GENERAL SYSTEMS THEORY
A MATHEMATICAL APPROACH

International Federation for Systems Research International Series on Systems Science and Engineering
Series Editor: George J. Klir State University of New York at Binghamton Editorial Board Gerrit Broekstra Erasmus University, Rotterdam, The Netherlands John L. Casti Santa Fe Institute, New Mexico Brian Gaines University of Calgary, Canada

Ivan M. Havel Charles University, Prague, Czech Republic Manfred Peschel Academy of Sciences, Berlin, Germany Franz Pichler University of Linz, Austria

Volume 8

THE ALTERNATIVE MATHEMATICAL MODEL OF LINGUISTIC SEMANTICS AND PRAGMATICS Vilém Novák CHAOTIC LOGIC: Language, Thought, and Reality from the Perspective of Complex Systems Science Ben Goertzel THE FOUNDATIONS OF FUZZY CONTROL Harold W. Lewis, III FROM COMPLEXITY TO CREATIVITY: Explorations in Evolutionary, Autopoietic, and Cognitive Dynamics Ben Goertzel GENERAL SYSTEMS THEORY: A Mathematical Approach Yi Lin PRINCIPLES OF QUANTITATIVE LIVING SYSTEMS SCIENCE James R. Simms INTELLIGENT ROBOTIC SYSTEMS: Design, Planning, and Control Witold Jacak

Volume 9

Volume 10 Volume 11

Volume 12 Volume 13

Volume 14

IFSR was established “to stimulate all activities associated with the scientific study of systems and to coordinate such activities at international level.” The aim of this series is to stimulate publication of high-quality monographs and textbooks on various topics of systems science and engineering. This series complements the Federation’s other publications.
A Continuation Order Plan is available for this series. A continuation order will bring delivery of each new volume immediately upon publication. Volumes are billed only upon actual shipment. For further information please contact the publisher. Volumes 1–6 were published by Pergamon Press.

GENERAL SYSTEMS THEORY
A MATHEMATICAL APPROACH

YI LIN
Slippery Rock University Slippery Rock, Pennsylvania

KLUWER ACADEMIC PUBLISHERS
NEW YORK, BOSTON , DORDRECHT, LONDON , MOSCOW

eBook ISBN
Print ISBN

0-306- 46962-6 0-306-45944-2

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

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: and Kluwer's eBookstore at:

http://www.kluweronline.com http://www.ebooks.kluweronline.com

Preface
As suggested by the title of this book, I will present a collection of coherently related applications and a theoretical development of a general systems theory. Hopefully, this book will invite all readers to sample an exciting and challenging (even fun!) piece of interdisciplinary research, that has characterized the scientific and technological achievements of the twentieth century. And, I hope that many of them will be motivated to do additional reading and to contribute to topics along the lines described in the following pages. Since the applications in this volume range through many scientific disciplines, from sociology to atomic physics, from Einstein’s relativity theory to Dirac’s quantum mechanics, from optimization theory to unreasonable effectiveness of mathematics to foundations of mathematical modeling, from general systems theory to Schwartz’s distributions, special care has been given to write each application in a language appropriate to that field. That is, mathematical symbols and abstractions are used at different levels so that readers in various fields will find it possible to read. Also, because of the wide range of applications, each chapter has been written so that, in general, there is no need to reference a different chapter in order to understand a specific application. At the same time, if a reader has the desire to go through the entire book without skipping any chapter, it is strongly suggested to refer back to Chapters 2 and 3 as often as possible. The motivation to write this book came from the strong influence of historical works by L. von Bertalanffy, George Klir, and M. D. Mesarovic, and the book On Systems Analysis, by David Berlinski (MIT Press, Cambridge, Massachusetts, 1976). Berlinski’s book and challenges from several scholars really made me decide to write such a book with strong applications in different scientific fields in order to justify the very meaning of existence for a general systems theory. At the same time, one of the important lessons we have learned from the severaldecade-old global systems movement, started and supported by many of the most powerful minds of our modern time, is that senseless transfer of statements (more specifically, theoretical conclusions or results) from one discipline to another makes people feel that general systems theory is a doubtful subject. To keep such unnecessary situations from occurring, we develop each application with rigorous
vii

viii

Preface

logical reasoning. Whenever a bold conclusion is deduced, some relevant gaps in the reasoning process will be pointed out right on the spot or in the final chapter (“Some Unsolved Problems in General Systems Theory”). On the other hand, doubtful people will be as doubtful as they can no matter what facts or evidence are out there to show them their doubt is unfounded. For example, more than 100 years ago, when naive set theory was first introduced and studied, many first-class mathematicians did not treat it as a serious theory at all. Furthermore, Cantor, the founder, was personally attacked by these scholars. As a consequence, he was hospitalized and eventually died in a psychiatric hospital. Today, set theory has succeeded in a great many areas of modern science, including the entire spectrum of mathematics, when the central idea of infinity is employed in systems science, we can still hear doubters saying things like: Infinity? One can be sure that in an infinitely long period of time, a monkey will produce the great Beethoven’s music! (A note: according to results in set theory, this statement is not true!) The structure of my theoretical development in this book is the “top-down” — formalization — approach, launched in 1960 by Mesarovic. This approach is characterized by the following: (1) All concepts are introduced with minimal mathematical structures. (2) Additional mathematical conditions are added when necessary to display the richness of systems properties. At the same time, applicability is always used to test the mathematical conditions added. Calculus is all that is needed to comprehend this book, since all other mathematical techniques are presented at appropriate levels. Finally, I would to express my sincere appreciation to many individuals, too many to list. My thanks go to President Robert Aebersold and Vice President and Provost Charles Foust, Deans Charles Zuzak and Jay Harper of Slippery Rock University, Pennsylvania, whose academic support for the past several years was essential to finishing this book. I thank Dr. Ben Fitzpatrick, my Ph.D. supervisor, for his years’ teaching and academic influence, Professor Lotfi Zadeh, the father of fuzzy mathematics, for his keen encouragement, Professor Xavier J. R. Avula, President of the International Association for Mathematical and Computer Modeling, for his personal influence and education on professional perfection for the past several years. I hope you enjoy using and referencing this book, and your comments and suggestions are welcome! Please let me hear from you — my e-mail address is jeffrey.forrest@sru.edu. Yi Lin

Principia Scientia (St. (Oxford. (London). Springer-Verlag (London). Switzerland. International Federation for Systems Research (Vienna. MCB University Press (Bradford. and I am grateful to the copyright owners for permitting me to use the material. Belgium). Poland). Kluwer Academic Publishers (Dordrecht. and New York). Pergamon Journals.K.Acknowledgments This book contains many research results previously published in various sources. Louis). World Scientific Press (Singapore and New Jersey). England).). and Wroclaw Technical University Press (Wroclaw. Austria). ix . They are International Association for Cybernetics (Namur. Ltd. Netherlands). U. Ltd. Taylor and Francis. Hemisphere (New York). Gordon and Breach Science Publishers (Yverdon.

. . . . 4. . . . . . . . . . . . Several Aspects of Systems Theory . . . . . . . . . . . . . . . . . . . . . Centralizability and Tests of Applications . . . . Growth of the Polish School of Mathematics . . . . . Ordered Sets . A Few Thoughts . . .4. . . . . . . . . . . . . . . . . . . . . . . 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. . . . . . . . . . . . . . . . . . . . . . . . 1. . . . . . . . . 2. . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . Constructions of Ordinal and Cardinal Number Systems . . . .3. . .6. . . . . . . . 2. . A Set and Its Classification . Introduction . . . Some Philosophical Issues . . . . 2. . . . . . . . . . . . . . . Crises of Naive Set Theory . . . . . . . . . . . . . .4.5. . . . . . . . . . . . . . . . . . . . . . . . . . . .1. . . . . . .2. . . . . Centralized Systems and Centralizability . . . . . . . . . . 3. . . . . . 4. . . . 3. . . . . . . . . . . 4. xi 1 2 6 14 15 15 19 23 32 42 56 59 59 65 77 90 96 97 97 99 100 104 106 108 . . . . . . . . . . . . . . . . . . . . . . . . Well-Founded Sets . . . . . . . . . . . . . . . . 4. . . . . . . .4. . . . . . . 2. . . . . . . . . . . Several Tests of Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . .1. 3. . . . . . . . . . . . . . 3. . References for Further Study . The Appearance of Nicolas Bourbaki . . . . . . . . . 3. . . Some Related Problems and a Few Final Words . . . . . . . . . . Historical Background . .5. . .Contents 1. . . . . . . . . . . . . 1. . . . . . 4. . . . . Families of Sets . . . . . . . . Naive Set Theory . . . . 2. . . . . . . . . . . . . . . . .2. .6. . .2. . . . . . 1. . . . . . . . . . . . . . . . . Axiomatic Set Theory . . . . 4. . . . Well-Ordered Sets . . . . . . . . . . . . . . . . . . . .2. . . . . . . . . . . . . . .5. . . . . . . . . . . . . . . . . . . . . . .1. Arithmetic of Cardinal Numbers . . . . . . . . . .3. . . . . . . . . . . . . . . . . . . . . . .3. . . . . . . . . . . . . . Operations of Sets . . . .1. . . 4. . .3. . . . . .

. ..4... ... . 9. What Is Mathematics? . .. Fundamental Equations of Generalized Systems . . The Concept of General Systems . .. ... 7. . ......... . .. . .......3.. .. .. .8. . ........... .. ...6.. . .... . . . ... ....5.. .. . .. .4......... ... Some Epistemological Problems 8.. . . . 9. . ... .. .. Controllabilities 111 111 112 115 119 121 121 122 124 125 129 134 135 135 136 138 140 143 147 150 157 161 163 163 166 171 174 176 177 188 191 191 200 204 214 224 235 ... . ... .. ..... ... .. ..... . A Mathematical Foundation for the Laws of Conservation . 8. . . .. .. . . ... . . ......... ....... Constructions of Systems . .. Hierarchies of Systems ....... ... . ...... .... 7... ..3.. . .. ..... ... ..... .... . .. .. .. 8... 8. . . .... .. . ..... . A Theoretical Foundation for the Laws of Conservation ... . .... Elements............ ....... .. ... . ..2. ...... Multilevel Structure of Nature . .. ..... .. 6... .1.. ...... ..... .. ......1..... . .. . .... .2....... .... . Introduction . .. 6...... ......7. .. Unreasonable Effectiveness of Mathematics: A New Tour ..... A Mathematics of Computability that Speaks the Language of Levels . .5. Final Comments and Further Questions . .. 5.. .. 8.. .... .. . . .... ... 7.......... Mathematics from the Viewpoint of Systems .. .2.... ........ ..... .. ... .. . ... .... . .. .4.. . ..... . .. ... ... .. .1..... ... Operation Epitome Principle 7... .. . 5..3.. . 9.... . . . A Description of the State of Materials ... ... ..4... Some Partial Answers and a Conjecture ... . 7. . . .. 5..... ........ ...... .. . . .... ... .... .. . ..xii Contents 5. ... .. . Hu’s Counterexample and Analysis .. . .. Introduction . . 8.. .. ... . .. . ...2...... 9. ... . .... 7... Some Final Words . .. .. . Some Hypotheses in Modern Physics . . General Systems: A Multirelation Approach .. . ... .. .. 7. 7.. .. ...... . ...3. ......1... ... . . .... A Non-Archimedean Number Field . 6. ...... . ...... ....1..5...4. .. 8.... .... .... .... .. . . .. .. .. .. . . .. . ..... .. .. 7. Generalization of Fundamental Equations . . . ...... . and Laws of Conservation .... . . .. .. Brief History of Atoms. ... . . .. 7.7... . .. .. Generalized Principle of Optimality .. Applications ... .. ..... The Vase Puzzle and Its Contradictory Mathematical ... ... 6... ... Applications .... . . . 6.. . .2.... ..3... ..... Mappings from Systems into Systems ..... .. . . .... ..... . ......6.... .. ... . ...6..... .. . .. .. 6. ... . A Brief Historical Note .. Bellman’s Principle of Optimality and Its Generalizations .5. .... Conclusion 8. Modelings . ..... 6..... 5. .. .. .... ... ...9...... . .. 9. . .... ... Structures of Systems .. . . ..6. .. A Few Final Words .. ... . .... . . . 9. The Construction of Mathematics ... . . .. .. .. ... 9. ... . .....

. . . . ... . . . . . . .. . . . .. . Some Unsolved Problems in General Systems Theory . . .. . . . . .. . . . . . .. . .2. . .. . . . .. . . . .. .. . . .. . . . .. . . . . . . . . . . . . . .. . .. . . . . . . . Decomposability Conditions . .3. . . .. . . . . . . . . . .. . . . . . . Feedback-Invariant Properties . .6. . . .. . .. .. .. . . . . . 9. . . . . . . . . . . . . . . . .. . . .. 11. . .. . . . .. . ..8. . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . .2.. . . . . . . . . . .5. Mathematical Modeling and the Origin of the Universe . .. . . . . . . . . . . . . . 12. . . . . . .. . . . .. . 11. . .. Laws of Conservation and the Multilevel Structure of Nature . . . . ..3. . . . 245 254 255 255 263 271 280 291 309 311 311 317 321 325 335 344 345 347 347 348 351 356 358 363 367 369 379 . . . .6. 10. . . 10. . . .. . .. . . . . . . .. . Integral Calculus . .. . . . . . . . .. . . Introduction . . . . .. . . . . . . . . . . . .Introduction . The Concept of Systems . . . . . . . . . 10. . . . .. .6. . . .3. . . . . . . . . .. . . . . . . . . . . . . Schwartz Distribution and GNS . . . . . . . . .. . . . . . . . . . . .. . .. .5. . . . . . . . . . . . . . . ... . . . . . . . . .. . References for Further Study . . .. 11. .. .. . . . . .. . . . .. . . . . . . . . . Feedback Transformations . . .. . . . . .. . . . .. . .. Limit Systems . . . . . . . ... . . .4. . .1. . . . . . 10. References Index . . . . . .. . . . 11. . . . . .. . . Set-Theoretic General Systems Theory . . 12. . . . . ....4. . . . . . .. .Contents xiii 9. .. 10.. .. . . 11. . . 12. . Differential Calculus . . . .1. . .. .. . . . . . . . .2. 12. . . Some Leads for Further Research . . . . . . . . Chaos and Attractors . . . . . . 11. . . . . . . . .. . Some References for Further Study .. . . . . .. . . Continuity . . Analytic Foundation for an Applicable General Systems Theory . . . . .. . . . . . 10.. 12.. . . . . . . .. . . . . . . .4. 10. .. 12. .. . . . .. . . A Few Final Words . . . . . . . Systems of Single Relations . . . . . . . . . .. . . .5.. . . . . . . . . . . . . . . . . .. . . . . .7. . . .7. . . .. . . . . . . . . . . .. .. . ... . . . . . . . . . . . . . . A Bit of History . . . . . 12. . . . . . . . . .. . . . . . . . . Calculus of Generalized Numbers .. . . . . . . . . . . . . . . .. . . . . . . 11. . . . . .. . . . . . . . . . . .. . . .. 12. . . .l. . . . 11. . . . . . . . Decoupling of Single-Relation Systems . . . . . . ..7. . . . .. . . . . . .

In particular. a new scientific theory — science of sciences — was born (Tian and Wang. von Bertalanffy (1934) wrote: since the fundamental character of living things is its organization.” From this statement it can 1 . the customary investigation of the single parts and processes cannot provide a complete explanation of the vital phenomena. administrators of scientific research are interested in whether science itself can be studied as a social phenomenon. That is. yet the boundaries of disciplines become blurred. Parallel with the science of sciences some natural questions arise.” what really are their characteristics? The most important and most obvious characteristic is that the forest of specialized and interdisciplinary disciplines can be easily seen. Because of the rapid progress of science and technology. is that the esthetic standard of scientific workers has been changing constantly. 1985). We believe that the attempts to find a foundation for the theoretical point at a fundamental change in the world picture. the “overall” trend of modern science and technology is to synthesize all areas of knowledge into a few major blocks. the amount of which has been increasing in a geometric series? If traditional research methods and esthetic standards are employed without modification. considered as a method of investigation. 1980). researchers and administrators of scientific research have been faced with unprecedented problems: How can they equip themselves with the newest knowledge? How can they handle new knowledge. we call “organismic biology” and. as evidenced by the survey of an American national committee on scientific research in 1985 (Mathematical Sciences. “the system theory of the organism. In this new environment. which is not obvious but dominates the development of modern science and technology. Can the concept of wholeness or connection be used to study similar problems in different scientific fields? Can the point of view of interconnections be employed to observe the world in which we live? In the second decade of this century. Another characteristic. This investigation gives us no information about the coordination of parts and processes.CHAPTER 1 Introduction Although people often talk about “modern science” and “modern technology. as an attempt at an explanation. Thus the chief task of biology must be to discover the laws of biological systems (at all levels of organization). then today’s unpublished scientific results could become outdated tomorrow. This view.

The objects of systems are classified into several categories: motivators. Tested in the last seventy some years. sensors and effectors do what they are supposed to do. policies.1. systems identification.2 Chapter 1 be seen that parallel to the challenge of modern science and technology. It employs the concepts of a black box and a white box (a white box is a known object. Systems methodology in Zadeh’s viewpoint is that systems theory is an independent scientific discipline whose job is to develop an abstract foundation with concepts and frames in order to study various . Through a set of rules. Scholars in the field understand it in different ways. Quastler (1965) wrote: generally speaking. In the following. see (Blauberg et al. (2) A system’s efficiency is stimulated by its internal movements and the reception of information about its environment. stability. For a detailed treatment on the history of systems theory. 1968). showing that research problems. systems optimization. systems methodology is essentially the establishment of a structural foundation for four kinds of theories of organization: cybernetics.. can be represented as white boxes. and controllability. Sensors are the elements of the system that receive information. systems analysis. appearing in the aforementioned theories on organizations. please refer to (von Bertalanffy. In this chapter we analyze the historical background for systems theory and several important research directions. reflecting the efficiency of the system’s given input and output). However. Quastler proved that the common structure of the four theories of organization can be described by the following laws: (1) Interactions are between systems and between systems and their environments. and their environments as black boxes. loosely speaking. learning and adaptation. Zadeh. systems control and programming. and information theory. Zadeh (1962) listed some important problems in systems theory as follows: systems characteristics. signal representation. 1977)]. By using these objects. let us first look at typical opinions of the two important figures in modem science: H. 1. Historical Background Before discussing the historical background of systems theory. decision theory. this concept has been widely accepted by the entire spectrum of science and technology [for details. and effectors. and effectors are the elements of the system that produce real reactions. let us first talk about what systems methodology is. signal classification. objects needed by the system to produce. which is formed in certain way. systems liability. 1972. Quastler and L. The characteristics of his opinion are that the main task of systems theory is the study of general properties of systems without considering their physical specifics. von Bertalanffy. and regulations. systems synthesis. game theory. sensors. a new concept of “systems” had been proposed formally. systems classification. their fundamental points are the same.

are still among the problems of today’s systems theory research. Nicholas of Cusa. in the way of the natural philosophers of the nineteenth century. where Descartes’ second principle says: divide each problem into parts as small as possible. seek interactions and connections between phenomena. systems theory is a theory of mathematical structures of systems with the purpose of studying the foundation of organizations and systems structures. thus romantically anticipating the ecosystems of modern parlance. That is. and was first introduced formally in the second decade of this century in biology (von Bertalanffy. Only in modern times have some new contents been added to the ancient systems thinking. can be traced to the time of Yellow Emperor about 2800 years ago. For example. such as “the whole is greater than the sum of its parts. 1924). what is a “whole”? What does “the sum of its parts” mean? All these problems have not been studied in all classical branches of science. supraindividual organizations of higher order than the usual objects of observation — for example. For a more comprehensive study. etc. Leibniz’s hierarchy of monads looks quite like that of modern systems. elaborated. antithesis. and synthesis. For example.” the order and goal directedness of living things. his mathesis universalis presages an expanded mathematics which is not limited to quantitative or numerical expressions and is able to formulate much conceptual thought. For example. treating each human body as a whole. Gustav Fechner. The methodology of studying systems as wholes adequately agrees with the development trend of modern science. Hegel and Marx emphasized the dialectic structure of thought and of the universe it produces: the deep insight that no proposition can exhaust reality but only approaches its coincidence of opposites by the dialectic process of thesis. we can still find many explanatory terms concerning the concept of systems. problems contained in it. the ideas and thinking logic of systems have a long history. only a few names are listed. Even though the concept of systems has been a hot spot for discussion in almost all areas of modern science and technology. Chinese traditional medicine. and Aristotle’s statement “the whole is greater than the sum of its parts” has been a fundamental problem in systems theory. mankind has been studying and exploring nature by using the thinking logic of systems. namely to divide the object of consideration into parts as small as possible and studying all of them. In the history of science. linking Medieval mysticism with the first beginning of modern science. see (von Bertalanffy. the opposition or indeed fight among the parts within a whole which nevertheless forms a unity of higher order. that profound thinker of the fifteenth century. and . life communities and the entire earth. and to observe and comprehend more and bigger pictures of nature. over the centuries. Here. 1972). since these branches have been established on Descartes’ second principle and Galileo’s method. introduced the notion of coincidentia oppositorum. Therefore.. as all new concepts in science. known as the author of the psychophysical law. Even though Aristotelian teleology was eliminated in the development of modern science.Introduction 3 behaviors of different kinds of systems. although the word “system” was never emphasized.

In a certain sense. and a higher level of difficulty. what changes have occurred in modern science? Under the name “systems research”. the world is not a pile of infinitely many isolated objects. Should we continue to use Descartes’ second principle and Galileo’s method? Yes. 1962)]. First. so one feels the tendency of synthetic development in scientific activities. von Bertalanffy described the scientific revolution in the sixteenth century as follows: “(The Scientific Revolution of the sixteenth to seventeenth centuries) replaced the descriptive-metaphysical conception of the universe epitomized in Aristotle’s doctrine by the mathematical—positivistic or Galilean conception. they have been extremely effective in scientific research and administration. which could be treated individually. for two reasons. 1967. This quest has been recognized in the human struggle for survival with nature. can we describe the change in today’s science and technology as follows? At the same time of continuously using Descartes’ second principle and Galileo’s method. p. From this superficial discussion. On the other hand. and has been studied at various points in time by using the languages of different historical moments.” One Soviet expert described the progress of modern science as follows (Hahn. This synthetic development requires the introduction of new concepts and new thoughts in the entire spectrum of science. but based on the known knowledge base.” Based on this description. we witnessed the appearance of the key concept “systems” in scientific research. In (von Bertalanffy. where all problems and phenomena could be decomposed into causal chains. That has been the foundation for all basic theoretical research and modern laboratory activities. research methods and results in various disciplines have been intertwined to influence overall research progress. where not every problem or phenomenon can be simply described by a single . all of this can be considered as the center of the concept of “systems . they are progressing in all directions with more depth. systems methodology was introduced in order to deal with problems of order or organization. modern science and technology are not Utopian projects as described by Popper (1945).4 Chapter 1 Galileo’s method implies simplifying a complicated process as basic portions and processes [for details see (Kuhn. 185): refining specific methods of systems research is a widespread tendency in the exploration of modern scientific knowledge. reknitting every corner for a new world. mathematical laws. However. more applicability. That is. they won victories for physics and led to several technological revolutions. 1972). Ackoff (1959) commented that during the past two decades. It is a reappearance of some ancient thought and a modern form of an ancient quest. just as science in the nineteenth century with forming natural theoretical systems and processes of science as its characteristics. with the appearance of the concept. it can be seen that the concept of systems we are studying today is not simply a product of yesterday. many branches of modern science have shown the trend of synthetic development. In addition. the vision of the world as a teleological cosmos was replaced by the description of events in causal. Second.

societal changes have brought more pressing demands for new construction materials. and the information highway. In the history of scientific development. and automatic controllers. Here the relation between people and machines becomes more obvious.. or new construction materials. The design of steam engines. and many other components. increased speed of communication furthers scientific development to a different level. cordless equipment. but when dealing with the designs of missiles. consisting of men. economic. that hinted that history has reached a point where all aspects of science and technology have been maturely developed so that each rational combination of information or knowledge could result in unexpected consumable products. . and political problems are intertwined to form a giant. the research progress of the three-body problem in mechanics is an adequate example. and uncountable financial. machines. the need which arose in the development of technology and requirement for higher level production. On the other hand. chemical reactions. agriculture. At this junction. transportation blockages. the exploration of nature has always moved back and forth between specific matters or phenomena and generalities. which has resulted in increased communication of new scientific results. has to be in place. It was the great political. people. etc. landing humans on the moon. etc. self-controlled equipment from domestic temperature controllers to self-directed missiles. for example. There have had been many advances in technology: energies produced by various devices. which includes the combination of various techniques. In the following. and crimes committed by teenage gangs. A great many problems in production require locating the optimal point of the maximum economic effect and minimum cost in an extremely complicated network. computers. Scientific theories need foundations rooted deep inside practice while theories are used to explain natural phenomena so that our understanding can be greatly enhanced. but politicians are also using similar (systems) methods to seek answers to problems like air and water pollution. studying problems with multicausality or multirelation will become more and more significant. aircraft. motors. technical and personnel arrangement success of the American Apollo project. and business. From these examples. as human race advances. machines. a collective effort. social. automobiles. can all be handled by specially trained engineers. military affairs. we discuss the technological background for systems theory — that is. such as steam engines. decline of inner cities. complicated system. it can be seen that the development of technology forces mankind to consider not only a single machine or matter or phenomenon. combining many different aspects of knowledge. This kind problem not only appears in industry. Also. but also “systems” of machines and/or “systems” of matter and phenomena. So. electronic technology.Introduction 5 causal relation. The study of either an isolated part or a single causality of problems can hardly explain completely or relatively globally our surrounding world. The fundamental characteristics of this world are its organizational structure and connections of interior and exterior relations of different matters.

Thus. 1976. not all scientific workers have this kind of optimistic opinion. as described previously. To this end. 1982). Some scholars believe that it is an omen that systems theory itself is facing a crisis [see (Wood-Harper and Fitzgerald. 1. appearing in the world of learning and production. another new scientific and technological revolution will soon arrive. systems thinking and methodology have seeped into all corners of science and technology. General Systems Methodology General systems methodology is a theoretical methodology whose purpose is to realize the ideal model of a systems theory described by von Bertalanffy (1968). 1982) for details]. since the concept of systems was first introduced by von Bertalanffy. Quastler (1965) and Zadeh (1962) — that is. Here we look at the taxonomy of different approaches. Based on the idea and logic of establishing such a general systems theory. made available by Wood-Harper and Fitzgerald based on the field of application of each approach. One group follows the direction pointed out by Quastler and Zedah.1. there exists a tendency that bigger or more accurate products with more profits are designed and produced.6 Chapter 1 In production. 1.2. Lilienfeld. all areas of learning have been faced with complexity. scholars in systems science have been divided into two big groups. and “systemality. establish a general theory applicable to explain phenomena in various specific systems.” This tendency betokens a sharp change in scientific thinking. for example. it is not hard to see that because of systems concepts. Berlinski. totality. However. each application of systems theory points out the fact that the relevant classical theory needs to be modified somehow [see (Klir. 1970. Because of such a wide range of applications. 1970). In fact. it would require modification of the relevant .2. since if a conclusion were obtained by applying results from general systems theory. However. Several Aspects of Systems Theory In the past half century or so. Only time and history will have the right to tell who is right and who is wrong. under the name “systems methodology.” many problem-oriented systems approaches have appeared. many scholars have tried to employ known results in general systems theory to solve practical problems (Klir. By comparing Descartes’ second principle and Galileo’s method with systems methodology and considering the development tendency. and the other contains scholars with the desire to discover general principles while solving practical and specific problems. under different interpretations. 1978)]. see (Wood-Harper and Fitzgerald. Berlinski (1976) and Lilienfeld (1978) thought that all these attempts were extremely unsuccessful. For the original work.

and Zadeh dreamed about. Before solving a problem. The Traditional Approach The traditional approach was introduced by the American National Computer Center. it can be seen that it is natural for the application of results from general systems theory to bring about such a phenomenon.2.2. Human Activity Systems Approach Checkland (1975) tried to establish a systems theory in order to solve problems with unclear meanings and incomplete definition of structures. and business.. On the other hand. if an application of results of general systems theory requires a complete rebuilding of the structure and rules of the relevant theory. Mumford believes that men must be considered as a part of the system under consideration. because of this phenomenon. administration of organizations. 1. with its emphasis on problems without definite structures. 1. It can be classified into general systems methodology. because we are using a different methodology than our forefathers used.2. This methodology has been applied to many different problems and obtained some obvious consequences. from the previous discussion. and much easier to use to solve practical problems. . To a certain degree. this method requires effort to explore and test the problem structure. we surely need to study the correctness of the application of the results from general systems theory. However. 1. this method is accepted by many systems analysts. He believes that this kind of system appears in environmental protection. He also attempted to look for a specific solution among many candidates of solutions so his approach would be more meaningful in the sense of application. Checkland established his methodology based on general systems theory and modified some aspects of systems theory so that the results would be more applicable in the real world. and complicated environments. The Participant Approach The participant approach emphasizes the importance of the researcher(s) in systems analysis (Mumford et al. because when practical problems are being solved often the effort is difficult to define and causes controversy. At the same time.3.4. many scholars hope to locate the reason why this phenomenon ever appeared and to fix the problem so that the resulting general systems theory will be more like what von Bertalanffy.Introduction 7 theory. some systems analysts believe that considering applications of general systems theory is not practical. Also. 1978). either as participants of the activities of the system or as designers of the system.2. Quastler.

all these models must possess some common properties. However. From the aforementioned classification of systems approaches due to WoodHarper and Fitzgerald. This implies that the given set of data can be the initial point for studying properties of the underlying structure. if the relations between the data have been established. consequently discoveries of new structures of the systems can be expected. then we will be able to comprehend the essential properties of the system under consideration.8 Chapter 1 Its basis is that certain problems are solvable with computers. This method was developed to solve problems contained in classical research methodologies. Also.2.2. then the essential structure of the system. and there is no need to define various models for different applications. Structured Systems Analysis Approach The structured systems analysis approach was described individually by de Marco (1980) and Gane and Sarsons (1979). how can the concept of subsystems be used to coordinate a large group of analysts and to solve complicated problems concerning big machines and large systems? This method offers new tools for analysis and literature editing. For instance. all the systems approaches listed are applications of general systems theory. data analysis can be considered as the “median” steps leading to the ultimate understanding of the structure of organization. Each application is considered individually. These concepts have allowed the literature of known systems to be edited more clearly. Therefore. represented by the set of data. The hypothesis is that if we can. For the same set of data. This process is employed to complete the systems design. classify the data. It has been attracting the attention of more and more businesses and organizations. Based on research on the properties of systems of interest. it can be seen that the basic assumptions of various systems methodologies can be considered as axioms or as results of general systems theory.5. At the same time. data dictionary. 1. Data Analysis Approach Data analysis is established on the principle that the fundamental structure of systems is data [see (1978) for details]. the input is designed. and problems are usually solved by finding and designing the optimal subsystems. under certain conditions. concepts such as data flow diagrams. According to the desired output.6. will appear naturally. the condition of optimality is determined. while the other three approaches concentrate on problem solving with the hope that the results will . 1. and constructed English have been introduced. See (1978–1979) for details. For example. In a certain sense. the first three approaches emphasize understanding of systems structures and theoretical analysis of systems. different mathematical models can be established.

They believed that a system consisted of a set of objects. and a finite sequence of relations defined on the domain. . 1983). the existence of subsystems is relative and dependent on each other and conditioning each other. see (Klir. such as control systems theory. while the others can be called applied systems techniques.2. especially. von Bertalarnffy asked how we can define a system. and information systems theory. That is. A system possesses three important features: (i) order or levels. That was why von Bertalanffy believed that general systems theory deals with formal characteristics of systems structures so that its results can be applied to research topics in various scientific fields. the goal of general systems theory is to establish a higher-level abstract theory without touching on any specific physical properties. automata. as a theory. In (Systems. Even though it follows from the previous discussion that systems have a long history and that many problems discussed in modern systems theory have had been considered by many great thinkers from different angles in the past several centuries with different languages. For detailed discussion on various systems approaches. (ii) structure of a set. (4) study special systems models logically and methodologically. each system consists of at least two subsystems. (2) classify systems and find their properties in the most general sense. In (von Bertalanffy. we still have not obtained an ideal definition for the concept of systems! As a matter of fact. and that it can be used not only in physical systems but in any “whole” consisting of interacting “parts” (von Bertalanffy. 1969–1976). Tarski (1954–1955) defined the concept of a system with relations as a nonempty set. 1967. the first three approaches can be termed theoretical systems methods. that is. (3) model systems behaviors. it is still relative new. as a methodology of general systems. 1972). The first three tasks constitute the theoretical basics for the concepts of special systems. (iii) relativity and wholeness. and some relations between the attributes of the objects. “system” is a generalization of the concept of “structure” in physics but differs from a structure.Introduction 9 allow us to understand these systems on a theoretical level. Uyomov proved that the attributes connecting the objects of a system can be considered as new objects of the system. 1. some relations between the objects. even though the idea of systems has permeated all of modern science. as stated in (Department of Business Administration. pp.7. that is. Therefore. 1970). we discuss the foundations of various systems approaches and some research directions in general systems theory. called the domain of the system with relations. That is. 125–126). To conclude this chapter. Hall and Fagen (1956) discussed the definition of systems. Some Research Directions in General Systems Theory Among the main tasks of general systems theory are (1) define the meaning of “systems” and related concepts.

1989d). does not contain any results. If that is a fact. contains something inappropriate? The authors also pointed out that there is no need to be concerned because the entire research literature of systems theory contains no results. However. In this book. produced with this definition of systems. Mesarovic and Takahara (Mesarovic and Takahara. given by Mesarovic. S is an ordered pair (M. to study the general principles of structures and structures with specific properties. Lin and Ma. for details [see (Lin.2. ten years later. does it imply that it is still too early to establish such a theory? Yi Lin and Yonghao Ma (Lin. and only if. 1987. general systems theory becomes a theory of relations. which constitute the object. initiated by Newton and Leibniz three centuries ago. even with special features. Hall and Fagen’s definition of systems can be simplified as follows: A system is a set of objects and some relations between the objects. 1979). generalized principles of dynamics can be applied to the totality of molecules or biological objects. For instance.2. developed on the definition of systems by Lin and Ma. parts. based on calculus. and has been used for research and description. Classical systems theory is a mathematical theory. is a different way of thinking and reasoning about . 1. Catastrophe theory. For example. 1975) pointed out that general systems theory. and relations between the parts of a structure. uncertainty. Therefore.R ) of sets. Ma and Lin. Lin and Ma.1. 1987) studied the attempts at defining systems. The referee(s) of (Ma and Lin.10 Chapter 1 In this way. 1987) claimed that the definition introduced by Lin and Ma will produce a good quantity of results. Mesarovic (1964) used the language of set theory to define a system as a relation. 1.” makes application of systems theory. these principles can find applications in chemistry and biology. Catastrophe theory. Does this imply that the definition of systems. 1987. That is. 1987. which has been evidenced by results contained in (Lin. Here a relation r belongs to R means that there exists an ordinal number n (nonzero) such that r is a subset of the Cartesian product M n.2. the reader will be led to some of these results. is the totality of molecules constituting the material a set? Is the collection of all relations between molecules in the material a set? To deal with these uncertainties. and introduced a definition of (general) systems as follows: S is a (general) system if. synthesized all opinions into one with a consideration of future development. contain many uncertain factors. Yi Lin introduced the following epistemological axiom: The existence of an object is determined by the existence of some particles of a certain level and of some relations between the particles. Classical systems theory. 1987. Its results have also been used to solve concrete problems. Because of the generality of the problems studied. There are many references here. one can still obtain some formal properties of the structure under concern. contained in the concept of “sets.7. We now briefly introduce several research directions in general systems theory. Lin. one of which is (Wonham. Here. 1989d)]. for a given experimental material. where M is the set of all objects of S and R is a set of relations defined on M.7.

and the system is an input—output “black” box. Consisting of a body of new concepts and techniques. abstraction and generalization.2. This theory has found many applications. In (Rescigno and Segre. 1. The mathematical difficulty in dealing with structures with three or more parts becomes obvious. For more information. Fuzzy mathematics is a theory dealing with the rapprochement between the precision of classical mathematics and the pervasive imprecision of the real world.g. see (Poston and Steward. many scholars do not see it as a part of systems theory. The fundamental concept here is fuzzy sets. in an object’s shape. including hydraulics. Therefore. 1. this theory accepts fuzziness as an all-pervasive reality of human existence. Fuzzy mathematics. this study has been applied to research of literal catastrophes (e.4.5. translation of languages. 1966). internal information transportation of systems. which has opened many new frontiers in psychology.7. Cybernetics. etc. Rescigno and Segre first studied compartment theory. philosophy. above all.2. between which there appear processes of transportation. and change in ideas themselves. The basic idea is that the problem or structure under consideration can be described as a whole consisting of parts satisfying certain boundary conditions. the parts can be chained together by transportation. linguistics. downfall of an empire) and to quiet changes. As suggested by its name. Cybernetics has often been used to describe the formal structure of some “action processes. The literature in this research area is plentiful. dancing of sunlight on the bottom of a pool. (e. which are classes with boundaries that are not sharp in which the transition from membership to nonmembership is gradual rather than abrupt. cybernetics can still offer some understanding about that structure.7. decision making with uncertainty. economics. such as “chains” or “nipples. Milsum. and other fields. Laplace transformation and networks and graph theory have been successfully used.. At present. collapse of a bridge. we are unable to design machines that can compete with humans in the performance of tasks such as recognition of speech. ecology.Introduction 11 “discrete. sociology. 1978). physiology. 1. and I will not list any here. operations research. containing fuzziness. 1966). This stems from a fundamental difference between human intelligence.” Even though the structure of the system under concern is not clearly given.g. Because of its usefulness.7. electricity.2. For more details. Compartment theory. and impacts on the environment of controlled systems. management science. cybernetics has been applied in many areas.” “jump” or “discontinuous” changes in a course of events. see (Bayliss. based on precision. comprehension of meanings.” That is. summarization of information. 1966.3.. or transportation appears between a central part and its surrounding parts. and. transition from wakefulness to sleep). Cybernetics is a theory on systems and their environments. . change in a system’s behavior. political science. In this area. markets. and machine intelligence.

which is defined by an expression similar to one with negative entropy in thermodynamics. 1983). One goal of this research is how to apply optimal strategies to confront other players (or the environment). Besides the quantitative study of structures.7. the characteristics of the models. since it deals with the behaviors of (n) ideal players with the ability to reason and make decisions. Quastler (1955) believes that information can be considered as a measure of the structure of organization. political science.2. game theory is a systems theory. Based on the opinion of Shannon and Weaver (1949). (Rashevsky. The goals of this research are twofold: (1) to abstract and rigorously explain the adaptive processes of natural systems. 1. for example. 1956)]. Information theory.7. 1989). especially directed graph theory. whose goal is to win more than to lose. computer science. biology. 1. Although information theory is very important in the engineering of communication. psychology. game theory is about the study of systems with special reaction forces.6. 1. including biology [see. That is. and the classic is (Holland. According to Quastler. the concept of chaos. 1.10.12 Chapter 1 1.2. which has led to the new science called Chaos. Graph analysis. which has several different meanings.2. developed in graph theory are similar to those of compartment theory.2. Game theory. Graph theory. so it establishes relations with the theory of open systems. some scholars have .7. This method has been successfully applied in many different areas. Navier–Stokes equation and chaos.7. the balance between efficiency and efficacy necessary for survival in many different environments. including economics. The central theme of research here has been robustness. information theory is established on the concept of information. 1975).7.9. its study has recently given some new understanding. On one hand. A good reference is (Shubik. GAS are search procedures based on the mechanics of natural selection and natural genetics. linguistics. Even though the Navier–Stokes equation has been successfully applied in weather forecasting (short term only). describes in detail the construction of relations between systems. Combined with matrix theory in mathematics.8. People are trying to find and understand chaos. On the other hand. has created a burst of research in almost all areas of science.7. A good introductory book on this subject matter is (Goldberg. This approach has led to important discoveries in both natural and artificial science. and (2) to design artificial systems software that retains the important mechanisms of natural systems. etc.2. There exist many methods useful to this research. They combine survival of the fittest among string structures with a structured yet randomized information exchange to form a search algorithm with some of the innovative flair of human search. Genetic algorithms. many problems deal with the organization and topological structures of the systems of interest. research results in other areas of application have not yielded any convincing consequences. Genetic algorithms (GAs) have been applied in many areas.

for details.7.7. but also opens up research areas in which. Here I list only two references (Cohen and Steward. see (Rapoport. .2. computeraided design uses computer simulation as a substitute for laboratory experiments. sense perception only gives information of this external world or of “physical reality” indirectly. 1975). For example. Lin and Qiu. or compartment theory. contain nonlinear equations.2. 1987. described by differential equations. As a matter of fact. we do not know if the solution exists. The main structure dealt with is net structures of systems. it is either an illusion of a concept of a higher dimensional space or a misunderstanding in computer-aided calculation. however. p.11. or graph theory. For example. 1994. Statistics. have been introduced. This method not only saves time and human labor. algebraic methods can be introduced so that concepts.2. not to solving practical problems. For example.7.7. 1987). For related references. Ma and Lin. At the same time. specific experiments in laboratories can be replaced by computer simulations. Network theory can be considered as part of set theory. On the other hand. at least until recently [see (Lin and Ma. which was described by a system of more than 100 nonlinear differential equations. is related to the concept of systems and related topics. this method has often been used in the study of markets and populations. Most of this research. 1989d. we can only grasp the latter by speculative means. All research along this line was originated by Mesarovic (1964. In fact. introduced by Lorenz in the study of simplifying Navier–Stokes equation. there is no way to solve the systems. see (Mesarovic and Takahara. 1949). carried out since 1964. Lin. has nothing to do with modem chaos. Networks. All the general definitions and properties of structures have been given in terms of set theory.2. 1–24). the language and methodology of category theory can be employed to classify structures.14.Introduction 13 shown that the original chaos. If the structures. On the other hand. 1. Einstein said (1922. Because of the use of the language of set theory. computer simulation becomes an important method. 1987. Hess (1969) employed this method to compute a 14-th step chemical reaction in human cells. in general. Set theory. mathematically.13. It follows from this that our notion of physical reality can never be final. 60): The belief in an external world independent of the perceiving subject is the basis of all natural science. the mathematical structure or model obtained by computer simulation can be evaluated by experimental data. pp. 1987)]. Since. 1. such as linear and nonlinear structures. open and closed systems can be described by axioms using the language of set theory. It has been applied to neural networks. 1. Gleick. Simulation. Because of this reason. 1.12. This kind of system cannot be solved by known mathematics.

Successful applications of statistics can be found in almost all applied areas of human endeavors.. if it can be expressed by finite steps of logic rules and calculations. 1. One such example is the Turing machine (Turing. 1936).1.7. then this process can be simulated by a machine. One of the best references here is Box. and consequent discussions have been intensive.14 Chapter 1 Based on this understanding. the human exploration of nature might be able to go to a higher level (Yablonsky. . It can be foreseen that the development of modern science and technology will be the effective motivation and method to test the duration and significance of ideas and thoughts in general systems theory. Hunter. This theory is not of what event causes another event (i. 1967).. and its ability to be corrected and learn (Minsky. causality) but about what event will imply that another will occur. That is why new ideas have been appearing quickly. Researchers are trying to find more practical examples and general properties. collected with reference to some prior mathematical analysis. 1. the Turing machine is an ideal machine which can print “1” and “0” on an infinitely long tape. The theory of automata concerns an ideal automaton with input and output. some scientific workers have simplified the question in which administrators of scientific research are interested to the following: Can science be studied as a system? This is an important and practical problem since if this question can be studied deeply with reasonable outcome. and Hunter’s classic (Box et al. Theory of automata. In ordinary language. no matter how complicated a process can be. 1978). It can be shown that. statistics is a mathematical theory of how to predict and comprehend “reality” according to a small sample.5.e. the future of this theory can be seen. A Few Thoughts It can be seen that systems theory is very young. 1984).3. and are testing and rethinking the theory. Based on the thoughts of general systems theory. From this fact.2.

M ⊆ N denotes inclusion. the vertices of a die. as a connecting link between mathematics and philosophy. and topology.3} and N = {2. Two sets M and N are equal. in symbols M = N . as well as through the generality of its concepts. m ∈ M means that m is an element of the set M.” For example. of finite. is one of the greatest creations of the human mind. or all elements of M also belong to N. Everyone who studies it is fascinated by it. if they contain the same elements. are designated by enclosing their elements in braces or by a symbol like {x : Φ (x)}. or at least first rendered the possibility for their further development. the points on a circle. M is a subset of N . that is. More important. We will present the basic features of Cantor’s naive set theory. For example. a set of infinitely many elements. however. theory of real functions. 2.1. Finally. The basic concepts and methods of proofs will be used in the development of general systems theory. is that the theory has become of the greatest importance for the entire spectrum of mathematics. It has given rise to new branches of mathematics.1. which means that the set consists of all the elements satisfying the proposition Φ .) The notation M ≠ N means that the sets M and N are different. welldistinguished objects (called the ‘elements’ of M) of our perception or of our thoughts. a set of infinitely many elements. if M = {1. a set of 8 elements.2. then M = N . the 38 students in a classroom constitute a set of 38 elements. in general.CHAPTER 2 Naive Set Theory Set theory. the totality of even numbers. Further. 15 . such as the theory of point sets. the theory of sets has had influence on the investigation of the foundation of mathematics.3}. whereas m ∉ M denotes that m is not an element in M. created by Georg Cantor (1845–1918). A Set and Its Classification According to Cantor a set M is “a collection into a whole. (Sets.

1. Sets M and N are equipollent if there exists a bijection of M onto N . or do not satisfy the previous two conditions. Then the set of all integers can be written as the sequence {0. –2. ..2. Another coarse classification of sets distinguishes the collection of all sets into finite. .16 Chapter 2 Proposition 2. precisely one natural number corresponds to it and vice versa. The set of all rational numbers is denumerable..1. ƒ(x) ≠ ƒ (y) (2. ƒ(m ) indicates the element in N which is assigned to the element m. .1) The mapping ƒ : M → N is a bijection (or a bijective mapping) if ƒ is both surjective and injective. Then for each element m ∈ M. A set X is denumerable iff it can be written as a sequence {x0 . The set of all integers is denumerable. Theorem 2. i.1.. say. x2 . This fact implies that an infinite set X is denumerable iff to every element x of the set. and h. 2. –n.. or injective) if for arbitrary distinct elements x and y ∈ M.2) (2. –3. denoted by ƒ : M → N. . The equipollence relation between sets classifies the collection of all sets into classes such that each class contains all the equipollent sets.1. –1. 1 –1. The mapping ƒ : M → N is 1–1 (or one-to-one.. ƒ. x1 ..3) Theorem 2. g . . x3 . 1. A mapping (or a function) from a set M into a set N is a rule under which each element in M is assigned to an element in N .2. they contain infinitely many elements and are not equipollent to N. 3.} (2. . or are equipollent to the set of all natural numbers N.1. n.. . lower case letters. 3. –2. Generally.1..e. denumerable and uncountable sets according to whether the sets contain a finite number of elements. . Let ƒ be a mapping from the set M into the set N. M = N iff M ⊆ N and N ⊆ M. The mapping ƒ : M → N is onto (or surjective). . .}. Proposition 2. there exists an element m ∈ M such that ƒ(m) = n. if for each element n ∈ N . 2. Proof: Let us write the integers as follows: 0. –3. will be used to indicate mappings.

then the open internal (0. From Proposition 2. and only once. .b 1 . Proof: We prove the theorem by a diagonal method. r1 . called the second. The set of all points in the closed interval [0. then all fractions with denominator 3. Thus.b 2 . The following theorem shows that not every set of real numbers is denumerable. then all fractions with denominator 2. The collection of these rational numbers is thus written as a sequence (2.. (2.5) If we denote this sequence by {r 1 .b 3 . }.6) We now write each of them as infinite decimal numbers by putting in enough zeros . We can write all whole numbers in order of magnitude (i. .4) If we write the numbers in the order indicated by the arrows (leaving out numbers which have already appeared). all numbers with denominator 1). . etc. . – r3 .e. } is the set of all rational numbers.–r 2 .1.2 it follows that the points in (0.3. and the denumerability of this set is established.r3 . 1) can be written as the sequence b 0 . 1] is uncountable. 1] is denumerable. – r 1 . r 2 . . we have the rows of numbers (2. or Cantor.r 2 . diagonal method. r3 .Naive Set Theory 17 Proof: Let us first deal with the set of all positive rational numbers. . 1) is also denumerable. then obviously {0. Theorem 2.. Suppose that the closed interval [0.1. .. then every positive rational number certainly appears.

defined by b = 0. We say that m is not larger than n. . Let m and n be two cardinalities. . The cardinality assigned to each denumerable set is denoted by ℵ0 (read aleph zero). .1. The equality X = Y holds iff X and Y are equipollent. Suppose that X and 2 are . to every set X the cardinality of X is assigned and is denoted by X. and let  X  = m and  Y  = n. We must now show that X and 2 are X not equipollent.1. 1). We prove this by contradiction. which is called the power set of X.4. . Then X < 2 X (2. if there exists a one-to-one mapping of X into Y. } . .18 Chapter 2 whenever necessary. or that n is larger than m. Let X be a set and 2X the set of all subsets of X. Generally.9) Proof: Define a mapping ƒ: X → 2 X by letting ƒ(x) = {x} for each x ∈ X.8) such that Then b is not contained in the sequence {b 0 . We say that m is smaller than n. and we write m < n or n > m.b 2 . The following fact says that for any fixed cardinality m there exists a set whose cardinality is larger than m. This contradiction implies that the set of all points in the interval (0. contradiction.b 1 . 1) is uncountable. and the cardinality assigned to the set of all real numbers is denoted by c (called the continuum ). or that n is not smaller than m.b 0 b 1 b2 b3 . (2. so  X ≤ 2 X. if m ≤ n and X and Y are not equipollent. Theorem 2. ℵ 0 < c. For a finite set X the cardinality of X is equal to the number of elements of X. Corollary 2. X Then ƒ is one-to-one. The sets X and Y are of the same cardinality if they are equipollent.1.7) Now we construct a number b ∈ (0. Then (2. and we write m ≤ n or n ≥ m.

and  X < 2 X. k 3 . } be a denumerable set and Y ⊂ X. Z may not contain any element.x 1 . then y ∈ A = h ( y ).. Operations of Sets Let X be a set. in symbols. then y ∉ h ( y ) = A. in particular. another contradiction.11) There now exist two possibilities: (1) y ∈ A or (2) y ∉ A. So y ∈ A. This is followed again by a first element. called the empty set. etc. let y 2 = x k2 . Z = X – Y.. Then in X there exists a first element x k1 which belongs to Y . If case (1) holds. This implies that y ∉ A.. of itself. When Y is not a proper subset of X. The union of finitely or infinitely many sets is defined to be the set of those elements which belong to at least one of the sets. Therefore. the bijection h cannot exist. in X. according to whether Y is finite. } comprises precisely the elements of Y. denoted by x k 2 . which belongs to Y . Proof: Let X = {x 0 . In this case. If case (2) is true. X 2 . k 2 .10) Since h is onto. A set Z of those elements of X which do not belong at the same time to the subset Y is called the complement of Y with respect to X. There then exists a bijection h from X onto 2X .1.. X1 . if Y is a subset of X and X ≠ Y. Define a subset of X as follows: (2.12) . Every subset of a denumerable set is at most denumerable. .. A set Y is called a proper subset of X. The process does or does not terminate. contradiction. Theorem 2. Ø.x 2 . 2. we introduce an ideal set. the possibly terminating sequence {k1 . The union U of an at most denumerable number of sets X0 . It is a subset of every set and. is written in the form (2. it follows that there exists an element y ∈ X such that h (y) = A (2. Since X contains all the elements of Y. .2.Naive Set Theory 19 equipollent.2. denoted Y ⊂ X.. The empty set is classed with the finite sets. . let y 1 = x k1 . Y is finite or denumerable. Therefore.

and C be sets.22) . B.2.2 [De Morgan’s Laws].2.2 [Associative Laws].17) (2. x ∈ A – (B ∩ C) iff x ∈ A and x ∉ B ∩ C.21) (2. (2. Proposition 2. (2. Suppose that A.14) Let I be a collection of sets. Then (2. Let A.1 [Commutative Laws]. Let A and B be two sets.2.18) Proof: We prove only the first equality.16) Theorem 2.3 [Distributive Laws]. Proposition 2. Let A. For the intersection D of two sets X 1 and X 2 . iff x ∈ ( A – B) ∪ (A – C). iff either x ∈ A and x ∉ B or x ∈ A and x ∉ C.20) The proof follows immediately from the definitions of union and intersection. Then. iff x ∈ A and either x ∉ B or x ∉ C.19) The proof follows immediately from the definitions of union and intersection.20 Chapter 2 The intersection of arbitrarily many sets is defined to be the set of those elements which belong to each of the aforesaid sets. For any element x. B.13) and for the intersection of at most denumerably many sets. Then (2. The union U and the intersection D of the sets in I is written (2. we write (2. Then (2. and C are sets.2. Proposition 2.15) and (2. B. and C be sets.

Proposition 2. Proposition 2. and D. we have the following results. (2. Let A be an arbitrary set.” For the union and intersection of a collection I of sets. C.2. iff x ∈ A and iff x ∈ A and there exists an X ∈ I such that x ∉ X. For arbitrary sets A. Then The following formulas illustrate the analogy between the inclusion relation and the “less than or equal to” relation in arithmetic: Proposition 2.23) The proof is straightforward and is omitted.6. B. Let A.7 [De Morgan’s Laws].Naive Set Theory 21 The proof is left to the reader. B and C be sets. The following identities are given without proof.2.5. the following are true: where → means “imply.2. Proposition 2.4 [Laws of Tautology]. iff there exists X ∈ I such that iff . Then Proof: (1) For any element x.2. For any set A.

Then (1) (X 1 ∪ X2 ) × Y = (X 1 × Y) ∪ ( X2 × Y).d ). it follows that { c} ∈ ( a. (2) Y × (X 1 ∪ X2 ) = (Y × X1 ) ∪ (Y × X2 ). one can check that the theorem holds for case (iii). For an ordered pair ( x.y). it is necessary and sufficient that a = c and b = d. For example. y) and y the second term (or coordinate) of (x. x ∈ A – ∪ X ∈ I X iff x ∈ A and for each set X ∈ I . we obtain a = c = b = d .y}}. y) is the set {{x}. then obviously X × Y = . Similarly.b }. and Y be sets.22 Chapter 2 (2) For any element x. Let X1 .b). (i) {c} = {a} or (ii) {c} = {a. Equality (ii) holds if a = b = c. In order that ( a.b ) and {c.2.d } = {a. iff either x ∈ X 1 or x ∈ X 2 and y ∈ Y. Hence. {x. Theorem 2.y) such that x ∈ X and y ∈ Y. ( A – X). we have the following results: Proposition 2. X2 . x ∈ A–X . iff for each set X ∈ I. This product is denoted by X × Y . Certain properties of Cartesian products are similar to the properties of multiplication of numbers.b) = (c. (6) Y × (X 1 – X2 ) = ( Y × X1 ) – ( Y × X2 ). If c = b . then (ii) holds and this case has already been considered. The Cartesian product of two sets X and Y is defined to be the set of all ordered pairs (x. then a = c and b = d. iff x ∈ Let x and y be two elements. We then have c = a and either c = b or d = b . We call x the first term (or coordinate) of ( x .3. It remains to show that the theorem holds for cases (i) and (iv).y) ∈ (X1 ∪ X2 ) × Y iff x ∈ X 1 ∪ X 2 and y ∈ Y. Proof: The sufficiency part is clear.8 [Distributive Laws]. Proof: We prove only the first equation. (x.d ).b}. Equalities (iii) and (iv) are then equivalent and it follows that c = d = a. The ordered pair (x . .2. From the definition of ordered pairs.y). If X = or Y = . x ∉ X . and (iii) {c. suppose that (a. To prove necessity. that is. in which case the theorem holds. (4) Y × ( X 1 ∩ X2 ) = (Y × X 1 ) ∩ (Y × X 2 ).d} = {a} or (iv) { c. If d = b.d} ∈ ( a. which proves the theorem.b ) = (c. (5) (X 1 – X2 ) × Y = ( X1 × Y ) – ( X2 × Y). (3) (X 1 ∩ X 2 ) × Y = (X 1 × Y) ∩ (X2 × Y ).

then f – 1 = {( x. that is. Example 2. iff either ( x . y 1 ). A subset f ⊂ X × Y is called a function or mapping from X into Y if for any x ∈ X and any y 1 and y 2 ∈ Y. That is. Proof: By definition.2. y ) ∈ X 2 × Y.x 1 ) and ( y.y ) : ( y.y ) ∈ f then x 1 = x 2.Naive Set Theory 23 iff either x ∈ X 1 and y ∈ Y or x ∈ X 2 and y ∈ Y. In this case. the subset ƒ –1 is a partial function from Y into X iff for any y. [0. f is called a partial function from X to Y. The notation X ~ Y indicates that the sets are equipollent. (a) The sets of the points of the intervals [0. . The proof is straightforward and is omitted. f is one-to-one.y ) and ( x 2 . The following result says that the equipollence relation is reflexive. ( x . y) ∈ X 1 × Y or (x . and (0. When each x ∈ X belongs to a pair in f . the element y is denoted f(x ). we illustrate it by some examples. If f : X → Y is a function.4. (y. if f (x1 ) = y = f ( x2 ).y 2 ) ∈ f implies that y 1 = y 2 = y. the function f is defined on X. (2) Symmetric property: A ~ B implies B ~ A. (x.y ) ∈ ( X 1 × Y )∪ ( X2 × Y). 1).1.3. 2. x1 and x 2 . if (x 1 . The function f is written f : X → Y. Since the concept of equipollence is of fundamental importance.3. (0. Theorem 2. and C the following holds: (1) Reflexive property: A ~ A. B.x 2 ) ∈ f –1 implies that x 1 = x 2 . and transitive. 1]. 1]. then x1 = x 2 .x ) ∈ f } is a partial function from Y into X iff f is one-to-one. (3) Transitive property: (A ~ B) and (B ~ C) imply A ~ C. symmetric. For arbitrary sets A. Arithmetic of Cardinal Numbers Let X and Y be two sets.1. Theorem 2. iff ( x. 1) are equipollent to each other.3. Thus. Otherwise.

they contain equally many elements. . . 1) ~ [0. }. A half-line can be obtained by central projection from an interval erected at right angles to it and open above (Fig. and only if. 1). . . let us prove that (0. This proves our assertion. 6.24 Chapter 2 Figure 2.1). A full line obtained from a bent open interval. First. Infinite sets Figure 2. (c) Two finite sets are equipollent if. We denote the points of the first interval by x and those of the second by y.2). 1].2. 4. A half-line obtained by central projection. . and set up the following correspondence: It is evident that we have already set up a 1–1 correspondence between the two intervals. 2. (d) An infinite set can be equipollent to one of its proper subsets. open interval (Fig. A full line can be obtained similarly from a bent. 2.2. We can show analogously that [0.1. This is shown by the equipollence of the sets {1. 1] ~ (0. 1).3. From this it follows finally that [0. . } and {2. 1) ~ (0. . (b) A half-line and an entire line are equipollent to an interval.

Likewise. construct a function g(z) defined on the interval 0 ≤ z ≤ 1 with the property that. precisely one of the following cases can occur: (a) X is equipollent to a subset of Y. The proof of this fact is obtained by a procedure which is again essentially the second diagonal method.3. Theorem 2. Proof: Assume that f : X → Y and g : Y → X are both 1–1 functions. Denote by fz (x) the function thus assigned to the point z. however. This property can therefore be used to distinguish finite and infinite sets independently of any enumeration. Now. in particular. it must coincide with an fu (x ). This. (a)(1). 1] is neither denumerable nor equipollent to the continuum. The assumption that the set of our functions is equipollent to the continuum has thus led to a contradiction. at every point z. vis.. For. There are four combinations of these two pairs of cases. Case (a)(1) is discussed next. Suppose that the set of functions were equipollent to the continuum. (b)(2). When two sets X and Y are given. Since g ( z) is also a function defined on [0. then X ~ Y.2 [Bernstein’s Equipollence Theorem]. 1]. It is clear not only that there are infinitely many distinct functions on [0. (2) Y is equipollent to no subsets of X. Thus.Naive Set Theory 25 thus exhibit in this respect an entirely different behavior from that of finite sets [see example (c)]. 1] correspond in a 1–1 manner. g( u) = f u (u ). g(z ) ≠ f z (z). Then it would be possible to make the function f (x) and the points z of [0. exactly one of the following possibilities can take place: (1) Y is equipollent to a subset of X. Case (b)(2) means that X and Y are not comparable. (b) X is equipollent to no subsets of Y. (e) The set of all real functions defined on the interval [0. 1] but also that there exist uncountably many. If each of two sets X and Y is equipollent to a subset of the other. at the point x = 0 alone. We will show that this case never occurs. and define inductively . there remains to be proved only that the set of functions is not equipollent to the continuum. hence. is excluded by the definition of g( z). In the two middle cases  X < Y  and  Y < X . the functions can already assume the uncountably many values between 0 and 1. Let A0 = X – g (Y ). (b)(1). B0 = Y – f (X). (a)(2).

Repeating this argument. .2.. there exist infinitely many other cardinal numbers.3. where x ∈ p κ –1 .26) Then h is 1–1 from X onto Y.1. then t = ƒ(z) for some z ∈ X (otherwise t ∈ B 0 ). since A 0 ∩ A n = and B 0 ∩ B n = . and z ∈ A implies t ∈ B. . Corollary 2. then B n + 1 ∩ B m +1 = since ƒ is 1–1 and if B n ∩ Bm = . Then no two of the sets (2.24) follows. h –1 (t) = g (t ).2 because pι + 1 = p(p ι ). if t ∈ Y – B. (2. which establishes the equipollence between X and Y.27) are equipollent.p κ –1 . the set p κ –1 is equipollent to a subset of p ι . p ι +1 is equipollent to some subset of pι . In addition.1 says that besides the finite cardinal numbers. We know from Section 2. then ƒ(z) ∉ B (since ƒ (z) ∈ Bm implies that m ≠ 0 and z ∈ A m – 1 ). . namely to the subset of singletons {x}.3. and if m is odd then h – 1 (t ) = ƒ – 1 (t). then A n +1 ∩ A m +1 = since g is 1–1. .1 and Theorem 2. similarly. h is from A onto B. Proof: Let p κ be the κ th set in the sequence (2. }. Eq.27) and suppose that there exist κ and ι such that κ > ι and p κ is equipollent to a subset of p ι . . ƒ is 1–1 onto between X – A and Y – B since if z ∈ X – A. the number ℵ0 . Now. define A = A n and B= Bn and define a 1–1 function h from A onto B by if z ∈ A m and m is even if z ∈ Am and m is odd (2. Theorem 2.3. Let A be an arbitrary set. In the sequel. . The set p κ – 1 is clearly equipollent to a subset of p κ . and the continuum c. So. Thus. Now.1 that to each set X a cardinality (or cardinal number) is assigned. we conclude that each of the sets p κ .24) If A n ∩ A m = . Then by induction on n we have (2. we use p(X) to indicate the power set of the set X. for each n ≠ 0.1. but this contradicts Corollary 2. So.1.25) Then if t ∈ Bm and m is even. we extend h as follows: (2.26 Chapter 2 where = {0. by induction. .

Starting with the set of nonnegative integers which has power ℵ0 .30) This procedure continues infinitely.3. multiplication. . . let X * and Y* be two disjoint sets with X*  =  X  and |Y* | = | Y |. .28) no two of which are equipollent by Theorem 2. ∪ Proof: Assume that there exists an X ∈ A such that It follows that there exists a 1–1 function ƒ such that By assumption of the theorem there is a set Y ∈ A which is not equipollent to any subset of X. .3 and 2. Then the union X ∈ A X is not equipollent to any X ∈ A nor to any subset of X ∈ A.3. Then it is not hard to see that X* ∪ Y * ~ X ∪ Y . p(B). p (p(p(S) ) ) .31) For the definition to have a meaning.3. The contradiction thus shows that Theorems 2. which coincides with We now begin to define the operations of addition. has a cardinal number different from each of the sets in sequence (2.4. . p (p(p(B) ) ) .3 we obtain the sequence of sets S .4 says the union where for each n ∈ . .3. Therefore.3. In this way we obtain infinitely many distinct cardinal numbers. (2. and exponentiation for cardinal numbers.Naive Set Theory 27 Theorem 2. p (p(B)). we can construct the sets (2.29) no two of which are equipollent and none of which is equipollent to any of the sets in sequence (2. ƒ(Y ) ⊂ X 1 and consequently Y ~ ƒ (Y ) ⊂ X. Let m and n be cardinal numbers and X and Y two disjoint sets such that  X  = m and  Y  = n. Then define m + n =X ∪Y  (2. Let the family A have the property that for every X ∈ A there exists a set Y ∈ A which is not equipollent to any subset of X. p (p(S)). it is necessary to show that the sum m + n does not depend on the choice of the sets X and Y. We see that the hierarchy of distinct infinite cardinal numbers is comparably richer than the hierarchy of finite cardinalities.3. (2. We may still obtain other cardinal numbers by constructing the family B = (S) and a new sequence B . Since we have that is.28).4 give us some ideas about how many distinct infinite cardinal numbers exist. . p(S). Theorem 2. Again applying Theorem 2. We obtain in this way a new infinite number of distinct cardinalities.3.28) and from each of their subsets.

however. where i = 1. and n3 . Theorem 2. (2.3. The cardinal number m is the product of n1 and n 2 . Thus. The addition of cardinal numbers is commutative and associative.5. such that m i ≤ n i .3. associative.2.37) (2. Then (2. and distributive over addition: (2. This gives m 1 + m 2 ≤ n1 + n2 . that is.6.32) (2.3. Let mi and ni be cardinal numbers.36) (2. for arbitrary cardinal numbers n1 .36) is an immediate consequence of the relation A×B~B×A Equation (2. Then by hypothesis there exist 1–1 functions by Define a new function g : (2. for every n ∈ ℵ0 = ℵ 0 + ℵ 0 . The multiplication of cardinal numbers is commutative. i = 1. Theorem 2. if every set of power m is equipollent to the Cartesian product A1 × A 2 . 2.38) Proof: Equation (2. always infer from m1 < n 1 and m2 ≤ n 2 that m 1 + m 2 < we have n < ℵ 0 and ℵ 0 ≤ ℵ 0 but n + ℵ 0 = n 1 + n 2 . 2. For example.33) The proof follows from the definition of addition of cardinal numbers. We cannot. n2 . i = 1.34) Proof: Let Xi and Yi . be four pairwise disjoint sets such that and .7.35) We then check that g is a one-to-one function from X1 ∪ X2 into Y 1 ∪ Y2 .37) follows from [A × (B × C)] ~ [(A × B) × C] . denoted m = n1 ⋅ n 2 .28 Chapter 2 Theorem 2.

where  A  = n. where g(z) ∈ X and h (z) ∈ Y. let m. (2.42) (2. and AB = {ƒ :ƒ is a function from B into A} .48) (2. (2. Then. and q. . . Y. .45)–(2. . and  Z = q.47) (2.44) Proof: Suppose that X.45). In other words. . .49) The proofs of Eqs. where a 1 .40) (2.47). n.40)–(2. .  Y  = p.an ).Naive Set Theory 29 Equation (2. a2 . We give only the arguments of the important formulas Eqs.45) (2. ƒZ).38) is shown by the following: Denote the n-fold product m ⋅ m ⋅ ⋅ ⋅ m by m n . For arbitrary cardinal numbers n. Then m n is the power of the set of all sequences of n-elements (a 1 . and Z are pairwise disjoint sets with  X  = n . (2. Then it is not difficult to show that this correspondence is a bijection from the set XY ∪ Z onto the set XY × X Z . and p be cardinal numbers. a n are elements of a n n set A of power m. For the proof of Eq.  B = p.3. we associate with every function ƒ ∈ X Y ∪Z the ordered pair of restricted functions (ƒY . .49) are straightforward. (2.46) (2. Generalizing this idea. (2. then ƒ (z) is.46).39) (read as m equals n raised to the pth power) if every set of power m is equipollent to the set AB .8. Theorem 2. h(z)).41) (2. for every z ∈ Z. notice that if ƒ ∈ (X × Y )Z . (2. Then m = np (2. showing Eq. To prove Eq.44) is equivalent to showing (2.48) and (2.  A  =  A  . p. an ordered pair (g(z). Then g ∈ X Z .43) (2.

50) expresses the transitivity of the relation ≤. Each function F ∈ (X Y ) Z corresponds in a manner described before to some function ƒ ∈ X Y ×Z .30 Chapter 2 and h ∈ Y Z .10. i. 1}A onto the set p(A). z 0 ) ≠ ƒ2 (y 0 . where y ranges through the set Y and z through the set Z and where ƒ takes values in X. (2. 2 m is the power of the set {0. To distinct functions ƒ1 and ƒ2 correspond distinct sets Xƒ1 and X ƒ2 . Finally.z) = g z (y). Theorem 2. g z ∈ X Y .e.e. then the corresponding functions F 1 and F 2 are also distinct. 1}A the set X ƒ ⊂ A. In fact. It follows that the correspondence of the function ƒ ∈ XY × Z t o the function F ∈ (X Y ) Z establishes the equipollence set X Y × Z with (X Y ) Z . h) determines a l–l mapping from the set (X × Y )Z onto the set X Z × Y Z . where gz = F(z). For each fixed z.54) mp ≤ np p implies m ≤ n p m implies p ≤p n Proof: Equation (2. we obtain a bijection from the set {0. if ƒ1 (y 0 .52) (2.54) are called the cancellation laws for the relation ≤ with respect to the operations of . Each such function is uniquely determined by the set X ƒ of those a ∈ A for which ƒ(a) = 1.50) (2. Let ƒ ∈ X Y ∪Z .47).. which proves Eq. i.z) is a function from Y into Y.51)–(2. so F ∈ (X Y)Z. The relation “ ≤ ” between cardinal numbers possesses the following properties: m ≤ n and n ≤ p m≤n m≤n m≤n m≤n imply implies implies m≤p m+p≤n+p (2.. ƒ(y. The proof is straightforward and is omitted here. (2. The function F defined by F(z) = g z associates with every z ∈ Z an element of the set X Y .47). Theorem 2. consisting of all functions ƒ whose values are the numbers 0 and 1 and whose domain is the set A. the function gz defined by g z (y ) = ƒ(y. If A has the power m. Proof: By definition. z 0 ). let us prove Eq.9.53) (2. Equations (2. then the power set p( A) has power 2 m . hence ƒ is a function of two variables y and z. Thus.51)–(2. It is easy to show that this correspondence of the function ƒ to the pair (g.3. 1}A . namely. associating with the function ƒ ∈ {0. then the elements F1 (z 0) and F2 (z0 ) of X Y are distinct.3. the laws converse to Eqs. In the arithmetic of natural numbers.54) express the monotonicity of addition. multiplication and exponentiation with respect to ≤.51) (2. If ƒ1 and ƒ2 are distinct functions from X Y∪Z . (2.

y. and exponentiation hold for the relation <.3.y ∈ (0.2)  = c. the cancellation laws with respect to addition. and this completes the proof.1.y thus determines precisely one z. conversely. then the second x-complex. and.y. precisely that pair which gave rise to z. Let the pair of decimal fractions x and y be given.y and the numbers z is thereby established. We now study a few properties of the cardinal numbers ℵ0 a n d c . then the first y-complex. These theorems hold in arithmetic provided that p > 1. Since the representation of our numbers by means of such decimal fractions is unique.3. followed by the second y-complex. Every number pair x.1]. and p = ℵ0 to obtain a counterexample. every z determines exactly one number pair x. n < ℵ 0 < c : The inequality follows from Theorem 2.Naive Set Theory 31 addition.y) : x. multiplication. Now we write the first x-complex.2). etc. In the arithmetic of arbitrary cardinal numbers all of the cancellation laws fail to fold: it suffices to let m = 2. (2) For each natural number n. in fact. the result follows.1) and [1. They follow without difficulty from the law of trichotomy. This decimal fraction shall be made to correspond to the number pair x. We write the numbers x. Theorem 2. On the other hand. (1) c = c + c: In fact.1) and (x. Then  X  =  Y  = c and X∪ Y  = c + c =  (0.y) ∈ X × X . Then X × X = {( x. and exponentiation. let X and Y be intervals (0. For arbitrary cardinal numbers m and n. Since neither x nor y exhibits only zeros from a certain point on. either m ≤ n or n ≤ m. A one-to-one correspondence between the number pairs x. y can be mapped on the set of decimal fractions z. which now we state without proof.11. the process can be continued without ending and gives rise to a nonterminating decimal fraction. n + c = ℵ 0 + c = c: In fact. (3) For each natural number n. by property 1 and Bernstein’s theorem. y and z by nonterminating decimal fractions. it is necessary to show that the set of decimal fraction pairs x. So. multiplication. The assertion is proved as soon as we show that the sets X and X × X are equipollent. n = 3. (4) c = c · c : Proof: Let X be the set of all points in (0. Let z ∈ (0. c ≤ n + c ≤ ℵ 0 + c ≤ c + c = c. 1]} has power c · c. .

32 Chapter 2 (5) For each natural number n. then x = y . Ordered Sets Since the elements of sets need by no means be numbers or points on the real number line.56) In fact. i.10. The ordered set X will be denoted by the ordered pair (X . if x ≤ y and y ≤ x. i. it follows that whence property 7 follows by applying Bernstein’s theorem.3. we have (2..10. from Theorem 2. It is necessary to remember that an ordering is . exists with its field X such that (1) For all x ∈ X. (3) For x. called an order relation and denoted by ≤. The concept of ordered sets will be formulated abstractly as follows: A set X is called an ordered set if a binary relation.4. cn = c: By induction and property 4 we obtain the proof. it follows from Bernstein’s theorem and property 4 that n ⋅ c= ℵ 0 ⋅ c = c. and even if the elements are numbers. ≤ ).” The field of the order relation ≤ is often said to be ordered without explicitly mentioning ≤.55) Thus.e. from Theorem 2. if x ≤ y and y ≤ z. 2. We have the following: for each natural number n > 1.y.3.e. then x ≤ z .z ∈ X. x = x. (6) For each natural number n. (2) For x. (7) Under the hypothesis that c = 2ℵ0 . n ⋅ c = ℵ 0 ⋅ c = c: In fact. the relation is antisymmetric. this is the continuum hypothesis. the relation is transitive.. it is not at all necessary for them to be ordered according to magnitude. y ∈ X. The notation x > y means the same as y < x and is read either “y precedes x” or “x succeeds y. (2.

