives
(like (6.2.2)).
Unfortunately, the achieved simplification very often is insufficient for
successive solving of (6.2.2). The point is that problem (6.2.3), espe-
cially in the case of multi-periodical nonlinear oscillations, could be far
from simple, from calculational point of view (this circumstance has been
clearly illustrated previously in Chapters 4 and 5). So the goal of the
second substep (see Fig. 6.2.1) is the accomplishing the asymptotic in-
tegration of equations (6.2.3). The hierarchical calculational technology,
which has been developed previously in Chapters 4 and 5, is proposed as a
methodical basis for this part of the discussed general calculational proce-
dure. So, equations (6.2.3), as before, can be regarded as another form
of the initial equation of the zeroth hierarchical level (see Fig. 6.2.1).
Below we perform the procedure of introducing the fast rotating phases

(see Chapter 4 for more details) as parameters:
where essence of all designations is obvious from what was set forth be-
fore in Chapter 4. Then according to the general procedure (see Chap-
ter 4) we separate the complete set of problem large parameters and
construct the hierarchical series
where all large parameters are determined by the standard method (see
(4.1.7) and corresponding commentaries). The asymptotic solutions of
the system (6.2.4) can be represented in the form of hierarchical sequence
of the Krylov–Bogolyubov substitutions (see Chapter 4, Subsection 1.4).
In particular, in the case of two-level hierarchical system the asymptotic

solutions for system (6.3.2) can be represented in the following form:
where all averaged values can be found from the equations for the first
hierarchical level:
All auxiliary functions could be taken into account by use

of the algorithms described in Chapter 4. For instance, result of such
calculations for the first (by the small parameter ) approximation
could be represented in the form
Taking into account the parametrical nature of the variable (see

Chapter 4), equations for some hierarchical level in the general
case can be written as (see Fig. 6.2.1):
where is some nonlinear functional matrix, the upper

index denotes that the corresponding values belong to the hierar-
chical level. The values and in (6.2.14) are, in principle, dependent
variables (because Lagrange formulation of the problem). However, they
are considered here (that is very important!) as parameters of the prob-
lem. According to what was set forth in Chapter 4, equations for any
hierarchical level (4.2.14) by themselves are the averaged
equations (6.2.3) (equations for the hierarchical level).
Then, keeping the parametrical form of the equations (6.2.14), a spe-
cial hypothesis is put in the basis of the considered step of the discussed
algorithm. Namely, it is supposed that equations (6.2.14) can be treated
as characteristics of some other equation with partial derivatives:
It should be specially pointed out that, in contrast to (6.2.14), the

variables and are the independent ones (because of the Euler form
of representation).
Comparing (6.2.15) and the initial equation (6.2.2) we find that both
these equations have analogous in principle mathematical structure, i.e.
hierarchical resemblance principle remains here. Because of their spe-
cial form, we call equation (6.2.15) the averaged (truncated) quasilinear
equation of the hierarchical level. Besides that, the second hierar-

chical principle is also satisfied (hierarchical compression principle — see
Chapter 2, Section 2 for more details). According to this principle the
initial hierarchical system (6.2.2) (zeroth hierarchical level) reproduces
itself on each hierarchical level in its general dynamical structure. So it
turns out to grow simpler as the hierarchy level grows higher. The other
two hierarchical principles are satisfied here, too.
The passage from the equation with partial derivatives (6.2.2) to equa-
tions of characteristics (6.2.3) (exact differential equations) is similar to
the passage from the Euler form of writing to the Lagrange form (about
the Lagrange and Euler formulations of the problem see in Chapter 1,
Subsection 4.9). Correspondingly, the passage from (6.2.14) to (6.2.15)
describes the passage like the passage from the Lagrange form to the
Euler one. However, it should be mentioned that we are talking about
some resemblance between both procedures only in the general case.
Exact equivalence takes place only in some particular cases (see below
Subsections 5–7). Mutual resemblance of these passages here is kept due
to the use of parametrical (with respect to the variables and ) form
of representation in corresponding differential equations.
It can easily be seen that equation (6.2.15) could be treated as ini-
tial equation (6.2.2) (zeroth hierarchical level), which is transformed to
the hierarchical level by means of the straight transformations (see
Fig. 6.2.1). So, taking into consideration the said above, we can af-
firm that eventual result of the first step of the discussed calculational
algorithm is the obtaining of averaged equation with partial
derivatives like (6.2.15) (i.e., the equation of hierarchical level).
The totality of calculational procedures of the first step we treat as the
straight transformation of the initial equation (6.2.2) into hierarchically
equivalent equation (6.2.15) for some level (see Fig. 6.2.1).
The main task of the second step of the discussed algorithm is the
solving of the equation (equations) (6.2.15). Analogously to the sit-
uation that takes place in the framework of the hierarchical methods
for exact differential equations (see Chapters 4 and 5), the transformed
equation for any hierarchical level (6.2.15) is also simpler than the
initial equation (6.2.2). Owing to this, the general procedure of solving
averaged equation (6.2.15) turns out to be simpler than the analo-
gous procedure for equation (6.2.2). In principle, maximal simplification
could be attained in the case when the straight transformation of ini-
tial equation (6.2.2) is accomplished up to the highest hierarchical level
But it should be mentioned especially that quite often there is
no strong necessity for performing of such ‘complete’ hierarchical trans-
formations. The point is that sufficiently simplified (because they are
averaged), solutions could be found on some of the lower hierarchical

levels
In what follows, it should be noted that any known analytical or nu-
merical calculational method can be applied for the solution of equation
(6.2.15). However, very interesting situation occurs in the case of appli-
cation for this purpose of the method of characteristics. As it will be
shown below in the next Section 3, this method has one specific feature.
Namely, the possibility of constructing solutions of a corresponding equa-
tion with partial derivatives, basing on the solutions of the equations for
characteristics. The characteristics, as was mentioned above, are exact
equations (equations with momentum derivatives). This means that in
that case, if the characteristics method is used, we, strictly speaking,
have no necessity for constructing of averaged equation (6.2.15)
(see the dotted line in Fig. 6.2.1). Their corresponding solutions can be
constructed immediately on the basis of solutions of equations for char-
acteristic (6.2.14) (see the following Sections for more details). But, in
any case, independently with the solution method used, we eventually
obtain the solutions of equation (6.2.15) in the form
Constructing solution like (6.2.16) completes the main task of the

second step of the discussed calculational algorithm.
Analyzing the form of solution (6.2.16) and possibilities of construct-
ing corresponding solutions of initial problem (6.2.2), we can observe a
new peculiar calculational situation. It realizes owing to the use of the
hierarchical procedure, which has been developed earlier for solving of
the exact equations (see Chapters 4, 5). However, let us discuss this
problem in more detail.
First of all we turn the reader’s attention to the following circum-
stance. In fact, the problems discussed earlier in Chapters 4 and 5 can
be classified as one-dimensional, because all dynamical variables change
in this case with the temporal coordinate only. (Here let us point out
that speaking further in this Chapter about one-, two- and more dimen-
sional problems, we will mean the total quantity of variables (the spatial
coordinates and the temporal one). In contrast, in other Chapters in
analogous situations we will talk about the spatial dimensionality only).
On the other hand, we used the same time for description of func-
tional dependencies on any hierarchical level, i.e.,
and so on. Spatial coordinates, vice versa, are dif-
ferent for different hierarchical levels. However, in the case of exact
equations this does not lead to any difficulties because we have the tem-
poral dependencies like on any hierarchical level. Owing

to this we obtained a chance of constructing solutions for any lower hi-
erarchical level, substituting the solutions for higher hierarchical levels
into the Krylov–Bogolyubov substitutions (4.2.2)
where all designations were given earlier in Chapter 4, and, therefore,

they are self-evident. Thus the relationships like (6.2.17) play double role
in the discussed situation. Namely, using them we have accomplished the
straight as well as the back transformations in the framework of method
described in Chapters 4, 5.
We have a different, in principle, situation in the case of differential
equations with partial derivatives. Here, besides the temporal functional
dependencies, we have also dependencies on spatial coordinates (see the
form of standard equations (6.2.1) and (6.2.2)). In contrast to the pre-
ceding case, variables of both these types are independent from each
other (the Euler description — see Chapter 1, Subsection 4.9). But as
was mentioned above, the spatial coordinates are different for different
hierarchical levels. This means that corresponding functions de-
pend — on each hierarchical level — on a proper set of variables for
this level. As before, the Krylov–Bogolyubov substitutions give us a
connection between different neighboring hierarchical levels:
i.e., it could be used for the straight transformations in this case also.
But in contrary with the above discussed situation with exact differen-
tial equations, here we can not immediately use the relationships like
(6.2.8) for constructing solutions of the initial problem (6.2.2) (i.e., for
performing the back transformations). For this we additionally should
have some special correlation between proper coordinates of the neigh-
boring hierarchical level like The correlation of such
type, as will be shown below, can be constructed really. Modernized in
such manner calculational procedures embody the main task of the third
step of the considered algorithm. Totality of all procedures, which allow
to accomplish the transformation of solutions from any hierarchi-

cal level to the zeroth hierarchical level (see Fig. 6.2.1), we call back
transformations.
The main idea of the back transformations is based on the following
observation. If we look at the Krylov–Bogolyubov substitutions (6.2.18)
we can see that difference between any variables of two neighboring
hierarchical levels ( and for example) is proportional to
and so on.
Taking this into account, let us expound the function
in the Taylor series in the vicinity of equivalent spatial point of
hierarchical level.
Substituting then all differences like (6.2.9) into this series we obtain
the required set of approximate correlation between any two neighboring
hierarchical levels. For instance, we can do it with respect to the differ-
ence between the first (averaged) and zero (initial) levels, respectively:
and so on.
Below we accomplish the chain of successive analogous back trans-
formations between each pair of neighboring hierarchical levels. As a
result, the back transformation procedure sought could be constructed
eventually. It allows to accomplish the back transformations from any
higher hierarchical level (including the highest one ) to initial
(zero) hierarchical level
Thus we can realize that key point of the discussed above calculational
technology is a special combination of the concept of characteristic and,
at the same time, the hierarchical algorithm for asymptotic integration

of exact differential equations described above in Chapters 4 and 5. In
other words, the method of characteristics is important part of the dis-
cussed method of averaged characteristics. That is why further we shall
pay special attention to the discussion of specific properties of the char-
acteristics, including the method of characteristics itself.
3. CHARACTERISTICS AND
THE METHOD OF CHARACTERISTICS
3.1 Method of Characteristics. The Scalar Case
We begin with the scalar differential equation of the first order because
it is the most convenient for the following:
where F is a given function of its arguments, is the unknown function

on arguments In this Subsection let us restrict ourselves with
study of the specific case of equation (6.3.1) which can be described by
the quasilinear differential equation of the type
where R are some given functions on It should be men-

tioned that equations of such type are rather widespread in electrody-
namics. Besides that, they are also typical for nonlinear problems of hy-
drodynamics, thermodynamics, quantum physics, etc.. So, our choice,
apart from merely methodical meaning, could have some practical sig-
nificance, too.
A few calculational schemes are known for solving equation (6.3.2)
using the method of characteristics [20–25]. Let us consider two of them.
The first scheme of constructing the general solution. The first

scheme of constructing the general solution of equation (6.3.2) we realize
in the form of the following four steps:
1) the system of equations that determines the characteristics of sys-
tem (6.3.2) is constructed:
2) independent integrals of the system (6.3.3) are found:

3) some arbitrary function of these integrals is constructed; this

function is set equal to zero (that is implicit solution of equation (6.3.2)):
4) further, the general solution of equation (6.3.2) should be obtained

by the solving of equation (6.3.5) with respect to the function
The second scheme of constructing the general solution. The

second scheme of constructing the general solution of equation (6.3.2)
differs with the first one only with the scheme of constructing of the
characteristics:
Then, steps 2–4 of the preceding scheme are performed.

The complete solution of the considered problem could be found only
in the case when initial and boundary conditions are given, i.e., the
Cauchy problem should be solved.
The Cauchy problem. The essence of the Cauchy problem is finding

the solution that satisfies the initial condition
on some surface The latter is given by the equation
This means that amongst interminable number of functions (6.3.5)

there should be chosen functions that satisfy the initial condition (6.3.7)
in case when condition (6.3.8) is satisfied.
A few algorithms of the Cauchy problem solution are known [20–25].
Now let us comment briefly two of them.
The first algorithm. For the sake of finding the solution

for equation (6.3.2) (for given initial conditions (6.3.7) or (6.3.8)) it is
necessary:
1) to construct the system for characteristic (6.3.3) or (6.3.6);

2) to find all independents integrals (6.3.4) for the characteristics;
3) to construct the algebraic system for the founded integrals and for
the initial conditions:
4) to exclude all variables and to obtain the equation of the type
5) substitution the left sides of equations (6.3.4) into (6.3.10) to obtain

the equation
6) solving (6.3.11) to find solution sought for initial

condition (6.3.7), (6.3.8).
The second algorithm. For the finding the solution of equation

(6.3.3)
with the given initial condition
it is necessary:
1) to construct the system of characteristics (6.3.3) or (6.3.6);
2) to find all independent integrals (6.3.4);
3) to replace the independent variable in integrals (6.3.4) with its
initial value
4) to solve equation (6.3.14) with respect to to obtain

the system:
5) to construct the relationship:
which gives the sought solution of the Cauchy problem in an implicit

form,
6) to solve (6.3.16) with respect to to find sought solution
for initial conditions (6.3.13).
We will widely use below in this Chapter the described algorithms
for constructing the calculational schemes by the method of averaged
characteristics. But considering the important role of the characteristic
method, let us give two simple examples for illustrating these algorithms.
The illustration example 1. The first scheme of constructing of gen-

eral solution and the second algorithm of solving of the Cauchy problem.
Let us find a solution of the equation
The initial condition is chosen in the form
Then, we follow the above-described scheme and algorithm:

1) the system for characteristics (6.3.3) is constructed:
2) the system integrals (6.3.4) are found:
3) substituting the initial conditions in integrals (6.3.20) we find:
4) solving this system with respect to we get:

5) relation (6.3.16) is constructed
6) accomplishing the replacement of the found constants in

(6.3.23) (6.3.21) we eventually have the solution of the problem in the
form
The illustration example: . The second scheme of constructing gen-

eral solution and the first algorithm of solving the Cauchy problem. We
shall look for solution of the equation
with the initial condition
Further we follow the scheme described above and algorithm:

1) the system for characteristic (6.3.6) is constructed:
2) the system integrals are found:
3) substituting the initial condition (6.3.26) in (6.3.28) we obtain:
4) it could be found after simple transformations of (6.3.29) that
5) substituting the expression for constants (6.3.28) in (6.3.30)

we get the problem solution:
6) which can be rewritten in more convenient form:

3.2 Method of Characteristics. The Vector Case

Above we discussed the simplest version of the method of character-
istics when the function sought is represented with a scalar. Now we
shall generalize these results for the case of a vector-functional variable
Therefore we start with the studying of the one-dimension problem,
which is described with a system with two independent variables
One-dimension problem. Let us write down the system of scalar

equations that describes some two-dimension problem in the following
form:
where and are some given functions.

The system (6.3.33) could be rewritten also in the matrix form:
where
are the corresponding vectors and vector-functions. It should be men-

tioned that the matrix form (6.3.34) is interesting also because it allows
simple generalization the discussed scheme on the case of more then two
(for instance, ) equations (6.3.33). In this case the matrices A and
B have size and the matrices and are determined as
column matrices.
Characteristics of the equations (6.3.33), (6.3.34) can be defined as
lines (which are given by the differentials of shift ) along which the
following condition is satisfied:
where E, in the general case, is the unity matrix of size Therein

the vector is not arbitrary. It should satisfy the relationship
The condition (6.3.37) is referred to as the relationship of character-

istics.
Canonical form. The particular case of equation (6.3.34)
is called the canonical form. Here The charac-

teristics are the lines that are described with the equations
where the characteristic inclinations are calculated as characteristic

roots of the matrix C
In the case we have to deal with multiple and real roots only, the
considered system is related with the parabolic type of equations with
partial derivatives. Analogously, the system could be classified as a
hyperbolic one in the case when all roots are complex.
Now let us make one more remark. The systems of the type
form the class of so called quasi-hydrodynamic (hydrodynamic) equa-

tions (see, for instance, equation (6.1.5)). Comparing (6.3.41) with
(6.3.33), (6.3.34) we can conclude that matrices A and B in the quasi-
hydrodynamic case are found to be dependent on the sought function
Nevertheless, as well known [20–25], in spite of this difference such
system can also be solved using the method of characteristics. Specific
feature of this calculational scheme is that the characteristics depend on
the solution considered. Or, in other words, the characteristics depend
in this case on the solution that is chosen for analysis, i.e., the line that is
a characteristic for certain solution is not the characteristic for another
one.
Two-dimensional problem. Later we shall discuss the calculational

situation when the number of variables is three respectively).
Let us write down the system of three equations with three unknown
functions
In this case the characteristics of the system (6.3.42) are the surfaces
for which the conditions
or
are satisfied. Here is the vector of the normal to the surface

S. Solutions for the characteristics could be obtained if the solutions
of equations (6.3.43) or (6.3.44) are substituted into the initial equation
(6.3.42). However, it should be mentioned that, in the general case, cal-
culational technology in this situation becomes somewhat complicated.
It becomes even more complicated in the four- and greater dimensional
cases.
The illustration example for the vector case. Let us consider the
problem of propagation of the sound wave within some physical medium
(see the corresponding qualitative picture in Section 3, Chapter 1). The
equation system that describes this process can be written in the form
where and are some constants. System (6.3.45) in the matrix form
can be written as
The equation for characteristics (6.3.36) is obtained in the form

Thus we obtain the characteristics of the system (6.3.45) as lines that

are described by the equation
It can easily be seen that they are straight lines:
In what follows, we find the relationship for the characteristics like in

(6.3.37). Analyzing the rank of the matrix in view of the relationship
(6.3.37)
we eventually have the condition
Substituting the characteristics (6.3.48) into (6.3.51) one can obtain

the following equations:
which could be easily integrated by the method described above:

where the arbitrary functions and are determined by the correspond-

ing initial conditions.
4. EXAMPLE: APPLICATION OF
CHARACTERISTICS
FOR A SIMPLEST SYSTEM WITH
OSCILLATATIVE RIGHT PARTS
As we can see, arrangement and calculational technology of the
method of averaged characteristics are far from simple. A practice shows
that their comprehension sometimes requires some additional training.
Therefore, this Section and the following Chapter are dedicated to dis-
cussion of illustrative examples which allow better understanding of the
main calculational peculiarities of the method.
The choice of the examples was done according to the following prin-
ciple. The first example (that is set forth below in this Section) have
purely abstract nature. Here a simplest system of two equations with
oscillatative right parts is chosen as a basic illustration object. The goal
of this example is to demonstrate the practical calculational technology
of the method using simplest means. The second example [17] (which is
set forth below in Chapter 7) has some other purpose. Namely, specific
features of the use of considered method in a real electrodynamic prob-
lem are shown there. This includes as well the problem statement (i.e.,
reducing the initial electrodynamic equations to the standard form) and
the solving of the obtained standard equations. The nonlinear theory of
two-stream instability in two-velocity relativistic electron beam is chosen
as a convenient illustration object for this.
4.1 Initial Equations

Let us write down the considered system of equations in the following
form:
where and are some unknown functions sought, are

given parameters of the problem.
4.2 Characteristics
According to the general procedure of the method of averaged char-
acteristics, we should construct on the first step the equations for char-
acteristics of equation (6.4.1) (see Fig. 6.2.1). For this we find the de-
terminant (6.3.36)
Thus we find for the system characteristics:
Using well known definition for momentum derivative
we can get (from equation (6.4.1)) the two following systems:
and
respectively.
4.3 Passage to the First Hierarchical Level

Below, regarding the system (6.4.5), (6.4.6) as equations for the zeroth
hierarchical level, we perform the passage to the first (one-fold averaged)
hierarchical level. For this we transform them into the standard form like
(6.3.2). The first stage of this transformation procedure is introducing
the new dependent variable (oscillation phase )
The following stage is the subdividing of all variables into slow and
fast ones. Taking into consideration the specific mathematical structure
of systems (6.4.5), (6.4.6), we can assume that the variables and can
be treated as the slow variables. (Here let us recall that slow variables are
the ones whose magnitudes do not essentially change during the system
period — see Chapter 1, Section 2). The phase correspondingly, plays
the role of the fast variable (because it changes greatly during the system
period). Then the large parameter of the system (hierarchical large
parameter) can be determined by the standard method (see Chapters 3–
5):
where
Differentiating the definition (6.4.7) by it is not difficult to obtain
the differential equation for the phase As a result of the performed
transformation we can reconstruct equations (6.4.5), (6.4.6) into the two
parametrical standard forms like (6.3.2):
and
Below we will accomplish the transformation of systems (6.4.9) into

the first hierarchical level. The formal solutions systems (6.4.9) can be
represented in the form of Krylov–Bogolyubov substitutions:
where the unknown functions and can be found by the use of

corresponding procedures of the Bogolyubov–Zubarev method described
in Chapter 4. The corresponding equations for the first hierarchical level
could be obtained analogously. Regarding this equations as character-
istics of corresponding averaged equations in partial derivatives (i.e.,
equations for the first hierarchical level — see Fig. 6.2.1), we can con-
struct them in the form
The solutions of truncated system (6.4.11) can easily be obtained:
where the initial (or boundary) conditions for the arbitrary functions
(solutions!) coincide with the analogous conditions for equiva-
lent functions for equations (6.4.1). According to the described in Sec-
tion 2, the back transformations should be performed (see Fig. 6.2.1) for
obtaining of the complete asymptotic solutions for the zeroth hierarchi-
cal level.
4.4 Back Transformations