1. The element a is called the maximum (respectively. Without confusion. x < a ).. then we say that the elements x and y are comparable. Thus. then we call Y a chain in X.4. a ≤ x). Proposition 2. 6. Let (X. . Proposition 2. i. .1.4. Let X be an ordered set with ordering ≤. is linearly ordered. 3. minimum) element of X provided that for every x ∈ X. . . . minimal) element in the ordered set X provided that no element x ∈ X satisfies x > a (respectively.2. however.4.57) if X = Y and for any x. .e. x ≤ a (respectively. the sets in Examples 2.4. we write X = Y instead. the two sets are identical. } is a linearly ordered set. (4) The set {0. ≤ ) be an ordered set and a ∈ X. y ∈ X. ≤ X ) and (Y. and then the odd integers. Let X be a set ordered by a relation ≤.3. (2. If Y ⊂ X and any two elements of Y are comparable.5. the subsets of ordered sets can always be regarded as ordered. . Two ordered sets (X. Then every subset of X is also ordered by ≤. The element a is called a maximal (respectively. If x and y are elements of X and either x ≤ y or y ≤ x. . . The same set may be ordered by many different order relations.7. first come the even nonnegative integers ordered according to increasing magnitude. ≤ Y ) are equal. (2) The set of natural numbers is ordered by the relation of divisibility. 1}.1(3) and 2. then the odd integers ordered according to decreasing magnitude.1. Hence. that is. x ≤ X y is equivalent to x ≤ Y y . An order relation ≤ of a set X is called a linear order relation if X ordered by ≤ becomes a chain.4. first come the even nonnegative integers ordered according to increasing magnitude. Example 2. if any two elements of Y are incomparable. An ordered set contains at most one maximum element and at most one minimum element. (3) The set of natural numbers in the order indicated by {0.Naive Set Theory 33 by no means an intrinsic property of the set. . then Y is called an antichain in X. 4. . 5. . as written. 2. otherwise they are incomparable.2. (1) Every family of sets is ordered by the inclusion relation. Without regard to an order.4. The next result is obvious.1(4) are to be considered distinct in the sense of ordered sets.

every ordered set is similar to itself. then Theorem 2. implies and implies X ~ Y.. The bijection h is called a similarity mapping from the ordered set (X. The same argument shows that an ordered set contains at most one minimum element. need have neither a maximum nor a minimum element. If a set X is equipollent to an ordered set (Y . Proof: Let h be a bijection from X onto Y. we can define coinitial sets and subsets. We define an order relation ≤ X on X as follows: For any if and only if Theorem 2. ≤ Y ).≤ Y ).58). which implies that a = a * . ≤ Y ) be two similar ordered sets with a similarity mapping h : X → Y. If x 0 is a minimum element from X. Then a * ≤ a and a ≤ a* . The following four fundamental properties result immediately from the definition: (1) (2) (3) If (4) i. then both possess maximum (minimum) elements or neither set has such an element.4. An ordered set. .4. (2. x 2 ∈ X. we write X Y instead of Eq.58) if there exists a bijection h : X → Y such that for any x 1 .4. The set of all real numbers is cofinal and coinitial with the set of all integers. Analogously.1. Without causing confusion. (2. Example 2. An ordered set A is cofinal with its subset B if for every x ∈ A there exists y ∈ B such that x ≤ y.e. x 1 ≤ X x 2 implies h(x 1 ) ≤ Y h (x 2 ). then for every x ∈ X. ≤ Y ). ≤ X ) onto the ordered set (Y. The subset B is a cofinal subset in A. An ordered set (X. ≤ X ) and (Y. If two ordered sets are similar.2. Proof: Let (X. ≤ Y ) . however. ≤ X ) is similar to an ordered set (Y.34 Chapter 2 Proof: Let a and a * be two maximum elements of a set ordered by an ordering ≤. then X can be ordered in such a way that X (Y .2.

by setting * < equal to >. 2. also be denoted by The order type of the set of natural numbers.. . we let s1 ≤ s 2 if and only if s 1 ≤ X s 2.59) It is again easy to see that the sum µ + γ is independent of the choice of the ordered sets (X . and it therefore orders X ∪ Y. If. if X= {1. . For example. we let s 1 ≤ s 2 if and only if s 1 ≤ Y s 2 . Let ( X. 2.62) The concept of addition of order types can be defined for arbitrary many order types. The ordered sum X + Y of the ordered sets (X. Generally. if s 1 and s 2 ∈ X. by similarity relation. Each order type will be denoted by a small Greek letter. 3. The addition of order types need not be commutative. into pairwise disjoint groups. 3. is denoted by ω . ≤ I ) be given. * µ denotes that order type which results from µ when the order of succession of the elements is reversed. then the order type of the set is denoted by * ω. ≤ X ) and (Y. and to each k ∈ I there (2.Naive Set Theory 35 This implies that for every also has a minimum element. .60) . ≤ Y ) be two disjoint = µ and = γ. ≤ Y ) be two disjoint ordered sets. The family of all ordered sets is divided. The order relation ≤ thus defined on the union is obviously transitive. An order relation ≤ on the union X ∪ Y is defined as follows: For any s1 and s 2 ∈ X ∪ Y. Let µ and γ be two order types and ( X . . ≤ Y ) is now understood to be the ordered set (X ∪ Y. ≤ Y ) are of the same order type if they are similar. . The sum of the order types µ and γ is ordered sets such that defined by (2. 2. γ . 3. if s 1 ∈ X and s 2 ∈ Y. where 1 is the order type of Y. . if s 1 and s 2 ∈ Y. . ordered according to increasing magnitude. when a new order relation is derived from the relation < defined for the original set. and the order type of (X. } and Y = {0} then X + Y = {1. ≤ ). . and π is valid: (2. } so that 1 + ω = ω . ≤ X ) and (Y. ≤ X ) will without confusion. then we let s 1 ≤ s 2 .e. but ω + 1 ≠ ω. . i. . ≤ X ) and (Y . on the other hand. 1. let an ordered set (I . In fact. ≤ Y ). ≤ X ) and (Y . Two ordered sets ( X. ≤ X ) and (Y. The associative law of order types µ . each of which contains all sets similar to each other.61) (2.0} and Y + X = {0. these numbers are ordered according to decreasing magnitude.

then 2 ⋅ ω = ω and ω ⋅ 2 is the order type of the ordered set {1. y 2 ) if and only if x 1 < x 2 .66) This set is not certainly similar to any ordered set of type w. ≤ k ). then the In the previous definition. then we let s 1 ≤ s 2 if and only if s 1 ≤ k s 2 . The sum of the order types µ k is defined by (2. (2.63) belong is ordered in the following manner: If two elements s 1 and s 2 of to the same X k . ≤ X ) and ( Y. 5. i. For if we choose K = ω and µ = 2. . y1 ) < (x 2 . 1 } ordered by 0 < 1. if µ k = µ for each k ∈ I and product of the order types µ ⋅ γ will equal the sum thus determined. ≤ Y ) such that = K and = µ . For multiplication there is no commutative law either. which is so important for calculation: (2. The concept of product of order types can also be defined as follows: Let K and µ be two ordered types of nonempty sets ( X. ≤ k ) . We do have. . . as for addition. 2. Then the ordered sum (2. 6. or when x 1 = x 2 then y l < y 2 . . say and then we let s 1 < s 2 if and only if k 1 < k 2 .67) .36 Chapter 2 corresponds an ordered set (X k . for multiplication.64) It is easy to see once more that this sum is independent of the choice of the ordered sets (X k . } (2. ≤ ) by letting (x 1 . s 1 and s 2 belong to different Xk ’s. however. . = γ .65) For γ = 0 we let µ ⋅ γ = 0. where 2 is the order type of { 0. ≤ k ) an ordered set such that µk = where {Xk : k ∈ I ) consists of pairwise disjoint sets. It is also evident from the second definition that the product of order types is independent of their particular set representatives. however. This second definition follows immediately from the first. 4.e. where the sets X k are pairwise disjoint. 2 ⋅ ω ≠ ω ⋅ 2. Hence. 3. Let µ k be an order type for each k ∈ I and ( Xk . Define an ordered set (X × Y. . if. .. the associative law. The order relation defined here is called a Then the product µ ⋅ K = lexicographical order.

70) For example. (y. i = 1.2.x). The right distributive law does not hold. where for any (zi . y i ). ≤ X ). z i ) (2.73) In fact. = π . i = 1.xi )) ∈ Z × ( Y × X ). (2. that is. = γ . ≤ Z ) be disjoint ordered sets such that = µ. If we now associate.Naive Set Theory 37 In fact. ≤ 1 ). (2. and z. and ( µ · γ ) · π is the order type of the ordered set (Z × (Y × X ). xi ) ∈ ( Z × Y ) × X.2.( yi . ≤ 2 ). (Y. x1 < x2 . y 1 < y2 . the order type = γ .y). we define a similarity correspondence between (Z × (Y × X) . ≤ Y ). y1 = y2 . Likewise. (2.71) and this differs from (2. ≤ 1 ) and (( Z × Y ) × X. ≤ Y ). y . or z 1 = z 2 . (Y .74) . ≤ 2 ). y 1 < y2 . or z1 = z2 . and (Z . however. ≤ X ). x)) with the element ((z. is valid. (2. let (X. the element (z.72) The left distributive law. and (Z. where for any ((zi . Then π · (µ + γ ) is the order type of the ordered set where for any (q i .69) if z 1 < z 2. µ · ( γ · π ) is the order type of the ordered set ((Z × Y ) × X. (2. and = π. ≤ Z ) be ordered sets such that = µ . Namely. or z 1 = z 2 and y 1 = y 2 and x 1 < x 2 . Then. for the same elements x. or z1 = z 2 . according to the second definition. let (X. This proves the assertion.68) if z1 < z2 .

It does not yet follow from the results derived thus far that to every cardinal number there corresponds at least one order type µ with  µ  = where  µ  means the cardinality of any of those ordered sets with type µ. Let the set of all order types which belong to a given cardinal number and call this set the type class belonging to Theorem 2. let (2.3.77) It remains to show that numbers ≥2 .79) Then distinct sequences (2. The set of all order types of class is equipollent to the set of all possible orderings of our set X. In other words. it be does not follow from the results derived that every set can be ordered. In fact. This completes the proof of the statement.38 Chapter 2 Now it is evident that the same ordered set is obtained with order type π · µ + π · γ .75) Proof: Let X be a set with cardinality Let µ be any order type in Then it is always possible to order X so that = µ.79).4. With any sequence of natural (2.80) . Then it can be seen that different order relations <1 and <2 on X correspond to different subsets S < 1 and S < 2 of X².4. (2.78) always give rise to distinct order types (2. For the class of all order types of denumerable sets. Therefore. For each order relation < on the set X.78) we form the order type (2. Proof: The previous theorem implies that (2. For any cardinal number the type class satisfies (2.76) Theorem 2. let S < = {(x1 .4. x 2 ) ∈ X ² : x 1 < x 2 }. Let γ = * ω + ω.

78) and (2. i = 1. etc. whereas does. X1 is mapped on X 2 and Y1 is mapped on Y 2 . It follows. If we now apply (a) to the equal sums α = β . and hence.81) follow from (a) finite sets. X 1 is mapped on X 2 and Y 1 is mapped on Y2 . . and X i .78) and (2.Naive Set Theory 39 also be a sequence of natural numbers. Consequently. (b) In fact. Thus. we get have no first element. This follows by induction from the following auxiliary consideration: Let X i and Y i be distinct ordered sets.80) coincide. which is impossible. under a similarity mapping of X 1 + Y 1 onto X 2 + Y 2 . by (a). whereas every element of Y 2 is preceded by infinitely many elements in X 2 + Y 2 . because Y 1 does not have a last element.2. are and so that. by (b). Then the relations (2. by induction it follows that the sequences (2.80) coincide. otherwise the subset Y 1 preceding x 1 in Y1 + X 1 would have to be mapped on a part of the elements of Y 2 . also in this case. Hence. i = 1. In like manner we see that no element of Y1 can correspond to an element of X 2 . that no element of Y1 can correspond to an element of X2 . In (b) under a similarity mapping of Y l + X 1 onto Y 2 + X 2 . likewise. no element of Y 2 can correspond to an element of X 1 because every element of X1 is preceded by only a finite number of elements in X 1 + Y l . for the case (a). Then we are to show that the sequences (2. and with it we form the order type Suppose that α = β . an element x 1 of X1 cannot correspond to an element y2 of Y 2 .2.

≤ X ) and ( Y. they can be written as sequences (2. In a densely ordered set no element has either a direct successor or a direct predecessor. for all i < n . All other densely ordered sets contain infinitely many elements. All unbordered. and no other y k . such that n 0 < Y n 1 when x 0 < X x 1 .5 [G. j < n such that between y i and y j .79) implies that any set with order type (2. for all i < n . In each case. Theorem 2. In fact. ≤ Y ) be two sets of the kind mentioned in the theorem.82) and (2. or between x 0 and x1 under ≤ X . or n 0 Y > n1 when x0 X > x 1 . as well as the empty set. denoted by n1 . Since they are denumerable. Proof: Let ( X. To this end. This is possible because Y is unbordered. (2) y n Y > yi . a mapping h from X into Y is defined. ≤ Y ) is unbordered and dense. are densely ordered. All one-element sets. then Let yn be the first element in (2. let x 0 correspond to any element n 0 of Y. is (3) there are i . say n 0 = y 0 . h is onto. there is an element n 2 with the smallest index in sequence (2. We say also that A is dense.79) has cardinality has at least the power of the set of all sequences (2. Now x 2 may appear before. Then there are three possibilities: (1) y n < Y yi . Suppose that h is not onto.78) gives rise to distinct order types (2. for k < n. The power of these sequences of numbers is A set A is densely ordered by an order relation < if < is linear and for any two elements x and y ∈ A there exists an element z ∈ A between x and y.79).4. We then look for the element in Y – {n 0 } with the smallest index.78). An ordered set is called unbordered if it is not empty and has no maximal or minimal elements. denumerable sets are similar to one another. In this way.83) We now show that X and Y with their orders ≤ X and ≤ Y can be mapped onto each other.40 Chapter 2 Since the totality of distinct sequences (2. . and since (2. We associate x1 with n1 . after.83) such that y n ∈ Y – h (X). Cantor]. dense. since (Y .83) which has the same position relative to n 0 and n 1 as x 2 has relative to x 0 and x1.

In other words. for all y ∈ Y }. let X* be the totality of elements of A which belong to X or precede an element of X. A similar argument shows that (2) and (3) do not hold. for all x ∈ X } and X = {a ∈ A : y A ≥ a .4. Every continuous set (A. Y ) be a cut in a linearly ordered set A.6. (A. Let x m be the first element in sequence (2.4. there is always a subset of the former which is similar to the latter: Let (A.4. The proof of Theorem 2. Given any dense set with at least two elements and any linearly ordered denumerable set. Then X ∩ Y contains at most one element. ≤ Q ). Theorem 2. Proposition 2.82) such that x m <X h –1 (yi ). it follows that there exists x ∈ X such that x < X h – 1 (yi ).5 yields the following result. and. If ( X*. ≤ A ) contains a subset Q which is similar to the set of all rational numbers. contradiction. ≤ A ) be a linearly ordered set. Y *) is a cut with a gap.4. Thus. h is a similarity mapping from (X. Then the definition of h implies that h (x m ) = yn . Therefore.3. the set (A .Naive Set Theory 41 If (1) is true. X ≠ ≠ Y) has a nonempty intersection X ∩ Y. ≤ A ) contains an element a which succeeds all elements of X while preceding all elements in Y. ≤ Y ) . Let (X. x = y. then it must be both x ≤ y and y ≤ x. A pair (X. defined by ≤Q = ≤A ∩ Q². and since (X. X* + Y * lacks at least one element a of A. and let Y* be the totality of elements of A which belong to Y or succeed an element of Y. Y ) is a cut in the ordered set A. of the cut. Proof: If there are x and y ∈ X ∩ Y . and Y the upper section. i < n. Theorem 2. there is at least one proper cut (X. Y ) in (Q. according to the construction of .7. Proof: According to Theorem 2. The set X is called the lower section. contrary to the assumption that (A. ≤ X ) is unbordered. the proof of which is left for the reader. We also say that A is continuous. ≤A) would have a gap. where ≤ Q is the restriction of ≤A on the set Q.6. In fact.84) be a decomposition of Q where the cut (X.. The set Q ordered by ≤ A has gaps. ≤ A ) contains a subset which is similar to the set of all real numbers in their natural order. i < n.e. ≤ X ) onto (Y . if Y = {a ∈ A : x ≤ A a. Y ) in A (i.4. Y ) determines a gap in Q Then (A. A set A is continuously ordered if A is linearly ordered and each proper cut (X. such that X ∩ Y = Let Q=X+Y (2. Hence. ≤ A ) is continuous.

due to (3). and hence. this element succeeds X * and precedes Y *. 2. For the set of all rational numbers can be mapped on U.Y ) and keep it fixed. according to Dedekind’s introduction of the real numbers.42 Chapter 2 X * + Y * . Proof: The set U is unbordered. so it also succeeds X and precedes Y. We have thus obtained a similarity mapping of the set of all real numbers on the set A.8. however.84) thus determines at least one element a of A. precisely one element of A. Every cut (2. Let ( A. which proves the theorem. implies the assertion. ≤ A ) be an ordered set with the following properties: (1) ( A. not more than one element of A. In other words. Let is a proper cut in with a gap} This set is also ordered by ≤ A and is similar to the set of all rational numbers together with their gap elements in the domain of real numbers. Well-Ordered Sets An ordered set ( X. ≤ X ) is well-ordered if it is linearly ordered and every nonempty subset of X contains a first element with respect to the relation ≤ X . Likewise. . consequently. Then (A. Theorem 2. we choose one of these elements a(X. (3) (A. ≤ A ) contains a denumerable set U such that between every pair of elements of A there exists at least one element of U. For every gap (2.7. Every gap in U determines. because of (3).5. because before and after every element of U there are still elements of A. elements of U. This. whereby with every gap in the domain of rational numbers there is associated precisely one gap of U and. ≤ A ) is continuous.4. (2) (A. according to the proof of Theorem 2. Consequently. R * is similar precisely to the set of all real numbers.4. at least one element a of A and.84). ≤ A ) is similar to the naturally ordered set of all real numbers. U is similar to the set of all rational numbers ordered naturally. ≤ A ) is unbordered. it follows from (2) and (3) that U is dense also.

3.0} is not well-ordered because it does not have a first element.2.5. Now it follows from (2.5.85) We have for every x ∈ P. Let p 0 be any element in P. then y ∈ B. it contains a subset of type * ω. We define by Since by letting induction a sequence p n < p n –1 .5.. for n > 1. This shows that {y ∈ A : y < x } ⊂ B. B ⊂ A. Theorem 2.5. Theorem 2.5.1. and if for every x ∈ A the set B satisfies the condition (2.}.5 [Principle of transfinite induction]. Theorem 2.. A function f : A → A is an increasing function if f satisfies the condition (2.. The set of all real numbers in the interval [0. Theorems 2. Proof: Suppose that This means that if y < x.1.3 imply the next results. Every finite linearly ordered set is well-ordered and so is the set = {0.2 and 2.3. Then there exists a first element in A – B. In every well-ordered set there exists a first element.1. this sequence is of type * ω. Let ( A.86) that x ∈ B. which contradicts the hypothesis that x ∉ B. Proof: Let P be a nonempty subset of A which contains no first element. The following two theorems are simple consequences of the definition.5.3.86) then B = A.4..2. it is necessary and sufficient that it contain no subset of type * ω.1.87) .5.. Theorem 2. Theorem 2. is not well-ordered because the subset (0. If a linearly ordered set A is not well-ordered. If a set A is well-ordered.5. ≤ ) be a linearly ordered set. Every element except the last element (if such exists) has a direct successor. in their natural order. In order that a linearly ordered set be well-ordered..Naive Set Theory 43 Example 2.2. The set {.1]. 1] does not have a first element. No subset of a well-ordered set is of type * ω. Let (2.

we have x ≤ A f (x ) and finally x ∈ B.87) it follows that Since y ∈ B.5. f (x ) ∈ A – O ( x ).5. or (ii) A is similar to a segment of B. If the well-ordered sets A and B are similar. where (A. This means that the element f (x) occurs after every element y of O(x ). Then. Corollary 2. we have y ≤ A f (y) and thus y < A f (x ). then there exists only one similarity mapping from A onto B.44 Chapter 2 Theorem 2. If a function f : A → A is increasing. where ( A. Proof: Let We now show that x ∈ B.5. By (2. (i) A and B are similar. is clearly increasing in A. Consequently.4 the set B equals A. Then the function f establishing the similarity from A onto would be increasing and would satisfy f (x ) ∈ that is. from Theorem 2. Let A and B be two well-ordered sets. f (x) < A x.7. y < A x. let y ∈ O ( x).5. Theorem 2.2. But this contradicts Theorem 2. then for every x ∈ A we have x ≤ A f (x ) . ≤ A ) is a wellordered set.88) Suppose that A and are similar. The initial segment is the set defined by (2.5. In fact.5.5. g (x ) ≤ B f (x ). This implies Corollary 2. . have instead of By the same argument f (x ) ≤ B g(x ).6.5. that is.5 we thus The function for every x ∈ A. Hence. Hence. ≤ A ) is a well-ordered set. By Theorem 2. Since x is the first element of A – O(x ). that is. Proof: Let x ∈ A. No well-ordered set is similar to any of its initial segments. Proof: Suppose that the sets A and B are well ordered by ≤ A and ≤ B and that there exist two functions f and g establishing similarity from A onto B. or (iii) B is similar to a segment of A.1.

90) Finally. Similarly.91) . x ∈ Z. either f ( Z) = B or else f(Z ) is a segment of B : To show this it suffices to observe that (2. then |A | ≤ | B | or | B | ≤ | A | . such that First we show that either Z = A or Z is a segment of A that is. we observe that f establishes the similarity between Z and f(Z ) . If A and B are well-ordered.2.7 implies that we can define a “less than” relation for ordinals. which contradicts the definition of a segment. Corollary 2. f (x ) < f (x ). we have just shown that x < x' ∈ Z implies that of therefore. let Since is a segment a ∈ A such that of the function f also maps onto a segment of B. is a segment Indeed. a ∈ Z . Therefore. there exists an In fact. because then that is. We write α ≤ β instead of α < β or α = β. By ordinal numbers (or ordinals or ordinalities) we understand the order types of well-ordered sets. Theorem 2.5. This relation will be denoted α < β or β > α. For any ordinals α and β one and only one of the following formulas holds: (2. We say that an ordinal α is less than an ordinal β if any set of type α is similar to a segment of a set of type β. thus. Theorem 2. respectively. and let (2.89) such By Corollary 2. is impossible.5. Case (4) and.Naive Set Theory 45 Proof: Let ≤ A and ≤ B be relations which well-order A and B. we have one of the following four possibilities: The first three possibilities correspond to those stated in the theorem. by definition of Z.5.5.3. Hence.8. for given x ∈ Z there exists only one segment that Let a function f : A → B be defined by f(x) = y for each x ∈ A.

5. then we would have β < α and then B would be similar to a segment of h(A). If the well-ordered sets A and B are of types α and β and if A is similar to a subset of B. Proof: Let A. then α < γ. By assumption. (2) O(a 1) is a segment of O(a 2 ) or O (a 1 ) = O (a 2 ). γ. Theorem 2. A is similar to a segment of B and B is similar to a segment of C. Associating the type of the segment O(a) with the element a ∈ A. It is easily seen that the following conditions are equivalent: (1) a 1 precedes a 2 or a 1 = a 2 .5. Proof: Let A be a well-ordered set of type α. β. Thus.5. This shows that the relation ≤ indeed orders W ( α) in type α.12. The set W ( α) consisting of all ordinals less than α is wellordered by the relation ≤.10. respectively. If α is not the smallest ordinal of A. Theorem 2. This contradicts Theorem 2. Proof: Let h : A → B be a mapping such that h establishes the similarity between A and h(A).46 Chapter 2 This theorem is a direct consequence of Theorem 2. and C be well-ordered sets of types α.5.5. and γ are ordinals such that α < β and β < γ. the type of W (α ) is α. Every nonempty set of ordinals has a first element under the relation ≤. A is similar to a segment of C. Moreover. we obtain a one-to-one mapping of A onto W ( α). then α ≤ β.5. then A ∩ W ( α) ≠ Then in the set A ∩ W ( α) there exists a smallest number β because W( α ) is well-ordered. β is the smallest ordinal in A. If the theorem were not true. (3) The type of O(a1 ) is not greater than the type of O(a 2 ). β.7. The following formulas can be proved without difficulty: Theorem 2. At the same time.5. B. Theorem 2.11. Proof: Let A be a nonempty set of ordinals and α ∈ A. .9. If α.

92) This implies that there exists no infinite increasing sequence of distinct remainders.15. µ < γ always implies ϕ (µ ) < ϕ (γ ). {0. Then every µ +1 ≤ σ .} = 3. Then the order type 0 of the empty set is a limit ordinal. Theorem 2. This implies that α = λ + n.5. This is a consequence of Theorem 2. So there exists only a finite number of m ∈ such that there exists a remainder of power m.16 there exist ordinals greater than all ordinals ϕ (γ ) where γ < λ. {0.5. Every decreasing sequence of ordinal numbers contains only a finite number of ordinal numbers.1} = 2. For example. an α-sequence ϕ o f ordinals will be called an increasing sequence if for any µ.. for every µ ∈ M. All order types of finite well-ordered sets are denoted by the corresponding natural numbers.12.5.5. Theorem 2. Theorem 2. Every set of the form A – O (a) is a remainder of A.5. A transfinite sequence of type α (or an α-sequence) is a function ϕ whose domain is W(α)..5. γ ∈ W ( α). it is a decreasing sequence. The smallest .13. Every set of ordinal numbers can be written as an increasing sequence. if µ < γ invariably implies ϕ ( µ ) > ϕ ( γ ) . If the values of ϕ are ordinals. there is an ordinal number which is greater than every ordinal number µ in M. {0} = 1.Naive Set Theory 47 An ordinal number is a limit ordinal if it has no direct predecessor. A – O(a 1 ) ⊂ A – O(a 2 ) if and only if (a1 ≥ a 2 ) (2. By Theorem 2. the sequence is also called a sequence of ordinal numbers.14.1. To every set M of ordinal numbers µ. where λ is a limit ordinal and n is finite ordinal (natural number). hence.5. µ < σ . Let ϕ be a λ -sequence of ordinals where λ is a limit ordinal.2.12. then the segment O(a) has no last element. Proof: By Theorem 2. Proof: Let α be an ordinal and A a set of type α.. This result follows from Theorem 2. Let α be an ordinal number.16. we can assume that M is an increasing sequence. In particular.5. Theorem 2. Thus O(a) is a limit ordinal λ. If n is the greatest such number and if A – O (a) is a remainder of power n. Let M * = { µ + 1 : µ ∈ M } and σ = U {W ( µ + 2) : µ ∈ M}. Clearly.14. Each ordinal can be represented in the form λ + n..

( α ≤ β ) → ( α + γ ≤ β + γ) (2. there exists ξ ∈ W(λ) such that η < ξ. consequently.93). For every ordinal ξ < α there exists an ordinal ϕ (ξ) in A such that the set { η ∈ A : η < ϕ (ξ)} is of type ξ.19.93) holds. is a segment of C + B.21 [The Second Monotonic Law for Addition]. The sequence ϕ (ξ) is clearly increasing and ϕ ( ξ) < λ for ξ < α because ϕ (ξ) ∈ A ⊂ W (λ ) . We have η < λ for all η ∈ A and. Hence. Since α < β. ξ < α .95) . Thus. The sum and product of two ordinal numbers are ordinal numbers.94) Proof: Let B and C be disjoint sets such that B = β and C = γ. This means that some ordinal in A is greater than ξ.20 [The First Monotonic Law for Addition]. This shows that λ is the least ordinal greater than all ordinals ϕ ( ξ ). since λ is a limit ordinal. µ < ξ ≤ ϕ(η ) for some η < α. Theorem 2. The ordered sum of ordinal numbers. in turn. ( α < β ) →( γ + α < γ + β ) (2. We now consider the arithmetic of ordinal numbers. The ordered sum C + A. We say that an ordinal λ is cofinal with a limit ordinal α if λ is the limit of an increasing α-sequence (2. Proof: Let A be a subset of W(λ) cofinal with W( λ ) and such that A = α. Suppose.12) is called the limit of the λ-sequence ϕ (γ ) for γ < λ and is denoted by lim γ < λ ϕ ( γ ). if ξ ∈ W ( λ) then ξ < λ . which proves Eq.5. which is of type γ + α. Let A be the set of all values of ϕ.5. Conversely.5. there exists an ordinal ξ < λ such that µ < ξ because the sets A and W (λ) are cofinal.5. B contains a segment A of type α. If µ < λ . Theorem 2. Thus. (2. Theorem 2. and by the definition of limit there exists ξ′ < α such that ξ < ϕ (ξ′ ).48 Chapter 2 such ordinal (see Theorem 2.18.5. which is of type γ + β. is an ordinal number. Theorem 2.93) Theorem 2. where the index set is also well-ordered.17.5. the sets A and W( λ) are cofinal. that Eq. An ordinal λ is cofinal with the limit ordinal α if and only if W(λ) contains a subset of type α cofinal with W( λ ) . These theorems are consequences of the definitions in the previous section. (2. γ + α < γ + β.

If α ≥ β . γ 1 = γ 2 by Theorem 2. The operation of ordinal exponentiation is defined by transfinite induction as follows: (2. γ is the base. (2. Thus. and let A be a segment of B. then there exists exactly one ordinal γ such that α = β + γ. where B = β and C = γ. Theorem 2.100) where λ is a limit ordinal. and C be the sets defined in the proof of Theorem 2.10 to the ordered sums A + C and B + C we obtain (2. To prove the uniqueness of γ suppose that β + γ 1 = β + γ 2 . Let A = α .97) cannot be replaced by < . .97). We say that γ α is the power of γ . Theorem 2.5. The Cartesian product A × C ordered lexicographically is a segment of B × C.22.Naive Set Theory 49 Proof: Let A.98) (2. Thus. B be a segment of A of type β .95). (α ≤ β) → (αγ ≤ βγ) (2. respectively. Let α and β be ordinals.23 [The First Monotonic Law for Multiplication].5. and γ = A – B. Clearly.5. and α the exponent.5. β .97) Proof: Let A. It follows from the identity 1 · ω = 2 · ω that ≤ in Eq.99) (2. which proves Eq.8. By (2. for γ > 0 (2. (2. and C be well-ordered sets of type α . B .20.5. and we assume that A ⊂ B.96) Proof: In fact. C × A ⊂ C × B . Theorem 2. B. γβ is the type of the Cartesian product B × C ordered lexicographically. Proof: Let A = α . α = β + γ. ( α < β ) → ( γ α < γβ) .24 [The Second Monotonic Law for Multiplication].5. Then by applying Theorem 2. and γ .94) this implies that γ 1 ≥ γ 2 and γ2 ≥ γ 1 .

By induction it follows that B = W(η + l). ς ∈ B.50 Chapter 2 Theorem 2. η . β .5. and γ . then W (ς) ⊂ B → ς ∈ B (2.25. ς 1 ∈ B.104) Because the sequence γ α . (2.. and γ with γ > 1.27. implication (2. In case (i).103) In fact. α < β → γα < γ β (2. thus. Finally. ξ + ς is a limit ordinal.105) Since for α < ς we have α ∈ W (ς ). the following three cases are possible: (i) ς = 0. Theorem 2. For any ordinals ξ. This proves Eq. i. (iii) ς is a limit ordinal > 0.5. thus.26. Hence. (ii) ς is not a limit ordinal. For any ordinals ξ . we have (2. in case (iii).5.e. η ∈ B. which implies that ς ∈ B.102) Proof: Given an ordinal ξ . Hence. by assumption. it follows that α ∈ B.106) . (2. ς = ς1 + 1 where ς1 ∈ W ( ς ). α < ξ + ς . For any ordinals α . Thus. let B denote the set of those ς ∈ W ( η + 1) for which We shall show that if ς ≤ η . and therefore which shows that ς ∈ B. and γξ+η = γξ · γ η γ . ( γξ )η = γ ξη (2. (2.102). Theorem 2. is increasing when γ > 1.103) is proved. Thus. η .101) This follows from the definition. because In case (ii).

Finally. σ 1)) = α. where Notice that there exists an ordinal pair ( β1 . ς ∈ B . Hence Eq. Proof: By transfinite induction it can be shown that for any ordinal α. if ς is a limit ordinal > 0. … . every ordinal number α can be uniquely represented in the form where n is a natural number. n. Let B denote the set of those ς ∈ W (η + 1) for which ( γ ξ )ς = γ ξη . Therefore. because (γ ξ ) 0 = 1 = γ 0 = γ ξ·0 . We now let ς be the smallest ordinal such that α < γ ς .5. where β1 < γ and σ1 < γ η1 . and 0 ≤ β i < γ . This contradicts then the definition of ς . i = 1. which implies γ ς ≤ α. For any given ordinal γ > 1. where ς 1 satisfies the condition that (γ ξ ) ς 1 = γ ξ ς 1 . … . In case (ii) we have ς = ς1 + 1. for i = 1. and γ λ ≤ α for λ < ς . For the same reason. It suffices to show that implication (2. σ1 ) ∈ W(γ ) × W(γ η1 ) such that O (( β 1 . w h e r e {β 1 } × O (σ1). β 2 < γ.103) holds.n. If ς were a limit ordinal. In case (i). Theorem 2. This formula in turn implies which shows that ς ∈ B. then we have ς ∈ B. there exist ordinals η2 < η1 . and ηi and β i . (2. and σ2 < γ η 2 such that . ς = η1 + 1. As before we consider cases (i)–(iii). 2 .Naive Set Theory 51 Proof: This argument is analogous to that of the previous theorem. 2. Thus.28.106) is proved. are ordinals such that η 1 > η 2 > … > η n .

This implies that there exists a relation on A well-ordering a subset of A into type α. 2 . β 1 = θ 1. Every set can be well-ordered. there exists a smallest ordinal β such that ϕ β = p. and Repeating this procedure. . p.5. Then there exists a set of all ordinals β . . where p is any fixed element which does not belong to A. we have n = p.52 Chapter 2 This procedure must stop within a finite steps because. and where j = 1 .107) (2. . . f (B) ∈ B. .5. and this contradicts Theorem 2. } would be a sequence of type * ω of ordinals numbers. This theorem is also of the greatest importance for the theory of cardinal numbers because it shows that all cardinal numbers are comparable with each other. Theorem 2.29 [Well-Ordering Theorem]. We can extend this function to the whole power set p (A) by letting = p. If n. we must have η 1 = ξ 1 . But this contradicts the definition of α.108) If for each then there would exist a transfinite sequence of type for all ξ < α we had α with distinct values belonging to A. We now define a transfinite sequence of type α by transfinite induction such that (2. η 2 . where β = the order type of some R ∈ C. Well-ordered sets owe their importance mainly to the fact that for each set there exists a relation which well-orders the set. such that for each nonempty subset B ⊂ A. . there exists a natural number n such that with the property that Now we show the uniqueness of the expression. A is the set of all terms of a transfinite sequence of type β whose terms are all distinct. and βk = θ k for k ≤ n.12. . Therefore. there exists a relation well-ordering A into type β. Consequently. Proof: Let f be a function defined on p (A) – where A is a nonempty set. otherwise. { η1 . By the definition of η1 and ξ1 . . Let α be the smallest ordinal greater than every ordinal β in this set. This implies that thus. Now let C denote the set of all relations R ⊂ A × A well-ordering their field. η k = ξ k . Therefore.

. The totality of ordinal numbers which represents the same cardinal number ℵ is designated the number class Q(ℵ ). etc.5. By Theorem 2. According to the well-ordering theorem.1. clear if every cardinal number is represented by the initial number belonging to it. and the initial number would be γ + 1. γ + 1 would not be the smallest ordinal number in its number class. then we denote the smallest transfinite cardinal number by ℵ0 . every transfinite cardinal number receives as index . every cardinal number can be represented by an ordinal number.4.5. Proof: For the first part of the assertion we have to show that every nonempty subset of cardinal numbers contains a smallest cardinal number. is well-ordered.31. This ordered set will prove to be actually well-ordered.Naive Set Theory 53 Theorem 2.11 and 2.30. This is. Theorem 2. it is practical to represent the cardinal numbers by means of ordinal numbers. Every set of cardinal numbers. however.5.30. There exists a cardinal number which is greater than every cardinal number in the set. If the family of all transfinite cardinal numbers is ordered according to increasing magnitude. every cardinal number can be represented by a well-ordered set. briefly. For this and similar considerations concerning cardinal numbers. where Q (ℵ ) contains a smallest ordinal. The second part of the assertion follows from Theorems 2. But then γ and γ + 1 would have the same cardinal number. contradiction. precisely one of the following holds: i. ordered according to increasing magnitude.5. every set of cardinal numbers can be ordered according to the increasing magnitude of its elements. and the corresponding cardinal number is the smallest cardinal number in the set. the next by ℵ1 . Proof: If some transfinite initial number were not a limit number. any two cardinal numbers are comparable with each other. i.e. Among these initial numbers then there is.8.5. a smallest one. Theorem 2. If A and B are two arbitrary sets. Every transfinite initial number is a limit number. The proof follows from the well-ordering theorem and Theorem 2.5.32.. it would be immediately preceded by an ordinal number γ.e. In general. which is called the initial number of Q(ℵ) or the initial number belonging to ℵ. as in every nonempty set of ordinal numbers.

Define a polynomial Since only a finite number of nonzero terms appear. Therefore. . Then ω 0 is the smallest transfinite limit number and is therefore the same as the ω introduced before. the cardinal number of the set of all solutions for all is at most We now need only show that Theorem 2.28 implies that Consequently. the cardinal number We now show that this set has at most the cardinal number Then by Theorem 2. For any ordinal µ > 0.109) The initial number which belongs to ℵµ is denoted by ω µ . we can represent ξ and η uniquely in the form (2.110) Proof: The set W ( ω µ ) has order type ω µ and is therefore a set of power is the power of the set Consequently.111) where γ 1 > γ 2 > .112) . > γ k and for all i ≤ k either m i > 0 or n i > 0. η by solving the equation But for every this equation has only a finite number of solutions.33.5.54 Chapter 2 the ordinal number of the set of transfinite cardinal numbers which precede the cardinal number in question.28 implies that for any ordinal α.28 and adding whatever Let necessary terms with coefficient 0.5. We therefore certainly obtain all number pairs ξ. (2. (2. (2. Theorem 2. it follows that and the proof of Theorem 2. . Therefore.5.5.

34. (1) Proof: We have invariably . for every γ < µ we also have γ + 1 < µ. i. Thus.5. These two results then yield the assertion in this case. we have α = ωβ. and (2) if µ is a limit number. . Then Theorem 2. Proof: and Theorem 2. on the other hand.35.e. Therefore. too.5. then. the sum is greater than every cardinal number which is less than . when α is a limit number.36. and. so . This implies that ω µ = ωβ for some fixed ordinal β. let . Theorem 2.Naive Set Theory 55 where σ < ω. on the one hand.5. If µ is a limit number. For Proof: Without loss of generality. in particular. For .. Therefore.

2.113) This implies that From this we have have Let be the order type of Q implies that and But Theorem 2.38. The number class Q ordered according to increasing magnitude of its elements.3. Proof: According to Theorem 2.113) implies that . has order type ωµ +1 and therefore has cardinality Proof: The elements of Q are the ordinal numbers γ with Hence. we have (2.5.35 implies that we cannot Then Eq.5. These contradictions are connected with the following concepts.37. Crises of Naive Set Theory Careful readers may have already found some inconsistencies in the development of naive set theory.5.9 and the hypothesis and hence Theorem 2. For ordinals γ and µ with µ ≤ γ + 1. in the sense of addition of well-ordered sets. . (2. Hence.56 Chapter 2 Theorem 2.6.

the concept of “set of all ordinal numbers” is a meaningless concept. The set of all subsets of X. From the list of paradoxes above.1. For example. This is the so-called Burali-Forti paradox. according to Theorem 2. The first case cannot occur because otherwise X would have an element. For further study of naive set theory. however. appear in each of them.” The concept “set” was introduced by Cantor as “a collection into a whole of definite.4. let A be the set of all cardinal numbers. however. Accordingly. and for this very reason X would be contained in the totality of all such sets. called the axiomatic set theory on ZFC. the reader is advised to consult (Kuratowski and Mostowski.5. i.5.16 implies that to every set of ordinal numbers there exists a still greater ordinal number. X. as for every set.33..5. namely.16 implies that there exists a cardinal number This contradicts Theorem greater than all numbers in A. well-distinguished objects. also leads to a contradiction. in X.” we may be in good shape to establish a set theory without contradictory propositions. Then Theorem 2. has a greater power than X. (d) The set of all sets. One of the two words is “set” and the other is “all. . For if X is such a set.e.Naive Set Theory 57 (a) The set of all cardinal numbers. The second assumption. For this set X. or it does not. and the meaning of “all” we have been using contradicts some results we get in set theory. (b) The set of all ordinal numbers. X would be a set which did not contain itself as an element. That is. or words. only the following two cases are conceivable: either X contains itself as an element. For example. 2. (c) The set of all sets which do not contain themselves as elements (Zermelo–Russell paradox or simply Russell’s paradox). In the next chapter. Theorem 2. For in this case. The concept “all” actually was never introduced in the context. we can see that two concepts.” The meaning of this definition might be too wide for us to avoid these paradoxes. 1976). Our analysis on the appearance of the paradoxes shows that if we can narrow the meaning of the concept of “set” and if we can avoid using the word “all. which contained itself as an element. we will try along this line to study properties of abstract sets and to set up a new theory of sets. it would have to contain the set of all its subsets.

i. The characteristic of a formal language is that there are precise rules of formation for linguistic objects. we establish a set theory which makes more precise the basic principles of naive set theory. list the axioms of ZFC. called first-order predicate calculus. 3.1. It is generally understood that some principles are correct and some are questionable.1 we introduce the language of axiomatic set theory. 59 .2 sketches the development of the ordinal and cardinal number systems. The idea of establishing a theory upon a set of axioms harks back to the time of Euclid. the word “system” has been understood in many different ways (see Chapter 4 for a detailed discussion). Some Philosophical Issues Most scientific workers have little need for a precise codification of the concepts of set theory. a language with many grammatical inconsistencies. or artificial. Martin’s axiom (MA) is introduced to study some related problems. excluding regularity.CHAPTER 3 Axiomatic Set Theory This chapter investigates how naive set theory can be rebuilt so that the resulting set theory will not contain the paradoxes in Section 2.e. In this chapter. Usually. Section 3.6. based on the ZFC axiom system (i. we build the theory on a few axioms.e. A major disadvantage of using an informal language is the possible misunderstanding of a sentence or word.4 develops the class WF of all well-founded sets and shows that all mathematics takes place within WF.. Axiomatic set theory. Section 3. based on the axioms of ZFC. For example. In Section 3.. principles. Zermelo–Fränkel axioms and the Axiom of Choice) is introduced. in science. we state our axioms in a formal. Section 3. language. and laws are stated in an informal language. In order to establish a rigorous set theory. such as English. axioms.3 covers some special topics in combinatorial set theory. and discuss some related philosophical problems. To accomplish this. in the past few decades.

60

Chapter 3

The basic symbols of our formal language include , =, and v j for each natural number j. Their grammatical meanings are as follows:

• means “and.” • ¬ means “not.” • ∃ means “there exists.” • ∈ means membership. • = means equality. • v j , for each natural number j, are variables.
The parentheses are used for phrasing. An expression is any finite sequence of basic symbols. The intuitive interpretation of the symbols indicates which expressions are meaningful; these are called formulas. Precisely, a formula is defined to be any expression constructed by the rules: (a) are formulas for any natural numbers i,j.

(b) If φ and ψ are formulas, so are for any i. To save ourselves the work of always writing long expressions, we will use the following abbreviations. (c) (d) (e) (f) (g) (h) Other letters from English, Greek, and Hebrew are used for variables. In this chapter, we follow standard mathematical usage of writing expressions mostly in English, augmented by logical symbols when this seems necessary. For example, we may say “there are sets x, y, z such that ,” rather than (3.1) A subformula of a formula φ is a constructive sequence of symbols of φ which forms a formula. The scope of an occurrence of a quantifier ∃ v i is the unique subformula beginning with that ∃ v i . For example, the scope of the quantifier ∃v 2 in formula (3.1) is An occurrence of a variable in a

Axiomatic Set Theory

61

formula is called bound if and only if it lies in the scope of a quantifier acting on that variable, otherwise it is called free. For example, in formula (3.2), (3.2) the first occurrence of v1 is free, but the second is bound, whereas v 0 is bound at its occurrences and v 2 is free at its occurrence. Note that since ∀ v i is an abbreviation for ¬ ∃ vi ¬, it also binds its variable v i , whereas the abbreviations , → , ↔ are defined in terms of other propositional connections and do not bind variables. We often write a formula as φ(x 1 , . . ., x n ) to emphasize its dependence on the variables x 1, . . , xn . Then, if y 1 , . . . ,y n are different variables, we use φ(y 1 , . . . ,y n) to indicate the formula resulting from substituting a yi for each occurrence of the free variable xi . Such a substitution is called free, or legitimate, if no free occurrence of an x i is in the scope of a quantifier ∃ y i . The idea is that the formula φ( y 1 , . . . , y n ) says about variables y 1 , . . . , y n what the formula φ(x 1 , . . . , x n ) said about x 1 , . . . , x n , but this will not be the case if the substitution is not free and some y i gets bound by a quantifier of φ . In general, we will always assume that our substitutions are legitimate. The use of the notation φ (x1 , . . . , x n ) does not imply that each x i actually occurs free in φ (x 1 , . . . , x n ); also, φ( x 1 , . . .,x n ) may have other free variables, which in this discussion we are not emphasizing. A sentence is a formula with no free variables. Intuitively, a sentence states an assertion which is either true or false. If S is a set of sentences and φ is a sentence, then S φ means that φ is provable from S by a purely logical argument that may quote sentences in S as axioms but may not refer to any intended “meaning” of ∈ . φ , where S is the empty set of sentences, we write φ and say that φ If S is logically valid. If ( φ ↔ ψ), we say that φ and ψ are logically equivalent. Let φ be a formula; a universal closure of φ is a sentence obtained by universally quantifying all free variables of φ. For example, let φ be the formula (3.3) Then ∀x ∀ y φ and ∀y ∀x φ are universal closures of φ . All universal closures of a formula are logically equivalent. In general, when we assert φ, we mean to assert its universal closure. Formally, if S is a set of sentences and φ is a formula, we define S φ to mean that the universal closure of φ is derivable from S. We call a formula φ logically valid if its universal closure is logically valid; we say two formulas φ and ψ are logically equivalent if φ ↔ ψ is logically valid. A set S of sentences is consistent, denoted by Con(S), if for no formula φ does φ . We can see easily that for any sentence φ, S φ if and only if S φ and S is inconsistent. In order to use formal deduction, we need the following important facts: (a) If φ. (b) If S is inconsistent, S φ , then there exists a finite S 0 ⊂ S such that S 0 there is a finite S 0 ⊂ S such that S 0 is inconsistent.

62

Chapter 3

As an exercise, we now list all axioms in the ZFC axiom system. Axiom 1. (Set Existence)

Axiom 2. (Extensionality)

Axiom 3. (Regularity)

Axiom 4. (Comprehension Scheme) For each formula φ with free variables among x, z, w1 , . . . , w n ,

Axiom 5. (Pairing)

Axiom 6. (Union)

Axiom 7. (Replacement Scheme) For each formula φ with free variables among x, y, A, w1 , . . . , wn ,

where the symbol ∃ ! means “there exists exactly one.” Axiom 8. (Infinity)

Axiom 9. (Power Set)

Axiom 10. (Choice (Axiom of choice or AC))

∀ ∃ R(R well orders A) A

Axiomatic Set Theory

63

In the following, we write the meanings of the axioms in English.

Axiom 1. (Set Existence) There exists a set. Axiom 2. (Extensionality) Two sets x and y are identical if they contain the same elements. Axiom 3. (Regularity) Each nonempty set x has an element y such that x∧y= Axiom 4. (Comprehension Scheme) For each given formula there exists a set y such that y = {x ∈ z : (x )}. and each set z,

Axiom 5. (Pairing) For any sets x and y, there exists a set z such that x ∈ z and y ∈ z. of sets there exists a set A which Axiom 6. (Union) For any given set consists of all elements belonging to at least one element set in Axiom 7. (Replacement Scheme) For each formula and any set A, if to each element x ∈ A there corresponds exactly one y satisfying (x,y), then there exists a set Y consisting of all y satisfying (x,y), for some x ∈ A. Axiom 8. (Infinity) There exists a set x satisfying (1) ∈ x, and (2) for each y ∈ x there exists a set S(y) ∈ x which contains all elements in y and y itself. Axiom 9. (Power Set) For any set x, there exists a set y, called the power set of the set x and denoted by p(x), such that all subsets of x are contained in y. (Choice) Any set can be well-ordered.

Axiom 10.

An interpretation of the language of set theory is defined by specifying a nonempty domain of disclosure, over which the variables are intended to vary, together with a binary relation defined on that domain, which is the interpretation of the membership relation ∈. If is a sentence in the language of set theory, is either true or false under a specified interpretation. In the intended interpretation, under which the axioms of ZFC are presumed true, the symbol x ∈ y is interpreted to mean that x is a member of y. Since we wish to talk only about sets and their elements in our domain of discourse, all

64

Chapter 3

elements of such a set should be sets also. Repeating this, we can see that the domain of our discourse consists of those x such that x is a set, and

We call such an x a hereditary set. If S is a set of sentences, we may show that S is consistent by producing an interpretation under which all sentences of S are true. Usually, ∈ will still be interpreted as membership, but the domain of discourse will be some subdomain of the hereditary sets. The justification for this method of producing consistency proofs is the easy direction of the Gödel completeness theorem, which says that if S holds in some interpretation, then S is consistent. The reason this theorem is true is that the rules then must be true under any of formal deduction are set up so that if S interpretation which makes all sentences in S true. If we fix an interpretation in which S holds, then any sentence that is false in that interpretation is not provable from S. Since and cannot both hold in a given interpretation, S cannot prove both and ; thus S is consistent. We now conclude this section with a few extremes of mathematical thought. For a more serious discussion, see (Fraenkel et al., 1973; Kleene, 1952; Kreisel and Krivine, 1967). Platonists believe that the set-theoretic universe has an existence outside the spirit of human beings. From this point of view, the axioms in ZFC are merely certain obvious facts in the universe, and we might have failed to list some other obvious but basic facts in the ZFC axiom system. Platonists are still interested in locating principles of the universe which are not provable from ZFC. A finitist believes only in finite objects. It is meaningless to consider any sort of concept like that of “whole.” If, say, “human knowledge” does not exist, the set of all rational numbers is not a practical concept; hence, the theory of infinite sets is a piece of meaningless garbage. There is some merit in the finitist’s position, since all objects in known physical reality are finite, so infinite sets may be discarded as figments of the mathematician’s imagination. Unfortunately, this point of view also discards a big portion of modern mathematics. The formalist can hedge his bets. The formal development of ZFC makes sense from a strictly finitistic point of view: the axioms of ZFC do not say anything, but are merely certain finite sequences of symbols. The assertion ZFC means that there is a certain kind of finite sequence of finite sequences of symbols, namely, a formal proof of Even though ZFC contains infinitely many axioms, notions like ZFC will make sense only when a particular sentence is recognized as an axiom of ZFC. A formalist can thus do mathematics just as a platonist, but if

Axiomatic Set Theory

65

challenged about the validity of handling infinite objects, he can reply that all he is really doing is juggling finite sequences of symbols. In this chapter, we will develop ZFC from a platonistic point of view. Thus, to establish that ZFC we simply produce an argument that is true based on the assumption that ZFC contains true principles.

3.2. Constructions of Ordinal and Cardinal Number Systems
The axiom on set existence says that our domain of discourse is not empty. Let z be any set. Axiom 3 (comprehensive scheme) implies that {x ∈ z : x ≠ x} (3. 4)

defines the existence of the empty set. By extensionality, such a set is unique. The Comprehension Axiom is intended to formalize the construction of sets of the form {x : P (x)}, where P(x) denotes a property of x. Since the notion of property is made rigorous via formulas, it is tempting to set forth as axioms statements of the form (3.5) where is a formula. Unfortunately, such a scheme is inconsistent by the famous Zermelo–Russell paradox (Section 2.6): Let be x ≠ x then this axiom gives a y such that (3.6) whence y ∈ y ↔ y ∉ y. To avoid this contradiction the Comprehension Axiom is given as follows: For each formula without y free, the universal closure of the following is an axiom:

We can also prove, not simply to avoid this paradox, that there is no universal set. Theorem 3.2.1. If ∀ x(x ∈ z ), by the Comprehension Axiom we can form which would yield a contradiction by the Zermelo– Russell paradox. Proof:

66

Chapter 3

We let A ⊂ B abbreviate ∀ x ( x ∈ A → x ∈ B ). So, A ⊂ A and ⊂ A. The example {x : x = x } shows that for a given formula (x ) , there need not exist a set {x : (x)}. Generally, this collection (or class) may be too big to form a set. The Comprehension Axiom says that if the collection is a subcollection of a given set, then it is a set. Axioms 4–8 of ZFC say that certain collections do form sets. By the Pairing Axiom, for given x and y we may let z be any set such that x ∈ z ∧ y ∈ z; then { v ∈ z : v = x ∨ v = y } is the unique (by extensionality) set whose elements are precisely x and y; we call this set {x,y}. Note that (x,y) = {{ x}, {x,y}} is the ordered pair of x and y. In the Union Axiom, we think of as a family of sets and postulate the existence of a set A such that each member Y of is a subset of A. This justifies our defining the union of the family, denoted by , by (3.7) When we let (3.8) If then and “should be” the set of all sets, which does not exist. Finally, we let For any sets A and B, we define the Cartesian product (3.9) To justify this definition, we must apply replacement twice. First for any y ∈ B, we have

so by replacement (and comprehension) we can define

Now

so by replacement we may define prod(A,B) = {prod(A , y) : y ∈ B } Finally, we define A × B = prod'(A,B ).

10) and (3. Note that our notation does not assume H ⊂ A × A. Lemma 3. H) (B. Since ƒ is an isomorphism it follows that ƒ(z) = z. so if (A. we say (A. ∀ B ⊂ A (B ≠ Ø → ∃ x ∈ B ∀ y ∈ B (xHy)). so (H – 1 ) = H.x. This mapping ƒ is called an isomorphism from (A.. or (A.11) We define H – 1 = {(x. Whenever H and K are relations and A and B are sets.H) is a well-ordering.H) for each x ∈ A –1 Proof: If ƒ : A → pred (A.H) such that H totally orders A.H) were an isomorphism. let pred(A. H is a relation such that (3. H) whenever B ⊂ A. H) and (B.2.14) and (3. H) is a total ordering so is (B.H) = { y ∈ A : yHx }.y) : (y. let z be the H -least element of {y ∈ A : ƒ(y) ≠ y }. Lemma 3. If (A.2. H) to (B.1. Let (3.e. Then for each y ∈ A with yHz.H) a well-ordering.Axiomatic Set Theory 67 A relation H is a set all of whose elements are ordered pairs.ƒ (y) = y. i. A function (or mapping) is a relation ƒ such that (3. We say H well-orders A. K) are isomorphic well-orderings. .13) (3. if (A. This contradicts the definition of z. A is a set. If x ∈ A.H) (pred(A. If (A.x.2. then the isomorphism between them is unique.12) A total ordering (or strict total ordering or linear ordering) is a pair ( A.15) where we write xHy for (x.x) ∈ H }.H) is a total ordering and every nonempty subset of A has an H-least element.H). then (A.K) if there exists a bijection ƒ : A → B such that ∀ x. that is. K).y) ∈ H .y ∈ A (xHy ↔ ƒ(x)Kƒ(y)).x.

K)).R) Proof: Let f = { (v. (B.w) : v ∈ A ∧ w ∈ B ∧ (pred( A.y. Theorem 3. then x is not ordinal since we have defined orderings to be strict. {Ø}}. y. then ∃x ∈ C ∀ y ∈ C(x ∈ y ∨ x = y). {Ø.H) and ( B. The set {{Ø}} is not transitive.2. where ∈x = {(y.K). Examples of ordinals are Ø. More formally. and z are ordinals. we write x (A. y∈ x . then y is an ordinal and y = pred(x. ∈ x ) is a well-ordering. ∈x ) (A.2. If x = {x}.3. H) and y ∈ x.K).K ). (4) If x. and that these segments cannot both be proper.K) be two well-orderings.2. {Ø} .w. .y.x. (pred (B.H) (pred (B. pred( x . (1) If x is an ordinal and y ∈ x. A set x is an ordinal if x is transitive and well-ordered by the relation ∈.2.H) (B.K)} Note that f is an isomorphism from some initial segment of A onto some initial segment of B. (2) If x and y are ordinals and x y. Examples of transitive sets are Ø. {Ø}}. x ∈ y . x ∈ y and y ∈ z implies that x ∈ z.v. {Ø. {Ø} . (3) If x and y are ordinals.z) ∈ x × x : y ∈ z }. (b) ∃ y ∈ B((A. (5) If C is an nonempty set of ordinals. Let (A. then x = y. Without causing confusion. A set x is transitive if every element of x is a subset of x. Then exactly one of the following holds: (a) (A.y).K)). let z be the H -least y ∈ A such that ƒ(y) ≠ g (y). ∈ x ).1 will occur.…. then exactly one of the following is true: x = y. Theorem 3. Then a similar contradiction as in the proof of Lemma 3. H) for (x. the assertion that x is well-ordered by ∈ means that ( x.H) (c) ∃ x ∈ A ((pred(A. y) for pred(x.H).68 Chapter 3 Proof: If f and g were different isomorphisms.

2. Now. (2) If X is a set of ordinals. ∀ α … instead of ∀ x (x is an ordinal → … ). Theorem 3.3. Thus.3(2). since x ∈ x would imply that ( x.2. the type (A.2.16) Let ƒ be the function with domain B such that for each a ∈ B. to prove existence. H) C is an ordinal (from Lemma 3.3 implies that the set of all ordinals. β ( α ≤ β → α ⊂ β ) . If (A.3). We may thus say. and let C = R(ƒ).3(5) implies that η ∩ { η } = Ø. Lemma 3.2.… to indicate ordinals.H) x)} (3.2. we have the following: If (A. For (5).4. H) is a well-ordering. let B = {a ∈ A : ∃ x(x is an ordinal ∧ ( pred(A. if it existed.2. Lemma 3.H) C. H). or B = pred( A. we note that the conclusion is. and. That no more than one holds follows from the fact that no ordinal can be a member of itself. X is the least ordinal in X. but the definition of η implies η ∈ η ∩ { η}. and that either B = A (in which case we are done). if X ≠ Ø ∩ ∪ . H) for some b ∈ A (in which case b ∈ B. then X is the least ordinal ≥ all elements in X.6 for details). γ . (2). then since x is well-ordered by ∈. denoted by η. and hence a contradiction). a.. e. H) C. Let x ∈ C be arbitrary.H) is a well-ordering. would be an ordinal. H) onto C.2 to show that at least one of the three conditions holds. Since ∈ orders the ordinals. there is an ∈ -least element x1 of x ∩ C. then Theorem 3. we use (1). Let A be a set of ordinals and ∀ x ∈ A ∀ y ∈ x(y ∈ A) then A is an ordinal. ∈x ) is not a strict total ordering (since x ∈ x x). and then x 1 ∩C = Ø Theorem 3. because if such an ordinal existed. If x ∩ C # Ø. f ( a) = the unique x.2. by (3).H) is defined to be the unique ordinal C such that (A.2. From now on we use Greek letters α. (1) ∀ α. equivalent to ∃ x ∈ C (x ∩ C = Ø).a. then there exists a unique ordinal C such that (A. and thus cannot exist.2.H). contradiction. we check that ordinal x such that (pred(A.g. Theorem 3.4. This fact takes care of the Burali-Forti paradox (see Section 2. and Theorem 3. Proof: The uniqueness of the ordinal C follows from Theorem 3. that ƒ is an isomorphism from ( B.4 implies that ordinals can be used as representatives of the order types of well-ordered sets. we write α < β for α ∈ β and use α ≥ β to indicate β ∈ α ∨ β = α.Axiomatic Set Theory 69 Proof: For(3). b. β.

Then ω is an ordinal and all smaller ordinals are successor ordinals or 0.. Therefore. Define α + β = type ( α × {0} ∪ × {1}. (3. An ordinal α is a successor ordinal (or isolated ordinal) if ∃ β( α = S (β)). Theorem 3. The following theorem shows that the elements of ω are the real natural numbers. The Axiom of Choice tells that every set can be counted by an ordinal. contradiction.6. and we have 0 = Ø 1 = S (Ø) 2 = S (2 ) 3 = S (2). where Theorem 3. and show that γ is a limit ordinal < ω. β Let α and β be two ordinals. So ω is a limit ordinal. γ: (1) (2) (3) . and hence ω is the least limit ordinal. H). For example. since if not it would be a natural number. natural numbers are those ordinals obtained by applying S to Ø a finite number of times. ) Let α be an arbitrary ordinal. (2) (3) (4) (Induction) Proof: For (4). we use natural numbers to indicate finite ordinals. let γ be the least element of ω \X. the Axiom of Infinity is equivalent to postulating the existence of a limit ordinal. For any α. β. It is a limit ordinal if α ≠ 0 and α is not a successor ordinal..17) Intuitively.2. Actually. Then S( α ) is also an ordinal.5. The successor S( α of the ordinal is defined by S ( α ) = α ∪{ α}. if X ≠ ω. The Peano postulates are (1) 0 ∈ ω . let ω be the set of all natural numbers and 0.70 Chapter 3 We use natural numbers to count finite sets. .2.

(1) (2) (3) (4) (5) If β is a limit ordinal.2. we write si for s(i). A² and that A × A are not the same. We define the multiplication α · β = type ( β × α. if s is a function with domain(s) = I. We define A n to be the set of all functions from n to A. H ). we call any collection of the form {x : φ (x)} a . In general. but there is an obvious 1–1 correspondence between them. Proof: The results are straightforward from the definition of addition. When D(s) is an ordinal α. In this case. w e may concatenate the sequences s and t to form a sequence of length α + β as for each follows: the function with domain α + β is defined by and for all We have seen that there need not exist a set of the form {x : φ ( x) }. and they sometimes provide useful motivation. we may think of I as an index set and s as a sequence indexed by I. and is the function s with domain n such Under these definitions. Informally. But there is nothing wrong with thinking about such collections. where H is the lexicographic order on β × α: (3. : γ: β. again. For any α.Axiomatic Set Theory 71 (4) (5) If β is a limit ordinal. need to remember that the multiplication of ordinals is not commutative and the distributive law (6) fails for multiplication on the right.7.18) Again. Theorem 3. (6) We. Recall that the addition of ordinals is not commutative. we get directly from the definition the basic properties of the multiplication of ordinals. Let A be a set. If D (t) = β. Let α and β be two ordinals. we may think of s as a sequence of length α.

8 [Transfinite Induction on ON].72 Chapter 3 class. we imply that A can be well-ordered. which we will call the cardinality of the set A. . To establish existence.2. As in Chapter 2. denoted by A. if g is a δ -approximation and g' is a δ ' Next. Regardless of the Axiom of Choice. where g is the δ -approximation for some δ > α . let G( α ) be the value g( α ). α is defined and is ≤ α for every ordinal α . we call g a δ -approximation if g is a function with domain g and (3.2. we can show that for each δ there is a δ -approximation (which then is unique). Proof: Fix α ∈ C. Then α β is defined by recursion on β by (1) (2) (3) If β is a limit.2. Generally.20) Theorem 3. Under the Axiom of Choice. if G1 and G 2 both satisfy Eq. (3. then C has a least element. if A can be well-ordered. Theorem 3.19) (3. then there exists a unique G : ON → V such that (3. Now. and there is then a least such α. let β be the least element of α ∩ C. A useful application of recursion is in defining ordinal exponentiation: Let α and β be two ordinals. In any statement involving  A. If F : V → V.21). then δ.21) Proof: For uniqueness. then A α for some α (Theorem 3. If C ⊂ ON and C ≠ Ø. If α is not the least element of C.4). We use V and ON to indicate the following two classes V = {x : x = x} ON = {x : x is an ordinal} (3.  A is defined for every set A. 1–1 functions are used to compare the size of sets.22) As in the proof of uniqueness. by transfinite induction on approximation.9 [Transfinite Recursion on ON]. we can show that by transfinite induction on α. Then β is the least element of C. A proper class is a class which is not a set (because it is too “big”).

A set A is finite if  A  < ω and to be countable if  A  ≤ ω.2. Proof: Statement (1) can be proved by induction on n and (2) follows by using Theorem 3. α × α = α ⊗ α < κ. Theorem 3. Theorem 3. δ) if and only if precedes Each has no more than predecessors under the ordering . so α ≤ β . Proof: We prove the theorem by transfinite induction on κ .2.1.2. Uncountable means not countable.11 (2). we have α = β.2. Proof: If κ is an infinite cardinal and κ = α + 1. Then for α < κ . Corollary 3. Thus. Cardinal multiplication and addition are defined in the following. They must be distinguished from ordinal multiplication and addition. β) ( γ .23) . then since 1 + α = α . Infinite means not finite. κ × κ = κ . by Bernstein’s equipollent theorem in the previous chapter. Then show n · m < ω by induction on m. so type lexicographically] whence (3.14. and α ~ α ⊂ β .2.10. (2) ∀α ( α ~ n → α = n) .13. ω is a cardinal and each n ∈ ω is a cardinal.6. Let κ and λ be two cardinals we define Theorem 3. Theorem 3.2. Every infinite cardinal is a limit ordinal. If n ∈ ω .11. Assume this holds for smaller cardinals. then (1) n n + 1.12. Define a well-ordering on κ × κ by ( α. κ = κ =  1 + α = α. then  β =  α . The rest follows from Theorem 3.This contradicts the definition of κ . If κ is an infinite cardinal. For Proof: First show n + m < ω by induction on m.2.2.Axiomatic Set Theory 73 Theorem 3. If α ≤ β ≤ α . Proof: β ⊂ α implies β ≤ α .

A cardinal number κ is a successor cardinal if there exists some ordinal α such that κ = α+ . the ordinal 2 ω is ω . Proof: Assume α ≥ ω . We now turn to cardinal exponentiation.17. but the cardinal 2ω = P( ω) > ω . ∀α∃κ( κ α and κ is a cardinal).2.2. (4) ωα is a successor cardinal if and only if α is a successor ordinal.2. Now.2. ∪ Let α be an ordinal and α + the least cardinal > α.15. . (2) ω α +1 = (ωα)+ .2.2. For example. Let S = {type(H) : H ∈ W }.74 Chapter 3 Corollary 3. Let AB = BA = {ƒ : ƒ is a function ∧ D(ƒ) = B ∧ R(ƒ) ⊂ A}. (3) ωα is a limit cardinal if and only if α is a limit ordinal. (2) α < β →ωα < ωβ . We define Theorem 3.16. then Proof: λ 2 ~ P( λ ) follows by identifying sets with their characteristic functions. use the proof of Theorem 3. Cardinal exponentiation is not the same as ordinal exponentiation. The cardinal κ is a limit cardinal if κ > ω and is not a successor cardinal. (3) For any limit ordinal Theorem 3. the sequence of transfinite cardinals ℵ α = ω α is defined by transfinite recursion on α by (1) ω0 = ω. Then (3. (1) Every infinite cardinal is equal to ωα for some α .24) where X Y implies that there exists a 1–1 function from X into Y. Let κ and λ be infinite cardinals then (1) (2) Proof: For (2). Theorem 3. Then S is a cardinal > α.13 to define a 1–1 mapping This gives a 1–1 mapping Therefore. If λ ≥ ω and 2 ≤ κ ≤ λ . Let W = {H : H ∈ P( α × α ) and H well-orders α} .

cf(cf( β )) = cf( β ).2.2.e. The Generalized Continuum Hypothesis is the conjecture that If α and β are ordinals and ƒ : α → β is a mapping. If κ .18. The cofinality of β . if β is a successor. and σ are any cardinals. Corollary 3.2. denoted by cf( β)..2.3. If α is a limit ordinal and ƒ : α → β is a strictly increasing cofinal mapping. then cf(α) = cf( β). Therefore. To show cf( α) ≤ cf( β ). and define a mapping ƒ recursively by Theorem 3. An ordinal β is regular if β is a limit ordinal and cf(β ) = β .20.2. then h : cf( β) → α is a cofinal mapping. Thus. Proof: cf(β ) ≤ c f (α) follows by composing a cofinal mapping from cf( α) into α with the mapping ƒ. cf(β) ≤ β . Theorem 3.3 implies that cf(β ) is regular for all limit ordinals β . cf( β ) = 1. The mapping ƒ maps α cofinally if D(ƒ) is unbounded in β . There is a cofinal mapping ƒ : cf(β ) → β which is strictly increasing Proof: Let g : cf( β) → β be any cofinal mapping. then and Proof: We can check that and The Continuum Hypothesis (CH) is the conjecture that 2ω = ω 1 . i. is the least α such that there is a mapping from α cofinally into β . .Axiomatic Set Theory 75 Theorem 3. and let h( ξ ) be the least η such that ƒ(η) > g ( ξ ). ∀ b ∈ β∃ a ∈ D(ƒ)( b ≤ a ). let g : cf( β) → β be a cofinal mapping. λ . Corollary 3.19.

Assume AC and GCH. Proof: Fix a cofinal mapping ƒ : λ → κ . the field of rational numbers. m ) is intended to represent n – m the equivalence relation ~ is defined by the following: (a. to construct the sets of of defining these sets from the natural numbers will do. based on ZFC.17. Theorem 3.26) Then h ∉ R(G).76 Chapter 3 Theorem 3. If κ is infinite and cf( κ ) ≤ λ .d) if and only if a + d = c + b . be the ring of integers.23 [König’s Theorem]. Proof: If ƒ mapped α cofinally into κ + for some α < κ + .2.21. then (3. . It suffices to show that there is a and Any reasonable way way. κ + is regular. κ λ > κ by Theorem 3.23. Theorem 3. For (3).2.2.14.2.2. For (2). If an ordinal β is regular.25) but the union of ≤ κ sets each of cardinality ≤ κ must have cardinality ≤ κ ⊗ κ = κ by Theorem 3. Then (1) (2) (3) Proof: (1) follows from Theorem 3.22. Let κ and λ ≥ 2 and at least one of them be infinite. and the field of complex numbers. where (n . then κ λ > κ .2. the field of real Let numbers.24. Theorem 3.2. but κ λ ≤ κ κ = 2 κ = κ + . but for definiteness we take = ω × ω/ ~. We show that G cannot be an onto mapping. Define h : λ → κ so that h( α) is the least element of (3. then β is a cardinal. λ < cf( κ ) implies that and each We conclude this section by showing that all of classical mathematics can be built upon ZFC. Let G : κ → λ κ .b) ~ (c.

Lemma 3.1. and operations + and · are defined as follows: (3. . Families of Sets For an infinite cardinal κ we say that two sets X and Y are k -disjoint if Theorem 3. To carry out this construction we need the following lemma. g in F (conceived as sets of ordered Let κ = ℵ α and pairs) are ℵ α -disjoint if We will construct a sequence of type ωα+1 of ℵ α disjoint functions. then there exists a function ƒ ∈ F which is ℵα -disjoint from each term ϕξ of the sequence.3. Two functions ƒ. Let where (x .3. Proof: It will be sufficient to find just one set A which can be decomposed as in the theorem and has power κ. because only if η ≥ ξ. y ) is intended to represent x / y. A 1–1 mapping of A onto A0 will then enable us to obtain the desired decomposition of A0 . then there exists S0 ⊂ p (A0 ) of power κ + consisting of m-disjoint sets and such that A0 = S0 . If κ ≥ ω 0 and  A0 = κ.3. If ϕ is a transfinite sequence of type γ < ωα+1 with range contained in F.28) where [( a.Axiomatic Set Theory 77 is the set of equivalence classes of the relation ~. Proof of lemma: To see this we notice that the range of ϕ has power ≤ ℵ α and thus can be represented as the range of a function Φ : ωα → F. .27) (3. (a) (b) 3. If ξ < Let ƒ (ξ) be the first ordinal in can hold Then ƒ ∈ F which is ℵ α -disjoint from every Φ η . with field operations defined in the usual way. b)] is the equivalence class of the relation ~ containing (a.1. We will need the following definitions. b ). and So is the set of left sides of all Dedekind cuts.

Since each element of A is an ordered pair of ordinals < ωα .1. Thus. which belong to F. If X ⊂ κ. { ƒ (A ): A ∈ } is an almost disjoint family of 2κ subsets of κ. Let A = R (g).78 Chapter 3 For each transfinite sequence ϕ denote by B(ϕ) the set of all functions in F which are ℵ α -disjoint from all terms of ϕ.29) for each ξ < ωα+ 1 . otherwise g ξ = ρ. Since If  X  = κ. An almost disjoint family ⊂ p( κ )such that and any two distinct elements of are almost disjoint. Theorem 3.3. Let κ ≥ ω 0 then (a) If ⊂ p ( κ ) is an almost disjoint family and   = κ . called the root of the ∆ -system. such that a b = r whenever a and b are disjoint members of is not Let Γ : p (F) → F ∪ {ρ}. disjoint family with no almost disjoint family Theorem 3. A maximal almost disjoint family is an almost properly containing it. provided that such functions exist.3.3. We claim that gξ ≠ ρ for each ρ < ωα+ 1. and is ℵα -disjoint from all gη . For an infinite cardinal κ two sets x. If we let ƒ be a 1–1 function from I onto κ . on the other hand. let Proof: Let I = { x ⊂ κ : x < κ } . A family of sets is called a ∆ -system. η < ξ. then then since if we fix β such that (3.30) Let then   = 2 κ and is an almost disjoint family of subsets of I. Otherwise there would be a smallest γ < ω α+ 1 ( g γ = ρ ) and g γ would be a sequence satisfying the assumptions of Lemma 3. then  Ax = κ also.3. infer that A ≤ ω α. If κ ≥ ω 0 and 2 <κ = κ. If X ≠ Y. This contradicts the assumption that ρ ∉ F. if there is a fixed set r. we Thus.1. where ρ ∉ F. y ⊂ κ are almost disjoint if x ∩ y  < κ. (b) There is a maximal almost disjoint family The proof follows from that of Theorem 3. Thus. then maximal. By transfinite induction. R ( g ) = ωα +1.2. then there exists an almost disjoint family ⊂ p ( κ ) with   = 2 κ . A = ωα . or a quasi-disjoint family. we obtain a function g : satisfying the equation (3.3. . such that + ⊂ p( κ ) of cardinality ≥ κ .

set-theoretic topology. If we now assume that CH fails. element of x. If x ∈ and ξ < ρ.31) Then α0 < θ and x (η ) < α 0 for all x ∈ and all η < ξ 0 . Let (3.32) Let = { x µ : µ < θ }.Axiomatic Set Theory 79 Theorem 3. Shoenfield. = θ and Under the assumption of CH or GCH. algebra. ( ∆ -System Lemma).2. (3.3. and analysis. we may assume Since what the elements of Then each x ∈ has some order type < κ as a subset of θ.3. let x ( ξ ) be the ξ th of α. For example. they are only of interest for ω < κ < 2 ω . For such κ. α <κ  < θ implies that fewer than θ elements of are subsets is unbounded in θ.3. . Assume   ≥ θ and ∀x ∈ (x < κ ) then ⊂ such that   = θ and forms a ∆ -system. we have Question 3. i. please consult (Martin and Solovay.e. 1970.1.2. called Martin’s axiom. does 2 κ = 2 ω? Question 3. Then Proof: By shrinking are as individuals is irrelevant. it is known by the method of forcing that neither question can be settled under the axioms ZFC + ¬CH. Now fix ξ 0 to be the least such ξ (ξ 0 may be 0). Since θ is regular. Rudin. there is some ρ < κ such that = {x ∈ : x has type ρ } has cardinality θ. Martin’s axiom is known to have numerous important consequences in combinatorics. pick x µ ∈ so that x µ (ξ 0 ) > α 0 and x µ ( ξ 0) is above all elements of earlier x γ . If κ < 2 ω . Let θ > κ b e regular and satisfy ∀α < θ (α <κ  < θ ). By transfinite recursion on µ < θ.4. Then   = θ and x ∩ y ⊂ α 0 whenever x and y are distinct elements in Since there is an r ⊂ α 0 and ⊂ with whence forms a ∆ -system with root r. questions arise about the various infinite cardinals κ < 2ω . 1977. We now fix such a ρ and deal with For each α < θ. If κ < 2ω . 1975). Since θ is regular and θ > κ. we calculated many exponentiations of cardinals in Section 3. Let κ be an infinite cardinal. In the following we show how these questions can be settled by using a new axiom. there is some ξ such that is unbounded in θ. For examples other than those given here. does every almost disjoint family A ⊂ p ( ω ) of size κ fail to be maximal? Since the answer to these questions is clearly “yes” when κ = ω.. we may assume   = θ. there is a if necessary. Thus.

D(ƒ G ) could be a very small subset of ω.” Elements of p are called conditions. denoted by p ⊥ q if ¬∃ r ∈ P (r ≤ p ∧ r ≤ q ). ≤ ) has the countable chain condition (ccc) if every antichain in P is countable. p and q ∈ P are compatible if (3. then q says more about ƒ than p does. in addition. (3.35) Define p ≤ q by q ⊂ p. For example. In fact. ≤ ) has ccc. then p ⊂ ƒ G .36) . An antichain in P is a subset A ⊂ P such that ∀ p. However. the elements of G are pairwise compatible. let (3. we are more interested in ƒ than G. A subset G ⊂ P is called a filter in P if Axiom 3. If (P. Now p and q are compatible if and only if they have the same values on the set D(p) ∩ D(q).1 [Martin’s Axiom]. ≤ ) such that P ≠ Ø and ≤ is a relation on P which is transitive and reflexive ( ∀ρ ∈ P ( ρ ≤ ρ)) . ≤ ) is a nonempty ccc partial order and is a family of at most κ many coinitial subsets of P.3.. If p ∈ P. let D ⊂ P be an antichain in P then  D  ≤ P = ω. MA( κ) is the statement: Whenever (P. for example. ≤ ) is a partial order. The partial order (P. If q ≤ p. then there is a filter G in P such that MA is the statement Example 3. we can make ƒG representative of typical functions. in which case p ∪ q is a common extension of both p and q. A partial order (P. Hence.80 Chapter 3 A partial order is a pair (P. ≤ ) is a partial order in the strict sense.34) They are compatible. So p says what f is restricted to D(p). i. the function p extends the function q. satisfies (3. If G is a filter in P. if it. by requiring that G intersect many coinitial sets. Ø is a filter and Ø = Ø the empty function. if we let ƒ = ƒG = G then ƒ G is a function with D(ƒ G ) ⊂ ω In this example.3. p ≤ q is read “p extends q.e. q ∈ A (p ≠ q → p ⊥ q ). A partial order (P.e. Let P be the set of all finite partial functions from ω to 2. we think of p as a finite approximation to ƒ.33) We often abuse notation by referring to “the partial order P” or “the partial order ≤” if ≤ or P is clear from context.1. and we say intuitively that p forces “p ⊂ ƒ” in the sense that if p ∈ G. i.

Let h : ω → 2 be a fixed mapping. which is impossible. then   = 2 ω . MA(2 ω ) is false. then ƒ G takes the value 1 somewhere.38) If G is a filter in P.3. (a) If κ < κ '.3. then ƒ G has a domain equal to ω. (b) MA( ω ) is true.1. since that would mean that R( G ) = ω 1 . G = {q ∈ P : ∃ n(q ≥ P n )}. Then G is a filter in P and G ∪ D n ≠ Ø for each n ∈ ω. Let E = {p ∈ P : ∃ n ∈ D (p)(p(n) = 1)}.2.3. Theorem 3.37) Then E h is coinitial in P.e. by induction on n. let G is a function with D( G ) ⊂ ω (3.3. Let G be the filter generated by {P n : n ∈ ω}. as in Example 3. Example 3. If ∀n ∈ ω ( G ∩ D n ≠ Ø). Let = {D n : n ∈ ω} ∪ {Eh : h ∈ ω 2 }. we might attempt to strengthen MA(k) by dropping this requirement.5. Theorem 3. . If α < ω 1 . P n ∈ P so that P 0 is an arbitrary element of P (since P ≠ Ø) and P n +1 i s any extension of P n such that P n + 1 ∈ D n . and if G ∩ E h ≠ Ø. So P 0 ≥ P 1 ≥ P 2 ≥ … .6(b) did not need the fact that P was ccc. Theorem 3. α < ω 1 . Proof: (a) is clear. then ƒ G ≠ h. and if G ∩ E ≠ Ø . P here is not ccc since the conditions {(0. then ƒ G is a function from ω to 2. Since the proof of Theorem 3. i. contradiction. This is possible since D n is dense. Let (3.3.6. Since each p ∈ P can be extended to a condition with n in its domain. Of course. then MA( κ') → MA (κ ). D n is coinitial in P. because we can view MA as saying that all infinite cardinals < 2ω have properties similar to ω. then.3. We claim that no filter G in P could intersect each D α . let = {D n : n ∈ ω } and define.. Define (3. This fact shows the next result. α ) : α ∈ ω1 } are pairwise incompatible.6(b) implies that MA follows from CH. For (b).39) Then D α is coinitial in P. and R ( ∪G) ⊂ ω 1 .Axiomatic Set Theory 81 for each n ∈ ω. If G is a filter in P and G ∩ D ≠ Ø ∀D ∈ . E is coinitial in P. but for k > ω this strengthening becomes inconsistent by the next example.

there is a filter G in P intersecting all sets in (3. In fact. F 2 ) ∈ P . since for each F ) is an extension of (s. Eq. F ) in E n.82 Chapter 3 We now proceed to show how MA is applied to answer Questions 3.41) Then the partial order has the ccc. Assume MA( k ). . For each x ∈ .F 1 ) ≤ ( s.y .3.3. Theorem 3. L e t sume that for all and all finite such that Proof: Define a partial order P by (3.44) If we pick E n. In fact.7.3. F ) are compatible.y is coinitial in P .1 and 3. they are compatible if and only if (3.45) Let Then In fact.F ) if and only if (3.2. which is impossible. then (s'.46) This implies that is infinite.42) Suppose that were pairwise incompatible. (3. (3.43) Then Dx is coinitial in P . for (s1 .40) where (s 1 . F 1 ) and (s2 . for each For any (3. with m > n. then By MA(k ). the sets s would all be distinct.3.42). If where and asThen there is d ⊂ ω Question 3. By Eq. F' ) and ( s. define (3.2 now can be answered by the following corollary.42) implies that If (3.