It is not difficult to reconstruct the first two Krylov–Bogolyubov sub-
stitutions (6.4.9) into the form (first approximation on the small param-
eter
Then we use the calculational scheme of back transformations de-

scribed previously in Section 2 (see (6.2.19), (6.2.20) and correspond-
ing commentaries). In particular, we note that difference between any
variables of the first and zeroth hierarchical levels is proportional to
Using this observation we can accomplish corresponding expansions

in the Taylor series in (6.4.13). As a result, we obtain the transformation
formulae like (6.2.20):
and so on.
After not very difficult calculations we have the complete approximate
solutions of the considered problem (here we restrict ourselves by keeping
only the terms of order in (6.4.15):
Thus the initial problem (6.4.1) is solved analytically by using the

method of averaged characteristics.
5. HIERARCHICAL METHOD OF
AVERAGED
QUASI-HYDRODYNAMIC EQUATION
As has been illustrated above, one of the key points of the method of
averaged characteristics is the transformation of an equation with partial
derivatives into some equivalent system of exact differential equations
(see Fig. 6.2.1 and corresponding commentaries). However, it should be
mentioned that such a type of straight transformation technology is not
the only one. Moreover, historically the first time the analogous, in prin-
ciple, idea (that is, however, based on the use of definition for momentum
derivative (3.2.3), (6.4.4)) was when it was proposed for nonlinear wave
resonant electrodynamic problems [14, 16, 26, 27]. This hierarchical ap-
proach is found to be suitable for solving motion problems mostly. The
Boltzmann kinetic equation (3.2.4), (6.1.1) has been used there as a ba-
sis (see Chapter 3, Subsection 2.2). That is why the developed method
is called the method of averaged kinetic equation (see Section 7 below in
this Chapter).
Unfortunately, the method of averaged kinetic equation turns out to
be inconvenient in some calculational situations. First, this takes place
quite often in the case of self-consistent problems. The matter is that
the field part of this problem is solved in three-dimensional space (be-
cause of the Maxwell equations), whereas Boltzmann’s equation (3.2.4),
(6.1.1) (the motion part of problem) is written for six-dimensional space.
The connection relationships (6.1.7)–(6.1.9) are used for the spatial co-
ordination of the field and motion problems. As a result the general
calculational scheme sometimes looks somewhat complicated.
Second, the taking into account of kinetic properties of the considered
plasma-like system quite often turns out not necessary (for instance,
in the theory high-current electron and ion beams in accelerators or
relativistic electron devices [1, 2, 28–30] — see Volume II). So the kinetic
approach, which is put in the basis of the method of averaging kinetic
equation, possesses superfluous possibilities in such situations. But the
cost of this superfluous generalization is the unjustified complication of
calculational procedures.
Some other version of the same idea had been proposed in [16] and
developed in [15]. The quasi-hydrodynamic equation (6.1.5), (6.3.41) is
used here as an initial basis. Because of this the calculational procedure
proposed is called the method of averaged quasi-hydrodynamic equation.
Inasmuch as the quasi-hydrodynamic equation is three-dimensional, the
above mentioned problem concerned the spatial coordination does not
arise there. Besides that, the kinetic plasma processes are not considered
directly in the framework of the quasi-hydrodynamic description (the

semi-qualitative estimation of influence of the kinetic effects was done
there only). Due to these peculiarities, the method of averaged quasi-
hydrodynamic equation in such cases provides essentially less labour
consuming nature of corresponding calculational procedures.
In what follows, let us note the following. In spite of their external
differences, the method of averaged characteristics, as well as the meth-
ods of averaged kinetic and quasi-hydrodynamic equations, actually have
very close general calculational ideology. The method of averaged char-
acteristics looks more general than the other two. However, this advan-
tage can be really attained in the case only when number of equations
in the considered system does not exceed three. At the same time, the
methods of averaged kinetic and quasi-hydrodynamic equations have no
such kind limitations. Therefore, all these method have the ‘right to
live’ in practice as well as optimal areas of their application.
Considering the above mentioned resemblance of the general calcula-
tional ideologies, we classify the methods of averaged kinetic and quasi-
hydrodynamic equations as some special modification of the method of
averaged characteristics.
5.1 Averaged Quasi-Hydrodynamic Equation

Let us choose quasi-hydrodynamic equation (6.1.5) as a basic equation
for description of an electron beam motion. So, as in the case of the
method of averaged characteristics, the main task is the representing of
the initial problem (6.1.5) in the hierarchical standard form like (4.1.5).
Comparing (4.1.5) and (6.1.5), we see clearly that, as before, formally the
problem actually consists in reducing the equation in partial derivatives
(6.1.5) to system of exact differential equations (4.1.5).
In contrast to the method of averaged characteristics, this time we
solve the straight and back transformation problems by means of the
passage from the Euler description to the Lagrange one and then in the
back direction. Here, the definition for momentum derivative (3.2.3),
(6.4.4) is put in the basis of these transformations. Below we express
from (6.1.5) in terms of Lagrange rotating phases It is
noteworthy that the transformation of variables drives us into the space
whose each point oscillates with all the frequencies of particle beam os-
cillations. It is obvious that this oscillation coordinate system is not
inertial because it moves with an acceleration.
Thus at the first stage of calculational we replace time coordinate
with particle oscillation phases that are considered as parameters.
Then we pass to the momentum derivative using (3.2.3), (6.4.4). This
way we can replace quasi-hydrodynamic equation (6.1.5) with the new

set of exact differential equations:
where sense of the functions and is obvious. For instance, can

be determined as a sum of electromagnetic forces acting on the particle
beam (including electric and magnetic Lorentz forces (see (4.1.4)), and
the forces of electromagnetic ‘friction’ and ‘viscosity’, respectively):
In the case that some or all phases of are wave, use of (4.1.2) yields
For hidden phases we have somewhat more complex definitions (see

Chapter 4, Subsection 1.3), and so on.
The quantities are functions of and obeying Maxwell’s
equations (6.1.6) and the particle/plasma continuity equation (6.1.10).
Now we classify the variables as slow and fast ones. In similar way
with ‘purely’ single-particle case (see, for instance, the example consid-
ered in Chapter 5) we take beam velocity and position vector as
slow variables. Moreover, the combination phases (see (4.1.9)) are
slowly varying quantities (this allows to choose the type of resonance
in the system — see Chapter 1, Subsection 2.3 and Chapter 4, Subsec-
tion 1.3), too. Analogously, amongst other (non-resonant) phases of the
set not included in combinations (4.1.9), we also find slowly vary-
ing variables. The rest of the phases of such type and the combination
phases (see (4.1.9)) are the fast ones.
Below we use the calculational hierarchical scheme described in Chap-
ter 4. After corresponding calculations the set (6.5.1) can be reduced
to the form coinciding (with respect to its general mathematics struc-
ture) with standard hierarchical form (4.1.5). In this case, dynamical
operators (2.2.7) for the first hierarchical level can be written as the
Krylov–Bogolyubov substitutions (see (6.2.6)):
and so on.
For equations of the first hierarchical level we have (see (6.2.7)):
and so on, where functions and

other similar functions are calculated using written above formulae of
averaging method (see Chapter 4).
Furthermore, with (3.2.3) we come back from the momentum deriva-
tive (with respect to time ) to a partial one (i.e., with respect to and
Euler’s vector-coordinates taking into account the equation (6.5.7).
As a result we obtain the new equation:
Comparing (6.5.8) and initial one (6.1.5) we find that both these
equations have analogous mathematical structure, i.e. hierarchical re-
semblance principle holds in the problem wave considered. Because of
special form of the latter we call equation (6.5.8) the averaged quasi-
hydrodynamic equation of the first hierarchical level. Besides that, the
second hierarchical principle is also satisfied (hierarchical compression
principle). This means that the original hierarchical system on each
hierarchical level is reproduced in its general dynamical structure, but
essentially simpler. The other two hierarchical principles are satisfied
here, too.
In the simplest case the system contains only two hierarchical levels
— the zeroth and the first ones. So, in order to obtain complete solu-
tion of initial non-averaged equation (6.1.5) we find solutions of trun-
cated equation (6.5.8) and substitute them into transformation formulae

(6.5.4), (6.5.5).

Let us recall once more (see (6.2.19), (6.2.20)) that the main idea of
the backward transformation is based on the fact that difference between
of solutions of two neighboring levels is (see also formulae (6.2.18) —
(6.2.20) and corresponding commentaries)
and so on, i.e., inversely proportional to the power of the scale parameter
Consequently, we can obtain approximate solutions of required
accuracy using corresponding expansions:
Here, in eventual results, they can be expressed in terms of the original

phases using expressions (4.1.9) and (6.5.9). If more than one hierarchi-
cal level exists in our system, asymptotic integration procedure is used
according to (6.5.8) as a new base for calculational. Or, in other words,
we again reduce (6.5.8) to (6.5.1), separate the next scale parameter
(in a hierarchical series (4.1.6)) and the next vector of fast phases
and so on, until we arrive at the last hierarchical level Then we
solve corresponding averaged quasi-hydrodynamic equation and
perform back transformations using the above scheme.
6. THE METHOD OF AVERAGED

CURRENT–DENSITY EQUATION
Below, let us discuss hierarchical method for calculational of nonlin-
ear current density in plasma-like wave systems. We call the discussed
calculational method the method of averaged current–density equation.
6.1 Averaged Current–Density Equation

We use relativistic quasi-hydrodynamic equation (6.1.5) and relations
(6.1.11) as initial set of equations. After trivial transformation, we ob-
tain the working form of current–density equation for the zeroth hierar-
chical level:
where subscript is omitted for simplicity, the quantities are

considered as predetermined functions of time and coordinates the
relativistic factor in (6.6.1) is represented in the form
The other parameters were defined earlier already. Then, using sim-
ilar (to the case of quasi-hydrodynamic equation method) calculational
scheme we write down the dynamical operators and corresponding equa-
tions for the first level of hierarchy:
where and are the slowly and fast varying averaged vector-phases,
which include as components the ordinary phases as well as combina-
tion ones (4.1.9); are functions, whose calculational
procedures are given in Chapter 4.
Further we construct the hierarchical series of large parameters. In
the case we restrict ourselves with taking into consideration the first
hierarchical level only (characterized by the large parameter ), we
can transform (6.6.5) into the Euler form (i.e., the form with partial
derivatives):
Here all phases are considered as pa-

rameters. The equation (6.6.8) is called the averaged current–density
equation.
In what follows we solve (6.6.8) considering the averaged current as
function of and It is obvious that equation (6.6.8) is essentially
simpler than initial equation (6.6.1) because it does not comprise depen-
dence on fast phases or During the next step of the discussed
calculational scheme we perform backward transformation from the
averaged space back to space. There we keep in mind
that functions and can be found according to the scheme
(6.5.9), (6.5.10).
In opposite case when the number of hierarchical levels that we take
into account exceeds two, we must turn to use multi-level version of
hierarchical scheme. Namely, we take equation (6.6.5) as a new basis
for calculations and separate out from the slowly varying vector phases
the fast phase of the second hierarchy The latter is characterized
by the next hierarchical scale parameter in hierarchical series
(4.1.6). Then equation (6.6.8) is reduced to standard form for the second
hierarchical level. Accomplishing an analogous transformation we obtain
the Krylov–Bogolyubov substitutions connecting the first and the second
hierarchical levels:
The two-fold averaged values are calculated by using the equation for
the second hierarchical level:
and so on.
Here is the averaged part of the vector from which new fast phase
was separated out. Considering the above reasoning, the meaning of
other functions is evident.
Following successive calculational scheme eventually we can reach the
ultimate (i.e. ) hierarchical level. It is important that
averaged current density function does not depend on any oscillating
phases:
and so forth.
Here, as before, subscript and superscript denotes the order of
multiplication of performed averaged procedure.
The obvious advantage of equation (6.6.14) lies in the fact that the lat-
ter is simpler mathematically than similar equations of lower hierarchical
levels, including the lowest (zero) level (6.6.1). We employ some calcula-
tional method (numerical or analytical) to find solutions of
( averaded) current density equation (6.6.14). Then performing
back transformations we obtain non-averaged function the
solution determining the dynamics of the zeroth hierarchical level equa-
tion.

The equation (6.6.14) can be solved by means of the successive ap-
proximation method or the characteristics method. Then, we substi-
tute the solutions, which are obtained for into Krylov–Bogolyubov
substitution (6.6.13). This allows to construct corresponding solution
for In turn, we substitute the latter into similar expression for
and so on. In the final stage of successive substitution procedure,
we have an analytical solution for non-averaged current density (6.6.3)
depending on corresponding averaged values of quantities involved. We
take into account algorithms analogous to (6.5.9) and (6.5.10). For this,
we rewrite (6.6.3) as
from which the difference
Consequently, we can write down the solution for non-averaged cur-

rent density as the expansion:
where and are initial values of these

functions (see definitions (4.1.9) and corresponding comments).
7. HIERARCHICAL METHOD OF
THE AVERAGED KINETIC EQUATION
7.1 Averaged Kinetic Equation
Similar calculational scheme can be constructed for the Boltzmann
kinetic equation (6.1.1) chosen as an initial equation. Let us discuss
main peculiarities and advantages of such calculational procedure.
First of all we note that characteristics of equations (6.1.1) coincide
with the single-particle Hamiltonian equations (4.1.3). So the method of
averaged kinetic equation, as well as the method of quasi-hydrodynamic
equation, can actually be considered as specific case of the method of
averaged characteristics.
By virtue of this or using relationship (3.2.3), we can reduce the gen-
eral calculational procedure to the form analogous to presented above
in the method of quasi-hydrodynamic equation. For this we express

the functions and (in the left-hand side of (6.1.1)) in
terms of Lagrange rotating electron phases used further as parameters.
Then using Liouville’s theorem and passing to momentum derivatives
we replace the kinetic equation (6.1.1) with the new set
The set (6.7.1)–(6.7.4) is similar to the set of quasi-hydrodynamic

equations (6.7.6), (6.7.7), with respect to general mathematical struc-
ture. Therein we employ further the same calculational scheme. That
is, vectors of slow and fast variables can be formed, the largest scale pa-
rameter (see the definitions (4.1.7), (4.1.10)) is separated out, and so
on. We obtain functional operator of the first hierarchical order (for-
mal solutions of the problem) in ‘usual’ form of the Krylov–Bogolyubov
substitutions:
and so forth, where corresponding averaged values are found from cor-
responding reduced (truncated) equations for the first hierarchical level:
Using (6.7.6) together with (6.7.7) and (6.7.8), we obtain averaged

kinetic equation of the first hierarchy:
where new the collision integral is
The corresponding relations for the second hierarchical level are found
in similar way. The next large parameter of hierarchical series (4.1.6)
serves as required scale basis. In a similar way we deal with sequence
of levels until hierarchical series (4.1.6) is exhausted. The correspond-
ing back transformations are accomplished according to the scheme de-
scribed above (see (6.7.9), (6.7.10), and other).
It is interesting to compare both the latter discussed hierarchical ap-
proaches (quasi-hydrodynamic and kinetic) in the beam motion prob-
lem. One can be sure that essential point is in the choice of specific
basis metrics of the problem space. In the quasi-hydrodynamic case
only three-dimensional space coordinates are involved. In the kinetic
case six-dimensional coordinate space is required. Therein the kinetic
approach seems more complicated for use. On the other hand, this ap-
proach has a wider area of application, inasmuch as it is more general.
Their common characteristic is that both cases contain some
set of equations much simpler mathematically than the initial
ones on final stage of calculations.
References
[3] A.G. Sitenko and V.M. Malnev. Principles of the plasma theory. Naukova
Dumka, Kiev, 1994.
REFERENCES 247
Moscow, 1973.
[6] A.F. Alexandrov, L.S. Bogdankevich, A. A. Ruhadze. Principles of plasma elec-


[8] B.E. Zshelezovskii. Electron beam parametric microwave amplifiers. Nauka,

Moscow, 1971.
[9] I.V. Savel’jev. Principles of theoretical physics. Nauka, Moscow, 1977.
[10] E.M. Liftshitz, L.P. Pitayevskiy. Physical kinetics. Nauka, Moscow, 1979.
[11] I.V. Dzedolik, V.V. Kulish. To the nonlinear theory of parametrical resonance
of electromagnetic waves in plasmas of a high-current relativistic electron flux.
Ukrainian Physical Journal, 32(11):1672–1677, 1987.

[13] V.V. Kulish, V.I. Savchenko. Method of averaged characteristics in nonlinear

electrodynamic problems. Gerald of Sumy State University, ser. Physics and
Mathematics, 2:5–12, 2002.
[14] V.V. Kulish, A.V. Lysenko. Method of averaged kinetic equation averaged ki-
netic equation and its use in the nonlinear problems of plasma electrodynamics.
Fizika Plasmy (sov. Plasma Physics), 19(2):216–227, 1993.
particles for electronics applications. the general hierarchical approach. Inter-
national Journal of Infrared and Millimeter Waves, 19(1):3–93, 1993.
[16] V.V. Kulish. Methods of averaging in nonlinear problems of relativistic electro-

[17] V.V. Kulish, V.I. Savchenko. Method of averaged characteristics in the nonlinear
theory of two-stream instability two-stream instability. Gerald of Sumy State
University, ser. Physics and Mathematics, 2:13–18, 2002.
[18] A.I. Olemskoi, A.Ya. Flat. Application of the factual concept in the condensed-
matter physics. Physics-Uspekhy, 163(12):101–104, 1993.
[19] R. Rammal, G. Toulouse, M.A. Virasoro. The gnats and gnus document prepa-
ration system. Reviews of Modern Physics, 50(3):765–788, 1986.
[20] F.G. Tricomi. Lezioni sulle equzioni a derivative partziali. Editrice Gheroni
Torino via Carlo Alberto, 1954.
[21] G.A. Korn, T.W. Korn. Mathematical handbook for scientists and engineers.
NY: McGraw Hill, 1961.
[22] A.M. Samoylenko, S.A. Kryvoshyja, N.A. Perestyuk. Differential equations:

examples and exercises. Vysshaja Shkola, Moscow, 1989.
[23] A.N. Tichonov, A.B. Vasil’jeva, A.G. Sveshnikov. Differential equations. Nauka,
Moscow, 1980.
[24] N.M. Matvejev. Integration methods integration of differential equations.

Vysshaja Shkola, Moscow, 1963.
[25] S.K. Godunov. Equations of mathematical physics. Nauka, Moscow, 1971.
Oxford, 1990.
Chapter 7
EXAMPLE: APPLICATION OF
CHARACTERISTICS IN NONLINEAR
THEORY OF THE TWO-STREAM
INSTABILITY
Again about instabilities and resonances. It is well known that two-
stream instability in electron beams (as well as its specific case — the
plasma-beam instability) is one of the most complex objects for studying
in electrodynamics [1–9]. Two main causes determine such situation.
The first cause is the uniqueness of a physical nature of the two-stream
instability. It should be mentioned that the concept of ‘instability’ in
electrodynamics is often confused erroneously with the concept of the
‘resonance’ (about these concepts see also in Chapter 1, Section 2).
Such situation could be explained by the following: most of the real
oscillation–wave systems, which are traditionally studied in electrody-
namics, are characterized by the presence of instabilities as well as reso-
nances, at the same time. On the other hand, according to the standard
definition, the resonance is usually treated as any steep increase of the
oscillation amplitude, but only in the case if the so called resonant condi-
tion (like in (1.2.13), for instant) is satisfied (see Chapter 1, Section 2 for
more details). This condition, in turn, is satisfied if the frequency of an
external (stimulated) perturbed force coincides with one of the proper
system frequencies (see, for instance, the definitions (1.2.13)). Or in
the other words, the simultaneous presence of external (stimulated) and
proper oscillations is necessary condition for realization of the resonance.
The weakly nonlinear models with harmonic (i.e., one-harmonic) ex-
ternal stimulated oscillations (or waves) are studied in practice most
often. The presence of such harmonic stimulated oscillation (wave)
‘imposes’ an evolving scenario for the resonant process. One of spe-
cific features of the resonant interactions is a relatively narrow resonant
frequency band around of the resonant point vicinity, where this phe-
nomenon appears remarkably (see Fig. 5.2.1b and corresponding com-
249
mentaries, and Chapter 1, Section 2). Roughly speaking, we can treat

such an interaction process as a quasi-harmonic one because the first
harmonic here is explicitly predominated in relevant Fourier spectrum.
One easily may make certain of this expanding corresponding solutions,
that describe such process, in the Fourier–Taylor series. As a result, it
could be found that such types of series, as a rule, are characterized by
a good convergence with respect to the harmonic amplitudes. Gener-
alizing, we can state that the discussed behavior of the Fourier–Taylor
series is typical calculational feature of the resonant problems.
The two-stream, as well as the plasma-beam instabilities, represents
some other kind of examples. Namely, they characterize the systems
where instability takes place, but the resonant conditions are not satis-
fied [4]. Or, in other words, these physical mechanisms can be treated
as instabilities, but not resonances. In contrast to the resonant systems,
any external stimulated oscillations are absent there. The instability
evolves only owing to the interaction between the system proper waves.
Hence, we cannot connect it with any resonant conditions like (1.2.13).
This means, in particular, that no external (stimulated) influences can
be ‘imposed’ at this time, and, consequently, no ‘external quasi-harmonic
scenario’ can occur in the system of the discussed type. As a result of
the absence of this ‘imposing’, in contrast to the resonant systems, the
wave process carry multiharmonical, of principle, nature. A large num-
ber of harmonics might be characterized by amplitudes of approximately
equal order (see further Figs. 8.7.1–8.7.3 and corresponding discussion).
Moreover, the situations when one of the higher harmonics of beam
waves becomes maximal could be realized, too (see further Fig. 8.7.3).
As a result, poor convergence of the Fourier–Taylor series which is con-
structed as a factorization of this process is typical calculational feature
of such kind of instability problems.
Thus summarizing, we can state that resonance is a specific partic-
ular case of instability. Or, that is the same, the concept of instability
includes the resonant as a specific particular case.
Then let us briefly discuss the second type of the above mentioned
causes, which determine the difficulties that arose in study of the two-
stream instability. These difficulties are connected with mathematical
side of the problem. The matter is that most widespread traditional
mathematical methods, such as the method of nonlinear dispersion equa-
tion [1–9], the method of slowly varying amplitudes (see below Chap-
ter 8) etc., in this case turns out to be ineffective very often. Overcoming
of such problems requires creation of a new methodical ideas and calcu-
lational schemes. One of such new approaches was proposed previously
(see Chapter 6) in the form of the method of averaged characteristics.
The demonstration of effectiveness of application of this method to the

two-stream instability nonlinear theory [10–12] is performed below in
this Chapter.
Two-Stream Superheterodyne Free Electron Lasers (TSFELs). In
what follows, let us discuss briefly the applied significance of the two-
stream instability, as a potentially useful practical phenomenon. We
have in view the possibility of constructing of the so called Two-Stream
Superheterodyne Free Electron Lasers (TSFELs) on its basis [13–25].
The TSFELs possess a number of unique properties that make them
very promising for various practical applications. The klystron-TSFEL
amplifiers [18, 20, 21, 23, 24] occupy a special place amongst the devices
of this type.
The simplest variant of scheme of the klystron TSFEL amplifier is
shown in Fig. 7.0.1.
The device works in the following way. Two-beam electron accelerator

1 generates two relativistic electron beams 2 and 3, respectively. Both
these beams are directed on the same input of first pumping system 4.
Later they form joint two-velocity electron beam 5. Electromagnetic sig-
nal (in the form of an amplified electromagnetic wave) is directed
on the same input of first pumping system 4. Two-velocity electron beam
5 is found to be weakly density modulated as a result of the nonlinear
parametric interaction electromagnetic wave with its plasmas. So
electron beam 5 then (within transit section 6) moves being modulated
by some (in the general case) frequency For simplicity the following
is accepted: This means that initial (input) signal in the
first pumping system transforms from the electromagnetic form into the
electron wave one.
The two-stream instability is evolving in electron beam 5 within tran-

sit section 6. Maximum amplification of the electron waves in this case
is achieved for the waves with the optimal frequency [15–25] (here
we note especially that this maximum does not have a resonant nature).
The key point of the discussed design is that the modulation frequency
is approximately equal to the optimal frequency In this case
amplitudes of the electron waves with the frequency increase because
of the two-stream instability. So the input signal exists in tran-
sit section 6 in the form of electron waves with frequency As a result
the depth of the beam modulation strongly increases. At the same time
the initial (input) electromagnetic signal is absorbed by absorber
system 7. Thus only strongly modulated electron beam is directed on
the input of second pumping system 8.
The generation (restoration) of the output electromagnetic signal
occurs within the second pumping system 8. This takes place due
to nonlinear interaction of strongly modulated electron beam 5 with cor-
responding pumping field of the second pumping system 8. This means
that the back transformation of the amplifying signal (from the electron
wave form into the electromagnetic one) realizes within the working bulk
of the second pumping system 8.
The worked-out electron beam is collected by the electron collector 9.
The amplified electromagnetic signal goes from the system output.
Thus the key part of the discussed design of klystron TSFEL am-
plifier is transit section 6, where the main signal amplification occurs.
Therefore, below (in this and the following Chapters) we will analyze
the nonlinear dynamics of the amplification process in this section in
more detail. We will do this in the framework of nonlinear theory of the
two-stream instability. So, let us turn to the calculational aspects of the
two-stream instability nonlinear theory.
1. PROBLEM OF MOTION
OF A TWO-VELOCITY ELECTRON BEAM
IN GIVEN ELECTROMAGNETIC FIELDS
Following the tradition, we divide the initial general self-consistent
problem into two formally independent parts. The first is the problem
about motion of a two-velocity relativistic electron beam in some given
(i.e., known) electromagnetic fields (it is the motion problem). The
second is the problem of generation (excitation) of those fields on a given
motion of the beam (field problem). Self-coordination of both these parts
gives the complete self-consistent solution of the initial problem. Let us
begin our study with the motion problem.
1.1 Statement of the Motion Problem

Let us consider the system of two-velocity relativistic intensive elec-
tron beam as the object of study. The beam model is assumed to be
a transversely unrestricted system. Velocities of the both beam com-
ponents (partial velocities) and are considered relativistic. It is
supposed that the immobile ion background compensates the averaged
spatial charge of the electron beam, as a whole. Systems parameters are
chosen so that all required conditions for achievement of the two-stream
instability are satisfied [7].
We choose relativistic quasi-hydrodynamic equation (6.1.5) as the ba-
sic motion equation for the electron beam that consists of two partial
beams:
where is the velocity vector for the partial beam component

is the effective frequency of electron collisions, is the
electron density for partial beam (here we suppose, for simplicity,
that is the relativistic fac-
tor for electrons of partial beam, is root-mean-square particle
thermal velocity, is the Euler’s (spatial) coordinate of an observation
spatial point, is the light velocity in vacuum, is its rest mass of
particles of the is their charge, is the intensity of electric
and is the induction of magnetic fields acting on electrons of the
beam. We consider that these fields are the same for both partial
components of the two-velocity beam.
Then we take into consideration the solely methodical purpose of this
illustration example. Let us somewhat simplify the considered model.
Namely, we will assume that the electron beam is ‘cold’ (i.e.,
external magnetic fields as well as internal are absent and
electric fields are represented by the beam electron waves and quasi-
stationary electric field only. They can be generated in the system as a
result of nonlinear interactions [6–8].
We take into account the supposition made above about the transverse
unrestrictedness of the model considered. This automatically means that
it is spatially one-dimensional. A simplest analysis shows that in this
case we may assume that all electron motion processes occur along
only:
As a result we obtain the fact that the electric field in the system
is longitudinal only. The intensity of the wave part of this longitudinal
field can be represented in the form of Fourier series:
where are the numbers of space charge wave (SCW) harmonics, N is

the number of harmonics that are taken into account, are slowly
varying (on time as well as on the spatial coordinate z) amplitudes,
is the wave phase of the SCW, is the cyclic frequency,
and is the wane number of the SCW.
Apart from the wave part (7.1.3) of the common electric field the
quasi-stationary electric field which can be generated in the system
as a result nonlinear wave interactions [6–8], exists here too. But a
corresponding analysis shows that we could neglect the influence of this
field on the accepted further precision of solving the motion problem.
Making all the suppositions accepted above we can rewrite motion
equation (7.1.1) in the simplified form:
The boundary condition is the following:
Thus the motion problem is stated, and then we can proceed with the
constructing procedure of asymptotic integration of equation (7.1.4).
1.2 Averaged Characteristics for the Motion

Problem
Let us use the method of averaged characteristics (see Chapter 6)
for this purpose. The constructing of characteristics for the equation
(7.1.4) is the first step on this way (see Fig. 6.2.1). Accomplishing the
passage to the parametric form of writing (where the phase is the
parameter), we obtain the characteristics of the equation (7.1.4) in the
form of standard system like (6.2.4):
Then we introduce the concepts of slow time
and slow coordinate
respectively. Than, accomplishing the following obvious transformations
it is not difficult to get the standard equation system:
where the function (see also the detail discussion of

analogous situation given previously in Chapter 5).
Then we perform the classification of variables into the fast and slow
ones, respectively. For this we should construct the hierarchical series
for the problem large parameters like (6.2.5). It can be easily seen that
in the discussed case only three large parameters of the problem can be
separated:
where
where are some normalizing parameters. However, corre-

sponding analysis shows that both these large parameters are charac-
terized by approximately equal order So,
further we could use only one of them. Let us choose the one that is
smaller:
The passage onto the first hierarchical level is the following step of
our calculational procedure. The Krylov–Bogolyubov substitutions are
used for performing these hierarchical transformations:
where the averaged variables are determined by the equations of the first
hierarchical level:
where All other auxiliary functions are found using

corresponding calculational procedures described earlier in Chapter 4.
Let us over again from discussion of the calculations accomplished in
the first approximation on
where is the averaged relativistic factor.

Then we turn the reader’s attention at the following specific calcula-
tional peculiarity. The choice of the approximation number at a practice
is determined usually by the physical features of a model studied. Let us
explain this thought in more detail using the here considered two-stream
model. The calculations performed in the first approximation allow
us to describe, as relevant analysis shows, the beginning of nonlinear
stage of the two-stream instability. Or, more exactly, we can ‘see’ in
this case the beginning only of a first distinctions from the simplest
quasilinear exponential-like law. The latter are characteristic for wave
amplitude dependencies obtained in the framework of simplest quasilin-
ear theory (see, for instance, [13–25]). However, the most characteristic
nonlinear phenomenon of the two-stream instability (the so called per-
cussive excitation of highest harmonics), as our analysis shows, can not
be described in the first approximation. Accomplishing of correspond-
ing calculations in the second approximation allows to solve this
problem effectively:
Substituting expressions (7.1.16), (7.1.18), (7.1.20) and (7.1.22) in

(7.1.15) we obtain the equations for the first hierarchical level in the
following form:
Regarding the first two equations (7.1.31) as characteristics of some

averaged quasilinear equation like (6.2.15), we can obtain this equations
in the form (see (6.2.15))
Now we should recall that the averaged parameter (see the third
of equations (7.1.31)) here plays the role of a parameter. Besides that,
it is interesting to point out the following. It is easy to be convinced
that, in spite of equation (7.1.32) being formally obtained in the first
approximation only, it turns out to be actually valid for the second
approximation too. This can be explained by that the zeroth re-
sults (7.1.14), (7.1.26), (7.1.28), and (7.1.30) are obtained in the second
approximation. It is obvious that the mentioned conclusion does not
concern the calculational results for oscillation parts of relevant expres-
sions.
Comparing (7.1.32) and (7.1.4), we see that the first of those equa-
tions has essentially simpler mathematical structure (because it contains
no oscillations in the right part). Solving (7.1.32) and performing corre-

sponding back transformation (see, for instant, (6.2.19), (6.2.20)) we can
obtain the complete solution of the initial problem (7.1.4). Let us begin
describing these calculational procedures with the back transformations.

Substituting (7.1.17), (7.1.21) and (7.1.23) in (7.1.14) we obtain the
formal solutions of the initial problem (7.1.4):
Relationships like (6.2.19) in the considered case can be written in the

form:
After accomplishing in (7.1.34) corresponding expansion in the Taylor

series (see (6.2.20)) solutions (7.1.33) could be rewritten as:
or, after corresponding calculational, in the form
where is the averaged relativistic

factor as a function of the averaged velocity
As can easily be seen, the sought complete solution of the initial prob-
lem (7.1.4) could be obtained after the determination of a specific form
of the function As was mentioned above, such expressions could
be obtained by solving of the averaged quasilinear equation (7.1.32).
1.4 Integration of the Averaged Quasi-Linear

Equation for the Beam Velocity
Now let us find solutions of equation (7.1.32)
for the boundary conditions:
where the function is determined in (7.1.5).

We will use the method of characteristics (see Chapter 6, Section 3) for
integration of the equation (7.1.37). Following the general calculational
scheme, we should construct characteristics for equation (7.1.37). The
peculiarity of the discussed situation lies in fact that these characteristics
are already known (see first two equations of (7.1.31)):
The integrals of the system (7.1.39) should be found at the next step
of the used calculational procedure. They can be found easily:
Substitution of boundary condition (7.1.38) in (7.1.40) yields
Excluding all varying values from (7.1.41), we obtain the functional

dependency, which binds together the constants
Then we replace the constants in (7.1.42) with the expressions

(7.1.40). As a result we obtain the sought solution of problem (7.1.37),
(7.1.38):
where are some arbitrary functions (see definitions

(7.1.38), (7.1.5)) of the arguments
As was mentioned above, the complete solutions of initial motion
problem (7.1.4), (7.1.5) can be found after substituting of the solution
(7.1.43) into (7.1.36). For example, let us choose the boundary condi-
tions in the following simplest form
It is obvious that the solution for the averaged velocity in this case
can be obtained as
2. FIELD PROBLEM. HIERARCHICAL

ASYMPTOTIC INTEGRATION OF
THE CONTINUITY EQUATION
The Maxwell equations (6.1.6) are used below as the basic equations
for the field problem. Except for that, continuity equation (6.1.10)
(which, as is well known, follows from the Maxwell equations) will be
used also. Therein the general field problem in the framework of the dis-
cussed theory, in turn, is subdivided into two separate, more particular
problems. They are the problem of continuity of the electron beam and
the field problem, in itself. Let us begin the study of the field problem
from the first one.
2.1 Continuity Equation of

the Two-Velocity Electron Beam
Using relationships (6.1.11), let us reformulate continuity equation
(6.1.10) in the more convenient form:
which we will solve with the following boundary conditions
Here is the concentration of electron plasma of partial beam.

Then we use some result of the preceding Section. Namely, using
solution (7.1.36), we construct the expression for partial derivative in
the right part of (7.2.1):
where the functional dependency is determined by relationship

(7.1.43). Substituting expressions (7.2.3) and (7.1.43) into (7.2.1), we
get an equation with respect to the sought function only. Then

we will use this equation as the basic one. The method of averaged
characteristic is used for finding of its solution.
2.2 Averaged Characteristics and

the Averaged Quasi-Linear Equation
Let us construct the equation system for characteristics — that is the
first step of the used calculational procedure. Separating the oscillation
phase as a parameter, we can rewrite this system in the standard
form like (6.3.2)
Analogously with the preceding motion problem, we introduce further

the concepts of slow time
and slow coordinate
respectively. Accomplishing set of required calculational procedures

(which where described previously repeatedly) we have the set of equa-
tions that can be written in the standard form like
where we have omitted the ‘prime’ symbol for simplicity; the function
Y and other notations are self-evident. Therein we consider
that really the function is known because it can be determined

from relationship (7.2.3).
Below we perform the analogue to the procedure of constructing the
problem large parameters of the preceding problem. As a result it can
be found that two of such large parameters could be constructed:
where
and are corresponding normalizing parameters (the large pa-

rameters must be dimensionless values). The less one of the pa-
rameters is chosen as the problem large parameter:
The asymptotic solutions of the problem (7.2.4) are constructed dur-

ing the next step of the used calculational scheme. We will seek these
solutions in the form of Krylov–Bogolyubov substitutions
where all averaged values are found from the equations of the first hier-
archy:
where
Then we use the standard hierarchy scheme described in Chapter 4
and already applied to the preceding problem. Restricting ourselves to
the first approximation (on only we obtain the following expressions
for the unknown functions in (7.2.11) and (7.2.12):
Analogously we have relevant results for the second approximation

Substituting (7.2.13), (7.2.16), and (7.2.19) into system (7.2.12), we

can rewrite the latter in the form:
Regarding equations (7.2.28) as characteristics equations of a cor-

responding averaged quasilinear equation with partial derivatives like
(6.2.15) (quasilinear equation of the first hierarchical level), we can eas-
ily construct such an equation:
Equation (7.2.29) can be solved by means of one of the known ana-

lytical or numerical methods. An example of such scheme of solution
will be given somewhat below. Here let us turn to the problem of back
transformation.
2.3 Back Transformation

Substituting the expressions (7.2.14), (7.2.18), (7.2.20) into Krylov–
Bogolyubov substitutions (7.2.11), we can construct the formal solutions
for the sought complete solutions of the initial problem:
Below we use the algorithm of back transformations described previ-

ously in Chapter 6, Subsection 4.2. First we turn our attention on the
fact that the differences
are proportional to (see (6.2.19) and corresponding comments). Us-

ing this, we perform corresponding expansion in the Taylor series (see
(6.2.20):
Taking into account the above-obtained result, after not very difficult
transformations we can rewrite expression (7.2.2) in the following form:
where is the averaged concentration of the electron plasma of

beam, it can be calculated from equation (7.2.29). Obtained here
expression (7.2.33) can be regarded as an approximate solution of initial
continuity problem (7.1.1). However, corresponding solution for the av-
eraged concentration should be found after eventual solving of
the considered field problem. We have in view the above mentioned elec-
tric field problem. The latter will be discussed somewhat below. Here,
we turn our attention to the solving of the averaged quasilinear equation
(7.2.29).
2.4 Characteristics of
the Averaged Continuity Equation
So, let us come back to averaged quasilinear equation (7.2.29)
which we will solve for the following boundary conditions:

The characteristics of equation (7.2.34) are
Solving equations (7.2.36) by standard means we can find the sought

solutions of the averaged quasilinear equation (7.2.29), (7.2.34).
3. FIELD PROBLEM. APPLICATION OF

CHARACTERISTICS FOR ASYMPTOTIC
INTEGRATION OF
THE MAXWELL’s EQUATIONS
3.1 Averaged Maxwell’s Equations
Taking into consideration one-dimension spatial nature of the con-
sidered model (see above Subsection 1.1) we can rewrite the Maxwell
equations in the following simplified form:
As well known in the electrodynamics, any of two last equations can

be chosen for calculational in the considered case. It is no difficult be
convinced that the other one always can be constructed by using the
continuity equation. Taking this into account, let us use further the
only equation
instead system (7.3.1). Here the functional expressions and

are found in the two preceding Subsections. We will solve
equation (7.3.2) for the boundary conditions
As before, the method of averaged characteristics for the asymptotic

integration of (7.3.2) is chosen as the basic one.
We regard (7.3.1) as an equation of the zero hierarchical level. Con-

structing a corresponding system for the characteristics, we perform the
passage to the parametrical form of the writing. For this we separate out
the fast oscillatative phase and construct a corresponding equation
for it. As a result we obtain:
Similarly with the preceding cases, the concepts of slow time
and slow coordinate
are introduced. Accomplishing required set of transformations, it is no

difficult to obtain:
Here as before we have neglected the double prime symbol for sim-
plicity.
As we did before, we separate out the large parameters of the problem:
where
where are corresponding normalizing factors. A smaller one of

the large parameters is chosen as the problem’s large parameter:
According to the general scheme of the method of averaged charac-

teristics, we seek the solutions of equations (7.3.4) in the form
where for determination of the averaged values we have the system of

equations for the first hierarchical level like
Here, let us recall once more that averaged spaces in this and the
previous problems are different. This means that the values and
differ from the analogous values in the preceding problems.
As before, we restrict ourselves by keeping only terms in calcu-
lational of the auxiliary functions in expressions (7.3.11), (7.3.12):
Substituting (7.3.13), (7.3.15) and (7.3.17) into (7.3.12) we obtain the

system of equations for the first hierarchical level in more specific form:
Then we regard these equations as characteristic equation of some

averaged quasilinear equation with partial derivatives:
that is equation with partial derivatives for the first hierarchical level.
We can treat equation (7.3.25) as the averaged Maxwell’s equation.
It should be mentioned that the values and in (7.3.25) are some
known functions in the averaged space of the considered field problem.
It should be emphasized that, in contrast to the situation which takes
place in the previous cases, the values and in the considered case
can be essentially different from the analogous values obtained in the
preceding Subsections.
The calculational scheme of solution of equation (7.3.25) will be pre-
sented somewhat later. Now let us turn to the problem of back trans-
formations.