3. Let ⊂ If MA( κ ).3.4. Corollary 3. then 2 κ = 2 ω . We shall find a filter H in Q intersecting each form of MA( κ ) to a suitable suborder P ⊂ Q with We first show that we can find P ⊂ Q such that (1) (2) D P is coinitial in P for each of size κ ( exists by Theorem 3. Iƒ MA. a family of at most κ many coinitial subsets of order of arbitrary cardinality. Proof: Fix any almost disjoint family Define that Φ is onto.3. Let ⊂ P (ω) be an almost disjoint family of size κ. Theorem 3. Then is not maximal. cf(2 ω ) > κ.3.3.3).1.Axiomatic Set Theory 83 Corollary 3.3. Then since is almost disjoint it follows that Proof: Let ω for all finite F ⊂ Thus. Let ⊂ P (ω) be an almost disjoint family of power κ. where Assume MA( κ ). and let ( Q.8. Theorem 3. Proof: If κ < 2 ω .2. then there is a d ⊂ ω such that and Proof: Apply Theorem 3. then 2 ω is regular.8 says (3) If p. MA( κ ) → 2 κ = 2 ω . and by applying the restricted Q. . where ω ≤ κ < 2 ω . MA(κ ) is equivalent to MA( κ ) restricted to partial orders of cardinality κ.3. be a ccc partial Proof: Assume the restricted form of MA( κ ).3.3. p and q are compatible in P if and only if they are compatible in Q.7 implies that there is an infinite almost disjoint from each member of Theorem 3.7 with Corollary 3. q ∈ P.3. so by König’s theorem. so Corollary 3.

for all D ∈ and (2) by closure under fD . ht( x. Then .48) and P is closed under g and each fD . (a) Let a partial order be defined by ht( x.3. ≤ ) such that. i. Statement (3) implies that P has the ccc. denoted by Levα ( T )..47) and let g : Q × Q → Q be such that (3. If x ∈ T. is Then ht( T ) is the least α such that Lev α A subtree of T is a subset T ' ⊂ T with the induced order such that (3.e.3. for each x ∈ T.49) If H is the filter in Q generated by G.e. is type For each ordinal α. the αth level of T. let G be a filter in P such that (3. then for each x ∈ T'.50) then H is a filter in Q intersecting each A tree is a partial order in the strict sense (T.1. (3. Example 3. so by applying the restricted MA(k ) to P. T' ). we abuse notation and refer to T when we mean Let T be a tree. Let P ⊂ Q be such that P satisfies (3) by closure under g g (P × P) ⊂ P and f D (P ) ⊂ P. denoted by ht(x.84 Chapter 3 For let f D : Q → Q be such that (3. i. (a) ht (b) If T' is a subtree of T. is well-ordered by ≤ . T ) = ht(x. T ) = 0 for all x ∈ T and ht( T ) = 1.51) Proposition 3.3. The proof is immediate from the definitions involved. the height of x in T. As usual. T )..

we refer to as the complete binary tree of height δ. At ω1 the situation is different. Proof: since if would be a chain of cardinality κ. A tree is defined as follows: and s ≤ t if and only if s ⊂ t. For any regular κ.Axiomatic Set Theory 85 (b) Let I be any set and δ an ordinal. If T is an ω -tree. By a similar argument. Then for each infinite chain in T. Theorem 3. then T has an infinite chain. If α < δ. Since  T  = κ and ht(T ) = κ.10 says that there are no ω-Aronszajn trees. we may inductively pick is an and is infinite.52) . For any infinite cardinal κ. We first continue with a general discussion of κ -Suslin trees and related concepts. each κ -Suslin tree is a κ -Aronszajn tree. and every element of T is ≥ some element of so that Lev 0 (T ). but there is always an ω 1 -Aronszajn tree. There is an ω1 -Aronszajn tree. Since each Lev α ( T ) is an antichain. For any regular κ. Thus. When κ is singular. a κ -tree is a tree T of height κ such that κ ). ht(T ) ≤ κ. regularity of κ implies that κ. Theorem 3. We confine our attention to the case when κ is regular. is infinite. Suslin trees were introduced by Kurepa (1936c).3. a κ -Aronszajn tree is a κ-tree such that every chain in T is of cardinality < κ. We think of elements of αI as sequences of and ht elements of I of length α. every κ-Suslin tree is a κ -tree.9. This tree is called the complete I -ary tree of height δ. Thus.3.11. but Theorem 3.10 [König’s Theorem]. This is possible Proof: Pick such that since Lev0 (T) is finite. Theorem 3.3. For any regular κ.3. Proof: Let (3. The existence of an ω1 -Suslin tree is independent of ZFC. κ -Suslin trees exist but are of little interest. a κ -Suslin tree is a tree T such that  T  = κ and every chain and every antichain of T have cardinality < κ. then Lev When I = 2. T is infinite.

T is a subtree of The ht(T) = ω1 . By (i). then T has a well-pruned κ -subtree. each s α ∈ T* so Lev α for each α < ω1 . Proof: Let T 1 be the set of x ∈ T such that (3. Thus. thus. is countable. T is not Aronszajn since T is not an ω1 -tree. t ∈ α ω. To verify Eq. since for every α < ω1 . (3. We will find s α for each α < ω1 such that Assuming such s α may be found. T* is a subtree of T. let (3. A well-pruned κ -tree is a κ -tree T such that and (3. Lev α (T ) is uncountable for However. if we set then (i) would hold for so and Then (iii) holds as well. T ) < α < κ.55) Then T 1 is clearly a subtree of T. take any and let it is here (iii) is used. If C were an uncountable chain in T.54) for T 1 . Let and the fact that each has cardinality .86 Chaper 3 Thus. T* is an Unlike T. Unfortunately. then C would be a 1–1 function from ω1 into ω.54) Theorem 3.3. we can define a subtree of T which is Aronszajn. If s.53) By (ii). every chain in T is countable.12. Given s α . Fix and and inductively define so that and Let Then and t is 1—1. let us fix x ∈ T 1 and By definition of T 1 α such that ht( x. there is a 1–1 function from α into ω. We now see how to construct the sequence {sα } by induction. define s ~ t if and only if the set is finite. Now suppose so that that we have s α for α < γ . If κ is regular and T is a κ -tree. where γ is a limit ordinal. T* is an ωl -tree since ω 1 -Aronszajn tree.

Since   < κ .13.3. each element in T is either a union on a finite intersection of some sets of the forms in (3. then (3. and each element of this set is above some element of Y. called intervals: For a.59). b ∈ X satisfying a < b . yn ∈ Lev α ( T ) with each y i > x. The set X is called a space. on the set X is the topology generated by the following sets. the elements of X are called points of the space. The order topology T. . The family T is also called a topology on X.Axiomatic Set Theory 87 κ. If κ is regular. and x ∈ T. . . 1975) is a pair (X.59) That is. Theorem 3. there are z i ∈ Lev β (T ) with z i > y i . then ∪ ∈ T. z n +1 ∈ Lev β (T ) with zn. (iii) If ⊂ T . y2 . and the subsets of X belonging to T are called open in the space. .T)and distinct y1 .58) and (3. T is a well-pruned κ -Aronszajn tree. there Y and this y is in T ¹ . . A similar argument is y ∈ Y such that shows that Now for every is a well-pruned subtree of T. T ) consisting of a set X and a family T of subsets of X satisfying the following conditions: (i) ø ∈ T and X ∈ T . (ii) If U ∈ T and V ∈ T . Then {z1 . If the theorem holds for n. A topological space (Engelking. z n +1 > yn . z2 . let (3. .56) Proof: For n = 2. this follows from the fact that {y : y > x } meets all levels above x and cannot form a chain. For i < n. z n + 1} establishes the theorem for n+1. For n > 2. Let ( X. . . then U ∩ V ∈ T.57)–(3. we proceed by induction. . Now let β > α be such that there are distinct z n . or the topology induced by the ordering ≤ . fix α > ht( x.57) (3. ≤ ) be a total ordering.

<).D )).D ) < min{h(C). <) (see Theorems 3. D ∈ L. It was well known that any total ordering (X. so each h(C ) is a limit ordinal. h( C) < ω 1 . If C.” SH arose in an attempt to characterize the order type of the real numbers (R. Clearly. Since T is Aronszajn. < ) . (b) X is continuous. Suslin (1920) asked whether (c) may be replaced by (c') X is ccc in the order topology.7).12. and (c) X is separable in the order topology is isomorphic to (R. it is separable if it has a countable subset A such that each nonempty open set intersects A. let C ( α) be the element of C on level α.D ) be the least α such that C (α ) ≠ D ( α).D )) < D (d( C. Suslin’s hypothesis (SH) states that “there are no Suslin lines. Pick (3. C ≠ D.3. Theorem 3. (c) and (c’) are equivalent. Let L = {C ⊂ T : C is a maximal chain in T } (3. We now show that (L. Suppose that is a family and α ξ so that of disjoint nonempty open intervals. It is easily verified that it is indeed a total order of L.4. h( D )}.88 Chapter 3 A topological space has ccc. there is an ordinal h (C ) such that C contains exactly one element from Lev α (T ) for α < h ( C ) and no elements from Levα (T ) for α ≥ h ( C ). We order L as follows: Fix an arbitrary total order < of T. if each collection of pairwise disjoint open sets of the space is countable. There is an ω1 -Suslin tree if and only if there is a Suslin line. Thus. let T be an ω 1 -Suslin tree. SH is equivalent to the statement that (a). under SH. For α < h( C). Since T is well-pruned. and (c’) characterize the ordering (R.4.3.61) . let d (C. Observe that d ( C. (b). then there is one satisfying (a) and (b). First. Proof: First. We have thus defined an order on L. ) is a Suslin line. we show that L has the ccc. A Suslin line is a total ordering (X. By Theorem 3. a maximal chain cannot have a largest element. ≤ ) such that in the order topology X is ccc but not separable. ≤ ) satisfying (a) X has no first and no last elements. Let C D if and only if C(d (C. and one can show that if there is a Suslin line.14.5 and 3. we can assume that T is well-pruned.60) If C ∈ L.

63) Let . and and or (3) if α < β . where a b. If A ⊂ T is an antichain. then by (3)(b) (3. elements of are of the form (a. thus. contradicting that T is a Suslin tree. z. Let D. |T| = ω 1 . F) contains no C ∈ L with h(C) < δ. T is a Suslin tree. let I be a maximal disjoint subfamily of (3. which is the smallest subset Y whose complement belongs to . so |A| ≤ ω.3. z. Let be the set of all nonempty open intervals of L. F) is a nonempty interval.13. with ξ < η → I ξ ≤ I η. Then is partially ordered by reverse inclusion: I ≤ J if and only if I ⊃ J. then the elements of A are pairwise disjoint. T is a tree and each β = Levβ (T ). respectively. ). but since x ∈ D ∩ F. we define α +1 as follows: For I ∈ . (D. we let T = By (1)–(3). Say they are ordered D E F then (D. To find T. (1) the elements of (2) are pairwise disjoint. b). is any maximal disjoint We now construct the sets β by induction. y. w ∈ Lev α( T) above x. I ∈ (a) I∪ = (b) ∈ then either ⊂ I and I \cl ≠ where cl( ) stands for the closure of . Finally. we first find for each β < ω1 so that for each β. there exists an α > δ with three distinct elements.62) so would contradict the ccc of L. { C : h ( C) < δ } is not dense in L. is a dense subset in L. Fix x ∈ Lev s ( T ). Given α . since if {I ξ : ξ < ω 1 } were such a chain. w. By Theorem 3. Thus. suppose that we are given a Suslin line ( L . On the other hand. We will define a subset T ⊂ s o that ≤ is a Suslin tree ordering on T. since (2) implies in particular that each β ≠ . subfamily of maximality implies is dense. We may assume that L is densely ordered by putting gaps in the set and no nonempty open subset of L is separable. E. it is sufficient to see that for each δ < ω 1.Axiomatic Set Theory 89 then {E ξ ( αξ ) : ξ < ω 1} forms an antichain in T. To show that L is not separable. F be elements of L containing y. T can have no uncountable chains. Assuming that this can be done.

Second question: Is there an “object” x such that x = { x}? This question is independent of the existence of physical reality.3. 3. By Μ Α (ω1 ).4. contradicting the fact that T is a Suslin tree. this will follow by maximality of ιφ we can show that for each J ∈ ∃K ∈ (K ⊂ J ). T) > α } is dense in P. and such a set x would clearly be an hereditary set. hereditary sets (Section 3. thenK1 does not contain the endpoints of I. so fix K 1 ∈ with K1 ⊂ J If I ∈ ∪α < γ α.90 Chapter 3 Finally. Let for each α < γ (3. we may assume that T is well-pruned. the reverse order of T. so and K1 ∩ E = then I ∩ K 1 = or K1 ⊂ I.3. It seems likely that we have not left any interesting mathematics behind by so restricting our universe. Well-Founded Sets Some questions about sets are relevant to applications of mathematics. and the set of all complex numbers. By Theorem 3. we assume γ is a limit ordinal and have defined the satisfying (1)–(3) for α < β < γ. Furthermore. since mathematical objects like the set of all real numbers. MA (ω 1 ) implies that there are no ω1-Suslin trees. Such an object x did not occur in the construction of mathematical This is also an important question in systems theory (see objects like and Chapters 8 and 12). Let E be the set of all left and right endpoints of all intervals in E is countable and J is not separable. We give two of the most important questions. ≥ ). the Axiom of Extensionality has embodied in it the assertion that all things in our domain of discourse are sets. P has ccc. are hereditary sets and have been defined explicitly within this domain in Section 3. Proof: Let (T. A detailed discussion on the relevance between the questions and applications of mathematics will be given in Chapter 8. . Since T has no uncountable antichains.1). and let P = (T .15. ≤ ) be an ω 1 -Suslin tree. (2) says that no J ∈ of .2. First question: Is there anything which is not a set? We have declared that our axioms of set theory deal with sets — in fact.12.64) be a maximal disjoint subfamily of Then (1) and (3) now hold for and let is disjoint from all members all α < β < γ. there is a filter G intersecting each D α . Now take K ∈ with K ⊂ K1 and K 1 \cl(K ) ≠ K ⊂ and K ∈ Theorem 3. in which case D α = {x ∈ T : ht(x. Then G is an uncountable chain. For β = γ.

1. Case 1. Let us define R ( α) for each α ∈ ON by transfinite recursion as follows: (a) R(0) = (b) (c) when α is a limit ordinal. and x ∉ R( α) for all α > β . (b) Proof: We use transfinite induction on α: We assume that the result holds for all β < α and prove it for α . WF = ∪ {R( α) : α ∈ O N}. This establishes (a) and (b) for all ordinals α. we study the position of the Axiom of Regularity relative to the axioms in axiomatic set theory. Hence. Let WF be the class of all well-founded sets. The rank of x. and x ∉ R( β). Let x ∈ WF. For each ordinal α: (a) R( α) is transitive. R (α) = {x ∈ WF : rank (x ) < α } Proof: For x ∈ WF .e. Theorem 3.. A set A is well-founded if there is an α ∈ ON such that A ∈ R ( α). We work in the axiom system of all axioms in ZFC except Regularity and define the class WF of well-founded sets by starting with and iterating the power set operation. p (R( β)) = R ( α) is transitive and R( β) ⊂ R (α). Statement (b) is immediate from the definition.Axiomatic Set Theory 91 In this section.4. (a) follows from the fact that the union of transitive sets is transitive. Case 2. Since R( β) is transitive. the result is trivial. For each ordinal α. if β = rank(x). the least ordinal α with x ∈ R ( α) must be a successor ordinal.4. Then we show that WF is closed under the other set-theoretic operations as well.2. rank(x) < α if and only if ∃β < α (x ∈ R ( β + 1)) if and only if x ∈ R( α). α = 0. denoted by rank(x). i. α is a limit. . α = β +1. Theorem 3. is the least β such that x ∈ R ( β + 1). Case 3. then x ⊂ R (b).

6. Similarly. then x × y. p(x ). Theorem 3. so rank( x) < α. As in (a). (a) If x ∈ WF. {x } ∈ R( α + 2) and ∪ x ∈ R( α + 1). so y x ⊂ R ( α + 3).4. ∀α ∈ ON (R( α ) ∩ ON = α ). x ∩ y. Thus the theorem holds for all ordinals α . then y ∈ R ( α + 1) = p (R( α )). by transfinite recursion. let α = ∪ {rank(x ) + 1 : x ∈ y }. For (b). Theorem 3. let α = max{rank(x). so α ∈ R ( α + 1 ) . so α ⊂ R( α).4. let α = rank( y). Then. the class WF excludes circularities like x ∈ y ∧ y ∈ x .( x .4 and 3.3. β ∈ R ( β + 1) ⊂ R ( α). each x ∈ y has rank < α . Theorem 3. . rank( α) = ∪ { β + 1 : β < α } = α .4. y}. then ∪ x. (b) If x . y } ∈ R ( α +2). rank(y )}. y) ∈ R ( α +3). and yx are all in WF. so y ⊂ R( α). More formally. For (b).92 Chapter 3 Theorem 3. { x. for β < α .4. and the rank of each of these sets is less than max{rank(x).g. By Theorem 3. there is no x ∈ WF such that x ∈ x . y ∈ WF. y ∈ R( α +l).4. Then x ⊂ R ( α ). we show. (b) rank( y ) = ∪ {rank(x ) + 1 : x ∈ y}.. from well-founded sets of smaller rank. Therefore. rank(y )} + ω.4. Furthermore. Proof: For (a). e. Thus. the class WF excludes sets which are built up from themselves. x ∪ y. It is clear that the following result holds. By (a) α ≤ rank( y). Likewise. If x ∈ y . and the rank of these sets is less than rank( x) + ω. so yx ∈ R ( α + 4). so p (x ) ⊂ p (R( α )) = R ( α + l).4. since this would yield rank(x) < rank(y ) < rank(x). so rank(y ) ≤ α . Proof: For (a). Any ordered pair of elements of x ∪ y is in R ( α + 3).3 says that the class WF is transitive and that we may think of the elements y ∈ WF as being constructed. then (a) ∀x ∈ y(x ∈ WF ∧ rank( x) < rank(y)).4. x ∈ R ( α ).3(b). Assume that the theorem holds for all β < α . We prove this by transfinite induction on α. Proof: What we need to show is that ∀α ∈ ON( α ∈ WF ∧ rank( α) = α ).2. Theorem 3. Each ordinal is in WF and its rank is itself.5. since from this we would have rank(x ) < rank(x). We leave the rest of the details for the reader. The proof follows from Theorems 3.4. so p (x) ∈ R(α + 2). and {x} ∈ WF . {x . If y ∈ WF . let α = rank( x ).

Proof: (a) is easy by induction on n. ) is a group . let a α be defined by transfinite recursion on α as follows: (1) a 0 = ω . Then the ordered sum of the sets R(n ).7. (a) ∀n ∈ ω (|R(n )| < ω ). × ) is any group. ω |R( ω + 1)| = 2 ω . Theorem 3. The sets and are all in R( ω + ω ). (2) Every topological space is homeomorphic to a topological space in WF.65) If (G. Generally.9. Theorem 3.4.Axiomatic Set Theory 93 Theorem 3. for each n we may identify R(n + 1) with R (n) 2 and order it lexicographically. Theorem 3. The proof follows from Theorem 3. let α = |G| and let ƒ be a 1–1 mapping from α onto G.6 and the definitions of these sets in Section 3.8. The following theorem is an example of showing that all mathematics takes place in the class WF. The cardinalities of the sets R( α) increase exponentially: |R( ω )| = ω . |R( ω + 2)| = a α.4. has order type ω. it is sufficient to see that R( ω ) is countable. etc. Proof: A group is an ordered pair (G.4. Theorem 3. To show this by induction. (b) | R ( ω) | = ω. Define an operation on α by ξ η = ƒ – 1 (ƒ( ξ ) × ƒ( η )). since ω ⊂ R ( ω ). × ) where × : G × G → G. For (b). The proof follows from transfinite induction on α. (2) a α +1 = 2a α .10. n ∈ ω .4.2. (3) For a limit γ . Then we have the following theorem.4. (1) Every group is isomorphic to a group in WF. |R( ω + 2)| = 2 2 . Then ( α. a γ = ∪{ a α : α < γ }.6 therefore implies that (3.4.

then x ∈ R ( α ). and pick a y ∈ X with rank( y) = α . If A ⊄ W F. Then y is ∈ -minimal in X by Theorem 3. let α = ∪ {rank( y) + 1 : y ∈ x }. then A ∈ WF.4. Thus. then z ∉ X.66) is called H -minimal in X.” which is the least transitive set containing A as a subset. Proof: Let X be a nonempty subset of A. Generalizing the notion of well-orderings. so y ∈ WF . (a) A ⊂ tr cl(A). (3. we have the following: A relation H is well-founded on a set A if and only if (3. Theorem 3. Theorem 3. Proof: It is sufficient to show A ⊂ W F .4. This contradicts the definition that y ∈ A\ WF. (b) tr cl( A) is transitive. ∈ is well-founded on A. (2) ∪n + 1 A= ∪ ( ∪ A ). n (3) tr cl( A) = ∪ {∪ n A : n ∈ ω}. (d) If A is transitive.66) The element y in Eq. (c) If A ⊂ T and T is transitive. × ). Here tr cl is the “transitive closure. Theorem 3. we define (1) ∪ 0 A = A.12. because if A ⊂ W F. so z ∈ WF .3(a).11. By recursion on n. let X = A \WF ≠ φ and let y be the ∈ -minimal in X.13. . If A is transitive and ∈ is well-founded on A.4. The same idea can be used to prove (b). then tr cl(A ) = A. so x ∈ R( α + 1) ⊂ WF .94 Chapter 3 isomorphic to ( G. Thus. We leave the proof of (b) to the reader. Let A be a set. then tr cl(A) ⊂ T.4. y ⊂ WF. If A ∈ WF. but z ∈ A since A is transitive. a relation H is well-founded on A if and only if every nonempty subset of A has an H-minimal element. Let α = min{rank( y) : Y ∈ X }. If z ∈ y.

Let V be the class defined by (3. The proof is straightforward from the definition of transitive closure.4. apply Theorem 3. (c) ∈ is well-founded on tr cl(A). (f) tr cl(A ) = A {tr cl(x ) : x ∈ A}.14. ∪ n A ∈ WF since WF is closed under the union operation (Theorem 3. Regularity has no application in ordinary mathematics. Proof: (a) ↔ (b) is immediate from the definition of well-founded relations.11.4. so A ∈ WF.12. Unlike the other axioms of ZFC.4.4.67) In other words. Theorem 3.15. The following are equivalent: (a) The axiom of regularity. since accepting it is equivalent to restricting our attention to WF. For (c) → (b). so A ⊂ tr cl( A) ⊂ WF. so tr cl(A ) ∈ WF. (b) tr cl(A ) ∈ WF. tr cl( A) ∈ WF.11. For (b) → (c). Thus. then by induction on n. but only that we restrict our domain of discourse to be just WF. each ∪ n A ⊂ WF . we discuss some of the consequences of adding the axiom V = WF to the axiom system of ZFC axioms except Regularity. By (c) and Theorem 3. V is the class of all sets. Theorem 3. If A ∈ WF . For any set A the following are equivalent: (a) A ∈ WF. This does not mean that the two classes V and WF are really identical. . ∈ is well-founded on tr cl(A). In the following. Proof: (a) → (b). (c) → (a). (c) V = W F. The proof of (b) → (c) is Theorem 3. then tr cl(x ) ⊂ tr cl(A).4. (b) implies that for any A. so A ∈ WF. (b) ∀ A ( ∈ is well-founded on A). If we are convinced by the previous theorems that all mathematics takes place in WF. so tr cl(A) ⊂ WF.6). it is reasonable to adopt as an axiom the statement that V = WF.Axiomatic Set Theory 95 (e) If x ∈ A.4.

1975. Since Regularity is equivalent to V = WF = ∪ { R( α) : α ∈ ON}. 1973.5. 1975. A is an ordinal if and only if A is transitive and totally ordered by ∈. 1967.16. 3. Kuratowski and Mostowski. we will see how the Axiom of Regularity can be used to study level structures of general systems. Jech. Then Regularity implies there exists an ∈-minimal element in X. Proof: To see that ∈ well-orders A. 1977. 1974. 1970.96 Chapter 3 where all mathematics takes place. In Chapters 5 and Chapter 9.4. Theorem 3. 1936c. let X be any nonempty subset of A. it gives us a picture of all sets being created by an iterative process. Shoenfield. 1970. 1976. References for Further Study There are many good books in the area of axiomatic set theory. Quigley. Kleene. Kurepa. Kunen. Suslin. Rudin. Kreisel and Krivine. Fraenkel et al. starting from nothing. 1978. Martin and Solovay. 1952. such as (Drake. 1980. 1920).. Engelking. .

Introduction Hall and Fagen (1956) introduced the concept of centralized systems. including the argument on why there must be a few people in each community who dominate others. public issues of contention. Lin (1988a) applied the concept of centralized systems to the study of some phenomena in sociology. The sets M and R are called the object set and the relation set of S. Applications of the concept in sociology. R). The leading part can be thought of as the center of the system. In the rest of this section. Then r is defined as follows: There exists an ordinal number n = n ( r ). and importance level of problems versus media. a function of r. Two real-life examples are given to illustrate the results obtained in this chapter. are listed. such that r ⊆ M n . and R is a set of some relations defined on M. A system is an ordered pair of sets.CHAPTER 4 Centralizability and Tests of Applications The concept of centralized systems is written in the language of set theory in order to take advantage of the rigorous mathematical reasoning. and some applications. all necessary concepts of systems theory and terminology of set theory will be listed in order for the reader to follow the discussions. since a small change in it would affect the entire system. In this chapter we look at the overall study of the concept of centralized systems. where a centralized system is a system in which one object or a subsystem plays a dominant role in the system operation. concerning the existence of factions in human society. Several interesting results were obtained. let r ∈ R be a relation of S. Here. related concepts. At the end some open questions are posed. 4. The concept of centralizable systems is introduced.1. S = (M . 97 . causing considerable changes. respectively. called the length of the relation r. such that M is the set of all objects of S.

injective.Ri ). where n is a fixed whole number. the mapping h : S 1 → S2 is also surjective. or bijective. (2) if is an (n – 1)th-level object system. For each relation is a relation on M2 with length n(r).R) is trivial (resp. By transfinite induction. In symbols. h will be used to indicate the class mapping and h is a mapping from the system S 1 into the system S 2 .S 1 is a partial system of S 2 if either (1) M 1 = M 2 and R 1 ⊆ R 2 or (2) M 1 ⊆ M 2 and ≠ there exists a subset R' ⊆ R 2 such that R1 = R 'M1 ≡ {ƒ : ƒ is a relation on M 1 and there exists g ∈ R' such that ƒ is the restriction of g on M 1 }. denoted by h : S 1 → S 2 . For set X. Without confusion. if h (R1 ) ⊆ R2 . i = 1. called the first-level object system. Let S i = ( M i . i = 1.R1 ) in M is a system. Note that a theorem followed by (ZFC) means that the theorem is true only under the assumption that the ZFC axioms are true. 2. The systems S i are similar if there exists a bijection h : S 1 → S 2 such that The mapping h is called a similarity mapping from S 1 onto S 2 . respectively. and a class mapping can be defined with the properties and for each x = where Ord is the class of all ordinals. injective. be two systems and h : M 1 → M2 a mapping. if h is a similarity mapping from S 1 onto S 2 .2. where ZFC stands for the Zermelo–Fränkel system with the axiom of choice..R) has n levels. Ri ). The system S 1 is a subsystem of S 2 if M 1 ⊆ M 2 and for each relation r1 ∈ R 1 there exists a relation r2 ∈ R 2 such that r 1 ⊆ r 2M1 . the inverse h – 1 is a similarity mapping from S 2 onto S 1 . A system S = (M. R' M1 = {ƒ : g ∈ R'(ƒ = g M 1 }. Evidently. where .98 Chapter 4 where is the Cartesian product of n copies of M. discrete) if M = R = Ø (resp. A mapping h : S 1 → S 2 is termed a homomorphism from S 1 into S 2 . X  denotes the . i = 1. if (1) each object S 1 = (M1 . or bijective. M ≠ Ø and either R = Ø or R = {Ø })..2. Given two systems S i = (Mi . called the nth-level object system of S. two classes . When h : M 1 → M 2 is surjective. A system S = (M. then each object is a system.

Centralizability and Tests of Applications

99

cardinality of X. A cardinal number α = ℵ λ is called regular if for any sequence of ordinal numbers { αβ < ℵ λ : β < λ} , the limit of the sequence is less than ℵλ . For details and other terminology of set theory, see (Kuratowski and Mostowski, 1976).

4.2. Centralized Systems and Centralizability
To use the mathematical reasoning to study centralized systems, Lin (submitted) redefined the concept in the language of set theory as follows. Definition 4.2.1. A system S = (M , R) is called a centralized system if each object in S is a system and there exists a nontrivial system C = ( M C , RC ) such that for any distinct elements x and y ∈ M , say x = (Mx ,R x ) and y = (My ,R y ), then the system C is called a center of S. M C = M x ∩ M y and Theorem 4.2.1 [ZFC (Lin and Ma, 1993)]. Let κ be an arbitrary infinite cardinality and θ > κ a regular cardinality such that for any ordinal number Assume that S = (M,R) is a system satisfying  M ≥ θ and each object m ∈ M is a system with m = ( M m ,R m ) and Mm < κ . If there exists an object contained in at least θ objects in M, there then exists a partial system S' = (M',R') of S such that S' forms a centralized system and  M'  ≥ θ. This result is a restatement of the well-known ∆ -lemma in axiomatic set theory (Kunen, 1980). The following question, posed by Dr. Robert Beaudoin, is still open. Question 4.2.1. A system S = (M,R) is strongly centralized if each object in S is a system and there is a nondiscrete system C = (M C , RC) such that for any distinct and elements x and y ∈ M , say x = (Mx ,R x ) and y = Give conditions under which a given system has a partial system which is strongly centralized. A system S 0 is n-level homomorphic to a system A, where n is a fixed natural number, if there exists a mapping h S0 : S 0 → A , called an n -level homomorphism, satisfying the following: (1) The systems S 0 and A have no nonsystem kth-level objects, for each k < n. (2) For each object S 1 in S 0 , there exists a homomorphism h S1 from the object system S 1 into the object system h S 0 (S 1 ). (3) For each i < n and each ith-level object S i of S 0 , there exist level object systems S k , for k = 0, 1, ... , i – 1, and homomorphisms h Sk , k = 1, 2, ... , i, such that S k is an object of the object system S k –1 and h S k is a homomorphism from S k into h S k– 1 (Sk ) , for k = 1, 2, ... , i.

100

Chapter 4

Definition 4.2.2 [Lin et al. (1998)]. A system S is centralizable if it is l-level homomorphic to a centralized system SC under a homomorphism h : S → S C such that for each object m in S the object systems m and h(m) are similar. Each center of S C is also called a center of S. Theorem 4.2.2 [Lin et al. (1998)]. A system S = (M,R) with two levels is centralizable if and only if there exists a nontrivial system C = ( MC ,R C ) such that C is embeddable in each object of S — that is, C is similar to a partial system of S1. Similar to the existence theorem of centralized systems (i.e., Theorem 4.2.1), we have the following result. Theorem 4.2.3 [Lin et al. (1998)]. Let κ and θ be cardinalities satisfying the conditions in Theorem 4.2.1. Assume that S is a system with an object set of cardinality ≥ θ and each object in S is a nontrivial system with object set of cardinality < κ . There then exists a partial system S' of S such that the object set of S' is of cardinality ≥ θ and S' forms a centralizable system. For detailed proofs of these theorems, please refer to Chapter 9.

4.3. Several Tests of Applications
In this section we show how the concept of centralizability has been applied in several research areas.

4.3.1. A Societal Phenomenon
Sociology concerns the environment in which we live and the problems encountered in human society. In (Lin, 1988a) the following social phenomenon was studied in the light of centralizability: the dominance of one person or a small group of persons over their family or neighbors. For example, in a relationship between two individuals, one of them must dominate the other. This is true in relationships between husband and wife, brothers, sisters, or even friends. To make any progress, let us first write a special case of Theorem 4.2.1. Theorem 4.3.1 [ZFC]. Suppose that S = (M,R) is a system such that  M ≥ c, where c is the cardinality of the set of all real numbers and that each object in S is a system with finite object set. If there exists such an element that belongs to at least c objects in M, there then exists a partial system B of S with an object set of cardinality ≥ c, and B forms a centralized system. Suppose that A is a collection of people and [A] < ω is the collection of all finite subcollections of people in A. Then for any finite subcollection x ∈ [A] < ω , three possibilities exist:

Centralizability and Tests of Applications

101

(1) There exists exactly one relation among the people in x. (2) There exists more than one relation among the people in x. (3) There does not exist any relation among the people in x. If situation 1 occurs, we construct a system S x = ( x , Rx ) as follows: in the relation set R x there exists only one element which describes the relation among the people in the finite set x. If situation 2 occurs, we construct a collection of systems {Sxi : i ∈ I x }, where I x is an index set that may depend on the finite i subcollection x such that for any i ∈ I x the relation set of the system Sx has cardinality 1 and for any fixed relation ƒ among the people in x there exists exactly i one i ∈ I x such that S x = (x,{ƒ}). Finally, if situation 3 occurs, we construct a system S x = (x, Ø ), where the relation set is empty. We now consider the system (M, Ø) with empty relation set, where M = {Sx : x ∈ [A] < ω , there exists at most one relation among the people in [A] < ω , there exists more than one relation among the people in x }. Theorem 4.3.1 says that if (1) the cardinality M is greater than or equal to c and (2) there exists at least one person who is an object in at least c elements in M, then there exists a subcollection M* ⊆ M such that (M*, Ø) forms a centralized system and M * ≥ c. Based upon the construction of the system (M, Ø), it follows that condition (1) means that there exists a complicated network of relations among the people in A. Condition (2) implies that there exists at least one person in A who has enough relations with the other people in A. However, when we study problems with people, how can we enumerate the relations that exist among these people? The difficulty comes from the notion of relation. To apply Theorem 4.3.1 correctly, we must rewrite each relation between people in the language of set theory. That is, each relation must be written in such a way that it consists of the following basic blocks only: “x is a set,” “x ∈ y,” “x = y ,” “and,” “or,” “if … then,” “if and only if.” For details see (Kuratowski and Mostowski, 1976). For the time being, the question is still open. However, no matter how complicated the relations among the people in a real community can be, Theorem 4.3.1 says that a person or a subcollection of people x ∈ [A] <ω who want to be a center in the collection A must possess the following background: (3) the people in x have many transverse and longitudinal relations with the others in A. Thus, if we suppose that in any social society there are uncountably many relations in the sense of set theory among the people, then from the point of view of centralizability, we can understand why there always exist factions in human society, namely because there is more than one center.

4.3.2. Public Issues of Contention
If history is seen as a moving picture, each moment of the moving picture shows a few public issues attracting major public attention. To prove this point, Vierthaler

102

Chapter 4

(1993) used the Reader’s Guide to investigate shifts and changes in American public concern for social problems. The research uses data available in the series of 52 volumes (1990–92) to consider historic variations in the series of volumes. Then Dr. Vierthaler looked at the topics identified as special “social problems” in the 92-year series to analyze the degree and types of “connectiveness” that exist among the specific social problems as a way to study changes in “American social problems consciousness.” Besides the data supporting the observation that at any moment in history some public issues seem to attract more attention than others, in (Lin and Vierthaler, 1998) a systems theoretical evidence was given to show that this observation is a historical truth. In this subsection we briefly describe how the systems theoretical evidence was developed. In the discussion of the previous subsection, we replace possibilities in 1–3 by the following: (1 ′) There exists exactly one social issue connnecting the people in x. (2 ′) There exists more than one social issue connnecting the people in x. (3 ′) There does not exist any social issue connnecting the people in x. where a social issue connecting the people in x means that the people in x are either affected by or attracted to a social issue. Now we consider the system (M, ) accordingly, and Theorem 4.3.1 will be applied. Note that when we study the collection of all social issues connecting the people at a fixed historical moment, there is no straightforward method to enumerate all social issues, both minor and major, because there are so many that we do not know how to enumerate them. As a result, social scientists use several methods to obtain indicators. Most recognized are social survey polls that repeat the question from 1935–1971, asking respondents: “What is the most important problem facing our country?” Other researchers use documents, such as the Readers’ Guide, a record of popular magazine articles about “social problems” from 1900–1991, or further original sources (Krippendorff, 1980) such as front page newspaper stories and the study of other texts over time to identify public issues of importance. So, a consequence of Theorem 4.3.1 is that because there always exist opposite central social issues, which divide the public, there will always exist opposite political parties in human society. In addition, different “positions” are possible on the same issue. hence, the basis to unify or divide depends on peoples’ interest and their ideas or positions toward how to best resolve a social issue. Thus, a public issue is a matter of importance to the public precisely because it addresses a matter of concern that is controversial — whether over the goals or the means considered plausible to attain the goal.

Centralizability and Tests of Applications

103

4.3.3. Public Issues: Importance Level of Problems versus Media
One school’s opinion says that there are “real” social issues at any given time in history which derive from major problems in the social structure that affects people, thus attracting public attention, and that the media just reports to the public truthfully about all aspects of these social issues. A different school’s opinion says that there is no such thing as major social issues at each given moment in history which attract public attention, because the mass media manufactures issues by exaggerated propaganda or by accentuating extreme ideological positions. For details, see (Lauer, 1976; Spector and Kitsuse, 1977). In (Lin and Vierthaler, 1998), the concept of systems was used to show that these two opposite theories are actually the two opposite sides of reality. They are unified in the name of systems. Let us see how to achieve this end. If reality is considered as truth, we then need to understand the meaning of truth first. In general, “truth” is a fuzzy concept. For example, the laws of nature are the truths about the world. A mathematical truth is a true statement in mathematics. Then what should be the general meaning of truth? Klir (1985) introduced a philosophical definition of systems. He says that a system is what is distinguished as a system. Based upon this understanding, it can be seen that for any system S = (M,R) and each relation r ∈ R , r is a true relation in S connecting some objects in the set M. So, r can be an S -truth. From Russell’s paradox (Kuratowski and Mostowski, 1976), we have the following theorem. Theorem 4.3.2 [Lin and Ma (1993)]. There is no system whose object set consists of all systems. This result means that we cannot consider a relation which is true with respect to all systems as its objects, and that for each given statement, if it is a truth, it then implies that there exists a system in which the given statement describes a relation. One interesting consequence of this understanding explains why there always exist different and opposite interpretations about each social issue. The reason is because for a given social issue, there always exist systems S 1 and S 2 such that the social issue is looked at differently in S1 than in S 2 . That is, the concept of “reality” is relative. The concept of truth was first studied in the name of systems in (Lin, 1990b). Let (M,R) be a system such that the object set M is defined as in Section 4.3.2 and R is a set of relations between the object systems in M. It is then reasonable to believe that this system describes all aspects or all “realities” about the collection of all social issues existing at a fixed moment of time. Then the fact that (M,R) is a system implies that M is a pile of isolated social issues and the relation set R describes how the historical moment (M,R) can be studied as an entity of continuity, connection, and movement. As pointed out in (Lin, 1990b), Klir’s definition of systems means the following.

104

Chapter 4

Axiom 4.3.1. That a system A possesses a property B means that there exists a system C, different from system A, such that C recognizes that A possesses property B. This axiom implies that consciousness and matter coexist. That is, the importance level of social issues catches the media’s interest; in turn, the media makes the issues more accessible to the public and so more public attention is created. Another interesting consequence of this axiom is that the media could do whatever possible to promote the importance level of a social issue. However, if the issue does not have many transverse and longitudinal connections with other social issues, no matter how hard the media tries, there will be a “plateau” over which the importance level or the level of public attention will never go. This might be the reason why during the 1992 U.S. presidential campaign, the percentage of supporters for then-president George Bush rose dramatically during the final days and then began to stall. Consequently, public support for Bush never reached a high level nor went beyond the observed level, which was lower than desired by that side of campaign. In the next two sections, two examples are used to elucidate the results obtained in this chapter.

4.4. Growth of the Polish School of Mathematics
A good book about the developing history of the Polish School of Mathematics is Kuratowski (1980), which is the source of the story used here. Briefly, in 1911 there were four distinguished professors of mathematics in the only two Polish universities: J. Puzyna, W. Sierpinski, S. Zaremba, and K. Zorawski. They had no common interests in the field of mathematics, since each worked in a different field: Puzyna in analytic functions, Sierpinski in number theory and set theory, Zaremba in differential equations, and Zorawski in differential geometry. As a consequence of the discussion in the preceding sections, they could not have students in common because there did not exist enough relations between them; this means that there was no possibility of organizing a mathematical school. For that to happen, there had to be a common set of problems on which a number of people were engaged so that the people would be working as a whole community. At the same time, no matter how small the group of people was at the beginning, it had to contain sufficiently bright and active people to attract the interest of other mathematical adepts by its topics; that is, the people in the group should be active enough to set up relations with other mathematicians outside. A few years later, two exceptionally talented and creative young mathematicians, Z. Janiszewski and S. Mazurkiewicz, were working in almost the same field, topology. Janiszewski received his doctor’s degree in Paris in 1912, Mazurkiewicz in Lwow in the same year.

Centralizability and Tests of Applications

105

In early 1918, some people spoke about a fairly strong Warsaw center of set theory, topology, and their applications under the direction of Professors Janiszewski, Mazurkiewicz, and Sierpinski. Since the arrival of Janiszewski and Mazurkiewicz, a community of the three mathematicians with common interests had been formed. The development of the hard-working community of mathematicians linked by common scientific interests was one of the essential factors leading to the creation of the Polish School of Mathematics. In the next two years the school was created and developed exactly as described by Janiszewski in the keynote address of the school. In 1918, in the first volume of a publication titles Polish Science, Its Needs, Organization and Development, Janiszewski published the article, “On the needs of mathematics in Poland.” He began with the assumption that Polish mathematicians could not afford “to be just the recipients or customers of foreign centers” but “to win an individual position for Polish mathematics.” One of the principal means suggested by him for attaining that end was the concentration of scientific staff in a relatively narrow field of mathematics but one in which Polish mathematicians had common interests and — what was more important — one in which they had achievements that counted on a world scale. Though mathematicians need no laboratories or expensive and sophisticated auxiliary equipment for their work, they do need a proper atmosphere; this proper atmosphere can be created only by the cultivation of common topics. For research workers, collaborators are almost indispensable, for in isolation they will in most cases be lost. The causes are not only psychological, such as lack of incentive, but an isolated researcher knows much less than those who work as a team. Only the results of research, the finished ripe ideas can reach an isolated researcher and then only when they appear in print, often several years after their conception. The isolated researcher does not know how and when they have been obtained, is far from those forges or melting pots where mathematics is produced, comes late, and must inevitably lag behind. Because of the concentration of scientific staff, Janiszewski suggested the establishment of a periodical devoted exclusively to the fields of mathematics connected with set theory and foundations of mathematics. Such a periodical, if published in a language known abroad, would serve a double goal: It would present the achievements of Polish mathematicians to the world of learning, while attracting foreign authors with similar interests. Such a journal, entitled Fundamenta Mathematicae, was founded, and the first volume appeared in 1920. That day was thought to be the inauguration of the Polish School of Mathematics. Although Fundamenta was conceived as an international journal, the first volume deliberately contained papers only by Polish authors. It was something of an introduction to the world of the newly risen school of mathematics. Fundamenta was a journal limited to just one field of mathematics. Publication in languages known abroad made the accomplishments of Polish mathematicians accessible to the scientific world on a large scale, and it was an indispensable

The editors had hit upon the very happy idea of presenting in a new edition the actual state of problems discussed in the first edition of that volume (i. A jubilee volume of twice the normal size was published. 17 years earlier). According to Tamarkin (Tamarkin. the work is by no means complete. where there exist strictly specialized mathematical journals. Fundamenta celebrated its twenty-fifth volume. The Appearance of Nicolas Bourbaki We can read about Nicolas Bourbaki. From this brief description. under the masterful guidance of its editors. 1980. which have been maintained since the first volume. 4. 1957).e. Thus. 1936). Bourbaki has also written articles for mathematical journals. In 1935. A considerable role has also been played by the problems sections. we can see how the idea of centralized systems was used extremely successfully in the establishment of a scientific center. Nicolas Bourbaki is the author of a wide-ranging textbook on mathematics written in French. both in individual papers and in the problem section. the papers in the jubilee volume showed the world that Fundamenta was a prestigious journal of the fields represented in it and showed the position of the Polish School of Mathematics. In this section we describe some stories useful to systems analysis. which appeared in 1937. 21 volumes totaling more than 3000 pages have been published to date. and the most distinguished scholars working in the fields represented in the journal were invited to publish their works in it.106 Chapter 4 condition for attracting the works of foreign mathematicians for publication in the journal and thus giving it an international character. Although the first volume appeared in 1939. The part played by Fundamenta in the development of fields represented in it is not limited to the publication of papers. Fundamenta Mathematicae immediately developed into a unique periodical that attracted international recognition and cooperation and whose history became the history of development of the modem theory of functions and point sets. others resulted in the appearance of many valuable papers. among them the Archiv der Mathematik. in many articles. Proof of this may be found by inspecting the new edition of the first volume. In contrast to the prevalent practice today.. In addition. in those days it was a novelty and met with undisguised skepticism on the part of many mathematicians. Halmos. a French School of Mathematics.5. Some problems became classics. The purpose of recalling the original problems was used to show the world of learning that Fundamenta was a journal that published important problems. he runs a seminar (the Séminaire Bourbaki) in Paris at which noted professors and researchers from all over France and from neighboring countries come together three times a year for a three-day . such as (Cartan.

R.O. the especially ingenious terminology.S. in an article for the Encyclopedia Britannica. Bourbaki introduced many new and important concepts. for they received a scathing letter signed by Nicolas Bourbaki in which he declared that he was not about to allow anyone to question his right to exist.. p. In about 1946.A. Bourbaki divided the class of all mathematical problems into several categories (Dieudonne. printed his opinion that Bourbaki was simply the pseudonym for a group of French mathematicians. the determination of Fermat primes or the irrationality of Euler’s constant) (II) Problems without issue (III) Problems that beget a method (IV) Problems that belong to an active and fertile general theory (V) Theories in decline for the time being (VI) Theories in a state of dilution The majority of Bourbaki’s topics belong to categories IV and III. an . For the problems in category III. his courage and resourcefulness in using the axiomatic method to develop the whole of mathematics. The publishers of Britannica soon found themselves in an acutely embarrassing position. and the clean and economical organization of ideas and style of presentation have fascinated many mathematicians. And to avenge himself on Boas. Bourbaki began to circulate the rumor that the mathematician Boas did not exist and that the initials B. Boas. Since the first notes. 2): (I) Stillborn problems (e. The study of problems in category IV ultimately (and perhaps only after a long time) reveals the existence of unexpected underlying structures that not only illuminate the original questions but also provide powerful general methods for elucidating a host of problems in other areas of mathematics. reviews.g. thenexecutive editor of Mathematical Reviews. In his work. Because there do not exist general rules in mathematics by which to judge what kinds of problems are important and what others are not of interest.Centralizability and Tests of Applications 107 mathematical conference held in the Institut Henri Poincaré. They had been asking: Who is Nicolas Bourbaki? How was the scientific research group formed? How are they working? Why did they choose this name? In short. 1982. P. mathematicians are extremely interested in the criteria on which Bourbaki decided what to do. and papers with Nicolas Bourbaki’s name appeared in some journals in the middle 1930s. Nicolas Bourbaki appeared in the world as if he came from a dim mist. Nearly 150 papers covering a broad range of mathematical topics have been delivered in this seminar and then to mathematical circles around the world. were simply a pseudonym for a group of editors of Mathematical Reviews.

each application is based on rigorous logical reasoning. it has been open in sociology for years. the study for the feasibility of this assumption will no doubt lead to an active research area in applied mathematics. since in applications of mathematics no one in the known literature has ever tried to address whether the setting of an application really makes the axioms in ZFC true. we can observe some similarities and differences. The study of the feasibility of the assumption will have impact on the entire research of applied mathematics. depend on the assumption that all axioms in ZFC are true. that the establishment of a network of relations is essential to achieving the desired superiority successfully. There may be some concepts among the fundamentals in Bourbaki’s textbook that have already become outdated. So far. Furthermore. how can scientists in the coming generations choose appropriate methods that will enable them to tower above the forest of scientific schools? This is an actual problem. 4. the study in this chapter is different in the following ways: First. Some Related Problems and a Few Final Words All applications of the concept of centralizability. But no matter how many thousands of methods exist to achieve superiority in science. As is well known. then the truthfulness of the mathematical theorems applied would be in doubt. no one seems to have developed a convincing method to enumerate human relations. he may feel obligated to start all over again. So. In the present situation. where there are uncountably many schools of thought in the world of learning. one point will never change—that is. any cap appearing in the logical reasoning .6. Now people are watching Bourbaki’s behavior with some further questions: What is the working direction of Bourbaki with newcomers? Will he still restrict himself to the mathematical foundation? What are his plans now? In comparing the appearance of Nicolas Bourbaki and the growth of the Polish School of Mathematics. After Bourbaki finishes the first part of his work. As for the question. the problems mentioned appeared. Second. discussed in this chapter.108 Chapter 4 examination of the techniques used to solve the original problems enables one to apply them (perhaps by making them considerably more complicated) to similar or more difficult problems without necessarily feeling that one really understands why they work. Mathematics has changed radically in the past 20 years (a development in which Bourbaki may have had a hand). the question of how to enumerate human relations in terms of the basic blocks of set theory is still unknown. The assumption and the question need to be addressed in order to make this piece of research more reliable. Compared with applications of systems theory in sociology. Because of this feature. according to the study in this chapter. if some of the axioms were not true in an application setting. done earlier by many scholars.

.Centralizability and Tests of Applications 109 is pointed out so that the applicability of the current approach of systems theory can be checked. at the same time. more and deeper insights in related areas can be obtained. the concept of feedback can play an important role here. if for some reason the approach needs to be modified. That is. Consequently. systems theory will accordingly be enriched.

and some important open questions are posed. named set theory. The conservation principle of ancient atomists is recast and proved mathematically. However. Lavoisier’s Axiom. as in certain cases the intuitive concept has been proved to be unreliable.CHAPTER 5 A Theoretical Foundation for the Laws of Conservation Finite divisibility of general systems is shown. According to Cantor. elements. A brief history of atoms. the word “set” means a collection of objects (or elements). and laws of conservation is given. the creator of set theory. 1. in axiomatic set theory the theory is based on a system of axioms (usually on the system of ZFC axioms). one of the greatest achievements of all time. The goal of this chapter is to establish a rigorous theoretical frame-work for the part of the world of learning on the laws of conservation. a current fashionable concept in modern science and technology. So. has found its way through general systems theory into the realm of scientific activities related to the search for order of the natural world in which we live. such a view is untenable. from which all theorems are obtained by deduction. where a fundamental particle is one which can no longer be divided into smaller ones. “Set” is one of the two primitive notions of the theory and is not examined directly about its meaning. Research in the foundation of mathematics. Hopefully. Based upon this result. we briefly list some basic terminology and concepts to fully understand the discussion in Section 5. introduced in 1789. this study will provide an early casting of a stone to attract many beautiful and useful gems in future research along the line described here. In the rest of this section. and systems. This chapter is based on (Lin. 1996). 5. is verified by using the concept of fundamental particles. Introduction The law of conservation of matter–energy.2. are joined under the name of general systems theory. based on which the law of conservation of matter–energy can be seen clearly. 111 .

2.2. Looking for a single basic reality. The concept of systems. That is.) 5. where M is the set of objects and R is a set of some relations on M. y ) is x. changing. refer to (Kunen. nebulous substance. Let x and y be two sets. denoted by (x . what would be the characteristics of nature’s “building blocks”? The ancient Greeks were intrigued by the diverse.112 Chapter 5 The ZFC axiom system is the Zermelo–Fränkel system with the axiom of choice. For convenience. used in the following discussion. It generalizes those concepts introduced by Whitehead (1978). Kuratowski and Mostowski. and Laws of Conservation The countless ideas in thousands of volumes in a library are arrangements of 26 letters of the alphabet. and Bunge (1979). Hall and Fagen (1956). 1980. In the search for order. and they looked for something unifying and eternal in the variety and flux of things. Brief History of Atoms. {x. Because of this and similar observations. 10 numerals. The set {{x}. The definition reads as follows: S is a (general) system. such that r is a subset of the Cartesian product of n copies of the set M. or supernatural intervention. Tarski (1954–1955). For details. Elements. if S equals an ordered pair ( M. are based on the assumption that the system of ZFC axioms is consistent. 1976). For details. each Ionian believed that all things have their origin in a single knowable element: water. (For those readers who are sophisticated enough in mathematics. But they were confident that the universe is one. temporary character of objects around them.C. denoted by Theorem (ZFC). the one-element Ionians (624–500 B. Mesarovic and Takahara (1975). and a few punctuation marks. The first term of (x. then there exists a set X in A such that All mathematical results. the great thinkers throughout the history have been wondering whether the rich profusion from objects around us are formed of a basic alphabet of nature itself in various combinations? If so. 1976). fire. the theory derived from ZFC axioms does not contain contradictory statements. one of the ZFC axioms: If A is a nonempty set of sets. and the second term is y.) had a naturalistic and materialistic bent. respectively. They sought causes and explanations in terms of the eternal working of things themselves rather than in any divine. was first introduced in (Lin. introduced in Section 5. we state the Axiom of Regularity. a relation r in R implies that there exists an ordinal number n = n(r ). see (Kuratowski and Mostowski. a function of r. y ). or some indeterminate. . The sets M and R are called the object set and the relation set of S. 1987). R ) of sets. mythological.y}} is called an ordered pair. air.

earth is male when it has the form of boulders and crags. they can be used to explain the large variety of objects they compose. female when it is cultivable land. It is the union and separation of atoms that is temporary and that results in the transformation of objects.C. That brings us to the very interesting question of the transformation of matter. tastes. every object is infinitely divisible. but by their various sizes and shapes. the previous four and wood.– ?). Anaxagoras (510–428 B. corrupted. Individual atoms.).C. With his philosophy. solid. the many products of the terrestrial four elements ceaselessly transform. The Egyptians had long recognized four elements with qualities of male or female. No matter how far an object is divided. it is male. To Leucippus and Democritus.C. The Leucippus–Democritus atom (500–55 B. claimed that there is not one basic element but four: earth. Philolaus (480 B. and transformed. always maintain their identity in uniting or separating.” Thus. Pythagoras and his followers (from 5th century B. a fundamental unit was lacking.) combined features of the Ionian single element. This element cannot be “generated. One main weakness of the concept is passing down the complexities of a large-scale object to unseen miniatures of itself. although all substances of the four elements are individually “generated. Among his physical ideas of matter. how can one substance develop into another? The contribution of Anaxagoras’ concept of seeds was its refinement. and indestructible. its idea of taking substances as they are and breaking them down minutely in order to know more about them. had five elements. He pointed out that. as early as the 12th century B.) in his “seeds” idea was unwilling to submerge the tremendous varieties in things into any common denominator. colors. For example. eternal. and fire. “Atom” means “not divisible” in Greek. . and seasons. air. the idea of Plato and Aristotle on the nature of matter might seem almost an anticlimax. Plato (427–347 B. on). The result is the principle of conservation of matter: Matter is neither created nor destroyed but is transformed. Eternally.) and Aristotle (384–322 B. or transformed. This term was intentionally chosen by Democritus to emphasize a particle so small that it could no longer be divided.C.. the universe as a whole is ungenerated and indestructible.C. Aristotle introduced a fifth element. corrupted. together with five virtues. If all substances are derived from unique seeds of themselves. Beneath the lunar sphere. ethereal substance makes up the heavens and all their unchangeable objects. and Empedocles’ four elements and yet was an improvement over all of them. water. He preferred to accept the immediate diversity of things as is. including Empedocles (490–435 B. what is left would have characteristics of the original substance. tones. Since these miniatures were infinitely divisible. after the atomists.C. the universe originally and basically consisted entirely of atoms and a “void” in which atoms moved. in him a conservation of matter concept can be found. when cloudy or sluggish. Anaxagoras’ seeds.A Theoretical Foundation for the Laws of Conservation 113 The four-element philosophers. However.).C. eternal. female. When air is windy.” This pure. the atom is indestructible and unchanging.C. The Chinese. Atoms are all of the same substance.

water. air. the Greek atomic theory. experimental techniques and knowledge had not yet reached the point where they could prevent the atomists from being eclipsed by Aristotle — and experimental techniques and knowledge are social accumulations. Why? From where does this heat of friction come? The calorists theorized that heat fluid is squeezed out from the interior of the blocks. Mayer (1814–1878) speculated that the calorists were wrong while observing venous blood color in different climates. Count Rumford (1753–1814) denied the calorists’ theory. a German physician. That is why we need to develop the fundamental theory and understanding about the existence and feasibility of the modern concept of elements and particles. or that they are more appropriate as states of matter that apply to every substance. becoming respectable in the Scientific Revolution. and fire can be resolved into simpler substances and that the Aristotelian elements are not elements after all. Sennert. Rubbing two blocks of wood together. had applied atomic theory to specific natural processes. developed an extensive nonmathematical atomic theory. the principle of conservation of energy may be regarded as one of the great achievements of the human mind. D. the total count of atoms in the universe was constant. techniques. The total amount of matter. and unifying principles. R. Lavoisier abstracted from his laboratory observations a law of conservation of matter: All matter existing within any closed system constitutes the universe for a specific experiment. Sir Humphrey Davy (1778–1829) corroborated Rumford by rubbing two pieces of ice together. a founder of the British Royal Society. but nature has a way of seeping through categories set for it. P. as it turned out. remained the same regardless of changes. P. it is merely forced to the surface by friction. Meanwhile. The atomists had a conservation principle of matter. Gassendi. influential French mathematician and philosopher. in 1929. His contemporary Newton used a corpuscular theory to explain his investigations of light. That is.” Let us now take a brief look at the development of the conservation law of energy. All these men claimed evidence that heat . where a “closed” system is one in which outside matter does not enter and inside matter does not leave. Boyle. And J. was among the first serious experimenters of the Scientific Revolution to use atoms to explain specific experiments.114 Chapter 5 Without the introduction of experimentation or the refinements of measurement. “Because of its practical use. In 1789. as he supervised the boring of cannon. However. Joule (1818–1889) argued against it as his paddles vigorously stirred water in calorimeters. and for its own intrinsic interest. caloric is neither created nor destroyed. Ship’s Surgeon J. therefore. did not have a chance against the authority of Plato and particularly that of Aristotle. eventually showed that earth. Heat of friction was the Achilles’ heel of the caloric theory. Men search for order through a few unifying principles. Bacon and Galileo had accepted the Greek idea of atoms. To ancient atomists. men gain detailed knowledge of their surroundings and develop more sophisticated concepts. makes the surfaces warmer. According to William Dampier. Experiments.

Watt’s steam engines. A Mathematical Foundation for the Laws of Conservation From the discussion in the previous section. and Joule established a quantitative connection between mechanical energy and heat through persistent. 5. please refer to Perlman’s work (Perlman. R. Conversely. a law of the conservation of matter–energy: The total amount of both matter and energy is always the same. Chemistry was based on the law of conservation of matter. by Joule’s time. Huygens proposed a conservation of kinetic energy for elastic collisions. transformed heat into mechanical energy. Black took seven-league steps with his conservation of heat principle. Actually.A Theoretical Foundation for the Laws of Conservation 115 of friction is heat produced by motion. Work concepts and swinging pendulums led to a principle of conservation of mechanical energy and to efforts toward perpetual-motion machines. and imagination to put the pieces together and present the whole picture. The models or understanding of the laws of conservation used in this section may not be those currently accepted by research physicists or chemists. What remained was for someone with enough insight. So. that heat is a form of motion (energy).3. Some limited aspects of the broad principle appeared as follows: Newton’s law of action and reaction led to conservation of momentum. Europeans of the Middle Ages used the powder to propel musket shot and cannonballs. not a form of matter. and energy into matter. von Helmholtz (1821–1894). conservation of energy as a broad principle was in the air. interest. a gun is an engine that converts the heat and expansive force of an explosion into mechanical motion. Physics was based on the law of conservation of energy. For a more detailed and comprehensive history. This was what the German physician J. The ancient Chinese invented gunpowder and used it in rockets. mathematical rigor. Experiments by Rumford and Davy pointed to heat as a form of energy. invented in 1763. It has been shown however that matter transforms into energy. and pointed applications of his countryman H. but they are surely some of the major steps . 1970). Mayer (1814–1878) did. it can be seen that all these laws of conservation were established based on experiments or intuitive interpretations of some related experimental results or simple observations. later reinforced by the technical language. precise techniques. a famous biophysicist. Now hen what? Where is the constancy of matter or energy? What happens to the indispensable “=” signs in the equations of physicist and chemist? What happens to physics and chemistry as “exact” sciences? Einstein combined the two conservation laws into one. In this section some mathematical results of general systems theory are introduced and used to establish a theoretical foundation for the laws of conservation.

the possibility (i) cannot be true.1 [ZFC]. That is. (5. 2. a system is what is distinguished as a system. . i= 1. Ri ). Each element in Mn is called an nth-level object of S0 .116 Chapter 5 along the path leading to the current ones. a system. for some i = 1. Y = S i = ( Mi . The theorem will be proven by contradiction. and (iii) Y = {Mi }. such that M stands for the set of all substances used in the reaction and R is the set . . for some ordinal number α. 3 . (5. If (i) holds. of different-level object systems of S . 1980). A system S n = (M n . . . . .1). Then each chain of object systems of S must be finite. This contradicts Eq. 1987) of a system S 0 = (M0 .1). {Mi . (ii) Y = S i . . for some i = 1. it follows that there exists a set Y in X such that (5. By doing so purposely. It is not intended for general audiences but for a few readers who have the adequate mathematical background. describing the matter of interest. again a contradiction of Eq. This contradicts Eq. 2. satisfying that the system S j is an nth-level object system of S i . Ri ) : i = 1. let S = ( M. such that for each pair i. . . Mi ∈ .R) be a systems representation of a chemical reaction process. satisfying that Si is an object of the system S i –1 . . . . .1). . j < α with i < j. If (ii) holds. . there exists an integer n = n (i. a proof of this result is given. (5. We then can do the following general systems modeling: For each chosen matter.3. 2. n – 1. . (ii) cannot be correct. Hence. Now define a set X as follows: From the Axiom of Regularity (Kunen. 2. A chain of object systems of a system S is a sequence { Si : i < α}. R n ) is an n th-level object system (Lin and Ma.Ri }}. such as {Si = ( Mi . . 3. such that the system S i is an object in M i . . Then we can prove the following result. I hope that a greater audience can be reached.R 0 ) if there exist systems S i = (M i . Therefore. Proof: For the sake of completeness of this chapter. can always be defined. As introduced by Klir (1985). for some i = 1. or environment. If (iii) holds. . real situation problem. such as a chemical reaction process. Theorem 5. }.R i ) = {{Mi }. (iii) cannot be true. . for i = 1. 2. . Suppose that there exists a chain of object systems of infinite length.1) There are now three possibilities: (1) Y = M i . 2.1 . 0 < i < n + 1. So. . . For example. which depends on the ordinal numbers i and j. These contradictions show that the assumption of a chain of object systems of infinite length cannot be correct. Suppose that S is a system. j ) .

In brief. Axiom 5. if the physical universe is studied as a system. a fundamental particle is a particle which can no longer be divided into smaller ones. indestructible. A change in which a substance loses its original composition and one or more new substances are formed is a chemical change. None of these changes is a change at the fundamental particle level.1. which are unstable. proves the feasibility of the Law of Conservation of Fundamental Particles.3. 1970. which states that the total count of atoms in the universe is constant.1 [Lavoisier’s Axiom (Perlman.e. each particle must be finitely divisible. and other changes. nothing is created! An equal quantity of matter exists both before and after the experiment . this conservation law says that all matter existing within any closed system constitutes the universe for a specific experiment.A Theoretical Foundation for the Laws of Conservation 117 of all relations between the substances in M. Rather each is a recombination of some higher level systems of the fundamental particles. since if a particle A can still be divided into smaller particles. p. based upon Klir’s definition of systems. are called nuclear reactions. A change where nuclei. . Based upon this understanding. The existence of fundamental particles is shown by Theorem 5.3. can be restated as follows: Law [Conservation of Fundamental Particles]. in turn. and nothing takes place beyond changes and modifications in the combination of the elements. That is. chemical. while take place in the nucleus of an atom. it follows that we have proven theoretically Lavoisier’s Axiom. 1970. and nuclear. the total amount of matter remains the same regardless of changes and modifications. The total count of basic particles in the universe is constant. the following theorem guarantees that for each given system S = (M.1 says that each chain of object systems must stop. if S = (M. At the same time.. A brief reasoning for the feasibility of this law is the following: There are three basic known changes in the universe: physical. It. the whole art of performing chemical experiments depends.R) is the systems representation of a chemical reaction defined as before. 414)]. therefore. Theorem 5. given in 1789.1. It can be seen that S represents the chemical reaction of interest. Upon this principle. . A then can be studied as a system. then a chemical reaction between the substances in M is just simply a change and modification of some relations between the fundamental particles. A change in which a substance entirely retains its original composition is a physical change. i. the conservation principle of ancient atomists (Perlman. the set of basic particles is eternal. Hence. In all the operations of art and nature. the total amount of matter remains the same regardless of changes. 414). Here. Changes of S represents changes of the objects in M and changes of relations in R. .3. p. the set of fundamental particles is uniquely determined. therefore. and unchangeable.3. shoot out “chunks” of mass and energy.R). Based upon this understanding and by virtue of Theorem 5.

That is. For each system S = (M. for all j > n + 1. the Law of Conservation of Matter–Energy becomes clear and obvious. Since energy can be in different forms. Each matter consists of fundamental particles.3. and energy into matter. The law states that the total amount of matter and energy is always the same. since each set Mi is uniquely defined. the set of all fundamental objects are unique. Let M 1 = {m ∈ M : m is not a system} and By mathematical induction.118 Chapter 5 Theorem 5. . and so does energy.R). matter can be transformed into energy. the proof given here is for completeness of this chapter. it confirms that each fundamental particle has no volume or size.1 guarantees that there is a natural number n such that Mj and Mj* equal φ. As a consequence of the Law of Conservation of Basic Particles. Theorem 5. Actually. Thus.1 also proves this fact. That is. Let us discuss some impacts of what has been done on the Law of Conservation of Matter–Energy.2 [ZFC]. the set consists of all fundamental objects of the system S and is unique. because if a particle has certain a volume. Proof: Again. the particle can be considered a system.3. it would be possible to chop it up into smaller pieces. Historically. this law was developed based upon the new development of science and technology: Energy and matter are two different forms of the same thing.3. assume that all sets Mi and Mi* have been defined for each i < n + 1 for a natural number n satisfying M i = {m ∈ M*– 1 : m is not a system} i and * Then Mn + 1 and Mn + 1 can be defined as follows: M n + 1 = {m ∈ M* : m is not a system} n and Theorem 5. it is not a fundamental particle. such as light.

contradicts the common notion that all things must change. Not only this. This idea is still developing. Only with the indispensable “=” signs in the equation can the theory. developed on the law of conservation of fundamental particles. A Few Final Words In the previous mathematical analysis of the laws of conservation. it is necessary to develop conservation equations in terms of fundamental particles and to write them in modern symbolic form. certainly modern physics. For example. the great idea. It took several thousand years to develop the Law of Conservation of Matter– Energy. .4. but surely the consequence will be revolutionary.A Theoretical Foundation for the Laws of Conservation 119 5. many important questions are left open. and technology could hardly exist. The law has been considered as one of the greatest syntheses of all time. contained in the conservation law. be made into one “exact science” with the capacity of prediction. chemistry. The full range of its implications is not yet clear. comparable to those of Ptolemy and Newton. Without the “=” signs of the conservation laws.

Mesarovic and Takahara introduced a concept of general systems. Since the theory of general systems has been established on set theory. This theory is used to study two problems in modern physics. Some important and fruitful leads for future research are posed in the final section. Hopefully. As the application of the systems concept has become more widespread. Based on the discussion. in the 1960s.CHAPTER 6 A Mathematics of Computability that Speaks the Language of Levels In this chapter we use the modern systems theory to retrace the history of important and interesting philosophical problems. Two difficulties appearing in modern physics are listed and studied in terms of a new mathematical theory. To meet the challenge of forming a unified foundation of the theory of systems. one shortfall of this “theoretical foundation” is that it is difficult or impossible to quantify any subject matter of research. This theory reflects the characteristic of the multilevel nature. and. based on Cantor’s set theory. the meaning of system has been getting wider and wider. 1991).1. based on Wang’s work (1985. Introduction Since the concept of systems was first formally introduced by von Bertalanffy in the 1920s. one is introduced. this work will make up the computational deficiency of the current theory of general systems. 6. the need to establish the theoretical foundation of all systems theories becomes evident. 121 . we show that the multilevel structure of nature can be approached by using the general systems theory approach initiated by Mesarovic in the early 1960s. In this chapter recent results in general systems theory are used to show the need for a mathematical theory of multilevels. Some elementary properties of the new theory are listed.

They sought causes and explanations in terms of the eternal working of things themselves rather than in any divine. Can this process go on forever? If the answer to the question is “yes. This term was intentionally chosen by Democritus to emphasize a particle so small that it could no longer be divided. Anaxagoras’ “seeds. the n -dimensional Euclidean spaces. nebulous substance. human minds have been searching for the organic structure of multilevels of the universe since at least the time of the one-element Ionians (624–500 B. For more on this. the plane. see (Perlman. first introduced formally by Mesarovic in the early 1960s. 1970). Ionians had a naturalistic and materialistic bent. and yet was an improvement over all of them. Using this idea inductively. in the search for order.). each Ionian believed that all things have their origin in a single knowable element: water. He was unwilling to submerge the tremendous varieties of things into any common denominator.. the universe originally and basically consisted entirely of “atoms” and a “void” in which atoms move. . or some indeterminate.” then “the world would be infinitely divisible”?! If the answer to the question is “no. air. The opposite idea — that the world is made up of fundamental particles — appeared from 500 to 55 B.C. . The Leucippus– Democritus atom combined features of the Ionians’ single element. etc. “Atom” means “not divisible” in Greek.2. For example.C. . This structure of general systems is multilevel. Mesarovic’s definition of a general system is as follows (Mesarovic and Takahara.C. we see that an element S1 of the system S could be a system. No matter how far an object is divided. or supernatural intervention. This section shows that general systems theory. based on Cantor’s set theory. fire.” and some thoughts of other schools. mythological. To Leucippus and Democritus. every object is infinitely divisible. With his philosophy. Looking for a single basic reality.122 Chapter 6 6. and preferred to accept the immediate diversity of things as is. what is left would have characteristics of the original substance. an element S 2 of system S 1 could be a system.). since elements in the nonempty sets V i could be systems also. Multilevel Structure of Nature Even though the major part of classical and modern mathematics deals with subject matters of one level. can be seen as a new research approach to the multilevel structure of nature.” does that mean the world is made up of fundamental elements? The idea that the world is infinitely divisible can be found in Anaxagoras’ “seeds” philosophy (510–428 B. 1975): A system S is a relation on nonempty sets Vi : where I is an index set. . such as problems on the real number line.

such that for each pair i. such that S i is an object in Mi – 1 . 1994. This new definition of general systems generalizes that of Mesarovic. Wu and Klir independently defined the concepts of pansystems and systems at a much higher level. respectively. .2.Ri ). i. that can be either finite or infinite. We note that the concept of general systems was defined in more general terms by many other authors as well. 2.A Mathematics of Computability that Speaks the Language of Levels 123 In the search for an answer as to which of the two opposite ideas could be right. There is no system such that the object set of the system consists of all systems. For each relation r ∈ R. if M1 = M 2 and R 1 = R 2 .R ) of sets. r is defined as follows: there exists an ordinal number n = n (r) . System S 1 is a partial system of system S 2 if M1 ⊆ M2 and for each relation r1 ∈ R 1 there exists a relation r 2 ∈ R 2 such that r 1 = r 2M1 = r 2 ∩ ( M1 ) n(r 2 ) . where (ZFC) means that we assume all the axioms in the ZFC axiom system (Kuratowski and Mostowski. Theorem 6. if S is an ordered pair ( M. where R is a set of some relations defined on M. depending on r.2. Then each chain of object systems of S must be finite. and M and R are called the object set and the relation set of S. System S 1 is a subsystem of S 2 .R 0 ) if there exist systems Si = (Mi .j < α with i < j . For more details see (Wu. introduced the following definition of general systems: S is a (general) system. R 0 ) be a system and Sn = ( Mn . Assume the length of the empty relation is 0. 1993) can be proved. i = 1. Two systems S i = ( Mi .1 [Russell’s paradox]. and called the length of r. For each system S there exists exactly one set M(S) .2. A chain of object systems of a system S is a sequence {S i : i < α } for some ordinal number α. Then it is impossible for S0 to be a subsystem of Sn f o r each integer n > 0. such that r ⊆ M n . if M1 ⊆ M2 and for each relation r 1 ∈ R1 there exists a relation r 2 ∈ R 2 such that r 1 ⊆ r 2M1 . . n( ) = 0. consisting of all basic objects in S. Rn ) is an nth-level object system of a system S 0 = ( M0 . Many important structures of systems can now be studied that otherwise could not have been with the one-relation approach. the following mathematical results (Lin and Ma. Theorem 6. 2.2 [ZFC]. Suppose that S is a system. n – 1. Let S 0 = ( M0 . there exists an integer n = n(i. Each element in M n is called an nth-level object of S 0 ..2. Each element in M is called an object of S. For example. . of different-level object systems of the system S. Klir. 1976) are true. A system Sn = ( Mn . Theorem 6. 1985).3 [ZFC]. .e.j ) such that the system S j is an nth-level object system of the system Si .R i ). Ma and Lin (1987). i = 1. where a basic object in S is a level object of S which is no longer a system. 1 ≤ i ≤ n. R n ) an nthlevel object system of S 0 . Based on these notations. Theorem 6. are identical. .4 [ZFC].

have been winning victories one after another. As for the photon. even it is assumed that m0 = 0. contradicting observations. was introduced as follows: δ : . We can then do the following general systems modeling: For each chosen matter. Einstein’s relativity and Dirac’s quantum mechanics.1. the theories. a function. Dirac’s δ Function In Dirac’s well-known work (1958). please refer to Chapter 9.1) because in classical mathematics on the real number line.124 Chapter 6 For detailed treatment of these theorems. R ) be a systems representation of a chemical reaction process. the rest mass m0 should be zero. the inertial mass m of a moving particle is related to its rest mass m 0 as follows: (6. As introduced by Klir (Klir. real situation problem. 0 ÷ 0 is undetermined. Some Hypotheses in Modern Physics Without using a mathematics of multilevels. such as a chemical reaction process. However. (6.4. .1) where c stands for the speed of light and v is the particle speed. or environment. However.2.3. later named Dirac’s δ function. the fallacy involved in the calculation still remains in Eq. or from Eq.2. In this section two such hypotheses are listed to further support the need for a mathematics of multilevels. 6.1) m = ∞ . (6.3. some hypotheses in modern science seem to be contradictory. let S = ( M . It can be seen that S represents the chemical reaction of interest. 6. for example. can always be defined. 6. established on the hypotheses.3 and 6. 1996) or in Chapter 5 the same methodology was used to show that a theoretical foundation can be established for the Law of Conservation of Matter– Energy. The Rest Mass of a Photon According to Einstein. For example. a system is what is distinguished as a system.2. Based upon this modeling and Theorems 6.3. describing the matter of interest. are two of these theories. where is the set of all real numbers. it follows that we have argued theoretically that the world is finitely divisible. a system. 1985). since its speed is v = c. In (Lin. such that M stands for the set of all substances used in the reaction and R is the set of all relations between the substances in M.

Schwartz introduced the concept of distributions. 1968) studied the concept of generalized power series. is introduced. the shortfall of Robinson’s nonstandard analysis is that infinitesimals and infinities can only be shown to exist and have no concrete expressions of any tangible form. Laugwitz (1961. it is impossible to have such a function. The concept of the GNS is based upon (Wang. Independently. etc.2) where ƒ (x ) is a function on with compact support and is infinitely differentiable. including modem measure theory. Algebraically. established in the early 1960s by Robinson (Davis. first introduced the infinitesimal and infinitely large numbers formally and legally. in classical mathematics. The concept of the GNS was first introduced in the early 1980s by Wang (1985. the international mathematics community became desperate to come up with a feasible explanation for the strange δ function. see (Richards and Youn. 1977). In 1950. some classical mechanics problems. With nonstandard analysis. can be greatly simplified and become more intuitive. based on which the δ function was understood as a linear functional defined on the “elementary space” consisting of real-valued functions with some special mathematical properties. nonstandard analysis. However. This theory broke through the limitation of the classical concept of real numbers. Based on the argument of the existence of infinitesimals and .. 1991) without reference to Laugwitz’s work and Robinson’s nonstandard analysis. called a generalized number system (GNS). 1990). based on mathematical logic. This function has played an important role in quantum mechanics. which successfully solved an open problem in general topology. The GNS is really a non-Archimedean number field and has found many successful applications in modern physics. 1964). since each function ƒ(x ) satisfying the first condition must possess the property that Because of the magnificent role that Dirac’s δ function had played in the success of quantum mechanics. A Non-Archimedean Number Field In this section a number field. However. algebra. where infinities are usually considered ideal objects.4.A Mathematics of Computability that Speaks the Language of Levels 125 satisfies the conditions: δ (x ) = 0 whenever x ≠ 0 δ (0) = ∞ (6. Historically speaking. such as mathematical analysis. many classical theories. 6. norm spaces. For details. the GNS is a subfield of Laugwitz’s generalized power series. 1962.

126

Chapter 6

infinities given in Robinson’s theory, the GNS has the computational advantage, since each number here, whether ordinary real, infinitesimal, or an infinity, has an exact expression in terms of the real numbers in the classical sense. A generalized number x is a function from to where is the set of all integers and is the set of all real numbers, such that there exists a number called the level index of x, satisfying x ( t ) = 0 for all t < k, and x (k (x )) ≠ 0 Then x can be expressed as a generalized power series as follows: (6.3)

(6.4) where 1 ( t ) stands for the function from to such that 1 (t ) ( s) = 0 if s ≠ t, and 1(t ) (t ) = 1 and x (t ) ≡ x t . Intuitively, the right-hand side of Eq. (6.4) is the “graph” of the function x : . The set of all generalized numbers will be written as GNS. We now define operations of generalized numbers as follows: (1) Addition and subtraction.: Let x,y be arbitrary from GNS. Define (6.5) (2) Ordering.: The order relation between x and y ∈ GNS is the lexicographical ordering; i.e., x < y iff there is an integer k 0 such that x (t ) = y (t) if t < k 0 and x (k 0) < y (k 0). (3) Scalar multiplication.: Let c ∈ and x ∈ GNS define (6.6) Because of the ordering relation defined in 2, we can now think of GNS as a system of numbers of different level; and the function 1( t) will be called the “unit element” of the tth level.

A Mathematics of Computability that Speaks the Language of Levels

127

(4) Multiplication.: Let x,y ∈ GNS. Then the product x × y or xy is defined by (6.7) Specifically, 1 ( t )

× 1( s ) = 1 (t+s ) .

(5) Division.: For x,y ∈ GNS, if there exists a z ∈ GNS such that y×z=x then the generalized number z is called the quotient of x divided by y, denoted by x ÷ y or x / y. Theorem 6.4.1. If y ≠ 0 ≡ ∑ t 0 × 1( t) , the zero function from quotient x ÷ y always uniquely exists. to then the (6.8)

Proof: Assume that the known and the unknown generalized numbers X, y, and z are

and

We need to find z such that (6.9) with k (y ) ≤ u ≤ s – k (z) and k (z) ≤ v ≤ s – k (y). This second inequality implies that we can define k (z ) = k (x ) – k(y ). Hence, replacing s by k(x ) in Eq. (6.9), we have

That is,

128

Chapter 6

Therefore, we have

When s is replaced by k(x) + 1 in Eq. (6.9),

Therefore,

By induction, we can continue this process and define each z(m), for m ≥ k (z) = Thus, the generalized number z is defined based on the given x and y. The uniqueness of z can be shown by observing that the level index of z cannot be less than To this end, assume that for some nonzero whole number n. Then replacing s by k(x) – n , in Eq. (6.9), we obtain

contradiction. Therefore, the whole number n must be zero. That is, the generalized number z must be unique. Theorem 6.4.2. For each x ∈ GNS such that x(k(x)) ≥ 0 and k(x) is an even number, there exists a unique y ∈ GNS such that y² = y × y = x. This generalized number y is called the square root of x, denoted Proof: Assume the unknown generalized number y is

Then the equation y × y = x is equivalent to that for any integer s, (6.10) where That is, When s = k (x), Eq. (6.10) becomes We can then define

Since

. When s = k(x) + 1, Eq. (6.10) becomes

A Mathematics of Computability that Speaks the Language of Levels

129

By applying induction on the index k(y) + i, the generalized number y can be well defined. To show the uniqueness of y, it suffices to see that k(y) cannot be less than k(x)/2, as defined. To this end, assume for some nonzero whole number n. Replacing s by k (x) – 2n, changes Eq. (6.10) to

contradiction. That is, the definition of the generalized number y is unique. The proof of Theorem 6.4.2 can be generalized to obtain 6.4.3. Theorem 6.4.3. For each natural number n ≥ 2 and each x ∈ GNS such that k (x ) is a multiple of n, then (1) If n is even and x(k(x)) ≥ 0, there exists a unique y ∈ GNS such that

(2) If n is odd, there exists a unique y ∈ GNS such that y n = x. The generalized number y in (1) or (2) will be called the nth root of x and denoted by Since y × y = In general, k(y n ) = n k(y), where xi ∈ G N S, i = 1,2, ... , n, and That is, if k(x) is not a multiple of n, there is no y ∈ GNS such that y n = x. In this sense, then Theorem 6.4.3 is the best we can get. To conclude this section, let us observe the following facts. Theorem 6.4.4. (GNS, +, ×, <) forms an ordered non-Archimedean field. If and k and h are integers satisfying k < h, then the generalized numbers contained in I (h) are infinitesimals, compared with those contained in I(k ) ; conversely, nonzero generalized numbers in I( k ) are infinities compared with those contained in I(h) .

6.5. Applications
In history, “laws” of the nature have been frequently expressed and treated with the rigor and beauty of mathematical symbols. For the most part, this application has been very successful. However, in the past 40 some years, theoretical physicists

130

Chapter 6

have been faced with difficulties in mathematical calculation. According to Wang (1991), such difficulties arise mainly because the modem world of physics is multilevel in character, whereas applicable mathematics still remains at the stage of single level where the only infinities are denoted by – ∞ or + ∞ , and are considered in our macroscopic world as ideal objects and treated with a different set of rules, compared with those of ordinary real numbers. Now, in GNS, the field is isomorphic to I (0) , which represents measurable quantities within the macroscopic world. Also, all numbers in GNS, whether infinities or infinitesimals, are treated with the same set of rules. In this section we show how the difficulties listed in Section 6.3 can be resolved with GNS.

6.5.1. The Hypothesis on the Rest Mass of a Photon
The difficulty in the calculation of m = 0 ÷ 0, discussed in Section 6.3, can be overcome as follows. Suppose that the rest mass m 0 of a photon is

and that the inertial mass m in Eq. (6.1) is a finite real number; that is, m ∈ I(0) has the representation

Now the velocity of a photon is

where v 2 < 0, since c is assumed by Einstein to be the maximum speed in the universe. Equation (6.1) can then be written as

where c is understood in the same way as m ∈ I (0) . By substituting the GNS values of m, m 0 , 1 (0) , v, and c, into this equation, and by comparing the leading nonzero That gives the terms (i.e., the coefficients of 1(2) ), we obtain relation between v 2 and m 1 .

A Mathematics of Computability that Speaks the Language of Levels

131

In this argument, the existence of the square root as follows:

can be seen

and

So

and

From Theorem 6.4.2, it follows that

exists.

6.5.2. A New Look at Dirac’s d Function
We show that, as a real-valued function is a function from a subset of I (0) to a subset of I (0) , Dirac’s δ function can be considered as a function from GNS into GNS. Therefore, it is a more natural representation. At the same time, the integral sign in Eq. (6.2) is no longer a symbol without much mathematical meaning. For convenience, boldfaced letters will be used for GNS-related concepts and lightface letters for real-valued concepts. For example, f( x) is a real-valued function, and f(x) is a function from a subset of GNS to a subset of GNS . Without loss of generality, let y = f(x) be a function from GNS to GNS. We define the integral of the function inductively as follows. Let E = {x ∈ GNS : f(x) ≠ 0}. Case 1: Suppose that for each x ∈ E, x –m = 0 for all m ∈ the set of all natural numbers, and that y = (… , y –m , … ,y 0 , … ,y n , … ), where, in general, {0}. y k = y k (x) is a function of all x m , m ∈ Step 1: Define a subset H (0) ⊆ as follows: x0 ∈ H (0) iff y – n = 0 if n > 0 and y 0 = y 0 (x 0 ) a function of x 0 (6.11)

132

Chapter 6

and let (6.12) If the real-valued function f (0) is Lebesgue integrable on then – H (0) has Lebesgue measure zero, denoted have all been defined. Step n: Assume the real numbers define a subset H ( n ) ⊆ as follows: x n ∈ H ( n ) iff For fixed y – m = 0 if m > n and y– n = y –n (xn ) a function of x n and let (6.14) If the real-valued function is Lebesgue integrable on denote the (6.13)

integral on by is meaningful — that is, if there are only up to countably many nonzero summands and the summation is absolutely convergent—then the sum is denoted a( n) . Definition 6.5.1 [Wang (1985)]. If is convergent and equals a, then the function f(x) is GNL-integrable and we write

Case 2: Partition the set E as follows: E k = {x ∈ GNS : k(x) = – k}. Consider the right-shifting of , and the left-shifting of f(x):

where

Using the method developed in Case 1, denote Definition 6.5.2 [Wang (1985)]. If ∑ k ak is convergent, then the function f(x) is GNL-integrable and we write

0.1 [Wang (1985)]. (6. . Proof: Pick a real-valued function g: entiable with bounded support and satisfies such that g is infinitely differ- Define for x = ( .. x1 .) otherwise Then for x = (. 0. Then it can be shown that 6. let g: be arbitrary and infinitely differentiable with bounded g(x ) dx = 1. x2 . . . x1 . 0... Theorem 6.x 2 . ... g(x) is a Dirac δ function according to Eq.5. Sup( g ) = {x ∈ Then the GNL-integrable function g: GNS → GNS can be defined if x = (. Let ƒ: differentiable and define f : GNS → GNS as follows: be infinitely This function is called the GNS function induced by ƒ(x). .x 1.5. .2).1 holds. Let ƒ and f be defined as before. x2 .A Mathematics of Computability that Speaks the Language of Levels 133 We are now ready to look at Dirac’s δ function. . ) otherwise That is. ..) otherwise .. Then there is a GNL-integrable function g: GNS → GNS such that In particular. .. g(x) ≠ 0 } is bounded in and support: that is..

1. Not only this. some open problems in relevant fields might be answered. 1985.. for each n = 0. . distribution theory. we hope that the study of GNS will furnish a convenient tool for a computational general systems theory.. Wang. many studies can be carried out along the lines of classical research. please refer to Chapter 11 6. Some Final Words If the ordered field GNS is substituted for we can see that the theory and applications of GNS will have a great future. quantum mechanics. .. etc. As shown in (Wang. such as mathematical analysis. and some new fields of research might be opened up. and that g (t)dt = 1. since has been applied in almost all areas of human knowledge and been challenged numerous times. Most importantly. 1991). it follows This completes the proof.134 Chapter 6 Since (f)(x)g(x))– n = 0. on GNS.2.6.. For a more rigorous treatment of GNS. we also expect that when one is studying topics from various disciplines with GNS.

1. and “the initial and terminal states” for the beginning and ending vertices v1 and v 2 . each subpath must be an optimum path connecting the initial and terminal vertices of the subpath. a “subpolicy” a subpath. Bellman’s principle has become the cornerstone in the study of the shortest-path problem. in a weighted graph. The principle says (Bellman. 1957) that each subpolicy of an optimum policy must itself be an optimum policy with regard to the initial and the terminal states of the subpolicy. However. The principle implies that for each chosen optimum path from v1 to v 2. economics. etc. respectively. equipment-replacement models. Furukawa. See (Bellman and Zadeh. Wu. the resource allocation problems.CHAPTER 7 Bellman’s Principle of Optimality and Its Generalizations In dynamic programming Bellman’s principle of optimality is extremely important. travelingsalesman problem. a number of practical optimization problems have been put forward in management science. 1981a. the optimum options are no longer maxima or minima of some set of real numbers. For example. and consequently. In this case. A Brief Historical Note During the last half century or so. an “optimum policy” stands for a path with the smallest weight. 1980. Mitten.” “optimum policy. For details see Dreyfus and Law (1977). 7. One feature common to all these problems is that the performance criteria are not real numbers. Wu. inventory problem. a “policy” means a path.” “subpolicy. since the early 1970s. Henig.” and “state” are are primitive and are not given specific meanings. Many dynamic programming models with nonscalar-valued performance criteria have been established. 1980. 1982). Baldwin and Pilsworth. Here. an optimum path from a vertex v1 to vertex v2 i s the one with the smallest sum of weights of the edges along the path. and many other fields. the terms “policy. 1970. 1983. 135 . 1974. these terms are understood in certain ways comparable to the situation of interest. In each application.

then the sum of local optimization orders is less than or equal to the global optimization order.2 [Wu and Wu (1986)]. it was shown by Qin (1991).1. Question 7. Hu (1982) constructed an example to show that the principle does not hold in general. 7.1. between the fundamental equation approach and Bellman’s Principle of Optimality? To seek the answer to these questions. Therefore. . as many scholars have realized. In what range does the fundamental equation approach hold? How can its validity be proved? What relations are there. while the fundamental equation approach offers one an effective method to obtain the desired optimal policy. Since the validity of Bellman’s principle is limited. A subpolicy of the optimum policy is no longer optimum. it is the fundamental equation approach. scholars began to question how far the validity of the principle can be carried. In fact.2. if any. is the fundamental equation approach also limited? Again. not the principle of optimality. that makes up the theoretical foundation of dynamic programming.1. Hu’s counterexample shows that the fundamental equation approach does not always work. Under what characterizations can a given problem be solved by dynamic programming? At the same time.136 Chapter 7 Figure 7. the following natural question arises. Thus. The optimum principle gives one a property of the optimum policy. the following questions are natural: Question 7. several approaches of systems analysis are presented. Since the application of Bellman’s principle is no longer restricted to classical cases. Hu’s Counterexample and Analysis Let us first look at Hu’s counterexample. that if the goal set is strong antisymmetric.1 [Hu (1982)].

For any a0 . where W ↑ G² is the set of all mappings from G² = G × G into W. define where Then Bellman’s principle of optimality can be written as follows.Bellman’s Principle of Optimality and Its Generalizations 137 Example 7. For any natural number n. Theorem 7. representing the set of all vertices. then for any 0 ≤ i ≤ j ≤ n . ƒ 〈4 〉 (A. since B2→ C → D 1 → E is. with A → B 1 → C → D 1 → E being its policy. but its subpolicy B 2 → C → D 2 → E is not optimum. Hence. } be the set of all possible weights.Applying these equations to Hu’s counterexample. the set of all subsets of W. then each network or graph or weighted graph is simply an element g ∈ W ↑ G ². Now let W ={0. b ∈ G.3) where z is the zero in W.2. the fundamental equation on (g.3 . into W such that (7.E ) ≠ g 〈 4〉 ( A. b) is given by the following functional equations: where 2 ≤ m ≤ n. 7..E). For any g ∈ W ↑ G ². a natural number n.1. a. ° If G is a nonempty set.2.. θ1 = minimum: p(W) → W be the optimum option. we ask when will the weight obtained from the fundamental equation approach be the same as the global optimum weight? . That is.1 [Bellman’s Principle of Optimality]. and a. Then the policy A → B2 → C → D2 → E is optimum.1.E ) =1. However. which is a mapping from the power set p( W ) of W. elements (7. in this example.2. Define an optimum path as one where the sum of all arc lengths (mod4) is the minimum. one obtains: ƒ 〈 4〉 ( A. Consider the network in Fig.1) and let θ2 = {+} (mod4) : W² → W be the binary relation on W such that ° (7.2) where stands for the ordinary composition of mappings. n . an ∈ G.1.. it can be easily seen that g 〈 4〉 ( A E ) = 0 and A → B 2 → C → D 2 → E is the global optimum policy with weight 0. which has weight 0.

In general. let W and G represent two nonempty sets.1). and for all for all ≠ X ∈ p (W ) and x ∈ W . In this section. implies w2 = w3 and implies w 2 = w3 whenever w1 is not the zero element in W. For any element g ∈ W ↑ G 2 and any a 0 . w 2 . w 3 ∈ W. (7. θ1 : p (W ) → W be the optimum option satisfying Eq. or simply zero if for each w ∈ W and for each The optimum option θ 1 is X-optimum-preserving for X ∈ p (W ) if for any The option θ 1 satisfies the optimum-preserving law if θ1 is X-optimum-preserving for all The relation θ 2 is distributive over θ1 if ≠ X ∈ p ( W ) and any x ∈ W. .3. 1984a). Generalized Principle of Optimality By formalizing the discussion in the previous section. define where An element z ∈ W is called a zero element. w l . w 2 . Wu and Wu (1984a) obtained one sufficient and three necessary conditions for Bellman’s principle to hold.138 Chapter 7 7. all results are based on (Wu and Wu. The relation θ 2 is associative if for any The relation θ 2 satisfies the cancellation law if for any w1 . w3 ∈ W. and θ2 be a mapping from W 2 into W. a n ∈ G.

. G. The generalized principle of optimality is a sufficient condition of the conservation from a global optimization to local ones. Then in the system ( W. since and and then by the optimum-preserving law of θ1 . .Bellman’s Principle of Optimality and Its Generalizations 139 Theorem 7. ai ) ≠ z and ω (a j aj +1 . θ 2).. . In the following section.3.1 [Generalized Principle of Optimality (Wu and Wu. by the cancellation law of θ 2 . . it follows that By the associative law and the distributive law over θ1 of θ 2 . n is a natural number.2. 1984a)]. and For any 0 ≤ i ≤ j ≤ n. Proof: Suppose that g ∈ W ↑ G 2 . one has one has that ω (a 0 a 1 . cancellative and distributive over θ1 .1) holds. Assume that θ1 satisfies the optimum-preserving law and that θ 2 is associative. That is. it is shown that the conditions in the generalized optimality principle are also necessary except for the condition of associativity. θ1. Bellman's Principle of Optimality (Theorem 7. a n ) ≠ z. Since Therefore.

From the assumption that the Generalized Principle of Optimality holds. a n ∈ G satisfy Eq. . it follows that (7. it follows that (7. then θ 2 possesses the following property of cancellation: for any 0 ≤ i < j < k ≤ n and any (7. a1 . it follows that That is.4. . and a 0 .1 [Wu and Wu (1984a)]. .3. (7.140 Chapter 7 7.5) By Eq. n a natural number. .3).4) implies and implies and from the assumption Proof: Since that the Generalized Principle of Optimality holds.4) and the associative law of θ2 . (7. Some Partial Answers and a Conjecture Theorem 7.6) .1) holds for a chosen g ∈ W ↑ G 2 .4. If the Generalized Principle of Optimality (Theorem 7. Assume that θ2 is associative.

(7.8) and the associative law of θ2 . The proof of the other identity is similar and is omitted. an ∈ G satisfy Eq.7) and (7. . Theorem 7. Assume that θ1 satisfies the optimumpreserving law.7) The assumption that θ1 satisfies the optimum-preserving law implies that (7.8) Then. n os a natural number. it follows that just what is desired. . it follows that (7.5) and (7. θ2 is associative. (7.3). and Proof: Since it is assumed that the Generalized Principle of Optimality holds. (7. . If the Generalized Principle of Optimality (Theorem 7.2 [Wu and Wu (1984a)].1) holds.Bellman’s Principle of Optimality and Its Generalizations 141 By Eqs.6). from Eqs.3. and a 0 . then θ2 possesses the following distributivity over θ1 : for 1 ≤ i ≤ n – 1. . a 1 . . it follows that The proof of the other equation is similar and is omitted.4.

If the Generalized Principle of Optimality (Theorem 7.4.142 Chapter 7 Theorem 7. In the system (W. . by associativity and the distributive law over θ1 of θ 2. We conclude with the following conjecture. Therefore. the Generalized Principle of Optimality holds iff there is an ordering relation ≤ on W such that (i) for any x. y ∈ W. one and only one of the following must be true: (ii) the optimum option θ1 is defined by θ1 = superior or θ1 = inferior under the ordering relation ≤ (iii) the ordering relation ≤ and the relation θ2 are consistent.1. a1 .1) holds. .3. then θ1 possesses the following optimum-preserving property: for each 1 ≤ i ≤ n – 1. . That is. G. . an ∈ G satisfy Eq. Proof: Since the Generalized Principle of Optimality is assumed to be true.3). x ≤ y iff x θ2 w ≤ y θ2 w and w θ2 x ≤ w θ2 y for any w ∈ W. and a0 . θ 2 ). θ1 .4. it follows that The proof of the other identity is similar and is omitted. .3 [Wu and Wu (1984a)]. (7. Conjecture 7. Assume that θ 2is associative and distributive over θ 1 and n is a natural number.

it is obvious that Suppose that for some and any a and b ∈ G. n. Here. In this section. the mappings θ1 and θ 2 express the recursiveness of operation.9) is a generalization of the fundamental equations in dynamic programming (Bellman. θ 2 ). which in turn shows that Eq. It is now sufficient to show that Since θ 2 satisfies the associative and distributive laws. one has . then for any and a. under the conditions. and θ 2 the associative and distributive laws over θ1 . However. mathematical induction will When m = 1.1 [Wu and Wu (1986)]. G . b ∈ G . and a. θ 2 ). for any g ∈ W ↑ G 2 . Theorem 7. b ). defined in the same way as in the previous section. θ1 . That is.5. a sufficient condition and two necessary conditions are established under which the fundamental equation approach holds. a . the weight computed from the fundamental equation equals that of a global optimum. the following functional equations are called a fundamental equation on (g. for any x ∈ G. In the system (W. or simply a fundamental equation: (7. Bellman’s Principle of Optimality has been generalized and three necessary conditions given. 1957). Generalization of Fundamental Equations In the previous two sections. θ1 .9) for 2 ≤ m ≤ n. (7. The presentation here is based on Shouzhi Wu and Xuemou Wu (1986). the following is true: Proof: For any be used on the index m to show that and a. one should keep in mind that the optimum principle only states a property of the optimum policy. if θ 1 satisfies the optimum-preserving law. b ∈ G . while the fundamental equation approach offers an effective method for one to obtain the desired optimal policy. G.5.Bellman’s Principle of Optimality and Its Generalizations 143 7. a natural number n. For a given system (W. b ∈ G .

. the following equation holds: Now the assumption that θ1 satisfies the optimum-preserving law and the Claim guarantee that Thus. .cn – 1 ∈ G. which is what we desired. for any x1 ∈ G. then for any and a and b ∈ G. Since and . Proof of claim: First note that for any x1 ∈ G.144 Chapter 7 Therefore. it follows that (7. Claim. . If θ1 satisfies the optimum-preserving law. let us see why the claim is true.10) Assume that for some c1 . . one has Hence. c 2 . the following is true: Therefore. To finish the proof of the theorem.

b ∈ G. (7. it is true that then θ 2 possesses the following distribution over θ1 : for any g and a. Proof: For any and a. it follows that Therefore. from the Claim it follows that From the assumption that θ 2 satisfies the associative law. The following results show that the conditions in Theorem 7. if for any and a. θ 2 ) with θ 1 satisfying the optimum-preserving law and θ 2 satisfying the associative law. In the system (W. Theorem 7.11) Now.2 [Wu and Wu (1986)].5.Bellman’s Principle of Optimality and Its Generalizations 145 and θ1 satisfies the optimum-preserving law.1 are also necessary. (7. G .11) and the assumption that θ1 satisfies the optimumpreserving law imply that This ends the proof of the theorem.12) . and since θ 1 satisfies the optimum-preserving law. b ∈ G. θ 1 .10) and (7. Eqs.5. b ∈ G. it follows that This equation implies at once that (7.

b ∈ G . In the system (W. the (7.13).13) By Eqs. θ1 . the following is true: Proof: Since for any and a.146 Chapter 7 Now. G . it is true that and θ 2 satisfies the associative law and the distributive law over θ1 . (7. b ∈ G. it is true that then θ1 possesses the following optimum-preserving property: for any g ∈ W ↑ and a and b ∈ G . it finally follows that This ends the proof of the theorem. from the assumption that for any Thus. b ∈ G.5. it then follows that as desired. following is true: and a. if for any g ∈ W ↑ G 2. and a.3 [Wu and Wu (1986)].12) and (7. Theorem 7. θ 2 ) with θ 2 satisfying the associative law and the distributive law over θ1 . .

defined by For a subpolicy with initial state of x i ∈ Xi and terminal state xj ∈ X j . the set E consists of all available policies. · ) : W → W are strictly order-preserving. The generalized result is called the “operation epitome principle. let Xi be the set of all states of the ith stage. (x. The gain of the policy e1 – N = (e1 .6. y ∈ W 2 .” For an N -stage sequential decision process. The binary relation g ⊆ W 2 establishes the comparison between criteria. 2. In addition. e N ) ∈ E is measured by the function ƒ : E → W. the gain of e i – j is defined as follows: and . for any x . . the mappings F ( · . . e 2 . where X0 .y) ∈ g iff x ≤ y. y ∈ W. i ≤ N . i = 0. w) < F (y. . w). w2 . Operation Epitome Principle This section presents a different approach to systems analysis. Further. That is. . 1. y).Bellman’s Principle of Optimality and Its Generalizations 147 7. . X 1 . . then the multiplication operation can also be the mapping F. . x) < F (w. Intuitively speaking.w i ) ∈ W i . . (2) For each (w1 . if W contains all positive real numbers. . (3) For each fixed w ∈ W. and 1 ≤ k ≤ i. Define a mapping so that the following conditions are satisfied: (1) F (w) = w for each w ∈ W. . . . X N are some fixed sets. . Bellman’s Principle of Optimality once again is generalized. For example. which is due to Qin (1991). and x < y iff F (w. and the set W is made up of all the criteria used in the decision-making process. .. h i j : E i × Xj → W is the gain (function) of decisions. N. . With this approach. x < y iff F (x. w) and F(w. it can be shown that the addition operation of real numbers can be the mapping F. Let and define a binary relation g ⊆ W 2 by the following: for any x.

the subpolicy ei–j = (ei .ƒ . respectively. g). If X ⊆ W is a nonempty subset. x) : x ∈ X }. E .148 Chapter 7 Now. we first formalize the previous discussion. Ei – j .6. e 2 . . if x ≠ w . Lemma 7. then there exists at least one x ∈ X = ƒ (E) such that x ≠ w and ( w. Then. and E is the set of all possible policies. E. For any subset X ⊆ W. g) such that E and W are nonempty sets. x) ∉ g. the stationary set (of the policy e) is defined If STY(e) = . ƒ : E → W is a mapping and g ⊆ W 2 is a binary relation. ei +1 . . These sets are called the criteria set and the policy set of (W. and the element e ∈ E is a generalized optimal solution (or policy) of the model (W . To prove this new version of Bellman’s principle. Intuitively speaking.E. defined by I = {(x. This system is called the optimization model of the general systems. by the definition of g * X it follows that. define g * X = {w ∈ X : for any x ∈ X.ƒ.1 [Bellman’s Principle of Optimality]. g). For any policy e1 – N = (e1 . Let e ∈ E be such that w = ƒ (e) ∈ ƒ (E) – g * ƒ(E) . there exists exactly one policy w ∈ W and the rule matching the criterion e with the policy w is described by the function ƒ : E → W. Consider the system (W. for each policy e ∈ E. Theorem 7. . for any x ∈ X. x) ∈ g . . w ∈ X and w ∈ ( g I )(w ) = . E : ƒ.ƒ.ƒ. if the equation STY(e1 – N ) = holds in the system (W.1 [Qin (1991)]. The binary relation g ⊆ W 2 defines how to compare criteria used. then (w. .g). . ej ) of the policy e1 – N satisfies the condition that the equation STY(e i – j ) = holds in the system (W. Bellman’s Principle of Optimality can be written as follows. then for any natural numbers i and j satisfying 0 ≤ i ≤ j ≤ N. g). . Proof: For each element w ∈ g * X. x ≠ w implies (w.6. then where I is the diagonal of the Cartesian product X2. That is. for each chosen criterion e ∈ E. . e N ) ∈ E. The set W consists of all the criteria used in an optimization problem of interest. then ƒ(e) is called a generalized extreme value of ƒ(E).14) If X = ƒ (E) and w ∈ X – g * X . x) ∉ g } (7. Then the element w ∈ W is called a generalized extreme value of the set X = ƒ(E).

2. This proves that Conversely. F(x) ≠ F (y). that is. w) : W → W are strictly order-preserving. define the mapping p 1 : E 1 → E. then for any x ∈ X. then x ∈ g (w). Now. and x ≠ w. (ii) For each fixed w. by p 2 (e 2 ) = (e'. by p1 (e1 ) = (e 1 .·) and F(·. g ⊆ E 2 a binary relation. ( W . Let W and E = E 1 × E2 be nonempty sets. A mapping F : X → W . which depends on fixed e'.F(y)) ∈ g if and only if x ≠ y . i = 1.y) ∈ g Theorem 7. That is. ƒ(e) ∈ g * ƒ (E).e' ) for each e 1 ∈ E 1 For any e' ∈ E 1 define the mapping p 2 : E 2 → E. where X ⊂ W. the mappings F(w.2 [Operation Epitome Principle (Qin.Bellman’s Principle of Optimality and Its Generalizations 149 contradiction. Proof: For any e' ∈ E 2 . then a mapping can be defined by and for each e = (e1 . is called strictly order-preserving if the following property holds: for any x. which depends on fixed e'. x ≠ w must which is impossib1e. If a mapping satisfies (i) F(w) = w.6.e2 ) for each e 2 ∈ E2 .y ∈ X. for each policy e ∈ E the stationary set (of the policy e) is defined If STY(e) = . 1991)]. then ƒ(e) is called a generalized extreme value of ƒ(E). if for imply (w. e2 ) ∈ E such that i = 1.x) ∉ g. So. (F(x). (x. mappings. 2 where  X  is the cardinality of the set X. x) ∈ g. for each w ∈ W. since otherwise. and h i : E i → W.

A binary relation g ⊆ W 2 is strictly strongly antisymmetric (Qin.z ∈ g. To this end. n ∈ the conditions (x .x) ∈ g n hold simultaneously. 1. it is easy to see that (7.7.6. . x) ∉ g for x . then (y.16) Equations (7.150 Chapter 7 Now the desired mapping is induced by and Since the mapping each is strictly order-preserving. 7. Now. (iii) (Transitivity) If (x. 2. 1991). 1991).y.y) and (y. . Bellman’s Principle of Optimality (Theorem 7. if the following conditions are satisfied: (i) (Antireflexivity) (x. recall that a relation g ⊆ W 2 is called a partial ordering (Kuratowski and Mostowski.15) On the other hand. sequence It can be seen that the concept of strong antisymmetry is a generalization of the concept of partial ordering relations. . one has that.1) is clear. the relation gn ⊆ W 2 is defined A binary relation g ⊆ W 2 is strongly antisymmetric (Qin. if there is no infinite such that wi ≠ w i +1 and (wi .16) imply the desired inequality.y) ∈ g and x ≠ y . E.z) ∈ g .y) ∈ g m and (y. . (ii) (Antisymmetry) If (x.y ∈ W such that natural numbers m. 1976). w i + 1 ) ∈ g for each i = 0. for (7.y ∈ W. z) ∈ g for x. Fundamental Equations of Generalized Systems For an optimization model (W. then (x.x) ∉ g for any x ∈ W. g) of general systems and a natural number n ∈ N.15) and (7. if for any there do not exist distinct elements x.ƒ.

and natural numbers m. w 2 ∈ W such that and such that w1 ≠ w 2 . Suppose that this statement is not true. g is strongly and that (b. Sufficiency. . the following example shows that a strongly antisymmetric relation g ⊆ W 2 may not be a partial ordering relation. That is. If a set W is finite.7.Bellman’s Principle of Optimality and Its Generalizations 151 First. 2. Let W = and for any a. . . Example 7.7. since the transitivity relation is not satisfied.1 [Qin (1991)]. 2 . 2 . . Suppose W is finite and g ⊆ W 2 is strongly antisymmetric. there exist w1 . a > b. for i = 1. . . . n ∈ From the transitivity of g. m + n –1 and (xi . b) ∈ g ⊆ W 2. g must be strongly antisymmetric. However. It can be shown that (a. it is not a partial ordering relation on W. there exist elements w1i . b) ∈ g m for m ∈ iff a + m ≥ a + 1 > b > a. such that and That is. Proof: Necessity. iff b + 1 > . . n ∈ such that Thus. and w 2j . for n ∈ antisymmetric. (a. There then exists an infinite sequence such that y i -1 ≠ yi and (yi –1 .y i + 1) ∈ g for each i ∈ contradicts the assumption that g is strictly strongly antisymmetric. then a binary relation g ⊆ W 2 is strictly strongly antisymmetric iff the relation g is strongly antisymmetric. . there exists a finite sequence such that x i ≠ x i + 1. This last condition contradicts the antisymmetry of g. Assume that g ⊆ W 2 is not strictly strongly antisymmetric. Theorem 7. yi ) ∈ g for each i ∈ . w 2 ∈ W and m . . if a relation g ⊆ W 2 is a partial ordering. m – 1. . it follows that and w 1 ≠ w 2 . . b ∈ W . Second.1. Hence. for i =1. Therefore.a ) ∈ g n . the relation g ⊆ W 2 is a partial ordering but not a strongly antisymmetric relation. From the definition of strong antisymmetry it follows that there exist w1 .x i + 1 ) ∈ g. . n – 1. . . iff b + m ≥ b + 1 > a > b . and j = 1. . Suppose that g ⊆ W 2 is strictly strongly antisymmetric but not strongly antisymmetric. m + n –1 Now an infinite sequence can be defined as follows: and which It is easy to observe that y i ≠ y i +1 and (y i . i = 1.

Question 7. such that w k = y k + p . Therefore. For any there exists such that w = ƒ(e).x) ∉ g. This condition some e ∈ E l.17) implies that based on which Eq. ( yq .x) ∉ g. if 0 < k < n – 1 and w n = y p .(7. x ≠ w implies ( w. i. . .1 [Qin (1991)]. for i = 1. y p +1) ∈ g so that q < p + 1. For any subset X ⊆ E. .18) . To this end. If not..e. is an infinite set. define (7. we have shown that last condition means that Conversely. There then exist p. This ends the proof of the theorem. n. It can be seen that (wi – 1 . then p > q. which contradicts the assumption. and that for any That is.17) and W* = g * W Lemma 7. Construct an example to show that the concept of strong antisymmetry is different from that of strictly strong symmetry. we have shown that implies that (7. w n .1.152 Chapter 7 It is true that the sequence satisfies the condition that if (y q . then q ≥ p. yp +1 ) ∈ g. . The Therefore.7. (7. we show that the elements differ from one another. Equation Proof: For any (7. w i ) ∈ g. p > q. The contradiction thus implies that g ⊆ W 2 must be strictly strongly antisymmetric.14) implies that w = ƒ (e) ∈ ƒ (E1 ) and that for any x ∈ ƒ (E1 ). To complete the proof. contradicting the fact that p < q. 2 . and (yp . w 2 . . for n = q – p + 1. y p ) ∈ g . .14) implies that for implies (w. . q ∈ such that p < q . there exists a subsequence such that Now take the finite sequence w0 . q ≤ p. and w 0 = wn . which contradicts the fact that g is strongly antisymmetric. contradicting that W is finite. Hence. elements of yp = y q . The proof of the previous theorem actually says that strictly strong antisymmetry implies strong antisymmetry. w 1 . and that when the set W is finite the converse implication is also true. suppose that the do not differ. for any Eq. Therefore. . Hence.7.

From Eq. w ∈ g * X and w ∈ (g ∪ I)(x). define a subset X(x) ⊆ W by Then where I is the diagonal of the Cartesian product X². w ∈ (g ∪ I) (x) and for any w' ∈ X(x). 2. i. . Assume that there exists a subset Ø ≠ X ⊆ W such that g * X = Ø.14) it follows that there exists x1 ∈ X such that x ≠ x 1 . Thus. w") ∈ g . It is clear that Therefore.7. (w.3 [Qin (1991)]. For any element x ∈ W and subset X ⊆ W. x) : x ∈ X }. the proof is complete. so w ∈ g * X. w") ∈ g. x 1 ) ∈ g. then g* X ≠ Ø Proof: The proof is by contradiction. This condition contradicts the assumption g is strictly strong antisymmetric. then x ∉ g * X = Ø. w ∈ g * X(x). and Ø ≠ X ⊆ W. x i + 1 ) ∈ g. . . which contradicts w ∈ g * X(x). y ≠ w implies y ∉ X. Proof: For each w ∈ g * X(x). for an element w ∈ (g * X ) ∩ ((g ∪ I) (x)). (x. an infinite sequence x1 . It follows that w ∈ X. (x1 . w ≠ y implies y ∉ X. one need only prove that w ∈ g * X. Therefore. . Thus. we have obtained that Conversely. w =X ∩ (g ∪I) (w ).. . The following inclusion relation is thus true: Lemma 7.2 [Qin (1991)]. and let x ∈ X. If g is a strictly strongly antisymmetric relation on the set W. there exists w" ∈ X(x) such that (w. . w" ∈ X(x). and x 1 ∉ g * X . w ∈ ( g ∪ I)(x). (7. w') ∉ g. has been obtained. i.Bellman’s Principle of Optimality and Its Generalizations 153 Lemma 7. Therefore. If w ∉ g * X. w' ≠ w. and that for any y ∈ ( g ∪ I) (x ). for any y ∈ (g ∪ I)(x). 3. for each i = 1. and x 2 ∉ g * X. that there exists x 2 ∈ X such that x 1 ≠ x 2 .7. x 3 .e. Hence. To show that w ∈ ( g * X) ∩ (g ∪ I)(x).. satisfying (x i . and so on.e. x 2 ) ∈ g. x 2 . defined by I = {(x. . there exists w" ∈ X such that w ≠ w" and ( w.

and G ⊆ p(W). That is. such that and and satisfying. Thus. That is.2 [Qin (1991)]. Let g be a strictly strongly antisymmetric relation on the set W. for any Y ∈ G 2 . one has that y ∈ X. the power set of W. for any X ∈ G 1 . w ∈ g * X. . Then Proof: For an element it follows from Lemma 7.19) Conversely. and Y ∈ G 2 . and and Therefore. and for any w ≠ y ∈ g (w). for any X ∈ G1 and Y ∈ G 2 .7. for an element Partition the set for any X ∈ G 1 .6.154 Chapter 7 Theorem 7. On the other hand. which implies that It has thus been shown that (7. Hence y ∉ g * X.1 that Partition the subset such that satisfying. for any it can be seen that for any and so y ∉ g * Y.

for any X ∈ G l . for any element w ∈ g * F(X). F(z)) ∈ g. Hence. It suffices to show that there is no F(w2 ) ∉ F (X) satisfying w = F(w 2 ) and F (w2 ) ≠ F (y).7. there exists y ∈ g * X such that w = F(y) ∈ F(X) and such that for any z ≠ y. E. g ) of general systems. Suppose this condition is not satisfied. Define A mapping ƒ : E → W is called p -resolving. Theorem 7. Then from the strictly orderpreserving properties of F. Again. From the arbitrariness of the element w. i. Z ) ∈ g.7. F (w 2 )) ∈ g. Suppose there is such an element F( w 2 ) ∈ F (X). so z ∉ X. It then follows that (y.20). ƒ. it follows that F(y) ≠ F(z) and (F(y).1 and 7. z) ∈ g..3 that for any Y ∈ G2 . y ∈ g * X. it follows that Conversely.20) Combining Eqs.. which contradicts the assumption that y ∈ g * X and implies that w ∈ g * F(X).7. which in turn implies that (7. This contradicts the assumption that w ∈ g * F(X). proved the theorem. X. then Proof: For an element w ∈ F(g * X). w) ∈ g and y ≠ w 2 . w ∈ g * X. and T be nonempty sets with g ⊆ W ² given. It suffices to show that there is no z ∈ X such that z ≠ y and (y. w ∈ F (g * X ) .e.e.19) and (7.3 [Qin (1991)]. (7. if there exists a mapping F : W ∪ W² → W satisfying . Lemmas 7.Bellman’s Principle of Optimality and Its Generalizations 155 Hence. the arbitrariness of the element w guarantees Let W. ( y. there exists an element y ∈ X such that w = F (y). It follows from Then one obtains that That is. In the optimization model (W. i. if F : W → W is a strictly order-preserving mapping and X is a subset of W. such that (F(y).

ƒ .ƒ.ƒ .E. the mappings F(w.1 it follows that Theorem 7.E. it follows from Theorem 7.156 Chapter 7 (i) F (w ) = w for each w ∈ W. (7.4 holds with W strongly antisymmetric if W is finite. E . (ii) For any (iii) For any w ∈ W. g) are called the zerothorder optimization solutions of the system (W. The first-order optimization solutions of the system (W. By Theorem 7. E .7. (ii) ƒ : E → W is p-resolving. the nth-order optimization solutions . g).7. If the optimization model (W. E.2 1) and Proof: It follows from Theorem 7. Theorem 7. ƒ .ƒ. (iii) then (7.g) are the optimal solutions of the subsystem of (W. Let (W.7. The optimal solutions of (W .g) be a given optimization model of general systems with the sets W and E nonempty. Eq. ·) and F(·. E . w) : W → W are strictly order-preserving.2 that For fixed e1 ∈ E1 .4 [Qin (1991)]. E.3 that Therefore. Inductively.7.21) is valid. g) .ƒ.g ) of general systems satisfies (i) for some given nonempty sets X and T.g ) obtained by removing all zeroth-order optimization solutions from the original system (W . ƒ .7.

7. . ƒ .7. and that θ2 is associative but not distributive over θ1 . Theorem 7. which is a function of the optimization solution e and is denoted by n = Oo (e). For details one has that with A → B 1 → C → D 1 → E its policy. . (ii) Consider the network in Fig.ƒ. The proof is left to the reader.E.ƒ. (n – 1)st-order optimization solutions from the original system (W .g).5.8.g).E.2.y ∈ W. Here. E. Then A → B 1 → C 2 → D is an optimum . If the optimization model (W. If e ∈ E is an n th-order optimization solution of ( W. by Theorem 7. first-order. it follows that the fundamental equations will not give the global optimum solution.2. E. At the same time.Bellman’s Principle of Optimality and Its Generalizations 157 of (W .2 says that Bellman’s Principle of Optimality does not hold.4. and with A → B 2 → C → D2 → E the global optimum policy.g) of general systems satisfies the conditions (i) W and E are nonempty.y ) = min{x .ƒ. θ2 ( x.5 [Qin (1988)]. 7. is called the optimization order of the solution e. That is. Therefore. . then the sum of the optimization orders of all subpolicies is less than or equal to the optimization order of the policy itself. Applications (i) In Hu’s counterexample it can be shown that θ1 satisfies the optimumpreserving law. for any x. . Let θ1 = maximum and θ 2 = ∧ . Define W0 = W and W i = W i – 1 – W i*– 1 for each i ∈ Each element e ∈ W i* is called an ith-order generalized extreme value of W. Theorem 7. (ii) ƒ : E → W is p-resolving. the number n. y}. and (iii) g ⊆ W 2 is strongly antisymmetric.g) are the optimal solutions of the subsystem obtained by removing all zeroth-order.

It can be shown that θ2 is associative but does not satisfy a cancellation property. but θ 1 does not possess an optimum-preserving property.1 it follows that the Generalized Principle does not hold in this case. For example. θ2 ) can be defined as follows: W = set of all possible weights θ1 = optimum choice rule θ2 = operation between weights G = the set of all states of the process Define a network g ∈ W ↑ G 2 such that. y ∈ G. and θ1 is defined as θ 1 (X ) = min{ X – {min{ X}}} = the second minimum in X for each X ∈ p (W) with  X  > 1. 7.G. That is. since its weight is 3 and the weight of B1 → C 1 → D is 2.y) = the zero element in W. The principle of optimality does not hold. From Theorem 7. If the process has n stages. but its subpath B1 → C 2 → D is not optimum. If a state y is not accessible from a state x.y ) = w iff the cost (or profit) from the state x to the state y is w.158 Chapter 7 Figure 7. θ1. path with weight 2. (iii) Consider the network in Fig. It can be seen that θ2 is associative and distributive over θ1 . a n ) is simply the cost (or profit) of a global optimum decision.2 again with θ1 = subminimum and θ2 = +. Then g 〈n〉 ( a 0 . (iv) For a sequential decision process. the system (W.2.4. g (x. while its subpolicy B1 → C 1 → D with weight 4 is not optimum. If . Then A → B1 → C l → D is the optimum path with weight 6. then define g(x. a0 is the initial state and an is the terminal state. θ2 is the regular addition. for any x.

Thus. but it does not satisfy the optimum-preserving law.G. See the network in Fig. inventory. then a i . .5.5. being recursive is equivalent to having the fundamental equation approach hold. +) satisfy the Generalized Principle of Optimality (Theorem 7. θ1 . . then for any 0 ≤ i ≤ j ≤ n. a i + 1 . . θ2 ) does not imply the recursiveness of the system. Thus. Intuitively speaking. +). 1977). (W. The optimum-preserving property of the system (W. G . and distributive laws over θ1 . . the Generalized Principle of Optimality says that if θ1 satisfies the optimum-preserving law and θ2 satisfies the associative. .G . a 1 . . θ2 ) satisfies the optimum-preserving law if for any g ∈ W ↑ G 2 . θ1 . Proposition 7. .Bellman’s Principle of Optimality and Its Generalizations 159 then a 0 . θ1 . . a system (W. G. +) the desired optimum solutions can be obtained by applying fundamental equations.8. . For a subpolicy ai .G . . . +) and ( max.a 1 . θ 2 ). n ∈ . θ 1 . a n ∈ G. So. θ1 is the maximum. 7. and can be solved by dynamic programming.1). For convenience. Proof: It suffices to construct an example to show that a recursive system ( W. is recursive if for any g ∈ W ↑ G 2 . a j .3. n ∈ . In the shortest-path. it can be shown that “min” and “max” satisfy the optimum-preserving law and that “+” satisfies the cancellation.8. G . a n is the optimum sequential decision required. +) and ( max. It is evident that θ 1 satisfies the optimum-preserving law and θ 2 satisfies the associative and distributive laws over θ1 . Bellman’s Principle of Optimality holds in these optimization problems. the following is true: if the zero element of W. associative. Theorem 7. G. and traveling-salesman problems (Dreyfus and Law. where 0 ≤ i < j ≤ n. . The following propositions and theorem reveal the relation between the fundamental equation approach and the generalized principle of optimality. . It can be shown that ( min.2 [Wu and Wu (1986)]. . the system of interest is ( max. θ 2 ) satisfies the optimum-preserving law means that the generalized principle of optimality holds. Let W is the set of all nonnegative real numbers. according to Theorem 7.1. cancellation. In the equipmentreplacement models and the resource allocation problems (Dreyfus and Law. . θ2 ) does not imply that it satisfies the optimum-preserving law. . and distributive laws (over “min” or “max”). The recursiveness of the system (W. 1977). G be a subset of real numbers. θ1 . (W. a j must be optimum with regard to the initial state a i and the terminal state aj . as defined in Section 7.θ2 ) is recursive. and a 0 .1 says that in ( min. The system (W. +). and θ 2 is the minimum.3. θ1. θ2 ) does not satisfy the optimum-preserving law. and That is. θ 1 . the system considered is ( min. At the same time. . Proposition 7.2.1 [Wu and Wu (1986)]. ai + 1 .

That is. and the generalized principle of optimality holds.5. 7. (Fig. in this special system the fundamental equation will give a global optimum solution.3. Theorem 7. In the system (W. A. 7. from the fundamental equation [Eq. Proof: Again. Optimum-preserving law does not imply recursiveness. it can be computed that ƒ 〈3〉 (A.D )-recursiveness. The result is a consequence of Theorems 7.8. Since { the system does not possess (g .1 [Wu and Wu (1986)]. A.3. θ2 ). the system possesses a (g . θ 1. However. and distributive laws over θ1 both the optimum-preserving law and the recursive law hold.160 Chapter 7 Figure 7. of which all subpolicies are optimum.3).1 and 7.3) satisfies g 〈3〉 (A. (7. Consider the system (W. θ1.1.3.3. θ 2 ) with θ1 satisfying the optimum-preserving law and θ2 satisfying the cancellation.G . it suffices to show the proposition by constructing a counterexample.D ) = 4 and A → B 1 → C 1 → D and is the optimum policy.D ) = 5 with A → B2 → C2 → D its corresponding policy.G.9)]. associative. That is. where W = the set of all nonnegative real numbers subminimum of X if  X > 1 θ1 (X ) = if X = {w} w θ2 = + (regular addition) Then it can be seen that the element g ∈ W ↑ G 2 (Fig. .D ) optimum-preserving property.

Denardo. Conclusion This chapter generalizes the fundamental idea of dynamic programming to the widest sense available today in the research of systems science and pansystems analysis. In these works. 1967. the method of classical dynamic programming may not apply. especially those with nonscalar-valued performance criteria. where for any nonempty sets W and G each element g ∈ W ↑ G 2 is called a generalized weighted network with weights in W and nodes in G. However. there have appeared some studies and applications of vector-valued dynamic programming (Brown and Strauch.9. 1965. the main point of dynamic programming is to use the fundamental equations to compute a desired optimal solution. in many practical optimization problems. +) their underlying structure of systems is much more complicated than ( and ( max. the first approach generalized the ideas of Bellman’s Principle of Optimality and of the fundamental equations in dynamic programming to generalized weighted networks. These facts are helpful in showing the universal significance of the approaches presented in this chapter. The second approach will surely open up a new era of research to study the idea of optimization in the theory of general systems. especially with those out of the ordinary. Furukawa. It may be expected that in dealing with optimization problems of the most general sense. tensors. In this chapter. +). Practically. . In those cases. 1980. and so on. 1983). Henig. Actually. such as vectors. Bellman’s Principle of Optimality in the dynamic vector-valued models has also been discussed and similar results obtained. matrices. In addition to the set of real numbers being used as the criteria set.Bellman’s Principle of Optimality and Its Generalizations 161 7. Bellman’s Principle of Optimality has been generalized to the Generalized Principle of Optimality and the Operation Epitome Principle. min. the Generalized Principle of Optimality and the Operation Epitome Principle will play important roles. For example. generalized dynamic programming assume all other possible sets as criteria sets. a criterion has been established to decide whether the approach of fundamental equations can still be employed.

1. Then the structure of mathematics is written as a formal system in the third section. where we are looking for our “brothers. The reason is that a theory is not considered a real scientific theory if it has not been written in the language and symbols of mathematics. some impacts of the puzzle are mentioned. painting. applications of mathematics and the origin of the universe are discussed. no matter where we direct our vision — whether to the depths of outer space. This new understanding is used to model and to analyze the states of materials and some epistemological problems in the fourth and fifth sections.” to slides beneath microscopes. Indeed. The structure of mathematics is studied in the second section. mathematics is compared with music. problems concerning the material structure of human thought and the origin of the universe are posed. 8.CHAPTER 8 Unreasonable Effectiveness of Mathematics: A New Tour It is well known that the spectacular success of modern science and technology is due in large measure to the use of mathematics in the establishment and analysis of models for the various phenomena of interest. More specifically. The application of mathematics brings life to all scientific theories. In this chapter. What Is Mathematics? Mathematics is not only a rigorously developed scientific theory. and poetry. many great thinkers had thought of this question. Mickens (1990) asked why this is the case. It is mathematics that makes us better understand the phenomena under consideration. 1988). In the concluding section. some problems related to the structure of mathematics. dealing with real and practical problems. In the sixth section a vase puzzle is given to show that mathematical modeling can lead to contradictory conclusions. Based on this fact. but it is also the fashion center of modern science. In fact. or to the realm 163 . in the first section. where we are trying to isolate the fundamental bricks of the world. when studying natural systems (Graniner. since such a theory does not have the capacity to predict the future.

but is it not the same with all the noblest arts? Organically arranged colors show us the beautiful landscape of nature so that we more ardently love the land on which we were brought up. but supreme beauty — a beauty cold and austere. like that of sculpture. investigate and predict the world. which is the touchstone of the highest excellence. the generations have gradually created an ordered cosmos.164 Chapter 8 of imagination where much human ignorance is comforted. and. History and reality have repeatedly showed that mathematical foliage is always young and flourishing. a long second-best. First. rightly viewed. possesses not only truth. Remote from human passions. at least. where pure thought can dwell as in its natural home. is to be found in mathematics as surely as in poetry. yet sublimely pure. Philosopher Bertrand Russell (Moritz. a perpetual compromise between the real and the possible. without the gorgeous trappings of painting and music. Mathematics is not like any other theories in history. without appeal to any part of our weaker nature. which have faded as time goes on and for which new foundations and new fashions have had to be designed by coming generations in order to study new problems. it is true. space. and sometimes comfort . What is best in mathematics deserves not merely to be learned as a task. Real life is. The true spirit of delight. and capable of a true perfection such as only the greatest art can show. has it not the esthetic character although the senses take no part in it? Only the privileged few are called to enjoy it fully. adepts find in mathematics delights analogous to those that painting and music give. and brought again and again before the mind with ever-renewed encouragement. and time. of our nobler impulses can escape from the dreary exile of the natural world. and the joy they thus experience. many unrecognized truths in mathematics have astonished truth pursuers by its sagacity. Although mathematical abstraction has frightened many people. let us see how some of the greatest thinkers in history have enjoyed the beauty of mathematics. the exaltation. but the world of pure reason knows no compromise. no barrier to the creative activity embodying in splendid edifices the passionate aspiration after the perfect from which all great work springs. but to be assimilated as a part of daily thought. the sense of being more than man. mathematics is a splendid abstractionism with the capacity to describe. no practical limitations. above all. they are amazed when a new discovery discloses for them an unlooked for perspective. Animated combinations of musical notes sometimes bring us to somewhere between the sky with rolling black clouds and the ocean with roaring tides. many diligent plowers of the great scientific garden have cultivated today’s magnificent mathematical foliage. In fact. 1942): Mathematics. remote even from the pitiful facts of nature. to most men. etc. and where one. 1942): it [mathematics] ought to incite the philosopher to search into the notions of number. All through history. Mathematician Henri Poincare (Moritz. They admire the delicate harmony of number and forms.. we can always find the track and language of mathematics.

We know that there are the same amount of books and apples. for example. every one will agree that the symbol “2” displays a figure of nature and insinuates the marrow of human thought. and musical notes can express fantasies from the depths of the human mind. 2n and 2 n + 1 can be put together to form a pair. That is because any even natural number can be denoted by 2n. was formally named in the second decade of this century. and any odd natural number can be described by 2n + 1. it can be shown that the totality of even natural numbers and the totality of odd natural numbers have the same number of elements. for example. The objects. there are two apples and three books. Notice that the symbol “2” manifests the same inherent law in different things — quantity. You might ask: Because mathematics. With the help of this kind of reasoning and abstraction. From this fact an ordering relation between the concepts of “2” and “3” can be expected. for each natural number n. it has been used to explore some problems man has been studying since the beginning of history. However. How can colors be used to display atomic structures? How can musical notes be used to imitate the association of stars in the universe? With mathematical symbols and abstract logic. In the previous example. How does mathematics depict the world for us? Mathematics has neither color nor sound.Unreasonable Effectiveness of Mathematics: A New Tour 165 us in a fragrant bouquet of flowers with singing birds. By using intuitive pictures and logical deduction. book and apple. is so universally powerful. one of the theories concerning these problems. What is “2”? Consider two books and two apples. Colors can describe figures of nature. because by pairing a book and an apple together. It is still not known whether the theory can eventually depict and answer the aforementioned problems. Systems theory. Thus. we get two pairs. there must be a book left. are completely different.” just as human investigation about nature goes from daily life into the microcosm and outer space. can it be used to study human society. After forming two pairs. It is because mathematics is the combination of the marrow of human thought and aesthetics that it can be developed generation after generation. the meaning of the concept “2” is the same. and the theoretical foundation began to be set up in the sixth decade. according to the previous description. research along this line began not too long ago. Let us see. each of which contains exactly one apple and one book. and communication and transportation between cities? The answer to this question is “yes. the relationship between the stars and the attraction between an atomic nucleus and its electrons are written in detail in many books. human relations. the domain of human thought is enlarged from “finitude” into “infinitude.” Even though the motivation for studying these kinds of problems appeared long before Aristotle’s time. Nevertheless. Otherwise. such as: (1) Does there exist eternal truth in the universe? (2) Does the universe consist of fundamental particles? Systems theorists have eliminated . where n is a natural number. the concept of numbers. but with its special methods — numbers and forms — mathematics has been showing us the structure of the surrounding world.

1 Music and poetry used to quicken you The mathematics. sir. The following two poems by Shakespeare (Moritz. . the mathematical world has been constructed on a few basic names. with its awesome blazing scenes of color. the moon. Kunen. Then K is defined by induction in the following manner. she is not ignorant. The logical language K used in mathematics can be defined as follows. study what you most affect 2 I do present you with a man of mine Cunning in music and in mathematics To instruct her fully in those sciences Whereof. and the logical language. The Construction of Mathematics Analogous to paintings. 1942) will give mathematics a better shape in your mind.2. a thirst for knowledge of the future. nor to music. I know. For instance. where is no pleasure ta’en: — In brief. the combination of vital basic particles. which is infiltrated with exclamations and admirations about nature. and studying each of them — and have begun to utilize systems methods to study problems with many cause—effect chains. Even though gentle breezes have never carried any soft mathematical songs. for details see (Kuratowski and Mostowski. which possesses intoxicatingly melodious sounds. 1976. 8. man has firmly believed that mathematics shows truth since the first day it appeared. you will be attracted by the delicately arranged symbols to somewhere into mathematical depths. axioms. where a formula is an expression (or sentence) which may contain (free) variables. Mathematics is analogous neither to painting. K is a class of formulas. and social systems of human beings are examples of such problems. and the pursuit of ideals. and the earth.166 Chapter 8 the traditional research method — dividing the object under consideration into basic parts and processes. in which the most beautiful symphonies in the world can be written with a few basic notes. and to music. which are based on a few colors. the famous three-body problem of the sun. It resembles poetry. and the metaphysics Fall to them as you find your stomach serves you No profit grows. If you have time to taste the flavor of mathematics meticulously. 1980).

on which the whole classical mathematics can be built and which is called the ZFC axiom system. (b) If φ and ψ belong to the class K . y ) holds for some x ∈ A. ( φ and ψ ). The set S is denoted by Axiom 4. and (c). A well-known set of axioms. elements y for which the condition φ (x. x = y as well as all expressions differing from them by a choice of variables. (d) Every element of K arises by a finite application of rules (a). Axiom 7. then there exists an element Y ∈ A such that Y consists exactly of all the elements of X and the set X itself. Axiom 6. then they are equal or identical. (Power sets) For every set A there exists a set of sets. Axiom 2. . there exists a set B which has exactly one element in common with each set belonging to A. contains the following axioms. then ψ ). which are the only undefined primitives in the mathematics developed in this section. then for every set A there exists a set B which contains those. Axiom 1. (Unions) Let A be a set of sets. (Replacement for the formula φ ) If for every x there exists exactly one y such that φ (x . There exists a set S such that x is an element of S if and only if x is an element of some set belonging to A. and only those. denoted by p(A). where the concept of “set” and the relation ∈ of membership are primitive notations. then the expressions (for any v . (Choice) For every set A of disjoint nonempty sets. if X ∈ A.Unreasonable Effectiveness of Mathematics: A New Tour 167 (a) Expressions of the forms given below belong to K : x is a set. (Empty set) There exists a set such that no x is an element of Axiom 3. (Infinity) There exists a set of sets A satisfying the conditions: ∈ A. y) holds. then so do the expressions ( φ or ψ ). x ∈ y. φ ) and (there exists v such that φ ) belong to K. and not φ ). (b). which contains exactly all the subsets of A. (Extensionality) If the sets A and B have the same elements. (if φ . (c) If φ belongs to K and v is any variable. The formulas in (a) are called atomic formulas. Axiom 5. ( φ and ψ are equivalent).

b}. there exists a set X which contains only a and b. (Pairs) For arbitrary a and b. Axiom 2. 0 is not the successor of any number. The entire arithmetic of natural numbers can be derived from the Peano axiom system. then the successor of n also has the property P. the set X is denoted by X = {a . Axiom P5 embodies the principle of mathematical induction and illustrates in a very obvious manner the enforcement of a mathematical “truth” by stipulation. of sets as follows: From Axioms 2 and 4. 0 is a number. (Regularity) If A is a nonempty set of sets. In the construction of the sequence we understand the symbol 0 in P1 as the notations of “number” in P2 as any element in the sequence and of “successor” in P2 as the power set of the number. we can construct a sequence Now it can be readily checked that the sequence P 1. then every number has the property P. The successor of any number is a number. P3. Axiom 6. satisfies Peano’s axioms: P5. and (b) whenever a number n has the property P. then there exists a set X such that X ∈ A and X ∩ A = where X ∩ A means the common part of the sets X and A. (Subsets for the formula ø) For any set A there exists a set which contains the elements of A satisfying the formula ø and which contains no other elements.168 Chapter 8 Axiom 8. If P is a property such that (a) 0 has the property P. No two numbers have the same successor. The following axioms can be deduced from Axioms 1–8. P2. The construction of elementary arithmetic on Peano’s axiom system begins with . P4.

In terms of addition and multiplication. m) is the ordered pair of the natural numbers n and m. In fact. This incompleteness of the definitions of subtraction and division suggests an enlargement of the number system by introducing negative and rational numbers. the construction of the sequence Because of P3 (in combination with P5). 7 – 10 and 7 ÷ 10 are undefined.Unreasonable Effectiveness of Mathematics: A New Tour 169 the definition of the various natural numbers — that is. its true propositions flow not merely from the definition of the concepts involved but also from the axioms that govern these various concepts. Multiplication of natural numbers may be defined by the following recursive definition. a definition of addition can be established. m ). d) belong to that for any two ordered pairs (a. in contradistinction to addition and multiplication. which is defined such Consider a subset Z of the power set by d) ( a .e. and by P4. Thus. Z is the desired set on which the definition of subtraction of integers can be defined for each pair of elements from Z. then we may now say that the propositions of the arithmetic of the natural numbers are true by virtue of the stipulations which have been laid down initially for the arithmetic concepts. is the where Consider the Cartesian product collection of all natural numbers. the element A can be denoted . m) one element A ∈ Z containing the pair ( n. which expresses in rigorous form the idea that a product nk of two natural numbers may be considered as the sum of k terms. the inverse operations (i. (c.. Then it can be seen that each is contained in exactly one element belonging to Z and no ordered pair in two elements in Z have an element in common. subtraction and division are not defined for every pair of numbers. each of which equals n (8. subtraction and division) can then be defined.. it does not lead back to 0 either. and (n. which expresses in a precise form the idea that the addition of any natural number to some given number may be considered as a repeated addition of 1. and each element in Z is there exists exactly called an integer. for example. where the set Z is called a set of all integers. As the next step.1) The two stipulations of this inductive definition completely determine the sum of any two natural numbers. defined as before. b). for each ordered pair (n. However they cannot always be performed.e. the latter operation is readily expressible by means of the successor relation (that is. none of the elements in will be led back to one of the numbers previously defined.2) Within the Peano system of arithmetic. b ) and ( c. Now. the same set A in Z if and only if a + d = b + c. i. which is a set by Axiom 5. This definition of addition goes as follows: (8. the power set relation). If we call the axioms and definitions of an axiomatized theory the “stipulations” concerning the concepts of that theory.

. It is important to note that the subset Z+ = ∈ Z : there exists a number x ∈ such that b + x = a} of Z now serves as the set of natural numbers. finally the huge system of mathematics as here delimited rests on the narrow basis of the ZFC axioms: Every concept of mathematics can be defined by means of the two primitives in ZFC. A well-known result discovered by Gödel shows that the afore-described mathematics is an incomplete theory in the following sense. because Z + satisfies axioms P1–P5. without posing a single new axiom. On the basis thus obtained.3) (8. of derivatives and integral can be introduced. The much broader system thus obtained is still incomplete in the sense that not every number in it has a square root. a remarkable achievement in systematizing the content of mathematics and clarifying the foundations of its validity. in the sense just characterized. affect the results we have outlined because the most unreasonably effective part of mathematics in applications is a substructure of the mathematics we have constructed. without introducing any new postulates or assumptions. the various arithmetical and algebraic operations can be defined for the numbers of the new system. by means of nothing more than the principles of formal logic. and more generally. It is another remarkable fact that the rational numbers can be obtained from the ZFC primitives by the honest toil of constructing explicit definitions for them. and subtraction can be defined as (8. in most cases. multiplication. and every proposition of mathematics can be deduced from the ZFC axioms enriched by the definitions of the nonprimitive terms. The various arithmetical operations can then be defined with reference to these new types of numbers. Again.170 Chapter 8 by follows: Then. Similarly. cube root. the concepts of function. other propositions can be expressed in pure ZFC language which are true but which cannot be derived from the ZFC axioms. however. Even though all those propositions in classical mathematics can indeed be derived. addition. rational numbers are defined as classes of ordered pairs of integers from Z.. Remark. These deductions can be carried out. . This suggests further expansions of the number system by introducing real and complex numbers. this enormous extension can be affected by only definitions. and the validity of all the arithmetical laws governing these operations can be proved from nothing more than Peano’s axioms and the definitions of various arithmetical concepts involved. .5) for any elements and in Z. not every algebraic equation whose coefficients are all numbers of the system has a solution in the system. This fact does not. thus. of limit.4) (8. from ZFC. . and the familiar theorems pertaining to these concepts can be proved.

For details see (von Bertalanffy. it follows that X is a set. respectively. such that r is a subset of Mn . otherwise the natural number 26 is not a set (from the discussion in the previous section). where R is a set of some relations defined on the set M. then any formal language can be described as a system. From the finiteness of the collection X and Axiom 7. Claim. i. called the length of the relation r. Each element in M is called an object of S . and M and R are called the object set and the relation set of S. This kind of investigation gives us no information about the coordination of parts and processes. In this definition. so instead of the collection X we can use the set 26. a system consists of a set of objects and a n =n collection of relations between the objects. Assume that the length of the empty relation is 0.. Proof: Assume that L is a formal language. = 0.3. Thus. suppose that L is English which does not contain sentences with grammatical mistakes. each relation r ∈ R is defined as follows: There exists an ordinal number n = n(r). where an English grammar book B is chosen and rules in B are used as the measure to see if an English sentence is in L or not. Mathematically speaking. the chief task of modern science should be a systematic study of the world. Customary investigation of the isolated parts and processes cannot provide a complete explanation of the vital phenomena. a finite sequence of letters. M is a set which contains all words in L. for convenience. that is. Mathematics from the Viewpoint of Systems The concept of systems was introduced formally by von Bertalanffy in the 1920s according to the following understanding about the world: The world we live in is not a pile of uncountably many isolated “parts”.e. Then M is the union Hence. . 1934). R) of sets. where M n indicates the Cartesian product of n copies of M. If by a formal language we mean a language which does not contain sentences with grammatical mistakes. and any practical problem and natural phenomenon cannot be described perfectly by only one cause–effect chain.Unreasonable Effectiveness of Mathematics: A New Tour 171 8. Each word in L consists of a finite combination of letters. Let X be the collection of all 26 letters in L. S is said to be a system (Lin. Let M be the totality of all finite sequences of elements from X. The basic character of the world is its organization and the connection between the interior and exterior of different things. In ordinary language. 1987) if and only if S is an ordered pair (M.

then the ordered pair ( M. Question 8. we still have not found any method outside ZFC which can be used to show that the system ( M. a theorem of Gödel shows that it is impossible to show whether the systems description (M. Then K must be a finite collection. n} is a finite index set.3. Concerning this. i. 1989c). T K) is a systems description of mathematics. T K) is . Question 8. (M.1. Here. any mathematical truth is a (M. i. and definitions of some nonprimitive terms. If Q be the totality of finite sequences of elements from M.3. e. then each relation r ∈ R can be understood as an S -truth. Each element in Q is called a sentence of L. From this discussion about mathematical truths the following question is natural. If there existed two mathematical statements derivable from ZFC axioms with contradictory meanings..e. Let (M.. any two statements derivable from ZFC will not have contradictory meanings) or not in ZFC..1. If each statement in K is assumed to be true.2. as the relation set? The following natural epistemological problem was asked in (Lin. From Klir’s (1985) definition of systems — that a system is what is distinguished as a system — and Claim 1. K ) can be seen as a systems description of the formal language L. Therefore.172 Chapter 8 Each sentence in L consists of a finite combination of some elements in M. T K)-truth.K) be the systems representation of the formal language L in the preceding proof. On the other hand. Then (M.. Let K ⊂ Q be the collection of all sentences in the grammar book B. Write M as a union of finitely many subsets: where J = {0.2. Elements in M (0) are called nouns... and it follows from Axiom 7 that K is a set. T K) of mathematics is consistent (that is. T K)-truths? Generally. what is the meaning of a system with contradictory relations.1 and 8.3. Questions 8.e.. a finite sequence of elements from M.3.g. i. Q is a set.R) is given. then Thus.. elements in M(2) adjectives.e. it follows that if a system S = (M. etc..2 cannot be answered. elements in M(1) verbs. the relation r is true among the objects in the set M. it follows that generally. would they still be ( M . K ) constitutes the English L. How can we know whether there exist contradictory relations in a given system? From Klir’s definition of systems. the principles of formal logic. and T the collection of Axioms 1–6’ in the previous section. it is derivable from the ZFC axioms.

mathematics based on ZFC) in which we do not know whether there exist propositions with contradictory meanings. As a consequence of our discussion. it assumes very nearly the shape of the mathematical curve known as the catenary. For example. satisfy all the theorems deduced from them. let us understand by 0 the origin of a half-line. 1980). 1976. of the nonprimitive mathematical terms) provided that the ZFC axioms are true. At the same time. (Note that this partial structure of mathematics constitutes the theoretical foundation for almost all successful applications of mathematics. . we cannot. six.. However. This means that there are systems (for example. we must go back to Axioms 1–6’.” into true statements. and by a number any point which is either the origin or can be reached from it by a finite succession of steps each of which leads from one point to its successor.. problems. All we can assert so far is that any specific interpretation of the primitives which satisfies the axioms —i. “set” and the relation of membership “ ∈ . this fact implies that perhaps not every system is consistent or that not every system has no contradictory relations. and therefore. Question 8.3. . strictly speaking. This description serves to distinguish it from flowers with three. It can then be readily seen that all Peano’s axioms as well as the ensuing theorems turn into true propositions. Why is mathematics so “unreasonably effective” when applied to the analysis of natural systems? In order to discuss this question. answer) the following problem: Suppose someone holds his 10-inch watch chain by its ends with his fingers at the same height and are 4 inches apart. Therefore. But the partial structure of mathematics developed on the basis of Peano’s axioms has several — indeed infinitely many — interpretations which will turn Peano’s axioms. Kunen. and predicting practical phenomena. refer to the ZFC axioms as propositions which are true or false. . petals.” which have not been assigned any specific meanings. respectively. That means that in practice a subsystem of ( M . For detailed discussion. a hibiscus flower has five petals. which contain the primitives “0. How far below his fingers will the chain dip? Now. the following amazing question can be asked (this question was not originally posed by me). for they contain free primitive terms.Unreasonable Effectiveness of Mathematics: A New Tour 173 inconsistent. say. The application of mathematics has showed us that mathematics is extremely effective in describing. T ∪ K ) can always be found to match a situation under consideration. turns them into true statements—will also satisfy all theorems deduced from them.” “number. by the successor of a point on that half-line the point 1 inch behind it. see (Kuratowski and Mostowski.) For example. counting from the origin. this equation can be used to predict (or. When a watch chain is suspended from its ends. The mathematical word “five” provides a certain description of the flower. . and future events.3. solving.e.” and “successor. The equation of the catenary is where a is a constant. the whole structure of mathematics might be said to be true by virtue of mere definitions (namely. four. at this juncture.

and and t ∈ W satisfying . because each molecule in X must belong to one of the unit cells. if X is crystalline. 8. We will give two examples to show how abstract mathematical structures developed on the basis of the ZFC axioms but outside Peano’s axiom system can be used to describe and study problems in materials science and epistemology. If X is crystalline. in addition. Given a linearly ordered set W .4. A Description of the State of Materials An arbitrarily chosen experimental material X can be described as a system S x = ( Mx . for any proper for the set M x such that for any subset M* of M x . Rx ).s. where M x is the set of all molecules in the material and R x contains the relations between the molecules in Mx . so we have the following. it can be shown that every progression of elements of any kind provides a true interpretation of Peano’s axiom system. we feel that mathematics is so unreasonably effective as applied to the study of natural problems. there exists a relation ƒ ∈ R x such that ƒM* ≠ ƒ. because of this fact. we understand something more about the nature. a family of mappings is given such that for any r. If.1. a time system S over W is a function from W into a family of systems. where each system S w is called the state of the time system S at the moment w.4.R w ). and at the same time. The experimental material X is crystalline if for the system S x = (Mx . for any unit cell Y in X define MY as the set of all molecules contained in Y.R x ) as follows. and there exists a cover A and B i n there exists a one-to-one correspondence h : A → B satisfying where ƒY indicates the restriction of the relation ƒ on the subset for any ordinal Y and number β and any For example. This example illustrates that mathematics permits many different interpretations.4. for each w ∈ W. Definition 8. then all cells in X have the same molecular spatial structure.1. That means that the molecules in X are aligned spatially in a periodic order. Then x = { MY : Y is a unit cell in X} is a cover of M x . and all unit cells look the same. which describe the spatial structure in which the molecules in the material are arranged.174 Chapter 8 although the interpretation given to the primitives is certainly not any of those given in the previous section. in everyday life as well as in the investigation of laws of the nature and from each different interpretation. say S (w) = S w = (M w . More generally. It therefore follows that the main characteristics of crystalline materials are described by Definition 8.

as soon as the nuclei grow large enough.Unreasonable Effectiveness of Mathematics: A New Tour 175 where is the identity mapping on Mt . + ∞ ) and R p = [0. their energy will be lower as more and more molecules join them.1.4. This kind of locally ordered region is called a nucleus. any chemical reaction between the molecules during the process will be neglected.4. Then at the state w. However. where w = ( t. We establish a systems theory model of crystallization process of the the material X. Let be the set of all real numbers. so there would be a tendency for the molecules to get together to form a short-range order.4.. In the state of time t. For details. In the previous mathematical model. the activity of the molecules decreases. The crystallization process of X is the process that X changes from an amorphous state to an ordered state or a crystalline state.1 is “yes. and pressure P. 1987). when temperature is lower than the melting temperature of the material. Although at certain temperature and pressure values. respectively. F w ). the relation set. Define an “order” relation “<” on the Cartesian product as follows: For any if and only if either (i) (ii) and y 1 > y 2 . see (Ma and Lin.1.e. then S is called a linked time system Each is called a linkage mapping of the linked and denoted by time system.1 when the material X is actually in a melting state. RT (= ) indicate temperature and Rt = [0.e. which means that the attraction or interaction between molecules increases. pressure imposed on X goes up. which is caused by the fluctuation of local energy. temperature T. consisting of all relations describing spatial structure of the molecules in X. i. or (ii) and x 1 < x 2 . many nuclei can be formed at the same time they will probably disintegrate because the ratio of the surface and the volume of each nucleus is still too great and the surface energy is too high. Question 8. Then it can be shown that < is a linear order. The process is divided into two steps.4. where M is the set of all molecules in X so that a linked time system where id M : M → M is the identity mapping on M. The first step is called nucleation. The second step is called growth. X can be described as a system (M. It is assumed that the size of molecules in X will not change as temperature and pressure change. The molecules in the nearby region would prefer to join those nuclei so that they will grow larger and larger. P ).” This assumption is based on . That means that the state of S satisfies the property in Definition 8. does there exist a relation ƒ ∈ R x satisfying the property in Definition 8. T.. is Fw . where is the systems representation of X? We assume that the answer to Question 8. If X is going to crystallize as time goes on. there then is a state such that X is crystalline. i. + ∞) (the interval of real numbers from 0 to positive infinity) indicate time and pressure. and temperature goes down.

and for any i. The basic structure of the material determines all the physical properties that X has. 1987). or gas). 1989c) the following epistemological problems were discussed: (i) the feasibility of the definition of the theory “science of science. is said to be (3) A sequence a chain of object systems of a given system S if S 0 is an i 0 th-level object system of S. With this model many phenomena about materials can be explained naturally. see (Lin and Qiu. the following results can be shown: (1) There is no system whose object set consists of all systems.” (ii) the existence of fundamental particles in the world. . if there exist systems Si = (M i . The background in which the . there exists a natural number ij > 0 such that S j is an ij th-level object system of Si . and every group is “topologically” isomorphic to the basic structure of X. 1987). there exist many pairwise disjoint groups of molecules in X such that each molecule in X is contained in exactly one of the groups. For details. Generalizing the model in (Lin and Qiu. 2 . Since then more and more scientists have been involved in its research. liquid. and (iii) the existence of absolute truths. n. i =1. n – 1. each chain of object systems of S must be finite. The theory of science of science appeared in the 1930s. i = 1 . because the relation ƒ will be observable under certain ideal conditions. where n is an ordinal number.176 Chapter 8 (i) A view of dialectical materialism—the internal movement of contradictions dominates the development of the matter under consideration (ii) The beauty of the systems theory model described earlier Under this assumption. The spatial geometric relation (called the structure of Y ) of the molecules in Y is termed the basic structure of X. Then for any system S. of systems. it can be seen that under certain ideal conditions any material may crystallize. . 1987) with examples. . if i < j. The meaning of topologically isomorphic was explained in (Lin and Qiu. .2. Some Epistemological Problems In the ZFC axiom system. there is no nth-level object system Sn of S such that S n = S for any natural number n > 0. Suppose Y is a cell in the crystalline material X. . .5.R 0 ).Ri ).j ∈ n. Then for any fixed system S. In (Lin. . 8. . the following is our model for the arbitrarily fixed material X: no matter what state X is in (solid. is an nth-level object system of a system (2) A system S 0 = (M0 . such that S i ∈ Mi – 1 .

the development of different aspects of science and technology have sped up. where a fundamental particle is a particle which cannot be divided into smaller particles? See (Lin.6. we can see that the theory of science of science cannot exist. Does this imply that the world consists of fundamental particles. Around this puzzle. the development of science will be sped up greatly. see (Lin. philosophy. roughly speaking. named the vase puzzle. 1989c). If the answer to the aforementioned question is “yes. For detailed discussion. including in general. Some unsettled problems are left open. there are many scientific achievements obtained independently by many different scientists. Some scientists think that the research object of the theory of science of science is the following: The theory of science of science is such a theory that instead of any real matter it studies science as a whole. Thus. Combining Claim 1 and result (2). is used to show that mathematical modelings can lead to contradictory inferences. the history and the present situation of each discipline. and the whole developing tendency of science. From Klir’s (1985) definition of systems and result (1). I think that for any given truth. The development of science shows a tendency to divide science into more and more disciplines while synthesizing results and methods in different fields to get new results and understanding about the world. it follows that we cannot consider a relation which is true in a system that contains all systems as its objects. Does this imply that there is no universal truth? I believe so.Unreasonable Effectiveness of Mathematics: A New Tour 177 theory appeared is that. the theory has to contain a study of itself. a natural problem arose: Can science be studied as a social phenomenon so that man can develop scientific research with purposes and improve its recognition of the natural world? This problem is very important to administrators of scientific research. the relationship between disciplines. . an environmental system is given in which the truth is a true relation.” then it partially means that we can save at least some valuable scientific labor from doing the same things by arranging them to do some pretargeted research. methodology. In this way. The Vase Puzzle and Its Contradictory Mathematical Modelings A conceptual problem. concerning proof methodology in mathematics and understanding mathematical induction arise. since human society entered the twentieth century. because so far each scientific achievement is obtained by either individual scientists or small groups of scientists. 1989c) for details. epistemology. Result (3) says that any system is built upon objects which are no longer systems. Hence. Because. and modeling time and space. some questions. 8.