We will follow the calculational schemes of back transformations that
were described more than once in this and the previous Chapters (see,
for example, (6.2.17)–(6.2.20), and corresponding comments). Namely,
we rewrite the Krylov–Bogolyubov substitutions in the form
and then the estimation of the differences could be represented as
Using the circumstance that differences (7.3.27) are proportional to

we accomplish corresponding expansion in the Taylor series in
(7.3.26). In what follows we take into consideration that
and As a result we can write:
or, after some transformation, we obtain the sought solutions for the
initial problem (7.3.2), which we can represent in the form like
3.3 Solving the Averaged Quasilinear

Equation for the Electric Field
Solutions for the averaged electric field can be found from
equation of the first hierarchical level (7.3.25):
for the boundary conditions
Let us use the method of characteristics (see previously Chapter 6,

Section 3) for the solving of the equation (7.3.30).
According to the scheme of the characteristics method, the system for
characteristics should formulated on its first step:
The first integrals of system (7.3.32) could be found in the form:
where
The constants could be found after substituting boundary con-

ditions (7.3.31) into (7.3.33):
Substitution of results (7.3.33) into (7.3.35) yields
where and are some arbitrary functions (of the stated argu-
ments) that do not contradict to the definition (7.3.32) and bound-
ary condition (7.3.31). Substitution of any of expressions (7.3.37) into
(7.3.29) solves the stated initial problem (7.3.1). However, obtained this
way solutions possess non-evident form. We should then reduce a non-
evident solutions like (7.3.29) to evident form (7.1.3). Corresponding
expressions for the harmonics in (7.1.3) for the beam space
charge waves should be formulated for this.
3.4 Truncated Equations for the Harmonic

Amplitudes of a Space Charge Wave
Let us turn to the problem of determining the complex amplitudes
We equalize for this definition for the electric wave field (7.1.3)
and the obtained result like (7.3.29). Then, we equalize the same coef-
ficients for equal exponents. As a result, the following equations could
be obtained:
where the expression designates the pro-

cedure of separation out the harmonic. We confine ourselves for
simplicity only accounting the nonlinear terms no higher second order.
The further generalization of obtained result (7.3.38) could be made
by simple keeping all higher nonlinearities, which appear in the second
approximation (i.e., on the lengths and time intervals
— see the following Subsection for more details).
System (7.3.38) could be analyzed simpler if we use the linear disper-
sion relationship for the two-stream instability. Let use rewrite for this
expressions (7.3.38) for the first harmonic in the linear approximation:
where is the plasma frequency for partial

beam, as before, is its relativistic factor. It is
readily seen from (7.3.39) that nontrivial solution in this case could be
got if the following condition is satisfied:
where This is the sought linear dispersion equation

[11–25] for the relativistic two-stream system. The correlation like
could be found from (7.3.40) (see Chapter 13, Volume II for more
details).
Let us shortly discuss dynamics of the harmonic amplitudes in the
case, if we neglect the influence of the quasi-stationary longitudinal elec-
tric field This means that we can neglect in (7.3.38) by the terms
proportional as well as the terms (because they
are also proportional
Then, accounting (7.3.39) and (7.3.40), we yield the following equation

system from (7.3.38)
System (7.3.42) describes the nonlinear dynamics of the harmonic

amplitudes. Analyzing its mathematical construction, we might be
surprised by their resemblance with corresponding results like (8.2.16)
which can be obtained in the framework of the method of slowly varying
amplitudes (see further Chapter 8). This means that we, in principle,
can regard the proposed calculational technique as a new specific version
of the slowly varying amplitude method. As relevant comparative anal-
ysis shows, the proposed new version has a number of very important
(from practical point of view) advantages. First of all, because it allows
to obtain the equations like to (7.3.42) for arbitrary harmonic number
The analogous procedure accomplished in framework of the traditional
slowly varying amplitude method turns out to be too labor-intensive in
such situation. Moreover, it cannot be really realized generally quite
often.
3.5 Some Commentaries for the Obtained Results

Let us discuss the above mentioned possibility of accounting formally
an unlimited number of the electron harmonics in the framework of the
method considered. Saying about this, we, however, have in view a
potential possibility only. The proposed algorithm indeed has no limi-
tations, in principle, on the number of harmonics considered. But this
does not means that we really should always deal with unrestricted num-
ber of harmonics. The matter is that the obtained solutions like (7.3.38)
possess the approximate nature. I.e., they are obtained with some given
precision, which, in our case, is determined by an accepted supposition
about the number of approximation (in the terms of the small param-
eter In this Chapter we have used the second approximation on
This circumstance determines the real number of harmonics, which
accounting makes sense for the accepted accuracy. It is necessary to in-
crease the calculational precision in the case if we would consider more
harmonics. For instance, to accomplish the analogous calculations, but
in the next approximation on However, let us discuss this topic in
more detail.
As was mentioned before, the peculiarity of the discussed type of
methods (see, for example, Chapter 4) is that the precision of solutions
which should be obtained, is determined beforehand. This means that
if we want to accomplish calculations in some approximation, the
method used guarantee the given precision for some first harmonics
only. This number of harmonics can be obtained from the following
obvious condition:
The precision of calculational of the harmonics cannot be

guaranteed in this case. Or, in other words, the precision of calcula-
tional of all higher harmonics can become lower than the pre-
cision that the given approximation can guarantee. Hence if we would
like to increase the number of considered harmonics we must increase
the number of approximation So really the number of harmonics
considered is not unlimited in the framework of the method of averaged
characteristics. But let us stress once more that this limitation is con-
nected with the precision of the problem only, and it has nothing to do
with specific method limitations.
Besides the limitations discussed above on the number of considered
harmonics, some others limitations exist in the considered theory. They
are the limitations on the electron pulse duration and maximal
length of the system respectively. These limitations arise from the
characteristic feature of the used hierarchical asymptotic methods (see,
for example, Chapter 4 for more details). Namely, they originate from
the fundamental theorems like the Bogolyubov’s theorem (see Chapter4,
Subsection 2.3). According to the latter, we can guarantee the given
precision of solutions for some limited time and some limited length
only:
REFERENCES 281
where are some normalization constants. This means that results

obtained, for instance, for the case are valid only for the time
and the length
respectively. Taking into account limitations similar (7.3.43)–(7.3.45) is

very important for a practice, especially in the cases when a numerical
analysis is accomplished.
References
[1] A. A. Ruhadze, L.S. Bogdankevich, S.E. Rosinkii, V.G. Ruhlin. Physics of high-
Kiev, 1994.
Moscow, 1973.
[7] A.F. Alexandrov, L.S. Bogdankevich, A.A. Ruhadze. Principles of plasma elec-
[9] B.E. Zshelezovskii. Electron beam parametric microwave amplifiers. Nauka,
Moscow, 1971.
[10] V.V. Kulish, V.I. Savchenko. Method of averaged characteristics in the nonlinear
theory of two-stream instability two-stream instability. Gerald of Sumy State
University, ser. Physics and Mathematics, 4, 2002.
[11] V.V. Kulish. Methods of averaging in nonlinear problems of relativistic electro-

[13] N.Ja. Kotzarenko, V.V. Kulish. About the effect of superheterodyne ampli-
fication electromagnetic waves in the plasma-beam system. Radiotechnika i
Electronika (Sov.), 25(11):2470–2471, 1980.
[14] Perekupko V.A., Silivra A.A., Kotzarenko N.Ja., V.V. Kulish. Patent of USSR
No. 8335259, 19. Priority of 28.01.80.
[15] G. Bekefi, K.D. Jacobs. Two-stream fels. J. Appl. Phys., 53:4113–4121, 1982.
[16] O.N. Bolonin, V.V. Kulish, B.P. Pugachev. Superheterodyne amplification

of electromagnetic waves in the system of two relativistic electron beams.
[17] M. Botton, A. Ron. Two-stream instability in fels. IEEE Trans. of Plasma

Sciences, 18(3):416–423, 1990.
[18] V.V. Kulish, V.E. Storizshko. Free electron laser. Patent of USSR No. 1809934.
Priority of 18.07.90.
[19] V.V. Kulish, B.P. Pugachev. To the theory of effect superheterodyne amplifica-
tion waves in plasma of two-stream system. Fizyka Plazmy (Sov.), 17(6):96–705,
1991.
[20] V.V. Kulish. To the theory of two-stream free electron lasers of klystron type.
[21] V.V. Kulish. To the theory of relativistic electron-wave free electron lasers.
[22] H. Wilhelmsson. Double beam free electron laser. Physica Scripta, 44:603–605,
1991.
[23] V.V. Kulish. Physics of two-stream free electron lasers. Gerald of Moscow State
University, ser. Physics and Astronomy, 33(3):74–78, 1992.
[24] V.V. Kulish, V.E. Storizshko. Free electron laser. Patent of USSR No. 1837722.
Priority of 13.10.92.
[25] V.V. Kulish. Superheterodyne electron-wave free electron laser. The Interna-
tional Journal of Infrared and Millimeter Waves, 14(3), 1993.
Chapter 8
WITH PARTIAL DERIVATIVES.
SOME OTHER ASYMPTOTIC METHODS
As follows from the materials of the two previous Chapters, the hier-
archical asymptotic methods, like the method of averaged characteris-
tics, could be quite effective for solving nonlinear electrodynamic wave
problems. At the same time, we have a possibility of seeing that these
methods have some limitations with respect to their areas of application.
In particular, it had been made clear that essential difficulties appear in
the case of asymptotic integration of systems with number of variables
more than three, etc.. As had been found, these limitations, predomi-
nantly, are determined by using some specific ‘improving’ of the method
of characteristics, which is used here as one of the key elements (see
Chapter 6, Section 3 for more details).
However, we note that the method of averaged characteristics is not
the only method of the class discussed. There is a number of other
methods which can also be effective, especially in the situations [1–12],
when the method of averaged characteristics is not expedient. Some of
them (such as the method of slowly varying amplitudes [1–5, 7–9]) are
well known and widespread in practical research. Others (such as the
Mitropol’skii method [10,11] or the method of hierarchical transformation
of coordinates [12]) are not as popular, although they can also be quite
useful in some specific calculational situations.
It should be mentioned that the traditional version of the method
of slowly varying amplitudes is found to be suitable only for special
calculational situations when the right-hand sides of some considered
equations are described by some periodic and quasi-harmonic functions
[11–15]. The systems of such a type are called Rabinovich standard sys-
tems. On the other hand, in studying the plasma-like systems researchers
very often face specific physical phenomena which are found to be far
283
behind the mentioned standard situation. For instance, they are the
nonlinear generation of quasi-stationary fields and currents (which are
non-periodic), the degeneration mechanisms of wave resonant interac-
tions, when the same resonance condition is satisfied for more than one
couple of waves, etc. [16–22]. This means that the method of slowly
varying amplitudes should be essentially improved (modernized) in the
case of its application for the electrodynamic nonlinear plasma-like wave
resonant problems. The main purpose of such modernization must be
relevant extending the application area for the mentioned standard ver-
sion.
The goal of this Chapter is to give for the readers some general infor-
mation about some of these ‘other methods’. Below, the main attention
is paid to the method of slowly varying amplitudes (including its so called
modernized version). Besides that, we will discuss some general ideas of
the Mitropol’skii method and the method of hierarchical transformation
of coordinates.
1. MAIN IDEAS OF THE METHOD OF

SLOWLY VARYING AMPLITUDES
The method of slowly varying amplitudes was been first proposed
and used by M.A. Leontovich [13]. The method was applied for solving
the known problem concerning nonlinear evolution of an electromag-
netic wave propagating in the Earth atmosphere. The main idea of the
proposed method is, in principle, analogous to the Van der Pol method
(see in Chapter 3, Section 4 for more details). Namely, the wave ampli-
tude owing to the weak nonlinearity of considered models should change
slowly (see the concept of weak nonlinearity in Subsection 2.1, Chap-
ter 1). Or, in other words, the wave amplitude in such a situation clearly
demonstrates properties of a slowly varying value. This allows us to re-
duce the initial problem of the wave propagation to the problem of the
slow amplitude evolution. In fact, the latter ‘slow’ problem is essentially
simpler than the initial ‘fast oscillatative’ problem. Owing to this and
to the simplicity of the discussed calculational idea in itself the method
of slowly varying amplitudes became rather popular in electrodynamics,
radiophysics, nonlinear optics, etc. [4–9]. This original (simple) version
of the discussed method is called the simplified version of slowly varying
amplitude method.
Further generalization and improvement of the method was accom-
plished by a number of authors. However, it should be mentioned that
A.V. Gaponov-Grehov, M.I. Rabinovich, and L.A. Ostrovsky [1–3] have
achieved most remarkable successes in this field. As a result of their
efforts the method acquired the required level of validity and strictness.
SOME ANOTHER ASYMPTOTIC METHODS 285
This ‘improved version’ of the method was named the rigorous version
of the method of slowly varying amplitudes.
Unfortunately, the essential complication of calculational procedures
turned out to be an unpleasant price for achieving the rigor and versatil-
ity mentioned. On the other hand, there are a lot of practical situations
in which these rigor and versatility are not required. That is why below
we will discuss both these versions.
1.1 General Calculational Scheme of

the Slowly Varying Amplitudes Method
The main calculational idea of the method of slowly varying ampli-
tudes is illustrated in Fig. 8.1.1. Analogously with the method of aver-
aged characteristics (see Fig. 6.2.1 and relevant commentaries) we can
separate out the three following stages of the asymptotic integration of
the equations like (7.1.1) (see Fig. 8.1.1). The first is the straight trans-
formation. The goal of this stage is the transformation of an initial
equation with partial derivatives (zeroth hierarchical level) into the so
called truncated equation for slowly varying amplitudes (the first hierar-
chical level). The second stage is obtaining the solutions of the truncated
equation. The third stage is constructing the solutions for the initial (ze-
roth) hierarchical level (on the basis of the solutions found earlier for the
first hierarchical level). Let us discuss some calculational peculiarities
of these stages in more detail. Such specific calculational technologies,
however, will be described somewhat below. Here we restrict ourselves
by discussions of the method general ideas only.
So we begin with the first stage of the calculational scheme considered.
Let us choose some dynamical wave system described by the following
differential matrix equation (see definition (3.3.1)):
which was previously called the general standard form (for the method
of slowly varying amplitudes).
Here A, B, C are square matrices of size is some
vector-function in Euclidean space with coordinates
i.e.
P is some linear differential operator in the space
R(...) is a given weak nonlinear vector-function, is some scalar vari-
able (for instance, the laboratory time). As a result of the straight
transformations we performed some equation for the first hierarchical
level (i.e., truncated equation) like the following
should be obtained. Here superscript (1) denote that this value belongs
to the first hierarchical level.
The main idea of the method discussed, as mentioned above, consists

in using the main property of the weak nonlinear wave systems (see
Chapter 1, Subsections 2.1 and 2.4). Namely, the dynamics of the weak
nonlinear wave system (8.1.1) is almost identical to the dynamics of
some equivalent linear system
over a few wave oscillations. Some remarkable differences in behavior of

both system (linear and weak nonlinear ones) appear after occurrence of
many wave oscillations in the system. This means that in principle, we
could choose basis solution (generated form of solution)
for description of the system nonlinear dynamics, which is obtained for

some equivalent linear system like (8.1.3). Here is
the complex amplitude of the s-th wave (see the concept of complex am-
plitudes in Chapter 1, Subsection 2.4), is the cyclic frequency, and
is the wave vector of the wave; is the vector of spatial co-
ordinate. Differences between these solutions lies only in the following:
the values, which are constants in the linear case (8.1.4), become slowly
varying in the corresponding nonlinear solutions. As can be easily seen,
we have two such constants that characterize the linear wave process
(8.1.4). Namely, they are the real amplitude and the initial phase
respectively (see Chapter 1, Subsection 2.4). It is important to note
that these values play the role of slowly varying real amplitudes and
slowly varying initial phases in the discussed method. Thus the concept
of complex amplitude (see Chapter 1, Subsection 2.4)
can also be used for description of the nonlinear system dynamics. In
connection with this we, to be exact, talk about the method of slowly
varying complex amplitudes. But, in fact, further we will use the short-
ened version of this term: the method of slowly varying amplitudes.
In what follows we should additionally apply some resonant condition
in the case of the wave resonant system.
Substituting solution (8.1.4) and assuming the amplitudes to be the
slowly varying functions
we can obtain the so called truncated equation for amplitudes (equation

for the first hierarchical level) like (8.1.2). Thus we consider that
Thus the first stage of the considered calculational scheme, in turn,

consists of the following five successive steps (substages):
a) constructing an equivalent linear equation (comparison equation);
b) solving of this linear equation;
c) formulation of the resonant conditions (in the case the system is
resonant wave one);
d) constructing of a form of the sought nonlinear solutions of the
initial problem on the basis of obtained linear solutions;
e) forming an equation for the slowly varying amplitudes.
The solving of the obtained truncated equation for slowly varying
amplitudes like (8.1.2) is the second of the three above mentioned stages
(see Fig. 8.1.1). Using of any suitable solution method, including the
method of characteristics (see Chapter 6, Section 3), can be utilized for

this.
The third stage is accomplishing of the back transformations. In con-
trast to the methods, similar to the method of averaged characteristic
(see Chapter 6), this calculational procedure is found here to be rather
simple and obvious. Indeed, the solutions of the initial problem (8.1.1)
can easily be obtained using known solutions for the slowly varying am-
plitudes by simple substitution of these solutions in the linear (quasilin-
ear ) solutions like (8.1.4).
In the following let us illustrate the calculational scheme described
above on the simplest example of the problem of parametric amplification
of a wave [4, 5]. Therein later in this Section we will restrict ourselves
by studying of the simplified version of the slowly varying amplitude
method.
1.2 Simplified Version of the Slowly Varying

Amplitude Method. Example: Effect of
Parametric Amplification of a Wave
The model. We choose the simplest quadratic-nonlinear spatially one-
dimensional model for illustration of the most characteristic features of
the slowly varying amplitudes method. It is supposed that two waves
of arbitrary physical nature are propagated along Let us assume
that the nonlinear dynamics of the first of them (one that is characterized
by some dimensionless value ) can be described by the equation
It can be easily seen that (8.1.7) is the simplest case of the standard
form (8.1.1). Here and are dimensionless time, dimensionless spa-
tial coordinate, and dimensionless phase velocity, respectively. Further
we will omit the prime in the values and for the sake of simplicity.
The second wave that is characterized by the value plays a role
of the pumping wave. The amplitude of this wave is considered to be
much greater then the first wave (so called approximation
of strong pumping [5, 7, 19]). We can suppose that, in contrast with
the first wave, the amplitude of the second wave could be considered
approximately constant. So we can consider the field of the pumping
wave as a given one. The functional dependency we choose in
the form
where all definitions are obvious considering (8.1.4).
The first stage of calculational procedure. According was had

been set forth in the preceding Subsection, we should pass to the first
hierarchical level (see Fig. 8.1.1) at the first stage of the described cal-
culational algorithm. In turn, the constructing of the equivalent linear
equation like (8.1.3)
is the first step of the first stage (first substage). We will seek for the
solution of (8.1.9) in the form (generating form of the solution)
where the dispersion law (see Chapter 1, Subsection 3.6 and below in
Chapter 8, Section 2) could be found after substitution of (8.1.10) into
(8.1.9). After trivial calculational this law could be easily found:
Below we consider that the relevant solution of the initial nonlinear

problem (8.1.7) should be sought in a form analogous to (8.1.10). The
difference is only in the fact that the complex amplitude in this case
turns out to be slowly varying value relative to the variables and
Let us simplify the model even more by supposition that it is spatially
homogeneous with respect to wave amplitudes and Then the
basic form of solution for the nonlinear problem can be written
where is the unknown (sought) slowly varying amplitude as a

function on time Corresponding truncated equation like (8.1.2) should
be constructed for determining of the slowly varying amplitude
Let us construct such equation.
Substitution of solution (8.1.11) into the left side of initial equation
(8.1.7) yields:
This means that expression (8.1.7) can be rewritten in the form
Two types of solutions for the equation (8.1.13) can be found there.
The first solutions are non-resonant ones. They are not interesting for
our study. Therefore, here we omit its analysis. The second are resonant
solutions. They are essentially more interesting as well from the physical
and calculational points of view. So let us discuss this case in more detail.
The resonant state of the system in the case of Lagrange description
can be determined (see introduction for this Chapter and Chapter 1,
Subsection 2.3) as a closeness of velocities of oscillatative phases change
for proper and stimulated waves. One can easily see that this definition
is equivalent to the closeness of phases of proper and stimulated waves in
the case of Euler description (because variables and are independent
values here). In our case the proper wave is connected with the phase
In turn, the stimulated waves are described by the terms in the right side
of expression (8.1.13). As can be easily seen, four combinational waves
with the phases that correspond to four stimulated waves,
can exist in the considered system. According to the given definition,
the resonant state of the system could be reached in the case when the
exponent exp in the left side of equation (8.1.13) is equal to one of
the combinative exponents exp in the right side. Simplest
analysis shows that interaction of two of the stimulated waves with the
phases only with the proper wave can have the resonant
nature. Thus the two remaining waves interact in the non-resonant man-
ner. Generalizing, we can say that both mentioned resonant conditions
could be written in the following form:
where is the sign function. Some another form of writing of the

resonant condition (8.1.14) is also used traditionally in literature (see,
for instance, [4–13]):
The expressions like (8.1.15) are called the condition of parametric

resonance (see the classification of resonances in Chapter 4, Subsec-
tion 1.3).
Below we take into consideration that in the case of occurrence of the
resonance we can neglect all other non-resonant terms in the right side
of equation (8.1.13). The explanation is quite simple. However, let us
discuss this topic in more detail. It is well known [4–9,19] that one of the
conditions like (8.1.15) usually corresponds to increase of amplitude
(on time in the considered case). This means that the non-resonant
amplitudes appear inevitably in the situation when the amplitude of this
resonant wave becomes essentially higher. So, in such case we can indeed
neglect the influence of all non-resonant waves comparing with resonant
waves.
Taking the latter into account and equating the expressions with the
same (resonant) exponents in the left and right sides of the equation
(8.1.13), we can formulate the sought equation for the slowly varying
amplitude
where are the Kronecker symbols.