. so that each piece has at most one label on it. Finally. Third. If a person were concerned only with problems in real life.1. the vase need not be infinitely large — actually any size will do. which is theoretically the collection of all the products that have been and will be produced from the assembly line. Specifically to study quality. Unfortunately. . Halmos. we draw a random sample of the products and use sample statistics to make inferences on the continuously expanding population. Second. different meanings are given to the concept. number-labeling can be done according to the steps in the puzzle.6. 1977.2. After the recursive procedure is finished. The vase puzzle. For convenience. In this section a conceptual problem is constructed to show that mathematical modeling does not always work. The Vase Puzzle and its Modelings The concept of infinity has been bothering mankind for centuries (Moore. into the vase. 199l). methodology. etc. The pieces of paper are labeled by natural numbers 1. In mathematics. Even though the phenomenon has been studied for some time. mathematical proofs. labeled from 10n – 9 through 10n into the vase. The following recursive procedure is performed: Step 1: Put the pieces of paper. we write the mathematical induction procedure. Jech. . Many of the greatest thinkers in history (Moore.1 shows how an infinite number of pieces of paper can be obtained. it is no longer the case. 8. For example. different models can give contradictory results. Suppose a vase and an infinite number of pieces of paper are available. 1991). he or she would say that the concept of infinity had no interest. the recursive procedure can be finished within any chosen period of time. First. .178 Chapter 8 8. how many pieces of paper are left in the vase ? Some comments are necessary. have been convinced that infinity is something we will never be able to understand because there are so many seemingly unsolvable puzzles about it. the total area of the infinite number of pieces of paper can also be any chosen size. . Fig. where n is any natural number 1. the study of quality control of the products of an automatic assembly line needs the concept. .3. Step n: Put the pieces of paper.2. For example. which guarantees that the vase puzzle is well defined.1. .6. (Davis. 1960. 8. This astonishing phenomenon forces us to rethink some principles and long-lasting questions in mathematical induction. labeled from 1 through 10.1. epistemology. then remove the piece labeled n. we often treat the ever-expanding population as an infinite population. we still hope that this presentation will cast the first brick in order to attract many beautiful jades. In other words.1. .6.3. then remove the piece labeled 1. To make the inferences more convincing. Question 8. philosophy. 1973).

To fit the recursive procedure in the vase puzzle into the format of mathematical induction.6) Equation (8. define the set M n of pieces of paper left in the vase as follows: M n = {x: x has a label between n and 10 n + 1 exclusively} (8.7) really tells how many pieces of paper are left in the vase after step n.3. Therefore. then the subset S equals the set N.2. then n+ 1 ∈ S .3 . A set-theoretical modeling. where n = 1. 1990). 8. An elementary modeling.Unreasonable Effectiveness of Mathematics: A New Tour 179 Figure 8. derived in the elementary modeling. Inductive step: If n ∈ S. 8. Obtain as many pieces of paper as needed out of a chosen total area of pieces of paper. Let S be a subset of the set N of all natural numbers.7) (8. . based upon mathematical induction. . For each natural number n.2. the answer to the question of the vase puzzle is “no paper is left in the vase. The answer is that infinitely many pieces of paper are left in the vase.6.. .6. If it can be shown that • • Initial step: 1 ∈ S.4. 8. by ƒ(n) = 9n (8.” This contradicts the conclusion. if the recursive procedure can be finished. . Based upon this modeling.1. let us define a function.8) . Mathematical induction.6. To answer the question of the vase puzzle.1. let us define a subset S of natural numbers as follows: S = {n : step n can be done and all pieces of paper with labels 1 through 10n have been in the vase at least once} It is left to the reader to verify that S is the same as N. See (Smith et al.1. the number of pieces of paper left in the vase should equal the limit of ƒ( n ) as n approaches ∞ .1.

for the following reasons: (1) The example is easy to describe. and the person in the second room moves to the third. The theory possesses a beauty analogous to painting.180 Chapter 8 Then after the recursive procedure is finished. at 3/4 minute it is at point 3/4. and so on ad infinitum. 1991)]. The hotel. and the harmony between numbers and figures [for details. (Moore. some interpretations can result in contradictory mathematical models. However. then the entire hotel will be only twice the height of the ground floor. each occupied at a particular time. 1990). if x is a piece of paper left in the vase. music. and the other concerns the concept of time. 1990d)]. In this section we chose the vase puzzle to emphasize the existence of the impassable chasm. One concerns the structure of space. see (Lin. need only occupy a finite amount of space. Hilbert used to present this paradox in his lectures.9) That is. what would a person looking at the hotel from above see? Paradox [Paradox of a Moving Particle]. the ZFC axioms. (2) The two contradictory modelings are readily comprehensible by readers with little background in mathematics. It is the assignment of a mathematical meaning to each object that causes problems. where will the particle be at 1 minute? . the set of pieces of paper left in the vase equals the intersection (8. at (2 n – 1)/2 n minute. Paradox [Paradox of the Hotel. As described in (Lin et al. 1991)]. while in applied mathematics each object of interest is always first given some mathematical meaning. and Port (1990) pointed out theoretically that there must be an impassable chasm between pure mathematics and applied mathematics. because different interpretations can be given to the same object. Ma. the particle is at point (2 n – 1 )/2 n . Suppose a hotel has infinitely many rooms. x then has a label greater than all natural numbers.6. and poetry. this raises the following question about the nature of the concept of space. Question 8. Suppose a particle moves from point 0 to point 1 on the real number line. for if the person in the first room moves to the second. This contradicts the assumption that each piece of paper put into the vase has a natural number label. Then a newcomer can be accommodated without anybody having to move out. If the roof of the hotel were removed. and in general.2 [Benardete (Moore. Thus. We give two more puzzles which will be relevant to forthcoming discussions. Lin. For if each successive floor is half the height of the one below it. say.. (3) The contradiction has very fruitful implications. The reason is that pure mathematics is established on a set of axioms. by the way. this will release the first room for the newcomer. and then conclusions are drawn based on the relations of those to the objects involved. is empty. If at the half-minute the particle is at point 1/2.

2. Explanation.6. 8. If another particle moves from point 2 to point 0 on the same real number line. The elementary modeling is incorrect because it does not use all known information.6. S is the area under ƒ (x ) between a and b. and the Vase Puzzle Some well-known theories and proofs of theorems in mathematics are used to argue that if explanation 2 were logically correct. This means that S (see Fig. the vertical lines x = a and x = b . Find the area of the region S under the curve y=ƒ(x ) from a to b. . We list two of the “explanations” given for the contradictory mathematical modelings. then a large portion of mathematics would be incorrect. and the x-axis. (Stewart.Unreasonable Effectiveness of Mathematics: A New Tour 181 Figure 8. Mathematical Proofs.6. which point of the sequence (8.10) should the particles meet first ? 8.2). Question 8. Therefore. is bounded by the graph of a function ƒ (where ƒ(x) ≥ 0). Theory of Cardinal Numbers. 1987)]. Explanation. 8. There is no way to finish the recursive procedure in the puzzle.3. "Experts" Explanations The vase puzzle has been presented to a number of scholars with expertise in different areas over the years. the question in the puzzle is invalid. Problem [The Area Problem.3.2.

and n rectangles.6. let us briefly go through an example. Therefore. Region S under graph of y = x ² from 0 to 1. The area of a polygon is found by dividing it into triangles.3. Find the area under the parabola y = x² from x = 0 to x = 1 (see Fig. 8. we first approximate it by areas of polygons and then take the limit of the areas of these polygons. Example 8. Figsures 8. Let S n be the sum of the areas of the n rectangles in Fig. 1] into subintervals of equal length and consider the rectangles whose bases are those subintervals and whose heights are the values of the function y = x ² at the right-hand endpoints of these subintervals.182 Chapter 8 Figure 8. to find the area of a region S. Solution. it is not so easy to find the area of a region with curved sides. 8. . Divide the interval [0. However.4–8.6 show the approximation of the parabolic segment by 4. Four-rectangle approximation of the area of the region S. To make the method more specific. Theoretically.3). The area of a triangle is half the base times the height. area is defined as the product of length and width. we have to answer the question: What is the meaning of the word “area”? It is easy to answer this question for regions with straight sides. 8. To solve this problem. For a rectangle.1.6. Figure 8.4.

either. 8.6.Unreasonable Effectiveness of Mathematics: A New Tour 183 Figure 8. Some readers may get confused here and do not see the similarity because the concept of limit is involved. we compute the area S n for n = 1. after all S n ’s are computed (based on mathematical induction we can do this). . the area A of the region S can be found by taking the limit: (8. how could we finish computing the S n ’s? If we are not able to. Eight-rectangle approximation of the area of the region S. . Figure 8.6 it appears that as n increases S n becomes a better and better approximation to the area of the parabolic segment. (8. we could not take the limit of the Sn 's. by taking the limit of the S n ’s we get the value for the area A of S. .2.3.11) From Figs. . In calculus.12) Now compare the idea applied in the area problem with that in the vase puzzle do we have similar recursive procedures? In solving the area problem.5.4 to 8. n-rectangle approximation of the area of the region S. If Explanation 2 were correct. .

then (8.3 [Halmos (1960)]. Look at the proof here. the proof given must be incorrect. let us look at the following well-known argument of Cantor. If the locations that the number A appears are the n 1 st term. say A. . the set has at least one cluster point c. . Some may argue that the proof above is not the only one for the theorem. then the recursive procedure in the vase puzzle can analogously be finished. n 2 nd term. a subsequence of is obtained such that xnk Thus.13) satisfying lim = A. When studying the concept of convergence of a sequence. Hence. 1/k → 0.6. there is an index n1 such that (8. This point must appear in the sequence at least once. suppose we have found such that Since the set (E – {c}) contains infinitely many points and the sequence x 1 ..6. . . there must be a point in the set (c – 1. must appear repeatedly infinitely many times. x n k has only n k terms.15) When k → ∞. If so. Example 8. If the subsequence can be obtained by mathematical induction.14) In general. c + 1) ∩ (E – {c}). we have a well-known theorem: Each bounded sequence of numbers must have a convergent subsequence. Then one of the numbers. . Therefore. . Case 2: Suppose is made up of infinitely many numbers. . Let us look at the proof of the theorem. We now show that there must exist a subsequence of which converges to c. Then the set E = is a bounded infinite set.184 Chapter 8 Example 8.” as posited in Explanation 2. we can pick a point (E – {c}) such that According to mathematical induction. That is.2 [Department of Mathematics (1978)]. Pick such a point y 1 in the set. One of the greatest proofs in set theory shows the following amazing result: The set of all real numbers that are greater than zero and smaller than 1 has greater cardinality than the set of all natural numbers. By a theorem of Weierstrass (1815–1897). we have proved is a subsequence of that has a convergent subsequence. if “there is no way to finish the procedure. x 2 . Since c is a cluster point of E. Thus. (8. lim = c. . Case 1: Suppose the bounded sequence is made up of a finite set of numbers.

” the desired number b can be constructed in the form of Eq. Once again. the existence of the number b does depend on the principle. (8. there is no way to construct b in Eq. By contradiction.18). Now let b be a number with expression (8.9}. the set and the set of all real numbers between 0 and 1 are infinite sets. This contradicts the assumption that each element of is in the list (8.5. That is. there exists a let a n = q if ƒ(n) = q for one-to-one correspondence ƒ : For any q ∈ Thus. is invalid.6.” there is no way to finish choosing the bn ’s.2.6.3. for each natural number n a number bn can be chosen.17).4. Conjecture 8. Hence.7. Statement (1) is true while statement (2) is not.18) such that each bk ∈ W and that for each m ∈ b m ≠ a mm . the number b needs to be constructed as follows: Initial step: Define b 1 being a number ∈ A – { a 11 }. the elements in can be written as a sequence some n ∈ (8. That is. (2) In the cases described. if “there is no way to finish the procedure.8. Even though this proof does not mention mathematical induction.17) where each a ij is a digit in the set W = {0. If this inductive process “can be finished.1. the principle of mathematical induction has been abused. To be more specific. (8. A seasoned mathematician can see that the principle of mathematical induction has been and will be “used” or “abused” in ways as we have described. the b k ’s have been defined for all k < n. By mathematical induction.18). suppose That is. Inductive step: Suppose for n ∈ Then let b n be a number ∈ A – {a nn }. one or both of the following is true: (1) Explanation 2 is incorrect. Consequently.16) Since each q ∈ satisfies 0 < q < 1. b ∈ and b is not in the list (8. it is possible to write the elements in as (8.1. Thus. . which we have just described.Unreasonable Effectiveness of Mathematics: A New Tour 185 Proof: Obviously.17). Cantor’s diagonal method.

he said.8). the phenomenon can either not be understood because of lack of knowledge or be understood at a more global level and more information is needed to be more specific. She knew the sum of the ages was 14. Connections with methodology. . methodology and epistemology. Age 1 1 1 1 1 2 2 2 2 1 3 3 Age 1 2 3 4 6 2 3 4 6 8 4 3 Age 72 36 24 18 12 18 12 9 6 9 6 8 Sum 74 39 28 23 19 22 17 15 14 18 13 14 Product 72 72 72 72 72 72 72 72 72 72 72 72 8. and the sum of their ages is the number of this room. Sonja. 1990) for details]. In this subsection. The product of their ages is 72. Factor said.. For example [see (Billstein et al.3.1. 8. Mr.6.6) and (3.1. Notice that the vase puzzle does not satisfy the general methodology of recognition. she knew that (2.6. when his students asked Mr.6. Based on the given information. When Sonja was told that the oldest child was good at chess. there would not be an oldest child.” Then. Sonja announced the correct ages of Mr.4. “I have three children. Then.6) could not be a possible combination because if the children were 2. Sonja constructed Table 8. told the teacher that she needed more information to solve the problem. 1. Some Threatening Impacts of the Vase Puzzle The first explanation is a great and fruitful resource of ideas that connect the vase puzzle with problems in philosophy. The information about ordering really made understanding more controversial rather than more specific. “My oldest child is good at chess. the class math whiz. Factor what his children’s ages were.” The children then asked for the door to be opened to verify the room number. but could not determine the correct answer from the two choices (2. and 6 years old. we discuss these connections.4.6. if known information about a phenomenon has not all been used.8.3. 6. which is 14. In general. Factor’s children as: 3.186 Chapter 8 Table 8.

It is known in analysis that the more data that are collected. Connections with epistemology. the modeling for the crystallization in (Lin and Qiu. the paradox of the hotel. 8. in turn.” It is conceptual and thought-provoking. Among the most common practices of modern scientific activities is the science and technology of data analysis. Therefore. which is continuous and complete. It challenges traditional mathematical reasoning. and an infinitely high building can be rebuilt in order to fit into a limited space. If the structure of hyperreal numbers is applied. the development of scientific theories and the understanding of realities have been hand in hand. 1987) would no longer be correct. push the development of the theories to a higher level. because by reassigning some numbers. because the ignorance of a fact will lead to a completely different prediction. For example. the commonly accepted model for time is the set of all real numbers. the vase puzzle also questions the accuracy of scientific predictions. we think the structure of the set of all hyperreal numbers (Davis. scientists establish hypotheses and develop theories.6. The vase puzzle. 1973). To know more about nature. the more misleading the conclusion that will be obtained. and an extraordinary amount of noise can easily conceal the actual state of a phenomenon. Through history. The property of completeness of real numbers was applied to model the process of crystallization of polymers (Lin and Qiu. At the same time. raise the interest of exploring the answer to the question. The vase puzzle is another example showing that the more facts we know. measuring and collecting each datum are subject to errors or noise.6. one could question the value and the meaning of scientific research and scientific predictions. about the continuity of space and time. let us describe the structure of the hyperreal numbers in some detail.4. The concept of hyperreal numbers is a product of nonstandard analysis.Unreasonable Effectiveness of Mathematics: A New Tour 187 Our vase puzzle does not fit into categories of problems considered by Polya (1973). Connections with philosophy. This theory deals with ideal elements and is a time-honored and significant mathematical . As for the first question. These theories further our understanding of nature and. Then we ask: Are the structures of space and time continuous? What is the meaning of volume? Maybe the answer to the second question is that the concept of volume is meaningless mathematically because one can prove that it is possible to chop a solid ball into small pieces and then reassemble these pieces into as many balls as one desires of the same size as the original ball (Jech. To make things more specific. 1977) describes and explains the hotel and moving particle paradoxes better than the commonly accepted one.4.3. The problem is neither a “perfectly stated” nor “practical. an infinite procedure can be finished within a randomly chosen time frame. once again.2. Studying nature and the structure of time and space is an ancient interest. It especially raises questions about the idea of mathematical modeling. the more confused we become. 8. an infinitely large piece of paper can be shrunk to any desired size. and the paradox of a moving particle. 1987).

s ∈ and infinitesimals x. × .3: Since a seemly smooth moving particle really jumps forward from a monad to another. Let  x be the absolute value of a hyperreal number x∈ Then  x = max{x.” and. – .2 and 8. The application of mathematics. and no guests live on these levels (no floor reaches these heights). the set of all hyperreal numbers has operations +.20) and – F. B. r < s .6. If is used to model time and the Cartesian product to model space. where F contains all finite hyperreal numbers. These ideal elements include “numbers” that are infinitely close to the numbers we are interested in and those that are infinitely far away from all real numbers.6. and – F includes all infinite hyperreal numbers. This neighborhood r ⊕ I is called the monad of r. 8.21) such that for any real numbers r. as shown by the examples. an ordered set of “blocks.2: Suppose the hotel roof is r feet high. Final Comments and Further Questions From our discussion it can be seen that there are roughly two methods to introduce new mathematical concepts. F can has a be obtained as follows: For each real number r ∈ the number r in neighborhood (8. I is the set of all infinitesimals. y ∈ I. Briefly speaking.6. For Question 8. brings life to the development . it could meet any point in the sequence (8.6. Each block has the same geometric structure as F and is termed a galaxy. -x}.” These blocks are densely ordered with neither a first nor last element.19) (8. It can be seen that all monads of real numbers are densely ordered. One is that in the study of practical problems. new concepts are discovered.7. The set – F of all infinite hyperreal numbers consists of F as the center “block. Questions 8. if and only if r + x < s + y. For (Benardete) Question 8.10) first. and is made up of three parts: (8.188 Chapter 8 theory. and the other is that in the search for relations between mathematical concepts. Geometrically. in either the positive or the negative directions. nothing can be seen if one looks at the hotel from above because what a person sees should be in the layers r – x for each x ∈ I and x > 0. new mathematical concepts are abstracted. ÷ and a linear ordering relation <. When the roof is removed.3 can be answered as follows: A. The set of all real numbers is a dense subset of F.

If all well-developed mathematical theories do not fit a specific situation under consideration. just as Genesis 1:1 states that “In the beginning. Does the human way of thinking have the same structure as that of the material world? On the other hand. mathematics often makes us better understand the phenomena under consideration. When facing practical problems. can we conclude that the universe we live in rests on the empty set. At the same time. mathematics. and void”? . a modified mathematical theory will be developed to study the situation. From these facts the following question arises.” History gradually shows that almost all natural phenomena can be described and studied with mathematics. is developed in the process of examining.2. God created the heaven and the earth. If all laws of nature can be written in the language of mathematics. verifying and modifying itself.1.7. the mathematical palace is built upon the “empty set. Question 8. just as all other scientific branches. And the earth was without form. That is to say.7. new abstract mathematical theories are developed to investigate them. predictions based on the newly developed theories are so accurate that the following question appears: Question 8.Unreasonable Effectiveness of Mathematics: A New Tour 189 of mathematics so that when dealing with real problems.

The Concept of General Systems Let A be a set and n an ordinal number.CHAPTER 9 General Systems: A Multirelation Approach Based on the previous chapter. the structure of quotient systems. layer structures. based on a set of given systems. Methods of constructing new systems. there exists an ordinal n = n (r). In the sequel.1. the finite chain condition. Generally speaking. The concept of systems is a generalization of that of structures. This structure shows whether a system can be seen as a set of systems or a single system only. it is time to study the concept of general systems mathematically. The concepts of controllabilities of general systems are introduced and studied in the sixth section. based on the methods of set theory. and centralized systems are studied. The second section consists of properties of relations between systems. hierarchies of systems (or. are introduced in the third section. say. and a characterization of centralizable systems. the existence of fundamental objects. large-scale systems) are studied. 9. In the fifth and the seventh sections. the structure of connected systems is studied and used to check the connectedness of newly constructed systems.” In this chapter. A n means either the Cartesian product of n copies of A or the set of all mappings from n into A. These methods serve as a mathematical foundation for the analysis of “whole–parts” relations: In the fourth section. such 191 . The chapter contains eight sections. and has a structure of “layers. From Chapters 2 and 3 it is easy to see that the two definitions of An are equivalent. Some references are listed in the last section. An n-ary relation r on A is a subset of the Cartesian product An . a function of r. a system consists of a set of objects and a set of relations between the objects. and some basic global properties of systems are studied in detail. a model of general systems is introduced. If r is a relation on A . In the first section.

It is trivial if M = Two systems S i = (Mi . Each element in M is called an object of S. are identical. Therefore. and M and R are called the object set and the relation set of S. Then a 1–1 correspondence h is defined such that h(x) = (x. for i = 1. Theorem 9. S is a (general) system if S is an ordered pair (M. Suppose that for a certain natural number n. R 0 ) if there exist systems S i = (Mi . ). if M1 = M 2 and R 1 = R2 The system S1 is a partial system of the system S 2 if M1 ⊆ M 2 and for each relation r 1 ∈ R1 there exists a relation r 2 ∈ R2 such that r1 = r 2M1 . R ) whose object set contains all systems. When r is the empty relation assume that n ( ) = 0. the system S 0 is a subsystem of the nth-level object system S n .e. R) of sets. the length of the relation is 0. respectively. and is defined by The system S1 is a subsystem of S 2 if M1 ⊆ M 2 and for each relation r 1 ∈ R 1 there exists a relation r 2 ∈ R 2 .1.192 Chapter 9 that r ⊆ An The ordinal number n = n (r) is called the length of the relation r. denoted S 1 = S 2 . Proof: Consider the class V of all sets.2. … . A system Sn = (Mn .. This is a restatement of the Zermelo–Russel paradox. . R 0 ) be a system and S n = (Mn . the Axiom of Comprehension shows that V is a set. Then S0 cannot be a subsystem of Sn for each natural number n ∈ Proof: The theorem will be proved by contradiction.1. If there is a system S = (M. then M is a set. For each x ∈ V define a system by (x . Let S 0 = (M0 . There is no system such that the object set of the system consists of all systems. 1 ≤ i ≤ n. Rn ) is an nth-level object system of S 0 = (M 0 .1. Ri).2. ). i = 1. satisfying r 1 ⊆ r 2M 1 .R i ).2 [ZFC].n–1. Each element in M n is called an nth-level object of S0 . i. R n ) an nthlevel object system of S 0 . contradiction. Theorem 9. The system S is discrete if R = or R = { } and M ≠ . such that S i is an object in M i –1. where r 2M1 is the restriction of r2 on M1 . where R is a set of some relations on the set M.

… . there exists exactly one set M(S) consisting of all fundamental objects in S. Theorem 9. Proof: Suppose that S = (M 0 . it follows that there exists a set Y ∈ X such that Y ∩ X = There now exist three possibilities: (i) Y = Mi for some i. 2. a function of i and j.1. (ii) Y = S i for some i. Each chain of object systems of a S must be finite. there exists an integer n = n (i. fore. Theorem 9. 2. such that S j is an nth-level object system of S i . 1. such that Si ∈ Mi –1 . … . These contradictions show that S 0 cannot be a subsystem of Sn . .4 [ZFC]. (iii) Y = {Mi } for some i.Ri ) be the systems. j).1. For any system S. R0 ). two sequences {*M i : i = 0.1. for i = 1. Define *M0 = {x ∈ M 0 : x is not a system} and are sets. of different-level object systems of S. and contradiction. where a fundamental object in S is a level object of the system S which is no longer a system. 2. If (ii) holds. . A similar proof to that of Theorem 9.n} and { : i = 0. . 2. If (iii) holds.General Systems: A Multirelation Approach 193 Let S i = (Mi . 1. such that for each pair i. contradiction.n – 1. Define a set X by From the Axiom of Regularity.3 [ZFC]. If possibility (i) holds. A chain of object systems of a system S is a sequence for some ordinal number α.n. i = 1.2 gives the following. Therecontradiction. … . . n} have been defined such that is not a system} . Then *M0 and Suppose that for a natural number n. j < α with i < j .

R m ) system satisfying  M and Mm < κ. Then (9. (9. Let S = ( M.R ) be a system such that.Rx ) and y = ( My .1. R ∈ p 4 ( M). y }} ∈ p 2 ( M ) and r ∈ p ( M ).3 guarantees that M( S) consists of all fundamental objects in S . . the length n(r ) = 2. ( x .1. and from the uniqueness of each *Mi it follows that M (S ) is uniquely determined by S. then there exists a partial system S' = ( M'. Therefore. . y ) = {{x }. R) is called centralized if each object in S is a system and there exists a nontrivial system C = ( MC . R' ) of S such that S' forms a centralized system and M'  ≥ θ. say x = ( Mx . Let Then Theorem 9.1. A system S = ( M. Let k be any infinite cardinality and θ > k a regular cardinality such that. Since { M} ∈ p (M) and ( M.y ∈ M.5. Proof: For each element (x . R ) p 4 ( M )).5.{x . y ) ∈ r.1) where p(X ) is the power set of X and pi +1 ( X) = p (p i ( X )) for each i = 1.1. 3 Question 9. . If there exists an object contained in at least θ objects in M.6 [ZFC].1. Ry ).1. .2.1) holds. R) is a  ≥ θ and each object m ∈ M is a system with m = ( M m . it follows that the inclusion in Eq.R C ) such that for any distinct elements x. for any Assume that S = (M. for any r ∈ R. Theorem 9. The system C is Theorem 9.194 Chapter 9 and We then define is not a system} and By mathematical induction a sequence can be defined. Give a structural representation for general systems similar to that in Theorem 9. then where called a center of the centralized system S.

we assume that  M = θ and that there is a common element in all the object systems in M . where R 1 is the restriction of the relation set R on M1 . Then the condition that there exists a common element in each system in M l implies that we can guarantee that ξ0 > 0. If x ∈ M l and ξ < ρ.Rx ) ∈ M. there is some ξ such that {Mx ( ξ ) : x ∈ M l } is cofinal in θ.2) . are irrelevant. Since θ is regular and θ > k. i.Ry ) are distinct objects in M2 . If S is a system with uncountable (or nondenumerable) object set. there exists a partial system S* of S with an uncountable object set and S* forms a centralized system. Corollary 9. Proof: It suffices from Theorem 9.R1 ) of S. there exists an forms a centralized system. Since for each α < θ. Since θ is regular.General Systems: A Multirelation Approach 195 Proof: Without loss of generality.e. for each object x = ( Mx . if each object in S is a system with a finite object set. for each x = (M x . and if there exists an object contained in at least ℵ1 objects in S. in θ. where R B is the restriction of the relation set R on B. there exists a ρ < k such that M 1 = {x ∈ M : M x has order type ρ } has cardinality θ. Let Then α0 < θ and Mx ( η) < α 0 for all x ∈ Ml and all η < ξ0 . Then Since the specific objects in each Mx . We now fix such a ρ and deal only with the partial system S1 = ( M1 . By transfinite induction on µ < θ.1.. Thus. Rx ) ∈ M. we may assume that Then.1. the object set M x has some order type < κ as a subset of θ. and whenever x = Then Let ( M x . Now fix ξ 0 to be the least such ξ. pick x µ ∈ Ml so that M xµ (ξ0 ) > α 0 and Mx µ ( ξ0 ) is above all elements of earlier x v .R x ) and y = ( My .1. let Mx ( ξ) be the ξth element of Mx .6 to show that This is clear because for each (9. For each implies that less than θ objects of the partial system is cofinal S 1 have object sets as subsets of α.

Tarski (1954–1955) defined the concept of relational systems. Klir (1985) introduced a philosophical concept of .2. many great thinkers used the languages of their times to study certain systems problems. They did not define mathematical meanings for objects nor for attributes of the objects. respectively. known as the author of the psychophysical law. A Brief Historical Remark Roughly speaking. elaborated in the way of the native philosophers of the nineteenth century supraindividual organizations of higher order than the usual objects of observation. His final model (Mesarovic and Takahara. 1972). more and more scholars have studied the concept of systems and related topics.1. Since then. In 1964 Mesarovic began to study the model of general systems in the language of set theory. Nicholas of Cusa. thus romantically anticipating the ecosystems of modem parlance. a profound thinker of the fifteenth century. Hall and Fagan (1956) described a system as a set of objects and some relations between the objects and their attributes. Rx ) and and The system C is called an S-center of S. the idea of systems appeared as long ago as Aristotle. see (von Bertalanffy.1. Von Bertalanffy began to study the concept of systems formally in biology. introduced the notion of the coincidentia oppositorum. say x = ( Mx . for example.R C ) such that for any distinct elements x and y ∈ M. 9. For example. The concept of systems was not introduced formally until the 1920s.R ) is strongly centralized if each object in S is a system and there is a nondiscrete system C = ( MC .1. for details. the sets C and E are called the composition and an environment of the system W. S ) is a system over T if and only if C and E are mutually disjoint subsets of T and S is a nonempty set of relations on C ∪ E . Then the ordered triple W = ( C . linking medieval mysticism with the first beginnings of modern science. Give conditions under which a given system has a partial system which is strongly centralized and has an object set of the same cardinality as that of the given system. 1975) reads: A system S is a relation on nonempty sets: (9. Bunge (1979) gave a model of systems as follows: Let T be a nonempty set.3) After considering the interrelationship between the systems under concern and some environments of systems. Leibniz’s hierarchy of monads looks quite like that of modem systems. Question 9. Gustav Fechner. E . Aristotle’s statement “the whole is greater than the sum of its parts” could be the first definition of a basic systems problem.196 Chapter 9 A system S = ( M . Later. For example.

T) be a topological space [for details see (Engelking. 1987).5) (9. It can be seen that the concept of algebras is a generalization of the concepts of rings. Structures Let A be a set and n a nonnegative integer. Ø. for every v < 0 1 ( τ ) we realize rv as an m v -ary relation (rv ) U on A. and the length of each relation in T is 1. 1978) is a triplet (A. where A is a nonempty set. We conclude this section with three mathematical structures related to the concept of general systems: structure. F. An n-ary relation r on the set A is a subset of A . So ( X . and the like. of an m v -ary relation. Any n -ary operation is an (n + 1)-ary relation. Now. An n -ary operation on A is a n mapping f from A n into A.1. For every v < 0 0 ( τ ) there exists a symbol f v of an n v -ary operation. . thus a general system. there exists an ordinal number n = n (U ) = 1 such that U ∈ X n .1. T) is also a system with object set X. and the concept of structures combines those of algebras and relational systems. applying the Axiom of Choice implies that the topology T can be well-ordered as (9. groups.7) where each open set is a 1-ary relation on X. and G-system. and relation set T.General Systems: A Multirelation Approach 197 general systems.6) If 0 1 ( τ) = 0.2. R). and (9. The concept of general systems discussed intensively in this chapter was first introduced by Lin and Ma (1987. (X. which is a relational system. 1975)]. A type τ of structures is an ordered pair (9. and for every v < 0 1 ( τ ) there exists a symbol rv . We show that any topological space is a relational system. 9.1. the concept of relational systems generalizes the idea of algebras. Therefore. Let (X .4) where 0 0 ( τ) and 01 ( τ ) are fixed ordinals and n v and m v are nonnegative integers. Therefore. T) becomes a structure. The fact that any topological space is a relational system shows that the research of relational systems will lead to the discovery of properties that topologies and algebras have in common. Then for each open set U ∈ T . U is called an algebra and if 0 0 ( τ ) = 0. L-fuzzy system. U is called a relational system. A structure U (Gratzer. For every v < 0 0 ( τ ) we realize f v as an n v -ary operation (f v ) U on A. Example 9.

≤ ) is called a lattice if for any elements a. then a' ≥ b'. 1990a) is based on the concept of fuzzy sets introduced in 1965 by Zadeh. denoted by ∧ A o r infA . Tarski. we consider that A is the such that mapping µ A in LX .8) and (9. A is an L -fuzzy set on X iff there exists a mapping µ A ∈ LX Without confusion. including Whitehead.9) We say that L is a lattice with order-reversing involution if there exists an operator ' defined on L such that for any a. Many great mathematicians worked in the field. Let X be a set. Henkin. Birkhoff.198 Chapter 9 A systematic introduction to the study of structures can be found in (Gratzer. . L-Fuzzy Systems The concept of fuzzy systems (Lin. assume that 0 and 1 are the minimum and maximum elements in L. has at most two values. Jonsson. or say A . and there exists exactly one least upper bound of a and b. Keisler. Chang. (9. An ordered set (L . where I is an index set. 9. such that We call a lattice distributive if the lattice. satisfies We call a lattice complete if for any nonempty subset A ⊆ L. denoted by a ∨ b . there exists exactly one least upper bound of the elements in A. 0 and 1. b ∈ L . In the following. there exists exactly one greatest lower bound of a and b. If µ A . in addition. 1978).3. denoted by ∨ A or supA.1. and there exists exactly one greatest lower bound of the elements in A. if a ≤ b. etc. We call a lattice completely distributive if for any a ∈ L and any subset {b i : and exist and satisfy i ∈ I} ⊆ L. then the L-fuzzy set A can be seen as a subset of X. b ∈ L. denoted by a ∧ b . a completely distributive lattice L with order-reversing involution is fixed.

…} be . and Thus. R ) = {{M}. then Let {A i : i ∈ I} be a set of L-fuzzy sets on X. The L-fuzzy system S1 is a subsystem of the L-fuzzy system S 2 if M l ⊆ M 2 and for any L-fuzzy relation r l ∈ R1 . i.1. Suppose that an L-fuzzy system S = (M . An example is constructed to show that the equation S 1 = S 2 will not follow from the condition that the L-fuzzy systems S 1 and S 2 are subsystems of each other.12) Proof: S = (M. Let = {1. y. Then are defined by and (9.11) for all x ∈ X. R }} ⊆ p(M ∪ R). S is an L -fuzzy system iff S is an ordered pair (M. (9. So holds. and for any (x. R ) is such that. If Y ⊂ X . i) ∈ r ⊆ M ² × L. 2.12) Suppose that S i = (M i . there exists an L-fuzzy relation r 2 ∈ R 2 such that Example 9.10) and (9.. Then A = B iff µ A ( x) = µ B (x ) for all x ∈ X . are L -fuzzy systems. for any r ∈ R . i = 1. R i ). Then (9. for any L-fuzzy relation r ∈ R. The ordinal n is called the length of the L-fuzzy relation r. {M. the inclusion in Eq. where R is a set of some L-fuzzy relations defined on the set M. n(r) = 2.2. R ) of sets. there exists an ordinal number n = n( r) such that r is an L-fuzzy set on the Cartesian product Mn . Theorem 9. A ⊂ B iff for all x ∈ X . 2.7.1. Suppose that the completely distributive lattice with order-reversing involution lattice L is the closed interval with the usual order relation.General Systems: A Multirelation Approach 199 Suppose that A and B are two L -fuzzy sets on X .e.

and a class mapping transfinite induction. 2. ≤ A ).4.14) and for all (9. but S1 ≠ S 2 . 2. R i ). = 1. 9. two classes can be defined with the following properties: (9. S is a G-system (Lin and Ma. for each i ∈ I. The concept of G-systems can be used to study more applied problems because it contains a more general meaning of relations. and We now define two L-fuzzy systems.200 Chapter 9 the set of all natural numbers.17) . and as follows: where for each (9. x) : x ∈ A }. and the set K is called the type of the system S.13) and for any (9. K i is a set partially ordered by ≤ i iff S is an ordered pair of The set K i is called the type of sets such that for any the relation r i . be two systems and h : M l → M2 a mapping. 1989).2. 9. 1989) of type where I is an index set and.16) and for each x = (9. where the ordering is defined by ≤ A = {(x. and a sequence defined by a i = 1 – 1 / i.15) Then it is easy to show that S 1 and S 2 are subsystems of each other.1. By i = 1. The interested reader is encouraged to consult (Lin and Ma. Mappings from Systems into Systems Let S i = (Mi . G-Systems In this part any set A will be seen as a partially ordered set (A .

General Systems: A Multirelation Approach 201 For each relation is a relation on M2 with length n (r) . h : S 1 → S 2 is also called surjective. When h : M 1 → M 2 is surjective.1. we show that is a similarity mapping. there exists a relation s ∈ R 1 such that r = h ( s). injective. . let [m1 ] and be such that and and so Therefore. A mapping h S 0 : S 0 → A is an n -level homomorphism from S 0 into A if the following inductive conditions hold. where and for each Proof: It is easy to check that Eh is an equivalence relation on the object set M1 . This implies that [m 1 ] = [m 2 ]. Then (9. there exists a relation s ∈ R1 such that r = µ (s). We next show that is well defined. Then is a similarity mapping. The systems S i are similar. This completes the proof that is a similarity mapping from S1 /Eh onto h (S 1 ) . if there exists a bijection h : S 1 → S 2 such that The mapping h is a similarity mapping from S1 onto S 2 . where n is a fixed natural number. Clearly. the inverse mapping h – 1 is a similarity mapping from S 2 onto S1 . R). In fact. contradiction. or bijective. and the system S 0 is n -level homomorphic to A. Without confusion. respectively. For any relation r ∈ R 1 /E h . Suppose that h : S 1 → S 2 is a homomorphism from S 1 into S2 . is a bijection. Theorem 9. we use h for h X. Then R/E = {µ (r) : r ∈ R }. Then h X indicates the restriction of the mapping on X. For any relation r ∈ h ( R 1 ). The quotient system S/E is defined (9. Let X be a subsystem of S1 .A mapping h : S 1 → S 2 is a homomorphism from S1 into S2 if Let E be an equivalence relation on the object set M of a system S = ( M. h(m1) = h (m 2 ).18) where µ : M → M /E is the canonical mapping defined by µ(m) = [m]. Let S 0 and A be two systems with no nonsystem kth-level objects for each k < n. This implies that Finally. injective. Therefore.2. Let be such that [m1 ] = [m2 ]. denoted h : S 1 → S 2.19) Thus. then m 2 ∈ [ m1 ]. where µ is the canonical mapping from M1 into M 1 /E h . Without confusion. and so h(m 2 ) = h (m 1 ). h will also be used to indicate the class mapping and h is a mapping from S 1 into S2. or bijective.

. and the system S 0 is n -level similar to A. k = 1. (2) For each i < n and each ith-level object Si of S 0 .21) . (b) If A is n-level similar to B and B is m-level similar to C. .i – 1. . Theorem 9. . Without loss of generality. then A is min{n. the object system m and h(m) are similar. Proposition 9. m}-level similar to C. B. . i. Let C be a center of S * . Each center of S C is also called a center of S. there exists a homomorphism h S l . k = 0. R) be a system such that each object in M is a system. If the system S = (M.2. Then S is centralizable. then A is min{n. A system S is centralizable if it is 1-level homomorphic to a centralized system S C under a homomorphism h : S → S C such that for each object m in S.2.202 Chapter 9 (1) For each object S 1 in S 0 . . The proof is straightforward and is omitted. RC ) such that C is embeddable in each object of S. and C are systems. Then C is embeddable in each object of S.1. For each the system C is embeddable in x. Let S = (M. (a) If A is n-level homomorphic to B and B is m-level homomorphic to C. . and each similarity mapping here is called an embedding mapping from C into S1 . When all homomorphisms in this definition are similarity mappings. i. there exists a centralized system S * = (M * . we can assume that the collection of all object sets of the objects in S consists of disjoint sets. Sufficiency.20) and let be the bijection defined by (9. and homomorphisms hS k . . .1. so there exists a one-to-one mapping such that is a partial system of x. such that S k is an object of the object system Sk –1 and hS k is a homomorphism from S k into h S k –1 (S k ) for k = 1. .) Proof: Necessity. Suppose that A. from the object system S 1 into the object system h S0 (S 1 ). Define a new as follows: system (9. the mapping hS 0 is termed as an n-level similarity. iff there exists a nontrivial system C = ( MC . . there exist level object systems S k .2. (C embeddable in an object S 1 in S means that C is similar to a partial system of S 1 . m}-level homomorphic to C.2. . R * ) such that S is 1-level homomorphic to S * .2.R) is centralizable.

i = 1. x * is a system with some element in M as its object set. Therefore. We now consider a system S with object set M * such that for any x * ∈ M * .2. R 1 ) into a system S 2 = (M2 . the concepts of centralized systems and centralizable systems are different. Theorem 9. Assume that S is a system with an object set of cardinality ≥ θ and each object in S is a system with an object set of cardinality < κ .2. Example 9. are groups.2. There then exists a partial system S' of S such that the object set of S' is of cardinality ≥ θ and S' forms a centralizable system. Let ω1 be the first uncountable ordinal number. for some natural number n > 0. S is a centralizable system.23) Proposition 9. iff it is continuous from S1 into S 2 .R 2 ) is S-continuous if for any relation r ∈ R2 .6.6 and 9. Let κ and θ be cardinalities satisfying the conditions in Theorem 9.1. such that  M = ω1 and for any x. and M a collection of n-element subsets of ω1 .1.1.6 that there exists an object belonging to at least θ objects in S is essential for the result to follow and that not every centralizable system is a centralized system. and for any x ∈ M there exists exactly one x * ∈ M * such that x is the object set of the system x * . the S-continuity of h implies that h is a homomorphism from S1 into S2 .2. are topological spaces. Then (i) If the systems S i . . (iii) If the systems S i . (ii) If the systems Si .2. x y = if x ≠ y. An example is constructed to show that the hypothesis in Theorem 9. It is clear that S has no partial system which forms a centralized system and that S is a centralizable system.2.2. Then the system h * (S) is centralized with the system C as a center.y ∈ M. (9.1. The proof follows from Theorems 9. are rings.3.2.General Systems: A Multirelation Approach 203 Then define (9.2. A mapping h from a system S 1 = (M1 . Consequently. Suppose that h : S1 → S 2 is a mapping. the S-continuity of h implies that h is a homomorphism from S1 into S2 . h is S-continuous. Let h * be a mapping defined by letting h * (x) = x * for each x ∈ M. i = 1.22) So C is a partial system of x * . i = 1.

2. g ƒ is S -continuous.2. and the relation r+ with +.24) From the hypothesis that the mapping ƒ is S-continuous.5. is defined to be the system (9. Then the composition g ° ƒ : S 1 → S 3 is also S-continuous. and z ∈ M 1 . (9. from the hypothesis that the mapping g is S-continuous we have (9.25) Therefore. Proof: For each relation r in the system S 3 .R) will be identified with the group (X.R) is defined as follows: R = {r + } and The system (X. A bijection h : S 1 → S 2 is a similarity mapping iff h : S 1 → S 2 and h – 1 : S 2 → S1 are S-continuous.n } is finite. denoted ⊕ {Si : i ∈ I} or if I = {1.1.2. y.1. let (X. . Theorem 9. from Eq. then h(x)+2 h(y) = h (z) This implies that h is a homomorphism from S 1 into S 2 . . For (ii). i = 1.25). Then Thus. . The proof is straightforward and is left to the reader. +) be a group. Suppose that ƒ : S 1 → S 2 and g : S 2 + S 3 are S-continuous mappings.26) .204 Chapter 9 Proof: (i) follows from the definition of S-continuity of mappings and Example 9. The proof of (iii) is a modification of that of (ii) and is omitted.3.2. The free sum of the systems Si . be two groups and h : S 1 → S 2 an S-continuous mapping.4. it follows that (9. if x + 1 y = z. Thus. . +). Theorem 9. for any x. ° 9. A system (X. Constructions of Systems be a set of systems such that M i Mj = for any Let distinct i and j ∈ I. Let .

29) . Hence. To show this. Then for any fixed i ∈ I and any relation r ∈ Ri . s∈R Theorem 9. The sets X and Y are called the input space and the output space of the system S. Suppose that {Si = ( Mi .3. If S1 is a free sum of some systems. and h : S 1 → S2 is a similarity mapping.Ri ) is an input–output system.1. Therefore. it can be assumed that any set of systems has a free sum (uniquely determined up to a similarity).1. A system S = (M.3.27) Up to a similarity. such that h (s) = r. and define (9. where {S i : i ∈ I } is a set of systems S = (M i . Proof: Necessity. there exists an i ∈ I such that r ∈ Ri . m n there are nonzero ordinals n = n(r) and m = m (r) such that r ⊆ X × Y . then so is S 2 . respectively. Let S 1 and S 2 be two similar systems. therefore. but in the proofs of some theorems in the future it will be tacitly assumed that the set under consideration consists of systems with pairwise disjoint object sets.28) because for each i ∈ I. for some i ∈ I. and for each r ∈ R2 there exists a relation i . S i is an input–output system with input space Mi X and output space M i Y. Then for each relation r in ⊕ {S i : i ∈ I}.Ri ) : i ∈ I} is a set of systems. Sufficiency.R i ). i ∈ I.R i ) is an input–output system with input space X i and output space Y i . Suppose that S i = (M i . Then the free sum ⊕{Si : i ∈ I} is an input–output system iff each system S i = ( M i . the free sum of the systems S i is unique. (9.R) is an input–output system if two sets X and Y are given such that X Y = M and for each relation r ∈ R there are nonzero ordinals n = n ( r) and m = m( r) such that r ⊆ X n × Y m .General Systems: A Multirelation Approach 205 The concept of free sums of systems can be defined for any set {S i : i ∈ I } of systems such that the collection of all object sets of the systems in the set consists of not necessarily pairwise disjoint sets. where I is an index set. for each i ∈ I. we simply take a set {* S i : i ∈ I } of systems with pairwise disjoint object sets such that * S i is similar to S i . Then i (9. Proposition 9. Proof: Suppose that S 1 = ⊕ {S i : i ∈ I }. Suppose that ⊕{Si : i ∈ I } is an input–output system with input space X and output space Y.

4.2 implies that the free sum is centralizable.3. Then the system S is also a free sum of some systems.R) and S i = (Mi .R i ) for each i ∈ I. Theorem 9.5. Proof: Let for some i ∈ I.Ø) is embeddable in each object in the free sum ⊕{S i : i ∈ I }. Theorem 9.30) Define (9.3.3. then each S i is centralized. The proof is straightforward and is omitted. i ∈ I.3. Suppose that {S i : i ∈ I } is a set of centralizable systems such and that that each system Si has a center for all i and j ∈ I. Then (9. Then the free sum ⊕ {S i : i ∈ I} is a centralizable system. .2. then S is also a free sum of some systems.32) Therefore.206 Chapter 9 Hence. Let h : S → ⊕ {Si : i ∈ I} be a homomorphism.3 1) From the hypothesis that h is a homomorphism it follows that (9. Theorem 9. It is easy to construct an example to show that the free sum of two centralized systems may not be a centralizable system. for i ∈ I.2. If the free sum ⊕{S i : i ∈ I} of nontrivial systems S i is a centralized system. Proof: Suppose that S = (M. Theorem 9. If a system S is embeddable in the free sum ⊕{S i : i ∈ I } of some systems Si . Proof: The theorem follows from the fact that each center of the free sum ⊕ {S i : i ∈ I } is a center of each system Si . Then the system ( ℵ .3. the free sum ⊕ {S i : i ∈ I } is an input–output system with input space and the output space Theorem 9.

let r be a relation in S 1 . *Ri ) as follows: *Mi = ƒ (Mi) and (9.1.7.8. and ƒ : S i → * S i is S-continuous. Proof: For each i ∈ I we define a system *S i = (*Mi .3. Suppose that h : S 1 → S 2 is a homomorphism from a system S1 into an input–output system S2 . Let ƒ : ⊕ {S i : i ∈ I } → S be an onto S-continuous mapping. there are nonzero ordinals n and m such that This implies that Theorem 9.37) Then S 2 is an input–output system with input space X2 and output space Y2 .33) follows and the S-continuity is S -continuous.3. Hence. Under what conditions will the system S in Theorem 9. Then S 2 is also an input–output system. (9. of the mapping ƒ : S i → * S i follows from Eq.3.35).6. Then S 1 is also an input–output system.3. Then h(r) ∈ R 2 .33) where *S i = (*Mi .36) (9.6 be a free sum of some systems? Theorem 9. Define and (9. Proof: Suppose that S 2 has input space X and output space Y. . Then there are partial systems {*S i : i ∈ I } of S such that (9. Proof: Suppose that S 1 has input space X 1 and output space Y1. Question 9. (9. for each relation Hence.35) Since ƒ From the hypothesis that ƒ is surjective.34) Then S 1 is an input–output system with input space X* and output space Y*. *Ri ). it follows that Then Eq.3. In fact. Define X 2 = ƒ (X 1 ) and Y2 = ƒ (Y1) (9. In fact. Let ƒ : S1 → S 2 be an S-continuous mappingfrom an input–output system S 1 into a system S2 . there are nonzero ordinals n and m such that Therefore. for each i ∈ I.General Systems: A Multirelation Approach 207 Theorem 9.

39) for all i ∈ I. S i ) . then if Thus. R) is a Cartesian product of the systems S i if there exists a family of S-continuous mappings {p i : . A system S = (M .9. S). We now suppose that S' = (M'. Therefore.3. Let Let {S i : i ∈ I } be a set of systems.S ) so that (9. S 2 ) = {h : S 1 → S 2 : h is a homomorphism} (9.S i ) : i ∈ I }. R) is a product of the systems S i if there exists a family of homomorphisms {Φ i ∈ Hom(S. a product of the systems or S i always exists. A system S = (M . S is a product of the systems S i .R') is a system and {Ψ i ∈ Hom(S'. If λ ' is another such homomorphism. i ∈ I. S i ) : i ∈ I}. Theorem 9. which is unique up to a similarity and denoted by Proof: For the set of systems {S i : i ∈ I}. they induce the same homomorphism. there exists a unique homomorphism λ ∈ Hom( S'. λ (m') = { Ψ (m') : i ∈ I } ∈ M.40) (9.208 Chapter 9 Let S 1 and S 2 be two systems. we define a system S = (M. λ ' = λ as mappings from the object set M' into the object set M.42) and (9.R) follows: as (9.43) where for any let for i ∈ I. Therefore. We define Hom(S 1 . and (9. For each set {S i : i ∈ I } of systems.S i ) : i ∈ I } such that for each system S' and a family {Ψ i ∈ Hom(S'.44) be the projection for each i ∈ I. It then can be shown that λ ∈ Hom( S ' .41) (9.38) Suppose that {S i : i ∈ I} is a set of systems. Define a mapping λ : S' → S as follows: For any object m' ∈ M'. Then pi ∈ Hom(S .

Then λ is S -continuous.General Systems: A Multirelation Approach 209 S → S i : i ∈ I } such that for each system S' and a family {q i : S' → S i : i ∈ I } of S-continuous mappings. Then it can be readily checked that each mapping pi : S → S i is S-continuous. there exists a unique S-continuous mapping λ : S' → S so that q i = p i λ for each i ∈ I. a Cartesian product of the systems S i always exists. Define a mapping λ : S' → S by (9. and {q i : S' → S i : i ∈ I } a family of S-continuous mappings. R i ) for each i ∈ I.47) for each object m' ∈ M'. In fact. Define a system S = ( M . which is unique up to a similarity and denoted by ° Proof: Let S i = (M i .3. Theorem 9. we show that all Cartesian products of the systems S i are similar to each other. R) as follows: (9. Let S' = (M'. .R') be another system. let r j ∈ R j be a relation such that Then To complete the proof. for an arbitrary relation r ∈ R. For each set {S i : i ∈ I } of systems.46) for any and define mappings p i : S → S i by letting for each i ∈ I.10.45) and (9.

In fact. Therefore.9 and 9. (9. S is the Cartesian product of the spaces S i . i ∈ I. and Eq. Uniqueness of the Cartesian product.50). λ 1 = λ 2 and λ 2 = λ 1 .1) commutes for each i ∈ I. When the systems S i are groups.9 is the direct sum of the groups S i .1.49).10 is called a Cartesian product of the systems S i is that when S i are topological spaces. we have (9. (9.49) and (9. Π{S i : i ∈ I} in Theorem 9.9 is the box product of the spaces S i . Eq. we prove Eqs.3.50).48) It suffices to show that (9.50) implies that λ 2 is 1–1. . Then there are S continuous mappings λ 1 : S → S * and λ 2 : S* → S such that the diagram (Fig.R*) be another Cartesian product of S i . For Eq.49) implies that the mapping λ 1 is –1 –1 1–1. let then Therefore.10 are called projections from the product systems into the factor systems. The mappings p i in the proofs of Theorems 9. Now. A similar argument shows The reason why the system S in Theorem 9. λ l and λ 2 are similarity mappings.3. (9. That is.3. (9. 9.50) for each m ∈ M and m * ∈ M*.210 Chapter 9 Figure 9.3. The system Π {S i : i ∈ I} in Theorem 9. Eq. for each i ∈ I.3.49) and (9. Let S* = (M*. (9.

56) for each object the product mapping is an embedding mapping. R) such that (9. denoted by is defined by where the object set M and the relation set R are defined as follows: There exists a bijection such that (9. is 1–1. For the 1-level product system it suffices by Theorem 9.52) The 1-level Cartesian product. For each Let and be an embedding mapping. denoted by defined by the following: There exists a bijection is the system (M. In fact. Proof: For each i ∈ I suppose that is a center of the system S i .2 we will not lose generality by assuming that each system Ci is embeddable in each object in Mi for i ∈ I.53) and (9. The 1-level product system.11.2. because if and are two distinct .54) Theorem 9.2 to show that the product system is embeddable in each product for all From Theorem 9. Then the 1-level product system and the 1 -level Cartesian product system are also centralizable systems.General Systems: A Multirelation Approach 211 Let be a set of systems such that each object in any one system Si in the set is a system again.5 1) and (9. first. Then the product mapping (9.55) defined by (9.3.2. Let be a set of centralizable systems.

(9. for each relation r in such that there exists a relation s in (9. it follows that there exists a relation for each i ∈ I.212 Chapter 9 objects in there exists an i ∈ I such that is an embedding mapping it follows that Therefore.60) implies that as follows: for each i ∈ I. In fact. since h i Therefore.59) From the hypothesis that each h i is an embedding mapping. for each relation r in i ∈ I.62) . (9.61) We now show that s satisfies Eq.57) Second. (9. such that for all i ∈ I and for each (9.60) Equation (9. Define a relation s in (9. such that (9.58) there exist relations In fact.58).

General Systems: A Multirelation Approach 213 where for each α < n (r) and i ∈ I.13. For each relation r in i ∈ I and a relation r i ∈ R i satisfying that Therefore. it follows that pi h is a homomorphism for every i ∈ I. ° Sufficiency. (9. h is S -continuous.12. it follows that each composition pi °h is S -continuous. Therefore.3. R) into a Cartesianproduct system is S-continuous iff the composition pi °h is S-continuous for each i ∈ I. A mapping h from a system S = (M.. Theorem 9.65) . Suppose that pi h is a homomorphism for every i ∈ I.3. From the fact that each projection is a homomorphism.5 and the proof of Theorem 9. Suppose that each composition pi ° h is S -continuous. Suppose that is a homomorphism.3. Proof: Necessity. i ∈ I. ° (9. Proof: Necessity. Sufficiency.10. i ∈ I. A mapping h from a system S = (M.63) The centralizability of the 1-level Cartesian product system can be proved in a similar way. From Theorem 9. ° For each i ∈ I we have We need to show that h(r) is a relation in since pi h is a homomorphism. Suppose that is S -continuous.2. We complete the proof of the theorem by observing that (9. Let there exist an for each i ∈ I. R) into a product system is a homomorphism iff the composition pi h is a homomorphism for ° every i ∈ I.64) This implies that i. Let r ∈ R.e. Theorem 9.

R) the following conditions are equivalent: (i) The system S is connected. (9. .4. . . .R) be a system and r ∈ R a relation.2. there exist a natural number n > 0 and n relations ri ∈ R such that (9.69) The system S is connected if it cannot be represented in the form S 1 ⊕ S 2 .68) This implies that the relation h(r) belongs to the relation set of 9. For every system S = (M. (ii) For any two objects x and y ∈ M.1.65) implies that (9.7 1) for each i = 1.66) Equation (9. denoted Supp(r). is defined by (9.70) and (9. Theorem 9. (9.67) So Eq. The support of r.4.66) implies that for every si ∈ R i with n (s) = n(s i ). n – 1.214 Chapter 9 and for each i ∈ I and s ∈ Ri . . (9. Structures of Systems Let S = (M. where S 1 and S 2 are nontrivial subsystems of S.

2. . n.79) for (9. such that (9.General Systems: A Multirelation Approach 215 Proof: (i) → (ii). Suppose that the system S is connected and there exist two objects x and y ∈ M such that there do not exist relations r i ∈ R.. thus r ∈ R 2 Therefore.77) (9.76) Then all natural numbers n.80) In fact. for each relation r ∈ R. i = 1. We prove by contradiction. . . there exist nontrivial subsystems S 1 and S 2 of S such Pick an object mi ∈ M i . . . that S = S 1 ⊕ S 2.78) (9. contradiction.2.75) and (9.n – 1. . it follows that there must be relations r1 . for any natural number n ≥ 1. Suppose that Then and and so . Thus. . (ii) → (i).72) and (9. if r ∉ R 1.s 1 ∈ R such that x ∈ Supp( r1 ) and y ∈ Supp (s1 ).74) Let and for each natural number n ∈ let (9. Suppose condition (ii) holds and S is disconnected. From the hypothesis that S is connected.73) for each i = 1. m ∈ We now define two subsystems and of S such that (9. Then our hypothesis implies that (9. The proof is again by contradiction.

An object m in a system such that the cardinality is isolated if there is not relation r ∈ R and . Then there are no relations rj ∈ R. R) the following conditions are equivalent: (i) The system S is connected. …. m ∈ and relations ∈ R such that (9.86) (9. for any fixed n ∈ such that (9. … . for every object p ∈ M .4. contradiction. Define the star neighborhood of the subset M* .4.216 Chapter 9 i = 1.83) By applying induction on n ∈ we can define Star 1 (M * ) = Star(M * ) and Star n+1 (M * ) = Star(Star n ( M *)) (9. Let M* ⊆ M .87) and Thus.81) and (9. it follows that there exist n. y ∈ M. Star Proof: (i) → (ii) follows from Theorem 9. For every system S = (M.84) Corollary 9. from the hypothesis that Star∞ ( {p })= M.82) for each j = 1.1(ii) holds and from Theorem 9. n – 1.1.4. j = 1.2.1 that S is connected. denoted Star (M * ). R) is a system.1 directly.2. (ii) For each object x ∈ M. (ii) → (i). where S = (M.85) (9. 2. it follows that Theorem 9. as follows: (9.n. For any two objects x.4.

n – 1. If S 2 is connected. Now there are two possibilities: If (1) holds. i = 1. There are relations that 1. Theorem 9. Suppose there exists a mapping f : S → D and such that Pick elements x ∈ From Theorem 9. R) is a connected system and D = is a disconnected system.3.1 it follows that there exist n ∈ and n relations ri ∈ R.General Systems: A Multirelation Approach 217 Corollary 9.2.4. There is no isolated object in a connected system. n. Let us pick two arbitrary objects for some fixed n ∈ such for each . Theorem 9. the S1 is also.90) Sinceƒ –1 (R D ) ⊇ R .4. we will have again a contradiction.3. 2. Then h(S1 ) = (h(M1 ). Let h : S 1 → S 2 be a homomorphism from a connected system S1 into a system S2. such that (9. The contradictions imply that every Supp mapping f : S → D with must satisfy Corollary 9. y ∈ M1 . … . Suppose that S = ( M. we must have Supp contradiction. Proof: Suppose that x. Suppose that h : S 1 → S 2 is an S-continuous mapping from a system S 1 into a system S2 .2.89) for each i = 1. If (2) holds. Pick an i with 1 ≤ i ≤ n such that (9.4. h(R1 )) must be a connected partial system of S 2 . …. 2. it follows that there exists a relation r ∈ RD such thatƒ(r i ) = r.4. 2. If there exists a mapthen only one of the ping f : S → D with following holds: Proof: We prove by contradiction.4.88) and (9.

it follows from Theorem 9.3. j ∈ Supp( rj ) = {j. for any two objects there exist relations ri .j – 1. … }. and for every pair S1 .218 Chapter 9 i = 1. n – 1. S 2 of subsystems of S with disjoint object sets such that C is a subsystem of S1 ⊕ S2 . Theorem 9.4. n.2.}. Then the system is disconnected. 3. ri = {(i . –2. 2. and S 1 is connected. R C ) and S i = (Mi .Ri ). where r i = {(i. –2. = {1. .92) for i = 1. Define a system S1 = ( .4. 2.1. ±2. ±1.1. then C is a subsystem of S1 or S 2 . … }. From the hypothesis that h is S -continuous. ±2.93) Pick elements x ∈ MC ∩ M1 and y ∈ MC ∩ M2 . n . and Supp(rk ) ∩ Supp(r k +1 ) = {k + 1}. … . … . it follows that there exist a natural number n > 0 and n relations r i ∈ R C . n – 1. i + 1)} for each i ∈ Then S 1 is connected. Let C = (MC . –1. it follows that for i = 1.j ∈ i ∈ Supp( ri ) = {i. 2. = {ri : i = –1. where = {0. where we assumed that i < j.1 that S1 is connected.95) . Since C is connected and from Theorem 9. Suppose that C is not a subsystem of S 1 nor of S 2. From Theorem 9. … . for k = i. … . . … . i + 1.4. .j + 1}.4. . . ±3. i = 1. Example 9.94) and (9. .4. We construct an example to show that the connectedness of the system S1 cannot guarantee the connectedness of the system S 2 in Theorem 9.91) and (9.4. R 1) as follows: R1 = {r i : i = 0. . Proof: We show the result by contradiction. … .1 it follows that S 1 is connected. 2. Let S 2 = ( R 2) be a system where is the set of all integers and where for each i = ±1.4. Thus. for i = 1. r i + l . i+ 1 ) } . 2. In fact. 2. (9. Let h : S 1 → S 2 be the identity mapping defined by h(i) = i for each i ∈ Then h is S-continuous. such that (9. Then (9. and S 2 is disconnected. such that i. … }. Therefore. i + 1}. …} and = {r i : i = 1. rj . If a subsystem C of a system S is connected.

98) and (9. S u p p ( r i ) ∩ M2 = Ø.4. If the set {Ci = (Mi . if r ∈ R1 . R) satisfies is connected. Choose an object z ∈ M i 0 . Therefore.2. then the system is connected. R) be a system and D ⊂ M a subset. is Corollary 9. R i ) : i ∈ I } of connected subsystems of the then the subsystem system S = (M. . The subset D is dense in the object set M if the following conditions hold: (i) For any relation r ∈ R with r ≠ Ø . Supp(r i) ∩ M1 = Ø. but not both.99) (9.R).100) Therefore. … . Let S = (M.4. there must be an i satisfying 0 < i ≤ n such that (9. m.4.General Systems: A Multirelation Approach 219 Therefore.97) (9. for each i ∈ I.1 implies that the system connected. Theorem 9. for k = 1. This implies that C is a subsystem of S1 or S 2 . it follows that there are natural numbers n and m and relations r k and q p ∈ R. If there exists an index i 0 ∈ I such that each system is connected. Supp(r) ∩ D ≠ Ø . contradiction.2. From the hypothesis that the systems are connected. Proof: Let us pick two arbitrary objects x and y from There are two indices i and j such that x ∈ Mi and y ∈ Mj . (ii) Each isolated object in M belongs to D. such that (9. if r ∈ R 2 . R i ) : i ∈ I } be a set of connected subsystems of a system S = (M.4. n and p = 1. Let { Ci = (Mi .5. Theorem 9.96) From the definition of subsystems it follows that there exists a relation r in the free sum S 1 S 2 such that It follows from the definition of free sums that r is contained either in R 1 or in R 2 . … .

(3) x.102) and (9. x and y are not isolated objects in M. There are three possibilities: (1) x.R* ) of S is a dense subsystem of S if the object set M* is dense in M. Let ry ∈ R be a relation such that (9. there are relations rx and ry ∈ R such that (9.4.104) (9.4. i = 1.105) and (9. (2) one of x and y is contained in M* and the other is contained in M – M* .108) .4.220 Chapter 9 A subsystem S* = ( M*. Theorem 9. . . If a system S = ( M. y ∈ M – M* . n. 2. . . Then it has to be that (9. .R*) is a connected dense subsystem of S. y ∈ M. such that (9. . Theorem 9. From the hypothesis that S* is connected. .6.107) Pick y* ∈ Supp( ry ) ∩ M*. If possibility (1) holds. . then S is connected. let qi be a relation in R such that ri ⊂ q iM*.1 implies that there are natural number n and n relations ri ∈ R*.R ) contains a connected dense subsystem. y are contained in M*. Then y is not an isolated object in M. . 2. such that (9. To show that S is connected. If possibility (2) holds. .106) Theorem 9. without confusion. i = 1. . let x ∈ M* and y ∈ M – M*. Proof: Suppose that S* = ( M*. it follows that there are a natural number n > 0 and n relations ri ∈ R*. n.101) It follows from the definition of dense subsets that there must be an object x* ∈ Supp( rx ) ∩ M* and an object y* ∈ Supp( ry ) ∩ M*. Therefore. 2.1 implies that S is connected. . n.103) For each i = 1. we pick two arbitrary objects x.

. ( mxi+1. 2. and ( m y n . y ∈ M Y and m x i . i = 1. n.y) ∈ E. R) is a system and E an equivalence relation on the object set M. R/E) i s connected iff for any two components X = (MX . Theorem 9.y) ∈ E. . i = 1.4. . . . RY) of S. m y i ∈ Mi . where µ : M → M/E is the canonical mapping. 2. . RX ) and Y = (M Y . n – 1 and Supp(q n ) ∩ Supp (ry ) ≠ Ø Applying Theorem 9. . . Suppose that the quotient system S/E = ( M/E. n – 1.109) For each i = 1. . . satisfying (x. Suppose that S = (M. for each i. 2. . there exist a natural number n and n components Si = (Mi . n – 1. m y i ∈ M i . . and (my n .m y i ) ∈ E. Then the quotient system S/E = (M/E. . Claim. . .2.R i ) of S. n. n –1 (9. for each natural number n. . . ( mx i +1 . Each maximal connected subsystem of S is called a component. 2. i = 1.7. R  Star ∞ ({ m})) of S is a component.111) and of the quotient system S/E are not identical. .m x 1 ) ∈ E. . n.112) (9. such that there exist x ∈ M X and y ∈ M Y and m x i . We then make a claim. n. 2. R i ) of S. .m x1 ) ∈ E. . 2. 2. . satisfying ( x. Then. RY ) of S such that there do not exist components S i = (Mi . . . let qi be a relation in R such that r i ⊂ q i  M * . i = 1. the partial system (Star ∞ ({ m}). The subsystems (9. . . . . for each object m ∈ M. . R X ) and Y = (MY. such that there exist x ∈ MX . Proof: Necessity. A connected subsystem S * of S is maximal in S if there is not connected subsystem S + of S such that S * is a subsystem of S + and S * ≠ S + .General Systems: A Multirelation Approach 221 and Supp(ri ) ∩Supp(r i + 1 ) ≠ Ø for each i = 1.1 we have showed that S is connected. R/E) is connected. i = 1. and there exist two components X = (MX . . . my i ) ∈ E. .4. Pick objects x ∈ MX and y ∈ M Y . Then x ∈ Supp( q 1 ) and y ∈ S u p p ( ry ) Supp (q i ) ∩ Supp ( qi +l ) ≠ Ø for each i = 1.110) (9.

. It suffices to show that Star∞ ({ µ (x)}) ≠ Star ∞ ( { µ (y)}). 2. . i = 1. Then there exist components (Myi . and m x i . for i = 1. .4. Suppose M/E ≠ Star ∞({µ (x)}). Then the Cartesian product system ∏ cp {Si : i ∈ I } is connected. Sufficiency. for some natural numbers n and m. .222 Chapter 9 This claim implies that the quotient system S/E is not connected. such that there are objects xi . . y ∈ M Y . . the connected component of S/E containing the object µ(m) must equal the whole system S/E. it suffices to verify that. R Y ). From Theorem . Let {S i : i ∈ I } be a set of systems such that there exists an i0 ∈ I such that the system Si 0 is connected. n and j = 1. Theorem 9. Choose and where m is an object in M. .4. . Suppose that for any two components X = (MX . R i ) of S . . To show that S/E is connected.R X ) and Y = (MY . i = 1. from the hypothesis that the quotient system S/E is connected. such that (x. m xi ) ∈ E. there exist a natural number n and components Si = (Mi . for i = 1. 2. . . n. . and . 2. for each object x ∈ M. . . i = 1.115) by Corollary 9. . where m is an object in M. n. . . . . 2. 2. . We now see the proof of Claim 1.R X ) and Y = (MY .R i ) of S. . . Suppose that (9. . contradiction which implies that M/E = Star∞ ({ µ( x ) } ) . for any n ∈ such that there exist m xi . m x i ∈ M xi and y j . m y i ∈ M i . (x i+1. x 1 ). . m y i ∈ M i . n. . . . . . n – 1. satisfying . n – 1. . . . This implies that there are two connected components X = (M X .114) for some x ∈ MX and some y ∈ M Y . i = 1. for j = 1. . 2. . satisfying . n and j = 1.8. i = 1. m – 1. This contradicts the assumption in the first paragraph of the necessary part of the proof. m. such that x ∈ M X and m ∈ M Y and for any two objects m x ∈ MX and m y ∈ M Y there do not exist connected components Si = (Mi . m yj ∈ M yj . . . . such that there exist x ∈ MX . . Star ∞ ({µ(x)}) = M / E (9. it follows that for each object µ (m) ∈ M /E. 2. z). . . m. . 2. . 2. 2. for each i = 1. RY ) of S. Then and and from the Cartesian product . We can then pick an object [m ] ∈ M/E – Star ∞ ({ µ (x)}). Proof: Pick objects system ∏cp {Si : i ∈ I}. . . . for each i = 1. To this end. . . n – 1. . In fact. 2.113) which we do by contradiction. n. It is clear that Star ∞ ({ µ (x)}) ⊂ M/E. 2.1. it suffices to show that (9.

. j = 1.4. Then each factor system Si is connected. . n –1. Let {S i : i ∈ I } be a set of systems such that the product ∏ {S i : i ∈ I } is connected. .4. Theorem 9. . .117) Therefore. n –1.121) for each j = 1. n.1 it follows that there are a natural number n > 0 and n relations rj ∈ R i 0 . such that (9. the relations satisfy and (9.2. 2.4.1 implies that Si is connected.General Systems: A Multirelation Approach 223 9.4. Proof: Pick two arbitrary objects xi and y i from a system S i for any fixed i ∈ I.119) for each i = 1. . .116) and (9. . …. it follows that there are a natural number n > 0 and n relations r j from the product system. such that x ∈ Supp ( r1 ) and y ∈ Supp(rn ) and (9. (9. 2. Then the relations pi (rj ) ∈ R i . Applying Theorem 9. . . .120) . There then are two objects x and y in the product system such that pi (x) = x i and p i (y) = y i . . satisfy (9. . From the hypothesis that the product system is connected.123) for each j = 1.122) and (9. …. .9. j = 1. 2.1 we have proved the theorem. j = 1. Theorem 9. . n. 2. . n. n–1. 2.

5. for any r. l t s T } or {S(t). ≤ ) . s ≥ t} of linkage mappings of S can now be defined.1. Let S be a nontrivial α -type hierarchy of systems. t are arbitrary elements in T satisfying s ≥ r ≤ t. s. t ∈ T with s ≥ t. l t s . T}.” For an α-type hierarchy S of systems. s ≥ t} of linkage mappings of S.e. for each t ∈ T. l t s . R t ) is a system.S( t) = S t = ( Mt . s ∈ T.1 [ZFC]. Rt ). From the Axiom of Choice it follows that there exists a choice function C : T → {M t : t ∈ T} such that C (t) ∈ M t for each t ∈ T.s ≤ T. s ≥ t} of linkage mappings is given. (9. denoted {S. ≤ ) is a function defined on T such that for each t ∈ T. An α -type hierarchy S of systems over the partially ordered set (T. An element α -type hierarchy S if lt s (xs ) = x t for all s. s ∈ T. and id St = id Mt is the identity mapping on the set Mt . is referred to as a linked α-type hierarchy (of systems) if a family {l t s : t.5. How many families of linkage mappings are there for a given α -type hierarchy of systems? Suppose {S. Question 9. An α-type hierarchy of systems S. each state S t is a nontrivial system. with r ≥ t such that and (9. T} is a linked α-type hierarchy of systems. where S t = ( Mt . Let be the subset of all threads. For any s. A family { l t s : t . The family { l t s: t..224 Chapter 9 9. and each mapping l ts a linkage mapping from S s into S t .125) . let l tr : S r → S t be a mapping from the system S r into the system St. Then there exists a family { l ts : t. t ∈ T. Proof: Suppose St = (Mt .5. T}. t ∈ T with s > t. i.s ∈ T. l t s .124) where r. R t ) is called a thread of the linked for all t ∈ T. Theorem 9. Without causing confusion. which will be called the thread set of the hierarchy { S. s ≥ t } is a family of linkage mappings of the α-type hierarchy S. Hierarchies of Systems Let (T. we omit the words “over the partially ordered set (T. let l t s (X) = C(t) for all x ∈ M s and l tt = id M t . ≤ ) be a partially ordered set with order type α. called the state of the α-type hierarchy S at the moment t.

130) . there exists a family { l ts : t. Suppose that S is an α -type hierarchy of systems over a partially ordered set (T.5. there exists exactly one α < κ such that x ∈ [ a α s ] Es . and an equivalence relation Et on the object set Mt such that Mt /Et = { [a] E t : a ∈ A t } and that for any distinct a. b ∈ At . where T * is a proper subset of T. s ≥ t} is a family of linkage mappings of the hierarchy of systems {S t : t ∈ T * }. there exist At ⊂ Mt satisfying  At  = κ '. t ∈ T with s > t.2. the relation ≤ t well-orders A t as follows: (9.  M t  ≥ κ'. Then for any nonzero cardinal number κ ' ≤ κ. Let κ be a cardinal number and S an α -type hierarchy of systems such that  M t  ≥ κ for each t ∈ T.129) for all s. and the Axiom of Choice. Proof: From the hypothesis that for any t ∈ T. For any x ∈ Ms . ≤ ).128) Then {l t s : t.126) Again. for any t ∈ T. where l t t = idS t . s ≥ t } of linkage mappings of S such that ∑ ≥ k'. t ∈ T * . Let D ⊂ T be a cofinal and coinitial subset in a partially ordered set (T.2 [ZFC]. s ∈ T. t ∈ T * with s ≥ t? Lemma 9. ≤ ) and { l t s : s.1 [ZFC]. Thus Question 9. (9. s ≥ t } is a family of linkage mappings.5. Under what of linkage mappings for conditions does there exist a family S such that (9. Then there exists a subset T * = D * ∪ D * of D such that _ + (a) ξ such that for some ordinals η and (9.5. and for any t ∈ T. for each t ∈ T. it follows that. Then define (9. from the Axiom of Choice it follows that there exists a set of well-order relations {≤ t : t ∈ T } such that.127) For any s.General Systems: A Multirelation Approach 225 Theorem 9. we now define a mapping l t s : S s → S t . s ∈ T.

let and d are incomparable under ≤ } (9. and d α is incomparable with d β under the relation <. Then D 0+ is a maximal antichain in D under ≤. dα < *d. β < η with α > β. ≤ * ). such that (i) {d α : α < η } is an antichain under the order <.226 Chapter 9 + then D β is a maximal – Then D β is a maximal (b) Let + antichain in D – Eβ for any β < η. note the construction of D + . 1 and D 0 ⊂ D such that and d 0 and d are incomparable under ≤ } If (9. α Suppose that sequences {d α : α < η } and {D0 : 0 ≠ α < η } have been defined. If is an antichain in D under ≤ . Let d 0 be the < * -initial element of D. for each d ∈ D 0 . and for all β < η.132) 1 otherwise. ≤ * ). By (iv) {dα : α < η + 1} is η an antichain under the relation ≤ . . and for each α < η. for all d ∈ D 0 . be defined by If η = ξ + 1 is isolated. d α is the initial element of the well-ordered set (D 0 .133) . (d) T * is cofinal and coinitial in T. If η If be the initial element of (D 0 . let d 1 be the initial element of (D 0 . If η is a limit ordinal. for each β with α < β < η.134) + If then D 0 is a maximal antichain in D under the let d η be the initial element of (D0η. * Proof: First. η d α and d are incomparable under ≤ for each d ∈ D 0 . < * ). From the Axiom of Choice it follows that there exists an order relation ≤ * on T which well-orders T as (9. < * ).131) where ≤ and ≤ * are generally different. let (9. For each α < η + 1. d α < * d for each d ∈ D 0 . (c) Let – antichain in D – E β for each β < ξ. α (iii) For each α < η. for some ordinal η > 1. β (iv) For each α < η . Then [d α : α < η + 1} order ≤ . (ii) for all α.

for any distinct a. such that if a subset with property (iv) can be defined in the way discussed earlier. t ∈ T * with Proof: Let T * ⊂ D be the subset defined in the proof of Lemma 9.5. There are two possibilities: (1) t ∈ TL and (2) t ∈ T R . (9. . Case 1: t ∈ TL . ≤ ). Using Let transfinite induction to consider sets such that x > d } and and applying the same technique we used before. exist such that properties (a)–(d) in the lemma hold. If such that From the If α = δ + 1 is isolated. where.t ∈ D with s ≥ r > t. sequences {dα : α < η } and with properties (i)–(iv) can be defined.137) and pick one t ∈ T. there exists there exist Axiom of Choice it follows that for any s ∈ T satisfying an equivalence relation Es on M s an A s ⊂ M s such that and satisfies. Then there exists a subset T * ⊂ D such that (i) T* is cofinal and coinitial in T. we can show that ordinals η and ξ and subsets and 0 ≠ β < ξ. it must be that Dη = Ø 0 + Then D 0 is a maximal antichain in (D. b ∈ As .136) and (9. Let us rewrite T = TL ∪ T R .General Systems: A Multirelation Approach 227 d α < * d. Theorem 9.5.135) for any s. and d α is incomparable with d ∈ D β under ≤ for any β. for all d ∈ D β . ≤ ) such that there exists a cofinal and coinitial subset is a family of linkage mappings for the D ⊂ T such that hierarchy of systems {S t : t ∈ D } satisfying (9. There then exists an ordinal α < ξ such that there exists and assume that α is the initial one in ξ with this property. Suppose that S is an α -type hierarchy of systems over a partially ordered set (T.138) of linkage mappings for S. Now we assume that eα < t. 0 0 with α < β < η + 1.1. such that (9.r. (ii) There is a family (iii) For any s. By transfinite induction.3 [ZFC].