Thus the main task of the first stage of the discussed calculational
procedure is fulfilled. So we can pass to the second stage (see Fig. 8.1.1)
that is finding the solutions of equation (8.1.16).
The second stage of the calculational procedure. Let us pass to

the real form of writing of equation (8.1.8). For this we use the definitions
of the amplitudes as complex values:
where Substituting (8.1.17) into (8.1.16) and separating the

real and imaginary parts in (8.1.16), we can obtain the following system
of real equations:
where is the phase mismatch. We divide the first of

equations (8.1.18) by
Then by carrying out some obvious transformations we reduce it to

the form
In the following we turn our attention to the second of equa-

tions (8.1.18). We multiply both parts of this equation by the value
and after simple transformations it can be easily obtained
that
Addition of equations (8.1.20) and (8.1.21) allows us to find the fol-

lowing system integral:
or, in the more acceptable form,
The integral (8.1.23) says that multiplication of the functions and

is constant for any instant of time, including the initial instant
It can easily be seen that in the case
where the constant for any time It is inter-

esting that, as analysis show, the effect of amplification of the first wave
(8.1.11) can be maximal namely in the case (8.1.24). Therefore, further
we restrict ourselves by the studying case (8.1.24) only. Accepting the
supposition we can rewrite system (8.1.18) as
because always (see (8.1.23)). Equations (8.1.25) can be easily

integrated:
As follows from (8.1.26), parametric resonant amplification takes place

for Correspondingly, we obtain the parametric resonant damp-
ing for the case
According to the ideology accepted in this book, solutions (8.1.25)
can be regarded as the solutions for the first hierarchical level. We must
pass to the initial (zero hierarchical level — see Fig. 8.1.1) in order to
obtain the solutions in the initial form. So, let us turn to the problem
of back transformations.
The third stage of the calculational procedure. The third stage of

the calculational procedure (back transformation) could be accomplished
very simply. We found the sought nonlinear solution for the zeroth
hierarchical level after substitution of solutions (8.1.26) into (8.1.17)
and then into (8.1.11):
Thus the problem stated above of the effect of parametric amplifica-

tion of a signal wave in some weak nonlinear non-dispersion medium is
solved.
2. TRADITIONAL VARIANT OF
THE SLOWLY VARYING AMPLITUDES
METHOD. RIGOROUS VERSION
As was mentioned above, the traditional version of the slowly varying
amplitude method is intended for application in studying weak nonlinear
wave resonant problems that occur in arbitrary distributed systems of
different physical nature. Thus traditionally it is supposed only wave
type processes realize. Standard system which describes such situation
can be represented by the second of equations (3.3.21) for In
the alternative case, we talk about the modernized version of the method
of slowly varying amplitudes discussed below in Section 3.
Then let us turn the reader attention that the simplified version of the
method of slowly varying amplitudes, discussed above in the preceding
Section, sometimes turns out to be too rough for practical applications.
Firstly, because the proposed calculational procedures of the straight

and back transformations (see Fig. 8.1.1) have not any satisfactory sub-
stantiation. This procedure can be effective, as a practice show. As
a rule such situation is characteristic in the cases when we have to do
with lowest nonlinear approximations (quadratic, as it takes place in the
above described example) [4, 5, 7]. In other cases we risk to lose some
terms in right parts of the truncated equations for amplitudes.
Secondly, because the simplified scheme very often does not give sat-
isfactory answer about the methods of analysis of systems with different
velocities of slow changing of amplitudes (and, sometimes, the medium
parameters) along the different spatial coordinates. This is important
inasmuch as such physical situations are not rare for real practice [4].
The rigorous version of the method of slowly varying amplitudes is in-
tended for namely such situations. Besides that, it allows more confi-
dently to perform the procedures of straight and back transformation.
2.1 Case of Spatially One-Dimensional Model

Let us start with the studying of the spatially one-dimensional model.
The standard system (8.1.1) is chosen as initial one. We consider trans-
versely unbounded homogeneous (in the plane XY) model in which par-
tial plane waves propagate along the In this case, all space-
coordinate dependencies of the vector U components can be written
as i.e., the model is spatially one-dimensional. Operator P in
(8.1.1) we choose in the simplest differential form
We neglect effects such as ‘nonlinear generation of the quasi-stationary

fields’ (see Chapters 12, 13, Volume II). Formally, this is equivalently
to neglecting slowly quasi-stationary varying part of the vector U in
(3.3.21), i.e., Assuming (which means that the wavy
fields are presented in the model only), one-dimensional version of initial
standard equation can be written in the so called Rabinovich’s standard
form [1–3]:
Here, besides the laboratory ‘ordinary’ time and the ‘ordinary’ co-
ordinate the slow time and slow coordinate are introduced addi-
tionally, i.e.,
Equation (8.2.2) can be represented in the scalar form
The next important step of the method of slowly varying amplitudes

(see above Section 1) is the procedure of forming the sets of slow and
fast variables. The so called generating form of solution (see (8.1.4)) can
be obtained from the equivalent linear system (see (8.1.3)):
We write the solution to equation (8.2.5) as The

system (8.2.5) has nontrivial solutions if the condition
holds. This formula is called the linear dispersion equation determining

linear dispersion properties of the system. The function is the
dispersion function. Solution of the equation (8.2.6) allows finding the
law of dispersion for the relevant waves of the system (see
definitions (1.3.9)–(1.3.11) and relevant commentaries). The waves de-
scribed by condition (8.2.6) relate to proper waves of the system. The
main characteristic of these waves is that they propagate through the
system after turning off the wave source. Electromagnetic waves in the
vacuum or some non-conductive medium, waves on water surface, acous-
tic waves of different types, etc., can serve as examples. Apart from the
proper waves in the system, as mentioned above in Section 2 there exist
stimulated (induced) waves, too. These waves are excited in every point
of the system by some distributed force. Properties of such waves are
completely determined by characteristic features of the acting force. In
contrast with the proper waves, the stimulated ones vanish after turn-
ing off the source, i.e., the stimulated force. The condition existing of
improper waves can be formulated as an alternative to (8.2.6).
i.e., cyclic frequency and wave number are independent quantities

for such wavy objects.
Using (8.2.6) solution for linear system (8.2.5) can be written as (see
above (8.1.4))
here are amplitudes to be found. Substituting (8.2.8) in (8.2.5) yield

a system of algebraic equations
where We construct the solution to the

system (8.2.9) in terms of Kramer formula with first equations
being linearly independent, i.e.,
The determinant
is obtained substituting the column of free terms for col-

umn of the matrix of the order Here is minor
of the element in the determinant i.e.
is the algebraic complement of the element in the determi-

nant
We construct solutions of initial standard system (8.2.2), (8.2.4) in
terms of (8.2.8), regarding the complex amplitudes as slowly varying
functions, i.e., as the function of second hierarchical level. Therefore, the
next stage of hierarchical procedure is constructing relevant dynamical
operators connecting dynamical variables of the first and the second
hierarchical levels. For this we use linear solutions (8.2.8) as the initial
basis. Besides that, we must define the wave resonance and account it.
Here we remind that nonlinear resonant processes are main causes of
slowly varying of dynamical variables at second hierarchical level (see
Chapter 1, Subsection 2.3).
Suppose proper waves (8.2.8) (here is finite) interact with the

wave of stimulated (induced) oscillations. We impose the wave resonant
condition for one of the stimulated wave only given by non-linearity of
the medium and proper waves (see illustration example (8.1.15) and
corresponding commentaries), i.e.,
where and are the frequency and the wave number of stimulated
wave, Within (8.2.7) resonant wave condition (i.e.,
closeness of the frequencies and wave numbers of proper and stimulated
waves) can be given as
The condition (8.2.14) holds only for a finite number of waves since
the system always has dispersion.
For the above reason we look for the solution of (8.2.4) for the slowly
varying amplitudes generally as
where is slowly varying complex amplitude of

wave normalized by Within above concept of functional hierar-
chical operator (2.2.7) we consider (8.2.15) as definition of the oper-
ator for the wave resonant problem (8.2.2), (8.2.4). The next
stage of hierarchical calculational is determining unknown variables
of the first hierarchical level (i.e., they are slowly varying functions
and For this purpose we use hierarchical re-
semblance principle, i.e., we construct the dynamical equation of the first
hierarchical level in the similar (in mathematical structure) to equation
(8.2.4) form (see also (8.1.2) and corresponding commentaries):
where are relevant elements of matrices of the first hier-

archical level, is the corresponding nonlinear functions of the first
hierarchical level. Then it is convenient to normalize equation (8.2.16)
with respect to matrix element in the way that this equation is
reduced to the form
where are nonlinear partial differential operators, yet unknown. Here

we took into consideration the definitions (8.2.3). Thus asymptotic solu-
tions to (8.2.16) reduce to expressions for unknown functions and
operators In order to find we substitute (8.2.15) into initial
system (8.2.4) and use (8.2.5), (8.2.17). Equating the coefficients near
the same powers of in the left- and right-hand parts yields a system of
interrelated equations
where
and so on, and is determined by (8.2.15) for Since is

periodic, the expressions for can be expanded in the Fourier series,
i.e.,
where are the phases. The first group of addends

relates to the proper waves of the system. The second sum originates
from the non-linearity of the medium and external wave fields that are
not synchronous with those of the first group (see (8.2.7)). For these we
have
The functions are determined by Fourier coefficients
The functions can be written as
We substitute (8.2.20) and (8.2.23) into (8.2.18), equalize coefficients

near the same exponential functions find the system of inhomogeneous
equations for
The dispersion function does not vanish for the improper (stimulated)
waves (see formula (8.2.21)). Then, within Kramer formulas, we find the
amplitudes
where is the algebraic complement of element in the determi-

nant It is more difficult to find because the determinant
vanishes. The solutions are bounded only provided the
vector is orthogonal to its null vector corresponding to Hermi-
tian conjugate matrix i.e.,
Functions containing unknown operators (and derivative-

dependent functions determined by preceding approximation),
are unknown. That is why relations (8.2.26) are equations for We
derive expressions for functionals in question substituting (8.2.19) and

(8.2.20) into (8.2.26). In the first approximation, we find from (8.2.19),
(8.2.20) that
Then using (8.2.26) we obtain
where
is derived from the functional relation
is group velocity (see definition (1.3.3) and corresponding comments),
The equation (8.2.29) can be rewritten in equivalent form:
Relations (8.2.30), (8.2.31) are obtained accounting for )

and where are arbitrary constants and are
algebraic complements of relevant matrix elements ii)
provided condition holds.
One can be confident that equation (8.2.31) is particular case (in the
first approximation) of the general equation of second hierarchical level
(8.2.16). This means that:
a) when the function is periodic in or (or in and simulta-

neously) the above calculational procedure can be used once more with
respect to the equation (8.2.31);
b) the basic hierarchical principles hold; hence, classic version of
slowly varying amplitude method is one of hierarchical methods.
Analogous situation realizes for higher approximations, too. We show
this for truncated equations of second approximation. In order to con-
struct the latter with respect to we find functions from the
system (8.2.24). To do this we employ the expression (8.2.25) and write
the functions as
where
Suppose the solution to equation (8.2.24) is
Moreover, we assume the conditions
Using (8.2.9), we find
and
is the solution of the first equation of (8.2.36). We employ the Kramer

rule and de l’Hopital rule for the indeterminacy of the type 0/0 to find
from the second equation (8.2.36) that
Thus after all the transformations we can write
Similarly, to deriving the expression for we employ (8.2.27) and

(8.2.28) to obtain
We substitute (8.2.41) into (8.2.26) with regard for (8.2.40) and use
(8.2.32). Thus we obtain equations for slowly varying wave amplitudes
accurate to the second-order terms in
Here is obtained similarly to (see formula (8.2.29). Deriving

(8.2.42), we use the relation
Within (8.2.31), the latter can be obtained from
Equations of higher approximations can be derived in a similar man-

ner.
The case of slowly varying parameters of the medium. Re-

call that above calculational scheme is constructed for the simplest one-
dimensional model. Equation (8.2.17) was derived with matrices A, B
and C independent of time and coordinates. However, this restriction is
not principal and is used only for simplicity. One can be confident that
above derivation of equations for slowly varying wave amplitudes can be
generalized onto the case with matrices A, B and C weakly dependent
of coordinates and time. Such situation can be treated as model with
slowly varying parameters of the medium (within which the considered
wavy processes occur). We carry out the procedure analogous to the
above derivation and obtain an equation for slowly varying amplitudes
in the first approximation with respect to i.e.,
In a similar way one obtains truncated equations of higher-order

approximations. This methodological aspect is illustrated in Chap-
ters 12, 13, Volume II, for isochronous free electron laser (the amplifier
of Dopplertron type).
The other similar generalization of discussed calculational scheme can
be realized, too. So, in the above calculational algorithm we assume that
a) the resonant waves have constant amplitudes (i.e., are neither
damped nor growing) in the linear approximation
b) the resonance band cannot contain simultaneously more than one
spectral component of each wave, i.e. wave resonant interactions in the
system are non-degenerated.
In electrodynamics of plasma-like systems there are more models with
violated assumptions than that with satisfied ones. For example, signal
wave and space charge wave in parametric free electron laser (see Chap-
ters 12, 13, Volume II) are damped in linear approximation due to dissi-
pation processes. The beam waves in the superheterodyne free electron
lasers are unstable (i.e., growing) even in the linear approximation. Sev-
eral space charge waves can be in the state of combination synchronism
simultaneously under the Compton interaction modes in the free elec-
tron lasers (both parametric and superheterodyne — see Chapters 12,
13, Volume II) and so on. It is easy to verify that above procedures
can be extended to these cases as well. However, the calculations be-
come too involved to be considered here without loss of consistency. To

avoid this, we illustrate such generalizations for several particular cases
in Chapter 12, Volume II.
2.2 Classification of Transversely Inhomogeneous

Models
Suppose the parameters of the model slowly vary in the transverse
plane. This may imply, for the physical characteristic cases in calcula-
tional practice, transverse non-uniformity of the pump and signal fields
in free electron laser, inhomogeneity of the density of relativistic electron
beam, etc.. Two distinctive possibilities occur:
a) transverse inhomogeneity (initial or developed) determined by the
amplitude derivatives (with respect to coordinates and ) of the same
order of magnitude as the derivative with respect to — the
model with moderate inhomogeneity,
b) the inhomogeneity much more pronounced in transverse directions

than in the longitudinal one — the model with strong inhomogeneity,
Respectively, the case
as almost transversely homogeneous model is equivalent to the situation

considered above in this subsection.
We note that ideology of constructing asymptotic solutions for models
(8.2.46) and (8.2.47) is similar to the above reasoning. That is why
we consider brief comments of calculational procedure reproducing the
previous cases.
2.3 Model with Moderate Inhomogeneity

Let us consider a non-one-dimensional model with quasi-plane waves.
In this case, operator P (compare with (8.2.1)) takes the form
where is nabla operator with the components

Then the standard form (i.e., (3.3.21) for ) reduces to
where is the slow radius-vector. The standard systems (8.2.50)

are also called Rabinovich standard form [1–3]. We consider the case of
constant matrices A, B, and C. In the scalar form equation (8.2.50)
can be written as
with be elements of matrices A, B, and C. We assume solu-

tions of (8.2.51) to be of the same form as (8.2.15), i.e.,
Here is the
wave phase. We carry out the same procedure as in previous subsection
to obtain the equation of the first approximation (see (8.2.28))
where
is the group velocity.

Equations of the second approximation can be constructed analo-
gously to (8.2.53); those for complex amplitudes are
where contains the vector of the

first approximation (see (8.2.51)).
Equations (8.2.56) are derived assuming derivatives
to be of the order of magnitude of while coefficients
of the equations are of the order of ~ 1. In some models, however, this
assumption does not hold. Nevertheless, the procedure can be easily
extended to such cases, too.
2.4 Method of Parabolic Equation

Let us discuss the method of parabolic equation [2, 4]. This method
is especially efficient in the study of effects produced by various mecha-
nisms of wave beam diffraction in nonlinear media [4].
Suppose the wave amplitude varies in the direction of wave prop-
agation more slowly than in the perpendicular plane (condition (8.2.47)).
Assuming waves propagating along axis, i.e., we change the
variables in equations (8.2.55):
At the same time,
The relevant amplitudes rates of change can be estimated as
Then within (8.2.56) and (8.2.57) equations (8.2.55) reduce to

where the terms of the order of are allowed for, and so on, and so forth.
3. MODERNIZED VERSION OF THE

SLOWLY
VARYING AMPLITUDE METHOD
3.1 Field Problem
As experience shows, attempts at using the conventional version of the
slowly varying amplitude method in non-linear wave resonant problems
of plasma-like systems show it is much more difficult task than it seems.
The first difficulty arises at the stage of problem formulating. In
particular, when one tries to describe adequately treated model in terms
of the traditional version of the slowly varying amplitude method. Thus
we should to reduce the Maxwell equations (1.4.23), (6.1.6) to standard
form like (8.2.2), (8.2.50). But two main problems arise here. The first
concerns the manner of selecting of the problem small parameter The
second is connected with the circumstance that the vector U, strictly
saying, is not periodic one (with respect to all fast phases) because it
contains null slowly varying harmonics of fields and currents. The latter
could be generated, for instance, as a result of nonlinear wave resonant
interaction in the system (see Chapters 12 and 13 for more details). All
this means that the considered method should be modernized in the case
of electrodynamics applications. The result of such improving is referred
as to the modernized version of the slowly varying amplitude method.
The first characteristic feature of the mentioned improving procedure
is the taking into account of the mentioned quasi-stationary part of act-
ing fields and currents (i.e., their null harmonics). The modernization in
this case consist in that we should introduce the concept of so called mod-
ernized standard system like (3.3.21) instead standard systems (8.2.2),
(8.2.50):
where is operator consisting of differential three-dimensional nabla-

operator as components, is a null harmonic of function
is its oscillatory part, and the vector U is defined in the following
manner
The second characteristic feature of the modernized version in the

case of field problem is the use of the methods described above in Chap-
ters 6, 7 for constructing the right parts of equations (8.3.1). We have in
view the methods of averaged characteristics, kinetic equations, averaged
quasi-hydrodynamic or current density equations.
The point is that a number of methodological problems are connected
with the method of separating the small parameter in the right hand
part of the Rabinovich standard forms (see (8.2.2), (8.2.50)). Usually,
some largest amplitude of interacted waves (after their corresponding
normalization) is taken to be at calculational practice [3–9]. How-
ever, this is important that such separation method, strictly speaking,
has no any acceptable mathematical substantiation. It is based on some
purely semi-qualitative considerations only. It should be mentioned that
it turns out to be tolerable, as a rule, in the low-order nonlinear theo-
ries only. But the practical calculational situation becomes unacceptable
quite often in the high-order nonlinear cases. Thus a number of calcu-
lational problems, which appear in practice, are results of incorrect use
of the standard slowly varying amplitude method. Including, the con-
vergence problems, the problem of determining the confidence interval
(over which obtained solutions are correct), the problem of losing some
terms in right parts of corresponding truncated equations, etc..
These problems could be solved in framework of the modernized ver-
sion by use the methods, described earlier in Chapter 6, which are used
for constructing the right parts of equation like (8.3.1). The methods
of averaged characteristics, or averaged quasi-hydrodynamic or kinetic
equations, or current density equations, can be used for this purposes.
They are applied for solving the motion problem (i.e., current density
and space charge problems, that is the same), which are characterized
by the rigorous method of separation of the small parameter
In turn, these solutions determine the mathematical arrangement of the

corresponding right hand parts of equations. Let us turn the reader at-
tention that the procedure of the small parameter separation is strictly
correct in this case (see Chapters 4 and 6 for more details). This means
automatically that the above noted difficulties are not typical for the
modernized version of the considered method.
Let us mention once more that concrete realization of the calculational
algorithm for construction of the modernized standard form like (8.3.1) is
determined essentially by the method of solving the motion problem. As
already mentioned, the latter, from the physical point of view, coincides
with the current density or space charge problems.
Let us assume that the method of averaging kinetic equation is chosen
as basic for calculational of the right parts of equations (8.3.1). This
means that the vector U in (8.3.1), (8.3.2) should be constructed in this
case in the form:
In what follows we turn to the averaged quasi-hydrodynamic or av-

eraged current density or the averaged characteristic methods. It is
remarkably that the space charge density also serves as one of coordi-
nates of vector U in this case:
But let us discuss this problem in more detail.
3.2 Current Density Problem

Let us choose the method of averaged current density equation as the
basic one. Taking into consideration the characteristic structure of the
Maxwell equations (1.4.23), (6.1.6), the vector-function
is some linear function of the vector of current density

where is the vector consisting of three-dimensional current density

vectors as components, is relevant square constant matrix;
and are the same vectors consisting of null-harmonics and
non-null-harmonics of the current density. Furthermore, we perform a
Fourier series and power series expansions of the function obtained
by means the averaged current–density method (see Chapter 6, Section 6
for more details):
where all notations are obvious. It is readily seen that by using definition
(8.3.5) and current density equation (6.6.1) we can complete system
(8.3.1) by the equation
where is some known function of current density vector

We construct asymptotic solutions for by means of the
previously described calculational procedures. Then, solutions for the
can be obtained in the form of relevant asymptotic convergent
series:
where
(see definition (8.3.6)).

All this means automatically that the second of the two above men-
tioned difficulties of the traditional version of the slowly varying am-
plitude method could be overcame, in principle, by the proposed way.
Indeed, taking into consideration the relationship
and substituting the obtained expansions like (6.6.17), (8.3.9) in modern-

ized standard form (8.3.1), we have the standard form with completely
determined right side. (Here, as earlier, is the largest scale
parameter of relevant hierarchical series like (4.1.6)).
3.3 Current Density Problem in Framework of

the Kinetic Approach
The calculational scheme for kinetic equation (Boltzmann equation
(6.1.1)) has a number of specific peculiarities. Let us discuss some of
them. According to (6.1.7), (6.1.8), the densities of current and the
space charge could be represented in kinetic case as
where we omit subscript for simplicity. By virtue of asymptotic repre-

sentations (6.7.5), (6.7.7) we can write (8.3.12), (8.3.13) in the form of
the power series
where notations for and are evident, and are the vectors
of non-averaged and averaged slow variables, respectively. So the earlier
discussed problem of separation out the problem small parameter

(see the preceding Subsection) here also finds a satisfactory solution.
Thus the motion beam problem has some another calculational pro-
cedure in the kinetic version of self-consistent theory. The distinction,
from the ‘point of view of the slowly varying amplitude method’ con-
cerns only structure of the vector U (8.2.5), (8.3.4). It is obvious that the
mentioned distinction is not important for general calculational scheme
of the considered method.
4. METHOD OF HIERARCHICAL
TRANSFORMATION OF COORDINATES
4.1 Main Idea of the Hierarchical
Transformations
Previously we discussed some version of application of the aver-
aged current–density equation method in framework of the modernized
method of slowly varying amplitudes. Analysis shows (see, in partic-
ular, Chapters 12, 13, Volume II) the calculational procedures of the
discussed class could be very promising for solving problems of various
nonlinear theory of wave dynamical systems. For instance, electrody-
namic systems with high intense electron or plasma beams very often
are characterized by complicated dynamical surface configurations. In-
cluding intensive electron beams in relativistic two-stream systems, rel-
ativistic high current electron devices (free electron lasers (FELs), can-
cerotrons, gyrotrons, etc.), various high-current acceleration systems,
such as linear induction accelerators, radio-frequency accelerators, EH-
accelerators, etc.. The considered hierarchical methods could be used for
overcoming the difficulties connected with such type of peculiarities. We
take in view, in particular, the partial variety of the hierarchical ideology
realization that is known as the method of hierarchical transformation
of coordinates [12].
The proposed approach can be especially effective for solving of the
earlier discussed nonlinear dynamical problems characterized by complex
non-stationary boundary conditions on the beam surface. Including,
the situations if the beam surface being a self-consistent function of the
solved problem (i.e., it is also the subject of determining). It should be
mentioned that this problem is rather old for the physics of intensive
high-current electron and ion beams. However, its has no acceptable
solution even today [14, 15]. This circumstance additionally stimulates
researcher for developing of new, of principle, conceptual ideas of such
kind.
Another topical problem of electrodynamics of plasma-like objects is

the problem of determination of equilibrium states of systems, which are
characterized by presence of some ‘controlled small-scale instabilities’ of
some another types. But let us to discuss this problem in more detail.
It is well known that the theory of stable equilibrium of charged par-
ticle beams was elaborated mainly for the unperturbed ‘electrodynam-
ically cold’ systems [14, 15], i.e. the systems within which any active
wave processes do not occur. Usually, talking about the ‘stable states of
beams’, namely, the ‘electrodynamycally cold’ systems are taken in view
namely. However, some wider treatment of the beam stability prob-
lem is known, too. Here we talk about the conditional stability of ‘hot’
charged particle beams, as whole. It is obvious that we have unstable,
in principle, electron beams in the interaction region of power devices
like FELs (see Section 1, Chapter 1 and Chapters 9–13, Volume II) or
other similar. But this instability carries small-scale character because
all processes here develop in the scale of a wave length (This leads,
in particular, to that the same small-scale oscillations characterize the
beam surface oscillations, too.) On the other hand, the beam, as a
whole, must be stable in the scale of transverse characteristic (design)
size also. Usually, it is typically for many traditional plasma-like sys-
tems: It is clearly that such ‘electrodynamically hot’ electron
beam possesses an averaged surface configuration different greatly from
the ‘cold’ one. Correspondingly, the determination of the averaged equi-
librium states in this case becomes very important because contrary to
‘cold’ system the latter in the ‘hot’ regime could be unstable with the
characteristic ‘large scale’ Really this means that the electron can
not free pass through the system working bulk, i.e. a normal work of
the device becomes impossible.
Let us show that rational solution for both discussed problems can be
got within framework of the method of hierarchical transformation of co-
ordinates. The essence of this calculational scheme consists in transform-
ing of the Maxwell equations from the three-dimensional non-averaged
space into a three-dimensional also, but averaged,
space The use of the transformation formulas
like (4.1.13), (4.1.15)
is characteristic feature of this method. Here is the vector consisting

of averaged slowly varying values of the first hierarchy (excluding com-
ponents of the vector is similar slowly varying vector of second
hierarchy, and so on, and is the largest hierarchy number.
The physical background of this transformation idea is the following.
The Lagrange or Euler description forms usually are used to describe
charged particle beams (see Chapter 1, Subsection 4.9 for more details).
Let us remind that the first approach includes the problems of motion
of individual particles as its inseparable part. Thus an observer is con-
nected with the observed particle and registers its coordinate and velocity
in time In contrast, one needs to fix a spatial point of radius-vector
and to register velocities of passed particles in the second approach. It
is important that the coordinate of observed space point always coin-
cides with electron coordinates in the first case. Hence the coordinate
depends on time as that is characteristic feature of
the Lagrange description. In contrast to this the variables and are
reciprocally independent ones in the second case. This can be explained
by that we have different particles in the same spatial point in different
instants of time in framework of the Euler description.
Let us suppose that all observed space is filled by charged particles.
Then, we may put in correspondence each spatial point for any instant
of time with some charged particle, i.e., we have different particles in
the same spatial point in different time But all these particles move
accordingly to the same motion dynamical law. Or, in other words, par-
ticle motion is described by the same differential equations. The differ-
ence concerns only the initial conditions, which are ‘personal’ for each
particle. Thus the wave nature of the studied object determines the co-
herentness of oscillations of different particles in different spatial points.
This means that any observer in the coordinate system oscillated with
all frequencies of particle motion should see the initial beam as a non-
oscillative ensemble of the particles. Hence, we can use the relationships
like for accomplishing the transformation of the spatial Euler’s
coordinates from the initial (non-averaged) to averaged space. These ob-
servations are put in the ideological basis of the method of hierarchical
transformation of coordinates. But let us discuss their in more detail.
Reverting to the above discussed beam problems we can reveal that
both: the boundary condition problem as well as the equilibrium prob-
lem can be solved essentially simpler in averaged coordinate
space. In this case, the equation system consisted of current–density

and Maxwell’s equations in the space does not contains
any oscillations. All averaged fields here are represented by some ef-
fective quasi-stationary fields’ only (see also about the effective fields
in Chapter 5, Subsections 2.5 and 2.6). Hence, the initial non-linear
resonant wave fast oscillation problem can be reducing to a spatially
‘smooth’ quasi-stationary problem. Because all small-scale oscillations
here are not presented already that the boundary conditions in this
averaged space are formulated in the terms of averaged smooth (i.e.,
non-oscillating) beam surface. It is obviously that the such av-
eraged problems are much simpler than the initial (zeroth hierarchical
level) fast-oscillative non-linear wave resonant problems.
It should be mentioned, however, that the method of hierarchical
transformation is not too developed today. The general idea and calcu-
lational scheme have been described only in references [12]. Therefore,
below in this Section we also confine ourselves by short discussion of the
general idea and calculational scheme of this method.
4.2 Hierarchical Equations

Let us successively transform standard system (8.1.1) into the
averaged coordinate space. Inasmuch as vector-functions contain
(as components) current density vector and with transformation for-
mulas (6.6.17), (8.3.1), (8.4.1) we can write for the the
following transformation equations of the first hierarchy:
where all the quantities have been defined earlier. Hence, we represent
the vector U as
where is the vector of averaged slow quantities. Consequently, we can

got the following presentation for modernized standard system (8.3.1)
on the first hierarchical level as before)
where is the differential operator constructed of the averaged three-

dimentional nabla operators as elements. We obtain for the second
hierarchical level
and so forth. Here notations of all the quantities are quite clear within
the context of above accomplished discussions. We proceed the proce-
dure of successive hierarchical transformations until the terminal hier-
archical level will be attained. Here we solve the constructed
averaged truncated equation
and obtain corresponding solutions for functions The important

peculiarity of such truncated equations is absence any periodic depen-
dencies in both: the slow and fast oscillatative phases. In this regard
equation (8.4.11) is quasi-stationary one, i.e., it describes only non-
oscillatory (quasi-stationary) dynamics of the system considered.
Then we discuss standard system (8.4.6), (8.4.7) in this av-
eraged space. We note that all resonances and oscillations are ‘hidden’
in transformation formula (8.4.10). Equation (8.4.11) in its mathemati-

cal structure is much simpler than non-averaged initial equation (8.3.8).
Hence, all excited electromagnetic fields in this averaged sys-
tem are quasi-stationary (i.e., non-oscillatative), too. We solve equation
(8.4.11) (by means some analytical numerical method) and construct
m-fold averaged the standard equation:
where subscript and superscript means multiplicity of hierarchi-

cal level, and is differential operator consisting of av-
eraged three-dimension nabla-operator Then we solve equa-
tion (8.4.12) and obtain relevant solutions for quasi-stationary vector-
function
We construct equations for functions of order of hierarchical
multiplicity at the following stage of calculational:
For the hierarchical level we obtain
and so on.
Following this way we eventually obtain the solution of initial problem

(8.4.6), (8.4.7).
Thus hierarchical procedure of asymptotic integration of the mod-
ernized standard system (8.3.1) in the discussed case consists of the
following steps:
1) the extending of averaged current density method on calculational
of
2) the transforming of coordinate and relevant differential operators
into averaged coordinate space (see below);
3) the solving of averaged equation for (see
(8.4.11)) and constructing averaged standard equation like
(8.4.12);
4) the solving equation (8.4.12) and forming the basis for obtaining
relevant solutions of hierarchical and so on;
5) the using successive inverse transformations into the initial (zeroth)
hierarchical level.
Two important circumstances should be mentioned especially. The
first is that solving standard system (8.4.16), (8.4.17) we use the method
of averaged characteristics or the slowly varying amplitude method. The
second concerns the level of the intensity of labour of these procedures.
The function in the case of ‘ordinary’ modernized slowly varying
amplitude method (or the averaged characteristics method), contains all
fast Euler’s oscillation phases and (the combination
phases are fast in framework of the Euler’s description). There-
fore, this function is rather complicated and corresponding calculational
procedure for U immediately becomes too complicated also.
Thus the main advantage of the method of hierarchical transformation
of coordinates is the following. Firstly, it allows to divide the initial too
complex problem into a number of much simpler partial ones. Thus
total labor-intensity of the calculations performed accordingly with the
discussed method is less than the labor-intensity of the slowly varying
amplitude method. Secondly, we got a promising way for solving the
above discussed problem of boundary conditions and equilibrium states.
4.3 Averaged Operator

At last, we give the transformation procedure of the
into the averaged This problem might
be formulated in the following manner:
It is obvious that
In what follows we use the Krylov–Bogolyubov substitutions (8.4.1):
where the method calculational of functions is described in Chap-

ter 4. Furthermore, we take into account that
i.e. we get the definition of functions through the non-averaged

coordinates by the method of back transformations (see Chapter 6,
Subsection 2.2 and others):
This means that using (8.4.21)–(8.2.24) we can write the following
i.e. required expression for the transformation function (8.4.24)

eventually can be represented as
5. MITROPOL’SKII METHOD
Previously, in this and preceding Chapters, we considered the asymp-
totic integration algorithm for a certain class of systems of weakly nonlin-
ear partial differential equations. The common property of the discussed
before methods is specific combination of the straight and back trans-
formations. However, some other technical realizations of calculational
schemes of such type are known also. In spite of that, some universal
basic ideas of the averaging method are employed here too, formally,
these algorithms essentially differs from the above discussed ‘classical’
calculational procedures. The discussed method is described in refer-
ences [10, 11]. Taking into consideration the surname of their principal
author, this method had been call as the Mitropol’skii method.
At the same time, the problems of electrodynamics of distributed-
parameter systems sometimes can be reduced to the standard form that
is characteristic for this method. So, let us further to discuss shortly
some main ideas and calculational peculiarities of the Mitropol’skii
method.
5.1 Reduction of a Partial Differential Equation

to the Standard Form with
Fast Rotating Phases
Initial equations. Let us consider a special case of general equation
(6.1.12) when this system of first-order differential equations may be
reduced to an equation of the second order that is given by
where are constant coefficients, is the small

parameter of the problem F is a given nonlinear function
of its arguments. We assume that the constant coefficients
satisfy the condition:
Let us show that in the case (8.5.2), the partial differential equa-
tion (8.5.1) reduces to a system of ordinary differential equations. The
asymptotic integration of these equations can be carried out by means
of Bogolyubov–Zubarev method (see Chapter 4).
We follow the procedure described in [10, 11], carry out standard

changes of variables, and thus reduce equation (8.5.1) to the canonical
form:
where is the new variable, and are constant coefficients, is a

known nonlinear function. For the sake of simplicity, in what follows we
omit the primes in the notation of the quantities which enter (8.5.3).
We impose the boundary and initial conditions by means of the equa-
tions
where and are sufficiently smooth functions which satisfy all

relevant conditions.
5.2 Basic Solutions

Let us find the form of the basic solutions of equation (8.5.3). To do
this, we linearize the equations in a manner similar to the procedure of
Sections 1 and 2. Namely, we put so that equation (8.5.3) takes
the form
We write the solutions of equation (8.5.6) (with boundary conditions

(8.5.4), (8.5.5)) in terms of Fourier series, i.e.:
where
is the normal oscillation frequency of the linear system,

are the numbers of Fourier harmonics, and are constant coeffi-
cients (amplitudes of the Fourier harmonics) determined by the condi-
tions (8.5.4), (8.5.5).
5.3 Truncated Equations

Similarly with the method of slowly varying amplitudes (see Sections 1
and 2), we assume that the solutions preserve their form (i.e., reproduce
(8.5.7)) for The distinction is that now the amplitudes and
are slowly varying functions. However, the calculational procedure
is modified as compared to the previous cases. Namely, in the general
case we write the solution of the equation (8.5.6) in the form:
where are unknown functions to be found. We substitute (8.5.9) in

the equation (8.5.6) with regard for the initial conditions (8.5.5). Then
we multiply the result obtained by and
integrate from 0 to Thus we come to an infinite system of ordinary
differential equations [10, 11]
The system (8.5.10) satisfies the initial conditions
where and are the Fourier coefficients of the functions and

respectively.
We introduce (instead of and ) new slowly varying complex
amplitudes and fast phases thus we have
Then we differentiate the last expression of (8.5.12) with respect to

and compare the expressions obtained to (8.5.10). As a result we have
the system of equations
Besides that, we observe that
where the asterisk symbolizes the complex conjugation operation.

We solve the system with respect to and and thus
obtain the required infinite (innumerable) system of ordinary differential
equations for the complex amplitudes, i.e.,
where
It is obvious that the system of equations of the form (8.5.15) is a

standard system with rotating phases. Therefore its asymptotic inte-
gration can be carried out in terms of one of algorithms described in
Chapter 4. The truncation procedure for the infinite system (8.5.15)
should be specialized for each problem under consideration.
6. EXAMPLES OF REDUCING OF
THE MAXWELL EQUATIONS TO
THE STANDARD FORM FOR
THE METHOD OF SLOWLY VARYING
AMPLITUDES
Maxwell equations can be reduced to hierarchical standard form like
to (8.1.1), (8.2.2), and (8.2.50). Let us demonstrate this at the examples
of kinetic and quasi-hydrodynamic versions of slowly varying amplitude
method. Thus we will take into account that both they are differed by
the definitions for the vector U (see (8.3.3) and (8.3.4)) only.
6.1 Kinetic Version

We start with Maxwell equations (1.4.23), (6.1.6) supplemented by
kinetic equation (6.1.1) and relevant definitions (6.1.7), (6.1.8). Com-
paring (1.4.23), (6.1.6) and (8.1.1), (8.2.2), and (8.2.50) we see that
Maxwell equations (1.4.23), (6.1.6) in their original form do not satisfy
standard equation requirement (8.1.1), (8.2.2), and (8.2.50). Therefore,
the first obvious step is to standardize the set of equations (1.4.23),
(6.1.6).
Let us represent the current density vector and the charge density
in (1.4.23), (6.1.6) as sums of linear and nonlinear parts, respectively:
In the most general case the linear parts and can be written
as
Explicit expressions for the matrices can be

easily derived from Maxwell’s equations.
Fields. We arrange the components of the vectors and as

a column vector
We have for the operator (see (8.1.1))
Maxwell’s equations (1.4.23), (6.1.6) in the terms of above notations

could be reduced to form like (8.1.1)
where
[0] are 3 × 3 zero matrices; (0) and {0} are three-dimensional zero-row
vectors and zero-column vector, respectively. The function can be
represented as
6.2 Quasi-Hydrodynamic Case

Here the difference with previous kinetic case is only that the vector
U includes the space charge density as a component (see (8.3.4)):
Accomplishing the sequence of the above described calculational pro-

cedures we can get the following expressions for the matrices A,B,C,
and the vector-function
where are known predetermined functions.
7. EXAMPLE: THE TWO-STREAM

INSTABILITY IN A TWO-VELOCITY
ELECTRON BEAM. THE METHODS OF
AVERAGED QUASI-HYDRODYNAMIC
EQUATION AND SLOWLY VARYING
AMPLITUDES
The nonlinear theory of the two-stream instability is discussed previ-
ously in Chapter 7. The calculational technology of the method of av-
eraged characteristic had been illustrated owing to this example there.
Let us turn once more to the problem about nonlinear evolution of the
two-stream instability in relativistic two-velocity system. However, we
will use, at that time, a specific combination of the methods of quasi-
hydrodynamic equation and slowly varying amplitudes for the same pur-
pose. The chosen manner of illustration allows us to compare calcula-
tional features of both these approaches. As will happen below, such
comparison could be very useful for deeper understanding calculational
technologies of the method of averaged characteristics as well as the
methods of quasi-hydrodynamic equation and slowly varying amplitudes,
respectively.
7.1 Statement of the Problem

Let us choose the statement of the problem analogous to the de-
scribed previously in Chapter 7, Subsection 1.1. The difference is only
that we take into account here a possible presence of longitudinal quasi-
stationary electric field. This field might be put to the model working
bulk (such type of external fields is referred to as the electrostatic sup-
port [19–22] — see also Chapters 11–13, Volume II). Apart from that,
the intrinsic electric field which can be generated as a result of non-
linear interaction of electrostatic waves is also taken into consideration

(see the corresponding discussion about the generated quasi-stationary
electric field in Chapter 7, Subsection 3.2). We will use instead defini-
tion (7.1.3) the following definition for the electric field in the considered
system
where describes the resulted slowly varying longitudinal

quasi-stationary electric field that acts at electrons (the generated as
well as the external ones). All other notations are the same that were
given in Chapter 7 (see definition (7.1.3)).
We also take into consideration the following. As is well known (see
further Chapter 13, Volume II) four electron waves can exist in the two-
velocity electron system. They are the increasing, decreasing (damped),
fast, and slow electron waves. Thus the increasing wave only exerts main
influence at the amplification process. We neglect other three waves at
the system input that allows to simplify the solving the problem. At
the same time, as the analysis shows, such supposition does not exert
essential influence at the eventual results.
The general initial self-consistent problem is divided into two more
particular ones. Namely, they are the problem about motion of elec-
tron beam in the given electromagnetic field (motion problem) and the
problem about excitation of these fields for the given motion of elec-
trons (field problem). The method of quasi-hydrodynamic equation (see
Chapter 6, Section 5) will be use for solution of the first problem. The
method of slowly varying amplitudes (see Sections 1 and 2) will be used
in the case of the field problem.
7.2 Motion Problem. The Averaged

Quasi-hydrodynamic Equation
The quasi-hydrodynamic equation in the form (6.1.5), (7.1.1)
is chosen as the basic for the motion problem. Here all definitions are
given for (6.1.5), (7.1.1). Similarly with the case discussed in Chapter 7
we assume for simplicity: v = 0, Then equation (8.7.2)

could be essentially simplify:
Let us remind that the function is considered as a given

one in framework of the motion problem.
We follow, as mentioned above, to the calculational scheme of the
method of quasi-hydrodynamic equation (see Chapter 6, Section 5). Ac-
cordingly with the method we should to rewrite further equation (8.7.3)
in the form of system of ordinary differential equations:
In what follows we perform the classification of variables into fast and

slow ones, respectively. Taking into consideration the previously given
definition for the latter (see, for instance, Chapter 1, Subsection 2.4),
we may regard the velocity and the coordinate as slow variables for
system (8.7.4). The variable we regard as the fast rotating phase
see, in particular, Chapter 4). This gives a possibility of
separating out the problem large parameters:
where and are the nor-

malization constants. As a result of corresponding transformations, we
can write equations (8.7.1) in the standard form like to (4.2.1):
where all notions in view (4.2.1) are self-evident. Characteristic feature

of the considered problem is that the vector of slow variables is here
the two-dimensional one:
Correspondingly, the vector-function in right part of the first of equa-

tions (8.7.6) is two-dimensional, too:
Thus
The solutions for the system (8.7.6) could be written in the form of
the Krylov–Bogolyubov substitutions like (7.1.10):
where the following equations of the first hierarchical level can be con-
structed for the averaged values:
The functions we find accordingly with the

algorithm described in Chapter 4, Section 2.
First approximation. First approximation on Then, accom-

plishing relevant calculations for the two-dimensional vector-function
we obtain:
where
Analogous calculations for the two-dimensional function
yield:
where Here let us recall that the vector-function

accordingly with the used calculational procedure, is the ‘integra-
tion constant’. In our case we used the conditions (4.2.9), (4.2.12) for its
rendering concrete. Analogous calculations for the functions
allow to got:
or
where the sum on in (8.2.4) is performed on The function

as well as the function is also the ‘integration constant’.
The method their determining is the same.
Second approximation. In what follows let us turn to calculate the

addends of the second approximation. This topic is very interesting
in the illustrative example considered because we could come to rather
unexpected results. In particular, we have that part of the second ap-
proximation addends in averaged equations (8.7.13) turn out to be equal
zero. These means that, in spite of the above obtained results
found in framework of the first approximation, they are valid for
system length and for the observation time (see the
discussion about the accuracy problem in Chapter 7, Subsection 3.5).
Later we will illustrate this peculiarity by concrete calculations.
Following accordingly with the calculational scheme used we obtain
corresponding expressions for the required functions in the second ap-
proximation Including, it could be found for the vector-function
As readily seen, the five terms (integrals) in right side of expressions

(8.7.26) determine the function Let us calculate successively each
of these terms.
We start with the first term in the function
because the integral on is equal zero. The second term is also equal
zero:
The third term is equal zero
because (see formula (8.7.8)) the function
does not depend on the large parameter

The fourth term in (8.7.26) is equal zero, too:
because the integral of function on equal zero (see corre-

spondent commentary to (8.7.19) concerning the choosing of the function
and do not depend on
The fifth term in (8.7.26) is also equal zero:
because, as mentioned above, the integral of function on

equal zero, and the function does not depend on
Thus we find finally that
In what follows let us turn to the calculating function Taking

into consideration the structure of expression (8.7.25), we will calculate
each of five integrals (terms) in the right side of (8.7.26). Then one is
readily convinced that the first term is equal zero:
because integrals of functions on equal zero, and other

factors do not depend on
The third term in (8.7.26)
is equal zero because the function does not depend neither on

the average phase nor of the parameter
The fourth and fifth terms in (8.7.26)
because Summarizing, we can write:
However, carrying out analogous calculations for the functions

and (see Krylov–Bogolyubov substitutions (8.7.12)) we can observe
that these addends are not equal to zero. Results of such calculations
are given in Appendix. This means that, In contrast to the averaged
components of motion, the oscillatative components calculated in the
first approximation are valid on the length Hence we should
use in this case the functions and for extending the ‘confidence
interval’ for the length (i.e., the length on which is guaranteed

the accuracy of obtained asymptotic solutions).
Thus the nonlinear averaged dynamical of the system in the second
approximation really is determined by correspondent functions calcu-
lated in the first approximation:
where But as it noted above

(and that follows from the results given in Appendix) this conclusion is
not valid for the oscillatative part of the problem.
Substituting expressions (8.7.29) into (8.7.13) we have the set of av-
eraged ordinary equations (equations for the first hierarchical level):
Carrying out the inverse passage from total derivatives to partial ones
in first of equations (8.7.30), and using the second equation, we obtain
the averaged quasi-hydrodynamic equation
It is readily seen that in the simplest case equation (8.7.2)

(which is obtained by the method of averaged quasi-hydrodynamic equa-
tion) can be reduced to equation (7.1.19) that is got by the method of
averaged characteristics. This once more illustrates the previously for-
mulated conclusion that the methods of averaged quasi-hydrodynamic
equation and averaged characteristics are not independent ones. The
difference is only that the partial derivatives in left side of initial equa-
tion in the case of the method of averaged characteristics can be ‘rolled
up’ into the total derivative immediately, i.e., without use of the concept
of characteristics.
Comparing equation (8.7.31) and initial equation (8.7.3) one can be
convinced that the first is essentially simpler because the right side of
the latter is non-oscillative. Various known solution methods (including
the method of characteristics — see Chapter 6, Section 3) could be used
for its integration. In the particular case of stationary model (when
the averaged velocity only, i.e., it does not depend on time
) equation (8.7.31) can be integrated, for instance, by the method of
separation of variables. In what follows, we assume that the considered
model is stationary and relevant solution is known.
7.3 Motion Problem. The Back Transformations

In what follows, let us to discuss the calculational peculiarity of the
back transformations. We suppose that relevant solution for averaged
quasi-hydrodynamic equation (8.7.31) is known. Then we simplify the
studied model assuming that the stationary state is established. Or,
in other words, it is considered that the averaged velocity does not
depend on time
The formal solutions of equation (8.7.31) could be written in the form
of the Krylov–Bogolyubov substitutions:
where required expressions for are given by formulas (8.7.29).