144) From the Axiom of Choice it follows that for any s ∈ T α .140) We now define mappings for all s.143) If α is a limit ordinal. for any s ∈ T α . then by definition α > 0. and the relation < eα well-orders as (9. with e α ≤ s < t < e δ as follows: For any x ∈ M t there exists exactly one a t i ∈ A t such that If s > e α .228 Chapter 9 and a set of well-order relations such that. let (9. and the relation < e α well-orders as follows: (9.147) .146) We now define mappings any x ∈ M j there exists exactly one for all such that as follows: For let (9.141) If s = e α. the relation < s well-orders Ms /E s as follows: (9. Let (9. there exists an As ⊂ M s – satisfying where q ∈ D 0 satisfies t < q. let (9. the relation < s well-orders M s /E s as follows: (9.142) Define mappings for all s ∈ T with e α < s < e δ as follows: (9. for any distinct Es on Ms exists such that be a set of order relations such that.145) where for all i.139) where follows: for all i < As. for any s ∈ T satisfying eα < s < e δ. An equivalence relation satisfies. t ∈ T.

define (9. each hierarchy S of systems over T is called a treelike hierarchy of systems.149) is a family of linkage mappings. A treelike hierarchy B of systems over a tree T is a partial treelike hierarchy of a treelike hierarchy A of systems over the same tree T if for each t ∈ T. then there exists exactly one d δ ∈ D – such that d β < t ≤ d δ . let (9. β < ξ such that there exist d α ∈ D α and d β ∈ D β . t ≥ s} of linkage mappings such that for any s. T'} be two hierarchies of systems. lt s . if t s is undefined where – above. A similar method can be used to define a family T. for each t ∈ T the set {y ∈ T : y ≤ x } is a chain). t ∈ T * with s ≥ t. . A mapping from the hierarchy H into the hierarchy H' is a set {φ. such that the diagram in Fig.e. A mapping {φ . each linked hierarchy {S.150) (2) If β is a limit ordinal. t' ∈ T' satisfying s' ≥ t'. for any s. There exists exactly one q ∈ D 0 such Then is a family of that t ≤ q ≤ s. linkage mappings for S such that Let H = {S. Then it is easy to check that Furthermore. 9. with d β < t < d α < s.. l t s .148) Define mappings for all s ∈ T α and all δ < α by (9.General Systems: A Multirelation Approach 229 If j = e α . + + Now let s ∈ T R and T L satisfying s > t..ƒt '} from H into H' is a similarity mapping if φ is a similarity mapping from T' onto T and each is a similarity mapping.2 commutes for any s'. then there exist α. Define for all s. and for each t' ∈ T' . It is easy to show that is a similarity mapping from H' into H. the state B t is a partial system of the state At .ƒt '} of mappings. t ∈ D * with s ≤ t.ƒt' is a mapping from Sφ ( t ') into S't' such that (9. where φ is a nondecreasing mapping from T' into T with φ (T') cofinal in T. Define δ (9. ≤ ) is a pseudo-tree (i. T } and H' = {S'. (1) If β = δ + 1. l t's' .e.152) i. t ∈ T L with s > t. When the partially ordered set (T. T} of systems is called a linked treelike hierarchy of systems.151) Case 2: t ∈ T R . Let α and β be the least ordinals with this property.

T } be a linked treelike hierarchy of systems such that the partially ordered set (T. lt s . (b) D * is a maximal antichain in D – Eβ . (ii) There exists a cofinal subset D ⊂ T such that l t s is an onto mapping for all s.5. By transfinite induction. Let {S.1 it follows that there exists D* ⊂ D such that (a) for some ordinal η > 0. T} Then there exists a nontrivialpartial linked hierarchy such that the state S' ( t ) is embeddable in the state S'( s ) for all t. Theorem 9.4 [ZFC].153) for each m ∈ M d0 . for all r ∈ T and t 0 ∈ T 0 satisfying r ≥ t 0 .t ∈ D ∪ T0 with s ≥ t .s ∈ T with s ≥ t. Let ≤ * indicate the order relation on T defined in the proof of Lemma 9.5. Proof: From Lemma 9.154) . where β (c) D * is cofinal in T.1.2. We define an equivalence relation E d0 on M d0 such that the equivalence class (9. ≤ *). l t s . ≤ * ) an equivalence relation E d on the object set Md of the state Sd can be defined such that the equivalence class (9.5. there exists at 0 ∈ T 0 such that t0 ≤ t.230 Chapter 9 Figure 9. From the hypothesis that l t s is onto for any s.t ∈ D ∪ T0 with s ≥ t. it follows that each linkage mapping l t 0 r is onto. A hierarchy mapping guarantees the diagram is commutative. for each element d in (D * . ≤ ) satisfies (i) T has a subset T 0 consisting of all minimal elements in T such that for each t ∈ T. of {S. for each β < η . Let d 0 be the initial element of the well-ordered set (D* .

5.r ∈ D * with r ≥ t . and define (9. hierarchies of systems.160) Then is a hierarchy of systems and a partial linked hierarchy of Let X be the set of all branches in the set D * T0 . for any t. Therefore. (9. for each d ∈ B . it follows such that that there exists an element (9. (9. By transfinite induction. for each [mr ] Er ∈ M r / Er . for all s.1 implies that (B. ≤ ) is well-ordered and AC.158) for all i.155) where [m t ] E t ∈ M t / E t satisfies l t r (mr ) = m t . can be defined such that for any branches C . From AC it follows that there exists a subset such that. be the projection. Let b be the initial element in B. called a branch. can induce a mapping from S r / E r onto S t / E t such that (9.t ∈ B with s ≥ t. In fact. (9.161) Here we used the hypothesis that T is a pseudotree. Let P d : M (B) → Md .B ∈ X and d ∈ C B.157) and that there exists an element such that (9.General Systems: A Multirelation Approach 231 The linkage mapping l t r . from the fact that (B. For any fixed element x ∈ M b . For any maximal linearly ordered set B ⊂ D * T0 .159) where l t s (gs ) = g t . there exists exactly one [m t] E t ∈ M t / E t such that (9.j ∈ B with i ≥ j.162) . Lemma 9. for each d ∈ D * we can define T0 . ≤ ) is well-ordered. for each x ∈ Mb . Therefore.156) and is well defined. for each B ∈ X .

where α is the For any t ∈ T – ( D * α least ordinal in η with this property. if n = 0 if n > 0 (9. T } of systems is called a linked time system. Suppose (T. We let 0 be the identity in T. For any integer n . An ordered triple (T. T is called the time set (or time axis) of the time system S. defined by x = max{ – x.166) if n < 0 . We take a partial system S' of the t system St defined by (9. choose s ∈ D * such that s ∈ D * . Then S ' d is a partial system of S d and is a nontrivial linked hierarchy of systems such that the state S’ (t ) = S't is embeddable in the state S’( s) = S’ . l t s. This kind of time system is called a system over a group.163) Then it is not hard to check that is a nontrivial partial linked hierarchy of {S . each linked hierarchy {S . and. T } such that the state S’ (t ) is embeddable in the state S’(s ) for any s. ≤ ) is an ordered Abelian group. +) is an Abelian group and (T.x } where –x is the additive inverse of x. each hierarchy S of systems over T is called a time system. ≤ ) is a linearly ordered set such that for any elements (9. ≤ ) is a total ordering. with s ≥ t. If the time system is also a linked time system. for any x ∈ T. and s > t. l t s.t ∈ T .232 Chapter 9 where B is a branch containing d. for any t. ≤ ) is called an ordered Abelian group if (T. +. we call the system a linked system over a group. When the partially ordered set ( T.165) (9. In this case.s ∈ D * T0 satisfying s t ≤ s. x indicates the absolute value of x. +.164) We conclude this section with a discussion of time systems over ordered Abelian groups. T0 ).

Suppose S – k p = S for each integer k satisfying 0 ≤ k < –n. for each t ∈ T.5. the left transformation of S.5. Then for any a. the result is clear. S np = S.5. Case 2: n < 0. denoted S +t 0 . by induction. Theorem 9. If n = –1. Then it follows that . Suppose that S is a system over a group. Then it follows that Thus. is defined by S – t0 (t) = S ( t – t 0 ) for each t ∈ T .General Systems: A Multirelation Approach 233 The following notation will be used: Let S be a system over a group and t0 ∈ T a fixed element. Then S (x) = S (x – p ) for each x ∈ T.6. If n = 0. is defined by S +t0 (t ) = S( t + t 0 ) for each t ∈ T. The right transformation of S. we have S +np = S for each n ≥ 0. and S np = S – [(– n) p ] if n < 0. If S is periodic with p ∈ T as its period. Let S be a system over a group T. where S np = S +np if n ≥ 0. let t = x – p. Proof: We show the theorem by induction on n. in the condition that S( t + p ) = S ( t ). and A system S over a group T is periodic if there exists p ∈ T such that p ≠ 0 and S +p = S.b ∈ T. The element p is called a period of S. denoted S –t 0 . Case 1 : n ≥ 0. This says that S – p = S . Suppose S +kp = S for each integer k satisfying 0 ≤ k < n. Theorem 9. then for each integer n.

Corollary 9.234 Chapter 9 Hence. In fact. If a system S over a group is periodic. lt s . for a periodic system {S.5.5. The proof is straightforward and is omitted.ƒx } is a similarity mapping from the hierarchy H into the hierarchy K. Then the hierarchy for any integer n. . Theorem 9. l t s .T } over a group. l t s .168) onto the ordered Then φ is a similarity mapping from the ordered set for each as set We next define mappings identity mappings. l t s. Theorem 9.5. T } over a group is periodic is similar to the with > 0 as a period. if n (9. l t s .7. In Theorem 9. Combining cases 1 and 2 completed the proof.5. then so is n for each integer n ≠ 0. if S is a periodic system over a group with a period such that for any t.2. Corollary 9. s ∈ T satisfying s ≥ t. hierarchy Proof: First we show that n < (n + 1) ≥ ( n + 1) for some integer n. T} . Suppose that a linked system {S. If is a period of a linked system {S.167) This contradicts the hypothesis that Second. T } over a group is periodic. This element in T is called a period of the linked system {S. A linked system {S. there then is a positive period for S. we define a mapping > 0. the hierarchies and are not similar. by induction we have showed that Snp = S for each integer n < 0.8.8. by (9.1. then for each integer n.5. the ordered sets replaced by each of the following pairs: and can be Generally. T } over a group. Then { φ .

174) Suppose V is a set and G a mapping such that G : Dom( X) × Ran( Y ) → V (9. y) and will be used congruently. In this section we study different concepts of controllability not only of input–output systems but also of general systems.173) and (9. we then can find some input such that the corresponding output satisfies the required properties. In the following.6. .169) (9. if we need to obtain some output of the system with certain predetermined properties. (9.171) for any j ∈ o ( Y ).170) for any i ∈ o( X). where and We now define (9. Let S = ( M. (9.General Systems: A Multirelation Approach 235 9. for any x ∈ Dom( X ) .176) The mapping cg is called a control function of S. Controllabilities The concept of controllability of systems was originally introduced as a property of input–output systems.172) and (9.175) For an element v ∈ V. the symbols (x .R ) be an input–output system with input space X and output space Y. the system S is v -controllable relative to G if and only if there exists a mapping cg : Dom( X ) → Dam( X ) such that. The main idea behind the concept says that an input–output system S is controllable provided that. there exists a y x ∈ Ran(Y ) such that (9.

For a subset V' ⊂ V. y) = v (9. Ran(Y)) ⊃ V'. we could choose various methods (denoted by x) for processing the wanted results. The input–output system S is controllable on V' relative to G : Dom(X) × Ran(Y ) → V iff there exists a mapping cg : Dom(X) → Dom( X) such that. suppose we are processing a chemical reaction.177) . y) = v Let V' ⊂ V be a subset. each v ∈ V' and any w ∈ Dom( X). Let cg : Dom(X) → Dom( X) be a control function of the system S. Theorem 9. Define a control function cg : Dom(X) → Dom( X) by letting cg(w) = x for all w ∈ Dom( X). Hence. y z ) = v The mapping cg is called a control function of S.6. y) = v.2. Proof: The necessity part of the proof follows from Eq. Sufficiency. (9. there exists a y ∈ Ran( Y ) such that G(cg(w). Proof: Necessity. G (x.1.176). Suppose there exists an element x ∈ Dom(X) such that G ({x } × Ran(Y )) ⊃ V'. y ) = G(x. we may add relative materials [denoted by cg(x)] together with those “choices” we have had. Then for each w ∈ Dom(X) there exists yw = y ∈ Ran(Y ) such that G(cg(w). for each v ∈ V' and each x ∈ cg(Dom( X)). To reach a collection of desired results (denoted by yx ) which satisfy an assigned target (denoted by v). in order to control the whole reaction procedure. Sufficiency. x) ∈ Ran(Y ) such that G(cg(x). Define a control function of S . For each element v ∈ V. cg : Dom(X) → Dom( X). there exists a y ∈ Ran( Y ) such that G (x.6. But.236 Chapter 9 For example. y) ∈ Dom(X) × Ran( Y ) such that G(x. Clearly. Let v ∈ G (Dom(X) × Ran( Y )). for each v ∈ V' and any x ∈ Dom( X). Then. as follows: cg(w) = x for all w ∈ D o m ( X). Then there exists (x. yw ) = G(x. y) = v. the input–output system S is vcontrollable relative to the mapping G iff v ∈ G(Dom(X) × Ran( Y )). there exists yz = y (v. We would like to deem the “input” (the relative materials) as “control” for the whole reaction procedure. Theorem 9. the input–output system S is controllable on V' relative to the mapping G iff there exists an element x ∈ Dom(X) such that G({x} × Ran( Y )) ⊃ V'.

there exists y ∈ M such that (x. we define the µ-initial section of r. Then for any v ∈ V' there exists (z. we study the concept of controllability of general systems. y) ∈ r. Suppose there exists a nonempty subset A ⊆ Dom( X) such that G(A × Ran(Y )) ⊇ V'. y) ∈ A × Ran( Y ) such that G(z. there exists a y ∈ Ran( Y ) such that G(z. S is controllable on V' relative to G.General Systems: A Multirelation Approach 237 Therefore.182) . This implies that. R) is a general system. For any relation r ∈ R and any ordinal number µ > 0.180) (9. Let cg : Dom(X) → p (Dom(X)) be a control function for S. Let (9. x) ∈ Ran(Y ) such that G(z. Sufficiency. Proof: Necessity. We define nonzero ordinal number such that µ + v = n(r) such that r ⊂ M v the µ-final section of r.178) Theorem 9. y) = v. where cg(x) ≠ Ø. there exists y = y(v. the input–output system S is collectively controllable on V' relative to the mapping G : Dom(X ) × Ran(Y ) → V iff there exists a mapping cg : Dom(X) → p(Dom(X)). denoted by E(r µ ). For V' ⊂ V. Let us define a control function cg : Dom( X) → p (Dom(X)) as follows: cg(x) = A for all x ∈ D o m ( X). the input–output system S is collectively controllable on V' relative to the mapping G : Dom(X ) × Ran( Y ) → V iff there exists a nonempty subset A ⊆ Dom(X) such that G(A × Ran(Y )) ⊇ V'. there exists y ∈ I(r µ ) such that (y.179) (9. as follows: for any x ∈ E(r µ ) ⊂ M .6. This completes the proof of the theorem. such that for any v ∈ V' and any x ∈ Dom( X).3. y) = v for some z ∈ cg( x ) (9. where v is a n ( r) . x) ∈ r.181) and and (9. as follows: v For any x ∈ I(r µ ) ⊂ M µ . For a subset V' ⊂ V. there exists an x ∈ Dom(X ) such that G(cg(x) × Ran(Y )) ⊇ V '. Hence. where p(Dom(X)) is the power set of Dom(X). denoted by I( rµ ). x) ∈ Ran(Y ) such that G(z. In the following. for any v ∈ V' and any x ∈ Dom(X). y) = v for some z ∈ cg( x). Examples can be given to show that not all systems are input–output systems. there exists y = y(v. Then for any v ∈ V' and each x ∈ Dom( X). Suppose S = (M. y) = v.

187) Let and Then (9.185) such that for each u ∈ W v there exists a yu ∈ E R such that for some (9. and a mapping (9.4.186) The mapping C v is called a control function of S.” Theorem 9. and (9. (9. such Then for each v ∈ V' there exist M v ⊂ M and a nonempty A v that (9. Let V be an another set and V' ⊂ V. Then S is generally controllable on V' relative to a mapping iff there exists a nonempty subset MV ' ⊂ V such that for some nonempty Proof: Necessity. there exist a set Wv. The general system S is generally controllable on V' relative to a mapping (9.187) we have G( A × E R) ⊃ V'. When S receives the commands. Therefore. Suppose S is generally controllable on V' relative to G.184) if.183) are used to indicate all the elements in I R and all the elements in E R. Suppose we are given a set of desired targets V. there will be some “well-organized” objects of S to accomplish the task.238 Chapter 9 If M * ⊂ M. an M v ⊂ M. Let S be a general system and V' ⊂ V a subset. The main idea of general controllability of systems is the following. there exists a set of “commands” given to a system S so that S can realize the desired target v. respectively. for each v ∈ V'. in which only the elements of M* appear as coordinates. .6.188) and from Eq. For each v ∈ V. there should be some “receivers” and “accomplishers” in S such that the whole process of executing the task is under a good control with a “fine” cooperation between the “receivers” and the “accomplishers.

General Systems: A Multirelation Approach

239

Sufficiency. Suppose there exists a nonempty subset M V' ⊂ M such that for Thus, for each v ∈ V' there some nonempty are subsets Av ⊂ A and B v ⊂ ER such that G(A v × Bv ) = {v} (9.189)

ON (m = x µ )}. Let W v be any Define M v ⊂ M by nonempty set and a mapping defined by C v ( u) = A v for each u ∈ W v . Then for each u ∈ W v there exists a y u ∈ Bv ⊂ E R such that G(z,y u ) = v This proves the assertion. Proposition 9.6.1. Suppose that S = (M, R) is an input–output system with input space X and output space Y, and that S is v-controllable relative to a mapping G : Dom(X) × Ran( Y ) → V. Then S is generally controllable on {v} relative to the mapping where G * is an extension of G. Proof: Because Dom(X ) ⊂ I R and Ran(Y ) ⊂ E R, we can define extensions for the mapping G. We now show the general controllability of S. For v ∈ { v} there exist a set Wv = Dom(X) and a mapping (9.190) defined by C v (u) = {cg(u)}, where cg : Dom(X) → Dom(X) is a control function for S. Hence, for each u ∈ W v there exists a y u ∈ Ran(Y ) ⊂ E R such that (9.191) where cg(u) ∈ Cv (u). Proposition 9.6.2. Let V' ⊂ V be a subset such that an input–output system S = (M, R) is controllable on V' relative to a mapping G : Dom( X) × Ran( Y ) → V , where X and Y are the input and the output spaces of S, respectively. Then S is generally controllable on V' relative to each extension G* of G. Proof: For each v ∈ V' there exist sets W v = Dom(X) and M v = M and a mapping (9.192) for some z ∈ Cv (u)

240

Chapter 9

defined by Cv (u) = {cg(u)} for each u ∈ Wv . Then for each u ∈ Wu , there exists yu = y(v, u) ∈ Ran( Y ) ⊂ E R such that (9.193) where cg(u) ∈ C v (u). Proposition 9.6.3. Suppose that S = (M, R) is an input–output system with input space X and output space Y such that S is collectively controllable on V' relative to a mapping G : Dom(X) × Ran(Y ) → V. Let be an extension of G. Then S is generally controllable on V' relative to G * . Proof: For each v ∈ V' there exist sets W v = Dom(X) and M v = M and a mapping (9.194)
defined by Cv (u) = cg(u) such that, for each u ∈ W v , there exists yu = y (v, u) ∈ Ran(Y ) ⊂ E R such that

G * (z, y z ) = G (z,yu ) = v for some z ∈ C v ( u ) = cg(u).

(9.195)

Theorem 9.6.5. Suppose that h : S1 = (M 1 , R 1 ) → S = (M, R) is an S-continuous mapping from a system S1 into a system S such that S is generally controllable on V' relative to a mapping Then S1 is generally controllable on V' relative to a mapping Proof: First, we construct a desired mapping Let v0 ∈ V be a fixed element. For any x ∈ I R 1 and y ∈ E R 1 , if there exist relations r x , ry ∈ R such that and then define G1 (x,y) = G (h(x),h(y)) If either rx or r y or both do not exist, we define G 1 (x,y) = v 0 (9.198) (9.197) (9.196)

General Systems: A Multirelation Approach

241

is well defined. It can be shown that Second, we show that S 1 is generally controllable on V' relative to Suppose that for each v ∈ V' there exist sets W v and M v ⊂ M and a mapping Cv : W v → p(I R(Mv )) such that, for each u ∈ W v , there exists a yu ∈ E R such that G (z,yu ) = v for some z ∈ C v (u). We now let h– 1 (Mv ) ⊂ M 1 and define as follows: For each u ∈ W v , a control function (9.199) Now it only remains to show that ¹ Cv is a control function of S1 . In fact, for each u ∈ W v there exists y u ∈ E R such that G( z,yu ) = v for some z ∈ C v (u) so there exists such that (9.200)

where to G1 .

Therefore, S 1 is generally controllable on V' relative

Theorem 9.6.6. Let {S i : i ∈ I } be a set of systems, where I is an index set, such that there exists an i 0 ∈ I such that the system S i 0 is generally controllable on V' relative to a mapping Then the Cartesian product system Πcp {S i : i ∈ I } is also generally controllable on V' (relative to a different mapping where it is assumed that ( M,R). The proof follows from Theorems 9.3.10 and 9.6.5. Theorem 9.6.7. Suppose that h : S = (M, R) → S 1 = (M1 ,R 1) is an embedding mapping from a system S into a system S 1 such that S is generally controllable on V’ relative to a mapping G : I R × E R → V. Then, S 1 is generally controllable on V’ relative to a mapping G 1 : I R1 × E R1 → V. Proof: We first construct a desired mapping Let v 0 ∈ V be a fixed element. For any x ∈ IR1 and y ∈ E R1 , if there exist x * ∈ I R and y * ∈ E R such that h(x* ) = x and h (y* ) = y, we define G 1 (x,y) = G (x* ,y* ) (9.201)

242

Chapter 9

From the hypothesis that h is 1–1, we see that Eq. (9.201) is well-defined, for the existence of x * and y * is unique. If one or both of the elements x * and y * do not exist, then we define G 1 (x,y) = v 0 (9.202)

We then have defined a mapping Second, we show that the system S1 is generally controllable on V' relative to the mapping G 1 . Suppose that for each v ∈ V' there exist sets W v and M v ⊂ M and a mapping such that, for each u ∈ W v , there exists a y u ∈ E R such that G(z,yu ) = v for some z ∈ C v (u). We now let h(M v ) ⊂ M 1 and define a control function as follows: For each u ∈ W v , (9.203) It remains to show that 1¹ C v is indeed a control function for S1 . For each u ∈ W v there exists a y u ∈ E R such that G(z,yu ) = v for some z ∈ C v ( u) Therefore, (9.205) where and (9.204)

Corollary 9.6.1. If a subsystem A = (M A , RA ) of a system S = (M, R) is generally then S is generally controllable on V’ relative to a mapping controllable on V' relative to a mapping Theorem 9.6.8. Let { Si = (Mi , Ri ) : i ∈ I } be a set of systems and {V i ⊂ V : i ∈ I} a set of nonempty subsets of V. If each system S i is generally controllable on the subset V i relative to a mapping then ⊕ {Si : i ∈ I } = (M , R) is generally controllable on {V i : i ∈ I} relative to a mapping where for each i ∈ I and any

Proof: From the definition of free sums of systems, it follows that (9.206) for any distinct i,j ∈ I. Let G be an extension of where the mappings {G i : i ∈ I}; i.e., G is a mapping from I R × E R into V such that for any then G(x,y) = G i (x,y).

General Systems: A Multirelation Approach

243

To show that the free sum ⊕ {Si : i ∈ I} is generally controllable on { V i : i ∈ I} relative to G, we pick an arbitrary v ∈ { V i : i ∈ I }. Then there exists an i ∈ I such that v ∈ V i . Therefore, there exist sets Wv and M v ⊂ Mi ⊂ {Mj : j ∈ I} and a mapping

(9.207) such that for each u ∈ Wv , there exists a yu ∈ E Ri ⊂ E R such that G(z,yu ) = v for some z ∈ C v ( u) This proves the theorem. In the following, we conclude this section with a discussion of controllabilities of time systems. Theorem 9.6.9. Suppose that {S , l t ∈ T such that
t s, T }

(9.208)

is a linked time system. If there exists a

(a) S t is a nondiscrete input–output system, (b) (c) for each s < t, for each s > t,

then each state Ss of the linked time system has a nondiscrete partial system which is an input–output system. Proof: Let S t = ( M t , Rt ) and X t and Y t be input and output spaces of the state S t , respectively. If s < t, let be a partial system of the system S s defined by Then is a nondiscrete input– output system with input space ls t (X t ) and output space ls t (Y t ). Finally, if s > t, let be a partial system of S s defined by R s : l t s(r) ∈ R t }. Then is a nondiscrete input–output system with input space X s = {x ∈ Ms : l t s (x ) ∈ X t } and output space Ys = {y ∈ M s : l t s (y) ∈ Yt }. Theorem 9.6.10. Suppose that {S, l t s , T } is a linked time system such that (a) T has an initial element t 0 . (b) Each lt 0 s is surjective with l t 0 s (Rs ) ∩ R t 0 ≠ Ø. (c) S t 0 is a nondiscrete input–output system with input space Xt 0 and output space Yt 0 . (d) St 0 is v-controllable relative to a mapping G : Dom(Xt 0 ) × Ran( Yt 0 ) → V.

244

Chapter 9

Then there exists a linked time system such that each state is a nondiscrete input–output system and is v-controllable relative to some mapping Gt . Proof: First we see the construction of the linked time system AC we well-order the time set T t0 , t1 , . . . , t α , . . . , α<  T  By (9.209)

where t 0 is the initial element in T. Suppose that the original order on T is ≤ , and in Eq. (9.209) T is well-ordered by a well-order relation ≤ * . Generally, ≤ and < * are different. Define Suppose that for some α <  T  the system have been defined such that for each and the mappings Now define and k < α and any i, r, j < α such that and and as follows: Let Et α be the equivalence relation on the object set M t a defined by and (9.210) where if t α > t β, define and For each β < α , (9.211) for each where If t α< t β , we define (9.212) for each where It is then not hard to prove that is a family of linkage mappings. By transfinite induction, a linked time system is defined such that is a nondiscrete input–output system with input space and each state and output space Second, we show that each state is v -controllable relative to some mapping (9.213) Let Dom and Ran exists x t 0 ∈ Dom (Xt 0 ) such that For any x t ∈ Dom(X t ) there (9.214)

General Systems: A Multirelation Approach

245

where cgt 0 is a control function of St 0 . Thus, there exists a y t 0 ∈ Ran(Y t 0 ) such that (9.215) Let then (9.216) and tem Therefore, the sysis v-controllable relative to the mapping G t . and If we assign

Theorem 9.6.11. Assume hypotheses (a)–(c) of Theorem 9.6. 10 and that S t 0 i s controllable on V’ ⊂ V relative to a mapping G : Dom (X t 0 ) × Ran (Y t 0) → V. Then is a nondiscrete such that each state there exists a linked time system input–output system and is controllable on V' relative to some mapping Gt . The proof is a slight modification of that of Theorem 9.6.10 and is omitted. Theorem 9.6.12. Assume hypotheses (a)–(c) of Theorem 9.6.10 and that St 0 is collectively controllable on V* ⊂ V relative to a mapping G : Dom (X t 0 ) × Ran (Y t 0 ) → V. Then there exists a linked time system such that each state Q t is a nondiscrete input—output system and is collectively controllable on V* relative to some mapping Gt . The proof is a slight modification of that of Theorem 9.6.10 and is left to the reader.

9.7. Limit Systems
We say that is a category if consists of a class of objects, denoted by obj pairwise disjoint sets of morphisms, denoted by Hom (A, B), for every ordered pair of objects (A, B), and a composition Hom (A, B) × Hom (B, C) → Hom (A, C), denoted by (ƒ, g) → gf, satisfying the following axioms: (i) For each object A there exists an identity morphism I A ∈ Hom (A,A) such that ƒ I A = ƒ, for all ƒ ∈ Hom (A, B) and I A g = g, for all g ∈ Hom (C,A). (ii) Associativity of composition holds whenever possible; that is, if ƒ ∈ Hom (A,B), g ∈ Hom (B,C) and h ∈ Hom (C,D), then h(gƒ) = (hg)ƒ (9.217)

homomorphisms whenever i ≤ j satisfying the univerwith sal mapping property. Universal mapping property of limit systems. homomorphisms) making the corresponding diagram commute whenever i ≤ j. in i. . contained in Fig. Let be an α-type hierarchy of systems in (respectively. contained in Fig. the universal mapping property.3.. denoted by lim or or lim . Thus.3 (respectively. for any two systems S1 and S 2 the equals the set of all S-continuous mappings from S 1 into S 2 . for two arbitrary systems S 1 and S 2 the set Hom of morphisms equals the set of all homomorphisms from S 1 into S 2 . a homomorphism). each linkage mapping l t s is an S -continuous mapping (respectively. The inverse limit of this hierarchy. a homomorphism making the diagram commute. there exists a unique S-continuous mapping (respectively.e. We give some examples of categories. Suppose is the category consisting of all general systems and all Scontinuous mappings between systems. is a system and a family of lim ← C St (respectively. 9. Universal mapping property of limit systems.4) satisfying: For every system X and S-continuous mappings (respectively. set Hom b. Thus. Figure 9. 9.4. a . lim S-continuous mappings (respectively. Suppose is the category consisting of all general systems and all homomorphisms between systems.246 Chapter 9 Figure 9.

9. Then it can be checked that the mappings αt are S -continuous and satisfy whenever t ≤ s.220) for every object and any i ∈ T. The uniqueness of β follows from the commutative diagram in the theorem or Fig. for each t ∈ T. follows that We now show that is S -continuous. R Y) be a system and S-continuous mappings such that whenever t ≤ s. the limit is unique up to a similarity.7. From the definition of S -continuity it follows that is S continuous.218) and (9. Let be a linked α -type hierarchy of systems in the category Then the inverse limit lim exists.General Systems: A Multirelation Approach 247 From the definition of inverse limits.3. . a family of We now let Y = (M Y .219) and define mappings α i : lim ← C St → Si by letting (9. Theorem 9. For any relation r ∈ R there exist j ∈ T and r j ∈ Rj such that Therefore. Then from the hypothesis that is well defined.221) whenever t ≤ s. it can be seen that if a linked hierarchy of systems has an inverse limit. Define a mapping by letting (9. Define a desired system lim Proof: Let (M.1. it for each y ∈ M Y .R) as follows: (9.

227) for each t ∈ T. then .223) and for any define n = n (r t )..R) is the inverse limit of the linked hierarchy Only two things need to be proven: (i) For any system and any set of homomorphisms such that (9. Let be a linked α -type hierarchy of systems in the exists. for t ∈ T. define a homomorphism from into S.R) into S t . and (9. there exists a homomorphism so that (9.7.7.224) and (9. Then lim ← C S t is a partial system of The proof follows from the proofs of Theorems 9. be the projection.224).225) Then. Thus. To prove (i). Let α t : M → M t .3.226) whenever t ≤ s.2.248 Chapter 9 Corollary 9. That implies that each α t is a homomorphism from (M.10. for each t ∈ T. Theorem 9. Suppose that is a linked α-type hierarchy of systems such that each linkage mapping l t s is S-continuous.1. Since for any relation Then β is . for any relation r ∈ R.e.1 and 9. (9. i.7.R ) as follows: (9. category Then the inverse limit lim Proof: We define a system lim ← h S t = ( M.222) and (9. for each t ∈ T. there exists such that r is defined by Eq. We now show that the system (M.

moreover. Thus. fs ' } of a linked hierarchy H = of of systems induces a mapping systems into a linked hierarchy H' = from lim ← H into lim ← H'. Let for all for all t ∈ T. for any relation and any so as mappings from the relation set into the relation This implies that for each t ∈ T. This implies that β ' = β on the set . systems into systems. If. The mapping ƒ : lim← H → lim ← H' in the proof of Theorem 9. is another homo(ii) The homomorphism β is unique.3.228) The element thus obtained is an object in lim ← H'. as homomorphisms from set R t for each t ∈ T. ƒ is also an onto mapping. so β ' and β induce the same homomorphism. we have for each t ∈ T. Each mapping {g . Hence.4. as mappings between sets. all mappings ƒs ' are onto. . Suppose Then morphism satisfying for each t ∈ T. Indeed.7.3 is called the limit mapping induced by {g.7. Hence. Theorem 9. On the other hand.ƒ s ' } is also one-to-one. Then ƒ is a mapping from lim ← H into lim ← H'. Let {g. define (9. where lim ← indicates either lim← C or lim ← h . for any s'. t' ∈ T' satisfying t' ≤ s'. then For the object x in lim← H define ƒ(x) = y in lim ← H'. Proof: For an object x = (x t ) t ∈T in lim← H and every s' ∈ T'.7.ƒ s '} be a mapping of a linked hierarchy of systems H = to a linked hierarchy of systems H' = If all mappings ƒ s ' are one-to-one.General Systems: A Multirelation Approach 249 Since for all . Theorem 9.ƒ s' } and is denoted by ƒ = lim ← {g. β ' = β as homomorphisms.ƒ s'}. the limit mapping ƒ = lim ← {g.

the limit mapping ƒ is S-continuous.3 is well defined. the limit mapping ƒ is one-to-one. Let r be a such that (9.5. is Theorem 9.ƒ s '} is a mapping of a linked hierarchy of systems H= into a linked hierarchy of systems where H and If all mappings ƒ s ' are S-continuous. Proof: Suppose lim and lim implies that the limit mapping relation in lim ←C H'.250 Chapter 9 be two distinct objects of the limit Proof: Let and and an s' ∈ T' satisfying system lim← H.230) It can be verified that xs does not depend on the choice of s'.7. For any s ∈ T choose an element s' ∈ T' such that s ≤ g (s') and define (9. that is. the limit mapping H' are in the category is also S-continuous. Take an s 0 ∈ T such that so it follows from the hypothesis that ƒs' is oneClearly. There exist j ∈ T' and Theorem 9. so that for any implies the element (9. to-one. That shows ƒ is an onto mapping. Suppose that {g.229) is well defined. . from the definition of S-continuity.231) Then Thus. It follows that s' ∈ T'. Take an object for any satisfying s' ≥ t'. which shows that Therefore. We now suppose that ƒ s' is a one-to-one mapping from S g (s ' ) onto for every from lim ← H'.7. and that ƒ(x) = y. that an element in lim ← H.

Proof: Applying Theorem 9.5 imply that f is also S-continuous. the limit mapping f is a similarity mapping from lim←C H o n t o lim← C H ' . the limit mapping f : lim ← h H → lim ← h H' is also a homomorphism. let where R * is defined by Eq. T } and H' = { S'. If all mappings ƒs ' are similarity mappings.234) Hence. Theorem 9.232) Then Therefore. in order to show that f –1 is S-continuous from lim ← C H ' onto lim ← C H.2. R'). Suppose that H = {S. the limit mapping f : lim is also a similarity mapping. T' } are two linked hierarchies of systems in the category and that {g. l t's' . it suffices to show that f is a homomorphism from lim ← C H into lim ← C H'. R ). we know that the limit mapping f is a bijection from lim ←C H onto lim ← C H'.6. f (r ) ∈ R'. . (9.7.7. T } and H' = { S'. such that (9. l t s .4. Let r ∈ R be a relation such that there exist j ∈ T and rj ∈ Rj satisfying (9.4 and 9.ƒs '} is a mapping from H into H'. T'} are two linked hierarchies of systems in the category and { g. Proof: It suffices to show that for any relation r in the system lim ← h H = (M. ƒ(r ) is a relation in the system lim ← h H' = ( M'.223). l t s . To this end. lt's' .7.7. Suppose that H = { S.233) Then (9. Theorems 9.7.ƒ s '} is a mapping from H into H'. If all mappings ƒs' are homomorphisms.General Systems: A Multirelation Approach 251 Theorem 9. Now.

(9.8. respectively. Let {S. it suffices to show that for any relation r' in the system lim ← h H’ = ( M'.7. T } is similar to lim← {S. and l t's' . . Pick an element where R' * is defined by Eq. Then lim ← {S. Corollary 9. an element The relation r ∈ R.7.236) Now for any t ∈ T choose a t' ∈ T' such that g( t' ) ≥ t. The result follows from Theorems 9. l t s .7. it follows that there exists exactly one relation rg (t ') ∈ R g ( t ') such that (9.223).R ) such that f (r ) = r'.7. The result follows from Corollary 9. T} be a linked α-type hierarchy of systems in category or If the partially ordered set T has a maximum element t0 ∈ T.2. (9.7.R' ). if all mappings ƒs' .238) satisfies f (r ) = r * . Define a relation r t ∈ Rt by (9. Then is well defined.4 and 9. Corollary 9. are similarity mappings. Under the same hypotheses as in Theorem 9.2. T' }.235) From the hypothesis that each ft ' is a similarity mapping from Sg ( t' ) onto S't' . T }. such that (9.3. where R * is defined by Eq. lt s .252 Chapter 9 Theorem 9.7. and l t s replaced by R't '.223) with Rt . T}. there exists a relation r in the system lim ← h H = ( M. the limit mapping f : is also a similarity mapping. l t s. T. T'. then the systems lim ← h {S.T } be a linked α -type hierarchy of systems in category or T' a cofinal subset of T. l t s.8.7.6 and 9. Proof: By Theorems 9. lt s .7.7. where lim ← indicates either lim ← h or lim ← C.7. defined by (9.7.237) It can be shown that the definition of r t does not depend on the choice of t'. Let {S. and S t0 are similar. lim← C {S. in addition. l t s .

where or and the object set M and the relation set R are lim← indicates defined as follows: There exists a bijection h :Σ → M such that (9.4. Let {S. l t s.4.7. for each t ∈ T.10. where each m s is again a system. T } be a linked α -type hierarchy of centralizable systems in the category If the 1-level inverse limit system is centralizable. Theorem 9.239) and R = { h ( r ) : r is a relation in lim S t} ← where lim← indicates either lim←c or lim ←h .9.1 and 9. or or when all relevant linkage mappings are homomorphisms.3.General Systems: A Multirelation Approach 253 Corollary 9. then the system Proof: The proof follows from the fact that if C t is a center of the system St . R).7. Let {S. Let {S. Theorem 9. lt s . T} is connected. T } is a linked α -type hierarchy of systems in the category Then the limit system lim ← h St is a partial system of ∏{ St : t ∈ T }. Then the inverse limit system lim ← C {S.7.7. The result follows from the proofs of Theorems 9. The 1-level inverse limit.9. then the inverse limit lim ←c Ct is a center of the 1-level inverse limit Theorem 9. exists.3. for each t ∈ T. Proof: This theorem follows from the fact that if Ct is a center of the system S t.7. l t s .7. is defined by lim ← St = (M . l t s . Let { S . T } be a linked α-type hierarchy of centralizable systems in the category If the 1-level inverse limit system exists. s ≥ t } of linkage mappings. l t s. lt s. there exists a set T. Suppose that {S.11. then the inverse limit lim ←h Ct is a center of the 1-level inverse limit . The result follows from Theorems 9. then the limit system is also centralizable.2 and 9. T } be a linked α -type hierarchy of systems such that for every element in the thread set Σ. denoted by or when all relevant linkage mappings are S -continuous. T } be a linked α -type hierarchy of systems in the category such that one of the states St is connected.

Rugh (1981 ) gave a systematic theory on nonlinear continuous and discrete time systems.254 Chapter 9 Theorem 9. T } is a linked α -type hierarchy of systems in the category such that for some index t 0 ∈ T. 1990e). Theorem 9. Klir (1985) and Lin (1987). Cornacchio (1972) showed what an important role the concept of continuity played in research of general systems theory and its impacts on engineering. such as free sum. and a study of controllabilities of nonlinear systems can be found in (Casti.8. Gratzer (1978). social and behavioral sciences. products and limit systems. 1989a). a more general setting of the concept of time systems can be found in Mesarovic and Takahara (1975). 1988b. Not all of the concepts of controllabilities are equivalent.6.1 implies this. The study of time systems can be found in many publications. Suppose that {S. Schetzen (1980) studied continuous time systems of multilinearity. l t s . Wonham (1979) presented a concept of controllability of linear systems of differential equations. References for Further Study The search for an ideal definition of general systems has been going on for decades: see von Bertalanffy (1972). Deng (1985) studied the concept of gray control. and natural sciences. 1990c. T} is also generally controllable (but relative to a diffrent mapping G * ). Ma and Lin. 1975. 1988d). 9. and their definition was generalized by Ma and Lin (1987). Periodic and order structures of families of systems were studied intensively in (Lin. Lin. computer science. Mesarovic and Tahakara (1975). 1985). The concepts of controllabilities presented here are based on research contained in (Mesarovic and Takahara.12. the state S t 0 = ( M t0 . .7. l t s. Studies on various controllabilities of systems can be found in almost all books on systems.7. 1987. were first given by Ma and Lin (1987. For example. Connectedness and S-continuity were first studied in (Lin.5 it follows that we need only show that the mapping is S-continuous. The constructions of new systems. Proof: From Theorem 9. 1987). Ma and Lin. For example.R t0 ) is generally controllable on V' relative to a mapping Then the inverse limit lim ← C {S.

a (general) system consists of a set of objects and a set of relations between the objects. Section 10.1. Section 10.CHAPTER 10 Systems of Single Relations As studied in Chapter 9.1 is devoted to the study of two very new concepts: chaos and attractor. we let X = M n and Y = M m . {r }). and the following section introduces important concepts. Chaos and Attractors A system S = ( M.2) is a binary relation then D is called a chaos of on Z. which can be described by one relation. the input–output system S in Eq.2) where the sets X and Y are called the input space and the output space of S. have many complicated structures and properties that we still do not completely understand. Let D ⊂ Z be an arbitrary subset. linear time systems. For example. In practice. and time invariably realizable systems. (10. If D 2 ∩ S = 255 . respectively. each human being and each factory are input–output systems. we will think of S as the binary relation r such that (10. {r }) is called an input–output system if there are nonzero ordinal numbers n and m such that (10. 10. In this chapter. Section 10. including MT-time systems. If we let Z = X ∪ Y. most systems we see are input–output systems. and study some feedback-invariant properties. stationary systems. instead of using the ordered pair ( M. and an example is given. many systems. we concentrate on systems of one relation and related concepts. In the real world.1) Without loss of generality.4 shows how “whole” systems can be decomposed into factor systems by feedback. The second section contains the general concept of linear systems and characterizations of feedback systems.5 studies the conditions under which “whole” systems can be decomposed.

Let D ⊂ Z. Therefore.4) where D ⊂ Z. D ⊂ S * D. D is Let COS(S) denote the set of all chaotic subsets on Z of the input–output system S. Then for each object d ∈ D and any y ∈ D . Proof: Necessity. Suppose D is a chaos of S. Then D is a chaos of the system S iff D ⊂ S * D. is S is not a subset of I where I is the diagonal of the set Z defined by I = {(x. Therefore.x) : x ∈ Z }. Sufficiency. for each d ∈ D.5) (10. Intuitively. d) ∉ S. Then the nonempty subset D is a chaos of S.1 implies I that each chaos of S is a certain subset of S– .1. (d. Sufficiency.3) holds. (d. . d ∈ S * D. Then D² ∩ S = Ø. I (10. Theorem 10. Suppose that (10. Then there exists d ∈ Z such that (d. the reason why D is known as a chaos is because S has no control over the elements in D. That is. Suppose D satisfies D ⊂ S * D.1. This implies that S ⊄ I. Theorem 10. Proof: Necessity.6) We denote the complement of SI by S – = Z – S I . Suppose that the input–output system S over the set Z has a nonempty chaos D.7) The following theorem is concerned with determining how many chaotic subsets an input–output system has. Let D = {d}. (10.1. The necessary and sufficient condition under which an input– output system S over a set Z has a chaotic subset D ≠ Ø.y) ∉ S Therefore. Therefore.2. Then D² a chaos of S. Let S ⊂ Z ² and define (10.1. If S ⊄ I.256 Chapter 10 S.3) ∩ S = Ø. Then Theorem 10. then (10. d) ∉ S.

it is known that every – chaotic subset of Z is a subset of S I . i. Let {x. and every two pairs do not have common elements. containing two pairs of elements in SI I with S-relations.3.e.x) ∈ S. y) ∈ S implies (y. Continuing this process. There are n/2 pairs of m –2 subsets from the elements in S – with S-relations.Systems of Single Relations 257 Theorem 10. We add the number of subsets which have been subtracted twice.8) where k = n/2. The total number of all subsets of S – is 2 m . some of the (n/2) × 2 subsets. – – then every subset of S I – {x. and there exists at most one such that either (x. have been added once more than they should have. Naturally. Among the elements – of S I . there are n/2 pairs of elements with S-relations.. when we add subsets. Let Then (10.y } forms a subset of S I not belonging to COS(S).y} ∪ {x. this number should be subtracted. There are a total of 2 m – 2 of this kind of subset. and obtain (10. every subset of S – . them are in fact the 2 I – same ones. has been subtracted twice.1. (ii) S is not a subset of I. (iii) (iv) (y. a subset of S I is a chaos of S if the subset – does not contain two elements of S I which have S-relations. Therefore. But. we subtract these (n/2) × 2 I m–2 – m subsets of S .9) However. (x. I – Under the assumption of this theorem. Suppose that an input–output system S ⊂ Z² satisfies the following conditions: (i) S is symmetric.x) ∈ S. we .y) ∈ S or Proof: According to the discussion prior to the theorem. the subsets of S – containing three pairs I – of elements of S I with S-relations. y} be a pair of elements in S–I with an S-relation. There are such sets.

S * D ⊂ D. Theorem 10. Suppose that D is an attractor of S.5.1 and 10. Suppose that S ⊂ Z² is an input–output system over the set Z and D ⊂ Z. That is. Let n = and k = n/2.1. Then D is an attractor of S. Then (10. 10.3. d ∈ D . Therefore.258 Chapter 10 have By using the same method. Figure 10. (d.1.1. Suppose that an input–output system S ⊂ Z² satisfies conditions (ii)–(iv) of Theorem 10.4.x) ∉ S. Proof: Necessity. .1.10) Let S be an input–output system on a set Z. Fig. Let d ∈ S * D be an arbitrary element. for every x ∈ D. By Eq. When S is a function. the constraint in Theorem 10.4). (10. Theorem 10. When S is not a function. iff D ⊃ S * D.3 that the system S is symmetric can be relaxed.2 shows that the graph of S outside the vertical bar D × Z must be contained in the horizontal region Z × D.1 shows that the graph of S outside the vertical bar D × Z overlaps the horizontal bar Z × D.2 show the geometric meaning of the concept of attractors. A subset D ⊂ Z is called an attractor where Figures of S if for each 10.

and S is a function from Z to Z.1. so there exists at least one object d ∈ D such that (d. Z – D ⊂ S * D.1. x ∉ S * D.. and S(x) contains at least one element for each x in Z – D.1. That is.e. It follows from Eq.2. when Z – D ≠ Ø .4) that for each object x ∈ Z – D. i. D is an attractor of S.Systems of Single Relations 259 Figure 10. S. (10. Sufficiency. ≠ Proof: Necessity. The necessary and sufficient condition under which an input– output system S ⊂ Z² has an attractor D ⊂ Z is that S is not a subset of I. D ⊄ S * D. does not have any attractor not equal to Z. This tells that the hypothesis that S has an attractor D such that D ⊂ Z implies that S ⊄ I. Then for each D ⊂ Z. D is an attractor of S. .5.x) ∈ S. according to Theorem 10. ≠ Figure 10.6. Suppose S has an attractor D such that D ⊂ Z. S (d) ∩ D ≠ Ø This implies that D is an attractor of S. By contra≠ diction. Suppose that D satisfies D ⊃ S * D. suppose S ⊂ I. Therefore. Theorem 10.

z) ∉ S. Theorem 10. Applying Theorem 10. Theorem 10. and define D = Z – {x}. Let (10.1. there exists an element y ∈ Z such that y ≠ d and ( d.5. according to Eq. ( y.3 shows the case when a subset D of Z is a strange attractor of an input–output system S. (10.11) and ATR(S) be the set of all attractors of the system S. Therefore. Choose distinct objects x. Let D ⊂ Z. for each object d ∈ S * DA .1. d ∉ A. In fact. it follows that the subset DA is an attractor of S.y) ∈ S. contradiction. where the square D × D is the only portion of the band Z × D over which the graph of S does not step.y ∈ Z such that ( x. for any z ∈ Z. ( y. This says that DA ⊃ S * DA . implies that the element y ∈ Z – A. Then n ATR (S ) = 2 (10. it follows that S has attractors. Figure 10.1.1. so d ∈ Z – A = DA .y) ∈ S then for any z ∈ Z.260 Chapter 10 Figure 10.3. . T h e hypothesis stating that if (x. Then.8. Otherwise.7.5. (d . D ⊃ S * D. This completes the proof of the theorem.12) Proof: For each subset A ⊂ S I –. Applying Theorem 10.z) ∉ S.4). x) ∉ S for every x ∈ DA . Suppose that an input–output system S ⊂ Z² satisfies the condition that if (x. Then DA ⊃ S * DA . A subset D ⊂ Z is a strange attractor of an input–output system S ⊂ Z ² if D is both a chaos and an attractor of S. Suppose S ⊄ I. Then D is a strange attractor of the input–output system S iff D = S * D.y) ∈ S . N o w it can be seen that there are 2n many such subsets D A in Z. Sufficiency.y) ∈ S then. let D A = Z – A . D is a strange attractor of the system S.

1. Therefore. Suppose S has a strange attractor D. However.y ) ∈ S.5. x ∈ D. It is clear that for each linearly ordered subset of .1. . The rest of the proof of the theorem follows from Theorems 10. 10. This implies that x ∈ S D. is also contained in . Sufficiency. and 10. x ∈ Z – S D. Thus.1. it follows that D 2 ∩ S = ø . This completes the proof that S * D = Z – S D. y ) ∉ D. Let x ∈ S * D be an arbitrary element.14) is not empty.Systems of Single Relations 261 The proof follows from Theorems 10. Let us denote (10. This leads to a contradiction.10.1.5. there exists at least one element y ∈ D such that (x . Proof: Necessity. D ∩ S( x ) ≠ ø. Then a necessary and sufficient condition under which there exists a strange attractor of the system S is that there exists a subset D ⊂ Z such that D 2 ∩ S = ø and for each x ∈ S I .8. i. Theorem 10. x ∉ Z – S D. (x. (iii) D is a strange attractor of S iff D = Z – S D. Suppose that S ⊂ Z 2 is a symmetric input–output system.1. From the definition. x ∉ S D. therefore. D ∩ S( x ) ≠ ø.2. If an element x ∉ S * D .9. if there exists an element x ∈ S I satisfying D ∩ S (x) = ø then x ∈ S * D. By Theorem 10.13) Then the following statements hold: (i) D is a chaos of S iff D ⊂ Z – S D. Then for any element y ∈ D.1.e. We conclude this section with a discussion of the existence of strange attractors of input–output systems. Suppose that S ⊂ Z 2 is an input–output system and D ⊂ Z.8. x ∈ S I .1. (ii) D is an attractor of S iff D ⊃ Z – S D. Let be the collection of subsets of Z defined by and Since D ∈ (10. Theorem 10. ordered by inclusion. which implies that x ∉ S * D..2 and 10. contains all maximal elements under the order relation of inclusion. At the same time. Suppose there exists D ⊂ Z such that D 2 ∩ S = ø and for each x ∈ S I . we must have x ∉ S * { x }.1. Proof: It suffices to show that S * D = Z – S D.

11.y ) ∉ S.z ) ∈ g and (z . Proof of (b). By Theorem 10. .10. the composition system g f has a strange attractor if and only if there is a subset D ⊂ Z such that (i′) (ii ′) For each It now suffices to show that (a) conditions (i) and (i’) are equivalent and (b) conditions (ii) and (ii’) are equivalent.1. If B ≠ S * B. Condition (i’) is true iff for any x . iff for any x ∈ D and any . i. iff there does not exist a z ∈ Z such that (x . i..2. pick an element x ∈ S * B – B and define (10. Besides. i. condition (ii’) holds iff for each x ∈ (g f ) I there exists d ∈ D such that (x . . Theorem 10. iff for each x ∈ (g f ) I there exists z ∈ f D such that ( x. it follows that (y. z ) ∈ g and (z . . for each y ∈ B. x) ∉ S for every y ∈ B. y ) ∈ ƒ. since for each for each A ∈ . y) ∈ ƒ . Then a necessary and sufficient condition under which the composition system g f has a strange attractor is that there exists a subset D ⊂ Z such that the following conditions hold: (i) (ii) For every is not a subset of (Z 2 – g) (x). Proof of (a).1.z ) ∈ g .262 Chapter 10 Let B be a maximal element in .1. Suppose that ƒ and g are two input–output systems on a set Z such that . iff for each x ∈ ( g f ) I there exists z ∈ f D such that (x .e. Therefore.z ) ∉ Z – g. (x. y ∈ D . (x.e. From the hypothesis that S is symmetric. Proof: Applying Theorem 10. we have x ∉ S I .. (i) holds. iff for each ..15) According to the definition of the element x. (ii) is true. it follows that B ⊂ S * B. d ) ∈ g f. This contradicts the assumption that B is a maximal element in We next discuss the conditions under which a composition of input–output systems has strange attractors. iff for each x ∈ (g f ) I there exist d ∈ D and z ∈ Z such that (x. y) ∉ g f .e.

2. . (10. x ) ∈ C × D (S).18) and (10. (2) There exist a linear space C over and a linear mapping ρ : C × D (S ) → Y such that (x . X and Y linear spaces over such that (i) ø ≠ S ⊂ X × Y. (x 2 . (iii) s ∈ S and α ∈ implify α · s ∈ S. Proof: First.Systems of Single Relations 263 10.y 2 ) ∈ X × Y and any α ∈ . x) = R 1 (c ) + R 2 (x ) for every ( c.x 2 ∈ D (S ) and any α ∈ .2. From the hypothesis that S is linear. and S an input–output system Here + and · are addition and scalar multiplication in X × Y.y ) ∈ S iff there exists a c ∈ C such that ρ(c . Feedback Transformations Let be a field.19) and thus x1 + x 2 and α x1 ∈ D(S). the domain D (S) is a linear subspace of the space X.17) The input–output system S is then called a linear system. (ii) s ∈ S and s' ∈ S implify s + s' ∈ S. respectively. (x2 .x ) = y. and defined as follows: For any (x 1 . it follows that (10.y 1 ). Here D ( S) is the domain of the input–output system S.1. Suppose that X and Y are linear spaces over the same field Then the following statements are equivalent: (1) S ⊂ X × Y is a linear system.y 1 ) . there exist y1 . (3) There exist a linear space C over and linear mappings R1 : C → Y and R 2 : D(S) → Y such that the mapping ρ in (2) is such that ρ (c .y 2) ∈ S.16) (10.y 2 ∈ Y such that (x 1 . Theorem 10. . For any x1 .

Define (10. y ) ∈ S for every x ∈ D (S). and C is a linear subspace Then C ≠ ø.22) We now choose an isomorphism I : Y/C ⊕ C → Y satisfying I (y.2. Three examples are given to show that the concept of linear systems is an abstraction of different mathematical structures. y ) ∈ S . (a) Suppose S is a system described by the following linear equations: . for each x ∈ D(S) and any y 1 .y ) ∈ S if and only if there exists c ∈ C such that ρ (c.y) ∈ X × Y is such that there exists c ∈ C with ρ ( c.x ) =y . so (x. This means that Define a linear mapping as follows: (10. (10. if .21) if (x. so . respectively. In fact. .264 Chapter 10 The proofs that (3) implies (1) and (2) implies (3) are clear because we can and define the desired mappings R 1 and R 2 by where 0x and 0c are the additive identities in the linear spaces X and C. Example 10. 0 y ∈ C because over .20) . Proof that (1) implies (2). y2 ∈ Y satisfying and thus .1. In fact. in fact. hence. so then . Suppose Then y = y 1 + c 1 for some c1 ∈ C. x) = y. respectively. Then σ is well defined. if (x . there exists c ∈ C such that On the other hand. C) = y for every y ∈ Y /C.23) This completes the proof of the theorem. Now the desired mapping ρ : C × D (S) → Y can be defined by Then ( x. Suppose S is a linear system and 0x and 0 y are the additive identities in X and Y. Consider the quotient space with addition +* and scalar multiplication ·* defined by and for any α ∈ and any (b 1 + C ) and Define a mapping by letting (10.

2. where. and bj . Let S ⊂ X × Y and S ƒ : Y → X be a linear system and a linear functional system. . are n variables.1(b). and Y a set of some linear transformations on X. An input–output system S ⊂ X × Y is a functional system if S is a function from the input space X into the output space Y. In Example 10.2.R) a system such that the object set M is a linear space over . Then S is a linear system if for each relation r ∈ R there exists an ordinal number n = n ( r) such that r is a linear subspace of M n . 2. Then Y can be studied as a linear system (X .j = 1.2 . where the object set M is the set of all real numbers. respectively. . b ]. .R). (b) Suppose a given system S is described by the differential equations for .Systems of Single Relations 265 where xi . m.j = 1. . b]. and is fixed such that where each i = 1. the concept of general linear systems is the following: Let be a field and S = (M. . i = 1. . m. and f ( t ) are continuous functions defined on the where interval [a. . .24) . m is the number of equations. . for each relation r ∈ R . and the relation r is defined by where y ∈ M is a fixed function such that (c) Suppose X is a linear space over a field . Then the feedback system of S by S ƒ is defined as the input–output system S′ such that and (10. S is not an input–output linear system.2. {r}). there exists a transformation y ∈ Y such that for any if and only if y(x 1 ) = x 2 . … . . The system S can then be rewritten as a linear system (M. We will study only input–output linear systems defined in the beginning of this section. Generally. .2. The system S can be rewritten as a linear system (M. m . . are the constraints of the system. n. . . . n ) are the coefficients of the system. (i = 1. {r}) in the following way: The object set M is the set of all continuous functions defined on [a. a i j ..

for each (x. Let = { S ⊂ X × Y : S is a linear system} ƒ = {Sƒ : Y → X : Sƒ is a linear functional system} and = { S' ⊂ X × Y : S' is a subset} (10. y ) ∈ S . This implies that (x . (x + S ƒ (y).24). Then. y) ∈ S '.2. y ) ∈ { (z – S ƒ (y). by the definition of feedback systems. (10. (10. The following theorem gives a concrete structure of each feedback system. Figure 10. iff (10. the relation in Eq. y) ∈ S '.28) Proof: Necessity. Theorem 10. Suppose that S ⊂ X × Y is a linear system and Sƒ : Y → X a linear functional system. Suppose S' is the feedback system of S by Sƒ .266 Chapter 10 Figure 10.24).4. Clear because Eq.2. which is called the feedback transformation over the linear spaces X and Y.27) (10. y) ∈ S. (10. (x – S ƒ (y).25) (10. The systems S and S ƒ are called an original system and a feedback component system.24) implies that for each element (x. Structure of feedback systems. Again. y) ∈ S}. Then a subset S' ⊂ X × Y is the feedback system of S by Sƒ . Sufficiency. y) : (z . (10. respectively.26) Then a feedback transformation is defined by Eq.4 shows the geometric meaning of the concept of feedback systems. . That is.28) implies the relation in Eq.

(10. a linear subspace) S' ⊂ X × Y. there exists a linear functional system S f : Y → X such that F(S. S f ) = S' ? Theorem 10. For each linear system S ⊂ X × Y and arbitrary linear subspace S' ⊂ X × Y. that is.2 implies that R(S') = R (S) and that . we need to produce a new linear structure (say.Sƒ ) = S '.2. For each linear system S ⊂ X × Y.2. Can we design a feedback process Sf : Y → X such that F ( S. Suppose there exists a linear functional system S ƒ : Y → X such that F(S.5 discusses the control problem: A linear structure (say.2. Theorem 10. Eq. where –Sf : Y → X is the linear functional system defined by (10. Then Theorem 10.Systems of Single Relations 267 Theorem 10. S is a feedback system of itself.32) Proof: Necessity.2. there exists a linear functional system Sf : Y → X such that F(S. Suppose that S.Sf).3. Proof: The desired linear functional system Sf : Y → X can be defined by (10. The proof is straightforward and omitted. system) S ⊂ X × Y is fixed. S' ⊂ X × Y are linear systems and Sf : Y → X is a linear functional system such that (10. defined by (10.2.28) says that S = F(S.31) for every y ∈ Y .29) Then F(S'. where N(W) indicates the null space of the input–output system W ⊂ X × Y. According to some predefined targets.5. –S f ) = S. Assume the axiom of choice.Sf ) = S.30) for every y ∈ Y. Theorem 10. Then.4.Sf ) = S' iff R(S') = R(S) and N(S') = N(S).

[i. The systems S.e. (10.37) Thus. .e. the mappings C and C' are well defined.yi ) ∈ S'. let (10. for each i ∈ I . for each element (x.39). We are now ready to check that for each y ∈ R (S') = R(S). For each element y ∈ R (S').38) satisfying claim. We now define a linear functional system (10.. (10.y) ∈ S ]. Let {yi : i ∈ I} be a basis in the linear space R(S'). where I is a finite or infinite index set. it follows that there exist mappings (10.. Suppose that S and S' have the same output space and the null space.34) From the Axiom of Choice.39) Therefore.40) .e. and (C(y).35) and (10. and Sf have the property that F(S. In fact. We define a linear functional system S f : Y → X such that F (S.33) and (10. (10. Then we make the following Claim. i.S f) = S'. Sf )= S'. R(S') = R (S) and N(S') = N (S). we have since (10. from Eq. S'. for an i ∈ I .268 Chapter 10 Sufficiency.36) such that [ i..

42) It now remains to show that for each pair (x. we have from Eq. (10. (10.24) holds. where In fact.yi ) ∈ S we then have . Then In fact. Then for a finite number of y i ’s from the basis {y i : i ∈ I}. Let (x. because (10. for every i ∈ I.43) This gives us the implication from the right to left in Eq.40). (10. Then (10.Systems of Single Relations 269 Now let (x. For each pair for some finite number of nonzero coefficients ci .44) . let each y i is from the basis {y i : i ∈ I}.24). where each ci is a nonzero number.y) ∈ X × Y.y) be an arbitrary element in S'. That is.41) the opposite implication of Eq. the relation in Eq. The opposite implication can be shown as follows.41) Hence. (10. (10. (10.

The feedback system F(S.6.270 Chapter 10 Figure 10. Let * be a system connecting operator. be input–output systems such that (10.49) and (10.48) Figure 10. Structure of cascade connections of systems. termed the cascade (connecting) operation. i = 1.51) . The operation * is defined as follows: Let Si .5 shows the geometric meaning of the cascade connection of systems.45) (10.47) where it is assumed that no algebraic properties are given on each set involved. This completes the proof of Theorem 10.2.5. 10.46) (10. The relationship between the two systems is displayed in Fig.5. We conclude this section with a discussion of a more general definition of the concept of feedback systems.3.50) where S ƒ is called a feedback component of S. The three systems satisfy S 1 * S 2 = S3 if and only if (10. We now let S and S ƒ be two input–output systems such that (10.2.Sƒ ) of S by S ƒ is defined by (10.

we first cascade S and Sƒ (i. A property of a linear system S ⊂ X × Y is called feedback invariant with respect to ƒ if the property also holds for the feedback system F(S.3.5 here the Axiom of Choice is not used. Let S ⊂ X × Y and S ƒ : Y → X be a linear system and a linear functional system. engineering. social science.2.2.Systems of Single Relations 271 Figure 10. 10. by feedback system we always mean the definition given by Eq. Without special explanation in this book. Sƒ )) = N (S) come from Theorem 10.1. The equations R(F(S. Feedback invariance has played a crucial role in characterizing homeostatic or explosive system behaviors in various areas such as biology. Proof: The linearity of the feedback system F(S. (10.6.3.24). S ƒ )) = R (S) and N(F(S. . Construction of general feedback systems. S ƒ )) = R (S) and N(F(S. Then the feedback system F(S.2. Those properties of original systems invariant under the feedback transformation F are of special interest in research on feedback systems. In constructing F(S.. S ƒ ) for each linear functional system S ƒ ∈ ƒ . S ƒ ) is a linear system over X × Y and R(F(S. Corollary 10. S ƒ ) follows from Theorem 10. Feedback-Invariant Properties Some properties of original systems may be kept by their feedback systems and some may not. respectively. and so on. Sƒ )) = N (S).e. find S * Sƒ) and then input one of the two outputs of the system S * Sƒ . Sƒ ) step by step.

Let S : X → Y and S ƒ : Y → X be linear functional systems such that S is surjective. Then D(F(S. where I : Y → Y indicates the identity mapping. x 2 ∈ D(F(S.55) . Let S ⊂ X × Y be a linear system and S ƒ : Y → X a linear functional system. Thus. i.. From the hypothesis that S is a surjective system. There then exists some y ∈ Y such that (x. it follows that there exists some x ∈ X such that S (x) = y. y) ∈ S. This proves that (I – S Sƒ ) : Y → Y ° Theorem 10. i. Sƒ )) such that (10.e. y) ∈ S.1. Then D(F(S. ) ∈ F(S.. Theorem 10.24).2. ∈ S. Hence. Necessity.3. Suppose the feedback system F(S. S i n c e D(F(S. S ƒ )) = D(S). (10. Then for each arbitrarily fixed Sƒ ∈ ƒ . Since (I – S °Sƒ ) : Y → Y is surjective. Proof: Sufficiency. The proof is clear from Eq.54) Thus is surjective. and let Sƒ ∈ ƒ be arbitrary.272 Chapter 10 Theorem 10. Pick an arbitrary y ∈ Y. S ƒ ) . Hence.e. Let S : X → Y be a linear functional system.3.e. for this x there exists some ∈ Y such that (x. Suppose the original system S is injective.. Sƒ ) is not injective. Sƒ )). there exist distinct x 1 . the feedback system F(S. It is clear that D (F(S. S ƒ )) = D(S) iff ( I – S ° Sƒ ) : Y → Y is surjective. . x ∈ D (F(S. Let x ∈ D (S) be arbitrary. there exists exactly one ∈ Y such that (10. ∈ S. Sƒ )) = X iff for each x ∈ X there exists a y ∈ R(S ) such that (x + S ƒ (y). i.53) That is. Sƒ )) ⊂ D (S).3. (10.3.52) This implies that (10. Proof: Sufficiency. I(y) = y for each y ∈ Y. Sƒ ) is injective iff the original system S is injective.

the feedback system F(S. Sƒ ) is injective.1. Therefore.e. (Mesarovic–Takahara time system). From Theorem 10. from the hypothesis that the system S : X → Y is injective. and α ∈ . In the rest of this section. linearity. Sƒ ). Range space. and bijectivity of original systems are feedback invariant with respect to ƒ .4. Proposition 10. Combining what have been obtained in this section. . we study feedback-invariant properties of MT-time systems.59) (10. it follows that x 1 + Sƒ (y) = x 2 + Sƒ (y). Proposition 10. null space. contradiction. For any linear system S ⊂ X × Y and any linear functional system Sƒ : Y → X.y) ∈ S.e. F(F(S.3. X ⊂ A T and Y ⊂ B T . surjectivity.3. Assume the input space X and the output space Y in the preceding discussion are linear subspaces of A T and B T . i.. (10.y). Therefore. Let A and B be two linear spaces over the same field . g ∈ A T (respectively. By Eq. Theorem 10.3. ∈ B T ). –S ƒ ) = S Applying the proof for sufficiency. i. We define A T = {x : x is a mapping T → A } and BT = {x : x is a mapping T → B } Then the sets A T and BT can be made linear spaces over elements ƒ. Each input–output system S ⊂ A T × B T is termed an MT-time system. x 1 = x 2 . Sƒ ) is a linear system over AT × B T.58) (10.57) as follows: For any (10. The proof is straightforward and is omitted.3. injectivity. Let T be the time axis defined as the positive half of the real number line. we have (x 1 + S ƒ (y). respectively. Suppose F(S. T = [0..Systems of Single Relations 273 for some y ∈ R (S).56) for each t ∈ T. Necessity.24). (ƒ + g)(t) = ƒ (t) + g (t) and ( αƒ )( t) = α · ƒ (t ) (10. i.2.e. (x 2 + Sƒ (y).1 implies that the feedback transformation defined in Section 10.2 is well defined on the class of all MT-time systems. we know that S is injective.. + ∞ ). the following theorem is evident.

The graph of x in the plane T × A.62) and (10. called the concatenation of x t and x't .63) For each y ∈ Y.66) for each ξ ∈ T τ . let (10.60) (10. or in general. Figures 10.64) where x t x' is a time function ∈ AT . 10. for example. In Figs. (10.65) Without loss of generality. we will always assume that D(S) = X and R (S) = Y . A subset S ⊂ X × Y ⊂ A T × B T is called a linear time system if S is a linear subspace of X × Y and the input space D (S) satisfies the condition: (10. and is °t defined by the following: For any s ∈ T. Let σ τ be the shift operator defined as follows: For each x ∈ X. For any t. for sets and vectors similar notation will be used. t' ∈ T with t < t'.61) Then the restrictions of a time function x ∈ X with respect to these time intervals are denoted by (10.7– 10.274 Chapter 10 Figure 10.7. for each etc. (10.8 .9 show the geometric meaning of the concept of shift operators. where τ ∈ can be any value if σ τ (x) is meaningful.

3. Let S ⊂ X × Y and Sƒ : Y → X be a linear time system and a strongly stationary linear functional time system. y) ∈ F(S.Systems of Single Relations 275 Figure 10. The graph of σ τ (x) with τ > 0. the application of σ τ implies that we move the graph of x τ units to the right.67) Theorem 10. Sƒ )} (10. Sƒ ) is strongly stationary.9 the dotted A × T planes indicate the locations of the A × T plane before the shift operation σ τ functions. when τ > 0. .5. y) : (x. Therefore. Then S = {(x + Sƒ (y). when τ < 0. Suppose that the feedback system F(S. respectively. Sƒ ) is strongly stationary iff the original system S is strongly stationary. Proof: Necessity.9.68) Figure 10. the application of σ τ tells us to consider the portion of the graph of x on the right of the vertical line t = τ .8. A linear time system S ⊂ X × Y is called strongly stationary if S satisfies the condition (10. The graph of σ τ (x) with τ < 0. Then the feedback system F(S. and 10.

we (10. implies that so Sufficiency.6. S f ) and S f are strongly stationary. A linear functional time system S : X → Y is Theorem 10.70) and (10.72) where so Eq. (10. This follows from Theorem 10. (10. Suppose S is precausal.71) shows that (10.276 Chapter 10 It suffices to show that for any now pick an arbitrary (x. it follows that (10.3 and the necessity part of this proof.73) where called causal if it is precausal. Then. Then for every t ∈ T. Then S is precausal iff Proof: Necessity. Let S ⊂ X × Y be a linear time system.3.2. A linear time system S ⊂ X × Y is called precausal if it satisfies the condition: (10.71) Let and Therefore.69) From the hypothesis that the systems F( S . Then for any This shows that when implies that because is the same as Hence.72) implies that Sufficiency. Eq.y) ∈ S and t ∈ T. Suppose implies that and This .

S ƒ ) is causal iff the composition time system is causal. (10.Systems of Single Relations 277 Theorem 10.76). (10. For any t ∈ T and any (10. (10. A linear functional time system S : X → Y is called time invariably realizable if it satisfies the condition (10. Suppose that the composition system is causal. Let S : X → Y and S ƒ : Y → X be linear functional causal time systems. Thus.8. It suffices to show that Eq. Proof: The argument of the necessity part follows from the fact that each composition of causal functional time systems is causal. for any t ∈ T and any (10.Sƒ ) is causal. Then S satisfies the relation in Eq.S ƒ ) ( x ). A linear functional time system S : X → Y is time invariably realizable iff for any t ∈ T and any x ∈ X (10.75) Therefore. where and This implies that the feedback system F ( S.3.78) .76).74) Let y = F (S.3. (10. We need to show that By Eq. Sufficiency.77) Proof: Necessity. implies that That is.24).76) where Theorem 10. (10.7. Then the feedback system F ( S. Suppose S is time invariably realizable.77) is derivable from the Then relation in Eq.

81) holds. the system is time invariably realizable.78) shows that (10. From realizable.85) .77) holds. Then Theorem 10. (10.3. (10.3. we have showed that the feedback system F (S. Suppose S satisfies Eq. Theorem 10. (10. Then (10.. Suppose F (S.278 Chapter 10 The last equality of Eq.8. Then the feedback system F (S. Then for any t ∈ T and any the hypothesis that S is time invariably realizable. (10.3.S ƒ ) is time invariably realizable.82) This implies that (10.76). Suppose that S : X → Y and S ƒ : Y → X are two linear functional time systems such that S and Sƒ are time invariably realizable. applying Theorem 10. Necessity. (10. (10.e. Therefore.84) Therefore. Suppose the system is time invariably Eq. (10. it follows from Theorem 10.8 that Eq.77) for any t ∈ T and any x ∈ X.9.79) Sufficiency.83) This equation is the same as (10.Sƒ ) is time invariably realizable iff for any t ∈ T and any (10. we see that S is time invariably realizable.S ƒ ) is time invariably realizable.80) and the relation in Eq. Proof: Sufficiency.81) i.3.80) Comparing Eq.8 says that (10.

Then a subset D ⊂ X is a chaos of S [iff D is a chaos of the feedback system F(S. such that D ⊂ X is a chaos. it follows that Eq. (10. iff D is a chaos. attractor.S ƒ ). respectively.12.89). This ends the proof. respectively.90) and (10. (10.3. it must be that Now which contradicts relation (10.87) can be written as (10.2. Pick an element Then Therefore. Theorem 10. We conclude this section with some discussion of chaos.Systems of Single Relations 279 Let Then from Eq. if (10.88) is the same as Eq. attractor.86) (10. So D must be a chaos of S.3. respectively. Sufficiency. Suppose D ⊂ X is a chaos of S but not a chaos of F (S. For each linear system S ⊂ X × X there exists a feedback component system Sƒ : X → X. attractor. of S. Then D ⊂ X is an attractor of S (iff D is an attractor of the feedback system F ( S. and feedback transformation. Pick an element Then This contradicts relation (10.89).S ƒ )).4.S ƒ ).88) Equation (10. Suppose that S ⊂ X × X and S ƒ : X → X are a linear system and a functional linear system.S ƒ )].S ƒ ) . The proof follows from Theorem 10. Suppose that S ⊂ X × X and Sƒ : X → X are a linear system and a functional linear system.85) we have (10. Theorem 10.11. (10. or strange attractor.3.3. respectively.81). Therefore. Suppose D ⊂ X is a chaos of F(S.8. if (10. or strange attractor of F ( S.89) Proof: Necessity.10.87) From the hypothesis that S is time invariably realizable and Theorem 10.91) . D must be a chaos of F (S. Theorem 10.S ƒ ) but not a chaos of S.

91) it follows that there exists such that Therefore.e. such that (10. We assume that the input space X and the output space Y can be represented in terms of the factor sets.94) and (10. there are sets . Suppose D ⊂ X . There then exists an element x ∈ X – D F such that (10. but not of the original system S.95) For each index i ∈ I. i.Sƒ ). Suppose D ⊂ X.e.96) be the projection defined by letting such that and We now can represent factor systems of S on the factor sets Xi and Yi .97) The system S can now be viewed as decomposed into a family of factor systems and S can be represented as a relation among its factor .4.. let (10. Decoupling of Single-Relation Systems Let S ⊂ X × Y be an input–output system. Sufficiency. contradiction.91). 10.92) From condition (10. satisfying condition (10.. satisfying condition (10.280 Chapter 10 Proof: Necessity.90). for each i ∈ I. where I is a fixed index set.S ƒ ). is an attractor of (S.93) From condition (10. i. Then there exists an element x ∈ X – D such that (10.90) it follows that there exists such that Thus. the factor system S i of S can be defined by (10. is an attractor of S but not of F(S. contradiction.