Substituting the latter into (8.7.32), we have:
(let us turn the reader attention that in the considered case, accord-
ingly with the (8.7.29),
Thus we got the expression for the averaged beam velocity as a
function of averaged values which, in turn, depend on
and implicitly. Accordingly with the calculational procedure of the
considered method, we should accomplish the back transformation for
obtaining solutions into the initial form (see Chapter 6, Subsection 5.2).
We perform for this the expansions of expressions (8.7.32) in the Tay-
lor series for two variables confining ourselves by only
accounting terms no higher
Substituting then expressions (8.7.29) and (8.7.33) into (8.7.34), we

find the looked for solution for the beam velocity in initial form
(i.e., for the zeroth hierarchical level)
7.4 Field Problem. The Method of

Slowly Varying Amplitudes
Analyzing the solution (8.7.35) obtained we see that in the accepted
approximation (the second, in our case — see above expressions (8.7.27),
(8.7.28)) is quadratic on the wave harmonic amplitudes The analo-
gous calculations for higher approximation orders (see, for instance, Ap-
pendix) would be do for obtaining nonlinearities higher than quadratic.
However, as it known from the theory of two-stream (and its particular

case — the plasma-beam) instability [14–18] the main source of non-
linearities here serves the field equations (Maxwell and continuity equa-
tions). Namely field nonlinearities mainly determine the multi-harmonic
properties of the two-stream instability (see relevant discussion in the
introduction to Chapter 7). In what follows, we will turn our attention
to the field problem, taking into account what has been said.
The system consisted of continuity equation (6.1.10), (7.2.1) and
Maxwell equations (6.1.6) is chosen as the initial one:
where, as before, corresponds to the partial electron beams, and

is responsible for the immobile ion background, the intensity of
longitudinal electric field is determined by definition (8.7.1). All other
notations are given earlier for expressions (6.1.10), (7.2.1).
Let us represent solutions for the concentration and beam velocity
(see solution (8.7.35) in the form of expansions in the Fourier series:
Thus the slowly varying amplitudes further are re-

garded as sought values, and they can be easily found by use of solutions
(8.7.35)
Here, as before,
Then we rewrite continuity equation in (8.7.36) in the form:
Here we have used equation (8.7.3), which, in turn, is rewritten in the

form:
Let us accept again the supposition about stationarity of amplitudes

in the considered model. This means that all amplitudes in (8.7.37),
(8.7.38) are depended on the longitudinal coordinate only. Therein
the expressions for partial derivatives on time for the concentration
and the beam velocity can be represented in the form:
Substituting expressions (8.7.41) and (8.7.42) into (8.7.40) we obtain

the equations for determining the slowly varying amplitudes in-
cluding the zeroth harmonics
where the values are

determined by expressions (8.7.37), (8.7.38), (8.7.1), (8.7.41), (8.7.42),
respectively.
Following on the analogous way it does not difficult to obtain relevant
equation for the slowly varying amplitudes of wave harmonics, including
the zeroth parts of the electric field:
where
Equations (8.7.35), (8.7.43), (8.7.44) forms a self-consistent system
which allows to analyze self-consistent nonlinear dynamics of the con-
sidered model. Numerical methods of analysis are used further for ac-
complishing of such analysis. Some results of this analysis are shown in
Figs. 8.7.1–8.7.3.
First of all, let us turn the reader attention that tree types of waves
can exist in the considered two-stream system. The first are the elec-
trostatic waves that determined by the wave part of expression (8.7.1).
The second are the electron-density waves defined by expression (8.7.37).
And, at last, the third are the kinematic waves (connected with particle
motion) determined by definitions (8.7.38). All these waves are closely
connected each with other, forming the only wave process. It should be
mentioned that the understanding of intrinsic wave structure of this pro-
cess is very important for its physical analysis. Unfortunately it is not a
rare event in the literature that ignoring this obvious circumstance leads
to some unpleasant results. Let us give one example which can obviously
illustrate such a kind of situation.
We take in view some works in references which have to do with

the analysis of influence of highest wave harmonics at the general wave
process. Sometimes the following erroneous calculational scheme is used.
It is studied dynamics of the first electrostatic wave harmonics under
influence of highest the kinematic wave harmonics. Relevant equations
for slowly varying amplitudes are constructed by use of the simplified
versions of the slowly varying amplitude method (see Section 1).
Let us to reproduce such a scheme of analysis for the case considered
of the two-stream model. Namely, we analyze the nonlinear dynam-
ics of the first harmonic of electrostatic wave amplitude (see equations
(8.7.44)), accounting therein the influence of high harmonics of the kine-
matic waves. The result of such analysis is illustrated in Fig. 8.7.1. There
the dependency of amplitude of intensity of first harmonic of electric field
on the non-dimension coordinate T (where L is the sys-
tem length) is shown. It is clearly seen that saturation of amplification
process occurs in the point on the level
In what follows, we turn our attention once more to such statement
of the problem being not correct. The point is that the excited high
harmonics of kinematic waves really should to excite (via the electron
density waves) a number of high harmonic for the electrostatic waves.
This means that calculating the dynamics of the first harmonics, we must
take into account influence of these electrostatic waves, too. Results of
such ‘improving’ calculations are shown in Fig. 8.7.2. Comparing the

results of Fig. 8.7.1 and Fig. 8.7.2, correspondingly, we see amplifica-
tion properties in the second case decrease more that 100 times! Hence
accounting the highest harmonics of electrostatic waves is always oblig-
atory in such types of researches. Besides that, this means also that the
analysis schemes of the first type (see Fig. 8.7.1) are not correct, at list
in the case of two-stream or plasma-beam systems.
The conclusion formulated above is also confirmed by the results
shown in Fig. 8.7.3. There the dependence of amplitudes of nine electro-
static harmonics on non-dimensional longitudinal coordinate T is shown
there. It is readily seen that most of these harmonics have approxi-
mately equal amplitude order. Moreover, one of harmonics (the second
one) has larger amplitude than the first harmonic even. It is obvious
that the neglect of influence of higher harmonics in such situation could
lead to the wrong results like that shown in Fig. 8.7.1.
References
[1] A.V. Gaponov, L.A. Ostrovsky, M.I. Rabinovich. One-dimentional waves in non-
linear dispersive media. Izv. Vysh. Uchebn., Ser. Radiofizika (Sov. Radiophys.),
13(2):169–213, 1970.
[2] M.I. Rabinovich, V.I. Talanov. Four lectures on principles of the theory of
nonlinear waves and wave interactions. Izd-vo LGU, Leningrad, 1972.
REFERENCES 347
[3] M.I. Rabinovich, D.I. Trubetzkov. Introduction in theory of oscillations and

waves. Nauka, Moscow, 1984.
[4] A.P. Sukhorukov. Nonlinear wave interactions in optics and radiophysics.

[5] N. Bloembergen. Nonlinear optics. Benjamin, Benjamin, 1965.
Kiev, 1994.

[8] L.A. Vainstein, V.A. Solntzev. Lectures on Microwave electronics. Sov. Radio,
Moscow, 1973.
[10] Ju.A. Mytropolskii, B.I. Moisejenkov. Lectures on application of the asymptotic

methods for solution of equations with partial derivatives. Pub. House of Inst.
of Math. Academia of Sciences of Ukraine, Kiev, 1968.
[11] Ju.A. Mytropolski, B.I. Moisejenkov. Asymptotic methods for solution of the
equations with partial derivatives. Vyscha Shkola, Kiev, 1976.
particles acceleration for electronics applications. the general hierarchical ap-
proach. International Journal of Infrared and Millimeter Waves, 19(1):3–93,
1998.

in the earth atmosphere. Izv. Akad. Nauk SSSR, ser. Fiz., (Bull. Acad. Sci.
USSR, Phys. Ser.), 8:6–20, 1944.
[16] V.V. Kulish, A.V. Lysenko, V.I. Savchenko. Application of the method of av-
eraging qusi-hydrodynamic equation for nonlinear problems of the theory of
two-stream free electron lasers. Gerald of Kyiv University, ser. Physics and
Mathematics, 4:471–480, 2000.
[17] V.V. Kulish, A.V. Lysenko, V.I. Savchenko. Method of asymptotic integration
of system with particular derivatives and its application for the problems about
motion of charged particles in external given electromagnetic fields. Gerald of
Sumy State University, ser. Physics and Mathematics, 3:5–12, 2001.
[18] V.V. Kulish, A.V. Lysenko, V.I. Savchenko. Nonlinear hierarchical theory of
two-stream instability in relativistic electron devices. Gerald of Sumy State
University, ser. Physics and Mathematics, 3:12–17, 2001.

dynamics. World Federation Pub. Inc, Atlanta, 1998.
the nonlinear problems of plasma electrodynamics. Fizika Plasmy (sov. Plasma
Physics), 19(2):216–227, 1993.
Appendix A
Results of calculations in the second
approximation
for Chapter 8, Subsection 7.2
349
The complete solution we obtain eventually in the form:
where the functions are the known functions of and (see