α (x. Proposition 10. If S = ∏ {Si : i ∈ I }. there are many interactions among subsystems. di ) ∈ Si . it follows that. respectively.1. i. For any (a i . Then each component system S i is also a linear time system. Proof: From the hypothesis that S is a linear system over the field A. (10. (10. S i is a nonempty system on Xi × Y i . and a complex system representation over {S i : i ∈ I }. for each i ∈ I.y) ∈ S .e. S ⊂ ∏{Si : i ∈ I }.98) because The argument of (2).2. Generally. If S is a linear system over a field A.Systems of Single Relations 281 by identifying with The systems S and S i are called the overall system is called and the factor (or component) system.e. We need to show two things: (1) For any for any (x .4. For any (ai . so S is a proper subset of the product of its factor systems ∏ {S i : i ∈ I }. ci ) and (b i . there exists a pair (x. S is called noninteracted We now consider what properties are transferable from the overall system S to its component systems Si or from the component systems Si to an overall system S.y 2 ) ∈ S such that systems (i.4. and (x 2 .y ) ∈ S i . ..y) ∈ S such that Therefore. c i ) ∈ S i and any α ∈ A. there are (x 1 . The argument of (1). y) ∈ S i and α ∈ A. each component system Si is then also a linear system over the field A. y1 ) and Therefore. Proposition 10.. Let S ⊂ ∏ {Si : i ∈ I } be an input–output system which has been decomposed into a family of factor systems = {Si : i ∈ I}. Suppose that S ⊂ X × Y is a linear time system that has been decomposed into a family of component systems {Si ⊂ Xi × Y i : i ∈ I }.99) because α(x.

y i ) ∈ S i Conversely. There then exists some (x i . we obtain the following: For each index i ∈ I and From Eq.1 to show that for any x.y ∈ D (S i ) for each i ∈ I. there exists some (x'. Proposition 10. because Proposition 10.2 and if the system S is precausal with D(S) = X. Consequently. In fact.y ) ∈ S such that such that we have That is.3. there exists some (x .4.2 and if the system S is strongly stationary. There then exists some element Furthermore. for and any t ∈ T. y) ∈ S Since which implies that Hence. v ∈ D (S) such that (10.y') ∈ S such that Hence. Under the same assumption as Proposition 10. Combining these arguments. there exist u. each component system S i is also precausal.4. . let Furthermore. Under the same assumption as in Proposition 10.y ∈ D(Si ) and any t ∈ T. we can show any that the component system Si is strongly stationary.100) Therefore. be arbitrary. there exists some (x. the concatenation of any x. then each component system Si is also strongly stationary. (10.282 Chapter 10 Proof: It suffices from Proposition 10.4. That is.4.66).4. Proof: Let such that such that where be arbitrary.4.

6 we have shown that each component system Si precausal.5. Therefore.6. y). It can then be seen that D ( Si ) = X i . In fact. let due to the assumption. Since it is assumed that S is Let arbitrary.y ∈ X satisfy and for every j ∈ I with j ≠ i. none of the component systems Si is functional. Let (x i .m) = (m. when the elements x. and Y 2 are all the same.y) and (w.yi ) and (wi . Then there exist (x. This implies that D(S) ≠ X. so. For any yi ∈ Y i there exists an element y ∈ Y such that From the hypothesis that S is bijective. Proposition 10. namely the set of all integers. z) ∈ S such and that Because S is functional. for any implies that from Theorem 10.Systems of Single Relations 283 Proof: Let i ∈ I be arbitrarily fixed. Let S be a functional linear system from X 1 × X2 into Y1 × Y 2 defined by S(n. Y 1 .3. y = z. it follows that there exists an element x ∈ X such that S ( x) = y.4.y) ∈ S such that . be arbitrary.n) for any (10. we have Thus (10.yi ) ∈ S i be and arbitrary.2} and the sets X1 . the overall system S is bijective. we can pick an element xi ∈ X i – D (Si ). and assume precausal. Since Proof: For any fixed i ∈ I.3. Therefore. Suppose that S ⊂ X × Y is a functional system that has been decomposed into a family of component systems with Then each component system S i is also functional.102) and Then it can be easily seen that neither S 1 nor S 2 is functional. We will construct an example to show that even though an overall system S is functional. Suppose the index set I = {1. because no x ∈ D (S) satisfies contradiction. If. there exist elements (x . Suppose S is a bijective function from X onto Y. moreover. S i is surjective. That is. is Example 10.101) Applying Theorem 10. each component system Si is also bijective. for That is. the component system Si is functional. X 2.4. (x . if D (Si ) ≠ Xi .1.

. We now let t ∈ T and x i ∈ X i be arbitrary.103) Let Eq. Hence. (10.104) Therefore.103).284 Chapter 10 that is. (10.6.1–10.4.. (10.4. There exists an element x0 ∈ X such that Since S is time invariably realizable. Suppose that S ⊂ X × Y is a functional linear time system that S has been decomposed into a family of component systems such that S is noninteracted.4. (10. (10. The following result combines Propositions 10.76).6. S i is bijective.e. we complete the proof that Si is time invariably realizab1e. from the proof of Proposition 10. If S is time invariably realizable. which implies that Then x 0 ∈ A ≠ Ø and each x ∈ A satisfies (10. for each fixed i ∈ I. it follows that D(S) ⊂ X. each component system Si is then also time invariably realizable.105) Applying Eq.4. from Eq. it follows that x = w.76) it follows that for t ∈ T. Proof: From the hypothesis that S is a functional system from X into Y. D( Si ) = X i .4. i. Thus. From the hypothesis that S is injective. Proposition 10.

Systems of Single Relations 285 Theorem 10. so there might be some properties which cannot be lifted up from component systems to the overall system S.4. such that S is decomposed into and S ƒ is decomposed into Then. (10. (4) S is causal if and only if each Si is causal. If S is decoupled by feedback. for each i ∈ I. we always hope to study S by studying each component system Si . Theorem 10. We say that S can be decoupled by feedback if there exists some functional system Sƒ ∈ ƒ such that (10. We now can use that representation to analyze and control S.4. Let S ⊂ X × Y be an input–output system which has been decomposed into a family of factor systems such that Then the following statements hold: (1) S is a linear system if and only if each Si is a linear system. respectively. (5) S is bijective if and only if each Si is bijective.1. (3) S is strongly stationary if and only if each Si is strongly stationary. Let S ⊂ X × Y be a linear system. In practice.4. The proof is not hard and is left to the reader. (6) The functional system S is time invariably realizable if and only if each S i is functional and time invariably realizable. Theorem 10.1 shows that the theoretical properties of many important systems are either hereditary from the overall system S to each component system S i or able to be lifted up from Si to S. Suppose that S ⊂ X × Y and S ƒ : Y → X are a linear system and a functional linear system. (2) S is a linear time system if and only if each Si is a linear time system. after applying appropriate feedback transformation.107) . The concept of decoupling by feedback shows the following actions: The original system S may not be noninteracted.106) where it is assumed that is the projection defined by relation (10.96).2. S is transformed to a system which has a noninteracted complex systems representation.

By the definition of decoupling by feedback. There then exists (x. we have (10.111) .3. be arbitrary. From this. there exists (x.108) implies that Let That is.108) Equation (10.110) gives Theorem 10. Suppose S can be decoupled by feedback.y) ∈ S such that (10.y) = (x i .110) Equation (10.109) Therefore. Suppose that S ⊂ X × Y is a linear system that can be decomposed into a family of component systems = {S i : i ∈ I}.4. Proof: Necessity. there exists a functional linear system Sƒ ∈ ƒ such that (10. Assume the Axiom of Choice.S ƒ) such that it follows that (10. Then S can be decoupled by feedback if and only if R(S) = Π {R(S i ) : i ∈ I } and N (S) = Π {N(S i ) : i ∈ I}. Since. y i ).286 Chapter 10 Proof: For each Π i (x .y) ∈ F(S.

Let S' ⊂ X × Y be the system defined by (10. by Theorem 10. A an m × m constant matrix.112) and (10. and : x is piecewise continuous} (10. Example 10.2. As an example of decoupling by feedback. we see that R(S) = R (S') and N (S) = N (S').5. C an n × m constant matrix. B an m × n constant matrix. we find there exists a functional linear system Sƒ ∈ ƒ such that (10. consider the multivariable system S defined by = Az + Bx y = Cz + D x z (0) = 0 (10.115) That is.2. (10.5.116) where z is an m × 1 variable vector.2.Systems of Single Relations 287 Therefore. S can be decoupled by feedback. By applying Theorem 10.113) Sufficiency. D an n × n nonsingular constant matrix.114) Since and due to our assumption.4.117) .

2. and : x i is piecewise continuous} is the input space and : y i is piecewise continuous} is the output space. n.116) in S is given by (10. satisfy and (10. define the system S i by = Az+Bi x i yi = C i z + D i x i z(0) = 0 (10.120) Di is a nonzero constant.119) (10. . . the inverse system S –1 of S is obtained as follows: (10. .122) (10.124) Thus.121) In fact. 2. i = 1.125) .288 Chapter 10 is the input space and : y is piecewise continuous} is the output space. Then the systems S and S i . since the matrix D is nonsingular. . B n ]. For each i = 1. the solution of the differential equation (10. . for any x ∈ X. .123) (10. D(S) = X. . n. .118) where B i is an m × 1 constant matrix such that B = [ B l B2 . C i is a 1 × m constant matrix such that (10. At the same time. . .

That is. a similar argument shows that R(S ) = D (S –1 ) = Y. . . (10.126) For the null spaces of the systems S and S i . . . . n. . . we can show that for each i = 1. . (10. . n. In fact. N(S ) = Π {N( S i ) : i = 1.127) and (10. R(S i ) = Y i . For the same reason.Systems of Single Relations 289 Therefore. 2.3. i = 1. 2. . S can be decoupled by feedback. n }.128) Therefore.4. . 2. . By Theorem 10. define .

(10.133) where the membership relation comes from the systems representation in Eq. Let z' be the solution of the differential equation in the representation in Eq.130) If we define a functional linear system Sƒ : Y → X as follows: (10. From the definition of Sƒ . (10. In fact. it follows that (10. ( x .290 Chapter 10 Then the system S d = S 1 × S 2 × . for any pair (x .130). y) ∈ X × Y.132) Therefore. × S n is represented by = α z + βx y = γ z + δx z(0)=0 –1 Since δ is nonsingular. y) ∈ S d iff the input x and the output y satisfy the systems representation in Eq. y ) ∈ S.129) is obtained as follows: (10.131). Then [ z z ' ]T is the solution of the differential equation in the systems representation in Eq. We show the last “if and only if” condition as follows. .130).130) iff (x + S ƒ (y ).125) of the inverse system S –1 . and let z be the solution of the differential equation in the systems representation in Eq. (10. First. suppose that the input x and the output y satisfy the systems representation in Eq. . (10.131) Then the original system S can be transferred into S d by this feedback component system S ƒ . . (10. (10.125). (10. S d (10.

135) Combining Eq.3 and 9.138) and the new system S ⊂ X × Y is defined by (10. Decomposability Conditions Intuitively speaking. From the systems representations in Eqs. 2.137).136) for i = 1. i = 1. (10. Sections 9. 3 if and only if and Suppose these three systems are connected in series.135) with the systems representation in Eq. x2 = y1 and x 3 = y 2 and are the input space and the output hold. defined as (10. However. the component systems S1. 2. 3. suppose space of a new system .131). and S3 completely disappeared. Let us now consider another construction of a new system S. S 2 . y) ∈ S. in Eq.Systems of Single Relations 291 Second. that is. we have (10.125) and (10. (10. y) satisfies the systems representation in Eq.137) Then is the composition of S 1 . which explicitly reflects the fact that . (10. Suppose that (10. (10. (10. The system S is a representation of the whole system.11 show the geometric meanings of and S.10 and 10. furthermore. S 2.5. for example. and S 3 . 10. are three input–output systems such that ∈ S i . We let (10.130). y) ∈ X × Y satisfies (x + Sƒ (y).139) Figures 10.6). Let us first consider the following example. suppose that a pair (x. (10.134) That is. a complex system consists of several subsystems (see.130) proves that the pair (x.

Structure of the system the whole system is composed of three component systems S i . 2. For the global system we emphasize the structure by writing The global system and the complex system S are intimately related to each other. Structure of the system S.11. i = 1. if we define two mappings and by h1 (x 1 . we identify with and denote the complex system S by S ⊂ S1 × S 2 × S 3 . y 2 . y 3 ) = y 3 . and S 3 . while S is a complex system representation of the whole system. 2. but to emphasize fact that S consists of the component systems S 1 .141) If we define two mappings and by then k 1 and k 2 satisfy the relations and (10.142) and (10. 3. i = 1.292 Chapter 10 Figure 10. The structure of is called a global system representation of the whole system. x 2 .10. S is the product system of Si . S 2.143) Figure 10. The complex system S is defined over X and Y.140) and (10. . x3 ) = x 1 and h 2 (y 1 . 3. then (10. In fact. In fact.

i = 1. Suppose that h = (h1 . Proposition 10.145) Therefore. Then the composition is a modeling relation from S1 to S3. (10. The proof is straightforward and is omitted. we can see that for any ( x.146) .1. Then the restriction of the product mappings h 1 × h2 and k 1 × k2 on S are mappings from S to S'. k 2 ) are modeling relations from system S1 to system S2 and from S 2 to system S3 . h 2 ) and k = (k1 .2.144) and for any (10.148) When I = {1.5. Generally. 2. 2. we denote the for any product mapping ƒ by ƒ 1 × ƒ 2 × … × ƒn .5. Suppose that h = ( h1 . . k 2 ) are two modeling relations from system S to system S'. The product mapping is defined by (10. Furthermore. and h i . be mappings such that h 1 : X → X' and h2 satisfying the condition (10. y) ∈ X × Y . . .Systems of Single Relations 293 Combining relations (10. i = 1. let S ⊂ X × Y and S' ⊂ X' × Y' be two input–output systems. S and are similar. respectively. h2 ) is called a modeling relation from S to S' and is denoted by Notice that h is actually a mapping from S into S'. h 2) and k = (k 1 . n} is finite. Suppose {X i : i ∈ I } and { Yi : i ∈ I } are families of sets and that {ƒi : X i → Y i : i ∈ I } is a family of mappings. if D ( S) = X and R(S) = Y. : Y → Y' (10. The following result shows the relation between the concept of modeling relations and that of product mappings.143). Proposition 10.147) Then h = (h1.140)–(10. then h 1 × h 2 S = k 1 × k 2 S if and only if h i = k i . . 2.

the mapping h 1 is not surjective. 2 ° ° but h' ≠ h". Due to Proposition 10.150) The last implication follows because h1 : X → X' is surjective and because D (S) = ' X. h" ) modeling relations from S' to 1 2 S" such that h' h = h" h. Then we h' h = h" h. Proof: Necessity. Pick mappings h'2 and h"2 : Y' → Y" such that h" = h"2 . then h is an epimorphism. That is.149) where. h 2 ) be a modeling relation from a system S to a system S' such that D(S) = X.1. Theorem 10. Then h = (h 1. and h 1 = h 2 = idX .2.5. Suppose h 1 and h 2 are surjective. h' 1 and h" are from X' to 1 X".1. S = { (x. if the modeling relation h is also surjective as a mapping (that is. Under the same assumption as in Theorem 10. Hence. We now see an example when the converse is not true. There exist mappings h' and h" such that 1 1 (10. h 1 (X) ≠ D (S') = X'. Let X = Y. we still use h for the mapping h 1 × h2 S and the modeling relation (h1 .5. h'2 ) and h" = (h" .2.1.5.2. Then ° ° (10.5. R(S) = Y. we 2 have h' = h". A modeling relation h = (h1 . Applying Theorem 10.5.5. A similar argument gives h 2 = h " . . The converse is not true. and R(S') = Y'. applying Proposition 10. h 2) : S → S' is called an epimorphism from the system S into the system S' if for any fixed system S" and modeling relations h' and h" from S' to S". Let S" be an arbitrary system over X" × Y" and h' = (h'1 .294 Chapter 10 The proof is straightforward and is omitted. Let h = (h 1. Sufficiency. Theorem 10. That is. h 2 ) : S → S'. S' = X × X. h is an epimorphism from S to S'. it follows that h is an epimorphism. h 1 × h2 S is surjective). whenever no confusion occurs. D(S') = X'. Then h is an epimorphism from S to S' if and only if h1 and h 2 are surjective. Proof: It can be clearly seen from the assumptions that the mappings h 1 and h 2 are surjective. Suppose h is an epimorphism from S to S' and. h 2 ) is an epimorphism but not surjective. for each arbitrarily fixed system S'' ⊂ X" × Y". h' ° h = h" ° h implies h' = h". without loss of generality. This contradicts the fact that h is an epimorphism from S to S'. x): x ∈ X }.

which contradicts the assumption that p ° h = (id X . k 2 ( y'1 ) = y1 . S is isomorphic to S'. then h is surjective as a mapping.5. y) and (u.. y) = h (u. i. Under the same assumption as in Theorem 10. A modeling relation h: S → S' is an isomorphism from S onto S' if and only if h is bijective from S onto S' as a mapping. y) = p ° h (u. and k 2 ( y'2 ) = y 2. It remains for us to show that h is injective. Case 1: Suppose that k1 (x'1 ) = x1 .. k2 ). That is. which contradicts the hypothesis that k is a modeling relation from S' to S. The contradictions imply that the modeling relation h does not have a right cancellation. h 2 ) has a right cancellation k = (k 1 . which contradicts the hypothesis that k is a modeling relation from S' to S. Let ° ( x'. Suppose h : S → S' is an isomorphism.e. Case 2: Suppose k 1 (x'1 ) = x 2 . v).Systems of Single Relations 295 Theorem 10. The converse is not true. Proof: Necessity. p h (x. It then follows that (k 1 .y2) ∉ S. . That is.5. suppose h : S → S' is not injective as a mapping. In this case. Suppose that S ⊂ X × Y and S' ⊂ X' × Y' are systems defined by (10.5.e. h has a right cancellation). y') = h(k(x'. h k = (idX . Then h ° k (x'. Proof: Let k : S' → S be a right cancellation of h . k2 ) (x'1 . id Y ). y')) = (x'. We then have that (k 1 .151) and (10. h is not a retraction. By contradiction.152) Let h 1 : D(S) → D(S') and h 2 : R(S) → R(S') be two mappings defined by If h = ( h1. V ) ∈ S such that h (x. Let p : S' → S be a left cancellation of h. idY ).y'2 ) = (x1 . We now construct an example to show the converse may not be true. A modeling relation h from a system S to a system S' is an isomorphism from S onto S' if h has both right and left cancellation. k 2 ) (x'1 . h is surjective.3. Theorem 10. There then exist distinct (x . if the modeling relation h is also a relation (that is. we have k 1 (x 1) = x 1 or x2 .1. ° id Y ) is the identity mapping on S. p ° h = (id X . y' ) ∈ S' be arbitrary.4.5. So. i. y').3 it follows that h : S → S' is surjective as a mapping. From Theorem 10.y'1 ) = (x2 y 1 ) ∉ S. v).

296 Chapter 10 Figure 10. (ii) If the modeling relation h is a section (that is. h has a left cancellation). but the converse is not true. but the converse is not true.3 we obtain Fig. then the modeling relation h is injective as a mapping from S to S'. Proof: (i) If h 1 and h 2 are injective. But. Therefore.5. Let S ⊂ X × Y and S' ⊂ X' × Y' be two systems defined by (10. (iii) If h is an isomorphism. Combining Theorems 10. then h is a monomorphism.5. Example 10. similar to the previous results.k 2 ) from D (S' ) × R ( S' ) → D( S ) × R (S ) be defined by (10.5. then h1 and h2 are injective.154) Define a modeling relation h = (h 1 . id Y' ). h2 ) from S to S' by letting h 1 (x1 ) = h 1 (x 2 ) = Let k = ( k 1 . define two systems S ⊂ X × Y and S' ⊂ X' × Y' such that (10. 10.156) .5. Theorem 10.12. but the converse is not true.5.4. S is not isomorphic to S' because   ≠ S' . the following hold: (i) If both mappings h 1 and h 2 are injective.12.5.153) and (10.5. We show that retraction does not guarantee an isomorphism. h is a monomorphism. then h is a section.1. Conversely. A modeling relation h = ( h 1 . we have the following.2 and 10.155) Then k is a modeling relation from S' to S and h k = (id X' . Under the same assumption as in Theorem 10. h 2 ) from S to S' is termed as a monomorphism from S to S' if h is injective as a mapping from S to S'. according S to Theorem 10. Then. Properties of the mapping h from S to S' with D(S ) = X and R ( S) = Y.1.

If then (10.163) and let the modeling relation h be defined as in the proof of (ii).161) This contradicts the hypothesis that k is a modeling relation from S' to S.4 because Theorem 10.157) by letting and Then h : S → S' is an injective modeling relation. Conversely.160) a contradiction. h 2 ) : S → S' is an isomorphism if and only if h 1 : D (S) → D( S' ) and h 2 : R (S ) → R (S' ) are bijective. but none of the mappings h 1 nor h 2 is injective.162) and (10. then either then (10. k 2 ). (ii) Let be a left cancellation of Then h 1 and h 2 are injective. Let (10. Then h 1 and h 2 are both injective.159) and define mappings h 1 and h 2 by and h2(y 2 ) = y'2 .5. k 2 ) of h as follows: But h cannot be an isomorphism from S to S' by and Theorem 10. and define .158) and (10. Define a left cancellation k = ( k 1 . (iii) We need only define two systems S ⊂ X × Y and S' ⊂ X' × Y' and a section h : S → S' such that h is not an isomorphism. A modeling relation h = (h 1 . Now if the modeling relation If h has a left cancellation k = ( k1 .5. define systems S ⊂ X × Y and S' ⊂ X' × Y' such that (10.6.Systems of Single Relations 297 and (10.

and where I is an index set.5. let h = (h1 .e. i. h = (h 1 .5 imply that the mappings h1 : D (S ) → D( S' ) and h 2 : R( S) → R (S' ) are surjective and injective.7.e. This gives Sufficiency. h 2 ) is a modeling relation from S to S'. for each Therefore. Then there are mappings k 1 : X' → X and k 2 : Y' → Y such that and (10. for any Then. Then the product mapping h 1 × h 2 reduces a S modeling relation from S to S'.k 2 ) is an isomorphism from S' to S. Suppose h is an isomorphism. From Theorem 10.5. In fact.3 and 10. h 1 × h 2 is a mapping from S to S' if and only if for any Proof: Necessity. respectively. Therefore.165) and are partitions of the sets S. is a covering of S. Sufficiency follows from Theorem 10. Theorem 10.5. Suppose h 1 × h 2S is a mapping from S to S'. h 2 (y ) ∈ S'(h 1(x )). of the input (ii) There is a relation Is ⊂ I x × I y with D (Is ) = I x and R( Is ) = I y such that are related by a one-to-one correspondence and ψ : I s → I with (10.298 Chapter 10 Proof: Necessity. in the future we can speak about isomorphic systems. The collection is referred to as an i. h 2 ) be an isomorphism from S to S'.6. if the collections Y. and In particular. the corresponding input–output system covering of S will be referred to as an input–output system partition.5. The system covering will be denoted by .164) Applying Theorem 10. it can be seen that if a system S is isomorphic to a system S'.6.5. Then Theorems 10.5. and S ⊂ X × Y and S' ⊂ X' × Y' be two systems. Let h 1 : X → X' and h 2 : Y → Y' be two mappings. it follows that k = (k1 . we have is. X. Suppose S ⊂ X × Y is an input–output system such that D( S) = X and R (S ) = Y. Suppose.4. then S' is also isomorphic to S. Then for any That x ∈ D (S) and any y ∈ S (x )... satisfies input–output system covering of S if satisfies the following two conditions: (i) There are coverings space X and the output space Y.

2. We now construct an input–output system S ⊂ X × Y which does not have a nontrivial partition. (10. partition) is termed a trivial covering (respectively.167) where x 1 .y 2 ∈ Y are distinct elements in X and Y. Each input–output system S ⊂ X × Y has an input–output system covering. In fact.169) To guarantee that for each i = 1.8. suppose S has a nontrivial input–output system partition Then the associated input–output system Is ⊂ I x × I y must satisfy Ix = 2 = Iy.166) The rest of the proof is clear. An input–output system covering (respectively. Suppose (10. if Ix > 2. partition) if Ix = 1 or I y = 1. This fact implies that if the input space X = {x 1 .5. in is nonempty. Proof: We define the collections and as follows: (10. Then S may not have a nontrivial partition. That is. not every covering is compatible with an input–output model structure. respectively.x 2 ∈ X and y 1 . . then the system covering would be a trivial covering. Generally.2. it can generate an approximation of the original input–output system. A similar argument shows that  and Then for each i = 1. for each (10.168) contradicts the condition that Iy has to be 2. Example 10. Theorem 10. respectively. In fact. Here.Systems of Single Relations 299 A system covering will be used when viewing a system macroscopically by ignoring its detailed structures. When it is compatible.x 2 } and {y 1 .x 2 } and the output space Y = {y 1 . there is an i ∈ I x such that X i ∈ contains neither x 1 nor x 2 . by contradiction.5.2.168) and each element Equation (10. the relation Is is a representation of the approximation and will be called an associated input–output system of the covering.y 2 }. Let S ⊂ X × Y be the system defined by (10.170) the sets X i and Y j must contain {x1 .y 2 }. contradiction. which is also an input–output system partition.2 and j = 1.

300 Chapter 10 When a system covering is a partition. suppose there exists (x. Let and i ∈ I } be an input–output system partition of a system S ⊂ X × Y. Suppose that h = ( h1 . Since (x.[y]) = [( x.y ) (x.176) .5. The definition of a system covering implies that (Xi × so there is Consequently.y ) ∈ S k for some k ∈ I (10. Theorem 10. j) ∈ I s .y ) ∈ S such we have that x ∈ Xi and y ∈ Y j . Conversely.175) A one-to-one correspondence ψ : Is → S/h 1 × h 2 is given by ψ ([x].10.172) Suppose (i . Since and are partitions.h 2 ) : S → is a modeling relation with D (S ) = X and R ( S) = Y. The uniqueness of existence of the associated input–output system Is comes from the fact that Let f: X → Y be a mapping from a set X to a set Y.9.5. we must have X i = X i' and Yj = Y j '.y ) ∈ S k hold. if the system Is exists. Theorem 10. X i ∈ and Yj ∈ In particular. An equivalence relation Ef can then be defined on X as follows: (10. we Let Since have x ∈ Xi ' and y ∈ Yj '.y )] (10.171) where Is is the associated input–output system of the partition. Then the following holds: and (10.174) The quotient set X/E f will be denoted by X/f. I s is unique for the given sets of and Proof: The following is clear: (10. the associated input–output system Is is characterized by the following results. There then exists an input–output system partition and is defined of S such that by (10.173) and (x.

e. the mapping ψ is injective. µ is surjective. Then the following statements hold: (i) µ = ( µ1 . h 2 ) : S → S' is a modeling relation. ψ is surjective. Theorem 10. 10.Systems of Single Relations 301 Figure 10.. Suppose that S ⊂ X × Y and S' ⊂ X' × Y' are input–output systems such that D(S) = X. there is an epi–mono decomposition of h.2. (iii) The diagram in Fig.13. Since holds.5. and R(S') = Y'.[y ]) = (h1 ( x) . it follows that that is. [y]) ∈ Is and be arbitrary such that and It suffices to show that (10. ψ is a mapping from Is into X × Y/h 1 × h2 .11. σ 2 ) : S/h → S' can be defined by σ ([x ].. Let ([x]. (iv) If h1 and h 2 are injective. Epi–mono decomposition of a modeling relation h. i = 1. We now need to show that for each i. . are injective. R(S) = Y. and h = (h 1 . we show that is a one-to-one correspondence. and Suppose Then i. that is.177) From the definition of equivalence class.e. Let be arbitrary. h2 ( y )) such that σ i . D(S') =X'.13 is commutative. µ 2 ) is an epimorphism from S to S/h. then σ is an isomorphism. (v) If h is surjective. (ii) A modeling relation σ = (σ1 . Proof: We first show that the correspondence ψ is well defined. In particular. Let µ 1 : X → X/h 1 and µ 2 : Y → Y/h 2 be canonical mappings. then the modeling relation µ is an isomorphism. there exists such that The last equation follows from the fact that ([x] × [y]) ∩ S ≠ Second.

y) = (µ 1 (x). If (x. of systems. an injective. From statement (i) and Theorem 10.180) .5. µ is surjective. y) ∈ ([x] × [y]) ∩ S. (ii) Since [x] = [x'] implies is well defined. y) ∈ S. we know that µ is an isomorphism. and S ⊂ X × Y satisfies the conditions is a collection (10. and be the diagonals of the product sets. For each system and each natural number n > 0. Proof: Let tively. [y]) ∈ S/h. the mapping µ 2 is also injective. [y]) ∈ S/h. and thus µ(x. [y]) ∈ S/h such that [x'] = [x] and [y'] = [y]. of S/h implies there exists such that and Hence. Since there is (x'. Proposition 10. µ 2 (y)) = ([x]. and σl and σ 2 are clearly injective. 2.5.302 Chapter 10 Proof: (i) It is clear that the canonical mappings µ 1 : X → X/h 1 and µ 2 : Y → Y/h 2 are surjective when defined by µ i (m) = [m] for i = 1. respec(10. says that µ 1 is injective. (v) The desired result can be shown as in the proof of (iv). Then the definition Consequently.178) and (10. S is called an isomorphic complex system representation of with such that and order n if there exist mappings is an isomorphism from S onto S'. (iii) The commutativity of the diagram in Fig. the modeling relation µ is an epimorphism. The mapping σ 2 is also well defined. That is. σ is a modeling relation. and an isomorphic complex system representation with order n.13 is clear from the definitions of µ and σ. B y Theorem 10. (iv) If h 1 and h 2 are injective.179) where we identify each element with The system S is called a surjective complex system representation of with and such that order n if there exist mappings S is called an injective complex system representation of with order n if there and such that ⊂ S and k 1 and k2 exist mappings are injective.3.6. Similarly. Suppose is a system. Suppose ([x].1.y') ∈ S for any ([x]. 10. has a surjective.5. (x.

Let S ⊂ X × Y be an injective complex system representation with order n. Since the associated input–output system is given by .Systems of Single Relations 303 We now define S ⊂ X × Y by letting (10. A system has an injective complex system representation S with order n over a class of systems iff there exist n partitions of for each i = 1. (10. . . n. (10. respectively. 2.2. Let be the projection defined by Eq. . . .183) Proof: Necessity. for arbitrary and for each i = 1. (ii) There exists an embedding where is an associated input–output system of Pi with σ1 i and σ2 i injective.96). (iii) Let and be the input and output spaces of respectively. h 2 ). . 2.. . . n. . we can relation. . Then Π i : S → S i is a modeling relation.5. .182) and (10. . such that the following conditions hold: (i) Pi is an input–output system partition of . Then. Theorem 10. × X n and Y ⊂ Y 1 × Y2 × . i = 1.181) and and and define mappings Then we have the desired results with respect to (h1 . × Y n . .n.12.184) We now check the conditions (i)–(iii) for these partitions Pi . Since define a partition Pi of by (10. where X ⊂ X1 × X 2 × . . . condition (i) holds. . . Let be an injective modeling is a modeling relation. Since Theorem 10. where and are partitions by definition and [ x j]i and [yj ] i are equivalence classes of and . n.10 implies that Pi is an input–output system partition.5. . . for each i = 1.2.

Consequently.188) Since. Suppose satisfies conditions (i)–(iii).187) Hence. Let X' = D(S') and Y' = R (S') and mappings and be defined by (10. condition (ii) also holds. A similar argument works for the collection . In the present case.191) for each and to S'.5. we have showed condition (iii).186) Suppose Since for each i. k 2 ) is an isomorphism from .2.11(ii). . (10. . . (10. Sufficiency. We will check condition (iii).185) Choose an equivalence class [x i ]i from each of implies for some for i = 1. Hence.189) Since e 1 is injective. Claim. implies (10. The mapping k = (k1 . Let be defined by (10.190) where we identify singleton sets with the elements. is empty or singleton. (10. n. .304 Chapter 10 there is an embedding mapping due to Theorem 10. The definition so (10. in general. Hence.

196) This ends the proof of Theorem 10. To show that k is surjective from onto S'. k1 is injective. we let It follows from the definition of k 1 that The condition (iii) implies that (10. let be arbitrary.197) has a surjective complex system repthen the . × S n .Systems of Single Relations 305 In fact.5.9 implies that Since condition (iii) implies that (10. Since there is an embedding σi from Si . . We now let S = S 1 × S 2 × .e. Theorem 10. The same argument can be used to show that k2 is also injective.12. for any Therefore.194) Consequently. k is a modeling relation from to S'. .192) we have That is. and (10. If a system resentation S with order n over a class following conditions hold: (i) has n input–output coverings (10.5.13. let be arbitrary.195) This completes the proof that k is an isomorphism from to S'. Theorem 10. an injective modeling relation from to S is to (10.5. To show k is one-to-one. Then the definition of S' implies that there exists such that (10.193) Hence.. and i.

. let and and (10. (10.200) Next let C i be the subset of C' defined by deleting empty sets and duplicated sets i and write (10. Since h is surjective. let Then C i' is a covering of for some be arbitrary.202) and (10.201) Then C i is an input–output system covering.203) Then Xi and Y i are coverings of the sets and respectively. In fact. and let ψ be the identity mapping on I s .198) In fact. and the projection. h 2 ) : S → be a surjective modeling relation. For each i let (10.306 Chapter 10 (ii) There is a surjective mapping from Si to an associated input–output system of C i. Next let the desired i i i relation between I x and I y b e I si . Proof: Let h = (h 1 . (10. Then (ii) holds.199) That is.

204) (i) has n input–output coverings (10. If the following conditions are true. .210) . Suppose that is a system and { Si ⊂ X i × Yi : i = 1. n } is a family of systems.209) where each singleton set is identified with its element. where and and (10. has a surjective complex system representation with order n.207) and (10. Let X' = D ( S') and be the mappings defined by and Y '=R (S ') and (10.206) (ii) There is a surjective modeling relation associated input–output system of C i.14. S ⊂ X × Y.5.Systems of Single Relations 307 Theorem 10. (iii) Let output spaces of and respectively.208) Proof: Suppose (i)–(iii) hold and Isi is an associated input–output system of C i . . . .205) (10. 2. Then for any from Si to an be the input and and (10. Define a system as follows: (10.

2. (10.212) Consequently. . i = 1 . .. 5. (ii) There is an isomorphism σi from an associated input–output system P i to S' i . of .5.12 and 10. and that is.308 Chapter 10 and Condition (iii) and the definition of S' ensure that h1 and h 2 are well defined. . We now show that h = (h 1 . If (x. For each i. Suppose is arbitrary. then the definitions of S' and h imply that ( h 1 (x ). A system has an isomorphic complex system representation S of order n over a family of systems if and only if there are partitions Pi .214) Then h and ° σ is a surjective modeling relation from S to Theorem 10. . (10.15. We now define σ : S → S' b y and (10.5. of such that the following conditions hold: (i) P i is an input–output system partition.14 and is left to the reader.14 is true. The proof is similar to those of Theorems 10.y) ∈ S' . h 2 (y )) ∈ and hence h is a modeling relation. × S' . n. where Si' ⊂ S i and S ⊂ S' × .213) The modeling relation h is surjective.211) The definition of an input–output system covering implies that for some (10. h2 ) is a surjective modeling relation from S to S'.5. 1 n (iii) Condition (iii) in Theorem 10. .

who began to study this concept in 1964 (Mesarovic. 1983b.” Section 10.1 is a combination of a result in (Lin. For recent research. Bunge (1979). 1964). The concepts of feedback transformation and feedback decoupling shed some light on the matter. see Takahara and Asahi (1985) and Saito (1986. They named the concepts panchaos and panattractor. Mesarovic and Takahara (1975). for example. Deng (1985). and Saito and Mesarovic (1985). The study of decomposability conditions enlightens the research of relations between “parts” and “whole. 1982b.Systems of Single Relations 309 10. given in this chapter was introduced by Saito and Mesarovic (1985). Some References for Further Study The concepts of chaos and attractors of single-relation systems were introduced by Zhu and Wu (1987). The concept of feedback systems has been a major research topic in systems theory and has been studied in many different ways by. because the concepts tell when the system under consideration can be divided into smaller systems. 1985). and some results are from Lin and Ma (1990).5 is mainly from Takahara’s excellent article (1982). 1984c. 1984b. 1987). in which the concept of general linear systems was studied in detail. . General systems of a single relation originated with Mesarovic.6. The concept of decomposition of systems shows its power in Section 10.2. Rugh (1981). Its final form was contained in the book by Mesarovic and Takahara (1975). Because all living and many nonliving systems are feedback systems. Klir (1985). 1984d. The definition of feedback systems.4. and systematically studied them under the name pansystems theory. Theorem 10. the method of studying these systems is important. 1983a. 1982c. For more information consult Wu’s work (1981b. Feedback transformation has been a really hot topic in recent years. 1982a. The theory appeared in the late 1970s. 1989b) and a fundamental theorem given by Mesarovic and Takahara (1975).

A comprehensive understanding of mathematical analysis is needed to understand this chapter. and f(x) a function from a subset of GNS to a subset of GNS . For example. integral calculus. differential calculus. 11. Introduction Let be the set of all integers. we will always assume that each real-valued function is from into and each generalized number-valued function is from GNS into GNS. and their applications in the well-known Schwartz distribution theory are introduced. and the set of all natural numbers. A generalized number x is a function from into such that there exists a number k = k(x) ∈ called the level index of x.1) into For convenience. continuity.CHAPTER 11 Calculus of Generalized Numbers As shown in Chapter 6.2) 311 . such that the concepts of limit. from considered a generalized number. ƒ(x ) represents a real-valued function. boldface letters are used for GNS-related concepts and lightface letters for real-valued concepts. Further research topics are also listed. and each such x is expressed as will also be (11. satisfying the condition (11. Without loss of generality.1. it is necessary to develop a more mature theory of mathematics of levels. denoted by 0. As before. the set of all real numbers. where is the set of all real numbers and GNS stands for the set of all generalized numbers. the zero function. In this chapter the system of generalized numbers is treated rigorously.

denoted x × y or xy. denoted by x ÷ y or x/y. . × .7) then the generalized number z is called the quotient of x divided by y. A2 x + ( y + z ) = ( x + y ) + z. y ∈ GNS is the lexicographical ordering.1. is defined as (11. Theorem 11. Let c ∈ and x ∈ GNS . z ∈ GNS.. and ordering relation of the generalized numbers are defined as follows: (1) Addition and subtraction. Let x and y be arbitrary from GNS .3) and x(t ) = x t .6) (5) Division. (GNS. The product of x and y.e. Proof: It suffices to show that (GNS. scalar multiplication. The operations of addition. y. then x ÷ y always uniquely exists. y ∈ GNS. +. for any x. × . A3 There exists 0 ∈ GNS such that x + 0 = x. if y ≠ 0 ∈ GNS. For x. The set of all generalized numbers will be written as GNS. and define (11. i. division. <) forms a non-Archimedean field. y ∈ GNS . Let x. for any x.5) (4) Multiplication. Define (11. if there exists a z ∈ GNS such that y×z=x (11. subtraction. for any x ∈ GNS. +.1. x < y iff there is an integer k0 such that and (3) Scalar multiplication. multiplication. The ordering relation between x.312 Chapter 11 where 1 (t) stands for the function from into such that if s ≠ t if s = t (11. <) satisfies the following axioms: A1 x + y = y + x.4) (2) Ordering. y ∈ GNS. As shown in Chapter 6.

M3 There exists 1 ∈ GNS such that x × 1 = x . y . N1 For each natural number n ∈ there exists x ∈ GNS such that n < x. R2 For any x. M1 x × y = y × x. there exists y ∈ GNS such that x × y = 1. D1 0 ≠ l. R5 For any x. there exists y ∈ GNS such that x + y = 0. M4 and R5. 0. for any x ∈ GNS . ) each natural number n ∈ such that k(n) = 0. . . 0. M2 x × (y × z) = (x × y ) × z. n. . . x < y and y < z imply x < z . . z ∈ GNS . (11. M4 For any 0 ≠ x ∈ GNS .x < y implies x + z < y + z . If u = 0 in Eq. if one to defined observes the following facts: In GNS 0 is the zero function from by 0(t ) = 0 for all t ∈ in GNS 1 is the unit element 1 ( 0 ) of the zeroth level.z ∈ GNS . for any x . x > y R4 For any x. if u ≠ 0. z ∈ GNS . one of the following must be true: x < y. and for treat n as the generalized number n = (. z ∈ GNS.Calculus of Generalized Numbers 313 A4 For any x ∈ GNS . we see that k(y) can be defined as –k(x). for any x ∈ GNS.8) where 1 = 1 (0) . y . x < y and z > 0 imply x × z < y × z . y ∈ GNS. and l(0) = 1. . for any x . 1 (u) = 0. Proof of Axiom M4: Assume that the fixed and unknown generalized numbers x and y have expressions Then the equation x × y = 1 to be satisfied by x and y is equivalent to (11. x = y.8). y. for any x . That is. y. y. z ∈ GNS . and . y ∈ GNS . It is straightforward to check all these axioms except. R1 x < x can never be true. R3 For any x. D2 x × (y + z) = (x + y) + (x × z).

k ∈ satisfy k < h. assume that k (y) = –k(x) –d. for s < k 0 .1. can be defined by By mathematical induction. so while other terms in the summation are the same as the corresponding terms in This ends the proof of axiom R5. To compare x × z and y × z. .8).2. It can also be shown that y is unique.314 Chapter 11 If u = 1. It then suffices to show that for all s < k 0 . To this end. Theorem 11. we get This ends the proof of Axiom M4. (11. By contradiction. in Eq. Eq. This contradicts the fact that Proof of Axiom R5: The assumption that x < y and z > 0 implies that there is t 0 ∈ such that x(t ) = y (t ) for all t < t0 . x( t 0 ) < y ( t0 ) and z( k (z)) > 0. and the generalized numbers in I( h) are infinitesimals compared with those in I( k) .8) becomes Hence. it suffices to show that there is no other way to define the level index k(y). then the generalized numbers in I(k) are infinities compared with those in I (h) . x(t 0) < y(t 0 ) implies since if v = k(z). let k 0 = t 0 + k( z). (x × z) (s) = (y × z) (s) and (x × z) (k 0 ) < (y × z) (k 0 ). for some d ∈ Replacing u by – d. Let If h. the generalized number y is defined. u has to equal t 0 . To this end. If s = k 0 = t 0 + k (z). (11. one has since v ≥ k (z ) implies u < t 0 and x( u) = y( u).

Eq. (11. . Then (1) k ( x + y ) ≥ min{ k( x). for any integer s. it suffices to see that k( z ) cannot be less than k(x) + k (y). (11. y ∈ GNS such that y ≠ 0 ≠ x. Proof: The argument for (1) is straightforward and is left to the reader. the definition of k (z) is unique. Theorem 11. When s = k (z ).Calculus of Generalized Numbers 315 The proof is straightforward and omitted.9) gives contradiction. (11. Assume that z = x × y and that Hence.1. To show the uniqueness of the index number k( z).9) becomes By applying induction on the index k( z ) + i. x × y = z is equivalent to. assume k(z) = k( x ) + k( y) – n. for some nonzero whole number n. (2) k( x × y) = k (x) + k (y).3. We can now define k(z) = k(x ) + k (y ). That is. k( y)}. (11. Eq. The proof of (2). Let x. (3) k(x / y) = k (x ) – k (y) . To this end. the generalized number z can be well defined. This ends the proof of (2). Replacing s by k(x) + k(y) – n in Eq. as defined.9) where k (x) ≤ m ≤ s – k (y ) and k ( y) ≤ n ≤ s – k (z ).9) becomes When s = k(z ) + 1.

z(k(z)) = x (k(x))/y(k(y)). Then replacing s by k(x) – n in Eq. Thus. When s is replaced by k(x) + 1 in Eq. (11. (11. 10). Assume that Then these numbers satisfy (11. one has Therefore. Thus. one has That is. .316 Chapter 11 Proof of (3). The whole number n must therefore be zero. k(z) must be unique. Let z = x/y. replacing s by k(x) in Eq.10) with k(y) ≤ u ≤ s – k(z) and k(z) ≤ v ≤ s – k (y). (11. and y can be found. The uniqueness of k(z) can be shown as follows: Assume that k(z) = k (x) – k(y) – n for some nonzero whole number n. 10). x .10) gives contradiction. we can continue this process and define each z(m) for m ≥ k (z) = k(x) – k (y). the relation between the generalized numbers z. By induction. Hence. This second inequality implies that we can define k(z) = k(x) – k(y). That is.

n ) continuity is quasi-(0. imply that Therefore. which is the ordinary definition of continuity of real-valued functions. for any x ∈ E imply that then f(x) is quasi-(m.2 implies that when the function value is infinity. one can still study the concept of continuity.1. there exist real numbers δ f > 0 and δ g > 0 such that for any x ∈ GNS . Continuity Let E be a subset of GNS and f : E → GNS a function. for any x ∈ GNS satisfying one has . 0) continuity. n ) continuity. if for any real number ε > 0 there exists a real number δ > 0 such that. n ) continuous at point a (Wang. A special case of quasi-(m. If n < m . n ∈ . If f(x) and g (x) are quasi-(m. then so are the functions h(x) = f (x) + g (x) and k(x) = f (x) – g (x). 1985). Proof: Since f(x) and g (x) are quasi-(m. Theorem 11. by the definition of quasi-(m. for any real number ε > 0. n ) continuous at point a . imply that and that for any x ∈ GNS . That is.1. n) continuous at a point a. the concept of continuity is expanded to include the study of continuity of infinities. For a fixed a ∈ E and fixed m.2. Theorem 11.Calculus of Generalized Numbers 317 11.2.

for some real number M. respectively. n 2 ) continuous at a. n2 ) continuous at a point a. Proof: Since f(x) and g(x) are quasi-(m. n 1 + k(g(a))}. n) continuous at point a. it has been shown that h(x) and k(x) are quasi-(m. imply that and k(f(x)) = k (f(a)) ≤ n2 . it follows that for any real number ε > 0 there exist real numbers δ f > 0 and δ g > 0 such that. and for any x ∈ GNS. where n = min{k(f(a)) + n 2 . and that imply that and Therefore. n ) continuous at the point a. from the definition of quasi-(m. n 1 ) and quasi-(m. for any x ∈ GNS satisfying .2. Then the product function h(x) = f(x)g(x) is quasi-(m. Theorem 11. n ) continuity. respectively. Assume that the functions f(x) and g (x) are quasi-(m.2. n 1 ) and (m.318 Chapter 11 and That is. for any x ∈ GNS.

Then. k (f(a)) – k (g(a))}. n f ) continuous at a point a and that g(x) is quasi-(m. imply that for some fixed positive and k(g(x)) = k (g(a)). the quotient function f(x)/g(x) is quasi-(m. for any x ∈ GNS. Proof: From the assumption that g(x) is quasi-(m.2. k (g(a))) continuous at a. k(g(a))) continuity of g(x) at a guarantees the existence of m and M. where numbers m. . n ) continuous at a.3. Assume that f(x) is quasi-(m. Theorem 11. n ) continuous at a. M ∈ The quasi-(m.Calculus of Generalized Numbers 319 one has and It has thus been shown that h(x) = f (x)g(x) is quasi-(m. k (g(a))) continuous at a such that g (a) ≠ 0 . it follows that for any real number ε > 0 there exists a real number δ g > 0 such that. where n = min{nf – k(g(a)).

it follows that for any real number ε > 0. . n f ) continuous at a.1.3 it follows that and that Therefore.320 Chapter 11 Now from the assumption that f(x) is quasi-(m. there exists a real number δf > 0 such that for any x ∈ GNS. imply that and Now let x ∈ satisfy Then from Theorem 11. the quotient function is quasi-(m. n ) continuous at a.

Let S ∈ be fixed such that for fixed m. and the ( m. let us look at the concept of 〈 ε– δ〉 limits in GNS first. for any real number ε > 0 there exists a real number δ > 0 such that for any x ∈ GNS .13) and (11. (11. n) limit of f (x ) at a is L.n)-differentiable at a. for any x ∈ GNS. Differential Calculus In order to simplify the discussion.n )-derivative of the function f at a (Wang.15) It is obvious that the (0. Assume that f : GNS → GNS is a function and a ∈ GNS a fixed generalized number.3. In this case.12) imply that (11. 1985). Eq.11) is read as the quasi-(m.11) stands for the fact that for fixed integers m and n. n ∈ for each real number ε > 0 there exists a real number δ > 0 such that. (11. imply that In this case.Calculus of Generalized Numbers 321 11. 0)-derivative of f is the ordinary derivative of real-valued functions. f is (m. For a fixed L ∈ (11.n )-derivative will be written as (11. .14) Then the generalized number S × 1 (n–m ) is called the (m.

n) continuous at a.3.3. there exists a real number δ > 0 such that. for any x ∈ GNS. imply that Since f is (m . then f is quasi-(m. imply that and So . It suffices to show that there exists a real number δ > 0 such that. for any x ∈ GNS.322 Chapter 11 Example 11. Define It can be shown that the (1.1. let ε be an arbitrary real number such that 0 < ε < 1.n) -differentiable at a point a ∈ GNS.–1)-derivative of g : GNS → GNS is Theorem 11. Let g : such that be an infinitely differentiable real-valued function is bounded and dt = 1. n)-differentiable at a. there exists an S ∈ such that for the chosen ε > 0. Proof: To show that f is quasi-(m.1.n) continuous at a. If a function f : GNS → GNS is (m.

(2) For any x. f is quasi-( m.15). (11. then (2) If s is a positive integer and then where m = k(x ) and n = sm. the axioms of a non-Archimedean ordered field (see the proof of Theorem 11. Proof: (1) From Eq.2.1) can be used to prove that .n ) continuous at a. (1) If f (x) = c ∈ GNS is a constant function.3. a ∈ GNS.Calculus of Generalized Numbers 323 and That is. Theorem 11.1. it follows that for any fixed a ∈ GNS.

Then is a constant and f'(m.17) . One problem with the concept of (m. .yn . .3. Suppose y = f ( x) = (. . Then. . in general yk depends on an infinite number of variables x0 .y 0 . . y– m .324 Chapter 11 Therefore. . . such that for each generalized number ε > 0. x – 2 . . . . . there exists a generalized number δ > 0 satisfying.n)-derivative is that the derivative is a generalized number in I (n – m ) . . . Theorem 11. As in the calculus of real-valued functions.n ) ( x ) The proof is straightforward and is omitted.16) If there exists a generalized number.x – 1 . . for all ∆ x ∈ GNS with (11. Suppose that c ∈ exist. . x 2 . In (Wang. let ∆ x and ∆ y represent increments of x and y such that where Let a ∈ GNS be fixed. That is. Let for each i ∈ (11.n) (x) and g'(m. denoted f'(a).) is a function from GNS into GNS. this problem was corrected by the following methods. . for each fixed index k ∈ y k is a function of x.x 1 .3. 1991). . for m = k(a) and n = sm .

Integral Calculus 11. we briefly look at some elementary definitions and properties of Lebesgue integration. (11. the concept of derivative of f( x ) at a ∈ GNS is has the box product topology. the theory established was often incomplete and also its applications were not flexible. as the theory grew more mature and more applications were found. With this ordering relation on GNS. For example. Symbolically. denoted by lim . denoted f'w (a) = K. the weak derivative of y = f(x). it was found that based on the consideration of continuous functions. Lebesgue Measures For convenience and completeness of this chapter. If for each real with the property that number ε > 0. However. denoted by on GNS can be defined as follows: x y iff for each i ∈ x i < yi . For example.4. For x. under certain circumstances the harmony between the two theories was destroyed. see (Kuratowski and Mostowski. such that whenever one has for each s ∈ then. In classical mathematical analysis. functions which are continuous everywhere are basically the center of the discussion of interest. a different ordering relation. 11.1.4. for each j ≥ k ( δ). it also seriously affected the development of the theory itself.18) Remark.Calculus of Generalized Numbers 325 Then f'(a ) is called the derivative of the function y = f(x) at a. there exists δ j > 0. even though it was restricted to the study of continuous functions. the theory of integration of continuous functions was developed with the theory of linear algebra as its background. a function from GNS into GNS. However. 1976). the concept is different from of limits introduced in the space L 2 . This background has had great impact on the establishment and development of the theory. it seemed that it was no longer sufficient to consider the everywhere continuous functions only. at the same time. K is called the weak derivative of f( x) at a. introduced so that the topological space For more about this topology. Topologically speaking. y ∈ GNS. 1991): Suppose K ∈ GNS is fixed. can be introduced as follows (Wang. At first. which is the center of the theory of real analysis.

. based on this fact. in applied situations. where L² stands for the space of all the functions which are Lebesgue square-integrable on This new concept of convergence has widespread applications. and A 1 . The open interval A in the space is assumed to be the set of points . (2) on each subinterval. . For example. Then is defined by ’s are open intervals and . Let E be For any interval A. and the coefficients of the differential equations in the models are not absolutely accurate. and therefore there is no way for one to talk about integrating each terms and summing the results. a sequence of open intervals such that . called the volume of A. the series must be uniformly convergent.” This is how the theory of Lebesgue integration was born. ƒ(x) cannot vary “too much” on each of these subintervals. where a i and b i are constants. That is. it means that some of the inequalities in its definition can contain the equals sign. Because of this. . on each of which ƒ(x) does not vary “too much. one should look at boundary problems from the viewpoint of approximation. there is no way that one can look at ƒ(x) as approximately a constant. (4) taking the limit to achieve the desired accuracy. In the theory of Riemann integrals. . . The exterior measure of E. denoted m * E. An . Similar problems exist in different areas of the old theory of integration. the limit of a sequence of integrable functions may not be integrable at all. . For instance. A2 . There are many good reasons for introducing new concepts.  A stands for a set of points in . . that these possess no practical meaning? Obviously not! since boundary problems are often approximations of some realistic situations. For this process to work. Without uniform convergence. If ƒ (x ) is discontinuous. The function is then considered to be not integrable. The essence of Riemann integrations involves (1) cutting the interval into finitely many subintervals. they are idealized mathematical models. It is unreasonable to restrict ourselves to the range of continuous and differentiable functions. looking at the function value ƒ(x) as a constant. the requirement of uniform convergence is often not satisfied or is so complicated that it only causes more trouble.” no matter how small each subinterval is. if it has been shown that the boundary problems of partial differential equations of some problems in physics do not have differentiable solutions. can one claim. the concept of generalized solutions was introduced.326 Chapter 11 the concept of lim in mathematical analysis. then we should try some new methods. Therefore. If A is an interval. Most arise because in the definition of definite integrals. (3) improving accuracy of the approximation by increasing the number of subintervals. to integrate each term of a given series. If ƒ(x) is “extremely discontinuous. On the other hand. . . too much emphasis is put on the requirement of continuity. partition the interval into small subsets. ƒ(x) must be “fundamentally” continuous.

x n ) be a function defined on a subset E ⊆ . . An . since a simple function can only have a finite number of different function values. One equivalent condition for a subset A ⊆ measurable is given in the next theorem.4. If Let ƒ(x ) be a real-valued function defined on a measurable set E ⊆ there is a partition of the subset such that on each E i the function ƒ(x ) is a constant. . the collection of all bounded subsets of is classified as measurable or not is measurable if B ∩ A is measurable measurable. .A 2 . Hence..Calculus of Generalized Numbers 327 Let A be an interval such that A ⊇ E.4. the limit of a sequence of simple functions may not be a simple function at all. A subset A ⊆ is measurable iff. a sequence of open intervals such that .. . A function ƒ(x ) is of simple functions. denoted as m * E.2. . . Therefore. In this case. for any subset T ⊆ . and . Theorem 11. . . by If m * E = m * E. and . An unbounded set B ⊆ to be for each chosen interval A. Assume that A and B are measurable subsets in respectively. Theorem 11. the exterior measure of A is called the measure of A and is denoted by mA. and A 1 . defined on measurable on E if there is a sequence E.. Define the interior measure of E. such that Let ƒ(x ) = ƒ ( x1 . However. ƒ(x) is a simple function. . It is obvious that the sum and product of simple functions are still simple functions. . Then C = A × B is a measurable subset in m(C) = mA × mB. . Define and The set is called the graph of ƒ(x). x 2 .1. the set E is Lebesgue measurable or simply measurable.

For . be a measurable subset with mE < + ∞.4. It is obvious that is also a partition.328 Chapter 11 Theorem 11. Assume that and are two partitions such that each E i* and E j** are measurable. respectively. Define which are called the small sum and the big sum of the function with respect to the partition . one has bmE ≤ s ≤ S ≤ B m E where b and B are the greatest lower and the least upper bounds of ƒ( x) on E . b i ≤ ƒ(x) ≤ B i for each x ∈ Ei .. A function ƒ(x) defined on a measurable subset E ⊆ is measurable iff the graph G( E. It can be shown that if D* is a finer partition than D and if s * and S * are the small sum and the big sum of ƒ( x) with respect to D * . the small sum and the big sum are always between bmE and BmE one can thus define Then where of ƒ(x) on E. Let us see how the Lebesgue integral of ƒ(x) is defined.e. then s ≤ s* ≤ S * ≤ S No matter what kind of partition D is.3. i. and ƒ (x) a Case 1: Let E ⊆ bounded function defined on E. which is a finer partition than either of the previous two. If and called the lower integral and the upper integral . then ƒ(x) is integrable and written as . Obviously. let b i and Bi be the greatest lower bound and the least upper bound of ƒ(x ) on Ei .ƒ) of the function is measurable in the (n + 1) dimensional space Now it is time to look at the concept of Lebesgue integrals.

and ƒ ( x) a function defined on E. then ƒ(x) is integrable on E. and ƒ( x) ≥ 0 a nonnegative function defined on E. To define the integral of ƒ(x) on E .Calculus of Generalized Numbers 329 Case 2: Let E ⊆ be a measurable subset with mE < + ∞ . {f (x)} n is integrable on E. define . If for each n ∈ . For let ƒ(x) be a function defined on each k ∈ . is a bounded function on E and satisfies . define Then If ƒ(x) is integrable on each . and we write For a general function ƒ(x) defined on E. then ƒ(x) has integral value on E. which is denoted by If the integral values of f + (x) and f – (x) are finite. . Case 3: Let E be a measurable subset of Define if x ∈ E otherwise Then F (x ) is a function defined on the entire space . let for any fixed n ∈ . assume ƒ( x) ≥ 0 for all x ∈ . ƒ (x) is integrable on E. Then {ƒ(x )} n . Without loss of generality. First. define when ƒ(x ) ≥ 0 when ƒ(x ) < 0 and when ƒ( x) ≤ 0 when ƒ(x ) > 0 If f +(x) and f – (x) have integral values on E which are not both infinities.

. . For a given function f : GNS → GNS. Integrals on GNS Let us now turn our attention to the concept of integrals of functions defined on GNS.2. y n . . the definition of integrals of functions defined on GNS. . Case 1: Suppose that for each x ∈ E. Define a new function f * : GNS → GNS as follows: if x ∈ E otherwise Without loss of generality. Define when ƒ(x ) ≥ 0 when ƒ(x ) < 0 and when ƒ(x ) ≤ 0 when ƒ(x) > 0 If as and are meaningful. . ƒ( x ) is integrable on . define the integral value of ƒ(x ) If these two integrals are finite. 11. where. y –m . new concepts of definite integrals and indefinite integrals will be studied.19) . with GNL standing for generalized number and Lebesgue. x – m = 0. Let f : E → F be a function from a subset E of GNS into a subset F of GNS. . y 0 . After that.4. To make this chapter self-contained.). . it can be assumed that each function f from a subset of GNS into a subset of GNS is a function from GNS into GNS . called a GNL-integral. let E = {x ∈ GNS : f (x) ≠ 0 }. studied in Chapter 6 will be reintroduced.330 Chapter 11 as the integral value of ƒ(x) on . When this is value is finite. in general. . . for all m ∈ and that y = f(x) = (. . let ƒ ( x ) be a general function defined on . ƒ (x ) is integrable on Second. y k = y k (x ) is a function of all Step 1: Define a subset H (0) ⊆ as follows: a 0 ∈ H (0) iff y– m = 0 where if m > 0 and y0 = y 0 (a 0 ) a function of a 0 (11. . . . of functions defined on GNS inductively as follows. We define the integral.

. . with x(k(x)) = a 0 . . denote the is meaningful. x n – 1 ). . (n ) For any fixed x0 . and set ( n – 1) Step n: Assume the real numbers a (0) .20) If the real-valued function ƒ(0) is Lebesgue integrable on then – H ( 0 ) has Lebesgue measure zero. . written is convergent and equals a. . . define a new function f (0) : GNS → GNS by . .22) If the real-valued function is Lebesgue integrable on . then Case 2: Partition the set E = {x ∈ GNS : f (x) ≠ 0 } as follows: where E k = {x ∈ E : k (x) = –k}.a have all been defined. consider the right-shifting of and the left-shifting of f(x): First. If the function f(x) is GNL-integrable. 0. a 0 . then the sum is denoted by a ( n ) . let (11.Calculus of Generalized Numbers 331 whenever x = (. .4. a (1) . x1 . define a subset H ⊆ as follows: an ∈ H ( n ) iff and where a function of a n (11. . x2 . . If that is — if there are only countably many nonzero summands and the summation is absolutely convergent. x n –1 . Definition 11. integral on by a ( n ) (x 0 . . .21) whenever let (11.1 [Wang (1985)]. . . . For each fixed k ∈ .). .

write such Step n: Let x 0 . write (GNS) Second. for each m ∈ Step 0: Define a subset H ( 0 ) ⊆ such that a 0 ∈ H ( 0 ) iff y– m = y – m (a 0 ) is a function of a 0 only for all m ≥ 0. Case 1: Suppose that for any x ∈ E the generalized number x takes the form x = (… . the concept of (G) integrals will be introduced in two parts. write (GNL) if f ( k ) is GNL-integrable. for any x ∈ . write Definition 11. s = 0. that an ∈ H – m Let be a function defined as follows: i f . ). 2. . In general. . then the function f(x) is A concept called (G) integrals was introduced by Wang (1985). . . x – m = 0. k(x) = 0. Define a new function f ( 1 ) : GNS → GNS by If f (1) is GNL-integrable. .332 Chapter 11 If f ( 0 ) is GNL-integrable. That is. iff y = y – m (a 0 ) is a function of a 0 only for all m ≥ n . Let As before. x 1 . Let f (0) : be a function defined by If each y– (s ) (a 0 ) is Lebesgue integrable. x 0 . x n – 1 be fixed. . 0.4. Define a subset H ( n ) ⊆ ( n ). x 1 . .2 [Wang (1985)]. . Let f : GNS → GNS be a function and E = {x ∈ GNS : f (x) ≠ 0 }. while the second part generalizes the first. The first part is a special case. . . If GNL-integrable and written is convergent. 1. . define That is. . . . x n . .

. denote it by for s = 0. . . define Second. 2. set (G) .4. denoted Case 2: Suppose that for some x ∈ E there exists m ∈ Partition the set E = {x ∈ GNS : f (x) ≠ 0} as such that x– m ≠ 0. For each s ≥ 0. 1. where for each First. 2. . Definition 11. 1. if the sequence is convergent. then the function ƒ(x) is (G)-integrable. write and if exists. . . . let (G) shifting of E 1 and a function f ( 1 ) : GNS → GNS by and If f ( 1 ) is (G)-integrable.3 [Wang (1985)]. s = 0.Calculus of Generalized Numbers 333 If each y– (n + s ) (a n) is Lebesgue integrable. define and exists If there are only a finite number of natural numbers s such that a– s ≠ 0. . define the rightIf f ( 0 ) is (G)-integrable.

. is convergent to a generalized num- The following result is clear. a 1 .integrable.4. it is then (G)-integrable and (11. If ber a. define the right-shifting of E n and a function f ( n ) : GNS → GNS by and If f ( n ) is (G)-integrable. (G) F -integral) of f is defined as follows: (11.26) (11. have been defined as before. Definition 11.334 Chapter 11 Inductively. .27) . Now a generalized number a n + 1 can be defined as follows: First. a 2 . . then we can make the following definition. Suppose that a convergence relation among functions defined on GNS has been chosen such that there exist a sequence and a function f : GNS → GNS such that (11. . then the function f is (G)-integrable.23) Now let F be a family of (GNL)-integrable (resp.4. a n have been defined. an . Theorem 11.4 [Wang (1985)]. a 2 .. . assume that for n ∈ the generalized numbers a 0 .. If f : GNS → GNS is (GNL). a 1 . . let (G) By mathematical induction it follows that if all the generalized numbers a 0 .4. .24) The (GNS) F -integral (resp. .25) (11. . . . (G)-integrable) functions.

The concept of indefinite integral of a function f : GNS → GNS was studied by Wang (1985) as follows..1. … . A Nontechnical Introduction to Schwartz Distribution The theory of distributions was developed to correspond to situations presented to the world of learning by physical experiences which are not adequately covered by the classical y = ƒ(x) notion of a function in mathematical analysis. and by : GNS → GNS is defined 11. 1951).5. the derivative of a . a – m .28) where it is assumed that the limits exist. Then the indefinite integral of y = f(x) is defined as where a = (… . pertaining to the phenomena of nature. together. has been widely used to tie many important and different instances. For example. That is. as introduced and codified by Laurent Schwartz (1950. a n . the theory has a coherence and a power that the theory of mathematical analysis lacks. Here. Assume that (GNL) exists and that the function value of f(x) depends only on the real variable x k ( x ) .e.Calculus of Generalized Numbers 335 and (11. There are many aspects of this conceptual power.” The class of all distributions or generalized functions includes many objects which are not functions at all in the sense of classical mathematical analysis. the operation of taking a derivative applies without restriction to distributions. the word “physical” really means “phenomenological” — i.5. rather. Schwartz Distribution and GNS 11. … . occurring in different parts of the world of learning. The reason why it is necessary to study the theory of distributions is not mere generality. a 0 . The concept of distributions. …) is fixed. A “distribution” is a kind of “generalized function.

like pulses or mesas (Figs. the operation of taking a derivative in the theory of distributions eliminates the defect of mathematical analysis.2). The motivations underlying the term “test function” are that test functions serve as tools in studying other functions. Let us now see how the idea of test functions can be used to evaluate an unknown function of interest.b]. (b) φ has compact support. Since continuous functions are distributions. φ(x) vanishes outside of some interval [a.1 and 11. where many continuous functions have no derivatives.b] .1. A mesa function supported on [a. the temperatures have a density φ (T) supported on [a. Assume that a chemist needs to test the properties of a certain substance at a fixed temperature t0.e. His objective is to find some law.1]. So. he gathers a batch of the stuff with temperatures distributed in a pulse φ (T ) near T = t 0 . There is no doubt that in reality some temperatures appear more often than others. Let us see how the test function φ is used to determine the unknown function f. can be constructed in an infinitely differentiable manner. 11. i.336 Chapter 11 Figure 11. . A pulse function supported on [0. say f (T).b] . Mathematically speaking. A function is a test function if (a) φ is infinitely differentiable. and a Figure 11..2. the law will normally be written as a function. and that curves with certain crude geometrical shapes. distribution always exists and is another distribution. Hence.

Then the action of a test function φ on f is defined to be The following result supports this example. written if converges the sequence of the k th derivatives (a) For each k ∈ uniformly to zero.b] .. (b) The φ n have uniformly bounded supports. be and g: Theorem 11. is called the elementary space. Definition 11. i. If φ is chosen differently with different temperature ranges. A distribution T is a mapping from the set D of all test functions into the real or complex numbers.5.5. such that the following conditions hold. there is an interval [a.1 [Richards and Youn (1990)]. he might say that by enough experiments φ we learn the “law” described by f. The set D of all test functions together with the convergence relation. Let f: then for all be piecewise continuous. In the language of the chemist.1. (b) (Continuity) If Example 11.Calculus of Generalized Numbers 337 “weighted average” is natural to consider: Now φ is used to measure f.e. Similarly. φ n → φ means ( φ – φ n ) → 0. where T ( φ ) is written as (a) (Linearity) D and constants a and b. For any φ ∈ D define It can be shown readily that T is a distribution. defined earlier. . such that every φn (x ) vanishes outside of [a. Then f = g. A sequence of test functions converges to zero. independent of n. eventually the structure of f will be known.5. Let ƒ : be such that dx exists and is finite for any finite interval [a.b ].b]. Let f : continuous and satisfy that every test function φ has the same action on f and g.1 [Richards and Youn (1990)].

2 [Richards and Youn (1990)]. Let f(x ) be a piecewise continuous function such that Denote fa (x ) = af ( ax ). the δ function is defined by the equation Now shift the δ function through a distance +a. In terms of distributions.2. the following theorem is interesting. and That is. For a different discussion on the δ function. see Chapter 6. . Then where the convergence is in the sense of distribution. written as Tn → T if is Based on this concept. The following may be a little surprising: and in general. Dirac’s δ function δ is a symbol which satisfies the following conditions: (11. This should give the function δ ( x – a ).5. Theorem 11.29) where φ ∈ D is any test function. convergent to T.338 Chapter 11 Example 11. Let be a sequence of distributions and T a distribution.5. the shifting δ(x – a ) evaluates the test function φ at x = a.

refer to the splendid work by Richards and Youn (1990).5.30) and.3.2. such that for each k ∈ det (11. GNS Representation of Schwartz Distributions be functions from Let linearly independent if for any constants into These functions are the equation implies that Otherwise. (11. the functions are linearly depenF is linearly independent if each dent. k = 1.5.. Since f n+1 is not the zero function.. Thus. none of the functions such that in F is the zero function. there are sequences of real numbers.. the real numbers have been defined such that Eq. pick a real number is different from all previously chosen and that implies such that such that .2.. For linearly independent continuous functions 1.. For a set F of functions from into finite subset of F contains linearly independent functions.Calculus of Generalized Numbers 339 For a more comprehensive nontechnical introduction to the theory of Schwartz distributions. Lemma 11.1.2. 11. if then Proof: Since the set F = is assumed to be linearly the equation independent for any constants implies that Therefore.... there exists a real number Assume that for a natural number n.30) holds and that Now we define the real numbers the desired conditions are satisfied.

30) holds true and if then Lemma 11. For any a...340 Chapter 11 The existence of previously chosen that comes from the fact that f n +1 is continuous and that all are a finite number. induction guarantees that there are sequences of real numbers. there exists a countable subset F ⊆ D such that for any φ ∈ D. i. Now pick a real number such The existence of such a number comes from the fact that fn and fn + 1 are linearly independent...e. The elementary space D of all test functions is separable. b ∈ . k = 1. there exists a F such that φn → φ.31) Then define is a test function with support [0.1].2. such that for each Eq.5. which implies that can be different from all The continuity of f n and f n +1 guarantees that the previous By mathematical induction it can be shown that the real numbers can be chosen one after another so that each one is different from all previous and that for each i satisfying Therefore. (11. sequence Proof: Let h : be defined by (11.2..

. for n ∈ .b (x) pn (x) Φ a. and a and b are rational}. define Let G = {p n (x) Φ a. and a and b are rational numbers}.5. Let p n (x) be a polynomial of degree n with all rational number coefficients.b (x). Theorem 11.5.32) Here f is called the GNS function induced by ƒ.Calculus of Generalized Numbers 341 Then φ a. n ∈ . Then p n (x) φ a.33) where f : GNS → GNS is induced by ƒ. Let T be a distribution and F as in Lemma 11. for each ƒ ∈ F.2. define a function f : (11.b is a test function with support [a. Proof: Let For any rational numbers a and b.b (x ) is a test function. It is left to the reader to show that for any φ ∈ there ⊆ F such that φ n → φ .3 [Wang (1985)]. (11. Then F ⊆ is countable. Then there exists K : GNS → GNS such that. exists a sequence For each infinitely differentiable function ƒ : GNS → GNS as follows: → .b (x) : p n (x) is a polynomial of degree n with all rational number coefficients.b (x) : p n (x) is a polynomial of degree n with all rational coefficients. Then the families F and G can be put into 1–1 correspondence by matching up pn (x) φ a. b]. Let F = {pn (x) φ a.

35) satisfying (a i . That is. . . say Let d 1 . (11. … . . .ƒk are linearly independent. so that each power of x appears at most once. assume where Φ ai . whenever i ≠ j. b i ) ≠ ( a j .bi ( x) is defined by Eq. We obtain S. b i ) ≠ ( a j .34) then becomes (11.37) according to the powers of x appearing in the polynomials Pn i (x). .342 Chapter 11 Now we show that the family G is linearly independent. .2 . Eq. (11. Equation (11. .34) becomes (11. k. . . dk . d 2 . Thus. That completes the proof that ƒ1 . .d 2 . i = 1.34) It suffices to show that d 1 = d2 = . bj ). b j ). and the polynomial (11. assume and that (a i . k linear equations in k variables d1 .36) where the coefficients cij ’s are rational. = d k = 0. Case 2: For each i with 1 ≤ i ≤ k. . … .35) Now choose k different values of x for which it is easy to evaluate the e-functions. . whenever i ≠ j. from which it can be shown that the system has zero solution only. d k be constants satisfying (11. Pick an arbitrary finite subset H ⊆ G. Case 1: For each i with 1 ≤ i ≤ k. d1 = d 2 = . (11.ƒ2 . .37) Now group the terms of Eq. = d k = 0.

Eq. .2. (11. (11. Eq. .30) holds for n = m 0 + 1.30) is true. The only difference which might occur is that some constant coefficients may show up in front of some di ’s in Eq.35). .35). = d k = 0.3. . That is. .ƒ2 . there must be a real number c 1 such that (11. Define a function K : GNS → GNS g(x) ≠ 0} is bounded and as follows: for m ≥ 0. .39) (2) When n = 2. (11. the argument in Case 1 implies that d1 = d 2 = . ƒ1 .38) From Lemma 11. Since G is countable. . This completes the proof that the family G is linearly independent. n 2 . Hence the system (11.30). n k } + 1 many equations. enumerate the elements in G as (11. the following . (3) In general. satisfying Eq. . . . Therefore. . . .2 . . (11. each of which looks similar to Eq. . . . m 0 + 1 has A unique solution: system with unknowns (1) (1) (11. .37) is equivalent to n = max{n 1 .40) has a unique solution which is denoted c1 and c 2 . .41) where Choose a real-valued infinitely differentiable function g(x ) satisfying {x ∈ : g(x)dx = 1. Hence.37) is true iff the coefficient of each power of x equals zero. k = 1. (11. Since Eq.5.1 it follows that there are sequences for k = 1.ƒk are linearly independent. (11.Calculus of Generalized Numbers 343 Equation (11. (1) Since (0) of real numbers. assume that for a fixed have been defined for k ≤ m + 1 ≤ m 0 + 1.

at the time criticized and not trusted by many. introduced by Frechet (1945). he discussed some algebraic problems of his nonArchimedean field and applications in Voronese’s non-Archimedean geometry (Voronese. see Eq. his work on ordered algebraic structures is a continuation of Levi-Civita’s formal power series. let K · F = {K · f :ƒ ∈ F }. (11. 1936a. This completes the proof of the theorem. Yasui (1975). one need only to consult the p-adic topology and Stone duality theorem in Boolean algebra (Nyikos and Reichel. Stevenson and Thron (1971). To see this. given in Lemma 11.33). Later. non-Archimedean topology has been applied in research of abstract distances. The study of nonArchimedean topological structures has been valuable in the study of dimension theory and in research in algebra and algebraic geometry. (11. Hans Hahn has been commonly recognized as the first person to study ordered algebraic structures. (11.32). Hodel (1975). Juhasz (1965). Theorem 11. In fact. 1975). Husek and Reichel (1983). 1891). As for the study of non-Archimedean analysis. The purpose of Levi-Civita’s work was to establish an extremely rigorous “arithmetic” foundation for Voronese’s geometry. including Hausdorff. 11.5. The concept of linear uniformity. Colmez (1947). 1936b. introduced by . The generalized number system. In particular.3. Kurepa (1934. 1896. Doss (1947). Voronese. and the theory of ω µ -additive topological spaces.5. and Nyikos and Reichel (1976). are related to the general study of non-Archimedean topology. Hilbert employed Levi-Civita’s work on formal power series in his work on foundations of geometry. 1956).41) that K : GNS → GNS satisfies Eq. A Bit of History Research in non-Archimedean fields has been closely related to those of non-Archimedean topology and non-Archimedean analysis. of ele- where T is the distribution used to determine K. studied by many. f : GNS → GNS is the GNS function induced by ƒ. Sikorski (1950). For each arbitrary test function ƒ ∈ mentary space.4 [Wang (1985)].2 and for the function For the countable subset F ⊆ K : GNS → GNS introduced in Theorem 11.5. introduced and studied by Frechet (1946). where for each ƒ ∈ F. Also.344 Chapter 11 It then follows from the definition of (GNL)-integrals and Eq. Shu-Tang Wang (1964). etc. Then the following result is not difficult to see. Italian differential geometer Levi-Civita (1892) introduced and studied the concept of formal power series.6.

division. (3) Develop workable and concrete formulas for the operations of multiplication.Calculus of Generalized Numbers 345 Wang (1985).3. (2) Find the relation between the ordering relations defined in Sections 11. Wang did not have a chance to reorganize his work. was first studied with the definition of multiplication operation (by Wang) in 1963. real analysis. the paper was not published and got lost. similar to the transfer principle in nonstandard analysis (Davis. The major difference between the study of formal power series and that of generalized number system is that even though algebraically the generalized number system is a subfield of the formal power series. through whom Levi-Civita’s formal power series was introduced to Wang. the concept of Taylor expansion. His goal was to study strange functions. (4) Introduce the concept of continuity based on the ordering relation defined in Section 11. which appear in modern physics. along with the successes which have been or will be accomplished in the realm of systems science and its applications. As shown in Chapter 6. Until 1977. one can be sure that the theory of generalized numbers will be developed a great deal. Some Leads for Further Research Comparing the study of the generalized number system with those contained in mathematical analysis. Wang.3. defined in Section 11. Also. It is obvious that the GNS field is not complete in the sense of Dedekind cuts. on GNS. governing the relation between a set of properties of real numbers and a set of properties of generalized numbers.1. it is necessary to (1) Study the ordering relation. and complex analysis. Unfortunately. introduced the concepts of various integrals and applied the theory of GNS to the study of various problems in modern physics. 11. . it can be seen that understanding the GNS is still at the germinating stage. and taking roots. More work needs to be carried out in order for us to gain a deeper understanding of nature and the world in which we live. and studied in some detail in this chapter.1 and 11.7. (5) Establish a law. due to the Cultural Revolution in China. Particularly. Dirac’s δ function in quantum mechanics and the Heaviside function in quantum field theory have been given concrete representations in GNS. Wang learned of Laugwitz’s work (1968). In 1978. he discussed formal power series from the angle of classical mathematical analysis. For instance. In Laugwitz’s research. 1977). and all elementary functions were extended to the GNS field.