Chapter 8, Section 7).
Index
averaged current–density equation, 242 case of small oscillations, 181, 188, 201
Cauchy problem, 223, 225, 226
ability to model principle, 38 chaotic system, 37, 60
action postulate, 41 characteristic root, 228
algorithmic complexity, 45, 46 Cherenkov instability, 194, 195
amplitude, 11, 14, 20–22, 24, 28, 91, 92, closed hierarchical system, 59, 61, 115
113, 129, 160, 164, 172, 187, closed system, 32, 36, 44
194, 199, 201, 202, 249, 258, collision integral, 80, 208, 246
279, 284, 287–289, 291, 297, collision scale parameter, 80
306, 308–310, 344–346 combination oscillations, 22
analysis problem, 64 comparison equation, 10, 287
analytical–numerical methods, xvii, 1, 52, comparison system, 10
89 complementarity postulate, 40
angular acceleration, 9, 30 complex system, 37, 39–42, 44, 49, 50, 68,
angular displacement, 30 208
angular velocity, 30, 34, 35 condition of parametric resonance, 291
approximation of strong pumping, 288 conditions of compatibility, 213
artificial system, 60 conservation laws, 32
autonomy postulate, 39 continuity equation, 211, 238, 262, 263,
averaged current–density equation, 314 271, 342
averaged equilibrium state, 315 convergence problem, 310
averaged kinetic equation, 246, 348 coupled resonance, 17
averaged Maxwell’s equation, 275 cyclic frequency, 11, 15, 20, 23, 91, 193,
averaged quasi-hydrodynamic equation, 254, 287, 295
237, 239, 240, 339, 340
averaged quasilinear equation, 259, 261, definition of resonance, 14
268, 270, 271, 275 determined system, 36, 60
averaging operator, 63, 76, 77, 92, 96 dielectric permittivity, 34
averaging operator on spatial coordinates, Dirichlet–Jordan theorem, 77
76 dispersion equation, 27, 207, 250, 278, 295
averaging operator on time, 76 dispersion function, 27, 295, 300
dispersion law, 27, 289
Bhatnahar–Gross–Krook collision model, dispersion relation, 27, 109, 278
210 distributed system, 72, 78, 82, 293
Big Bang point, 116 distribution function, 79, 208, 209, 211
bucket, 179, 181, 182, 198, 201, 202 drift motion, 182
bunch, 196, 203, 204 drifting standing electromagnetic wave,
160
canonical form, 208, 228, 323 Duffing’s equation, 11, 146, 189, 194
canonical variable, 33, 168 Duffing’s parameter, 189
351
dynamical hierarchy, 101, 115 general standard form, 83, 285
dynamical operator, 176, 241, 296 generalized hierarchical equation of the
dynamical parameter, 54 first order, 73, 96, 120
dynamical system, xix, 18, 32, 35, 36, 50, generated form of solution, 286
52, 57, 59–63, 65, 67, 68, 71, generating equation, 10, 91, 99
73, 75, 83, 111, 175, 314 gravitational field, 174, 175
dynamical variables, 50, 56, 61, 83, 296 gray zone, 143
Grebennikov’s method, 113
effect of electron beam modulation, 159 group velocity, 25, 301, 307
effective potential, 176 grouping mechanism, 196, 197
electron clusters, 197
electron oscillation phase, 13, 168 Hamilton’s function, 33
electrostatic support, 329 Hamiltonian equations, 33, 244
elliptic function, 179 Hamiltonian formalism, 33
energy conservation law, 32, 44, 178 harmonic oscillations, 21, 22
energy parameter, 179 Heisenberg’s uncertainty principle, 41
equation of nonlinear pendulum, 10, 178, hidden oscillation phase, 12
179, 186, 188 hierarchical analogue of the second ther-
equation of stimulated nonlinear pendu- modynamic principle, 59, 67
lum, 194
hierarchical analogue of the third thermo-
equations of comparison, 99
dynamic principle, 187
equations of higher hierarchical level, 4
hierarchical asymptotic analytical-
equations of the first hierarchical level, 81,
numerical methods, xvii, xix,
93, 94, 127, 139, 184, 239, 256, xx, 4
332
hierarchical military unit, 114
equivalent linear system, 81, 286, 295
hierarchical oscillation system, 13, 111
essentially nonlinear system, 11
hierarchical principles, xx, 50, 58, 60, 63,
Euler variable, 35
73, 74, 83, 84, 113, 114, 176,
Euler's variable, 34, 112
218, 239, 302
explicit oscillation phase, 13
hierarchical series, 53, 54, 57, 58, 61, 90,
extended combinative phase, 186
extended coordinate vector, 86 114, 117, 119, 215, 240, 242,
246, 255, 313
extended nonlinear dynamical vector-
function, 86 hierarchical series in the normalized di-
mensionless form, 58
fast oscillation phase, 13 hierarchical system, xviii, xix, 36, 39, 44,
fast rotating phase, 12, 102, 107, 108, 111, 49–58, 60, 62, 64, 67, 69, 71,
119, 130, 138, 145–147, 167, 72, 78, 95, 107, 108, 111, 114,
168, 171, 215, 331 116, 130, 182, 192, 195, 218,
fast variable, 90–92, 101, 102, 119, 134, 239
233, 245, 295 hierarchical theory of oscillations and
fast varying combination phase, 16 waves, 17
field of forces, 31 hierarchical transformations, xviii, 89,
field oscillation phase, 13 218, 256, 318
first integral of motion, 178 hierarchical tree, 51, 52, 66–69, 78, 114–
first law of thermodynamics, 44 116, 185, 196
force, 9, 12–14, 29, 30, 32–34, 51, 110, hierarchy, xviii, xix, xxi, 4, 5, 13, 17,
113, 167, 176, 177, 194, 199– 49, 52–54, 56–58, 61, 64, 73,
202, 249, 295 84, 90, 101, 111, 114, 117, 118,
four-wave parametric resonance, 16 213, 218, 241, 242, 246, 265,
Fourier–Bessel series, 188 316, 317
free electron laser (FEL), 1, 4, 5, 13, 18, highest hierarchical scale parameter, 4
22, 77, 94, 161 holographic principle, 52, 59
functional operator, 64, 73, 75, 84, 88, 96, homogeneous equation, 213
245
improper wave, 27
general hierarchical principle, 58, 60, 64, information entropy, 44, 59, 61, 115
83 initial phase of oscillations, 11
INDEX 353
instability, 14, 49, 249–252, 258, 282, 315, modernized version of the method of
329, 342 slowly varying amplitudes, 293
instantaneous acceleration, 29 moment of inertia, 9, 30
instantaneous velocity, 29, 208 moment of momentum, 30, 32
intensity vector, 34 moment of momentum conservation law,
inverse problem, 64, 65, 71, 72 32
momentum, 5, 29, 30, 32, 33, 79, 80, 111,
kinetic energy, 19, 31, 32, 115, 198–200, 164, 166, 167, 208, 212, 214,
203 219, 232, 236, 237, 239, 245
kinetic equation, 59, 79–81, 83, 208, 236, momentum conservation law, 32
244, 245, 310, 311, 313, 326 motion integral, 32, 174, 178
klystron, 251, 252 motion integral of the first hierarchical
Kramer’s formula, 296, 300 level, 174
Krylov–Bogolyubov’s substitution, 88, multi-frequency nonlinear resonant
96, 117, 120, 127, 150, 153, oscillation–wave problem, 3
155, 173, 189, 215, 220, 221, multi-frequency nonlinear resonant
233, 234, 239, 242, 244, 245, oscillation–wave systems, 3
256, 265, 268, 275, 321, 332, multi-harmonic periodic oscillations, 22
338, 340 multi-resonance system, 17
multiple resonance, 16
lag-effect, 24, 26
Lagrange oscillation phase, 34
Langmuir’s electron wave, 159 natural hierarchical system, 51, 54, 55, 60,
Lantzos' method, 112 69
large particle, 34, 197, 204 negentropyness, 37
large scale parameter of the problem, 103 new fast oscillatative phase, 184
linear differential equation, xviii, 10, 91 Newtonian formalism, 33
linear frequency, 202, 203 non-bound resonances, 17
longitudinal wave, 26 non-harmonic oscillations, 21, 22
lumped system, 71, 82 non-isochronous oscillator, 181
non-resonant case, 146, 151
magnetic permeability, 34 non-resonant zone, 143
many-particle description, 35 nonlinear dynamical system, 10, 143
mathematical pendulum, 9, 12, 13 nonlinear equation, xviii, 10, 91
Maxwell’s equations, 33, 83, 210, 211, nonlinear oscillatative system, 9
236, 238, 262, 271, 309, 311, nonlinear oscillation dynamical system,
315, 317, 326, 327, 342 10
method of averaged characteristics, xx, 7, normalized combinative phase, 188
72, 89, 207, 213, 222, 225, 231, normalized energy, 181
232, 235–237, 244, 250, 254, normalized potential function, 179, 180
271, 273, 280, 283, 285, 320, normalized time, 188
329, 339
method of averaged current-density equa- one-particle description, 35
tion, 240 open system, 36, 38, 60
method of averaged kinetic equation, 79, oscillation, 21
236, 244
oscillation amplitude, 12
method of averaged quasi-hydrodynamic
oscillation phase, 12
equation, 236
overdetermined system, 212
method of characteristics, 85, 207, 213,
219, 222, 227, 228, 261, 276,
283, 288, 340 parabolic type, 228
method of exact solutions, 186 parametric amplification, 288, 293
microstate, 42 parametric electron-wave lamp, 113
Miller–Gaponov potential, 176, 177, 198, parametric resonance, 16, 113, 193
199 parametric resonant damping, 293
moderately nonlinear system, 11 partial solution, 212
modernized standard system, 309, 317, particle ensemble, 196, 197, 209
320 passage motion of particle, 198
percussive excitation of highest harmon- Russian matryoshka, 53
ics, 258
period of oscillations, 14
period of the oscillations, 3 saturation of amplification, 20, 345
perturbed model, 201, 203, 204 SCW, 28, 254
phase grouping, 204, 205 second law of dynamics, 9, 29, 31, 33, 177
phase mismatch, 291 second law of thermodynamics, 44
phase velocity, 25, 26, 160, 161, 164, 194, secular term, 99, 100, 120, 129, 133, 190
288 sefirot, 51
physical definition of the concept of God, self-consistent system, 344
68 self-modeling principle, 50, 52, 56, 59, 116
potential energy, 31, 32, 177, 179, 198, 199 self-organization, 50, 204
potential field, 32, 174 self-resemblance, 50, 51
potential function, 32, 176, 177, 179 separatrix, 179, 181, 182
potential well, 174, 177, 179, 199, 202 shortened form, xvii
principle of hierarchical resemblance, 59, simplified version of slowly varying ampli-
61, 83 tude method, 284
principle of information compression, 58, single particle problem, 163
64, 83, 93 slow electron wave, 330
principle of physicality, 38 slow resonance, 17
principle of quasi-stationary interaction, slow variable, 102, 112, 114, 118, 136, 147,
200 167, 169, 170, 233, 238, 313,
problem small parameter, 61, 189, 309, 331
314 slowest dynamical variables, 4
proper field of the zeroth hierarchical slowly varying amplitude, xix, xx, 7, 20,
level, 174 21, 72, 82, 89, 207, 208, 250,
proper oscillations, 13, 249 279, 283–285, 287–289, 291,
proper wave, 27, 109, 250, 290, 295, 297, 293, 295, 297, 302, 305, 309,
299 310, 312, 314, 320, 324, 326,
pumping wave, 14, 26, 161, 288 329, 330, 342, 343, 345
purposefulness, 37, 42, 49, 50 slowly varying amplitude method, 320
purposefulness principle, 38, 42 slowly varying combination phase, 16, 137
slowly varying cyclic frequency, 21
quasi-harmonic oscillations, 21 slowly varying initial oscillation phase, 20
quasi-potential field, 175–177, 180 small oscillations, 10, 203
quasilinear resonance, 16, 193 small parameter of the problem, 16, 91,
186, 189, 322
Rabinovich’s standard form, 85, 307, 310
space charge wave, 254, 277, 305
relationship of characteristics, 228
relativistic factor, 15, 111, 172, 210, 241, spatially one-dimensional model, 288, 294
253, 258, 261, 278, 344 standard equation of zeroth hierarchical
resonance, 13, 14, 18–20, 38, 108, 113, level, 6
114, 130, 134–139, 150, 161, standard system with partial derivatives,
170, 192, 197, 200, 238, 249, 213
250, 284, 291, 296, 305 standing electromagnetic wave, 158, 164,
resonances of the same hierarchy, 17 176, 185
resonant case, 143, 146, 195 static hierarchical system, 54
resonant condition, 15, 16, 18, 20, 143, static system, 36
144, 150. 195, 197, 200, 249, stimulated Duffing’s equation, 146
250, 287, 290, 297 stimulated oscillations, 13–15, 146, 249,
resonant curve, 18, 19 250
resonant point, 18–20, 135, 249 stochastic system, 36, 60, 79, 80, 208
resonant zone, 144 straight problem, 64
rigorous version of the method of slowly strong hierarchical series, 111
varying amplitudes, 285, 294 strong synchronism, 160
rotating motion of the nonlinear pendu- structural complexity, 45, 46
lum, 198 structural hierarchical scale parameter, 53
rotating vector phase, 102 structural hierarchy, 53, 54, 57, 62
INDEX 355
structural operator, 62, 63, 73, 75, 84, 85, uncertainty postulate, 41
99, 109, 111 underdetermined system, 212
successive approximation method, 244 uniqueness, 37, 60, 249
successive approximations, 128 unperturbed model, 198, 204
surface wave, 27
synchronous condition, 162, 164, 170 Van der Pol’s method, xix, 93, 94, 125,
synthesis problem, 64 126, 284
system definition, 35 Van der Pol’s variables, 91, 146, 189, 194
system for optical signals transformation vector of slow variables, 101, 117, 118,
into microwave signals, 159 137, 145, 146, 171, 331
system property, 36, 39, 41, 44, 204 vicinity of the resonant point, 18
system with constant rotation frequency, volumetric wave, 27
171 vortex field, 31, 32
system with fast rotating phases, 102,
103, 111 wave energy density, 28
system with slow and fast variables, 100 wave number, 15, 24–26, 165, 193, 295,
297
thermodynamic probability, 43, 44 wave period, 15, 23, 24
third law of thermodynamics, 44 wavelength, 24, 26, 27
threshold, 41 weak nonlinear dynamical system, 11
transversal wave, 23, 26 weak nonlinear mathematical problem, 21
tree of life, 51, 55, 66 weak nonlinear oscillations, 11–13, 21
two-level hierarchical system, 95, 101, 216 weak predictability, 37
two-multiple resonance, 16 weakly nonlinear equation, 10
two-stream instability, 14, 29, 208, 231,
249–253, 258, 278, 329, 342 zero-level equation, 61
46. P.P.J.M. Schram: Kinetic Theory of Gases and Plasmas. 1991 ISBN 0-7923-1392-5
47. A. Micali, R. Boudet and J. Helmstetter (eds.): Clifford Algebras and their Applications in
Mathematical Physics. 1992 ISBN 0-7923-1623-1
48. Quantum Geometry. A Framework for Quantum General Relativity. 1992
ISBN 0-7923-1640-1
49. M.H. Mac Gregor: The Enigmatic Electron. 1992 ISBN 0-7923-1982-6
50. C.R. Smith, G. J. Erickson and P.O. Neudorfer (eds.): Maximum Entropy and Bayesian Methods.
Proceedings of the 11th International Workshop (Seattle, 1991). 1993 ISBN 0-7923-2031-X
51. D.J. Hoekzema: The Quantum Labyrinth. 1993 ISBN 0-7923-2066-2
52. Z. Oziewicz, B. Jancewicz and A. Borowiec (eds.): Spinors, Twistors, Clifford Algebras and
Quantum Deformations. Proceedings of the Second Max Born Symposium Poland,
1992). 1993 ISBN 0-7923-2251-7
53. A. Mohammad-Djafari and G. Demoment (eds.): Maximum Entropy and Bayesian Methods.
Proceedings of the 12th International Workshop (Paris, France, 1992). 1993
ISBN 0-7923-2280-0
54. M. Riesz: Clifford Numbers and Spinors with Riesz’ Private Lectures to E. Folke Bolinder and
a Historical Review by Pertti Lounesto. E.F. Bolinder and P. Lounesto (eds.). 1993
ISBN 0-7923-2299-1
55. F. Brackx, R. Delanghe and H. Serras (eds.): Clifford Algebras and their Applications in
Mathematical Physics. Proceedings of the Third Conference (Deinze, 1993) 1993
ISBN 0-7923-2347-5
56. J.R. Fanchi: Parametrized Relativistic Quantum Theory. 1993 ISBN 0-7923-2376-9
57. A. Peres: Quantum Theory: Concepts and Methods. 1993 ISBN 0-7923-2549-4
58. P.L. Antonelli, R.S. Ingarden and M. Matsumoto: The Theory of Sprays and Finsler Spaces
with Applications in Physics and Biology. 1993 ISBN 0-7923-2577-X
59. R. Miron and M. Anastasiei: The Geometry of Lagrange Spaces: Theory and Applications.
1994 ISBN 0-7923-2591-5
60. G. Adomian: Solving Frontier Problems of Physics: The Decomposition Method. 1994
ISBN 0-7923-2644-X
61. B.S. Kerner and V.V. Osipov: Autosolitons. A New Approach to Problems of Self-Organization
and Turbulence. 1994 ISBN 0-7923-2816-7
62. G.R. Heidbreder (ed.): Maximum Entropy and Bayesian Methods. Proceedings of the 13th
International Workshop (Santa Barbara, USA, 1993) 1996 ISBN 0-7923-2851-5
63. Z. Hradil and Quantum Optics and Fundamentals of Physics. 1994
ISBN 0-7923-3000-5
64. M. Evans and J.-P. Vigier: The Enigmatic Photon. Volume 1: The Field B(3). 1994
ISBN 0-7923-3049-8
65. C.K. Raju: Time: Towards a Constistent Theory. 1994 ISBN 0-7923-3103-6
66. A.K.T. Assis: Weber’s Electrodynamics. 1994 ISBN 0-7923-3137-0
67. Yu. L. Klimontovich: Statistical Theory of Open Systems. Volume 1: A Unified Approach to
Kinetic Description of Processes in Active Systems. 1995 ISBN 0-7923-3199-0;
Pb: ISBN 0-7923-3242-3
68. M. Evans and J.-P. Vigier: The Enigmatic Photon. Volume 2: Non-Abelian Electrodynamics.
1995 ISBN 0-7923-3288-1
69. G. Esposito: Complex General Relativity. 1995 ISBN 0-7923-3340-3
70. J. Skilling and S. Sibisi (eds.): Maximum Entropy and Bayesian Methods. Proceedings of the
Fourteenth International Workshop on Maximum Entropy and Bayesian Methods. 1996
ISBN 0-7923-3452-3
71. C. Garola and A. Rossi (eds.): The Foundations of Quantum Mechanics Historical Analysis
and Open Questions. 1995 ISBN 0-7923-3480-9
72. A. Peres: Quantum Theory: Concepts and Methods. 1995 (see for hardback edition, Vol. 57)
ISBN Pb 0-7923-3632-1
73. M. Ferrero and A. van der Merwe (eds.): Fundamental Problems in Quantum Physics. 1995
ISBN 0-7923-3670-4
74. F.E. Schroeck, Jr.: Quantum Mechanics on Phase Space. 1996 ISBN 0-7923-3794-8
75. L. de la Peña and A.M. Cetto: The Quantum Dice. An Introduction to Stochastic Electro-
dynamics. 1996 ISBN 0-7923-3818-9
76. P.L. Antonelli and R. Miron (eds.): Lagrange and Finsler Geometry. Applications to Physics
and Biology. 1996 ISBN 0-7923-3873-1
77. M.W. Evans, J.-P. Vigier, S. Roy and S. Jeffers: The Enigmatic Photon. Volume 3: Theory and
Practice of the B(3) Field. 1996 ISBN 0-7923-4044-2
78. W.G.V. Rosser: Interpretation of Classical Electromagnetism. 1996 ISBN 0-7923-4187-2
79. K.M. Hanson and R.N. Silver (eds.): Maximum Entropy and Bayesian Methods. 1996
ISBN 0-7923-4311-5
80. S. Jeffers, S. Roy, J.-P. Vigier and G. Hunter (eds.): The Present Status of the Quantum Theory
of Light. Proceedings of a Symposium in Honour of Jean-Pierre Vigier. 1997
ISBN 0-7923-4337-9
81. M. Ferrero and A. van der Merwe (eds.): New Developments on Fundamental Problems in
Quantum Physics. 1997 ISBN 0-7923-4374-3
82. R. Miron: The Geometry of Higher-Order Lagrange Spaces. Applications to Mechanics and
Physics. 1997 ISBN 0-7923-4393-X
83. and A.S. Shumovsky (eds.): Quantum Optics and the Spectroscopy of Solids.
Concepts and Advances. 1997 ISBN 0-7923-4414-6
84. A. Sitenko and V. Tartakovskii: Theory of Nucleus. Nuclear Structure and Nuclear Interaction.
1997 ISBN 0-7923-4423-5
85. G. Esposito, A.Yu. Kamenshchik and G. Pollifrone: Euclidean Quantum Gravity on Manifolds
with Boundary. 1997 ISBN 0-7923-4472-3
86. R.S. Ingarden, A. Kossakowski and M. Ohya: Information Dynamics and Open Systems.
Classical and Quantum Approach. 1997 ISBN 0-7923-4473-1
87. K. Nakamura: Quantum versus Chaos. Questions Emerging from Mesoscopic Cosmos. 1997
ISBN 0-7923-4557-6
88. B.R. Iyer and C.V. Vishveshwara (eds.): Geometry, Fields and Cosmology. Techniques and
Applications. 1997 ISBN 0-7923-4725-0
89. G.A. Martynov: Classical Statistical Mechanics. 1997 ISBN 0-7923-4774-9
90. M.W. Evans, J.-P. Vigier, S. Roy and G. Hunter (eds.): The Enigmatic Photon. Volume 4: New
Directions. 1998 ISBN 0-7923-4826-5
91. M. Rédei: Quantum Logic in Algebraic Approach. 1998 ISBN 0-7923-4903-2
92. S. Roy: Statistical Geometry and Applications to Microphysics and Cosmology. 1998
ISBN 0-7923-4907-5
93. B.C. Eu: Nonequilibrium Statistical Mechanics. Ensembled Method. 1998
ISBN 0-7923-4980-6
94. V. Dietrich, K. Habetha and G. Jank (eds.): Clifford Algebras and Their Application in Math-
ematical Physics. Aachen 1996. 1998 ISBN 0-7923-5037-5
95. J.P. Blaizot, X. Campi and M. Ploszajczak (eds.): Nuclear Matter in Different Phases and
Transitions. 1999 ISBN 0-7923-5660-8
96. V.P. Frolov and I.D. Novikov: Black Hole Physics. Basic Concepts and New Developments.
1998 ISBN 0-7923-5145-2; Pb 0-7923-5146
97. G. Hunter, S. Jeffers and J-P. Vigier (eds.): Causality and Locality in Modern Physics. 1998
ISBN 0-7923-5227-0
98. G.J. Erickson, J.T. Rychert and C.R. Smith (eds.): Maximum Entropy and Bayesian Methods.
1998 ISBN 0-7923-5047-2
99. D. Hestenes: New Foundations for Classical Mechanics (Second Edition). 1999
ISBN 0-7923-5302-1; Pb ISBN 0-7923-5514-8
100. B.R. Iyer and B. Bhawal (eds.): Black Holes, Gravitational Radiation and the Universe. Essays
in Honor of C. V. Vishveshwara. 1999 ISBN 0-7923-5308-0
101. PL. Antonelli and T.J. Zastawniak: Fundamentals of Finslerian Diffusion with Applications.
1998 ISBN 0-7923-5511-3
102. H. Atmanspacher, A. Amann and U. Müller-Herold: On Quanta, Mind and Matter Hans Primas
in Context. 1999 ISBN 0-7923-5696-9
103. M.A. Trump and W.C. Schieve: Classical Relativistic Many-Body Dynamics. 1999
ISBN 0-7923-5737-X
104. A.I. Maimistov and A.M. Basharov: Nonlinear Optical Waves. 1999
ISBN 0-7923-5752-3
105. W. von der Linden, V. Dose, R. Fischer and R. Preuss (eds.): Maximum Entropy and Bayesian
Methods Garching, Germany 1998. 1999 ISBN 0-7923-5766-3
106. M.W. Evans: The Enigmatic Photon Volume 5: O(3) Electrodynamics. 1999
ISBN 0-7923-5792-2
107. G.N. Afanasiev: Topological Effects in Quantum Mecvhanics. 1999 ISBN 0-7923-5800-7
108. V. Devanathan: Angular Momentum Techniques in Quantum Mechanics. 1999
ISBN 0-7923-5866-X
109. P.L. Antonelli (ed.): Finslerian Geometries A Meeting of Minds. 1999 ISBN 0-7923-6115-6
110. M.B. Mensky: Quantum Measurements and Decoherence Models and Phenomenology. 2000
ISBN 0-7923-6227-6
111. B. Coecke, D. Moore and A. Wilce (eds.): Current Research in Operation Quantum Logic.
Algebras, Categories, Languages. 2000 ISBN 0-7923-6258-6
112. G. Jumarie: Maximum Entropy, Information Without Probability and Complex Fractals. Clas-
sical and Quantum Approach. 2000 ISBN 0-7923-6330-2
113. B. Fain: Irreversibilities in Quantum Mechanics. 2000 ISBN 0-7923-6581 -X
114. T. Borne, G. Lochak and H. Stumpf: Nonperturbative Quantum Field Theory and the Structure
of Matter. 2001 ISBN 0-7923-6803-7
115. J. Keller: Theory of the Electron. A Theory of Matter from START. 2001
ISBN 0-7923-6819-3
116. M. Rivas: Kinematical Theory of Spinning Particles. Classical and Quantum Mechanical
Formalism of Elementary Particles. 2001 ISBN 0-7923-6824-X
117. A.A. Ungar: Beyond the Einstein Addition Law and its Gyroscopic Thomas Precession. The
Theory of Gyrogroups and Gyrovector Spaces. 2001 ISBN 0-7923-6909-2
118. R. Miron, D. Hrimiuc, H. Shimada and S.V. Sabau: The Geometry of Hamilton and Lagrange
Spaces. 2001 ISBN 0-7923-6926-2
119. The Landscape of Theoretical Physics: A Global View. From Point Particles to the
Brane World and Beyond in Search of a Unifying Principle. 2001 ISBN 0-7923-7006-6
120. R.M. Santilli: Foundations of Hadronic Chemistry. With Applications to New Clean Energies
and Fuels. 2001 ISBN 1 -4020-0087-1
121. S. Fujita and S. Godoy: Theory of High Temperature Superconductivity. 2001
ISBN 1-4020-0149-5
122. R. Luzzi, A.R. Vasconcellos and J. Galvão Ramos: Predictive Statitical Mechanics. A Nonequi-
librium Ensemble Formalism. 2002 ISBN 1-4020-0482-6
KLUWER ACADEMIC PUBLISHERS - DORDRECHT / BOSTON / LONDON

(Fundamental Theories of Physics) V. Kulish - Hierarchical Methods - Hierarchy and Hierarchical Asymptotic Methods in Electrodynamics. Volume 1 - Kluwer Academic Publ (2002)

Uploaded by

Copyright:

Available Formats

(Fundamental Theories of Physics) V. Kulish - Hierarchical Methods - Hierarchy and Hierarchical Asymptotic Methods in Electrodynamics. Volume 1 - Kluwer Academic Publ (2002)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Fundamental Theories of Physics) V. Kulish - Hierarchical Methods - Hierarchy and Hierarchical Asymptotic Methods in Electrodynamics. Volume 1 - Kluwer Academic Publ (2002)

Uploaded by

Copyright:

Available Formats

Hierarchical Methods

Fundamental Theories of Physics

An International Book Series on The Fundamental Theories of Physics:

Editorial Advisory Board:

KLUWER ACADEMIC PUBLISHERS

©2004 Kluwer Academic Publishers

Print ©2002 Kluwer Academic Publishers

All rights reserved

Created in the United States of America

Visit Kluwer Online at: http://kluweronline.com

3.4 Transverse and Longitudinal Waves 26

1.3 Self-Modeling Principle 51

5.2 Sewing Together of Resonant and

3.2 Duffing Oscillator 188

6. HIERARCHICAL SYSTEMS WITH

1 MAIN IDEAS OF THE METHOD OF

7 EXAMPLE: THE TWO-STREAM

Everybody is current in a world surrounded by computer. Computers

some well known conventional numerical or analytical methods. The hi-

The version of hierarchical theory [10, 11, 19–25], which is described

erators (UNIACs or EH-accelerators) and Free Electron Lasers (FELs))

[1] J.S. Nicolis. Dynamics of hierarchical systems. an evolutionary approach.

[2] R. Rammal, G. Toulouse, M.A. Virasoro. Ultrametricity for physicists. Reviews

[3] H. Haken. Advanced Synergetic. Instability Hierarchies of Self-Organizing Sys-

[9] Interpretation the hierarchy of nature: from systematic pattern to evolutionary

[11] V.V. Kulish. Methods of averaging in non-linear problems of relativistic electro-

[13] N.M. Krylov, N.N. Bogolyubov. Application of methods of nonlinear mechanics

[14] N.M. Krylov, N.N. Bogolyubov. Introduction to nonlinear mechanics. Kiev:

[15] M.A. Leontovich. To the problem about propagation of electromagnetic waves

[16] N.N. Moiseev. Asymptotic methods of nonlinear mechanics. Nauka, Moscow,

[17] E.A. Grebennikov. Averaging method in applied problems. Nauka, Moscow,

[22] V.V. Kulish. Hierarchical approach to nonlinear problems of electrodynamics.

[27] C. Brau. Free electron laser. Boston: Academic Press, 1990.

[29] A.N. Kondratenko, V.M. Kuklin. Principles of plasma electronics. Energoat-

[35] J. Weiland, H. Wilhelmsson. Coherent nonlinear interactions of waves in plas-

GENERAL IDEAS, CONCEPTS,

1. ESSENCE OF THE HIERARCHICAL

be essentially different from some characteristic scale parameter (the

In experiments these periods are usually characterized by values

The simplest analysis shows that the general calculational situation

Inequality (1.1.3) means that we should take many more than

1.2 Essence of the Hierarchical Approach

where is the corresponding vector (electron coordinates,

function, and are the phases of the electron oscillations in pumping

It is clear that the above discussed calculational problem, concern-

1.3 Classification of the Hierarchical

effective for solving of a number of complex nonlinear wave resonant

2. BASIC CONCEPTS AND DEFINITIONS

2.1 Nonlinear Oscillations and Nonlinear Systems

where is the of moment of the force vector

is the free fall acceleration vector, is its space radius-

is the moment of inertia with respect to the is the vector

collinear to the normal to the drawing plane), is the magnitude

where is the pendulum length. After corresponding transformations

where the physical meaning of the constant

Accounting the first term in (1.2.7) only, we reduce the equation

The solution to (1.2.8) can be obtained easily using Euler’s substitu-

where is the angular amplitude, is the phase of lin-

where The nonlinear pendulum equation (1.2.5) as well as