(10) Use GNS representations of Schwartz’s distributions to explore new explanations of equations. (13) Explore applications of the theory of GNS in other branches of human knowledge. which do not make sense in classical mathematical analysis. more puzzles in current quantum mechanics can be explained more rationally. in general. the products of distributions might not exist? (12) Redevelop quantum mechanics based on GNS so that. hopefully. (8) Generalize the Fundamental Theorem of Calculus to the case of GNS. definite and indefinite integrals of given functions from GNS into GNS . . linking GNS differential calculus to GNS integral calculus. maximum and minimum theorem. (7) Find detailed formulas to compute various derivatives. etc. The future success of the theory of GNS may well depend on various successful applications. (9) Develop an integral calculus of GNS based on Lebesgue integrals.346 Chapter 11 (6) Develop the results of generalized numbers corresponding to theorems in mathematical analysis. appearing in the theory of distributions. (11) Employ GNS representations of Schwartz’s distributions to solve open problems in the theory of distributions. Can the theory of GNS be used to explain why. such as intermediate value theorem. such as the problem of defining products of distributions.

Led by von Bertalanffy and supported by several of the most powerful minds of our time. and notably unsuccessful applications of general systems theory [for details see (Berlinski. Technical difficulties of introducing an ideal definition for general systems. 12.” (3) Based on historical experience and 347 . developed on the concept of general systems. 1978)]. a global systems movement has been going on for several decades. under the name of systems analysis. Starting in the early 1960s. Along the development path of the systems movement. The main part of this chapter is also contained in (Lin et al. many scholars from different disciplines have tried to first define the concept of general systems and then apply the concept and the theory. developed in various disciplines.CHAPTER 12 Some Unsolved Problems in General Systems Theory In this chapter some open problems in set-theoretic general systems theory are listed. 1997).1. to solving problems in various research areas. Introduction Von Bertalanffy introduced the concept of systems in biology in the late 1920s (von Bertalanffy. 1934). In this circumstance. senseless transfer of results from one discipline to another have really made many scholars doubt the value of “general systems theory. upon which a general systems theory could be developed so that this theory would serve as the theoretical foundation for all approaches of systems analysis. and it has been developed a great deal in the last 70 years or so. scholars began to establish a theory of general systems in order to lay down a theoretical foundation for all approaches of systems analysis developed in different disciplines. (2) Without an ideal definition of general systems in place. “general systems theory” would be meaningless. Lilienfeld. have taught many painful lessons. Among them are: (1) There might not exist an ideal definition for general systems.. A common feature of these problems is that their studies will lead to some new understanding in several research areas and the introduction of new concepts. 1976.

and calumnies. existence of absolute truths unreasonable effectiveness of mathematics. For example. prejudices.). For example. No matter how suspicious some scholars are. a profound thinker of the fifteenth century. nebulous substance (Perlman. linking medieval . or supernatural intervention. meeting challenges and self-renewing are among the most important motivations for a newborn theory to grow. The idea of systems appeared at least as early as the one-element Ionians (624–500 B. have taught us that no new theory would be born without suffering from suspicions. the need to develop a mathematics of multilevels and its application in Einstein’s relativity theory. philosophy of mathematical modeling. The Concept of Systems The fundamental characteristics of a system are the unification of “isolated” objects. the development histories of Cantor’s set theory. history will still proceed at its own pace. or some indeterminate. to mature. in the search for order. one can see that the study of general systems theory will surely impact the entire spectrum of human knowledge. Nicholas of Cusa.348 Chapter 12 the “unreasonable effectiveness” of mathematics (Wigner. each Ionian believed that all things have their origin in a single knowable element: water. How the systems movement is regarded is each scholar’s opinion and position.C. To meet the challenge of showing the need and power of the concept and theory of general systems. the real challenge from scholars — namely to show the need and power of the concept of general systems and its theory it is necessary to obtain results which have not been nor could be. and the structure of layers. From this brief list. mythological. several fruitful definitions of general systems have been introduced since the early 1960s. In fact. meaning of a system with contradictory relations. the need to study an applied set theory. the Law of Conservation of Matter–Energy. the origin of the universe. the material foundation of human thought. why the concept of systems was not studied more intensively until now. To this end. They sought causes and explanations in terms of the eternal working of things themselves rather than in any divine. Ionians had a naturalistic and materialistic bent. Einstein’s relativity theory. In history the idea of systems has been used by many great thinkers to study different problems. 12. multilevel structure of the world. air. Aristotle’s statement “the whole is greater than the sum of its parts” could be an early definition of a basic systems problem. Even though abstract general systems theory has been established through the efforts of many scholars. fire. etc.2.. Looking for a single basic reality. 1970). we list some the following problems: infinite or finite divisibility of the material world. achieved without using the concept and its theory — has not been very well met. relations between the objects. and to gain a visible position in scientific history. 1960). including Mesarovic and Takahara.

a new concept. i ∈ I. After 70 years of testing. as an attempt at an explanation. supraindividual organizations of higher order than the usual objects of observation. Leibniz’s hierarchy of monads looks quite like that of modern systems. known as the author of the psychophysical law. every object is infinitely divisible. 1975): A (general) system S is a relation on nonempty sets Vi : where I is an index set. 1977). elaborated. Using this idea of layers inductively. With his philosophy. The Leucippus–Democritus . … Can this process go on forever? If the answer is “yes. in the early 1960s introduced the mathematical definition of (general) systems. This view. von Bertalanffy (1934) wrote Since the fundamental character of the living thing is its organization. we shall call “organismic biology” and. we can see that an element S 1 of S could be a system. and the structure of layers. what is left would have characteristics of the original substance. considered as a method of investigation. and preferred to accept the immediate diversity of things as is. an element S2 of the system S 1 could be a system. based on Cantor’s set theory. Here. To establish a theoretical foundation for all approaches of systems analysis. Gustav Fechev. In the second decade of this century. Mesarovic. 1972). developed in different disciplines. as follows (Mesarovic and Takahara. thus romantically anticipating the ecosystems of modern parlance (von Bertalanffy. We list only a few of the great thinkers. could be systems too. “system” was introduced.C.” Here. relations between the objects. i ∈ I.. We believe that the attempts to find a foundation for theoretical biology point at a fundamental change in the world picture. No matter how far an object is divided. in the way of the naive philosophers of the nineteenth century. S represents a relation between objects. and elements of the sets Vi . Thus the chief task of biology must be to discover the laws of biological systems (at all levels of organization).C. introduced the notion of the coincidentia oppositorum. The opposite idea that the world is made up of fundamental particles appeared during 500–55 B.” then “the world would be infinitely divisible”! If the answer to the question is “No. This structure of general systems reflects the unification of “isolated” objects. the customary investigation of the single parts and processes cannot provide a complete explanation of the vital phenomena. this concept has been widely accepted by scientists in all disciplines (Blauberg et al. are the objects of the system.). elements in the nonempty sets Vi . “the system theory of the organism. This investigation gives us no information about the coordination of parts and processes. He was unwilling to submerge the tremendous varieties of things into any common denominator.Some Unsolved Problems in General Systems Theory 349 mysticism with the first beginnings of modern science.” does that mean the world is made up of fundamental elements? The idea that the world is infinitely divisible can be found in Anaxagoras’ “seeds” philosophy (510–428 B.

.. see (Lin. computer technology. respectively. The theory deals with general systems.E. Anaxagoras’ seeds. “Atom” means “not divisible” in Greek. and yet was an improvement over all of them. 1989c). etc. { x > y. mathematical logic. and shengke (means survival and vanquishing). Starting in 1976. which is one of the definitions with the widest meaning among all definitions of systems. the universe originally and basically consisted entirely of atoms and a “void” in which atoms move. Closely related to the concept of systems are the following questions: (1) What is the meaning of a system with two contradictory relations. it follows that generally these two questions cannot be answered. This fact implies that there are systems. relations. transformation. developed on the ZFC axioms is consistent. a theorem of Gödel shows that it is impossible to show whether the systems representation of mathematics. reveals one fact: History is in a special moment.” Pansystems analysis is a new research of multilevels across all disciplines. Xuemou Wu and his followers have studied many different theories. where the accumulation of knowledge has reached such a level that each discovery of a relation between different areas of knowledge will materially produce a useful or consumable product.g.g. climate control of giant buildings. so that we do not know if propositions with contradictory meanings exist or not. symmetry. generalized calculus. designs of satellites. the ordered triple W = ( C . Klir’s definition contains the most general meaning of the concept of systems originally posed by von Bertalanffy. the system of set theory on ZFC. Concerning this. Klir (1985) introduced a philosophical concept of general systems.” Theoretically. which reads “a system is what is distinguished as a system. and mechanical structures into one solid body of knowledge. To Leucippus and Democritus. e. C and E are called the composition and an environment of the system W . the theory blends philosophical reasoning. Based on research of these emphases. . The discussion here gives rise naturally to the following question: Why was the concept of systems not studied more before now? One answer is that the development of modern technology. Upon consideration of the interrelationship between the systems under concern and some environments of the systems.350 Chapter 12 atom combined features of the Ionians’ single element.. called the emphases of pansystems. for example. S ) is a system over T if and only if C and E are mutually disjoint subsets of T and S is a nonempty set of relations on C ∪ E. For a detailed discussion of the systems representation of mathematics. under the name “pansystems. e. x ≤ y } as the relation set? (2) How can we know whether there exist contradictory relations in a given realistic system? From Klir’s definition of systems. Bunge (1979) gave a model of systems as follows: For nonempty set T. This term was intentionally chosen by Democritus to emphasize a particle so small that it could no longer be divided. This means that maybe not every system is consistent or that not every system has no contradictory relations. and some thoughts of other schools.

mathematics . Russell’s paradox implies that there does not exist any statement (or relation) which describes a fact among all systems. does Russell’s paradox say that there is no universal truth or absolute truth? It is well known in the mathematics community that no theorem is true in all axiom systems. the chief task of modem science should be a systematic study of the world. 1972). and any practical problem and phenomenon cannot be described perfectly by only one cause–effect chain. 1972). it should be recognized that the world we live in is not a pile of innumerable isolated “parts”. the following is true: There does not exist a system whose object set consists of all systems. each relation r ∈ R can be understood as an S -truth. At the same time.. respectively. The sets M and R are called the object set and the relation set of S. study each isolated part (Kline. as Klir’s definition of systems says. if and only if S is an ordered pair ( M. etc. Countless examples have shown that mathematics is extremely effective when applied to the analysis of natural systems. because their methods emphasize separating the problem and phenomenon under consideration into isolated parts and processes. more and more scientists and administrators are forced to study problems with many cause–effect chains (i. and the connection between the interior and the exterior of different things. But according to von Bertalanffy. Hence.3. Some examples are the many-body problem in mechanics. and not only those. Descartes’ and Galileo’s methods have been very successfully applied.Some Unsolved Problems in General Systems Theory 351 By rewriting Russell’s famous paradox in the language of systems theory.e.R ). In the history of science and technology. That is. 12. Descartes and Galileo developed individually the following methods about scientific research and administration: Divide the problem under consideration into as many small parts as possible. The elements in M are called the objects of S. Now. The basic character of the world is its organization. However. Mathematical Modeling and the Origin of the Universe It is well known that almost all laws of nature can be studied in the language of mathematics. Descartes’ and Galileo’s methods need to be modified somehow. They guaranteed that physics would win great victories one by one (von Bertalanffy. 1987). because their methods are still widely used for research of fundamental theories and in modern laboratories. In the study of such problems. and simplify the complicated phenomenon into basic parts and processes (Kuhn. where S is a system. due to the tendency of modern science toward synthesis and the transverse development of modem technology. Thus. If any structure in the world can be studied as a system. 1962).R ) of sets such that R is a set of some relations defined on the set M (Lin. for any given system S = (M. the body structure of living things (but not the particles constituting the living things). systems). r is true among the objects in S.

called the vase puzzle. At the same time. at this juncture. six.R) of sets. The elements in M are called objects of S and M and R are called the object set and the relation set of S. they form a basis for which the structure of a mathematical theory can be developed independently of their genesis.. T ∪ K ) with T the collection of all axioms in the ZFC axiom system (Kuratowski and Mostowski. for our pleasure. we discuss some problems related to the structure of mathematics. applications of mathematics. . will show that different mathematical understandings could be contradictory. “set” and the relation “∈ ” of membership. . to wide branches of learning. which have not been assigned any specific meanings. In this section. we feel that mathematics is unreasonably effective as applied to the study of natural problems. All we can assert so far is that any specific interpretation of the primitives which satisfies the axioms — i. By a (general) system. In either case. for better or for worse. Wigner (1960) asserted that “the unreasonable effectiveness of mathematics” is a mystery whose understanding and solution have yet to come: The miracle of the appropriateness of the language of mathematics for the formulation of the laws of physics is a wonderful gift which we neither understand nor deserve. and the origin of the universe.K). four. where M is a set containing all words in the language and K is the collection of all grammatical rules in a fixed grammar book (Lin. The following example. It has been shown that each formal language can be described as a system (M. T ∪ K ) of mathematics to the situation under consideration.e. However. 1976).T ∪ K ). and each theory can be described as a system ( M. Its postulates come from pure thought or experience. . once formulated. As an example. 1989c). of the nonprimitive mathematical terms) if the ZFC axioms are true. where M and K are defined as before and T is a set of basic principles upon which the theory is developed (Lin. where R is a set of some relations defined on M. we cannot refer to ZFC axioms as propositions which are either true or false.. though perhaps also to our bafflement. Examples can be given to show that the free primitive terms “set” and ∈ permit many different interpretations in everyday life as well as in the investigation of laws of nature and from each interpretation we understand something more about nature. turns them into true statements — will also satisfy all the theorems deduced from them. This description serves to distinguish it from flowers with three. we mean an ordered pair S = (M. Now the application of mathematics is an activity of matching a subsystem of the systems representation (M .352 Chapter 12 seems to be an intellectual activity separate from the development of science and technology. We should be grateful for it and hope that it will remain valid in future research and that it will extend. . The mathematical word “five” provides a certain description of the flower. respectively. The whole structure of mathematics might be said to be true by virtue of mere definitions (namely. 1989c). petals. even though these new understandings may not necessarily be correct. mathematics can be described as a system (M. For example. for they contain free primitive terms. a hibiscus flower has five petals.

among the most common practices of modern scientific activities is the science and technology of data analysis. . Step n: Put the pieces of paper. the contradiction does not exist. one could argue that since there is no way to finish the recursive procedure in the puzzle. existence of subsequences. then remove the piece labeled n. we define the set Mn of pieces of paper left in the vase after the nth step is finished. The reason is that pure mathematics is established upon a set of axioms. for each natural number n. Vase Puzzle Suppose a vase and an infinite number of pieces of paper are available. However. . including the concept of limits.2. and poetry.” and then conclusions are drawn based on the relations of these meanings to the objects involved. the number of pieces of paper left in the vase should be equal to the limit of the function ƒ(n) as n → ∞ . Ma. whereas in applied mathematics each object of interest is always first given some “mathematical meaning.. for n = 1. if the recursive procedure can be finished. It is known in analysis that the more data that are collected. how many pieces of paper are left in the vase? First. The pieces of paper are labeled with natural numbers 1. As described in (Lin et al. labeled from 10 n – 9 through 10 n. .1. a large portion of mathematics would be incorrect. it can be seen that the contradiction has very fruitful implications. because different interpretations can be given to the same object. into the vase. .3. some interpretations can result in contradictory mathematical models. and Cantor’s diagonal method in set theory. Therefore. Second. define a function ƒ(n) = 9n. labeled from 1 through 10. the phenomenon can either not be understand because of the lack of knowledge or be understood at a more global level and more information is needed to be more specific. First. An extraordinary amount of errors can easily conceal the actual state of a phenomenon. Third. which tells how many pieces of paper are left in the vase after step n. the answer is that infinitely many pieces of paper are left in the vase. the vase puzzle does not satisfy this general methodology of recognition. no piece of paper is left in the vase. 1990). Second. into the vase. then remove the piece labeled 1. The following recursive procedure is performed: Step 1: Put the pieces of paper. If we go back to the vase puzzle. After the procedure is completed. The theory possesses a beauty analogous to painting. .2. the set of pieces of paper left in the vase equals the intersection That is.Some Unsolved Problems in General Systems Theory 353 12. such that each piece has exactly one label. This is due to the accumulation of errors. the more misleading conclusion could be. . music. The vase puzzle is another example showing .. However. if information about a phenomenon has not all been used. and a harmony between numbers and figures. It is the assignment of a mathematical meaning to each object that causes problems. Lin.3.3. Information about the ordering really made understanding more controversial rather than more specific. After the recursive procedure is finished. and Port (1990) pointed out theoretically that there must be an impassable chasm between pure and applied mathematics. if this argument is correct. since they all require completion of a similar recursive procedure. Hence. say ZFC.

which are continuously coming out of several assembly lines. The third question is. let us take a more general stand. has greatly influenced all current approaches of systems analysis. Otherwise. Because of its promise. Now instead of looking at the concept of mathematical modeling exclusively. Lin (1988c) analyzed the points on which each application of general systems theory is based and pointed out that these are the most probable places from which unexpected troubles arise. many people have tried to apply general systems theory to solving practical problems (Klir. The second question is. for detailed discussion on this. Lilienfeld. it cannot be assumed to be infinite either. so in applications of general systems theory ∈ can be defined in different ways. Wood-Harper and Fitzgerald (1982) thought that the reason for this lack of success was that the very generality of the theory made it difficult to use and to develop a methodological solution. Another problem we have to face before any mathematical modeling is established is. At the same time. determining if a collection of some objects is a set is always the first problem we encounter before we can draw any useful conclusions based on theoretical results. It is a known fact that human eyes can offer the brain wrong information. then A is a set. the collection cannot be assumed to be finite. How do we know anything? Descartes had pondered this question for a long time. Therefore. 1976. 1970). the more confused we become. The difficulty is that the collection of interest is not fixed and is dependent on time. 1978). because the ignorance of a fact will lead to a completely different prediction. 1972. This is apparently a source of confusion in applications of the theory. for details see (Kline. derived from the work of von Bertalanffy (1968). It is also a fundamental problem. 305). since in applications no one ever doubts the reasonableness of the statement that the collection of living organisms is a set. because in each application of general systems theory. we can see that if A is a collection of objects and there exists a one-to-one correspondence h : A → X for some set X. we are studying the collection of all products of a certain kind. Surely. This is another source of confusion in applications of general systems theory. is the collection of all living organisms in a certain experimental “container” a set? This problem seems to be simple. By using axioms in the ZFC system. What does the formula x ∈ y mean? Here ∈ is a primitive notation. A well-known fact is that the basic concepts of general systems theory were introduced in the language of set theory. the vase puzzle also questions the accuracy of scientific predictions. what does the statement “x is a set” mean? It is well known that some collections of objects are not sets. however.354 Chapter 12 that the more facts we know. p. General systems theory. It is natural to consider the following: How can we establish a one-to-one correspondence between a set and a collection of objects that we are studying? For example. an elementary question is. After comparing several definitions of . What is a set? For example. we would run into the same situation as described in the vase puzzle. in applications. see (Berlinski. This attempt has been notably unsuccessful.

even as Genesis states that “in the beginning.” Since a system is what is distinguished as a system. 1990d). mathematics can be built upon the “empty set. do we have a procedure to select additional models so that they produce the most desirable and most reasonable conclusions on the material system of interest? When faced with practical problems. Therefore.Some Unsolved Problems in General Systems Theory 355 systems. is: How can we bring in the idea of “noise”? For instance. Another interesting and important question.” History gradually shows that almost all natural phenomena can be described and studied with mathematics. based upon the language of formal logic. When can a collection of objects in an application be studied as a set so that the results in set theory and general systems theory can be used to make inferences? Can we introduce a procedure to check when a given collection of real matter is a “set”? Studying this question will lead to the creation of “applied set theory. it is necessary to develop a method to check whether a collection of objects and a membership relation defined on the collection satisfy the ZFC axiom system. according to Klir (1985). And the earth was without form. each study of human relations is subject to the correction of the human noise. Ø. current achievements in modern quantum mechanics strongly support the fact that the laws of nature. related to the concept of systems modeling and systems identification. if any. human noises are different from those observed in any other systems containing no human beings. At the same time. another difficulty we have to face to apply general systems theory is how to write a realistic relation. . it has never been checked whether the collection of objects and the membership relation between the objects satisfy all the ZFC axioms. as pointed out by Masani (1994). God created the heaven and the earth. Only when ∈ makes all ZFC axioms true propositions can general systems theory and mathematics be applied to a situation. could we conclude that the universe we live in rests on the empty set Ø. such as a human relation. Lin (1989d) proposed the following axiom: The existence of any real matter is known by the existence of all particles in some level of the matter and by some relations between the particles in the matter and some particles in an environment of the matter. and void?” In applications of general systems theory and mathematics. are written in terms of probability. predictions based on the newly developed theories are so “accurate” that the following question appears naturally: Does the human way of thinking have the same structure as that of the material world? As shown in (Lin. In some sense. After several systems models have been established. We thus ask: If all laws of nature could be written in the language of mathematics. as one or several relations in set theory. new abstract mathematical theories are developed to investigate them. Also. a collection of cows cannot be considered a set. As mentioned by Kunen (1980).

of object systems S i ∈ M i–1. if there exist systems S i = (M i . a system S n = (M n . 1987). R n ) is an n th-level object system of a system S 0 = (M0 . One inference. based on the structure of layers. The last conclusion is subject to three possibilities: (i) The world is made up of fundamental particles. where a sequence of systems. such that S i ∈ M i –1 .356 Chapter 12 12. An object x is an n th-level object of a system S 0 = (M0 . such that S i ∈ M i –1 .j ∈ n. (ii) There exists a particle X such that X can be divided into smaller particles and the collection of all smaller particles in X is not a set. R ) there exists a unique set M(S ) consisting of all fundamental objects in S. the concept of systems characterizes the organizational structure of the world. with n an ordinal number. i = 1. Under the assumption that the ZFC axiom system is consistent. the previous result implies that if M is a set. n. Therefore. Laws of Conservation and the Multilevel Structure of Nature As shown in Section 12. is a chain of object systems of S if S 0 is an i 0 th-level object system of S. including the structure of layers. for i = 1. … . A fixed atom A can be thought of as a system S consisting of the set of all electrons and the nucleus in A and a set of some relations between the electrons and the nucleus.2. it can be shown that for any system S = (M. then it is called a fundamental object of S 0 . n. n . for i = 1. i ∈ n. where n is a positive integer. R 0 ). it can be considered as an ordered pair (M. and for any i.Ri ). that can be made from the concept of systems is that the world is finitely divisible. (iii) The ZFC axiom system is not consistent. If the object x is no longer a system.j) such that the system S j is an n (i. then X consists of fundamental particles. if i < j. 2. … . and x ∈ M n . Based on this discussion. if a particle X can be divided into smaller particles. this result says that each system is built on fundamental objects. … . R 0 ) if there exist systems S i = (Mi . if M is a set. Intuitively speaking. The set M(S) is constructed as follows: For each object x ∈ M (S). such that S 0 ∈ M .j)th-level object system of Si . Generally. and R is a set of some relations between the particles in M. n – 1. the following has been shown: For any system S each chain of object systems of S must be finite. 2. The collection of all electrons and the nucleus in A is a set because of the finiteness of the collection. where a fundamental particle is a particle which cannot be divided into smaller particles? We know that the world consists of atoms.… .1. there then is a natural number n(i. for i = 1. and x ∈ M n . Hence. . Hence. there exists a chain for some natural number n = n(x).4. where M is the totality of the smaller particles in X. X can be described as a system. 2. R). where n is a natural number (Lin and Ma. The previous result then says that each system must be finitely divisible. R i ). Does this result imply that the world is made up of fundamental particles.

we can be convinced that the ZFC axiom system is consistent. let S = (M. Upon this principle. people gradually began to believe that there did not exist any fundamental particles in the world at all. a system is what is distinguished as a system. nothing is created! An equal quantity of matter exists both before and after the experiment … and nothing takes place beyond changes and modifications in the combination of the elements. and consequently.e. smaller structures were found (i. a system describing the matter of interest can always be defined. Because all classical mathematics. Hence. (Perlman. due to Lavoisier given in 1789.Some Unsolved Problems in General Systems Theory 357 Man has spent hundreds of years on the first possibility. so the total amount of matter remains the same regardless of changes. then each fundamental particle must be so small that it has no size. such that M stands for the set of all substances used in the reaction and R is the set of all relations between substances in M. Based on this understanding. the whole art of performing chemical experiments depends. At this time. the following axiom.. (2) There is a particle X such that X can be divided into smaller particles and the collection of all smaller particles in X is not a set. We can do the following systems modeling: For each chosen matter. which has been the “unreasonably” effective part of mathematics when applied to analyze practical problems. On the other hand. the conservation principle of ancient atomists. (Notice that there is no way to prove this yet!): Now the reasons are: (1) There are some mistakes in the application of systems theory. what the meaning of particles is becomes an important problem. if the foregoing discussion is correct and if the world is made up of fundamental particles. p. Thus. scientific achievement disproved that. For example. or environment. has been proven theoretically: In all the operations of art and nature. It can be seen that S represents the chemical reaction of interest. But. atoms). it can still be cut into smaller pieces. 1980). When molecules were first found. Can we conclude that the world is made up of fundamental particles? This depends on whether the second or third possibilities and the following reason are true: There are some covert mistakes in the application of the result of systems theory. they were thought to be the fundamental bricks of the natural world. Here it is shown that each system is made up of fundamental objects. Otherwise. 1970. As introduced by Klir (1985). can be recast as follows: . Based on this understanding and the result of systems theory that each chain of object systems of a given system must be finite. such as a chemical reaction process. because each fundamental particle cannot have a size. once again. all modern science and technology can be developed on ZFC (Kunen. Before long. 414). real situation problem. which states that the total count of atoms in the universe is constant. under the assumption that the world consists of fundamental particles. R ) be a systems representation of a chemical reaction process. people really believed that the smallest bricks of the world were found. Changes of the system represents changes of the objects in M and changes of relations in R.

All matter consists of fundamental particles. As mentioned in Section 12. R ) of sets such that R is a set of some relations defined on the set M (Lin. the total amount of matter remains the same regardless of changes and modifications. as already discussed and so does energy. respectively. and energy into matter. 1987).2. this general theory will establish a theoretical foundation for all approaches of systems analysis developed in various disciplines. Let S be a (general) system if S equals an ordered pair (M . 12. Only with the indispensable “=” in the equations can the theory. Let us discuss some impacts of what has been done on the Law of Conservation of Matter–Energy. matter can be transferred into energy. therefore. it confirms that each fundamental particle has no volume or size. The number n is called the length of r. Historically. become an “exact science” with the capacity of prediction.1. such as light.358 Chapter 12 12. a component of the system. Now. there exists an ordinal number n = n(r) such that r ⊆ M n . n . where S = (M. submitted) The total count of fundamental particles in the universe is constant. a more appropriate definition for general systems should be an ordered pair of a set of attributes of some objects and a set of relations among the attributes of the objects. developed on the Law of Conservation of Fundamental Particles. Law of Conservation of Fundamental Particles (Lin. Set-Theoretic General Systems Theory This section focuses on general systems theory developed on ZFC set theory. the Law of Conservation of Matter–Energy becomes clear and obvious. The elements in M are called the objects of S and M and R are the object set and the relation set of S. It is = 0. One of the many important open problems is to develop conservation equations in terms of fundamental particles and to write them in modern symbolic form. Since the existence of an object is recognized through its attributes or characteristics. For a relation r ∈ R. The law states that the total amount of matter and energy is always the same. then the definition of general systems would be reasonable.4. However. if one just simply uses the attributes here as the objects. That is.5. an object in M is recognized as a building block. as a consequence of the Law of Conservation of Fundamental Particles. R ) is a system. it can be real matter or a conceptual entity. Since energy can be in different forms. assumed that the length of the empty relation is zero. As pointed out by Mesarovic and Takahara (1989). that is. this law was developed based on the new development of science and technology: energy and matter are two different forms of the same thing.

From these examples it is natural to consider the problem: Find properties of networks. 1998. Banach spaces. in addition. mappings from systems into systems. Hilbert spaces. since only beauty can make the theory assimilated into our daily thoughts process and brought again and again before the mind with ever-renewed encouragement. If in. algebras. hierarchies of systems.Some Unsolved Problems in General Systems Theory 359 One goal of systems theory. metric spaces. where for any r ∈ R. Is it possible that a structural representation for general systems can be given similar to the one given earlier? The characteristic of this representation is that only ordinal numbers and power set operations are involved. For centralized systems. Beauty ensures that the theory will be handed down from generation to generation. R C ) satisfying M C ≠ such that for any distinct elements x.2. networks. and systems of axioms and formal languages are all systems.y ∈ M. for κ an infinite cardinality and θ > κ a regular cardinality such that. Lin and Vierthaler. Along this line are several open problems. n(r) = 2. fuzzy structures. say x = (M x . 1988a. 1995). semiordered spaces. topological spaces. manifolds. is to possess the same beauty as mathematics possesses. . Riesz spaces.R x ) and y = (M y . such that in general systems theory they have the same appearance. there exists a partial system S' = (M'. automatic machines. it is worth asking: the question below: Under what conditions is the partial system S' a strongly centralized system? It can be seen that the concept of systems is a higher-level abstraction of mathematical structures. S is called strongly centralized. etc. M C = M x ∩ My and where and The system C is called a center of S . abstract automatic machines. it has been shown that under the assumption that ZFC axiom system is consistent.3. fuzzy sets and fuzzy relations. Since this result has found several interesting applications in sociology and epidemiology (Lin. then for each system S = (M. normed rings. for any α < θ . S is a subset of the set p ²(M) ∪p(p(M) ∪ p 4 (M)).R ). constructions of systems. This result is a restatement of the general ∆ -lemma in set theory (Kunen. the general structure of a system. based on this definition.R) where |M| ≥ θ and |M m | < κ . vector spaces. where p¹(X) = p (X) is the power set of X and p i +1 (X ) = p(p i (X)) for each i = 1.. To achieve this goal. algebraic systems. n-tuple relations.…. and where there exists an object contained in at least θ objects in M .. topological spaces.R y ). It can be shown that for any system S = (M . A system S = (M.R) is called centralized system if each object in S is a system and there exists a system C = (M C . etc. R m ). 1980). normed spaces. Banach algebras. semigroups. algebras. Frechet spaces. for each object m ∈ M a system with m = ( Mm . Lin and Forrest. systems of axioms.R') of S s u c h that S' forms a centralized system and |M'| ≥ θ . need to be studied systematically. For example. vector spaces.

. such that in systems analysis they have the same appearance. are rings in algebra. λ : S' → S ) so that for each i ∈ I It can be shown that for each set {S i : i ∈ I} of systems. … . denoted by h : S 1 → S 2 .2.. which are unique up to a similarity and denoted by and respectively. is a relation on the set M 2 with length n(r). a product and a Cartesian product of S i always exist. etc. … ) where ON stands for the class of all ordinal numbers.2. if there is a family of morphisms (resp. A system S = (M . i = 1.2. such as groups.2.S i ) : i ∈ I } (resp. h will also be used to indicate the class mapping and h is a mapping from S 1 into S 2 . R 2 ) is S-continuous (Ma and Lin.. and that if the systems S i . In this case. A mapping h : S 1 = (M 1 . When the systems S i are algebras. h is S-continuous iff h is a continuous mapping from the topological space S 1 into the topological space S 2 . the system is the box product of the spaces S i ... Let Hom(S 1 . and a class mapping can be defined with the following properties: and for each x = (x 0 .x1 . i = 1. R ) is a product (resp. S-continuous mappings) { Φ i ∈ Hom(S .. two classes i = 1. {q i : S' → S i : i ∈ I} of S continuous mappings) there is a unique morphism (resp.S i ) : i ∈ I ) (resp. Suppose {S i : i ∈ I } is a set of systems. rings. .360 Chapter 12 Let S i = (M i . For each relation r ∈ R .S) (resp. {p i : S → S i : i ∈ I }) such that for each system S' and a family {Ψ i ∈ Hom(S'. i = 1. S 2 ) = {ƒ : S 1 → S 2 : ƒ is a morphism}. A mapping ƒ : S 1 → S 2 is a morphism from a system S 1 into a system S 2 if ƒ (R1 ) ⊆ R . are topological spaces. such as groups. By transfinite induction. R 1 ) → S 2 = (M2 . Cartesian product) of S i . This example shows the need to find the topological properties of topological spaces and algebraic properties of algebras. be two systems and h : M 1 → M 2 a mapping. The reason why the system is called the Cartesian product of S i is that when S i are topological spaces. Without causing confusion. S-continuous mapping) λ ∈ Hom( S'. It then can be shown that if the systems S i . this system is the Cartesian product of the spaces S i . 1990e) if for any relation r ∈ R 2 .x α .R i ). the S-continuity of h implies that h is a homomorphism from the ring S 1 into the ring S 2 ..

An α -type hierarchy S of systems [over the partially ordered set (T. is the following: Suppose S is an α-type hierarchy of systems over a partially ordered set (T. s ≥ t } of linkage mappings for S such that ts = l t s for all s.Some Unsolved Problems in General Systems Theory 361 rings. the hierarchy is linked. there exists a family {lt s : t. An interesting question is. If. Its main characteristic is that the transition history of the health and disease patterns of industrialized nations have been repeated. It has been shown (Lin. For a fixed t0 ∈ T. for any r. which describes the transition of health and disease patterns. ≤ ) be a partially ordered set with order type α. Conversely. in addition. t ∈ T * with s ≥ t? As shown by Omran (197l). t are arbitrary elements in T satisfying s ≥ r ≥ t and id S t = id Mt is the identity mapping on the set Mt .t y p e hierarchy S of systems. . Mt ≠ Ø where S t = (Mt . s. how many families of linkage mappings are there for a given α -type hierarchy of systems? Or. To be specific. t ∈ T with r ≥ t.R t ). there is strong evidence for the existence of the epidemiological transition. ≤ ) is called a system over a group. let l t r : S r → S t be a mapping from the system S r into the system St . 1989a) that for each α -type hierarchy S of systems satisfying that for each t ∈ T. To model this important concept. where T * ⊂ T. The family { l ts : t. the right transformation of S. That is. denoted S + t 0 . which is similar to the question of expanding a given function. are there for a given α -type hierarchy of systems? Another question. This fact gives rise to the following problem: Find topological structures and algebraic structures such that in systems analysis they appear to be the same. ≤ ) is an ordered Abelian group. Under what conditions does there exist a family { ts : t. s ∈ T. namely that infectious diseases are gradually replaced by noncommunicable diseases as the economic and industrial development of a region takes place. s ≥ t } is a family of linkage mappings of the hierarchy of systems {S t : t ∈ T * }. A hierarchy S of systems over (T.s ∈ T. etc. Lin and Lin (1995) employed the concepts of systems over groups and of periodic systems over groups. the system is termed a linked system over a group. the system ∏ i ∈I Si is the direct sum of the algebras S i . such that and where r.s ∈ T. is defined to be S – t0 (t ) = S (t – t 0 ) for each t ∈ T. S(t) = S t = ( Mt . +.s ∈ T * .s ≥ t } is termed a family of linkage mappings of the α-type hierarchy S. For an α. ≤ )] is a function defined on T such that for each t ∈ T. or are yet to be repeated. 1989a). suppose (T. denoted S – t0 . and the left transformation of S. how many families of linkage mappings with certain properties. the box product in topology and the direct sum in algebra have the same structure. such as S-continuous. Rt ) is a system (Lin.. ≤ ) and {lt s : t. +. it is natural to ask: When can a system be studied as a topological space or a specific algebra? Let (T.s ≥ t} of linkage mappings of S. and each lt s is termed a linkage mapping from the system S s into the system S t . in developing regions or nations.

Let be a field. (ii) s ∈ S and s' ∈ S imply that s + s' ∈ S .y 1 ) + (x 2 . ƒ = {Sƒ : Y → X : S ƒ is a linear functional system}. ( x 2 . where + and · are addition and scalar multiplication in X × Y. Let S i = p i ( S ) for i ∈ I. The element p is called a period of S. Assume that the input space X and the output space Y can be written and For each i ∈ I. Let = {S ⊂ X × Y : S is a linear system}. αy 1 ). How to study these systems is an important question.y 1 ). consult Takahara and Asahi (1985). defined by: for (x 1 . for any x ∈ X and y ∈ Y. A system S over a group T is periodic if there exists p ∈ T such that p ≠ 0 and S + p = S. X and Y linear spaces over the field .y 2 ) ∈ X × Y and any α ∈ and α · ( x1 . y2 ) = (x 1 + x 2 . y 1 + y 2 ) any (x1 . and many nonliving systems are feedback systems. 1985). and Saito (1986. Then the feedback system of S by Sƒ is defined as the input–output system S' such that and The systems S and Sƒ are called an original system and a feedback component system.362 Chapter 12 is defined by S +t 0 (t ) = S (t + t 0 ) for each t ∈ T. 1995). 1975). and = { S' ⊂ X × Y : S' is a subset}. For recent research. the projection p i = ( p ix . An input–output system S ⊂ X × Y is a functional system if S is a function from the input set X into the output set Y. Let us take a brief look at the concepts of feedback transformation and feedback decoupling and several related open questions. see (Saito and Mesarovic. Then a feedback transformation is defined by the last equivalence relation for each The transformation F is called the feedback transformation over the linear spaces X and Y. and S an (input– output) system satisfying the following conditions: (i) Ø ≠ S ⊆ X × Y . For details. 1987). respectively. The concepts of feedback transformation and feedback decoupling have brought some light because they tell when the system under consideration can be divided into smaller systems. respectively. (iii) s ∈ S and α ∈ imply α · s ∈ S. Feedback has been a hot topic in recent years since all living. As pointed out in (Lin and Forrest. Let S ⊂ X × Y and S ƒ : Y → X be a linear system and a linear functional system. Then S is called an (input–output) linear system (Mesarovic and Takahara. respectively.y 1 ) = ( αx1 . there is a need to study multiperiodic systems where a system S over a group T is multiperiodic if there exists a subset D ⊆ T such that S + P = S for each p ∈ D.p iy ) : X × Y → Xi × Y i is defined by p ix : X → X i and p iy : Y → Y i satisfying p ix (x ) = x i and p iy ( y ) = y i . Then the system S is decomposed into a family of factor systems If then A linear is identified with system S ∈ is decoupled by feedback (Saito and Mesarovic. 1985) if there is S ƒ ∈ ƒ such that .

n. respectively. A. n × m. The family of factor systems consists of n one-dimensional linear systems of the following form: = Az + Bix i yi = C i z + Di x i z (0) = 0 for i = 1. If some properties of the original system are not kept by feedback systems.Some Unsolved Problems in General Systems Theory 363 Under the assumption that ZFC is consistent. For example. based on the solutions of the one-dimensional factor systems? 12. and D are constant matrices of size m × m. 1994) that if S ∈ is decomposed into a family of factor systems then S is decoupled by feedback iff and where and Based on this result. and n × n. Many questions are left open. and the input space Xi and the output space Yi are given by Xi = Yi = {r : [0. do the input space X and the output space Y have to be the same dimension in order to apply the theory? Find examples of systems such that some of their important properties are not kept by feedback systems. that to show the need for the concept of (general) systems it is absolutely . B. he pointed out a fact.2. and the input space X and the output space Y are X = Y = { r : [0. where B = [B1 B2 … B n]. it has been shown (Lin. … . 1994) gives a concrete procedure to decouple the linear system = Az + Bx y = Cz + Dx z(0) = 0 where z is an m × 1 variable vector. Is it possible to construct the solution of the n-dimensional linear system described earlier. C. C = [C1 C2 … C n ] T.6. which could also be considered a challenge to systems science. ∞ ) → : r is piecewise continuous} where is the set of all real numbers. 1992). ∞ ) → : r is piecewise continuous}. (Lin. such that D is nonsingular. can the properties be studied by using the concept of pullback? For the concept of pullback. m × n. see (Negoita. called the state of the system. D i is a nonzero constant. Analytic Foundation for an Applicable General Systems Theory During a recent and fruitful conversation with Professor Pesi Masani.

introduced the concepts of infinitesimals and infinities formally and legally. has a computational advantage. since each number. some classical mechanics problems. where infinities are considered ideal objects. y ∈ GNS. Throughout history. infinitesimal. such as mathematical analysis. algebra. such that there exists a number called the level index of x. into where is the set of A generalized number x is a function from all integers and the set of all real numbers. Historically speaking. first introduced by Shu-tang Wang in the late 1970s and early 1980s. the non-Archimedean number field. and 1 (t) ( t ) = 1. is an excellent starting point. then the generalized number z is called the quotient of x divided by y. has an exact expression in terms of the real numbers in the classical sense. write x as a generalized ( c x)(t ) = c x(t ) for t ∈ power series where 1 (t) ∈ GNS is defined by 1(t ) (s ) = 0. However. if there exists a z ∈ GNS such that y × z = x. a n d k = k (x) ∈ x ( k (x)) ≠ 0. nonstandard analysis. and define and x(k 0 ) < y ( k0 ). the shortfall of Robinson’s nonstandard analysis is that infinitesimals and infinities can only be shown to exist and have no concrete expressions of any tangible form. (x ± y )( t) = x( t ) ± y ( t ). and x ∈ GNS. y ∈ GNS. whether ordinary real number. x < y iff there is a k 0 ∈ I such that x (t ) = y (t ). can be greatly simplified and become more intuitive. for t < k. For the most part. normed spaces. y ∈ GNS. for t ∈ for any x. y ∈ GNS.364 Chapter 12 necessary to obtain some results which have not or could not be achieved without using the concept of systems. (3) Scalar multiplication: let c ∈ (4) Multiplication: for x ∈ GNS. if s ≠ t.. The product x × y or xy is defined by (5) Division: For x. “laws” of nature have been frequently expressed and treated with the rigor and beauty of mathematical symbols. it will be shown that the nonArchimedean number field. established in the early 1960s by Robinson (1977). denoted by x ÷ y or x /y. many classical theories. or infinity. based on mathematical logic. Let x. To meet this challenge and to make up a deficiency of the current theory of general systems analysis. Based on the argument of the existence of infinitesimals and infinities given in Robinson’s theory. This theory broke through the limitation of classical concept of real numbers. satisfying x (t) = 0. called the generalized number system (GNS) by Wang (1985). if t < k 0 . etc. Let GNS stand for the field consisting of all generalized numbers and the zero function 0 : with the following operations: (1) Addition and (2) Ordering: subtraction: for any x. . where no quantitative study based on general systems analysis can be carried out. With nonstandard analysis.

So Eq. the rest mass m 0 should be m 0 = 0. theoretical physicists have often been faced with difficulties in mathematical calculation.1) that m = ∞ .Some Unsolved Problems in General Systems Theory 365 this kind of application of mathematics has been very successful. This is because in classical mathematics on the real number line. Now. in GNS the field {x ∈ GNS : x (t) = 0 if t ≠ 0}. we look at the calculation of the rest mass of a photon. Furthermore. According to Wang (1991). since c is assumed by Einstein to be the maximum speed in the universe. in the past 40 years. However. are treated with the same set of rules. contradicting observations. whether infinities or infinitesimals. one would obtain from Eq.1) can be written . that is.1) where c stands for the speed of light and v is the particle speed. m = 0 ÷ 0 is undetermined. the inertial mass m of a moving particle relates to its rest mass m 0 as follows: (12.1) is a finite real number. (12. Suppose the rest mass m 0 of a photon is and that the inertial mass m in Eq. However. (12. all numbers in GNS. injustice involved in the calculation still remains. compared with those of ordinary of all real numbers is isomorphic to (0) = real numbers. Since the photon’s speed is v = c. Otherwise. and are considered in our macroscopic world as ideal objects and are treated with a different set of rules. To give the reader a taste of how GNS can be applied to the study of physics. even it is assumed that m 0 = 0. m ∈ has a representation (0) Now the velocity of a photon is where v 2 < 0. while applicable mathematics still remains single level where the only infinities are denoted by – ∞ or + ∞. which represents the quantities measurable within the macroscopic world. (12. This difficulty can be overcome by using generalized numbers. According to Einstein. such difficulties arise mainly because the modern world of physics is multilevel in its character.

definite and indefinite integrals of given functions from GNS into GNS. governing the relation between a set of properties of real numbers and a set of properties of generalized numbers. 1998). more puzzles in the current quantum mechanics can be explained more rationally. m0 . (6) Develop the results of generalized numbers. it is necessary to (1) Study the ordering relation. 1 (0) . (4) Redevelop quantum mechanics based on GNS so that. such as defining products of distributions. linking GNS differential calculus to GNS integral calculus. we see that our understanding of the GNS is still superficial. maximum and minimum theorem. hopefully. and c into this equation and comparing the leading nonzero terms. corresponding to theorems in mathematical analysis. (5) Establish a law. That gives the relation between v 2 1 and m 1 . (10) Use GNS representations of Schwartz’s distributions to explore new explanations of equations. For technical details. The future success of the theory of GNS may well depend on various successful applications. appearing in the theory of distributions. etc. such as the intermediate value theorem. and taking roots. division. By substituting the GNS values of m. 1977). (9) Develop an integral calculus of GNS based on the same mechanism of Lebesgue integrals. (3) Develop workable and concrete formulas for the operations of multiplication. real. (7) Find detailed formulas to compute various derivatives. Comparing the generalized number system with those contained in mathematical analysis. (2) Explore applications of the theory of GNS in other branches of human knowledge. similar to the transfer law in nonstandard analysis (Davis. one obtains (–2c)m 2 v 2 = m 2 . (8) Generalize the Fundamental Theorem of Calculus to the case of GNS. (11) Employ GNS representations of Schwartz’s distributions to solve open problems in the theory of distributions. More work needs to be carried out to gain deeper understanding about nature and the world in which we live.366 Chapter 12 where c is understood in the same way as we did with m ∈ (0) . see (Lin and Wang. For instance. and complex analysis. v. Can the theory of GNS be used to explain why in general the products of distributions might not exist? . which do not make sense in classical mathematical analysis. It is obvious that the GNS field is not complete in the sense of Dedekind cuts.

A Few Final Words To face challenges. Only with these two tasks being accomplished. among the keys for its future success are publicity and practicalities. self-dedicating. technology.Some Unsolved Problems in General Systems Theory 367 12. more large-scale combination. more accuracy. Even though the general systems enterprise has not been. general systems theory has been set off from the grayer background. . it will definitely become a powerful force sweeping through the world of learning. Research and application of general systems theory are a natural extension of his intelligent and comprehensive pursuits for the past several thousand years. developing and self-completing. Any one with a prepared mind can achieve a great deal in the development and application of such a theory. The general systems enterprise has been. self-rebeling and self-surpassing. In the scientific garden with various theories which will stand the test of time and history. and more unification (blending philosophy. etc. self-realizing. will the scientific community be truly moved so that research of general systems theory and applications moves to a higher level. and will not be. physical science. more pragmatism. and will continue to be. They are also creatures of interactions between them and the physical world and reduced reflections and enlarged images of the history of the world of learning. Man is both a reflection reduced with the size of history and an enlarged image of history..7. pursuing eternity. He is the creature of interactions between himself and the physical world. self-coquetting. Based on the history of general systems theory and its current development. the study of general systems theory must further the research of the concept of (general) systems and related topics so that the theory possesses more systems. mathematics. into one body of knowledge).

New York. Bellman. Bayliss. — (1934).. Woodge) Oxford University Press. Borntäger. Benjamin/Cummings. and L.. 369 . Reidel. and Model Building. L. 145–150. S. Moscow. H. Inform. Einführung in Spengler’s Werk. The history and status of general systems theory. 6 . A. Dynamic Programming. Philosophical and Methodological Problems. San Francisco. Yudin (1977). 1–23. 4 . Oxford. decisions and organizations. and R. Pilsworth (1982). T. social science. Princeton University Press. Wiley. Treatise on Basic Philosophy. W. Systems Theory. Berlinski. von Bertalanffy.. (1924). 125–136.. General Systems. A.. CA. V. Cambridge. New York. MIT Press. Hunter. B 141–B 164. MA. George Braziller.. and J. Vol. J. A Problem Solving Approach to Mathematics for Elementary School Teachers. 4 : A World of Systems. Soc.. Living Control Systems. 12. and E. Sadovsky. R. Klir (ed. J.. G... W. In: G. Trends in General Systems Theory. Redwood City. Holland. Dynamic programming for fuzzy systems with fuzzy environment. Bunge. S. Games. V. — (1967). Data Analysis. New York (1962). Baldwin. Anal. Harper Torch Books. Management Sci. J. R. 21–41. E. General systems theory: Application to psychology. General Systems Theory. Dynamic programming in multiplicative lattices. 4th ed. Billstein. Libeskind.. Berlin (1928). 624–637. D. Anal. Progress Publishers. On Systems Analysis. Decision making in fuzzy environment. (1979). I. and B. (1976). — (1972). (1959). 85. Dordrecht. Blauberg. Lott (1990). and J. Princeton. J. Box. Literaturblatt Kolnische Zeitung. R. Brown. Modern Theories of Development (Transl. N. Math. G. R. (1966). Statistics for Experimenters: An Introduction to Design. L. New York. Appl. Strauch (1965). May. 17. Hunter (1978).References Ackoff. — (1968). NJ. B. German Original: Kritische Theories der Formbildung. pp. L. Zadeh (1970). Sci. Appl. E. Bellman. Math. (1957).).

North-Holland. Nicolas Bourbaki and contemporary mathematics. Dirac. (1947). Frechet. DC. Paris. pp. Contraction Mappings in the Theory Underlying Dynamic Programming. Bourbaki. V. .). Sci. (1985). Res.. Klir (ed. London. Amsterdam. — (1946). 223. The Art and Theory of Dynamic Programming. 337–339. Fraenkel. Intell. London. Acad. Analyzing business information needs. R. J. Ellis.. (1977). Oxford University Press. C. Warszawa. de P. Y. R. In: R. Progress in Cybernetics and System Research. Wuhan. Denardo. A. J. (1958). British Computer Society. Academic Press.-L.370 References Cartan. Department of Mathematics. No. Law (1977). J. Furukawa ( 1980). Amsterdam. Oper. E. and A. Einstein. 5 . P. Hemisphere. Friede. J. 1121–1129. 372–373. N. Comacchio. (1945). (1975). Department of Business Administration of Wuhan University (1983)... Academic Press. C. ( 1967). North-Holland. F. 244. Trappl and F. Espaces a escart generalise regulier. Wiley-Interscience. The development of systems thinking by systems practice—A methodology from action research programme. (1982). A Panorama of Pure Mathematics: As Seen by N. H. 9. (1978). and A. (1975). 121–131. C.. Ji-Lin University (1978). Fundamentals of Modern Administration. The World as I See It. New York. New York. M. De l’ecart numerique alecart abstrait. BCS Conference Papers. Dieudonne. Viking Press. A. Trends in General Systems Theory. Engelking. Casti. Deng. Foundations of Set Theory. R. Paris. 221. Maddison (ed. In: R. Davis. Covici. 271–279. J. (1980). Wuhan. The Principles of Quantum Mechanics. II. SIAM Rev. (1922). Doss. Applied Non-Standard Analysis. Acad. 1087–1088.. 175–180. In: G. Sci. La notion d’uniformite et les ecart abstrait. Paris. 5 . New York. Vol.). A. and I. (1947). Non-Linear Systems Theory. S.). Checkland.. Mathematical Analysis (in Chinese). Sur les espaces ou la topologie peut etre definis a l’ aid d’unecart abstrait symmetrique et regulier. 278–283. Orlando. Academic Press. Sci. Acad. Cohen. Levy (1973). (1974). New York. Wiley. H. Portugese Math. Washington.. People’s Education Press. General Topology. Math. R. Dreyfus. Grey Control Systems. Stewart (1994). Data Analysis for Information Systems Design. (1972). Polish Scientific Publishers. V. Set Theory: An Introduction to Cardinals. New York. (1985). Math. Knowledge Press. B. Drake. R. R. Bar-Hillel. Colmez. The Collapse of Chaos. A. N. Topological concepts in the mathematical theory of general systems. Hua-zhong Institute of Technology Press. New York. C. Characterization of optimal policies in vector-valued Markovian decision process. P. J. M. Hawika (eds. 2 .

SIAM J. . 15.. E. Math. Hess. and T. A. NJ. 61.. Husek. No. Control Optim. — (1985). Springer-Verlag. Mat. Introduction to Meta-Mathematics. Gleick. Topological characterizations of linearly uniformizable spaces. B. Nicolas Bourbaki. Plenum Press. Content Analysis: An Introduction to Its Methodology. Kline. NJ M. Structured Analysis: Tools and Techniques. S. Kuhn. (1952). J. Naive Set Theory. 178–191. 88–99. MA. D. May. 8. Van Nostrand. I. Sage Publications. 220–230. Viking Press. Vector-Valued Dynamic Programming. Combinatorial Algorithms. Halmos. 129–145. (1978). 87. L. Germany). J. (1962). University of Chicago Press. Set Theory. Princeton.. (1982). Adaptation in Natural and Artificial Systems. Universal Algebra. and H. Englewood Cliffs. C. J. — (1978). Academic Press. Krippendorff. 237–245. I.. Mag. Princeton. Oxford University Press. AddisonWesley. Budapest. Math. M. Reichel (1983). North-Holland. CA. Genetic Algorithms in Search. Kleene. (1975). Topology Appl. 1. Sec. General Systems. Amsterdam. T. Ann. Hall. Hodel. New York. V. Hahn. NJ. Modelle Enzymatischer Prozesse. (1972). 173–188. C. Klir.. Deutsche Z. Optimization and Machine Learning. (1987). T. Henig (1983). Nova Acta Leopoldina (Halle. Elements of Mathematical Logic. J. The Axiom of choice. (1967). Amsterdam. E. P. University of Michigan Press. Beverly Hills. Prentice-Hall. 219– 229. Sarsons (1979). North-Holland. D. (1969). Amsterdam. American Elsevier. (1975). New York. Holland. 18–28. Phil. Mathematical Thought from Ancient to Modern Times. J. Juhasz (1965). G. Reading. New York. Jech. and R. Scientific American. New York. Addison-Wesley. M. Goldberg. (1970). C. Fagen (1956). C. Fund. New York. Univ. (1989). R. R. T.. An Approach to General Systems Theory. Extensions of metrization theorems to higher cardinality. Graniner. New York. G. Krivine (1967). E. Hu.. K. Aktuelle entwicklungstendenzen der soziologischen theorie.References 371 Gane. — (1960). Chicago. (1980). Architecture of Systems Problem Solving. Oxford. (1957). The Structure of Scientific Revolutions. Gratzer. (1988). North-Holland. Van Nostrand.. Kreisel and J.. (1973). Sci. Chaos: Making a New Science. 15. Definitions of systems. 3. Untersuchungen über ω µ -metrisierbare raume.

Sur les classes (E) et (D). Anwendungen unendlichkleiner zahlen II. Reine Angew. Publ. II. Oxford. North-Holland. Computer Model. 11. Acad.. North-Holland. 51–66. D. J. — (1988). J. Kuratowski. Math. — (1987). and A. (1980). Pergamon. 11. New York. R. Paris. 473–488. Reine Angew. Laugwitz. Systems Sci. Methodol. 4. — (1962). Int. R. Set Theory: With an Introduction to Descriptive Set Theory. Cybernet. 79. Sur lecart abstrait. Systems. — (1989). Acad. R. Astr:. Le probleme de Souslin et les espaces abstraits. Epistem. 1765–1815. Sci. Sys. Int. Eine nichtarchimidische erweiterung anordeneter korper. 19. Wiley. 1st Veneto Sc. (1961). Introducing Systems Analysis and Design.... 1979). Systems Sci. 124–132. 1875–1889. Lilienfeld. Kuratowski. — (1989). — (1989). 53–60. C. 20. K. — (1936). (1978). K. Sci. Math. Vols. Defining social problems: Public and professional perspectives. 95–104. Math. 203. Set Theory: An Introduction to Independence Proofs. 202. T. 22–34. 24. Amsterdam. Ser. J. Centralized Systems and Their Existence. Acad. K. The concept of fuzzy systems. Y. (1980). J. Multi-level systems. Anwendungen unendlichkleiner zahlen I. 185–187. Lauer (1976). — (1936). Sci. 738– 742. C. An application of systems analysis in sociology. 37. 267–278. Tableu ramifies densembles: Espaces pseudo-distancies. 7 a . 1563–1565. 208. 19. Math. Amsterdam. Int. Modelling. (1892). — (1988). On periodic linked time systems. 1049—1052.. 9. Kybernetes: Int. A Half Century of Polish Mathematics: Remembrances and Reflections. Levi-Civita. A model of general systems. textbf19. Social Problems.372 References Kunen. 6. Mostowski (1976). Letted Art.. Cybernetics and Systems.. Some properties of linear systems. R. . Can the world be studied in the viewpoint of systems? Math. Synthese: Int. Paris. Phil. Order structures of families of general systems. D.. — (1989).. Sugli infinitied infinitesimi attuali quali elementianalitici. No... C. Atti. 5. 105–134. Univ. Fiz. — (1968). — (1936). Manchester. (1934). 225–236. 1299–1310.. Math. 20. L’Hypothese de Ramification. Barry Lee (1978. J. A multi-relation approach of general systems and tests of applications. Sci. — (1990). 45–51. — (1956). Beograd. Kurepa. 198. Nachr. I. J. Cybernetics and Systems. 927–937.. The Rise of Systems Theory. 122–30. J. Lin. (submitted). 20. Ser.. National Computer Center. 207. Sci. D. — (1988). Glasnik Mat.

Systems. 24. P. Cybernetics and Systems. — (1990). J. J. No. Model. Li (1997). General Systems.References 373 — (1990).. No. 6. Systems Cybernet.. A. 33. Lin. 143–154. and S. 52–60. Internal Cohen extensions. Y. Lin and Y. — (1990). Y. Appl. 5–17. Lin. Comput. Systems Sci.. Q. Ma. Notices of the American Mathematical Society. D.-H. Centralizability and its existence.-P. — (1994). 647–664.. 125–134. Abstracts of 8th International Congress of Cybernetics and Systems. Model. and D. Sci. 287–303. Limit properties of hierarchies of general systems. Systems.. and 591–605. Model.-H. Mickens (ed. Englewood Cliffs. Math. Model. Z. Solovay (1970). Mathematics and Science. Cybernet. 1. Port (1998). 135–141. Math. Structured Analysis: Systems Specifications. Comput. New York. Math.-T. — (1988). Model. 5. and R.. Lin. 4. Math. 14. and Y. — (1990). Developing a theoretical foundation for the laws of conservation... .. 58–63. Lin (1987).. 5–132. R. 52–57. 2.. Y. J General Systems. Some properties of linked time systems. 463–479.. Mathematical sciences: a unifying and dynamical resource.. Comput. Simu. No. Systems. 493– 498. Kybernetes: Int. Wang (1998). World Scientific Press. Ann. — (1989).. Kybernetes: Int. Systems analysis of the morphology of polymers. Vierthaler (1998). Int. Y. J. Logic. Lett. 157–160. Several epistemological problems related to the concept of systems. (1985). Connectedness of general systems. Forrest (1995). Masani. Model. Y. E. Lin. General feedback systems. Cybernet. June 11–15. Existence of major diseases and modeling of epidemiological transition: A general systems theoretic approach. 26. 2. 14. — (1990). and K. Qiu (1987). Int. Continuous mappings between general systems. (1996). 23. T. Developing a mathematical theory of computability which speaks the language of levels. Int. Remarks on the definition of systems.. Singapore. Y. 16. Lin. Y.). A few philosophical remarks on general systems theory. 32–38. Math. Lin. and R. Systems Eng. Comput. 143–178. 3. Ma. NJ. Math. System — a unified concept. J. Model. A systems identification of social problems and its application to public issues of contention during the twentieth century.. Sci. Lin. A. de Marco (1980). and Y. — (1993). 13. 9. Comput. General Systems. (1994). A few systems-colored views of the world. Math. Martin.. In: R. 375–406. P. 28. Feedback transformation and its applications. Some unsolved problems in general systems theory I and II. M. — (1990). J. Ma (1987). Prentice-Hall. 24. and E. Systems Anal. Cybernet. Math. 13. Y. Y. Remarks on analogy between systems.. 923–931.-H. 18. Hu. The scientific methodology in the light of cybernetics. Y.

Morowitz (eds. 509–538. Princeton. 93. London. Minsky. (1974). Moritz. Pitman. London. — (1989). Mumford.. Honderich (ed). K. Urbana.). Englewood Cliffs. Catastrophe Theory and Its Applications. Omran. Mitten. — (1965). On the structure of zero dimensional spaces. Jiangsu Inst.. A. Belmont. New York. Popper. and W. The epidemiologic transition: A theory of the epidemiology of population change. On Mathematics and Mathematicians. J. F. Technol. (1970). In: T. Princeton University Press. Cybernetics and Applied Systems.374 References Mesarovic. Routledge. General Systems Theory: Mathematical Foundations. Moore. — (1976). D... C. H. A. No. 28. New York. 4. Reichel (1975). 49. General principles of systems analysis. Views on general systems theory. (1964). New York.D. and Y. Computation. L. Berlin. E. T. Abstract Systems Theory. Quigley. Pansystems Research about Dynamic Programming. H. Proceedings of the 2nd Systems Symposium at Case Institute of Technology Wiley. 43–46. Impact Sci. (1973). New York. Singapore. Theoretical and Mathematical Biology. Management Sci. New York. On uniform spaces with linearly ordered bases II. Preference order dynamic programming.. Mickens (1990). Math. D. W. (1970). 62–68. 1–10.. A participative approach to computer systems. New York. In: M. C. G. (1971).. Information Theory in Biology.. Academic Press. Pullback versus feedback. D.). Routledge. Quastler. Mathematics and Science. 21. (1955). Negoita (ed. 73–79. Negoita. NJ. Perlman. The Open Society and Its Enemies. New York. World Scientific. Dynamic programming and pansystems operation epitome principle. 2. M. M. V. Mesarovic. Mesarovic (ed. . 2nd ed. CA. Appleton-Century Crofts. Polya. Qin.. The Atom and the Universe. and H. — (1991).. Prentice Hall. Wadsworth. A.. Finite and Infinite Machines. Hawgood (1978). The University of Illinois Press. IL. McGraw-Hill. (1966).. (1942). The Infinite: The Problems of Philosophy — Their Past and Present. (1988). London and New York. Marcel Dekker. Takahara (1975). 37. and I. No. (1991). Milband Memorial Fund Q. In: T. 235–253. Math. Manual of Axiomatic Set Theory.-G. How to Solve It: A New Aspect of Mathematical Method. P. G. 239-247. L. In: C. V. G. Fund. (1992). M. (1967). G. Poston. Indag. Nyikos. J. R. 2. Soc. J. Springer-Verlag. Biological Control Systems Analysis. Dover.. (1945). S.). H. Stewart (1978). 120–136. Land. Milsum. E. Waterman and H. NJ. R. J. Guizhou Sci.

Vol.. Pacific Grove. 1. Theorie des Distributions. Pacific Grove. 1305–1315. Constructing Social Problems. 12 (1950). . Drug and Tracer Kinetics... Martin’s axiom. Topology and life: In: search of general mathematical principles in biology and sociology. CA.. CA. A meaning of the decoupling by feedback of linear functional time systems. CA. II. Int. 1. Rashevsky. Suslin. Math. Paris. Brooks/Cole. Shubik. Saito.. J. A.. M. Schetzen. Remarks on some topological spaces of higher power. 3rd ed. E. Andre (1990). (1920). C. Moscow. Wiley. J. A Transition to Advanced Mathematics. 47–61. Johns Hopkins University Press. No. F. Blauberg. Int. and W. I. Bull. 11. 7–17. I. Stevenson. 9. (1982). R. (1975). (1983). MA. (1980). Rugh. V. Biophys. The Volterra and Wiener Theories of Non-Linear Systems. W. Game Theory in the Social Science: Concepts and Solutions. 3. Barwise (ed. and W. J. No. (1956). 3. Fund. University of Illinois Press. IL. (1987). and G. Smith. M. Cummings. Systems Sci. 183–196. Sikorski. Youn (1990). no. Mesarovic (1985). Hermann. 125–136.. Thron (1971).). Systems Research. 17. M. Vol. 123–138. Results on ω µ -metric spaces and ω µ . (1950).. D. Problems 3. 610–617. 171–178. (1977). General Systems. J. and J... L. Systems Sci. Baltimore. Stewart.Technical Introduction.. MIT Press. Takahara. Yearbook (1969–1976). Fund.. T.. Amsterdam. I. R. M. Cambridge. Rudin. 273–281. 82. Am. and M.References 375 Rapoport.proximities. Shoenfield. — (1987). I–III. 73. 37. Shannon. MA. Calculus. In: J. Segre (1966). Cambridge University Press. Math. M. (1986). Handbook of Mathematical Logic. Martin’s axiom.. Weaver (1949). — (1951). No. (eds. Monthly. 223. Blaisdell. General Systems. Richards. Math. Menlo Park. Hermann. (1949). Int. I. I–VIII. M. D. Urbana. 57-65. Eggen. Paris. Outline of a probabilistic approach to animal sociology. Waltham. The Mathematical Theory of Communication... and R. (1981). T. Large Scale Systems. (1950). Cambridge. Feedback transformation of linear systems. I. Spector. Theorie des Distributions. Theory of Distributions: A Non. Math. Decomposability conditions for general systems. 1017–1027. A.). Math. NorthHolland. N. Y. Brooks/Cole. Structure of the class of linear functional time systems under output feedback transformation. St. Schwartz. J. J. Nonlinear Systems Theory: The Volterra/Wiener Approach. 18. 11. Fund. et al. and H. W. Rescigno. Vols.. Kitsuse (1977). New York. 1. Saito. J.

Intor noad alcune osservazioni sui segmeni infiniti e infinitesimi attuali. II. Wu. 15. Multicriteria dynamic programming. In: G. Math. Sci. J. A taxonomy of current approaches to systems analysis. J. 23. Soc. 125–164. Linear Multi. Comput. Wonham. 58. 2.Variable Control: A Geometric Approach. Ser. 1. Y. Soc.. J. 12–16.. (1993). 2. M. No.. Pure Appl.. No. 1. The unreasonable effectiveness of mathematics in the natural sciences. — (1896). 310–19. Tian.. 42. Sci. 300. E. Wu (1986). pp. Pansystems methodology: Concepts. 33–56. Singapor. (1936). S. 423–432. Exploration. M. Gauss. and X. 2. 2.additive spaces. 1–14. F.. 181–184. No. M. A. Wood-Harper. Proc. — (1982). — (1981).-T. Wetensch. Exploration. Systems Sci. (1979). theorems and applications III. Vierthaler. C. 4. pp. Twenty-five volumes of Fundamenta Mathematica. The Mathematical Heritage of C. Wholesale content analysis: Collecting long-term data to study large-scale societal changes of historic importance. W. Generalized number system and its applications I. A.-M. chemistry. Math. Wang (1964). 1955). 55. 2. Sinica. theorems and applications I. New York. (1936). Model. Math. theorems and applications II. D. Turing. 810–826. Ser. with an application to the Entscheidungs problem. Remarks on ω µ. Bull. World Scientific. A. E. Contributions to the theory of models I. Int. Ann. Springer-Verlag. Tarski. Optimal control problems with nonscalar valued performance criteria and dynamic programming. Sci... 93–106.-M. — (1985). 2. Wu and X. Sci. Exploration. III. (1981). Math. Proc. Sci. . F. Sci. 47. P. Nederl. and G. 1185–1206. Asahi (1985). X72–X76. Kybernetes. 56–64. 2. Causality invariance of general systems under output feedback transformations. Generalized fundamental equations in pansystems network analysis. 25. 203–15.. 1. Rassias (ed. Fund. Scientific Press. 9. (1980). G. 101–112. logic and economics). No. and T. (1891). London Math. Fondamenti di Geometria a Piu Dimensioni. Pansystems methodology: Concepts. X. 16. — (1982). 582–588..376 References Takahara. Comput. 10. — (1991). (1960). 572–581. P. Beijing. Tsukuba J. Exploration. Fitzgerald (1982).. Am. A. S. Proceedings of the 8th Triennial World Congress of the IFAC. — (1982). C. Wigner. Akad. Padova. Pansystems methodology: Concepts. (1954. A non-Archimedean number field and its applications in modem physics. No.. Voronese.). Wu. 814–822. Math. A. 13. Tamarkin. The Course of Science of Sciences..-Z. Math. 42. T. Comm. Wang (1980). On computable numbers. No. Pansystems analysis — a new exploration of interdisciplinary investigation (with appendices on discussions on medicine. 57. 123–132.

theorems and applications V.. 995–1010. A. Nos. Mech. 1. and X. theorems and applications VI. Y. Gvishiani (ed. From pansystems philosophy to pansystems poetry. No. S.metrizable spaces. 339–344. theorems and applications VII. Exploration. No. Exploration. — (1984). 3. No. Appl. 231–235. Zadeh. Sci. 8. 91. 50.. Exploration. Exploration. In: J. No.150. 20. Exploration.. theorems and applications VIII. Mech. 159–180. 125–137. (1975). 5. . 4 97–104. Yasui. L. No. Sci. Tianshui Shichuan Xue Bao. 3. Hunan Science and Technology Press.Some Unsolved Problems in General Systems Theory 377 — (1983). From circuit theory to systems theory. (1962).-Z. An approach on fixed pansystems theorems: Panchaos and strange panattractor. Japon. Approximation Transformation Theory and the Pansystems Concepts in Mathematics. Systems Research: Methodological Problems. No. Math. 13. Pansystems methodology: Concepts. — (1984). Proc. — (1983). Pansystems methodology: Concepts. 1–14. Pansystems methodology: Concepts. — (1984). 1. Math. 30.. Zhu. — (1984). X. Generalized principle of optimality in pansystems network analysis. On ωµ . IRE. No. 1. 4. (1984).-D. 9–23. 856–865. Sci. I. Wu (1987). Math. Sci. 4. Appl.-M. — (1985). pp. 107–116. — (1994). Pansystems methodology: Concepts. Kybernetes. 3–4. Investigation and applications of pansystems recognition theory and pansystemsoperations research of large-scale systems. M. theorems and applications IV. Pansystems methodology: Concepts. New York. 1.). Wu and X. Sci. The development of science as an open system. Yablonsky. 4. 211– 228. Oxford University Press. 5. Wu (1984). 4..-M.

20 Decision theory. 345 Densely ordered set. 344 AC. 12. 16 Black box. 299 Atom. 256 Class of all sets. 19 Complex system representation. 9 Axiom of choice. 168 Associated input–output system. 344 Almost disjoint family. 285 Dedekind cut. 123 Chaos. 34 Cofinality. 62 Axiomatic set theory. 309 Decoupled by feedback. 332 (m. 61 Continuous everywhere. 33 Chain of object systems. 10 Causal relation. 16 Derivative. 78 Ancient atomists. 11 Complement. 113 Conservation of mechanical energy. 40 Density. 115 Consistent set. 114 Cartesian product of two sets. 137. 59 Beauty of mathematics. 6 Component system. 75 Coinitial. 8 De Morgan’s Laws. 2 Data analysis approach. 57 Calorists. 135 Bernstein’s Equipollence Theorem. 164 Compartment theory. 5 Causal system. 22 Catastrophe theory. 10 Cofinal. 116. 2 Decomposability. n) -derivative. 325 Burali-Forti paradox. 321 1–1 mapping. 325 Continuously ordered set. 281 Concatenation. 274 Conservation of matter. 276 Chain. 113 Atomic theory. 41 Continuum hypothesis. 148 Bellman’s principle of optimality. 114 Attractor. 75 Crystalline material. 344 Box product topology. 95 Classical systems theory. 336 Denumerable set. 292 Complexity. 41 Cybernetics. 164 Bellman’s Principle of Optimality. 25 Big sum. 27 Addition of order types. 5 Arithmetic of natural numbers. 175 Cut. 2 Boolean algebra. 114 Antichain.Index (G) integrals. 174 Crystallization process. 258 Attributes. 328 Bijective mapping. 16 Abstract distances. 33 Apollo project. 62 Addition. 35 Algebraic geometry. 34 Color. 325 379 .

11 Galileo’s method. 326 Factor system. 335 Infinite set. 67. 326 Genetic algorithm. 135 Equipollence. 59 Formal language. 314 Information highway. 3 Hu’s counterexample. 174 Equipment-replacement models. 136. 60 Foundations of geometry. 302 Injective mapping. 136 Human activity systems approach. 125 Generalized power series. 345 Height. 64 Gödel’s theorem. 5 Epimorphism. 133 GNL-integral.380 Descartes’ second principle. 16. 123 General systems methodology. 125 Infinity. 125 Divisibility. 337 Empty set. 4 Game Theory. 255 Input–output system. 84 Hierarchy of monads. 266 Feedback decoupling. 311 Generalized number system. 125 Generalized Principle of Optimality. 113 Function. 298 Interior measure. 292 GNL-integrable function. 294 Epistemology. 265 Fundamental equation. 309 Feedback transformation. 77 . 298 Input–output system partition. 80 Finite set. 266 Filter. 309 General system. 12 Global system representation. 281 Feedback component. 16 Input space. 5 Injective complex system representation. 327 Graph theory. 12 Index Heat of friction. 345 Elementary space. 135 Isolated ordinal. 327 Inventory problem. 75 Generalized number. 157 κ -disjoint sets. 62 Exterior measure. 2 General linear system. 125. 295 ith-order generalized extreme value. 314 Infinities. 19 Energies. 126. 188 Identical systems. 67 Functional system. 255 Input–output system covering. 172 Graph of a function. 123 Impassable chasm. 73 Finitist. 27 Extensionality. 16 Exponentiation. 344 Four-element philosopher. 7 Hyperreal number. 344 Formalist. 64 First-order predicate calculus. 64 Formula. 124 Dirac’s quantum mechanics. 346 Fuzzy mathematics. 341 Gödel completeness theorem. 114 Heaviside function. 344 Dirac’s δ function. 309 Feedback invariant property. 135 Einstein’s relativity. 180 Increasing function. 171 Formal power series. 12 Game theory. 111 Dynamic programming. 271 Feedback system. 265. 6 Generalized continuum hypothesis. 143 Fundamental particle. 73 Infinitesimal. 330 GNS function induced by f. 124 Distribution. 139 Generalized solution. 43 Indefinite integral. 124 Elementary functions. 111 Fundamental Theorem of Calculus. 3 Dimension theory. 70 Isomorphism.

150 Optimization model. 263 Linear time system. 115 Non-Archimedean analysis. 326 Russell’s paradox. 168 Platonists. 48 Limit ordinal. 67 Resource allocation problems. 87 Order type. 23 Partial order. 72 Ordinal number. 345 Quantum mechanics. 138 Order topology. 10. 174 Mathematical induction. 80 Participant approach. 27 Music. 317 Quasi-(m. 276 Problem-oriented systems approaches. 157 Objects. 166 Naive set theory. 33 Modeling relation. 123 nth-order optimization solution. 45 Ordinalities. 9 ω µ -additive topological spaces. 344 p-resolving mapping. 328 Lebesgue integration. 188 Monomorphism. 36 Products of distributions. 75 Regularity. 57. 7 Peano’s axioms. 6 Product mapping. 344 Non-Archimedean number field. 179 Mathematical proofs. n) limit. 296 MT-time system. 116. 135 Rest mass. 117 Law of Conservation of Fundamental Particles. 16 Open interval. 12 Network. 273 Multilevel structure. 117 Lebesgue integral. 62 Precausal system. 181 Maximal. 281 Nonstandard analysis. 311 Limit of the λ -sequence. 124 Riemann integral. 309 Paradox of a moving particle. 174 Monad. 327 Length of a relation. 16. 33 Medieval mysticism. 32 Ordered sum. 26. 33 Minimum. 344 One-element Ionians. 123 Level index. 255 Overall system. 125 Non-Archimedean topology. 326 Operation Epitome Principle. 166 Pansystems theory.Index 381 Lavoisier’s Axiom. 15 Navier–Stokes equation. 344 Linearly independent functions. 293 Molecular spatial structure. n) continuity. 35 Ordinal exponentiation. 35 Ordered algebraic structures. 62 Relation. 45 Original system. 112 One-to-one mapping. 180 Paradox of the hotel. 155 Painting. 80 Materials science. 266 Output space. 166 Power set. 47 Linear order relation. 64 Poetry. 281 p-adic topology. 60 Quantum field theory. 22 Ordered set. 339 Mapping. 293 Product of order types. 344 Ordered pair. 33 Maximum. 126. 346 Quasi-(m. 321 Regular ordinal. 13 Newton’s law of action and reaction. 346 Proper subset. 274 Linear uniformity. 3 Minimal. 125 nth-level object system. 344 Noninteracted system. 67 Martin’s Axiom. 10. 149 Optimization model. 19 Quantifier. 180 Partial function. 67 Linear system. 45 Ordinals. 148 Optimum-preserving law. 33 Linear ordering. 123 . 122 Multiplication. 325 Lebesgue measurable set.

6 Systems connected in series. 326 Voronese’s non-Archimedean geometry. 38 Unbordered set. 149 Strictly strongly antisymmetric relation. 67 Strictly order-preserving function. 138 Zeroth-order optimization solution. 345 Test function. 291 Systems description of mathematics. 326 Unit cell. 172 Systems movement. 1 Scientific revolution. 345 Transfinite Induction. 346 Science of science. 13 Small sum. 277 Topological space. 123 Successor. 7 Transfer principle. 34 Simple function. 59 . 325 Well-founded set. 4 Seeds. 2 Taylor expansion. 72 Traveling-salesman problem. 113 Self-controlled equipment. 336 Theory of automata. 150 Strongly antisymmetric relation. 61 Utopian project. 13 Stone duality theorem. 60 Subsystem. 15 Set existence.382 Index Schwartz’s distributions. 40 Uncountable set. 327 Simulation. 1 Zermelo–Russell paradox. 79 Type class. 14 Theory of distributions. 87 Total ordering. 9 Systemality. 91 Well-ordered set. 174 Universal closure. 4 Vase puzzle. 70 Successor ordinal. 57 Zero. 344 Weak derivative. 5 Time invariably realizable system. 335 Theory of science of science. 347 Systems research. 70 Surjective complex system representation. 4 Systems theory. 61 Set. 2 Wholeness. 62 Set theory. 78 ∆ -System Lemma. 135 Tree. 88 System with relations. 274 Shortest-path problem. 6 Traditional approach. 344 Strange attractor. 176 Three-body problem. 156 ZFC axiom system. 302 Surjective mapping. 52 White box. 150 Strongly stationary system. 88 Suslin tree. 42 Well-Ordering Theorem. 85 Suslin’s hypothesis. 178 Volume of a interval. 10 Shift operator. 72 Transfinite Recursion. 16 Uniform convergency. 135 Similarity mapping. 84 ∆ -system. 8 Subformula. 67 Totality. 16 Suslin line. 275 Structured systems analysis approach. 5 Sentence. 148 Statistics. 260 Strict total ordering. 328 Stationary set.

Sign up to vote on this title
UsefulNot useful